SMART_BRFSS_City_and_Country_Analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ***Libraries & Tools***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from scipy.stats import randint, uniform, chi2_contingency\n",
    "\n",
    "\n",
    "import time\n",
    "import json\n",
    "import math\n",
    "import re\n",
    "\n",
    "# NLP tools\n",
    "import nltk\n",
    "from wordcloud import WordCloud\n",
    "from nltk.tokenize import word_tokenize\n",
    "\n",
    "# Scikit-Learn\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.tree import DecisionTreeClassifier, export_graphviz, plot_tree\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.model_selection import train_test_split, RandomizedSearchCV, GridSearchCV, StratifiedShuffleSplit\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, make_scorer, classification_report, RocCurveDisplay, f1_score\n",
    "from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier, BaggingClassifier\n",
    "from sklearn.preprocessing import LabelEncoder, OneHotEncoder\n",
    "from sklearn.naive_bayes import GaussianNB, MultinomialNB, BernoulliNB\n",
    "from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer\n",
    "from sklearn.calibration import CalibratedClassifierCV, calibration_curve\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn import tree\n",
    "\n",
    "# Imbalanced learn library\n",
    "from imblearn.over_sampling import SMOTE, RandomOverSampler, SMOTEN, SMOTENC, BorderlineSMOTE\n",
    "from imblearn.under_sampling import RandomUnderSampler\n",
    "\n",
    "# Boosting ensemble methods\n",
    "from xgboost import XGBClassifier\n",
    "from catboost import CatBoostClassifier, Pool\n",
    "\n",
    "\n",
    "from itertools import product\n",
    "from IPython.display import Image\n",
    "import pydotplus\n",
    "import graphviz\n",
    "import shap\n",
    "\n",
    "SEED = 42"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to /home/georgematlis/AUTH/DWS/A\n",
      "[nltk_data]     DVANCED_TOPICS_IN_MACHINE_LEARNING/CDC-\n",
      "[nltk_data]     BRFSS/Downloaded_Files...\n",
      "[nltk_data]   Unzipping tokenizers/punkt.zip.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Download punkt package for wordcloud\n",
    "nltk.download('punkt', download_dir='/home/georgematlis/AUTH/DWS/ADVANCED_TOPICS_IN_MACHINE_LEARNING/CDC-BRFSS/Downloaded_Files')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Specify the path to the tokenizers\n",
    "nltk.data.path.append(\"/home/georgematlis/AUTH/DWS/ADVANCED_TOPICS_IN_MACHINE_LEARNING/CDC-BRFSS/Downloaded_Files\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ***Data Overview and Exploration***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Overview"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('MMSA_2021.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 227792 entries, 0 to 227791\n",
      "Columns: 152 entries, DISPCODE to MMSANAME\n",
      "dtypes: float64(151), object(1)\n",
      "memory usage: 264.2+ MB\n"
     ]
    }
   ],
   "source": [
    "data.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>DISPCODE</th>\n",
       "      <th>STATERE1</th>\n",
       "      <th>CELPHON1</th>\n",
       "      <th>LADULT1</th>\n",
       "      <th>COLGSEX</th>\n",
       "      <th>LANDSEX</th>\n",
       "      <th>RESPSLCT</th>\n",
       "      <th>SAFETIME</th>\n",
       "      <th>CADULT1</th>\n",
       "      <th>CELLSEX</th>\n",
       "      <th>...</th>\n",
       "      <th>_FRTLT1A</th>\n",
       "      <th>_VEGLT1A</th>\n",
       "      <th>_FRT16A</th>\n",
       "      <th>_VEG23A</th>\n",
       "      <th>_FRUITE1</th>\n",
       "      <th>_VEGETE1</th>\n",
       "      <th>_MMSA</th>\n",
       "      <th>_MMSAWT</th>\n",
       "      <th>SEQNO</th>\n",
       "      <th>MMSANAME</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1200.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>10100.0</td>\n",
       "      <td>90.030265</td>\n",
       "      <td>2.021000e+09</td>\n",
       "      <td>b'Aberdeen, SD, Micropolitan Statistical Area'</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1200.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>10100.0</td>\n",
       "      <td>29.643176</td>\n",
       "      <td>2.021000e+09</td>\n",
       "      <td>b'Aberdeen, SD, Micropolitan Statistical Area'</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1200.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>10100.0</td>\n",
       "      <td>37.000680</td>\n",
       "      <td>2.021000e+09</td>\n",
       "      <td>b'Aberdeen, SD, Micropolitan Statistical Area'</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1200.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>10100.0</td>\n",
       "      <td>49.702755</td>\n",
       "      <td>2.021000e+09</td>\n",
       "      <td>b'Aberdeen, SD, Micropolitan Statistical Area'</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1200.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>5.397605e-79</td>\n",
       "      <td>10100.0</td>\n",
       "      <td>15.396983</td>\n",
       "      <td>2.021000e+09</td>\n",
       "      <td>b'Aberdeen, SD, Micropolitan Statistical Area'</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 152 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   DISPCODE  STATERE1  CELPHON1  LADULT1  COLGSEX  LANDSEX  RESPSLCT  \\\n",
       "0    1200.0       NaN       NaN      NaN      NaN      NaN       NaN   \n",
       "1    1200.0       NaN       NaN      NaN      NaN      NaN       NaN   \n",
       "2    1200.0       NaN       NaN      NaN      NaN      NaN       NaN   \n",
       "3    1200.0       NaN       NaN      NaN      NaN      NaN       NaN   \n",
       "4    1200.0       NaN       NaN      NaN      NaN      NaN       NaN   \n",
       "\n",
       "   SAFETIME  CADULT1  CELLSEX  ...  _FRTLT1A  _VEGLT1A  _FRT16A  _VEG23A  \\\n",
       "0       1.0      1.0      1.0  ...       2.0       1.0      1.0      1.0   \n",
       "1       1.0      1.0      1.0  ...       2.0       1.0      1.0      1.0   \n",
       "2       1.0      1.0      2.0  ...       1.0       1.0      1.0      1.0   \n",
       "3       1.0      1.0      1.0  ...       1.0       1.0      1.0      1.0   \n",
       "4       1.0      1.0      2.0  ...       1.0       1.0      1.0      1.0   \n",
       "\n",
       "       _FRUITE1      _VEGETE1    _MMSA    _MMSAWT         SEQNO  \\\n",
       "0  5.397605e-79  5.397605e-79  10100.0  90.030265  2.021000e+09   \n",
       "1  5.397605e-79  5.397605e-79  10100.0  29.643176  2.021000e+09   \n",
       "2  5.397605e-79  5.397605e-79  10100.0  37.000680  2.021000e+09   \n",
       "3  5.397605e-79  5.397605e-79  10100.0  49.702755  2.021000e+09   \n",
       "4  5.397605e-79  5.397605e-79  10100.0  15.396983  2.021000e+09   \n",
       "\n",
       "                                         MMSANAME  \n",
       "0  b'Aberdeen, SD, Micropolitan Statistical Area'  \n",
       "1  b'Aberdeen, SD, Micropolitan Statistical Area'  \n",
       "2  b'Aberdeen, SD, Micropolitan Statistical Area'  \n",
       "3  b'Aberdeen, SD, Micropolitan Statistical Area'  \n",
       "4  b'Aberdeen, SD, Micropolitan Statistical Area'  \n",
       "\n",
       "[5 rows x 152 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Some columns have many null values\n",
    "isna_data = data.isna().sum()\n",
    "for column, count in isna_data.items():\n",
    "    print(f\"{column}: {count}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# If we were to drop all null values, the data would have no records left\n",
    "data.dropna().shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Hand-pick Variables\n",
    "Select some interesting columns from the dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "columns = [\n",
    "    # === List with all meaningful columns ===\n",
    "    [\n",
    "        # Sec 1 Health Status\n",
    "        '_RFHLTH',\n",
    "\n",
    "        # Sec 2 Healthy Days\n",
    "        '_PHYS14D',\n",
    "        '_MENT14D',\n",
    "\n",
    "        # Sec 3 Health Care Access\n",
    "        '_HLTHPLN',\n",
    "        '_HCVU652',\n",
    "\n",
    "        # Sec 4 Exercise\n",
    "        '_TOTINDA',\n",
    "\n",
    "        # Sec 5 Hypertension Awareness\n",
    "        '_RFHYPE6',\n",
    "\n",
    "        # Sec 6 Cholesterol Awareness\n",
    "        '_CHOLCH3',\n",
    "        '_RFCHOL3',\n",
    "\n",
    "        # Sec 7 Chronic Health Conditions\n",
    "        '_MICHD',\n",
    "\n",
    "        # Sec 8 Arthritis\n",
    "        '_DRDXAR3',\n",
    "        '_LMTACT3',\n",
    "        '_LMTWRK3',\n",
    "\n",
    "        # Sec 11 Tobacco Use\n",
    "        '_SMOKER3',\n",
    "        '_RFSMOK3',\n",
    "        '_CURECI1',\n",
    "\n",
    "        # Sec 12 Alcohol Consumption\n",
    "        'DRNKANY5',\n",
    "        'ALCDAY5',\n",
    "        '_RFBING5',\n",
    "        '_DRNKWK1',\n",
    "        '_RFDRHV7',\n",
    "\n",
    "        # Sec 14 HIV/AIDS\n",
    "        '_AIDTST4',\n",
    "\n",
    "        # Sec 15 Fruits & Vegetables\n",
    "        'FTJUDA2_',\n",
    "        'FRUTDA2_',\n",
    "        'GRENDA1_',\n",
    "        'FRNCHDA_',\n",
    "        'POTADA1_',\n",
    "        'VEGEDA2_',\n",
    "        '_FRUTSU1',\n",
    "        '_VEGESU1',\n",
    "        '_FRTLT1A',\n",
    "        '_VEGLT1A',\n",
    "        \n",
    "        # Sec 9 Demographics\n",
    "        '_INCOMG1',\n",
    "        '_EDUCAG',\n",
    "        'CHILDREN',\n",
    "        '_RFBMI5',\n",
    "        '_BMI5CAT',\n",
    "        '_BMI5',\n",
    "        'WTKG3',\n",
    "        'HEIGHT3',\n",
    "        '_AGE_G',\n",
    "        '_AGE80',\n",
    "        '_AGE65YR',\n",
    "        '_AGEG5YR',\n",
    "        '_SEX',\n",
    "        '_RACEPRV',\n",
    "        '_RACEGR3',\n",
    "        '_RACEG21',\n",
    "        '_RACE',\n",
    "        '_HISPANC',\n",
    "        '_MRACE1',\n",
    "        '_PRACE1'\n",
    "    ],\n",
    "\n",
    "    # === 1st subset of all the meaningful columns === \n",
    "    [\n",
    "        # Sec 1 Health Status\n",
    "        '_RFHLTH',\n",
    "\n",
    "        # Sec 2 Healthy Days\n",
    "        '_PHYS14D',\n",
    "        '_MENT14D',\n",
    "\n",
    "        # Sec 3 Health Care Access\n",
    "        '_HLTHPLN',\n",
    "        '_HCVU652',\n",
    "\n",
    "        # Sec 4 Exercise\n",
    "        '_TOTINDA',\n",
    "\n",
    "        # Sec 5 Hypertension Awareness\n",
    "        '_RFHYPE6',\n",
    "\n",
    "        # Sec 6 Cholesterol Awareness\n",
    "        '_CHOLCH3',\n",
    "        '_RFCHOL3',\n",
    "\n",
    "        # Sec 7 Chronic Health Conditions\n",
    "        '_MICHD',\n",
    "\n",
    "        # Sec 8 Arthritis\n",
    "\n",
    "        # Sec 11 Tobacco Use\n",
    "        '_SMOKER3',\n",
    "\n",
    "        # Sec 12 Alcohol Consumption\n",
    "        '_RFBING5',\n",
    "        '_RFDRHV7',\n",
    "\n",
    "        # Sec 14 HIV/AIDS\n",
    "        '_AIDTST4',\n",
    "\n",
    "        # Sec 15 Fruits & Vegetables\n",
    "        '_FRUTSU1',\n",
    "        '_VEGESU1',\n",
    "        \n",
    "        # Sec 9 Demographics\n",
    "        '_INCOMG1',\n",
    "        '_EDUCAG',\n",
    "        '_RFBMI5',\n",
    "        '_BMI5CAT',\n",
    "        '_AGE_G',\n",
    "        '_SEX',\n",
    "        '_RACE'\n",
    "    ],\n",
    "\n",
    "\n",
    "    # === 2nd subset of all the meaningul columns ===\n",
    "    [\n",
    "        # Sec 1 Health Status\n",
    "        '_RFHLTH',\n",
    "\n",
    "        # Sec 2 Healthy Days\n",
    "        '_PHYS14D',\n",
    "        '_MENT14D',\n",
    "\n",
    "        # Sec 3 Health Care Access\n",
    "        '_HLTHPLN',\n",
    "        '_HCVU652',\n",
    "\n",
    "        # Sec 4 Exercise\n",
    "        '_TOTINDA',\n",
    "\n",
    "        # Sec 5 Hypertension Awareness\n",
    "        '_RFHYPE6',\n",
    "\n",
    "        # Sec 6 Cholesterol Awareness\n",
    "        '_CHOLCH3',\n",
    "        '_RFCHOL3',\n",
    "\n",
    "        # Sec 7 Chronic Health Conditions\n",
    "        '_MICHD',\n",
    "\n",
    "        # Sec 8 Arthritis\n",
    "        \n",
    "\n",
    "        # Sec 11 Tobacco Use\n",
    "        '_SMOKER3',\n",
    "\n",
    "        # Sec 12 Alcohol Consumption\n",
    "        '_RFBING5',\n",
    "        '_RFDRHV7',\n",
    "\n",
    "        # Sec 14 HIV/AIDS\n",
    "        '_AIDTST4',\n",
    "\n",
    "        # Sec 15 Fruits & Vegetables\n",
    "        '_FRTLT1A',\n",
    "        '_VEGLT1A',\n",
    "        \n",
    "        # Sec 9 Demographics\n",
    "        '_INCOMG1',\n",
    "        '_EDUCAG',\n",
    "        '_RFBMI5',\n",
    "        '_BMI5CAT',\n",
    "        '_AGE_G',\n",
    "        '_SEX',\n",
    "        '_RACE'\n",
    "    ]\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Categorical and Continuous Columns\n",
    "We can use the one hot encoder to properly encode the values of each categorical variable that exhibits no apparent order."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "\n",
    "1: For categorical\n",
    "2: For continuous\n",
    "\n",
    "\"\"\"\n",
    "\n",
    "cat_con_variables = {\n",
    "    # Sec 1 Health Status\n",
    "    '_RFHLTH': 1,\n",
    "\n",
    "    # Sec 2 Healthy Days\n",
    "    '_PHYS14D': 1,\n",
    "    '_MENT14D': 1,\n",
    "\n",
    "    # Sec 3 Health Care Access\n",
    "    '_HLTHPLN': 1,\n",
    "    '_HCVU652': 1,\n",
    "\n",
    "    # Sec 4 Exercise\n",
    "    '_TOTINDA': 1,\n",
    "\n",
    "    # Sec 5 Hypertension Awareness\n",
    "    '_RFHYPE6': 1,\n",
    "\n",
    "    # Sec 6 Cholesterol Awareness\n",
    "    '_CHOLCH3': 1,\n",
    "    '_RFCHOL3': 1,\n",
    "\n",
    "    # Sec 7 Chronic Health Conditions\n",
    "    '_MICHD': 1,\n",
    "\n",
    "    # Sec 8 Arthritis\n",
    "    '_DRDXAR3': 1,\n",
    "    '_LMTACT3': 1,\n",
    "    '_LMTWRK3': 1,\n",
    "\n",
    "    # Sec 11 Tobacco Use\n",
    "    '_SMOKER3': 1,\n",
    "    '_RFSMOK3': 1,\n",
    "    '_CURECI1': 1,\n",
    "\n",
    "    # Sec 12 Alcohol Consumption\n",
    "    'DRNKANY5': 1,\n",
    "    'ALCDAY5': 2,\n",
    "    '_RFBING5': 1,\n",
    "    '_DRNKWK1': 2,\n",
    "    '_RFDRHV7': 1,\n",
    "\n",
    "    # Sec 14 HIV/AIDS\n",
    "    '_AIDTST4': 1,\n",
    "\n",
    "    # Sec 15 Fruits & Vegetables\n",
    "    'FTJUDA2_': 2,\n",
    "    'FRUTDA2_': 2,\n",
    "    'GRENDA1_': 2,\n",
    "    'FRNCHDA_': 2,\n",
    "    'POTADA1_': 2,\n",
    "    'VEGEDA2_': 2,\n",
    "    '_FRUTSU1': 2,\n",
    "    '_VEGESU1': 2,\n",
    "    '_FRTLT1A': 1,\n",
    "    '_VEGLT1A': 1,\n",
    "    \n",
    "    # Sec 9 Demographics\n",
    "    '_INCOMG1': 1,\n",
    "    '_EDUCAG': 1,\n",
    "    'CHILDREN': 1,\n",
    "    '_RFBMI5': 1,\n",
    "    '_BMI5CAT': 1,\n",
    "    '_BMI5': 2,\n",
    "    'WTKG3': 2,\n",
    "    'HEIGHT3': 2,\n",
    "    \n",
    "    # === The age variables can remain the same because their values represent a natural ordering ===\n",
    "    '_AGE_G': 1,\n",
    "    '_AGE80': 1,\n",
    "    '_AGE65YR': 1,\n",
    "    '_AGEG5YR': 1,\n",
    "    \n",
    "    '_SEX': 1,\n",
    "    '_RACEPRV': 1,\n",
    "    '_RACEGR3': 1,\n",
    "    '_RACEG21': 1,\n",
    "    '_RACE': 1,\n",
    "    '_HISPANC': 1,\n",
    "    '_MRACE1': 1,\n",
    "    '_PRACE1': 1\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Column Values that are Irrelevant\n",
    "Some columns/variables have values, such as \"Dont Know/Not Sure\", that are irrelevant to the analysis. These values are included in data mining methods but can be ignored in NLP mehtods. NLP methods handle text sequences of varying forms."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "irrel_col_values = {\n",
    "    # Sec 1 Health Status\n",
    "    '_RFHLTH': 9,\n",
    "\n",
    "    # Sec 2 Healthy Days\n",
    "    '_PHYS14D': 9,\n",
    "    '_MENT14D': 9,\n",
    "\n",
    "    # Sec 3 Health Care Access\n",
    "    '_HLTHPLN': 9,\n",
    "    '_HCVU652': 9,\n",
    "\n",
    "    # Sec 4 Exercise\n",
    "    '_TOTINDA': 9,\n",
    "\n",
    "    # Sec 5 Hypertension Awareness\n",
    "    '_RFHYPE6': 9,\n",
    "\n",
    "    # Sec 6 Cholesterol Awareness\n",
    "    '_CHOLCH3': 9,\n",
    "    '_RFCHOL3': 9,\n",
    "\n",
    "    # Sec 7 Chronic Health Conditions\n",
    "\n",
    "    # Sec 8 Arthritis\n",
    "    '_LMTACT3': 9,\n",
    "    '_LMTWRK3': 9,\n",
    "\n",
    "    # Sec 11 Tobacco Use\n",
    "    '_SMOKER3': 9,\n",
    "    '_RFSMOK3': 9,\n",
    "    '_CURECI1': 9,\n",
    "\n",
    "    # Sec 12 Alcohol Consumption\n",
    "    'DRNKANY5': 9,\n",
    "    'ALCDAY5': 900,\n",
    "    '_RFBING5': 9,\n",
    "    '_DRNKWK1': 99900,\n",
    "    '_RFDRHV7': 9,\n",
    "\n",
    "    # Sec 14 HIV/AIDS\n",
    "    '_AIDTST4': 9,\n",
    "\n",
    "    # Sec 15 Fruits & Vegetables\n",
    "    '_FRTLT1A': 9,\n",
    "    '_VEGLT1A': 9,\n",
    "    \n",
    "    # Sec 9 Demographics\n",
    "    '_INCOMG1': 9,\n",
    "    '_EDUCAG': 9,\n",
    "    'CHILDREN': 9,\n",
    "    '_RFBMI5': 9,\n",
    "    '_AGE65YR': 3,\n",
    "    '_AGEG5YR': 14,\n",
    "    '_RACEGR3': 9,\n",
    "    '_RACEG21': 9,\n",
    "    '_RACE': 9,\n",
    "    '_HISPANC': 9,\n",
    "    '_MRACE1': [77,99],\n",
    "    '_PRACE1': [77, 99]\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of rows with value 9 in column '_RFHLTH': 601\n",
      "Number of rows with value 9 in column '_PHYS14D': 4691\n",
      "Number of rows with value 9 in column '_MENT14D': 4011\n",
      "Number of rows with value 9 in column '_HLTHPLN': 9341\n",
      "Number of rows with value 9 in column '_HCVU652': 83992\n",
      "Number of rows with value 9 in column '_TOTINDA': 454\n",
      "Number of rows with value 9 in column '_RFHYPE6': 1031\n",
      "Number of rows with value 9 in column '_CHOLCH3': 15155\n",
      "Number of rows with value 9 in column '_RFCHOL3': 1639\n",
      "Number of rows with value 9 in column '_LMTACT3': 712\n",
      "Number of rows with value 9 in column '_LMTWRK3': 1664\n",
      "Number of rows with value 9 in column '_SMOKER3': 14005\n",
      "Number of rows with value 9 in column '_RFSMOK3': 14005\n",
      "Number of rows with value 9 in column '_CURECI1': 13478\n",
      "Number of rows with value 9 in column 'DRNKANY5': 15146\n",
      "Number of rows with value 900 in column 'ALCDAY5': 0\n",
      "Number of rows with value 9 in column '_RFBING5': 19641\n",
      "Number of rows with value 99900 in column '_DRNKWK1': 19412\n",
      "Number of rows with value 9 in column '_RFDRHV7': 19412\n",
      "Number of rows with value 9 in column '_AIDTST4': 11198\n",
      "Number of rows with value 9 in column '_FRTLT1A': 27859\n",
      "Number of rows with value 9 in column '_VEGLT1A': 32613\n",
      "Number of rows with value 9 in column '_INCOMG1': 50442\n",
      "Number of rows with value 9 in column '_EDUCAG': 1471\n",
      "Number of rows with value 9 in column 'CHILDREN': 23\n",
      "Number of rows with value 9 in column '_RFBMI5': 26341\n",
      "Number of rows with value 3 in column '_AGE65YR': 5586\n",
      "Number of rows with value 14 in column '_AGEG5YR': 5586\n",
      "Number of rows with value 9 in column '_RACEGR3': 6007\n",
      "Number of rows with value 9 in column '_RACEG21': 6007\n",
      "Number of rows with value 9 in column '_RACE': 6007\n",
      "Number of rows with value 9 in column '_HISPANC': 3241\n",
      "Number of rows with values in [77, 99] in column '_MRACE1': 7869\n",
      "Number of rows with values in [77, 99] in column '_PRACE1': 7869\n",
      "Size of the dataset after removing rows with specified values: 84002\n"
     ]
    }
   ],
   "source": [
    "for column, values in irrel_col_values.items():\n",
    "    if isinstance(values, int):  \n",
    "        num_rows = len(data[data[column] == values])\n",
    "        print(f\"Number of rows with value {values} in column '{column}': {num_rows}\")\n",
    "    elif isinstance(values, list):  \n",
    "        num_rows = len(data[data[column].isin(values)])\n",
    "        print(f\"Number of rows with values in {values} in column '{column}': {num_rows}\")\n",
    "\n",
    "# Calculating the size of the dataset after removing rows with specified values\n",
    "filtered_data = data.copy() \n",
    "for column, values in irrel_col_values.items():\n",
    "    if isinstance(values, int):  \n",
    "        filtered_data = filtered_data[filtered_data[column] != values]\n",
    "    elif isinstance(values, list):  \n",
    "        filtered_data = filtered_data[~filtered_data[column].isin(values)]\n",
    "\n",
    "# Calculate the size of the filtered dataset\n",
    "filtered_size = filtered_data.shape[0]\n",
    "print(f\"Size of the dataset after removing rows with specified values: {filtered_size}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some columns described in the 2021 SMART City and Country CodeBook are derived from other columns in the dataset.\n",
    "\n",
    "- DROCDY3_ Calculated variable for drink-occasions-per-day. DROCDY3_ is derived from ALCDAY5 by dividing the ALCDAY5 variable by 7 days per week or 30 days per month.\n",
    "\n",
    "- _CHLDCNT Calculated variable for number of children in household. _CHLDCNT is derived from CHILDREN.\n",
    "\n",
    "- HTM4 Calculated variable for reported height in meters. HTM4 is derived from the variable HTIN4 by multiplying HTIN4 by 2.54 cm per in and dividing by 100 cm per meter. HTM4 is derived from HEIGHT2 metric values by dividing by 100.\n",
    "\n",
    "- HTIN4 Calculated variable for reported height in inches. HTIN4 is derived from HEIGHT3. HTIN4 is calculated by adding the foot portion of HEIGHT3 multiplied by 12, to the inch portion. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Candidate Dataframe Creation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Check if the columns selected earlier exist in the dataset\n",
    "for col in columns[0]:\n",
    "    if col in data.columns:\n",
    "        print(col)\n",
    "    else:\n",
    "        print(f'{col} does not exist')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# How many null values each selected column has\n",
    "isna_data = data[columns[0]].isna().sum()\n",
    "for column, count in isna_data.items():\n",
    "    print(f\"{column}: {count}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create dataframes based on the selected columns\n",
    "# Question: Should we drop the rows with null values?\n",
    "dataframes = [\n",
    "    data[columns[0]].dropna().copy(),\n",
    "    data[columns[1]].dropna().copy(),\n",
    "    data[columns[2]].dropna().copy(),\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataframes[2].info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Categorical Columns: \n",
      "['_RFHLTH', '_PHYS14D', '_MENT14D', '_HLTHPLN', '_HCVU652', '_TOTINDA', '_RFHYPE6', '_CHOLCH3', '_RFCHOL3', '_MICHD', '_DRDXAR3', '_LMTACT3', '_LMTWRK3', '_SMOKER3', '_RFSMOK3', '_CURECI1', 'DRNKANY5', '_RFBING5', '_RFDRHV7', '_AIDTST4', '_FRTLT1A', '_VEGLT1A', '_INCOMG1', '_EDUCAG', 'CHILDREN', '_RFBMI5', '_BMI5CAT', '_AGE_G', '_AGE80', '_AGE65YR', '_AGEG5YR', '_SEX', '_RACEPRV', '_RACEGR3', '_RACEG21', '_RACE', '_HISPANC', '_MRACE1', '_PRACE1']\n"
     ]
    }
   ],
   "source": [
    "# Store all categorical only columns in a list\n",
    "object_cols = []\n",
    "for col, cat_con_var in cat_con_variables.items():\n",
    "    if cat_con_var == 1:\n",
    "        object_cols.append(col)\n",
    "\n",
    "print(f'Categorical Columns: \\n{object_cols}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "for dataframe in dataframes:\n",
    "    dataframe['_MICHD'] = dataframe['_MICHD'].replace({1.0: 0, 2.0: 1})\n",
    "    dataframe['_MICHD'] = dataframe['_MICHD'].astype('int64')\n",
    "\n",
    "    # Convert categorical columns to integers. Categorical columns have float numerical values representing their categories, e.g 1.0, 2.0, 3.0\n",
    "    for column, col_type in cat_con_variables.items():\n",
    "        if col_type == 1 and (column in dataframe.columns):  # Check if the column is categorical\n",
    "            dataframe[column] = dataframe[column].astype('int64')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use this if you want to utilize the one hot encoder\n",
    "# The one hot encoder will create sub-variables for each distinct value of a variable\n",
    "selected_cols = [col for col in set(object_cols) & set(dataframes[2].columns) if col != '_MICHD']\n",
    "\n",
    "dataframes[2] = pd.get_dummies(dataframes[2], columns=selected_cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "_MICHD\n",
       "1    153711\n",
       "0     13872\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataframes[2]['_MICHD'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(167583, 85)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataframes[2].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "32930"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Print the total number of duplicates for one of the categories of variable \"_MICHD\"\n",
    "dataframes[2][dataframes[2]['_MICHD'] == 1].duplicated().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataframes[2].info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Frequency of Variable Categories"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split variables into groups of five\n",
    "variable_groups = [dataframes[2].columns[i:i+5] for i in range(0, len(dataframes[2].columns), 5)]\n",
    "\n",
    "# Plot each group of variables\n",
    "for group in variable_groups:\n",
    "    plt.figure(figsize=(12, 6))\n",
    "    for idx, variable in enumerate(group):\n",
    "        plt.subplot(math.ceil(len(group)/2), 2, idx+1)\n",
    "        sns.countplot(data=dataframes[2], x=variable, hue=variable, palette='Set3', legend=False)\n",
    "        plt.title(f'Frequency of Categories in {variable}')\n",
    "        plt.xlabel('Categories')\n",
    "        plt.ylabel('Frequency')\n",
    "        plt.xticks(rotation=45)\n",
    "    plt.tight_layout()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Chi-Squared Test Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   Independent Variable  Chi-square Value        P-value\n",
      "0               _RFHLTH       7650.663883   0.000000e+00\n",
      "1              _PHYS14D       4268.662883   0.000000e+00\n",
      "2              _MENT14D        464.365123  2.515920e-100\n",
      "3              _HLTHPLN         96.618771   1.045925e-21\n",
      "4              _HCVU652       5983.420192   0.000000e+00\n",
      "5              _TOTINDA       1663.766148   0.000000e+00\n",
      "6              _RFHYPE6       6697.390558   0.000000e+00\n",
      "7              _CHOLCH3        262.036886   6.173480e-59\n",
      "8              _RFCHOL3       4740.451737   0.000000e+00\n",
      "9              _SMOKER3       2021.200796   0.000000e+00\n",
      "10             _RFBING5        651.940519  2.709661e-142\n",
      "11             _RFDRHV7        120.394056   7.190565e-27\n",
      "12             _AIDTST4        200.666633   2.665596e-44\n",
      "13             _FRTLT1A         35.432240   2.022957e-08\n",
      "14             _VEGLT1A        146.811814   1.318924e-32\n",
      "15             _INCOMG1       2093.487975   0.000000e+00\n",
      "16              _EDUCAG       1098.497436  1.602772e-236\n",
      "17              _RFBMI5        243.521990   6.710847e-55\n",
      "18             _BMI5CAT        323.220608   9.366207e-70\n",
      "19               _AGE_G       7845.859736   0.000000e+00\n",
      "20                 _SEX        880.096634  2.082702e-193\n",
      "21                _RACE        410.974811   8.404536e-84\n"
     ]
    }
   ],
   "source": [
    "# Create an empty DataFrame to store the results\n",
    "chi_squared_results = []\n",
    "\n",
    "\n",
    "for col in [col for col in columns[2] if col != '_MICHD']:\n",
    "    contingency_table = pd.crosstab(dataframes[2][col], dataframes[2]['_MICHD'])\n",
    "    chi2, p_value, dof, expected = chi2_contingency(contingency_table)\n",
    "    \n",
    "\n",
    "    # Append the results to the DataFrame\n",
    "    chi_squared_results.append({'Independent Variable': col,\n",
    "                              'Chi-square Value': chi2,\n",
    "                              'P-value': p_value})\n",
    "    \n",
    "    \"\"\"\n",
    "    print(f\"Chi-square test for independence between {col} and {'_MICHD'}:\")\n",
    "    print(f\"Chi-square value: {chi2}\")\n",
    "    print(f\"P-value: {p_value}\")\n",
    "    print(\"\")\n",
    "\n",
    "    # You can perform further analysis based on the p-value and significance level\n",
    "    # For example:\n",
    "    if p_value < 0.05:  # Assuming a significance level of 0.05\n",
    "        print(f\"The p-value is less than 0.05, indicating that {col} is likely dependent on {'_MICHD'}.\")\n",
    "    else:\n",
    "        print(f\"The p-value is greater than 0.05, indicating that {col} is likely independent of {'_MICHD'}.\")\n",
    "    print(\"\")\n",
    "    \"\"\"\n",
    "chi_squared_results_df = pd.DataFrame(chi_squared_results)\n",
    "print(chi_squared_results_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cramer's V Method for Correlation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Encode categorical variables\n",
    "encoded_df = pd.get_dummies(dataframes[2], drop_first=True)  # One-hot encoding\n",
    "\n",
    "# Function to calculate Cramér's V\n",
    "def cramers_v(x, y):\n",
    "    confusion_matrix = pd.crosstab(x, y)\n",
    "    chi2 = chi2_contingency(confusion_matrix)[0]\n",
    "    n = confusion_matrix.sum().sum()\n",
    "    phi2 = chi2 / n\n",
    "    r, k = confusion_matrix.shape\n",
    "    phi2corr = max(0, phi2 - ((k-1)*(r-1))/(n-1))\n",
    "    rcorr = r - ((r-1)**2)/(n-1)\n",
    "    kcorr = k - ((k-1)**2)/(n-1)\n",
    "    return np.sqrt(phi2corr / min((kcorr-1), (rcorr-1)))\n",
    "\n",
    "# Calculate Cramér's V for each pair of variables\n",
    "cramers_v_values = {}\n",
    "for col1 in encoded_df.columns:\n",
    "    for col2 in encoded_df.columns:\n",
    "        cramers_v_values[(col1, col2)] = cramers_v(encoded_df[col1], encoded_df[col2])\n",
    "\n",
    "\n",
    "# Create a correlation table\n",
    "correlation_table = pd.DataFrame(index=encoded_df.columns, columns=encoded_df.columns)\n",
    "for col1 in encoded_df.columns:\n",
    "    for col2 in encoded_df.columns:\n",
    "        correlation_table.loc[col1, col2] = cramers_v_values[(col1, col2)]\n",
    "\n",
    "# Create an empty dictionary to store associations\n",
    "association_dict = {}\n",
    "\n",
    "# Iterate over each variable in the correlation table\n",
    "for var1 in correlation_table.columns:\n",
    "    # Create an inner dictionary to store associations with other variables\n",
    "    inner_dict = {}\n",
    "    \n",
    "    # Iterate over other variables\n",
    "    for var2 in correlation_table.columns:\n",
    "        # Skip the current variable\n",
    "        if var1 == var2:\n",
    "            continue\n",
    "        \n",
    "        # Get the correlation value between var1 and var2 from the original correlation table\n",
    "        association_value = correlation_table.loc[var1, var2]\n",
    "        \n",
    "        # Store the association value in the inner dictionary\n",
    "        inner_dict[var2] = association_value\n",
    "    \n",
    "    # Sort the inner dictionary based on correlation values\n",
    "    sorted_inner_dict = {k: v for k, v in sorted(inner_dict.items(), key=lambda item: item[1], reverse=True)}\n",
    "\n",
    "    # Store the inner dictionary in the outer dictionary\n",
    "    association_dict[var1] = sorted_inner_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_RFHLTH:\n",
      "  _PHYS14D: 0.3614064284613051\n",
      "  _MICHD: 0.21363829499934134\n",
      "  _INCOMG1: 0.1935867208013684\n",
      "  _TOTINDA: 0.18029964894301873\n",
      "  _MENT14D: 0.17143570592315746\n",
      "  _EDUCAG: 0.15204789025587417\n",
      "  _RFHYPE6: 0.13517998796401912\n",
      "  _SMOKER3: 0.10057197701484735\n",
      "  _BMI5CAT: 0.0977096647006075\n",
      "  _RFCHOL3: 0.09291320987420845\n",
      "  _AGE_G: 0.07929579365029235\n",
      "  _RFBMI5: 0.0684755915778215\n",
      "  _VEGLT1A: 0.06726165464569893\n",
      "  _RACE: 0.06659320210777403\n",
      "  _HCVU652: 0.06226586351954926\n",
      "  _FRTLT1A: 0.041714917712912046\n",
      "  _RFBING5: 0.04004045008712682\n",
      "  _AIDTST4: 0.03235917549083652\n",
      "  _HLTHPLN: 0.029593844701554028\n",
      "  _CHOLCH3: 0.021763368271458052\n",
      "  _RFDRHV7: 0.020303714527265388\n",
      "  _SEX: 0.018633063685266255\n",
      "_PHYS14D:\n",
      "  _RFHLTH: 0.36140642846130516\n",
      "  _MENT14D: 0.2259787574960069\n",
      "  _TOTINDA: 0.16818639229380167\n",
      "  _MICHD: 0.15954367057453467\n",
      "  _INCOMG1: 0.13090327181619776\n",
      "  _RFHYPE6: 0.09931600485496782\n",
      "  _EDUCAG: 0.0831936525736148\n",
      "  _RFCHOL3: 0.07530743213427221\n",
      "  _SMOKER3: 0.06985201978885339\n",
      "  _VEGLT1A: 0.06855089609141606\n",
      "  _SEX: 0.06622007388379528\n",
      "  _AGE_G: 0.06599047070433955\n",
      "  _BMI5CAT: 0.06594503808907168\n",
      "  _HCVU652: 0.05767722528567192\n",
      "  _RFBING5: 0.05024061790337511\n",
      "  _RFBMI5: 0.05002285790132909\n",
      "  _FRTLT1A: 0.046677185202963725\n",
      "  _AIDTST4: 0.045866544006086266\n",
      "  _RACE: 0.03337471515390333\n",
      "  _HLTHPLN: 0.02800163324181061\n",
      "  _RFDRHV7: 0.027919003212383598\n",
      "  _CHOLCH3: 0.024446030081562813\n",
      "_MENT14D:\n",
      "  _PHYS14D: 0.22597875749600685\n",
      "  _RFHLTH: 0.17143570592315746\n",
      "  _AGE_G: 0.14342705563066327\n",
      "  _HCVU652: 0.1417079186020869\n",
      "  _SEX: 0.1286422681342891\n",
      "  _INCOMG1: 0.10089409119423892\n",
      "  _AIDTST4: 0.10013423439921575\n",
      "  _TOTINDA: 0.08584669558639717\n",
      "  _SMOKER3: 0.0764505797086261\n",
      "  _RFBING5: 0.07373321434795055\n",
      "  _EDUCAG: 0.07013415426771347\n",
      "  _VEGLT1A: 0.059897925636334254\n",
      "  _FRTLT1A: 0.05816526783483396\n",
      "  _MICHD: 0.05246971284832731\n",
      "  _BMI5CAT: 0.05038990206022836\n",
      "  _RFDRHV7: 0.04520335285068367\n",
      "  _RFHYPE6: 0.044771136032782086\n",
      "  _HLTHPLN: 0.038496652169306485\n",
      "  _RACE: 0.03787587235163304\n",
      "  _RFBMI5: 0.02714307781069763\n",
      "  _CHOLCH3: 0.025887036039715147\n",
      "  _RFCHOL3: 0.025464141334490036\n",
      "_HLTHPLN:\n",
      "  _HCVU652: 0.7021287312428877\n",
      "  _RACE: 0.14313399097653334\n",
      "  _CHOLCH3: 0.14225712337988047\n",
      "  _INCOMG1: 0.13195709525132313\n",
      "  _EDUCAG: 0.1315967296117762\n",
      "  _AGE_G: 0.12379677865686037\n",
      "  _SMOKER3: 0.05618860796758344\n",
      "  _VEGLT1A: 0.04714693698157233\n",
      "  _RFHYPE6: 0.04393816078364641\n",
      "  _RFCHOL3: 0.04303841520117264\n",
      "  _MENT14D: 0.038496652169306485\n",
      "  _AIDTST4: 0.03816117448744557\n",
      "  _RFBING5: 0.03471115861574111\n",
      "  _SEX: 0.03140156531288751\n",
      "  _FRTLT1A: 0.030362616703895957\n",
      "  _RFHLTH: 0.029593844701554028\n",
      "  _RFDRHV7: 0.028471559116438915\n",
      "  _TOTINDA: 0.02807974883104035\n",
      "  _PHYS14D: 0.02800163324181061\n",
      "  _MICHD: 0.023761559432600807\n",
      "  _BMI5CAT: 0.01785023249549455\n",
      "  _RFBMI5: 0.01765441943785564\n",
      "_HCVU652:\n",
      "  _HLTHPLN: 0.7021287312428877\n",
      "  _AGE_G: 0.6719814604377053\n",
      "  _INCOMG1: 0.20170008737270237\n",
      "  _RFHYPE6: 0.19242481801997774\n",
      "  _MICHD: 0.18892458252528632\n",
      "  _AIDTST4: 0.1829835284142059\n",
      "  _RACE: 0.17153582526519806\n",
      "  _CHOLCH3: 0.14970510480058127\n",
      "  _RFCHOL3: 0.14602203365329747\n",
      "  _MENT14D: 0.1417079186020869\n",
      "  _RFBING5: 0.12991356593431772\n",
      "  _SMOKER3: 0.12495540329553781\n",
      "  _EDUCAG: 0.1247839130661196\n",
      "  _TOTINDA: 0.0713438518708386\n",
      "  _RFHLTH: 0.06226586351954926\n",
      "  _PHYS14D: 0.05767722528567192\n",
      "  _FRTLT1A: 0.05380370473882116\n",
      "  _BMI5CAT: 0.04791869653059837\n",
      "  _VEGLT1A: 0.04302335070099706\n",
      "  _SEX: 0.04119457802012976\n",
      "  _RFDRHV7: 0.03954008963578854\n",
      "  _RFBMI5: 0.03626961675601535\n",
      "_TOTINDA:\n",
      "  _RFHLTH: 0.18029964894301873\n",
      "  _PHYS14D: 0.16818639229380167\n",
      "  _INCOMG1: 0.16345523503531145\n",
      "  _EDUCAG: 0.16070175937431813\n",
      "  _VEGLT1A: 0.10502479067504433\n",
      "  _BMI5CAT: 0.1019770764090294\n",
      "  _MICHD: 0.09957980413088452\n",
      "  _RFHYPE6: 0.0917244332508616\n",
      "  _RFBMI5: 0.08657332559513206\n",
      "  _SMOKER3: 0.08640726896690365\n",
      "  _MENT14D: 0.08584669558639717\n",
      "  _FRTLT1A: 0.0828857341163244\n",
      "  _AGE_G: 0.07833276824934479\n",
      "  _HCVU652: 0.0713438518708386\n",
      "  _SEX: 0.06731317594979579\n",
      "  _RACE: 0.06167080515851505\n",
      "  _RFCHOL3: 0.05394259920127644\n",
      "  _RFBING5: 0.03881136371051177\n",
      "  _HLTHPLN: 0.028079748831040353\n",
      "  _RFDRHV7: 0.018141855648807406\n",
      "  _AIDTST4: 0.013873903018422779\n",
      "  _CHOLCH3: 0.0\n",
      "_RFHYPE6:\n",
      "  _AGE_G: 0.2494228005895958\n",
      "  _RFCHOL3: 0.20679772316732292\n",
      "  _MICHD: 0.1998822699273013\n",
      "  _HCVU652: 0.1924248180199777\n",
      "  _RFBMI5: 0.17903924349608533\n",
      "  _BMI5CAT: 0.1497490469743976\n",
      "  _RFHLTH: 0.13517998796401912\n",
      "  _INCOMG1: 0.10307944758142625\n",
      "  _PHYS14D: 0.0993160048549678\n",
      "  _TOTINDA: 0.0917244332508616\n",
      "  _CHOLCH3: 0.09030024939894991\n",
      "  _SMOKER3: 0.08391948081573296\n",
      "  _RACE: 0.0822536312894928\n",
      "  _EDUCAG: 0.079185769984511\n",
      "  _SEX: 0.05516826042698035\n",
      "  _RFBING5: 0.04726054930981198\n",
      "  _MENT14D: 0.044771136032782086\n",
      "  _AIDTST4: 0.044493778526897884\n",
      "  _HLTHPLN: 0.04393816078364641\n",
      "  _VEGLT1A: 0.02783074985121571\n",
      "  _FRTLT1A: 0.022814793849226193\n",
      "  _RFDRHV7: 0.013186443862152933\n",
      "_CHOLCH3:\n",
      "  _HCVU652: 0.14970510480058127\n",
      "  _HLTHPLN: 0.14225712337988047\n",
      "  _AGE_G: 0.10645887681532712\n",
      "  _RFCHOL3: 0.09180814278143222\n",
      "  _RFHYPE6: 0.09030024939894991\n",
      "  _SMOKER3: 0.06268917811925308\n",
      "  _RFBING5: 0.0498040406720533\n",
      "  _SEX: 0.049533734279837816\n",
      "  _MICHD: 0.03946728490894508\n",
      "  _EDUCAG: 0.03197525361978724\n",
      "  _FRTLT1A: 0.02960116607354173\n",
      "  _RFDRHV7: 0.029029739101836788\n",
      "  _RACE: 0.027520248785880184\n",
      "  _BMI5CAT: 0.02747736066797396\n",
      "  _MENT14D: 0.025887036039715147\n",
      "  _PHYS14D: 0.024446030081562813\n",
      "  _RFBMI5: 0.02342687892597176\n",
      "  _RFHLTH: 0.021763368271458052\n",
      "  _INCOMG1: 0.021160969526398992\n",
      "  _VEGLT1A: 0.016127734367114305\n",
      "  _AIDTST4: 0.008313578954389608\n",
      "  _TOTINDA: 0.0\n",
      "_RFCHOL3:\n",
      "  _AGE_G: 0.2095909144184728\n",
      "  _RFHYPE6: 0.20679772316732292\n",
      "  _MICHD: 0.16815297123566464\n",
      "  _HCVU652: 0.14602203365329747\n",
      "  _RFBMI5: 0.11111372421359493\n",
      "  _RFHLTH: 0.09291320987420845\n",
      "  _CHOLCH3: 0.09180814278143222\n",
      "  _BMI5CAT: 0.08194833448390741\n",
      "  _PHYS14D: 0.07530743213427221\n",
      "  _SMOKER3: 0.06937518939687554\n",
      "  _TOTINDA: 0.05394259920127644\n",
      "  _INCOMG1: 0.05033138057440106\n",
      "  _RACE: 0.04448004831956392\n",
      "  _HLTHPLN: 0.04303841520117264\n",
      "  _RFBING5: 0.04251037825408345\n",
      "  _AIDTST4: 0.04057477330208641\n",
      "  _MENT14D: 0.025464141334490036\n",
      "  _VEGLT1A: 0.024945459521937777\n",
      "  _EDUCAG: 0.02416743851045223\n",
      "  _FRTLT1A: 0.02407600166702732\n",
      "  _SEX: 0.020334594925421456\n",
      "  _RFDRHV7: 0.014001520324039696\n",
      "_MICHD:\n",
      "  _AGE_G: 0.21630579617609447\n",
      "  _RFHLTH: 0.21363829499934134\n",
      "  _RFHYPE6: 0.1998822699273013\n",
      "  _HCVU652: 0.18892458252528632\n",
      "  _RFCHOL3: 0.16815297123566464\n",
      "  _PHYS14D: 0.1595436705745347\n",
      "  _INCOMG1: 0.11158202893663714\n",
      "  _SMOKER3: 0.10971370422371139\n",
      "  _TOTINDA: 0.09957980413088452\n",
      "  _EDUCAG: 0.0808153177654357\n",
      "  _SEX: 0.07242768805490152\n",
      "  _RFBING5: 0.06227634528516087\n",
      "  _MENT14D: 0.05246971284832731\n",
      "  _RACE: 0.049037152102378585\n",
      "  _BMI5CAT: 0.043713033150034476\n",
      "  _CHOLCH3: 0.03946728490894508\n",
      "  _RFBMI5: 0.03804187124883185\n",
      "  _AIDTST4: 0.03443093117396074\n",
      "  _VEGLT1A: 0.02939600374108021\n",
      "  _RFDRHV7: 0.026579771450664473\n",
      "  _HLTHPLN: 0.023761559432600807\n",
      "  _FRTLT1A: 0.014124366528096708\n",
      "_SMOKER3:\n",
      "  _HCVU652: 0.12495540329553782\n",
      "  _EDUCAG: 0.1145215848149265\n",
      "  _MICHD: 0.10971370422371139\n",
      "  _AGE_G: 0.10858918801837908\n",
      "  _RFHLTH: 0.10057197701484735\n",
      "  _INCOMG1: 0.09732210233918918\n",
      "  _RFBING5: 0.09168508272229992\n",
      "  _RFDRHV7: 0.09121744077591848\n",
      "  _TOTINDA: 0.08640726896690366\n",
      "  _RFHYPE6: 0.08391948081573296\n",
      "  _AIDTST4: 0.08062621309373905\n",
      "  _MENT14D: 0.0764505797086261\n",
      "  _FRTLT1A: 0.07244228516040423\n",
      "  _SEX: 0.07039820414293561\n",
      "  _RACE: 0.07020796252442162\n",
      "  _PHYS14D: 0.06985201978885337\n",
      "  _RFCHOL3: 0.06937518939687554\n",
      "  _CHOLCH3: 0.06268917811925308\n",
      "  _HLTHPLN: 0.05618860796758344\n",
      "  _RFBMI5: 0.04633909252862637\n",
      "  _VEGLT1A: 0.04367425377160284\n",
      "  _BMI5CAT: 0.03400219846827467\n",
      "_RFBING5:\n",
      "  _RFDRHV7: 0.5742777468111351\n",
      "  _AGE_G: 0.15849860071464336\n",
      "  _HCVU652: 0.12991356593431772\n",
      "  _SEX: 0.09859678588069172\n",
      "  _SMOKER3: 0.09168508272229993\n",
      "  _INCOMG1: 0.09124363499258806\n",
      "  _FRTLT1A: 0.07715134370588378\n",
      "  _MENT14D: 0.07373321434795055\n",
      "  _VEGLT1A: 0.06850946722256404\n",
      "  _MICHD: 0.06227634528516088\n",
      "  _AIDTST4: 0.05842664053193804\n",
      "  _PHYS14D: 0.05024061790337512\n",
      "  _CHOLCH3: 0.0498040406720533\n",
      "  _RFHYPE6: 0.04726054930981199\n",
      "  _RFCHOL3: 0.04251037825408345\n",
      "  _RACE: 0.04027981632347155\n",
      "  _RFHLTH: 0.040040450087126825\n",
      "  _TOTINDA: 0.03881136371051177\n",
      "  _HLTHPLN: 0.03471115861574111\n",
      "  _EDUCAG: 0.032807082126146844\n",
      "  _BMI5CAT: 0.015778997483872705\n",
      "  _RFBMI5: 0.015073485705989672\n",
      "_RFDRHV7:\n",
      "  _RFBING5: 0.5742777468111351\n",
      "  _SMOKER3: 0.09121744077591847\n",
      "  _INCOMG1: 0.07033052541613223\n",
      "  _VEGLT1A: 0.06880145786633995\n",
      "  _FRTLT1A: 0.065454297799917\n",
      "  _RACE: 0.04538946980079897\n",
      "  _MENT14D: 0.045203352850683665\n",
      "  _HCVU652: 0.03954008963578854\n",
      "  _AGE_G: 0.037153336090860335\n",
      "  _EDUCAG: 0.033977940118562845\n",
      "  _SEX: 0.02923398409430249\n",
      "  _CHOLCH3: 0.029029739101836788\n",
      "  _HLTHPLN: 0.028471559116438915\n",
      "  _PHYS14D: 0.027919003212383605\n",
      "  _MICHD: 0.026579771450664473\n",
      "  _AIDTST4: 0.025234583203713475\n",
      "  _RFHLTH: 0.020303714527265385\n",
      "  _RFBMI5: 0.01911288332534394\n",
      "  _BMI5CAT: 0.01823064394797814\n",
      "  _TOTINDA: 0.018141855648807406\n",
      "  _RFCHOL3: 0.014001520324039696\n",
      "  _RFHYPE6: 0.013186443862152933\n",
      "_AIDTST4:\n",
      "  _AGE_G: 0.20708505535551572\n",
      "  _HCVU652: 0.1829835284142059\n",
      "  _RACE: 0.13832477447975616\n",
      "  _MENT14D: 0.10013423439921575\n",
      "  _SMOKER3: 0.08062621309373905\n",
      "  _INCOMG1: 0.07550759640078651\n",
      "  _RFBING5: 0.05842664053193804\n",
      "  _PHYS14D: 0.045866544006086266\n",
      "  _RFHYPE6: 0.044493778526897884\n",
      "  _RFCHOL3: 0.04057477330208641\n",
      "  _HLTHPLN: 0.03816117448744557\n",
      "  _MICHD: 0.03443093117396074\n",
      "  _RFHLTH: 0.03235917549083652\n",
      "  _BMI5CAT: 0.031768246680280804\n",
      "  _EDUCAG: 0.03164167409667444\n",
      "  _RFBMI5: 0.029822373380535793\n",
      "  _RFDRHV7: 0.025234583203713475\n",
      "  _FRTLT1A: 0.016631080436661904\n",
      "  _TOTINDA: 0.013873903018422779\n",
      "  _VEGLT1A: 0.013148774358180383\n",
      "  _SEX: 0.011494045366991728\n",
      "  _CHOLCH3: 0.008313578954389608\n",
      "_FRTLT1A:\n",
      "  _VEGLT1A: 0.3497812043723713\n",
      "  _TOTINDA: 0.0828857341163244\n",
      "  _EDUCAG: 0.07946237424182954\n",
      "  _RFBING5: 0.07715134370588378\n",
      "  _SMOKER3: 0.07244228516040423\n",
      "  _INCOMG1: 0.07033141664649455\n",
      "  _RFBMI5: 0.06961202956146942\n",
      "  _RFDRHV7: 0.065454297799917\n",
      "  _BMI5CAT: 0.06327991222086313\n",
      "  _MENT14D: 0.05816526783483396\n",
      "  _SEX: 0.056817376729589586\n",
      "  _AGE_G: 0.05590084458572916\n",
      "  _HCVU652: 0.05380370473882116\n",
      "  _PHYS14D: 0.046677185202963725\n",
      "  _RFHLTH: 0.041714917712912046\n",
      "  _RACE: 0.03415122589867544\n",
      "  _HLTHPLN: 0.030362616703895957\n",
      "  _CHOLCH3: 0.02960116607354173\n",
      "  _RFCHOL3: 0.02407600166702732\n",
      "  _RFHYPE6: 0.022814793849226193\n",
      "  _AIDTST4: 0.016631080436661904\n",
      "  _MICHD: 0.014124366528096708\n",
      "_VEGLT1A:\n",
      "  _FRTLT1A: 0.3497812043723713\n",
      "  _INCOMG1: 0.11733610335750481\n",
      "  _EDUCAG: 0.11298520332434593\n",
      "  _TOTINDA: 0.10502479067504433\n",
      "  _RACE: 0.09092335765428111\n",
      "  _RFDRHV7: 0.06880145786633995\n",
      "  _PHYS14D: 0.06855089609141606\n",
      "  _RFBING5: 0.06850946722256404\n",
      "  _RFHLTH: 0.06726165464569893\n",
      "  _MENT14D: 0.059897925636334254\n",
      "  _SEX: 0.0496040954643402\n",
      "  _HLTHPLN: 0.04714693698157233\n",
      "  _SMOKER3: 0.04367425377160284\n",
      "  _HCVU652: 0.04302335070099706\n",
      "  _AGE_G: 0.036906026840996614\n",
      "  _RFBMI5: 0.03556357986685795\n",
      "  _BMI5CAT: 0.03284109066084437\n",
      "  _MICHD: 0.02939600374108021\n",
      "  _RFHYPE6: 0.02783074985121571\n",
      "  _RFCHOL3: 0.024945459521937777\n",
      "  _CHOLCH3: 0.016127734367114302\n",
      "  _AIDTST4: 0.013148774358180383\n",
      "_INCOMG1:\n",
      "  _EDUCAG: 0.21736483137323562\n",
      "  _HCVU652: 0.20170008737270237\n",
      "  _RFHLTH: 0.19358672080136843\n",
      "  _TOTINDA: 0.16345523503531145\n",
      "  _HLTHPLN: 0.13195709525132313\n",
      "  _PHYS14D: 0.1309032718161978\n",
      "  _SEX: 0.11830634994056119\n",
      "  _VEGLT1A: 0.11733610335750481\n",
      "  _AGE_G: 0.11708373610311748\n",
      "  _MICHD: 0.11158202893663714\n",
      "  _RFHYPE6: 0.10307944758142625\n",
      "  _MENT14D: 0.10089409119423892\n",
      "  _SMOKER3: 0.09732210233918917\n",
      "  _RACE: 0.09175380165482222\n",
      "  _RFBING5: 0.09124363499258806\n",
      "  _AIDTST4: 0.07550759640078651\n",
      "  _RFBMI5: 0.07086016971213167\n",
      "  _FRTLT1A: 0.07033141664649455\n",
      "  _RFDRHV7: 0.07033052541613224\n",
      "  _BMI5CAT: 0.06577493156643704\n",
      "  _RFCHOL3: 0.05033138057440106\n",
      "  _CHOLCH3: 0.021160969526398992\n",
      "_EDUCAG:\n",
      "  _INCOMG1: 0.21736483137323562\n",
      "  _TOTINDA: 0.16070175937431813\n",
      "  _RFHLTH: 0.15204789025587417\n",
      "  _RACE: 0.1378512193204756\n",
      "  _HLTHPLN: 0.13159672961177624\n",
      "  _HCVU652: 0.1247839130661196\n",
      "  _SMOKER3: 0.1145215848149265\n",
      "  _VEGLT1A: 0.11298520332434593\n",
      "  _PHYS14D: 0.08319365257361483\n",
      "  _MICHD: 0.08081531776543571\n",
      "  _RFBMI5: 0.08029109092888015\n",
      "  _FRTLT1A: 0.07946237424182954\n",
      "  _RFHYPE6: 0.079185769984511\n",
      "  _MENT14D: 0.07013415426771347\n",
      "  _BMI5CAT: 0.06699213304723627\n",
      "  _AGE_G: 0.06411550152782822\n",
      "  _SEX: 0.03577649407196621\n",
      "  _RFDRHV7: 0.033977940118562845\n",
      "  _RFBING5: 0.032807082126146844\n",
      "  _CHOLCH3: 0.03197525361978724\n",
      "  _AIDTST4: 0.03164167409667444\n",
      "  _RFCHOL3: 0.024167438510452232\n",
      "_RFBMI5:\n",
      "  _BMI5CAT: 0.9999940327186039\n",
      "  _RFHYPE6: 0.17903924349608533\n",
      "  _AGE_G: 0.1139071986410422\n",
      "  _SEX: 0.11285530566352975\n",
      "  _RFCHOL3: 0.11111372421359493\n",
      "  _RACE: 0.09997380868084825\n",
      "  _TOTINDA: 0.08657332559513206\n",
      "  _EDUCAG: 0.08029109092888016\n",
      "  _INCOMG1: 0.07086016971213167\n",
      "  _FRTLT1A: 0.06961202956146942\n",
      "  _RFHLTH: 0.0684755915778215\n",
      "  _PHYS14D: 0.05002285790132909\n",
      "  _SMOKER3: 0.04633909252862637\n",
      "  _MICHD: 0.03804187124883185\n",
      "  _HCVU652: 0.03626961675601535\n",
      "  _VEGLT1A: 0.03556357986685795\n",
      "  _AIDTST4: 0.029822373380535793\n",
      "  _MENT14D: 0.02714307781069763\n",
      "  _CHOLCH3: 0.02342687892597176\n",
      "  _RFDRHV7: 0.01911288332534394\n",
      "  _HLTHPLN: 0.01765441943785564\n",
      "  _RFBING5: 0.015073485705989672\n",
      "_BMI5CAT:\n",
      "  _RFBMI5: 0.9999940327186039\n",
      "  _RFHYPE6: 0.1497490469743976\n",
      "  _SEX: 0.1335238702774991\n",
      "  _TOTINDA: 0.10197707640902942\n",
      "  _RFHLTH: 0.09770966470060749\n",
      "  _RFCHOL3: 0.08194833448390741\n",
      "  _AGE_G: 0.07622029817115206\n",
      "  _RACE: 0.06963210499750819\n",
      "  _EDUCAG: 0.06699213304723628\n",
      "  _PHYS14D: 0.06594503808907168\n",
      "  _INCOMG1: 0.06577493156643704\n",
      "  _FRTLT1A: 0.06327991222086311\n",
      "  _MENT14D: 0.05038990206022836\n",
      "  _HCVU652: 0.04791869653059837\n",
      "  _MICHD: 0.043713033150034476\n",
      "  _SMOKER3: 0.034002198468274666\n",
      "  _VEGLT1A: 0.03284109066084437\n",
      "  _AIDTST4: 0.031768246680280804\n",
      "  _CHOLCH3: 0.02747736066797396\n",
      "  _RFDRHV7: 0.01823064394797814\n",
      "  _HLTHPLN: 0.01785023249549455\n",
      "  _RFBING5: 0.015778997483872705\n",
      "_AGE_G:\n",
      "  _HCVU652: 0.6719814604377053\n",
      "  _RFHYPE6: 0.24942280058959576\n",
      "  _MICHD: 0.21630579617609447\n",
      "  _RFCHOL3: 0.20959091441847277\n",
      "  _AIDTST4: 0.20708505535551572\n",
      "  _RFBING5: 0.15849860071464336\n",
      "  _MENT14D: 0.14342705563066327\n",
      "  _HLTHPLN: 0.12379677865686037\n",
      "  _INCOMG1: 0.1170837361031175\n",
      "  _RFBMI5: 0.1139071986410422\n",
      "  _RACE: 0.11362212547030479\n",
      "  _SMOKER3: 0.10858918801837908\n",
      "  _CHOLCH3: 0.10645887681532713\n",
      "  _RFHLTH: 0.07929579365029235\n",
      "  _TOTINDA: 0.0783327682493448\n",
      "  _BMI5CAT: 0.07622029817115206\n",
      "  _PHYS14D: 0.06599047070433955\n",
      "  _EDUCAG: 0.06411550152782823\n",
      "  _FRTLT1A: 0.05590084458572916\n",
      "  _SEX: 0.05201261096512566\n",
      "  _RFDRHV7: 0.037153336090860335\n",
      "  _VEGLT1A: 0.036906026840996614\n",
      "_SEX:\n",
      "  _BMI5CAT: 0.13352387027749912\n",
      "  _MENT14D: 0.12864226813428914\n",
      "  _INCOMG1: 0.11830634994056119\n",
      "  _RFBMI5: 0.11285530566352978\n",
      "  _RFBING5: 0.09859678588069173\n",
      "  _MICHD: 0.07242768805490152\n",
      "  _SMOKER3: 0.07039820414293561\n",
      "  _RACE: 0.06814335840865235\n",
      "  _TOTINDA: 0.06731317594979579\n",
      "  _PHYS14D: 0.06622007388379528\n",
      "  _FRTLT1A: 0.056817376729589586\n",
      "  _RFHYPE6: 0.05516826042698035\n",
      "  _AGE_G: 0.052012610965125654\n",
      "  _VEGLT1A: 0.0496040954643402\n",
      "  _CHOLCH3: 0.049533734279837816\n",
      "  _HCVU652: 0.04119457802012976\n",
      "  _EDUCAG: 0.0357764940719662\n",
      "  _HLTHPLN: 0.03140156531288751\n",
      "  _RFDRHV7: 0.02923398409430249\n",
      "  _RFCHOL3: 0.020334594925421456\n",
      "  _RFHLTH: 0.018633063685266255\n",
      "  _AIDTST4: 0.011494045366991728\n",
      "_RACE:\n",
      "  _HCVU652: 0.17153582526519806\n",
      "  _HLTHPLN: 0.14313399097653334\n",
      "  _AIDTST4: 0.13832477447975616\n",
      "  _EDUCAG: 0.1378512193204756\n",
      "  _AGE_G: 0.11362212547030479\n",
      "  _RFBMI5: 0.09997380868084825\n",
      "  _INCOMG1: 0.09175380165482222\n",
      "  _VEGLT1A: 0.09092335765428108\n",
      "  _RFHYPE6: 0.0822536312894928\n",
      "  _SMOKER3: 0.0702079625244216\n",
      "  _BMI5CAT: 0.0696321049975082\n",
      "  _SEX: 0.06814335840865235\n",
      "  _RFHLTH: 0.06659320210777403\n",
      "  _TOTINDA: 0.06167080515851505\n",
      "  _MICHD: 0.04903715210237858\n",
      "  _RFDRHV7: 0.04538946980079898\n",
      "  _RFCHOL3: 0.044480048319563925\n",
      "  _RFBING5: 0.04027981632347156\n",
      "  _MENT14D: 0.03787587235163305\n",
      "  _FRTLT1A: 0.03415122589867544\n",
      "  _PHYS14D: 0.03337471515390333\n",
      "  _CHOLCH3: 0.027520248785880184\n"
     ]
    }
   ],
   "source": [
    "for var1, inner_dict in association_dict.items():\n",
    "    print(f\"{var1}:\")\n",
    "    for var2, association_value in inner_dict.items():\n",
    "        print(f\"  {var2}: {association_value}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dataset Sample Creation\n",
    "Create a sample for one of the dataframes for later feature extraction and classification. Another method we can use to take a sample is to split the original dataframe and apply a combination of oversampling and undersampling to the training set."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First way to create a sample: \n",
    "- Take all the samples from the minority class\n",
    "- Define a percentage of samples that you want to take and subtract the number of minority class samples; that will be the number of majority class samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Take all minority samples. Including duplicates\n",
    "minority_samples = dataframes[2][dataframes[2]['_MICHD'] == 0] \n",
    "num_minority_samples = len(minority_samples)\n",
    "\n",
    "\"\"\"\n",
    "Lets say that we want to take p% of the original dataframe as a sample. The first step we have to take is subtract the number of minority samples assuming we have already included them in our sample.\n",
    "The second step is to add the necessary number of majority samples in our sample until we reach the upper limit which is p% of samples. We can add some of the unique samples from the majority class \n",
    "to the sample or we can randomly pick some of them. Below, we add unique samples to the sample.\n",
    "\"\"\"\n",
    "num_majority_samples =  math.floor(dataframes[2].shape[0] * 0.25) - num_minority_samples\n",
    "majority_unique_samples = dataframes[2][dataframes[2]['_MICHD'] == 1].drop_duplicates()\n",
    "num_unique_majority_samples = len(majority_unique_samples) # Not needed\n",
    "\n",
    "majority_samples = majority_unique_samples.sample(n=num_majority_samples)\n",
    "\n",
    "dataframe_sample = pd.concat([minority_samples, majority_samples])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Second way to create a sample:\n",
    "- Perform stratified sampling using a percentage of samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Specify the fraction of samples you want to retain in the subset (e.g., 0.5 for 50%)\n",
    "subset_fraction = 0.20\n",
    "\n",
    "# Perform stratified sampling to create a smaller dataframe\n",
    "dataframe_sample = dataframes[2].groupby('_MICHD', group_keys = False).apply(lambda x: x.sample(frac = subset_fraction))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Third way to create a sample:\n",
    "- Undersample the majority class so we can keep all instances of the minority class, thereby reducing the overall size of the dataset. The downside of this approach is we lose valuable instances of the majority class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the undersampler\n",
    "undersampler = RandomUnderSampler(random_state=SEED)\n",
    "\n",
    "# Separate features (X) and target variable (y)\n",
    "x = dataframes[2].drop(columns=['_MICHD'])\n",
    "y = dataframes[2]['_MICHD']\n",
    "\n",
    "# Perform undersampling\n",
    "x_resampled, y_resampled = undersampler.fit_resample(x, y)\n",
    "\n",
    "# Construct the balanced DataFrame\n",
    "dataframe_sample = pd.concat([x_resampled, y_resampled], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(27744, 23)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataframe_sample.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "_MICHD\n",
       "0    13872\n",
       "1    13872\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataframe_sample['_MICHD'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataframe_sample.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Importance with Random Forests\n",
    "We can use the GridSearchCV, or the RandomizedSearchCV method with the Random Forest algorithm to evaluate the importance of each variable in assessing the risk of MI or CHD."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "target_column = \"_MICHD\"\n",
    "\n",
    "#y = dataframes[2][target_column]\n",
    "#x = dataframes[2].drop(columns=[target_column])\n",
    "\n",
    "y = dataframe_sample[target_column]\n",
    "x = dataframe_sample.drop(columns = [target_column])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=SEED, stratify=y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the cost matrix\n",
    "cost_matrix = [[0, 0.25],  # Cost of false positive \n",
    "               [1, 0]]  # Cost of false negative"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a custom scoring function using the cost matrix\n",
    "def custom_score(y_true, y_pred):\n",
    "    cm = confusion_matrix(y_true, y_pred)\n",
    "    cost = np.sum(cm * cost_matrix)\n",
    "    return -cost  # Minimize cost, so negative of cost is returned"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make the custom scoring function into a scorer object\n",
    "custom_scorer = make_scorer(custom_score, greater_is_better = False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Random Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the parameter distributions to sample from\n",
    "param_dist = {\n",
    "    'n_estimators': randint(50, 300),\n",
    "    'max_depth': [None] + list(range(10, 41)),\n",
    "    'min_samples_split': randint(2, 21),\n",
    "    'class_weight': [{0: 1, 1: 0.25}]\n",
    "}\n",
    "\n",
    "# Initialize a Random Forest classifier\n",
    "rf_classifier = RandomForestClassifier(random_state=SEED)\n",
    "\n",
    "# Initialize RandomizedSearchCV with custom scoring\n",
    "random_search = RandomizedSearchCV(estimator=rf_classifier, param_distributions=param_dist, \n",
    "                                   scoring=make_scorer(f1_score, average='weighted'), n_jobs=4, random_state=SEED)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Grid Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the parameter grid to search\n",
    "param_grid = {\n",
    "    'n_estimators': [100, 200, 300],\n",
    "    'max_depth': [None, 10, 20],\n",
    "    'class_weight': [{0: 1, 1: 0.25}]\n",
    "}\n",
    "\n",
    "# Initialize a Random Forest classifier\n",
    "rf_classifier = RandomForestClassifier(random_state=SEED)\n",
    "\n",
    "# Initialize GridSearchCV\n",
    "grid_search = GridSearchCV(estimator=rf_classifier, param_grid=param_grid, n_jobs=4, scoring=make_scorer(f1_score, average='weighted'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Initialize search, output the best parameters, and print feature importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-3 {\n",
       "  /* Definition of color scheme common for light and dark mode */\n",
       "  --sklearn-color-text: black;\n",
       "  --sklearn-color-line: gray;\n",
       "  /* Definition of color scheme for unfitted estimators */\n",
       "  --sklearn-color-unfitted-level-0: #fff5e6;\n",
       "  --sklearn-color-unfitted-level-1: #f6e4d2;\n",
       "  --sklearn-color-unfitted-level-2: #ffe0b3;\n",
       "  --sklearn-color-unfitted-level-3: chocolate;\n",
       "  /* Definition of color scheme for fitted estimators */\n",
       "  --sklearn-color-fitted-level-0: #f0f8ff;\n",
       "  --sklearn-color-fitted-level-1: #d4ebff;\n",
       "  --sklearn-color-fitted-level-2: #b3dbfd;\n",
       "  --sklearn-color-fitted-level-3: cornflowerblue;\n",
       "\n",
       "  /* Specific color for light theme */\n",
       "  --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, white)));\n",
       "  --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-icon: #696969;\n",
       "\n",
       "  @media (prefers-color-scheme: dark) {\n",
       "    /* Redefinition of color scheme for dark theme */\n",
       "    --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, #111)));\n",
       "    --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-icon: #878787;\n",
       "  }\n",
       "}\n",
       "\n",
       "#sk-container-id-3 {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 pre {\n",
       "  padding: 0;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 input.sk-hidden--visually {\n",
       "  border: 0;\n",
       "  clip: rect(1px 1px 1px 1px);\n",
       "  clip: rect(1px, 1px, 1px, 1px);\n",
       "  height: 1px;\n",
       "  margin: -1px;\n",
       "  overflow: hidden;\n",
       "  padding: 0;\n",
       "  position: absolute;\n",
       "  width: 1px;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-dashed-wrapped {\n",
       "  border: 1px dashed var(--sklearn-color-line);\n",
       "  margin: 0 0.4em 0.5em 0.4em;\n",
       "  box-sizing: border-box;\n",
       "  padding-bottom: 0.4em;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-container {\n",
       "  /* jupyter's `normalize.less` sets `[hidden] { display: none; }`\n",
       "     but bootstrap.min.css set `[hidden] { display: none !important; }`\n",
       "     so we also need the `!important` here to be able to override the\n",
       "     default hidden behavior on the sphinx rendered scikit-learn.org.\n",
       "     See: https://github.com/scikit-learn/scikit-learn/issues/21755 */\n",
       "  display: inline-block !important;\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-text-repr-fallback {\n",
       "  display: none;\n",
       "}\n",
       "\n",
       "div.sk-parallel-item,\n",
       "div.sk-serial,\n",
       "div.sk-item {\n",
       "  /* draw centered vertical line to link estimators */\n",
       "  background-image: linear-gradient(var(--sklearn-color-text-on-default-background), var(--sklearn-color-text-on-default-background));\n",
       "  background-size: 2px 100%;\n",
       "  background-repeat: no-repeat;\n",
       "  background-position: center center;\n",
       "}\n",
       "\n",
       "/* Parallel-specific style estimator block */\n",
       "\n",
       "#sk-container-id-3 div.sk-parallel-item::after {\n",
       "  content: \"\";\n",
       "  width: 100%;\n",
       "  border-bottom: 2px solid var(--sklearn-color-text-on-default-background);\n",
       "  flex-grow: 1;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-parallel {\n",
       "  display: flex;\n",
       "  align-items: stretch;\n",
       "  justify-content: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-parallel-item {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-parallel-item:first-child::after {\n",
       "  align-self: flex-end;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-parallel-item:last-child::after {\n",
       "  align-self: flex-start;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-parallel-item:only-child::after {\n",
       "  width: 0;\n",
       "}\n",
       "\n",
       "/* Serial-specific style estimator block */\n",
       "\n",
       "#sk-container-id-3 div.sk-serial {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "  align-items: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  padding-right: 1em;\n",
       "  padding-left: 1em;\n",
       "}\n",
       "\n",
       "\n",
       "/* Toggleable style: style used for estimator/Pipeline/ColumnTransformer box that is\n",
       "clickable and can be expanded/collapsed.\n",
       "- Pipeline and ColumnTransformer use this feature and define the default style\n",
       "- Estimators will overwrite some part of the style using the `sk-estimator` class\n",
       "*/\n",
       "\n",
       "/* Pipeline and ColumnTransformer style (default) */\n",
       "\n",
       "#sk-container-id-3 div.sk-toggleable {\n",
       "  /* Default theme specific background. It is overwritten whether we have a\n",
       "  specific estimator or a Pipeline/ColumnTransformer */\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "/* Toggleable label */\n",
       "#sk-container-id-3 label.sk-toggleable__label {\n",
       "  cursor: pointer;\n",
       "  display: block;\n",
       "  width: 100%;\n",
       "  margin-bottom: 0;\n",
       "  padding: 0.5em;\n",
       "  box-sizing: border-box;\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 label.sk-toggleable__label-arrow:before {\n",
       "  /* Arrow on the left of the label */\n",
       "  content: \"▸\";\n",
       "  float: left;\n",
       "  margin-right: 0.25em;\n",
       "  color: var(--sklearn-color-icon);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 label.sk-toggleable__label-arrow:hover:before {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "/* Toggleable content - dropdown */\n",
       "\n",
       "#sk-container-id-3 div.sk-toggleable__content {\n",
       "  max-height: 0;\n",
       "  max-width: 0;\n",
       "  overflow: hidden;\n",
       "  text-align: left;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-toggleable__content.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-toggleable__content pre {\n",
       "  margin: 0.2em;\n",
       "  border-radius: 0.25em;\n",
       "  color: var(--sklearn-color-text);\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-toggleable__content.fitted pre {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 input.sk-toggleable__control:checked~div.sk-toggleable__content {\n",
       "  /* Expand drop-down */\n",
       "  max-height: 200px;\n",
       "  max-width: 100%;\n",
       "  overflow: auto;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {\n",
       "  content: \"▾\";\n",
       "}\n",
       "\n",
       "/* Pipeline/ColumnTransformer-specific style */\n",
       "\n",
       "#sk-container-id-3 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-label.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator-specific style */\n",
       "\n",
       "/* Colorize estimator box */\n",
       "#sk-container-id-3 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-estimator.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-label label.sk-toggleable__label,\n",
       "#sk-container-id-3 div.sk-label label {\n",
       "  /* The background is the default theme color */\n",
       "  color: var(--sklearn-color-text-on-default-background);\n",
       "}\n",
       "\n",
       "/* On hover, darken the color of the background */\n",
       "#sk-container-id-3 div.sk-label:hover label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "/* Label box, darken color on hover, fitted */\n",
       "#sk-container-id-3 div.sk-label.fitted:hover label.sk-toggleable__label.fitted {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator label */\n",
       "\n",
       "#sk-container-id-3 div.sk-label label {\n",
       "  font-family: monospace;\n",
       "  font-weight: bold;\n",
       "  display: inline-block;\n",
       "  line-height: 1.2em;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-label-container {\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "/* Estimator-specific */\n",
       "#sk-container-id-3 div.sk-estimator {\n",
       "  font-family: monospace;\n",
       "  border: 1px dotted var(--sklearn-color-border-box);\n",
       "  border-radius: 0.25em;\n",
       "  box-sizing: border-box;\n",
       "  margin-bottom: 0.5em;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-estimator.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "/* on hover */\n",
       "#sk-container-id-3 div.sk-estimator:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-3 div.sk-estimator.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Specification for estimator info (e.g. \"i\" and \"?\") */\n",
       "\n",
       "/* Common style for \"i\" and \"?\" */\n",
       "\n",
       ".sk-estimator-doc-link,\n",
       "a:link.sk-estimator-doc-link,\n",
       "a:visited.sk-estimator-doc-link {\n",
       "  float: right;\n",
       "  font-size: smaller;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1em;\n",
       "  height: 1em;\n",
       "  width: 1em;\n",
       "  text-decoration: none !important;\n",
       "  margin-left: 1ex;\n",
       "  /* unfitted */\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted,\n",
       "a:link.sk-estimator-doc-link.fitted,\n",
       "a:visited.sk-estimator-doc-link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "div.sk-estimator:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "div.sk-estimator.fitted:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "/* Span, style for the box shown on hovering the info icon */\n",
       ".sk-estimator-doc-link span {\n",
       "  display: none;\n",
       "  z-index: 9999;\n",
       "  position: relative;\n",
       "  font-weight: normal;\n",
       "  right: .2ex;\n",
       "  padding: .5ex;\n",
       "  margin: .5ex;\n",
       "  width: min-content;\n",
       "  min-width: 20ex;\n",
       "  max-width: 50ex;\n",
       "  color: var(--sklearn-color-text);\n",
       "  box-shadow: 2pt 2pt 4pt #999;\n",
       "  /* unfitted */\n",
       "  background: var(--sklearn-color-unfitted-level-0);\n",
       "  border: .5pt solid var(--sklearn-color-unfitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted span {\n",
       "  /* fitted */\n",
       "  background: var(--sklearn-color-fitted-level-0);\n",
       "  border: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link:hover span {\n",
       "  display: block;\n",
       "}\n",
       "\n",
       "/* \"?\"-specific style due to the `<a>` HTML tag */\n",
       "\n",
       "#sk-container-id-3 a.estimator_doc_link {\n",
       "  float: right;\n",
       "  font-size: 1rem;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1rem;\n",
       "  height: 1rem;\n",
       "  width: 1rem;\n",
       "  text-decoration: none;\n",
       "  /* unfitted */\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 a.estimator_doc_link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "#sk-container-id-3 a.estimator_doc_link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "#sk-container-id-3 a.estimator_doc_link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "</style><div id=\"sk-container-id-3\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>RandomizedSearchCV(estimator=RandomForestClassifier(random_state=42), n_jobs=4,\n",
       "                   param_distributions={&#x27;class_weight&#x27;: [{0: 1, 1: 0.25}],\n",
       "                                        &#x27;max_depth&#x27;: [None, 10, 11, 12, 13, 14,\n",
       "                                                      15, 16, 17, 18, 19, 20,\n",
       "                                                      21, 22, 23, 24, 25, 26,\n",
       "                                                      27, 28, 29, 30, 31, 32,\n",
       "                                                      33, 34, 35, 36, 37, 38, ...],\n",
       "                                        &#x27;min_samples_split&#x27;: &lt;scipy.stats._distn_infrastructure.rv_discrete_frozen object at 0x7948058ab5b0&gt;,\n",
       "                                        &#x27;n_estimators&#x27;: &lt;scipy.stats._distn_infrastructure.rv_discrete_frozen object at 0x794805828be0&gt;},\n",
       "                   random_state=42,\n",
       "                   scoring=make_scorer(f1_score, response_method=&#x27;predict&#x27;, average=weighted))</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-7\" type=\"checkbox\" ><label for=\"sk-estimator-id-7\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;&nbsp;RandomizedSearchCV<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.model_selection.RandomizedSearchCV.html\">?<span>Documentation for RandomizedSearchCV</span></a><span class=\"sk-estimator-doc-link fitted\">i<span>Fitted</span></span></label><div class=\"sk-toggleable__content fitted\"><pre>RandomizedSearchCV(estimator=RandomForestClassifier(random_state=42), n_jobs=4,\n",
       "                   param_distributions={&#x27;class_weight&#x27;: [{0: 1, 1: 0.25}],\n",
       "                                        &#x27;max_depth&#x27;: [None, 10, 11, 12, 13, 14,\n",
       "                                                      15, 16, 17, 18, 19, 20,\n",
       "                                                      21, 22, 23, 24, 25, 26,\n",
       "                                                      27, 28, 29, 30, 31, 32,\n",
       "                                                      33, 34, 35, 36, 37, 38, ...],\n",
       "                                        &#x27;min_samples_split&#x27;: &lt;scipy.stats._distn_infrastructure.rv_discrete_frozen object at 0x7948058ab5b0&gt;,\n",
       "                                        &#x27;n_estimators&#x27;: &lt;scipy.stats._distn_infrastructure.rv_discrete_frozen object at 0x794805828be0&gt;},\n",
       "                   random_state=42,\n",
       "                   scoring=make_scorer(f1_score, response_method=&#x27;predict&#x27;, average=weighted))</pre></div> </div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-8\" type=\"checkbox\" ><label for=\"sk-estimator-id-8\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">estimator: RandomForestClassifier</label><div class=\"sk-toggleable__content fitted\"><pre>RandomForestClassifier(random_state=42)</pre></div> </div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-9\" type=\"checkbox\" ><label for=\"sk-estimator-id-9\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;RandomForestClassifier<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.ensemble.RandomForestClassifier.html\">?<span>Documentation for RandomForestClassifier</span></a></label><div class=\"sk-toggleable__content fitted\"><pre>RandomForestClassifier(random_state=42)</pre></div> </div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "RandomizedSearchCV(estimator=RandomForestClassifier(random_state=42), n_jobs=4,\n",
       "                   param_distributions={'class_weight': [{0: 1, 1: 0.25}],\n",
       "                                        'max_depth': [None, 10, 11, 12, 13, 14,\n",
       "                                                      15, 16, 17, 18, 19, 20,\n",
       "                                                      21, 22, 23, 24, 25, 26,\n",
       "                                                      27, 28, 29, 30, 31, 32,\n",
       "                                                      33, 34, 35, 36, 37, 38, ...],\n",
       "                                        'min_samples_split': <scipy.stats._distn_infrastructure.rv_discrete_frozen object at 0x7948058ab5b0>,\n",
       "                                        'n_estimators': <scipy.stats._distn_infrastructure.rv_discrete_frozen object at 0x794805828be0>},\n",
       "                   random_state=42,\n",
       "                   scoring=make_scorer(f1_score, response_method='predict', average=weighted))"
      ]
     },
     "execution_count": 136,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random_search.fit(x_train, y_train)\n",
    "#grid_search.fit(x, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameters: {'class_weight': {0: 1, 1: 0.25}, 'max_depth': 32, 'min_samples_split': 5, 'n_estimators': 153}\n"
     ]
    }
   ],
   "source": [
    "# Get the best parameters from the random search\n",
    "best_params = random_search.best_params_\n",
    "\n",
    "# Print the best parameters\n",
    "print(\"Best Parameters:\", best_params)\n",
    "\n",
    "# Get the best estimator (Random Forest classifier with the best parameters)\n",
    "best_rf_classifier = random_search.best_estimator_\n",
    "\n",
    "# Train the best classifier on the entire dataset\n",
    "best_rf_classifier.fit(x, y)\n",
    "\n",
    "# Extract feature importance\n",
    "feature_importances = best_rf_classifier.feature_importances_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Print the importance of each feature if original unaltered variables are used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Sort the features and their importance scores by importance in descending order\n",
    "sorted_features = sorted(zip(x.columns, feature_importances), key=lambda x: x[1], reverse=True)\n",
    "\n",
    "# Print all features and their importance scores in descending order\n",
    "print(\"\\nFeature Importance (Descending Order):\")\n",
    "for feature, importance in sorted_features:\n",
    "    print(f\"{feature}: {importance}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Print the importance of each feature if one hot encoder is used in descending order."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_AGE_G: 0.10378490525284366\n",
      "_INCOMG1: 0.09896199952245247\n",
      "_HCVU652: 0.07764165379308188\n",
      "_RFHYPE6: 0.07763948023634709\n",
      "_EDUCAG: 0.0578318636042885\n",
      "_RFHLTH: 0.05549590228974927\n",
      "_RFCHOL3: 0.05404574003870293\n",
      "_PHYS14D: 0.051028269222555\n",
      "_RACE: 0.049967238577657634\n",
      "_SMOKER3: 0.04949376816673063\n",
      "_MENT14D: 0.04317670659519978\n",
      "_BMI5CAT: 0.04116992926502239\n",
      "_AIDTST4: 0.038188030270938686\n",
      "_FRTLT1A: 0.03676601724944525\n",
      "_VEGLT1A: 0.03543090405782048\n",
      "_SEX: 0.03182151319394934\n",
      "_TOTINDA: 0.02727769554681295\n",
      "_RFBING5: 0.020129593026226612\n",
      "_RFBMI5: 0.016887414479694468\n",
      "_RFDRHV7: 0.013723627853188538\n",
      "_HLTHPLN: 0.012955427012098343\n",
      "_CHOLCH3: 0.006582320745193958\n"
     ]
    }
   ],
   "source": [
    "feature_importance_sum = {}\n",
    "\n",
    "for feature, importance in zip(x.columns, feature_importances):\n",
    "    feature_name = re.sub(r'(.*?)_\\d$', r'\\1', feature)\n",
    "\n",
    "    if feature_name in feature_importance_sum:\n",
    "        feature_importance_sum[feature_name] += importance\n",
    "    else:\n",
    "        feature_importance_sum[feature_name] = importance\n",
    "\n",
    "sorted_feature_importance_sum = sorted(feature_importance_sum.items(), key=lambda x: x[1], reverse=True)\n",
    "\n",
    "for feature, importance_sum in sorted_feature_importance_sum:\n",
    "    print(f\"{feature}: {importance_sum}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "features = [feat for feat, _ in sorted_feature_importance_sum]\n",
    "importance_sums = [imp_sum for _, imp_sum in sorted_feature_importance_sum]\n",
    "\n",
    "# Plotting\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.barh(features, importance_sums, color='skyblue')\n",
    "plt.xlabel('Importance Sum')\n",
    "plt.ylabel('Features')\n",
    "plt.title('Feature Importance Sum')\n",
    "plt.gca().invert_yaxis()  # Invert y-axis to display features from top to bottom\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Print the top features if original unaltered variables are used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_top_features = 23  # You can adjust this number based on your preference\n",
    "top_feature_indices = feature_importances.argsort()[-num_top_features:][::-1]\n",
    "top_features = x.columns[top_feature_indices]\n",
    "\n",
    "print(\"Top features:\")\n",
    "for feature in top_features:\n",
    "    print(feature)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Print the top complete features if one hot encoder is used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "top_features = [sorted_feature_importance_sum[i][0] for i in range(5)]\n",
    "\n",
    "complete_features = []\n",
    "\n",
    "for feature in top_features:\n",
    "  for column in dataframe_sample.columns:\n",
    "      if column.startswith(feature):\n",
    "          suffix = column[len(feature):].lstrip('_')\n",
    "          if suffix != '':\n",
    "            complete_part = feature + '_' + suffix\n",
    "          else:\n",
    "            complete_part = feature\n",
    "\n",
    "          complete_features.append(complete_part)\n",
    "\n",
    "top_features = complete_features\n",
    "for complete_feat in top_features: print(complete_feat)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ***Classification with Data Mining Methods***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Class Frequency of the Target Variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "category_labels = [\"Didn't Report having MI or CHD\", 'Reported having MI or CHD']\n",
    "sns.countplot(data=dataframes[2], x='_MICHD', order=dataframes[2]['_MICHD'].value_counts().index)\n",
    "plt.xticks(ticks=range(len(category_labels)), labels=category_labels)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generation of Synthetic Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we try different oversampling methods and for each variant we run the model and save the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "SMOTE_data = dataframes[2].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "SMOTE_data['_MICHD'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "SMOTE_data.drop_duplicates()['_MICHD'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "target_column = \"_MICHD\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### First variant: Borderline SMOTE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate synthetic data for the original imbalanced data\n",
    "resampled_data = BorderlineSMOTE().fit_resample(SMOTE_data.drop(columns = ['_MICHD']), \n",
    "                                        SMOTE_data['_MICHD'])\n",
    "SMOTE_data = resampled_data[0]\n",
    "SMOTE_data[target_column] = resampled_data[1].to_numpy()\n",
    "\n",
    "# This variable specifies the oversampling method we use. 1 is for the first method, 2 is for second and so on. Based on this variable the results are saved to the correct dataframe.\n",
    "method=1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Counts for the \"Had_CHD_MI\" variable\n",
    "sns.countplot(data=SMOTE_data,x='_MICHD')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 0: Reported having MI or CHD\n",
    "# 1: Did not report having MI or CHD\n",
    "\n",
    "SMOTE_data['_MICHD'] = SMOTE_data['_MICHD'].map({1.0: 0, 2.0: 1})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# In this dataframe we will keep the results for every oversampling variant and for all the methods we use. In rows, we have the metrics and in columns, we have the classifiers.\n",
    "BorderlineSMOTE_loss = pd.DataFrame(columns = ['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "BorderlineSMOTE_precision = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "BorderlineSMOTE_recall = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "BorderlineSMOTE_f1_score = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Second variant: SMOTE-Tomek"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate synthetic data for the original imbalanced data\n",
    "from imblearn.combine import SMOTETomek\n",
    "\n",
    "smote_tomek = SMOTETomek()\n",
    "resampled_data = smote_tomek.fit_resample(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'])\n",
    "\n",
    "SMOTE_data = resampled_data[0]\n",
    "SMOTE_data[target_column] = resampled_data[1].to_numpy()\n",
    "\n",
    "# This variable specifies the oversampling method we use. 1 is for the first method, 2 is for second and so on. Based on this variable the results are saved to the correct dataframe.\n",
    "method=2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Counts for the \"Had_CHD_MI\" variable\n",
    "sns.countplot(data=SMOTE_data,x='_MICHD')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 0: Reported having MI or CHD\n",
    "# 1: Did not report having MI or CHD\n",
    "\n",
    "SMOTE_data['_MICHD'] = SMOTE_data['_MICHD'].map({1.0: 0, 2.0: 1})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# In this dataframe we will keep the results for every oversampling variant and for all the methods we use. In rows, we have the metrics and in columns, we have the classifiers.\n",
    "smote_tomek_loss = pd.DataFrame(columns = ['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "smote_tomek_precision = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "smote_tomek_recall = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "smote_tomek_f1_score = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Third variant: ADASYN (Adaptive Synthetic Sampling)\n",
    "ADASYN is an extension of SMOTE that adaptively generates samples according to the local density of minority samples. It focuses more on regions where the class imbalance is more severe."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from imblearn.over_sampling import ADASYN\n",
    "\n",
    "adasyn = ADASYN()\n",
    "resampled_data = adasyn.fit_resample(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'])\n",
    "\n",
    "SMOTE_data = resampled_data[0]\n",
    "SMOTE_data[target_column] = resampled_data[1].to_numpy()\n",
    "\n",
    "# This variable specifies the oversampling method we use. 1 is for the first method, 2 is for second and so on. Based on this variable the results are saved to the correct dataframe.\n",
    "method=3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Counts for the \"Had_CHD_MI\" variable\n",
    "sns.countplot(data=SMOTE_data,x='_MICHD')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 0: Reported having MI or CHD\n",
    "# 1: Did not report having MI or CHD\n",
    "\n",
    "SMOTE_data['_MICHD'] = SMOTE_data['_MICHD'].map({1.0: 0, 2.0: 1})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# In this dataframe we will keep the results for every oversampling variant and for all the methods we use. In rows, we have the metrics and in columns, we have the classifiers.\n",
    "adasyn_loss = pd.DataFrame(columns = ['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "adasyn_precision = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "adasyn_recall = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "adasyn_f1_score = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fourth variant: SMOTE "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from imblearn.over_sampling import SMOTE\n",
    "\n",
    "resampled_data = SMOTE().fit_resample(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'])\n",
    "\n",
    "SMOTE_data = resampled_data[0]\n",
    "SMOTE_data[target_column] = resampled_data[1].to_numpy()\n",
    "\n",
    "# This variable specifies the oversampling method we use. 1 is for the first method, 2 is for second and so on. Based on this variable the results are saved to the correct dataframe.\n",
    "method=4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Counts for the \"Had_CHD_MI\" variable\n",
    "sns.countplot(data=SMOTE_data,x='_MICHD')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 0: Reported having MI or CHD\n",
    "# 1: Did not report having MI or CHD\n",
    "\n",
    "SMOTE_data['_MICHD'] = SMOTE_data['_MICHD'].map({1.0: 0, 2.0: 1})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# In this dataframe we will keep the results for every oversampling variant and for all the methods we use. In rows, we have the metrics and in columns, we have the classifiers.\n",
    "SMOTE_loss = pd.DataFrame(columns = ['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "SMOTE_precision = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "SMOTE_recall = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])\n",
    "SMOTE_f1_score = pd.DataFrame(columns=['XGBoost', 'Decision Tree', 'Naive Bayes', 'Random Forests', 'KNN'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "target_column = \"_MICHD\"\n",
    "\n",
    "\n",
    "# Utilize all the selected dataframe samples \n",
    "#y = dataframes[2][target_column]\n",
    "#x = dataframes[2].drop(columns=[target_column])\n",
    "\n",
    "\n",
    "# Utilize a dataframe sample \n",
    "y = dataframe_sample[target_column]\n",
    "x = dataframe_sample.drop(columns = [target_column])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use this spliting only when undersampling of the majority class has occured and there is no class imbalance\n",
    "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.25, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the cost matrix\n",
    "cost_matrix = [ \n",
    "    [0, 1],  \n",
    "    [1, 0]\n",
    "] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "tabular_based_classifiers = {}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ***XGBoost***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### XGBoost for undersampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_dist = {\n",
    "    'n_estimators': randint(100, 500),\n",
    "    'max_depth': randint(3, 10),\n",
    "    'learning_rate': uniform(0.01, 0.5),\n",
    "}\n",
    "\n",
    "xgb_classifier = XGBClassifier(objective='binary:logistic', random_state=SEED)\n",
    "\n",
    "random_search = RandomizedSearchCV(estimator=xgb_classifier, param_distributions=param_dist, n_iter=50, random_state=SEED,n_jobs=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameters found: {'learning_rate': 0.017818203370596968, 'max_depth': 3, 'n_estimators': 443}\n"
     ]
    }
   ],
   "source": [
    "random_search.fit(x_train, y_train)\n",
    "\n",
    "best_params = random_search.best_params_\n",
    "print(\"Best parameters found:\", best_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.74      0.79      0.77      3504\n",
      "           1       0.77      0.72      0.75      3432\n",
      "\n",
      "    accuracy                           0.76      6936\n",
      "   macro avg       0.76      0.76      0.76      6936\n",
      "weighted avg       0.76      0.76      0.76      6936\n",
      "\n",
      "Confusion Matrix:\n",
      "[[2783  721]\n",
      " [ 964 2468]]\n",
      "Total Cost: 1685\n"
     ]
    }
   ],
   "source": [
    "#best_xgb_model = random_search.best_estimator_\n",
    "\n",
    "xgboost_clf = XGBClassifier(n_estimators=443, max_depth=3, learning_rate=0.018, objective='binary:logistic', random_state=SEED)\n",
    "xgboost_clf = CalibratedClassifierCV(xgboost_clf, method='isotonic')\n",
    "xgboost_clf.fit(x_train, y_train)\n",
    "\n",
    "tabular_based_classifiers['XGBoost'] = xgboost_clf\n",
    "\n",
    "y_pred = xgboost_clf.predict(x_test)\n",
    "\n",
    "print(\"Classification Report:\")\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "print(\"Confusion Matrix:\")\n",
    "print(confusion_matrix(y_test, y_pred))\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(\"Total Cost:\", total_cost)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### XGBoost for oversampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the balanced dataset to a set of training and testing samples \n",
    "x_train, x_test, y_train, y_test = train_test_split(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'], test_size=0.25, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# Define the parameter grid to search\n",
    "param_grid = {\n",
    "    'n_estimators': [100, 300, 500],  # Number of trees in the forest\n",
    "    'learning_rate': [0.05, 0.1, 0.2],  # Learning rate\n",
    "    'max_depth': [3, 5, 7],  # Maximum depth of each tree\n",
    "    'min_child_weight': [1, 3, 5],  # Minimum sum of instance weight needed in a child\n",
    "}\n",
    "\n",
    "# Create an XGBClassifier object\n",
    "xgb_model = XGBClassifier(random_state=SEED)\n",
    "\n",
    "# Create GridSearchCV object\n",
    "grid_search = GridSearchCV(estimator=xgb_model, param_grid=param_grid, cv=3, scoring='neg_mean_squared_error', n_jobs=-1)\n",
    "\n",
    "# Perform grid search\n",
    "grid_search.fit(x_train, y_train)\n",
    "\n",
    "# Get the best parameters and the best estimator\n",
    "best_params = grid_search.best_params_\n",
    "best_xgb_model = grid_search.best_estimator_\n",
    "\n",
    "# Train the best estimator\n",
    "best_xgb_model.fit(x_train, y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = best_xgb_model.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "# Print classification report for additional evaluation metrics\n",
    "print('\\nClassification Report:')\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method == 1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost baseline', 'XGBoost'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision baseline', 'XGBoost'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall baseline', 'XGBoost'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score baseline', 'XGBoost'] = metrics_dict['f1-score']\n",
    "elif method == 2:\n",
    "    smote_tomek_loss.loc['Total Cost baseline', 'XGBoost'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision baseline', 'XGBoost'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall baseline', 'XGBoost'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score baseline', 'XGBoost'] = metrics_dict['f1-score']\n",
    "elif method == 3:\n",
    "    adasyn_loss.loc['Total Cost baseline', 'XGBoost'] = total_cost\n",
    "    adasyn_precision.loc['macro precision baseline', 'XGBoost'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall baseline', 'XGBoost'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score baseline', 'XGBoost'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost baseline', 'XGBoost'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision baseline', 'XGBoost'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall baseline', 'XGBoost'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score baseline', 'XGBoost'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Get feature importances\n",
    "importances = best_xgb_model.feature_importances_\n",
    "features = x_train.columns\n",
    "\n",
    "# Sort feature importances in descending order\n",
    "indices = np.argsort(importances)[::-1]\n",
    "\n",
    "# Plot\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.title(\"Feature Importances (XGBoost)\")\n",
    "plt.bar(range(x_train.shape[1]), importances[indices], align=\"center\")\n",
    "plt.xticks(range(x_train.shape[1]), [features[i] for i in indices], rotation=90)\n",
    "plt.xlabel(\"Feature\")\n",
    "plt.ylabel(\"Importance\")\n",
    "plt.tight_layout()\n",
    "\n",
    "# Save the plot as an image file\n",
    "plt.savefig(\"feature_importance_plot (XGBoost).png\")\n",
    "\n",
    "# Display the plot\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import shap\n",
    "\n",
    "# Explain the model's predictions using SHAP\n",
    "explainer = shap.TreeExplainer(best_xgb_model)\n",
    "shap_values = explainer.shap_values(x_train)\n",
    "\n",
    "# Save SHAP values to a file\n",
    "np.save('shap_values.npy', shap_values)\n",
    "\n",
    "# Plot the SHAP values\n",
    "shap.summary_plot(shap_values, features=x_train, feature_names=x_train.columns, show=False)\n",
    "plt.savefig('shap_summary_plot.png', bbox_inches='tight')  # Save the plot as an image\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def predict(y_prob, cost_m):\n",
    "    t_BMR = (cost_m[1][0] - cost_m[1][1])/(cost_m[0][1] - cost_m[1][1] - cost_m[0][0] + cost_m[1][0])\n",
    "    y_pred = np.greater(y_prob[:, 1], t_BMR).astype(int)\n",
    "    \n",
    "    return y_pred\n",
    "\n",
    "from sklearn.calibration import CalibratedClassifierCV\n",
    "\n",
    "# Create an XGBClassifier object\n",
    "xgb_model = CalibratedClassifierCV(XGBClassifier(n_estimators = 500, learning_rate = 0.1, random_state=SEED), method=\"isotonic\", cv=10)\n",
    "\n",
    "# Train the classifier\n",
    "xgb_model.fit(x_train, y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_prob = xgb_model.predict_proba(x_test)\n",
    "y_pred = predict(y_prob, cost_matrix)\n",
    "    \n",
    "# Print the confusion matrix\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal loss Cost: {total_cost}\") \n",
    "\n",
    "# Print classification report for additional evaluation metrics\n",
    "print('\\nClassification Report:')\n",
    "print(classification_report(y_test, y_pred, target_names = [\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method == 1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost MEC', 'XGBoost'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision MEC', 'XGBoost'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall MEC', 'XGBoost'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score MEC', 'XGBoost'] = metrics_dict['f1-score']\n",
    "elif method == 2:\n",
    "    smote_tomek_loss.loc['Total Cost MEC', 'XGBoost'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision MEC', 'XGBoost'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall MEC', 'XGBoost'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score MEC', 'XGBoost'] = metrics_dict['f1-score']\n",
    "elif method == 3:\n",
    "    adasyn_loss.loc['Total Cost MEC', 'XGBoost'] = total_cost\n",
    "    adasyn_precision.loc['macro precision MEC', 'XGBoost'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall MEC', 'XGBoost'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score MEC', 'XGBoost'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost MEC', 'XGBoost'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision MEC', 'XGBoost'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall MEC', 'XGBoost'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score MEC', 'XGBoost'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Calculate the scale_pos_weight based on the corrected cost matrix\n",
    "scale_pos_weight = cost_matrix[1][0] / cost_matrix[0][1]\n",
    "\n",
    "# Create an XGBClassifier object\n",
    "xgb_model = XGBClassifier(n_estimators = 500, learning_rate = 0.1, scale_pos_weight=scale_pos_weight, random_state=SEED)\n",
    "\n",
    "# Train the classifier\n",
    "xgb_model.fit(x_train, y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = xgb_model.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal loss Cost: {total_cost}\") \n",
    "\n",
    "# Print classification report for additional evaluation metrics\n",
    "print('\\nClassification Report:')\n",
    "print(classification_report(y_test, y_pred, target_names = [\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method == 1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost weighting', 'XGBoost'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision weighting', 'XGBoost'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall weighting', 'XGBoost'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score weighting', 'XGBoost'] = metrics_dict['f1-score']\n",
    "elif method == 2:\n",
    "    smote_tomek_loss.loc['Total Cost weighting', 'XGBoost'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision weighting', 'XGBoost'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall weighting', 'XGBoost'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score weighting', 'XGBoost'] = metrics_dict['f1-score']\n",
    "elif method == 3:\n",
    "    adasyn_loss.loc['Total Cost weighting', 'XGBoost'] = total_cost\n",
    "    adasyn_precision.loc['macro precision weighting', 'XGBoost'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall weighting', 'XGBoost'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score weighting', 'XGBoost'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost weighting', 'XGBoost'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision weighting', 'XGBoost'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall weighting', 'XGBoost'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score weighting', 'XGBoost'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ***Decision Tree Classifier***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Decision Tree plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from io import StringIO\n",
    "\n",
    "# Visualize the decision tree\n",
    "def show_decision_tree_classifier(model, features):\n",
    "  dot_data = StringIO()\n",
    "  export_graphviz(model,\n",
    "                  out_file=dot_data,\n",
    "                  filled=False,\n",
    "                  rounded=True,\n",
    "                  special_characters=False,\n",
    "                  impurity=True,\n",
    "                  precision=2,\n",
    "                  rotate=False,\n",
    "                  feature_names=features,\n",
    "                  fontname='arial',\n",
    "                  label='all',\n",
    "                  class_names=['1', '2', '3', '4', '5', '6'])\n",
    "\n",
    "  graph = pydotplus.graph_from_dot_data(dot_data.getvalue())\n",
    "  graph.write_png('DecisionTree.png')\n",
    "\n",
    "  return graph"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Decision Tree for undersampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_grid = {\n",
    "    'max_depth': [5, 10, 15, None],\n",
    "    'min_samples_split': [100, 200, 500],\n",
    "    'min_samples_leaf': [50, 100, 200]\n",
    "}\n",
    "\n",
    "grid_search = GridSearchCV(DecisionTreeClassifier(random_state=SEED), param_grid)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-1 {\n",
       "  /* Definition of color scheme common for light and dark mode */\n",
       "  --sklearn-color-text: black;\n",
       "  --sklearn-color-line: gray;\n",
       "  /* Definition of color scheme for unfitted estimators */\n",
       "  --sklearn-color-unfitted-level-0: #fff5e6;\n",
       "  --sklearn-color-unfitted-level-1: #f6e4d2;\n",
       "  --sklearn-color-unfitted-level-2: #ffe0b3;\n",
       "  --sklearn-color-unfitted-level-3: chocolate;\n",
       "  /* Definition of color scheme for fitted estimators */\n",
       "  --sklearn-color-fitted-level-0: #f0f8ff;\n",
       "  --sklearn-color-fitted-level-1: #d4ebff;\n",
       "  --sklearn-color-fitted-level-2: #b3dbfd;\n",
       "  --sklearn-color-fitted-level-3: cornflowerblue;\n",
       "\n",
       "  /* Specific color for light theme */\n",
       "  --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, white)));\n",
       "  --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-icon: #696969;\n",
       "\n",
       "  @media (prefers-color-scheme: dark) {\n",
       "    /* Redefinition of color scheme for dark theme */\n",
       "    --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, #111)));\n",
       "    --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-icon: #878787;\n",
       "  }\n",
       "}\n",
       "\n",
       "#sk-container-id-1 {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 pre {\n",
       "  padding: 0;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 input.sk-hidden--visually {\n",
       "  border: 0;\n",
       "  clip: rect(1px 1px 1px 1px);\n",
       "  clip: rect(1px, 1px, 1px, 1px);\n",
       "  height: 1px;\n",
       "  margin: -1px;\n",
       "  overflow: hidden;\n",
       "  padding: 0;\n",
       "  position: absolute;\n",
       "  width: 1px;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-dashed-wrapped {\n",
       "  border: 1px dashed var(--sklearn-color-line);\n",
       "  margin: 0 0.4em 0.5em 0.4em;\n",
       "  box-sizing: border-box;\n",
       "  padding-bottom: 0.4em;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-container {\n",
       "  /* jupyter's `normalize.less` sets `[hidden] { display: none; }`\n",
       "     but bootstrap.min.css set `[hidden] { display: none !important; }`\n",
       "     so we also need the `!important` here to be able to override the\n",
       "     default hidden behavior on the sphinx rendered scikit-learn.org.\n",
       "     See: https://github.com/scikit-learn/scikit-learn/issues/21755 */\n",
       "  display: inline-block !important;\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-text-repr-fallback {\n",
       "  display: none;\n",
       "}\n",
       "\n",
       "div.sk-parallel-item,\n",
       "div.sk-serial,\n",
       "div.sk-item {\n",
       "  /* draw centered vertical line to link estimators */\n",
       "  background-image: linear-gradient(var(--sklearn-color-text-on-default-background), var(--sklearn-color-text-on-default-background));\n",
       "  background-size: 2px 100%;\n",
       "  background-repeat: no-repeat;\n",
       "  background-position: center center;\n",
       "}\n",
       "\n",
       "/* Parallel-specific style estimator block */\n",
       "\n",
       "#sk-container-id-1 div.sk-parallel-item::after {\n",
       "  content: \"\";\n",
       "  width: 100%;\n",
       "  border-bottom: 2px solid var(--sklearn-color-text-on-default-background);\n",
       "  flex-grow: 1;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-parallel {\n",
       "  display: flex;\n",
       "  align-items: stretch;\n",
       "  justify-content: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-parallel-item {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-parallel-item:first-child::after {\n",
       "  align-self: flex-end;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-parallel-item:last-child::after {\n",
       "  align-self: flex-start;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-parallel-item:only-child::after {\n",
       "  width: 0;\n",
       "}\n",
       "\n",
       "/* Serial-specific style estimator block */\n",
       "\n",
       "#sk-container-id-1 div.sk-serial {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "  align-items: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  padding-right: 1em;\n",
       "  padding-left: 1em;\n",
       "}\n",
       "\n",
       "\n",
       "/* Toggleable style: style used for estimator/Pipeline/ColumnTransformer box that is\n",
       "clickable and can be expanded/collapsed.\n",
       "- Pipeline and ColumnTransformer use this feature and define the default style\n",
       "- Estimators will overwrite some part of the style using the `sk-estimator` class\n",
       "*/\n",
       "\n",
       "/* Pipeline and ColumnTransformer style (default) */\n",
       "\n",
       "#sk-container-id-1 div.sk-toggleable {\n",
       "  /* Default theme specific background. It is overwritten whether we have a\n",
       "  specific estimator or a Pipeline/ColumnTransformer */\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "/* Toggleable label */\n",
       "#sk-container-id-1 label.sk-toggleable__label {\n",
       "  cursor: pointer;\n",
       "  display: block;\n",
       "  width: 100%;\n",
       "  margin-bottom: 0;\n",
       "  padding: 0.5em;\n",
       "  box-sizing: border-box;\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 label.sk-toggleable__label-arrow:before {\n",
       "  /* Arrow on the left of the label */\n",
       "  content: \"▸\";\n",
       "  float: left;\n",
       "  margin-right: 0.25em;\n",
       "  color: var(--sklearn-color-icon);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 label.sk-toggleable__label-arrow:hover:before {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "/* Toggleable content - dropdown */\n",
       "\n",
       "#sk-container-id-1 div.sk-toggleable__content {\n",
       "  max-height: 0;\n",
       "  max-width: 0;\n",
       "  overflow: hidden;\n",
       "  text-align: left;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-toggleable__content.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-toggleable__content pre {\n",
       "  margin: 0.2em;\n",
       "  border-radius: 0.25em;\n",
       "  color: var(--sklearn-color-text);\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-toggleable__content.fitted pre {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 input.sk-toggleable__control:checked~div.sk-toggleable__content {\n",
       "  /* Expand drop-down */\n",
       "  max-height: 200px;\n",
       "  max-width: 100%;\n",
       "  overflow: auto;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {\n",
       "  content: \"▾\";\n",
       "}\n",
       "\n",
       "/* Pipeline/ColumnTransformer-specific style */\n",
       "\n",
       "#sk-container-id-1 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-label.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator-specific style */\n",
       "\n",
       "/* Colorize estimator box */\n",
       "#sk-container-id-1 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-estimator.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-label label.sk-toggleable__label,\n",
       "#sk-container-id-1 div.sk-label label {\n",
       "  /* The background is the default theme color */\n",
       "  color: var(--sklearn-color-text-on-default-background);\n",
       "}\n",
       "\n",
       "/* On hover, darken the color of the background */\n",
       "#sk-container-id-1 div.sk-label:hover label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "/* Label box, darken color on hover, fitted */\n",
       "#sk-container-id-1 div.sk-label.fitted:hover label.sk-toggleable__label.fitted {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator label */\n",
       "\n",
       "#sk-container-id-1 div.sk-label label {\n",
       "  font-family: monospace;\n",
       "  font-weight: bold;\n",
       "  display: inline-block;\n",
       "  line-height: 1.2em;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-label-container {\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "/* Estimator-specific */\n",
       "#sk-container-id-1 div.sk-estimator {\n",
       "  font-family: monospace;\n",
       "  border: 1px dotted var(--sklearn-color-border-box);\n",
       "  border-radius: 0.25em;\n",
       "  box-sizing: border-box;\n",
       "  margin-bottom: 0.5em;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-estimator.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "/* on hover */\n",
       "#sk-container-id-1 div.sk-estimator:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-1 div.sk-estimator.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Specification for estimator info (e.g. \"i\" and \"?\") */\n",
       "\n",
       "/* Common style for \"i\" and \"?\" */\n",
       "\n",
       ".sk-estimator-doc-link,\n",
       "a:link.sk-estimator-doc-link,\n",
       "a:visited.sk-estimator-doc-link {\n",
       "  float: right;\n",
       "  font-size: smaller;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1em;\n",
       "  height: 1em;\n",
       "  width: 1em;\n",
       "  text-decoration: none !important;\n",
       "  margin-left: 1ex;\n",
       "  /* unfitted */\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted,\n",
       "a:link.sk-estimator-doc-link.fitted,\n",
       "a:visited.sk-estimator-doc-link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "div.sk-estimator:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "div.sk-estimator.fitted:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "/* Span, style for the box shown on hovering the info icon */\n",
       ".sk-estimator-doc-link span {\n",
       "  display: none;\n",
       "  z-index: 9999;\n",
       "  position: relative;\n",
       "  font-weight: normal;\n",
       "  right: .2ex;\n",
       "  padding: .5ex;\n",
       "  margin: .5ex;\n",
       "  width: min-content;\n",
       "  min-width: 20ex;\n",
       "  max-width: 50ex;\n",
       "  color: var(--sklearn-color-text);\n",
       "  box-shadow: 2pt 2pt 4pt #999;\n",
       "  /* unfitted */\n",
       "  background: var(--sklearn-color-unfitted-level-0);\n",
       "  border: .5pt solid var(--sklearn-color-unfitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted span {\n",
       "  /* fitted */\n",
       "  background: var(--sklearn-color-fitted-level-0);\n",
       "  border: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link:hover span {\n",
       "  display: block;\n",
       "}\n",
       "\n",
       "/* \"?\"-specific style due to the `<a>` HTML tag */\n",
       "\n",
       "#sk-container-id-1 a.estimator_doc_link {\n",
       "  float: right;\n",
       "  font-size: 1rem;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1rem;\n",
       "  height: 1rem;\n",
       "  width: 1rem;\n",
       "  text-decoration: none;\n",
       "  /* unfitted */\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 a.estimator_doc_link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "#sk-container-id-1 a.estimator_doc_link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "#sk-container-id-1 a.estimator_doc_link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "</style><div id=\"sk-container-id-1\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>GridSearchCV(estimator=DecisionTreeClassifier(random_state=42),\n",
       "             param_grid={&#x27;max_depth&#x27;: [5, 10, 15, None],\n",
       "                         &#x27;min_samples_leaf&#x27;: [50, 100, 200],\n",
       "                         &#x27;min_samples_split&#x27;: [100, 200, 500]})</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-1\" type=\"checkbox\" ><label for=\"sk-estimator-id-1\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;&nbsp;GridSearchCV<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.model_selection.GridSearchCV.html\">?<span>Documentation for GridSearchCV</span></a><span class=\"sk-estimator-doc-link fitted\">i<span>Fitted</span></span></label><div class=\"sk-toggleable__content fitted\"><pre>GridSearchCV(estimator=DecisionTreeClassifier(random_state=42),\n",
       "             param_grid={&#x27;max_depth&#x27;: [5, 10, 15, None],\n",
       "                         &#x27;min_samples_leaf&#x27;: [50, 100, 200],\n",
       "                         &#x27;min_samples_split&#x27;: [100, 200, 500]})</pre></div> </div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-2\" type=\"checkbox\" ><label for=\"sk-estimator-id-2\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">estimator: DecisionTreeClassifier</label><div class=\"sk-toggleable__content fitted\"><pre>DecisionTreeClassifier(random_state=42)</pre></div> </div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-3\" type=\"checkbox\" ><label for=\"sk-estimator-id-3\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;DecisionTreeClassifier<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.tree.DecisionTreeClassifier.html\">?<span>Documentation for DecisionTreeClassifier</span></a></label><div class=\"sk-toggleable__content fitted\"><pre>DecisionTreeClassifier(random_state=42)</pre></div> </div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "GridSearchCV(estimator=DecisionTreeClassifier(random_state=42),\n",
       "             param_grid={'max_depth': [5, 10, 15, None],\n",
       "                         'min_samples_leaf': [50, 100, 200],\n",
       "                         'min_samples_split': [100, 200, 500]})"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid_search.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameters: {'max_depth': 15, 'min_samples_leaf': 100, 'min_samples_split': 100}\n"
     ]
    }
   ],
   "source": [
    "best_params = grid_search.best_params_\n",
    "print(\"Best parameters:\", best_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "dt_clf = DecisionTreeClassifier(random_state=SEED, max_depth=15, min_samples_split=100, min_samples_leaf=100)\n",
    "dt_clf = CalibratedClassifierCV(dt_clf, method='isotonic')\n",
    "dt_clf.fit(x_train, y_train)\n",
    "tabular_based_classifiers['DT'] = dt_clf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.72      0.83      0.77      3504\n",
      "           1       0.80      0.67      0.73      3432\n",
      "\n",
      "    accuracy                           0.75      6936\n",
      "   macro avg       0.76      0.75      0.75      6936\n",
      "weighted avg       0.76      0.75      0.75      6936\n",
      "\n",
      "Confusion Matrix:\n",
      "[[2918  586]\n",
      " [1145 2287]]\n",
      "Total Cost: 1731\n"
     ]
    }
   ],
   "source": [
    "y_pred = dt_clf.predict(x_test)\n",
    "\n",
    "print(\"Classification Report:\")\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "print(\"Confusion Matrix:\")\n",
    "print(confusion_matrix(y_test, y_pred))\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(\"Total Cost:\", total_cost)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Decision Tree for oversampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the balanced dataset to a set of training and testing samples \n",
    "x_train, x_test, y_train, y_test = train_test_split(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'], test_size=0.25, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def predict(y_prob, cost_m):\n",
    "    t_BMR = (cost_m[1][0] - cost_m[1][1])/(cost_m[0][1] - cost_m[1][1] - cost_m[0][0] + cost_m[1][0])\n",
    "    y_pred = np.greater(y_prob[:, 1], t_BMR).astype(int)\n",
    "    \n",
    "    return y_pred\n",
    "\n",
    "from sklearn.calibration import CalibratedClassifierCV\n",
    "\n",
    "# Create a Decision Tree Classifier object\n",
    "dt = CalibratedClassifierCV(DecisionTreeClassifier(random_state=SEED), method=\"isotonic\", cv=10)\n",
    "\n",
    "# Train the classifier\n",
    "dt.fit(x_train , y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_prob = dt.predict_proba(x_test)\n",
    "y_pred = predict(y_prob, cost_matrix)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method==1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==2:\n",
    "    smote_tomek_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==3:\n",
    "    adasyn_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    adasyn_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Create a Decision Tree Classifier object\n",
    "dt = DecisionTreeClassifier(random_state=SEED)\n",
    "\n",
    "# Train the classifier\n",
    "dt.fit(x_train, y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = dt.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method==1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==2:\n",
    "    smote_tomek_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==3:\n",
    "    adasyn_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    adasyn_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Get feature importances\n",
    "importances = dt.feature_importances_\n",
    "features = x_train.columns\n",
    "\n",
    "# Sort feature importances in descending order\n",
    "indices = np.argsort(importances)[::-1]\n",
    "\n",
    "# Plot\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.title(\"Feature Importances ()\")\n",
    "plt.bar(range(x_train.shape[1]), importances[indices], align=\"center\")\n",
    "plt.xticks(range(x_train.shape[1]), [features[i] for i in indices], rotation=90)\n",
    "plt.xlabel(\"Feature\")\n",
    "plt.ylabel(\"Importance\")\n",
    "plt.tight_layout()\n",
    "\n",
    "# Save the plot as an image file\n",
    "plt.savefig(\"feature_importance_plot ().png\")\n",
    "\n",
    "# Display the plot\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Calculate sample weights based on the corrected cost matrix\n",
    "sample_weights = np.zeros_like(y_train, dtype=float)\n",
    "sample_weights[y_train == 0] = cost_matrix[0][1]\n",
    "sample_weights[y_train == 1] = cost_matrix[1][0]\n",
    "\n",
    "# Create a Decision Tree Classifier object\n",
    "dt = DecisionTreeClassifier()\n",
    "\n",
    "# Train the classifier\n",
    "dt.fit(x_train , y_train, sample_weight=sample_weights)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = dt.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method==1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==2:\n",
    "    smote_tomek_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==3:\n",
    "    adasyn_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    adasyn_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ***Gaussian Naive Bayes***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### NB for undersampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix:\n",
      "[[2723  781]\n",
      " [1165 2267]]\n",
      "\n",
      "Total Cost: 1946\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.70      0.78      0.74      3504\n",
      "           1       0.74      0.66      0.70      3432\n",
      "\n",
      "    accuracy                           0.72      6936\n",
      "   macro avg       0.72      0.72      0.72      6936\n",
      "weighted avg       0.72      0.72      0.72      6936\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sample_weights = np.zeros(y_train.shape[0])\n",
    "sample_weights[np.where(y_train == 0)] = 1\n",
    "sample_weights[np.where(y_train == 1)] = 1\n",
    "\n",
    "# Create a Gaussian Naive Bayes Classifier object\n",
    "gnb = GaussianNB()\n",
    "\n",
    "gnb = CalibratedClassifierCV(gnb, method='isotonic')\n",
    "\n",
    "# Train the classifier\n",
    "gnb.fit(x_train , y_train, sample_weight=sample_weights)\n",
    "\n",
    "tabular_based_classifiers['GNB'] = gnb\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = gnb.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred))\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### NB for oversampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the balanced dataset to a set of training and testing samples \n",
    "x_train, x_test, y_train, y_test = train_test_split(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'], test_size=0.25, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Create a Gaussian Naive Bayes Classifier object\n",
    "gnb = GaussianNB()\n",
    "\n",
    "# Train the classifier\n",
    "gnb.fit(x_train , y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = gnb.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method==1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==2:\n",
    "    smote_tomek_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==3:\n",
    "    adasyn_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    adasyn_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost baseline', 'Decision Tree'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision baseline', 'Decision Tree'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall baseline', 'Decision Tree'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score baseline', 'Decision Tree'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sample_weights = np.zeros_like(y_train, dtype=float)\n",
    "sample_weights[y_train == 0] = cost_matrix[0][1]\n",
    "sample_weights[y_train == 1] = cost_matrix[1][0]\n",
    "\n",
    "# Create a Gaussian Naive Bayes Classifier object\n",
    "gnb = GaussianNB()\n",
    "\n",
    "# Train the classifier\n",
    "gnb.fit(x_train , y_train, sample_weight=sample_weights)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = gnb.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method==1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==2:\n",
    "    smote_tomek_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==3:\n",
    "    adasyn_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    adasyn_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost weighting', 'Decision Tree'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision weighting', 'Decision Tree'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall weighting', 'Decision Tree'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score weighting', 'Decision Tree'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def predict(y_prob, cost_m):\n",
    "    t_BMR = (cost_m[1][0] - cost_m[1][1])/(cost_m[0][1] - cost_m[1][1] - cost_m[0][0] + cost_m[1][0])\n",
    "    y_pred = np.greater(y_prob[:, 1], t_BMR).astype(int)\n",
    "    \n",
    "    return y_pred\n",
    "\n",
    "from sklearn.calibration import CalibratedClassifierCV\n",
    "\n",
    "# Create a Gaussian Naive Bayes Classifier object\n",
    "gnb = CalibratedClassifierCV(GaussianNB(), method=\"isotonic\", cv=10)\n",
    "\n",
    "# Train the classifier\n",
    "gnb.fit(x_train , y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_prob = gnb.predict_proba(x_test)\n",
    "y_pred = predict(y_prob, cost_matrix)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "if method==1:\n",
    "    BorderlineSMOTE_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    # Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "    BorderlineSMOTE_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    # Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "    BorderlineSMOTE_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    # Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "    BorderlineSMOTE_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==2:\n",
    "    smote_tomek_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    smote_tomek_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    smote_tomek_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    smote_tomek_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "elif method==3:\n",
    "    adasyn_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    adasyn_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    adasyn_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    adasyn_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']\n",
    "else:\n",
    "    SMOTE_loss.loc['Total Cost MEC', 'Decision Tree'] = total_cost\n",
    "    SMOTE_precision.loc['macro precision MEC', 'Decision Tree'] = metrics_dict['precision']\n",
    "    SMOTE_recall.loc['macro recall MEC', 'Decision Tree'] = metrics_dict['recall']\n",
    "    SMOTE_f1_score.loc['macro f1-score MEC', 'Decision Tree'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ***Random Forests***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### RF for undersampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_grid = {\n",
    "    'n_estimators': [100, 200, 300],\n",
    "    'max_depth': [None, 10, 20]\n",
    "}\n",
    "\n",
    "rf_classifier = RandomForestClassifier(random_state=SEED)\n",
    "\n",
    "grid_search = GridSearchCV(estimator=rf_classifier, param_grid=param_grid, n_jobs=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-2 {\n",
       "  /* Definition of color scheme common for light and dark mode */\n",
       "  --sklearn-color-text: black;\n",
       "  --sklearn-color-line: gray;\n",
       "  /* Definition of color scheme for unfitted estimators */\n",
       "  --sklearn-color-unfitted-level-0: #fff5e6;\n",
       "  --sklearn-color-unfitted-level-1: #f6e4d2;\n",
       "  --sklearn-color-unfitted-level-2: #ffe0b3;\n",
       "  --sklearn-color-unfitted-level-3: chocolate;\n",
       "  /* Definition of color scheme for fitted estimators */\n",
       "  --sklearn-color-fitted-level-0: #f0f8ff;\n",
       "  --sklearn-color-fitted-level-1: #d4ebff;\n",
       "  --sklearn-color-fitted-level-2: #b3dbfd;\n",
       "  --sklearn-color-fitted-level-3: cornflowerblue;\n",
       "\n",
       "  /* Specific color for light theme */\n",
       "  --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, white)));\n",
       "  --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-icon: #696969;\n",
       "\n",
       "  @media (prefers-color-scheme: dark) {\n",
       "    /* Redefinition of color scheme for dark theme */\n",
       "    --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, #111)));\n",
       "    --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-icon: #878787;\n",
       "  }\n",
       "}\n",
       "\n",
       "#sk-container-id-2 {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 pre {\n",
       "  padding: 0;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 input.sk-hidden--visually {\n",
       "  border: 0;\n",
       "  clip: rect(1px 1px 1px 1px);\n",
       "  clip: rect(1px, 1px, 1px, 1px);\n",
       "  height: 1px;\n",
       "  margin: -1px;\n",
       "  overflow: hidden;\n",
       "  padding: 0;\n",
       "  position: absolute;\n",
       "  width: 1px;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-dashed-wrapped {\n",
       "  border: 1px dashed var(--sklearn-color-line);\n",
       "  margin: 0 0.4em 0.5em 0.4em;\n",
       "  box-sizing: border-box;\n",
       "  padding-bottom: 0.4em;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-container {\n",
       "  /* jupyter's `normalize.less` sets `[hidden] { display: none; }`\n",
       "     but bootstrap.min.css set `[hidden] { display: none !important; }`\n",
       "     so we also need the `!important` here to be able to override the\n",
       "     default hidden behavior on the sphinx rendered scikit-learn.org.\n",
       "     See: https://github.com/scikit-learn/scikit-learn/issues/21755 */\n",
       "  display: inline-block !important;\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-text-repr-fallback {\n",
       "  display: none;\n",
       "}\n",
       "\n",
       "div.sk-parallel-item,\n",
       "div.sk-serial,\n",
       "div.sk-item {\n",
       "  /* draw centered vertical line to link estimators */\n",
       "  background-image: linear-gradient(var(--sklearn-color-text-on-default-background), var(--sklearn-color-text-on-default-background));\n",
       "  background-size: 2px 100%;\n",
       "  background-repeat: no-repeat;\n",
       "  background-position: center center;\n",
       "}\n",
       "\n",
       "/* Parallel-specific style estimator block */\n",
       "\n",
       "#sk-container-id-2 div.sk-parallel-item::after {\n",
       "  content: \"\";\n",
       "  width: 100%;\n",
       "  border-bottom: 2px solid var(--sklearn-color-text-on-default-background);\n",
       "  flex-grow: 1;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-parallel {\n",
       "  display: flex;\n",
       "  align-items: stretch;\n",
       "  justify-content: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-parallel-item {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-parallel-item:first-child::after {\n",
       "  align-self: flex-end;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-parallel-item:last-child::after {\n",
       "  align-self: flex-start;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-parallel-item:only-child::after {\n",
       "  width: 0;\n",
       "}\n",
       "\n",
       "/* Serial-specific style estimator block */\n",
       "\n",
       "#sk-container-id-2 div.sk-serial {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "  align-items: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  padding-right: 1em;\n",
       "  padding-left: 1em;\n",
       "}\n",
       "\n",
       "\n",
       "/* Toggleable style: style used for estimator/Pipeline/ColumnTransformer box that is\n",
       "clickable and can be expanded/collapsed.\n",
       "- Pipeline and ColumnTransformer use this feature and define the default style\n",
       "- Estimators will overwrite some part of the style using the `sk-estimator` class\n",
       "*/\n",
       "\n",
       "/* Pipeline and ColumnTransformer style (default) */\n",
       "\n",
       "#sk-container-id-2 div.sk-toggleable {\n",
       "  /* Default theme specific background. It is overwritten whether we have a\n",
       "  specific estimator or a Pipeline/ColumnTransformer */\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "/* Toggleable label */\n",
       "#sk-container-id-2 label.sk-toggleable__label {\n",
       "  cursor: pointer;\n",
       "  display: block;\n",
       "  width: 100%;\n",
       "  margin-bottom: 0;\n",
       "  padding: 0.5em;\n",
       "  box-sizing: border-box;\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 label.sk-toggleable__label-arrow:before {\n",
       "  /* Arrow on the left of the label */\n",
       "  content: \"▸\";\n",
       "  float: left;\n",
       "  margin-right: 0.25em;\n",
       "  color: var(--sklearn-color-icon);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 label.sk-toggleable__label-arrow:hover:before {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "/* Toggleable content - dropdown */\n",
       "\n",
       "#sk-container-id-2 div.sk-toggleable__content {\n",
       "  max-height: 0;\n",
       "  max-width: 0;\n",
       "  overflow: hidden;\n",
       "  text-align: left;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-toggleable__content.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-toggleable__content pre {\n",
       "  margin: 0.2em;\n",
       "  border-radius: 0.25em;\n",
       "  color: var(--sklearn-color-text);\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-toggleable__content.fitted pre {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 input.sk-toggleable__control:checked~div.sk-toggleable__content {\n",
       "  /* Expand drop-down */\n",
       "  max-height: 200px;\n",
       "  max-width: 100%;\n",
       "  overflow: auto;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {\n",
       "  content: \"▾\";\n",
       "}\n",
       "\n",
       "/* Pipeline/ColumnTransformer-specific style */\n",
       "\n",
       "#sk-container-id-2 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-label.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator-specific style */\n",
       "\n",
       "/* Colorize estimator box */\n",
       "#sk-container-id-2 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-estimator.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-label label.sk-toggleable__label,\n",
       "#sk-container-id-2 div.sk-label label {\n",
       "  /* The background is the default theme color */\n",
       "  color: var(--sklearn-color-text-on-default-background);\n",
       "}\n",
       "\n",
       "/* On hover, darken the color of the background */\n",
       "#sk-container-id-2 div.sk-label:hover label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "/* Label box, darken color on hover, fitted */\n",
       "#sk-container-id-2 div.sk-label.fitted:hover label.sk-toggleable__label.fitted {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator label */\n",
       "\n",
       "#sk-container-id-2 div.sk-label label {\n",
       "  font-family: monospace;\n",
       "  font-weight: bold;\n",
       "  display: inline-block;\n",
       "  line-height: 1.2em;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-label-container {\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "/* Estimator-specific */\n",
       "#sk-container-id-2 div.sk-estimator {\n",
       "  font-family: monospace;\n",
       "  border: 1px dotted var(--sklearn-color-border-box);\n",
       "  border-radius: 0.25em;\n",
       "  box-sizing: border-box;\n",
       "  margin-bottom: 0.5em;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-estimator.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "/* on hover */\n",
       "#sk-container-id-2 div.sk-estimator:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-2 div.sk-estimator.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Specification for estimator info (e.g. \"i\" and \"?\") */\n",
       "\n",
       "/* Common style for \"i\" and \"?\" */\n",
       "\n",
       ".sk-estimator-doc-link,\n",
       "a:link.sk-estimator-doc-link,\n",
       "a:visited.sk-estimator-doc-link {\n",
       "  float: right;\n",
       "  font-size: smaller;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1em;\n",
       "  height: 1em;\n",
       "  width: 1em;\n",
       "  text-decoration: none !important;\n",
       "  margin-left: 1ex;\n",
       "  /* unfitted */\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted,\n",
       "a:link.sk-estimator-doc-link.fitted,\n",
       "a:visited.sk-estimator-doc-link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "div.sk-estimator:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "div.sk-estimator.fitted:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "/* Span, style for the box shown on hovering the info icon */\n",
       ".sk-estimator-doc-link span {\n",
       "  display: none;\n",
       "  z-index: 9999;\n",
       "  position: relative;\n",
       "  font-weight: normal;\n",
       "  right: .2ex;\n",
       "  padding: .5ex;\n",
       "  margin: .5ex;\n",
       "  width: min-content;\n",
       "  min-width: 20ex;\n",
       "  max-width: 50ex;\n",
       "  color: var(--sklearn-color-text);\n",
       "  box-shadow: 2pt 2pt 4pt #999;\n",
       "  /* unfitted */\n",
       "  background: var(--sklearn-color-unfitted-level-0);\n",
       "  border: .5pt solid var(--sklearn-color-unfitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted span {\n",
       "  /* fitted */\n",
       "  background: var(--sklearn-color-fitted-level-0);\n",
       "  border: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link:hover span {\n",
       "  display: block;\n",
       "}\n",
       "\n",
       "/* \"?\"-specific style due to the `<a>` HTML tag */\n",
       "\n",
       "#sk-container-id-2 a.estimator_doc_link {\n",
       "  float: right;\n",
       "  font-size: 1rem;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1rem;\n",
       "  height: 1rem;\n",
       "  width: 1rem;\n",
       "  text-decoration: none;\n",
       "  /* unfitted */\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 a.estimator_doc_link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "#sk-container-id-2 a.estimator_doc_link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "#sk-container-id-2 a.estimator_doc_link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "</style><div id=\"sk-container-id-2\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>GridSearchCV(estimator=RandomForestClassifier(random_state=42), n_jobs=4,\n",
       "             param_grid={&#x27;max_depth&#x27;: [None, 10, 20],\n",
       "                         &#x27;n_estimators&#x27;: [100, 200, 300]})</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-4\" type=\"checkbox\" ><label for=\"sk-estimator-id-4\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;&nbsp;GridSearchCV<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.model_selection.GridSearchCV.html\">?<span>Documentation for GridSearchCV</span></a><span class=\"sk-estimator-doc-link fitted\">i<span>Fitted</span></span></label><div class=\"sk-toggleable__content fitted\"><pre>GridSearchCV(estimator=RandomForestClassifier(random_state=42), n_jobs=4,\n",
       "             param_grid={&#x27;max_depth&#x27;: [None, 10, 20],\n",
       "                         &#x27;n_estimators&#x27;: [100, 200, 300]})</pre></div> </div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-5\" type=\"checkbox\" ><label for=\"sk-estimator-id-5\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">estimator: RandomForestClassifier</label><div class=\"sk-toggleable__content fitted\"><pre>RandomForestClassifier(random_state=42)</pre></div> </div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-6\" type=\"checkbox\" ><label for=\"sk-estimator-id-6\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;RandomForestClassifier<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.ensemble.RandomForestClassifier.html\">?<span>Documentation for RandomForestClassifier</span></a></label><div class=\"sk-toggleable__content fitted\"><pre>RandomForestClassifier(random_state=42)</pre></div> </div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "GridSearchCV(estimator=RandomForestClassifier(random_state=42), n_jobs=4,\n",
       "             param_grid={'max_depth': [None, 10, 20],\n",
       "                         'n_estimators': [100, 200, 300]})"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid_search.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameters: {'max_depth': 10, 'n_estimators': 100}\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.74      0.82      0.78      3504\n",
      "           1       0.79      0.70      0.75      3432\n",
      "\n",
      "    accuracy                           0.76      6936\n",
      "   macro avg       0.77      0.76      0.76      6936\n",
      "weighted avg       0.77      0.76      0.76      6936\n",
      "\n",
      "Confusion Matrix:\n",
      "[[2879  625]\n",
      " [1018 2414]]\n",
      "Total Cost: 1643\n"
     ]
    }
   ],
   "source": [
    "# Get the best parameters from the random search\n",
    "best_params = grid_search.best_params_\n",
    "\n",
    "# Print the best parameters\n",
    "print(\"Best Parameters:\", best_params)\n",
    "\n",
    "# Get the best estimator (Random Forest classifier with the best parameters)\n",
    "best_rf_classifier = grid_search.best_estimator_\n",
    "\n",
    "best_rf_classifier = CalibratedClassifierCV(best_rf_classifier, method='isotonic')\n",
    "\n",
    "# Train the best classifier on the entire dataset\n",
    "best_rf_classifier.fit(x_train, y_train)\n",
    "\n",
    "tabular_based_classifiers['RF'] = best_rf_classifier\n",
    "\n",
    "# Use the best estimator to make predictions\n",
    "y_pred = best_rf_classifier.predict(x_test)\n",
    "\n",
    "# Print the classification report\n",
    "print(\"Classification Report:\")\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# Print the confusion matrix\n",
    "print(\"Confusion Matrix:\")\n",
    "print(confusion_matrix(y_test, y_pred))\n",
    "\n",
    "# Calculate the total cost based on the cost matrix\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(\"Total Cost:\", total_cost)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### RF for oversampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the balanced dataset to a set of training and testing samples \n",
    "x_train, x_test, y_train, y_test = train_test_split(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'], test_size=0.25, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def predict(y_prob, cost_m):\n",
    "    t_BMR = (cost_m[1][0] - cost_m[1][1])/(cost_m[0][1] - cost_m[1][1] - cost_m[0][0] + cost_m[1][0])\n",
    "    y_pred = np.greater(y_prob[:, 1], t_BMR).astype(int)\n",
    "    \n",
    "    return y_pred\n",
    "\n",
    "\n",
    "from sklearn.calibration import CalibratedClassifierCV\n",
    "\n",
    "\n",
    "# Create a Random Forests Classifier object\n",
    "rf_classifier = CalibratedClassifierCV(RandomForestClassifier(n_estimators=393, max_depth=16, random_state=SEED), method=\"sigmoid\", cv=10)\n",
    "\n",
    "# Train the classifier\n",
    "rf_classifier.fit(x_train, y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_prob = rf_classifier.predict_proba(x_test)\n",
    "y_pred = predict(y_prob, cost_matrix)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\")\n",
    "\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "BorderlineSMOTE_loss.loc['Total Cost MEC', 'Random Forests'] = total_cost\n",
    "# Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "BorderlineSMOTE_precision.loc['macro precision MEC', 'Random Forests'] = metrics_dict['precision']\n",
    "# Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "BorderlineSMOTE_recall.loc['macro recall MEC', 'Random Forests'] = metrics_dict['recall']\n",
    "# Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "BorderlineSMOTE_f1_score.loc['macro f1-score MEC', 'Random Forests'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Create a Random Forests Classifier object\n",
    "rf_classifier = RandomForestClassifier(n_estimators=393, max_depth=16, random_state=SEED)\n",
    "\n",
    "# Train the classifier\n",
    "rf_classifier.fit(x_train , y_train)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = rf_classifier.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\")\n",
    "\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "BorderlineSMOTE_loss.loc['Total Cost baseline', 'Random Forests'] = total_cost\n",
    "# Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "BorderlineSMOTE_precision.loc['macro precision baseline', 'Random Forests'] = metrics_dict['precision']\n",
    "# Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "BorderlineSMOTE_recall.loc['macro recall baseline', 'Random Forests'] = metrics_dict['recall']\n",
    "# Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "BorderlineSMOTE_f1_score.loc['macro f1-score baseline', 'Random Forests'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sample_weights = np.zeros_like(y_train, dtype=float)\n",
    "sample_weights[y_train == 0] = cost_matrix[0][1]\n",
    "sample_weights[y_train == 1] = cost_matrix[1][0]\n",
    "\n",
    "# Create a Random Forests Classifier object\n",
    "rf_classifier = RandomForestClassifier(n_estimators=393, max_depth=16, random_state=SEED)\n",
    "\n",
    "# Train the classifier\n",
    "rf_classifier.fit(x_train , y_train, sample_weight=sample_weights)\n",
    "\n",
    "# Predict the response for test dataset\n",
    "y_pred = rf_classifier.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix and a classification report\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\")\n",
    "\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "BorderlineSMOTE_loss.loc['Total Cost weighting', 'Random Forests'] = total_cost\n",
    "# Update BorderlineSMOTE_precision DataFrame with macro precision\n",
    "BorderlineSMOTE_precision.loc['macro precision weighting', 'Random Forests'] = metrics_dict['precision']\n",
    "# Update BorderlineSMOTE_recall DataFrame with macro recall\n",
    "BorderlineSMOTE_recall.loc['macro recall weighting', 'Random Forests'] = metrics_dict['recall']\n",
    "# Update BorderlineSMOTE_f1_score DataFrame with macro F1-score\n",
    "BorderlineSMOTE_f1_score.loc['macro f1-score weighting', 'Random Forests'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "import shap\n",
    "\n",
    "# Create an explainer object with the trained model and the training dataset\n",
    "explainer = shap.TreeExplainer(rf_classifier)\n",
    "\n",
    "# Calculate SHAP values for the test dataset\n",
    "shap_values = explainer.shap_values(x_test)\n",
    "\n",
    "# Visualize the SHAP values\n",
    "shap.summary_plot(shap_values, x_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ***K-Nearest Neighbors (KNN)***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### KNN for undersampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_grid = {\n",
    "    'n_neighbors': [3, 5, 7],\n",
    "    'weights': ['uniform', 'distance'],\n",
    "    'algorithm': ['auto', 'ball_tree'],\n",
    "}\n",
    "\n",
    "grid_search = GridSearchCV(KNeighborsClassifier(), param_grid, cv=5, n_jobs=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-10 {\n",
       "  /* Definition of color scheme common for light and dark mode */\n",
       "  --sklearn-color-text: black;\n",
       "  --sklearn-color-line: gray;\n",
       "  /* Definition of color scheme for unfitted estimators */\n",
       "  --sklearn-color-unfitted-level-0: #fff5e6;\n",
       "  --sklearn-color-unfitted-level-1: #f6e4d2;\n",
       "  --sklearn-color-unfitted-level-2: #ffe0b3;\n",
       "  --sklearn-color-unfitted-level-3: chocolate;\n",
       "  /* Definition of color scheme for fitted estimators */\n",
       "  --sklearn-color-fitted-level-0: #f0f8ff;\n",
       "  --sklearn-color-fitted-level-1: #d4ebff;\n",
       "  --sklearn-color-fitted-level-2: #b3dbfd;\n",
       "  --sklearn-color-fitted-level-3: cornflowerblue;\n",
       "\n",
       "  /* Specific color for light theme */\n",
       "  --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, white)));\n",
       "  --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, black)));\n",
       "  --sklearn-color-icon: #696969;\n",
       "\n",
       "  @media (prefers-color-scheme: dark) {\n",
       "    /* Redefinition of color scheme for dark theme */\n",
       "    --sklearn-color-text-on-default-background: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-background: var(--sg-background-color, var(--theme-background, var(--jp-layout-color0, #111)));\n",
       "    --sklearn-color-border-box: var(--sg-text-color, var(--theme-code-foreground, var(--jp-content-font-color1, white)));\n",
       "    --sklearn-color-icon: #878787;\n",
       "  }\n",
       "}\n",
       "\n",
       "#sk-container-id-10 {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 pre {\n",
       "  padding: 0;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 input.sk-hidden--visually {\n",
       "  border: 0;\n",
       "  clip: rect(1px 1px 1px 1px);\n",
       "  clip: rect(1px, 1px, 1px, 1px);\n",
       "  height: 1px;\n",
       "  margin: -1px;\n",
       "  overflow: hidden;\n",
       "  padding: 0;\n",
       "  position: absolute;\n",
       "  width: 1px;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-dashed-wrapped {\n",
       "  border: 1px dashed var(--sklearn-color-line);\n",
       "  margin: 0 0.4em 0.5em 0.4em;\n",
       "  box-sizing: border-box;\n",
       "  padding-bottom: 0.4em;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-container {\n",
       "  /* jupyter's `normalize.less` sets `[hidden] { display: none; }`\n",
       "     but bootstrap.min.css set `[hidden] { display: none !important; }`\n",
       "     so we also need the `!important` here to be able to override the\n",
       "     default hidden behavior on the sphinx rendered scikit-learn.org.\n",
       "     See: https://github.com/scikit-learn/scikit-learn/issues/21755 */\n",
       "  display: inline-block !important;\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-text-repr-fallback {\n",
       "  display: none;\n",
       "}\n",
       "\n",
       "div.sk-parallel-item,\n",
       "div.sk-serial,\n",
       "div.sk-item {\n",
       "  /* draw centered vertical line to link estimators */\n",
       "  background-image: linear-gradient(var(--sklearn-color-text-on-default-background), var(--sklearn-color-text-on-default-background));\n",
       "  background-size: 2px 100%;\n",
       "  background-repeat: no-repeat;\n",
       "  background-position: center center;\n",
       "}\n",
       "\n",
       "/* Parallel-specific style estimator block */\n",
       "\n",
       "#sk-container-id-10 div.sk-parallel-item::after {\n",
       "  content: \"\";\n",
       "  width: 100%;\n",
       "  border-bottom: 2px solid var(--sklearn-color-text-on-default-background);\n",
       "  flex-grow: 1;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-parallel {\n",
       "  display: flex;\n",
       "  align-items: stretch;\n",
       "  justify-content: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  position: relative;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-parallel-item {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-parallel-item:first-child::after {\n",
       "  align-self: flex-end;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-parallel-item:last-child::after {\n",
       "  align-self: flex-start;\n",
       "  width: 50%;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-parallel-item:only-child::after {\n",
       "  width: 0;\n",
       "}\n",
       "\n",
       "/* Serial-specific style estimator block */\n",
       "\n",
       "#sk-container-id-10 div.sk-serial {\n",
       "  display: flex;\n",
       "  flex-direction: column;\n",
       "  align-items: center;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  padding-right: 1em;\n",
       "  padding-left: 1em;\n",
       "}\n",
       "\n",
       "\n",
       "/* Toggleable style: style used for estimator/Pipeline/ColumnTransformer box that is\n",
       "clickable and can be expanded/collapsed.\n",
       "- Pipeline and ColumnTransformer use this feature and define the default style\n",
       "- Estimators will overwrite some part of the style using the `sk-estimator` class\n",
       "*/\n",
       "\n",
       "/* Pipeline and ColumnTransformer style (default) */\n",
       "\n",
       "#sk-container-id-10 div.sk-toggleable {\n",
       "  /* Default theme specific background. It is overwritten whether we have a\n",
       "  specific estimator or a Pipeline/ColumnTransformer */\n",
       "  background-color: var(--sklearn-color-background);\n",
       "}\n",
       "\n",
       "/* Toggleable label */\n",
       "#sk-container-id-10 label.sk-toggleable__label {\n",
       "  cursor: pointer;\n",
       "  display: block;\n",
       "  width: 100%;\n",
       "  margin-bottom: 0;\n",
       "  padding: 0.5em;\n",
       "  box-sizing: border-box;\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 label.sk-toggleable__label-arrow:before {\n",
       "  /* Arrow on the left of the label */\n",
       "  content: \"▸\";\n",
       "  float: left;\n",
       "  margin-right: 0.25em;\n",
       "  color: var(--sklearn-color-icon);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 label.sk-toggleable__label-arrow:hover:before {\n",
       "  color: var(--sklearn-color-text);\n",
       "}\n",
       "\n",
       "/* Toggleable content - dropdown */\n",
       "\n",
       "#sk-container-id-10 div.sk-toggleable__content {\n",
       "  max-height: 0;\n",
       "  max-width: 0;\n",
       "  overflow: hidden;\n",
       "  text-align: left;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-toggleable__content.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-toggleable__content pre {\n",
       "  margin: 0.2em;\n",
       "  border-radius: 0.25em;\n",
       "  color: var(--sklearn-color-text);\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-toggleable__content.fitted pre {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 input.sk-toggleable__control:checked~div.sk-toggleable__content {\n",
       "  /* Expand drop-down */\n",
       "  max-height: 200px;\n",
       "  max-width: 100%;\n",
       "  overflow: auto;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {\n",
       "  content: \"▾\";\n",
       "}\n",
       "\n",
       "/* Pipeline/ColumnTransformer-specific style */\n",
       "\n",
       "#sk-container-id-10 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-label.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator-specific style */\n",
       "\n",
       "/* Colorize estimator box */\n",
       "#sk-container-id-10 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-estimator.fitted input.sk-toggleable__control:checked~label.sk-toggleable__label {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-label label.sk-toggleable__label,\n",
       "#sk-container-id-10 div.sk-label label {\n",
       "  /* The background is the default theme color */\n",
       "  color: var(--sklearn-color-text-on-default-background);\n",
       "}\n",
       "\n",
       "/* On hover, darken the color of the background */\n",
       "#sk-container-id-10 div.sk-label:hover label.sk-toggleable__label {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "/* Label box, darken color on hover, fitted */\n",
       "#sk-container-id-10 div.sk-label.fitted:hover label.sk-toggleable__label.fitted {\n",
       "  color: var(--sklearn-color-text);\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Estimator label */\n",
       "\n",
       "#sk-container-id-10 div.sk-label label {\n",
       "  font-family: monospace;\n",
       "  font-weight: bold;\n",
       "  display: inline-block;\n",
       "  line-height: 1.2em;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-label-container {\n",
       "  text-align: center;\n",
       "}\n",
       "\n",
       "/* Estimator-specific */\n",
       "#sk-container-id-10 div.sk-estimator {\n",
       "  font-family: monospace;\n",
       "  border: 1px dotted var(--sklearn-color-border-box);\n",
       "  border-radius: 0.25em;\n",
       "  box-sizing: border-box;\n",
       "  margin-bottom: 0.5em;\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-0);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-estimator.fitted {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-0);\n",
       "}\n",
       "\n",
       "/* on hover */\n",
       "#sk-container-id-10 div.sk-estimator:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-2);\n",
       "}\n",
       "\n",
       "#sk-container-id-10 div.sk-estimator.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-2);\n",
       "}\n",
       "\n",
       "/* Specification for estimator info (e.g. \"i\" and \"?\") */\n",
       "\n",
       "/* Common style for \"i\" and \"?\" */\n",
       "\n",
       ".sk-estimator-doc-link,\n",
       "a:link.sk-estimator-doc-link,\n",
       "a:visited.sk-estimator-doc-link {\n",
       "  float: right;\n",
       "  font-size: smaller;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1em;\n",
       "  height: 1em;\n",
       "  width: 1em;\n",
       "  text-decoration: none !important;\n",
       "  margin-left: 1ex;\n",
       "  /* unfitted */\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted,\n",
       "a:link.sk-estimator-doc-link.fitted,\n",
       "a:visited.sk-estimator-doc-link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "div.sk-estimator:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link:hover,\n",
       ".sk-estimator-doc-link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "div.sk-estimator.fitted:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover,\n",
       "div.sk-label-container:hover .sk-estimator-doc-link.fitted:hover,\n",
       ".sk-estimator-doc-link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "/* Span, style for the box shown on hovering the info icon */\n",
       ".sk-estimator-doc-link span {\n",
       "  display: none;\n",
       "  z-index: 9999;\n",
       "  position: relative;\n",
       "  font-weight: normal;\n",
       "  right: .2ex;\n",
       "  padding: .5ex;\n",
       "  margin: .5ex;\n",
       "  width: min-content;\n",
       "  min-width: 20ex;\n",
       "  max-width: 50ex;\n",
       "  color: var(--sklearn-color-text);\n",
       "  box-shadow: 2pt 2pt 4pt #999;\n",
       "  /* unfitted */\n",
       "  background: var(--sklearn-color-unfitted-level-0);\n",
       "  border: .5pt solid var(--sklearn-color-unfitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link.fitted span {\n",
       "  /* fitted */\n",
       "  background: var(--sklearn-color-fitted-level-0);\n",
       "  border: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "\n",
       ".sk-estimator-doc-link:hover span {\n",
       "  display: block;\n",
       "}\n",
       "\n",
       "/* \"?\"-specific style due to the `<a>` HTML tag */\n",
       "\n",
       "#sk-container-id-10 a.estimator_doc_link {\n",
       "  float: right;\n",
       "  font-size: 1rem;\n",
       "  line-height: 1em;\n",
       "  font-family: monospace;\n",
       "  background-color: var(--sklearn-color-background);\n",
       "  border-radius: 1rem;\n",
       "  height: 1rem;\n",
       "  width: 1rem;\n",
       "  text-decoration: none;\n",
       "  /* unfitted */\n",
       "  color: var(--sklearn-color-unfitted-level-1);\n",
       "  border: var(--sklearn-color-unfitted-level-1) 1pt solid;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 a.estimator_doc_link.fitted {\n",
       "  /* fitted */\n",
       "  border: var(--sklearn-color-fitted-level-1) 1pt solid;\n",
       "  color: var(--sklearn-color-fitted-level-1);\n",
       "}\n",
       "\n",
       "/* On hover */\n",
       "#sk-container-id-10 a.estimator_doc_link:hover {\n",
       "  /* unfitted */\n",
       "  background-color: var(--sklearn-color-unfitted-level-3);\n",
       "  color: var(--sklearn-color-background);\n",
       "  text-decoration: none;\n",
       "}\n",
       "\n",
       "#sk-container-id-10 a.estimator_doc_link.fitted:hover {\n",
       "  /* fitted */\n",
       "  background-color: var(--sklearn-color-fitted-level-3);\n",
       "}\n",
       "</style><div id=\"sk-container-id-10\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>GridSearchCV(cv=5, estimator=KNeighborsClassifier(), n_jobs=4,\n",
       "             param_grid={&#x27;algorithm&#x27;: [&#x27;auto&#x27;, &#x27;ball_tree&#x27;],\n",
       "                         &#x27;n_neighbors&#x27;: [3, 5, 7],\n",
       "                         &#x27;weights&#x27;: [&#x27;uniform&#x27;, &#x27;distance&#x27;]})</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-22\" type=\"checkbox\" ><label for=\"sk-estimator-id-22\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;&nbsp;GridSearchCV<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.model_selection.GridSearchCV.html\">?<span>Documentation for GridSearchCV</span></a><span class=\"sk-estimator-doc-link fitted\">i<span>Fitted</span></span></label><div class=\"sk-toggleable__content fitted\"><pre>GridSearchCV(cv=5, estimator=KNeighborsClassifier(), n_jobs=4,\n",
       "             param_grid={&#x27;algorithm&#x27;: [&#x27;auto&#x27;, &#x27;ball_tree&#x27;],\n",
       "                         &#x27;n_neighbors&#x27;: [3, 5, 7],\n",
       "                         &#x27;weights&#x27;: [&#x27;uniform&#x27;, &#x27;distance&#x27;]})</pre></div> </div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-23\" type=\"checkbox\" ><label for=\"sk-estimator-id-23\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">estimator: KNeighborsClassifier</label><div class=\"sk-toggleable__content fitted\"><pre>KNeighborsClassifier()</pre></div> </div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator fitted sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-24\" type=\"checkbox\" ><label for=\"sk-estimator-id-24\" class=\"sk-toggleable__label fitted sk-toggleable__label-arrow fitted\">&nbsp;KNeighborsClassifier<a class=\"sk-estimator-doc-link fitted\" rel=\"noreferrer\" target=\"_blank\" href=\"https://scikit-learn.org/1.4/modules/generated/sklearn.neighbors.KNeighborsClassifier.html\">?<span>Documentation for KNeighborsClassifier</span></a></label><div class=\"sk-toggleable__content fitted\"><pre>KNeighborsClassifier()</pre></div> </div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "GridSearchCV(cv=5, estimator=KNeighborsClassifier(), n_jobs=4,\n",
       "             param_grid={'algorithm': ['auto', 'ball_tree'],\n",
       "                         'n_neighbors': [3, 5, 7],\n",
       "                         'weights': ['uniform', 'distance']})"
      ]
     },
     "execution_count": 199,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid_search.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameters: {'algorithm': 'auto', 'n_neighbors': 7, 'weights': 'uniform'}\n"
     ]
    }
   ],
   "source": [
    "best_params = grid_search.best_params_\n",
    "print(\"Best parameters:\", best_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_clf = KNeighborsClassifier(n_neighbors=7, algorithm='auto', weights='uniform')\n",
    "knn_clf = CalibratedClassifierCV(knn_clf, method='isotonic')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.71      0.80      0.75      3504\n",
      "           1       0.77      0.67      0.71      3432\n",
      "\n",
      "    accuracy                           0.74      6936\n",
      "   macro avg       0.74      0.74      0.73      6936\n",
      "weighted avg       0.74      0.74      0.73      6936\n",
      "\n",
      "Confusion Matrix:\n",
      "[[2809  695]\n",
      " [1137 2295]]\n",
      "Total Cost: 1832\n"
     ]
    }
   ],
   "source": [
    "knn_clf.fit(x_train, y_train)\n",
    "\n",
    "tabular_based_classifiers['KNN'] = knn_clf\n",
    "\n",
    "y_pred = knn_clf.predict(x_test)\n",
    "\n",
    "print(\"Classification Report:\")\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "print(\"Confusion Matrix:\")\n",
    "print(confusion_matrix(y_test, y_pred))\n",
    "\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(\"Total Cost:\", total_cost)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### KNN for oversampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the balanced dataset to a set of training and testing samples \n",
    "x_train, x_test, y_train, y_test = train_test_split(SMOTE_data.drop(columns=['_MICHD']), SMOTE_data['_MICHD'], test_size=0.25, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import cross_val_score\n",
    "\n",
    "# Define a range of k values to try\n",
    "k_values = range(1, 21)  # You can adjust this range as needed\n",
    "\n",
    "# Initialize lists to store mean cross-validation scores for each k\n",
    "mean_cv_scores = []\n",
    "\n",
    "# Iterate over each k value\n",
    "for k in k_values:\n",
    "    # Create a KNeighborsClassifier object with the current k value\n",
    "    knn_model = KNeighborsClassifier(n_neighbors=k)\n",
    "    \n",
    "    # Perform 5-fold cross-validation and compute the mean accuracy\n",
    "    cv_scores = cross_val_score(knn_model, x_train, y_train, cv=5, scoring='accuracy')\n",
    "    mean_cv_score = np.mean(cv_scores)\n",
    "    \n",
    "    # Append the mean cross-validation score to the list\n",
    "    mean_cv_scores.append(mean_cv_score)\n",
    "\n",
    "# Find the index of the k value with the highest mean cross-validation score\n",
    "best_k_index = np.argmax(mean_cv_scores)\n",
    "\n",
    "# Get the best value of k\n",
    "best_k = k_values[best_k_index]\n",
    "\n",
    "print(f\"The best value of k is: {best_k}\")\n",
    "\n",
    "# Create a KNeighborsClassifier object with the best value of k\n",
    "best_knn_model = KNeighborsClassifier(n_neighbors=best_k)\n",
    "\n",
    "# Train the classifier\n",
    "best_knn_model.fit(x_train, y_train)\n",
    "\n",
    "# Predict the response for the test dataset\n",
    "y_pred = best_knn_model.predict(x_test)\n",
    "\n",
    "# Print the confusion matrix\n",
    "print('Confusion Matrix:')\n",
    "print(confusion_matrix(y_test, y_pred).T)\n",
    "\n",
    "# Calculate and print the total cost\n",
    "total_cost = np.sum(confusion_matrix(y_test, y_pred).T * cost_matrix)\n",
    "print(f\"\\nTotal Cost: {total_cost}\") \n",
    "\n",
    "# Print the classification report for additional evaluation metrics\n",
    "print('\\nClassification Report:')\n",
    "print(classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"]))\n",
    "\n",
    "# Extract macro metrics from the classification report\n",
    "metrics_dict = classification_report(y_test, y_pred, target_names=[\"Reported having MI or CHD\", \"Did not report having MI or CHD\"], output_dict=True)['macro avg']\n",
    "\n",
    "# Update the performance metrics DataFrame with KNN results\n",
    "BorderlineSMOTE_loss.loc['Total Cost baseline', 'KNN'] = total_cost\n",
    "BorderlineSMOTE_precision.loc['macro precision baseline', 'KNN'] = metrics_dict['precision']\n",
    "BorderlineSMOTE_recall.loc['macro recall baseline', 'KNN'] = metrics_dict['recall']\n",
    "BorderlineSMOTE_f1_score.loc['macro f1-score baseline', 'KNN'] = metrics_dict['f1-score']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ***Classification with NLP Methods***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convert Original Data to Text Sequences"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sentences from Variable Values\n",
    "Create a small sentence from each variable value according to the 2021 SMART BRFSS Survey Codebook"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "sequences = {\n",
    "    # Sec 1 Health Status\n",
    "    '_RFHLTH': {\n",
    "        1: \"i think my health is good or very good\",\n",
    "        2: \"i think my health is fair or poor\"\n",
    "    },\n",
    "\n",
    "    # Sec 2 Healthy Days\n",
    "    '_PHYS14D': {\n",
    "        1: \"my health has not been good for zero days\",\n",
    "        2: \"my health has not been good for 1-13 days\",\n",
    "        3: \"my health has not been good for 14 or more days\"\n",
    "    },\n",
    "\n",
    "    '_MENT14D': {\n",
    "        1: \"my mental health has not been good for zero days\",\n",
    "        2: \"my mental health has not been good for 1-13 days\",\n",
    "        3: \"my mental health has not been good for 14 or more days\"\n",
    "    },\n",
    "\n",
    "    # Sec 3 Health Care Access\n",
    "    '_HLTHPLN': {\n",
    "        1: \"i had some form of health insurance\",\n",
    "        2: \"i didn't have some form of health insurance\"\n",
    "    },\n",
    "\n",
    "    '_HCVU652': {\n",
    "        1: \"i have some form of health insurance\",\n",
    "        2: \"i don't have some form of health insurance\"\n",
    "    },\n",
    "\n",
    "    # Sec 4 Exercise\n",
    "    '_TOTINDA': {\n",
    "        1: \"i have exercised during the past 30 days\",\n",
    "        2: \"i haven't exercised during the past 30 days\"\n",
    "    },\n",
    "\n",
    "    # Sec 5 Hypertension Awareness\n",
    "    '_RFHYPE6': {\n",
    "        1: \"i haven't been told that my blood pressure is high\",\n",
    "        2: \"i have been told that my blood pressure is high\"\n",
    "    },\n",
    "\n",
    "    # Sec 6 Cholesterol Awareness\n",
    "    '_CHOLCH3': {\n",
    "        1: \"i have had my cholesterol checked within the past five years\",\n",
    "        2: \"i haven't had my cholesterol checked within the past five years\",\n",
    "        3: \"i have never had my cholesterol checked\"\n",
    "    },\n",
    "\n",
    "    '_RFCHOL3': {\n",
    "        1: \"i have checked my cholesterol and it's not high\",\n",
    "        2: \"i have checked my cholesterol and it's high\"\n",
    "    },\n",
    "\n",
    "    # Sec 7 Chronic Health Conditions\n",
    "    \n",
    "\n",
    "    # Sec 8 Arthritis\n",
    "    \n",
    "\n",
    "    # Sec 11 Tobacco Use\n",
    "    '_SMOKER3': {\n",
    "        1: \"i smoke every day\",\n",
    "        2: \"i smoke some days\",\n",
    "        3: \"i'm a former smoker\",\n",
    "        4: \"i've never smoked\"\n",
    "    },\n",
    "\n",
    "    # Sec 12 Alcohol Consumption\n",
    "    '_RFBING5': {\n",
    "        1: \"i didn't drink in the past 30 days, but if i did, i didn't drink five or more drinks on an occasion\",\n",
    "        2: \"i did drink in the past 30 days and had five or more drinks on one or more occasions in the past month\"\n",
    "    },\n",
    "\n",
    "    '_RFDRHV7': {\n",
    "        1: \"i'm not a heavy drinker\",\n",
    "        2: \"i'm a heavy drinker\"\n",
    "    },\n",
    "\n",
    "    # Sec 14 HIV/AIDS\n",
    "    '_AIDTST4': {\n",
    "        1: \"i have been tested for HIV\",\n",
    "        2: \"i haven't been tested for HIV\"\n",
    "    },\n",
    "\n",
    "    # Sec 15 Fruits & Vegetables\n",
    "    '_FRTLT1A': {\n",
    "        1: \"i eat fruits one or more times per day\",\n",
    "        2: \"i don't eat fruits one or more times per day\"\n",
    "    },\n",
    "\n",
    "    '_VEGLT1A': {\n",
    "        1: \"i eat vegetables one or more times per day\",\n",
    "        2: \"i don't eat vegetables one or more times per day\"\n",
    "    },\n",
    "    \n",
    "    # Sec 9 Demographics\n",
    "    '_INCOMG1': {\n",
    "        1: \"my inclome is less than 15,000\",\n",
    "        2: \"my income ranges from 15,000 to 25,000\",\n",
    "        3: \"my income ranges from 25,000 to 35,000\",\n",
    "        4: \"my income ranges from 35,000 to 50,000\",\n",
    "        5: \"my income ranges from 50,000 to 100,000\",\n",
    "        6: \"my income ranges from 100,000 to 200,000\",\n",
    "        7: \"my income ranges from 200,000 and higher\"\n",
    "    },\n",
    "\n",
    "    '_EDUCAG': {\n",
    "        1: \"i didn't graduate high school\",\n",
    "        2: \"i graduated high school\",\n",
    "        3: \"i attended college\",\n",
    "        4: \"i graduated from college\"\n",
    "    },\n",
    "\n",
    "    '_RFBMI5': {\n",
    "        1: \"my BMI is less than 25.00\",\n",
    "        2: \"my BMI is greater than 25.00\"\n",
    "    },\n",
    "\n",
    "    '_BMI5CAT': {\n",
    "        1: \"i am underweight\",\n",
    "        2: \"my weight is normal\",\n",
    "        3: \"i am overweight\",\n",
    "        4: \"i am obese\"\n",
    "    },\n",
    "\n",
    "    '_AGE_G': {\n",
    "        1: \"my age is 18 to 24\",\n",
    "        2: \"my age is 25 to 34\",\n",
    "        3: \"my age is 35 to 44\",\n",
    "        4: \"my age is 45 to 54\",\n",
    "        5: \"my age is 55 to 64\",\n",
    "        6: \"my age is 65 or older\",\n",
    "    },\n",
    "\n",
    "    '_SEX': {\n",
    "        1: \"i'm a male\",\n",
    "        2: \"i'm a female\"\n",
    "    },\n",
    "\n",
    "    '_RACE': {\n",
    "        1: \"i'm white\",\n",
    "        2: \"i'm black\",\n",
    "        3: \"i'm american indian or alaskan\",\n",
    "        4: \"i'm asian\",\n",
    "        5: \"i'm native hawaiian or other pacific islander\",\n",
    "        6: \"i belong to some other race\",\n",
    "        7: \"i'm multiracial\",\n",
    "        8: \"i'm hispanic\"\n",
    "    }\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Target Variable: '_MICHD': {\n",
    "        1: \"i had MI or CHD\",\n",
    "        2: \"i haven't had MI or CHD\"\n",
    "    },"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### String Variable Values\n",
    "In the 2021 SMART BRFSS Codebook there are numerical values that correspond to string values, which represent a short description of the variable. The description of each variable value is used to create the small sentence. Here, we use the corresponding string values which might be the optimal and most practical choice of converting the numerical values to text. The reason why is obvious; we don't have to manually write small sentences for each variable value, we hust have to map each numerical value to its corresponding string value which already exists in the 2021 SMART Codebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "string_var_values = {\n",
    "    # Sec 1 Health Status\n",
    "    '_RFHLTH': {\n",
    "        1: \"Good or Better Health\",\n",
    "        2: \"Fair or Poor Health\"\n",
    "    },\n",
    "\n",
    "    # Sec 2 Healthy Days\n",
    "    '_PHYS14D': {\n",
    "        1: \"Zero days when physical health not good\",\n",
    "        2: \"1-13 days when physical health not good\",\n",
    "        3: \"14+ days when physical health not good\"\n",
    "    },\n",
    "\n",
    "    '_MENT14D': {\n",
    "        1: \"Zero days when mental health not good\",\n",
    "        2: \"1-13 days when mental health not good\",\n",
    "        3: \"14+ days when mental health not good\"\n",
    "    },\n",
    "\n",
    "    # Sec 3 Health Care Access\n",
    "    '_HLTHPLN': {\n",
    "        1: \"Have some form of insurance\",\n",
    "        2: \"Do not have some form of health insurance\"\n",
    "    },\n",
    "\n",
    "    '_HCVU652': {\n",
    "        1: \"Have some form of health insurance\",\n",
    "        2: \"Do not have any form of health insurance\"\n",
    "    },\n",
    "\n",
    "    # Sec 4 Exercise\n",
    "    '_TOTINDA': {\n",
    "        1: \"Had physical activity or exercise\",\n",
    "        2: \"No physical activity or exercise in last 30 days\"\n",
    "    },\n",
    "\n",
    "    # Sec 5 Hypertension Awareness\n",
    "    '_RFHYPE6': {\n",
    "        1: \"_RFHYPE6-No\",\n",
    "        2: \"_RFHYPE6-Yes\"\n",
    "    },\n",
    "\n",
    "    # Sec 6 Cholesterol Awareness\n",
    "    '_CHOLCH3': {\n",
    "        1: \"Had cholesterol checked in past 5 years\",\n",
    "        2: \"Did not have cholesterol checked in past 5 years\",\n",
    "        3: \"Have never had cholesterol checked\"\n",
    "    },\n",
    "\n",
    "    '_RFCHOL3': {\n",
    "        1: \"_RFCHOL3-No\",\n",
    "        2: \"_RFCHOL3-Yes\"\n",
    "    },\n",
    "\n",
    "    # Sec 7 Chronic Health Conditions\n",
    "    \n",
    "\n",
    "    # Sec 8 Arthritis\n",
    "    \n",
    "\n",
    "    # Sec 11 Tobacco Use\n",
    "    '_SMOKER3': {\n",
    "        1: \"Current smoker-now smokes every day\",\n",
    "        2: \"Current smoker-now smokes some days\",\n",
    "        3: \"Former smoker\",\n",
    "        4: \"Never smoked\"\n",
    "    },\n",
    "\n",
    "    # Sec 12 Alcohol Consumption\n",
    "    '_RFBING5': {\n",
    "        1: \"_RFBING5-No\",\n",
    "        2: \"_RFBING5-Yes\"\n",
    "    },\n",
    "\n",
    "    '_RFDRHV7': {\n",
    "        1: \"_RFDRHV7-No\",\n",
    "        2: \"_RFDRHV7-Yes\"\n",
    "    },\n",
    "\n",
    "    # Sec 14 HIV/AIDS\n",
    "    '_AIDTST4': {\n",
    "        1: \"_AIDTST4-Yes\",\n",
    "        2: \"_AIDTST4-No\"\n",
    "    },\n",
    "\n",
    "    # Sec 15 Fruits & Vegetables\n",
    "    '_FRTLT1A': {\n",
    "        1: \"Consumed fruit one or more times per day\",\n",
    "        2: \"Consumed fruit less than one time per day\"\n",
    "    },\n",
    "\n",
    "    '_VEGLT1A': {\n",
    "        1: \"Consumed vegetables one or more times per day\",\n",
    "        2: \"Consumed vegetables less than one time per day\"\n",
    "    },\n",
    "    \n",
    "    # Sec 9 Demographics\n",
    "    '_INCOMG1': {\n",
    "        1: \"Less than $15,000\",\n",
    "        2: \"$15,000 to less than $25,000\",\n",
    "        3: \"$25,000 to less than $35,000\",\n",
    "        4: \"$35,000 to less than $50,000\",\n",
    "        5: \"$50,000 to less than $100,000\",\n",
    "        6: \"$100,000 to less than $200,000\",\n",
    "        7: \"$200,000 or more\"\n",
    "    },\n",
    "\n",
    "    '_EDUCAG': {\n",
    "        1: \"Did not graduate High School\",\n",
    "        2: \"Graduated High School\",\n",
    "        3: \"Attended College or Technical School\",\n",
    "        4: \"Graduated from College or Technical School\"\n",
    "    },\n",
    "\n",
    "    '_RFBMI5': {\n",
    "        1: \"_RFBMI5-No\",\n",
    "        2: \"_RFBMI5-Yes\"\n",
    "    },\n",
    "\n",
    "    '_BMI5CAT': {\n",
    "        1: \"Underweight\",\n",
    "        2: \"Normal Weight\",\n",
    "        3: \"Overweight\",\n",
    "        4: \"Obese\"\n",
    "    },\n",
    "\n",
    "    '_AGE_G': {\n",
    "        1: \"my age is 18 to 24\",\n",
    "        2: \"my age is 25 to 34\",\n",
    "        3: \"my age is 35 to 44\",\n",
    "        4: \"my age is 45 to 54\",\n",
    "        5: \"my age is 55 to 64\",\n",
    "        6: \"my age is 65 or older\",\n",
    "    },\n",
    "\n",
    "    '_SEX': {\n",
    "        1: \"Male\",\n",
    "        2: \"Female\"\n",
    "    },\n",
    "\n",
    "    '_RACE': {\n",
    "        1: \"White only, non-Hispanic\",\n",
    "        2: \"Black only, non-Hispanic\",\n",
    "        3: \"American Indian or Alaskan Native only, Non-Hispanic\",\n",
    "        4: \"Asian only, non-Hispanic\",\n",
    "        5: \"Native Hawaiian or other Pacific Islander only, Non-Hispanic\",\n",
    "        6: \"Other race only, non-Hispanic\",\n",
    "        7: \"Multiracial, non-Hispanic\",\n",
    "        8: \"Hispanic\"\n",
    "    }\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Target Variable: '_MICHD': {\n",
    "        1: \"Reported having MI or CHD\",\n",
    "        2: \"Did not report having MI or CHD\"\n",
    "    },"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### New Text Dataset Creation\n",
    "We create text datasets based on the transformation of the numerical values to string/text values. Two datasets will be created:\n",
    "- Dataset A: This dataset will transform each variable value into a small sentence\n",
    "- Dataset B: This dataset will transform each variable value into its corresponding text value, according to the text values provided in \"2021 SMART BRFSS Survey Codebook\"."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The following algorithm implements the generation of text sequences from the original data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                Sentence  Classification\n",
      "0      i think my health is good or very good, my hea...               0\n",
      "1      i think my health is good or very good, my hea...               0\n",
      "2      i think my health is good or very good, my hea...               0\n",
      "3      i think my health is good or very good, my hea...               0\n",
      "4      i think my health is good or very good, my hea...               0\n",
      "...                                                  ...             ...\n",
      "27739  i think my health is good or very good, my hea...               1\n",
      "27740  i think my health is good or very good, my hea...               1\n",
      "27741  i think my health is good or very good, my hea...               1\n",
      "27742  i think my health is good or very good, my hea...               1\n",
      "27743  i think my health is good or very good, my hea...               1\n",
      "\n",
      "[27744 rows x 2 columns]\n",
      "\n",
      "                                                Sentence  Classification\n",
      "0      Good or Better Health, Zero days when physical...               0\n",
      "1      Good or Better Health, 14+ days when physical ...               0\n",
      "2      Good or Better Health, Zero days when physical...               0\n",
      "3      Good or Better Health, 1-13 days when physical...               0\n",
      "4      Good or Better Health, Zero days when physical...               0\n",
      "...                                                  ...             ...\n",
      "27739  Good or Better Health, 1-13 days when physical...               1\n",
      "27740  Good or Better Health, Zero days when physical...               1\n",
      "27741  Good or Better Health, 1-13 days when physical...               1\n",
      "27742  Good or Better Health, Zero days when physical...               1\n",
      "27743  Good or Better Health, 1-13 days when physical...               1\n",
      "\n",
      "[27744 rows x 2 columns]\n"
     ]
    }
   ],
   "source": [
    "# Create an empty list to store rows for the two datasets\n",
    "new_dataset_rows = [\n",
    "    [],\n",
    "    []\n",
    "]\n",
    "\n",
    "# Create an empty list to store the two datasets\n",
    "text_data = []\n",
    "\n",
    "# Iterate through each row in the original dataset\n",
    "for _, row in dataframe_sample.iterrows():\n",
    "    sentence_parts = [\n",
    "        [],  # To store parts of the sentence for each row for dataset A\n",
    "        []   # To store parts of the sentence for each row for dataset B\n",
    "    ]\n",
    "\n",
    "    # Iterate through each column in the original dataset\n",
    "    for column_name, value in row.items():\n",
    "        # Check if the current column is in the sequences dictionary. Both the sequences and the string_var_values dictionaries have the same keys.\n",
    "        if (column_name in sequences) and (value in sequences[column_name]):\n",
    "\n",
    "            # === Dataset A ===\n",
    "            # Get the description for the current value of the variable\n",
    "            description = sequences[column_name].get(value, '')\n",
    "            # Add a comma to the end of the description\n",
    "            description += ','\n",
    "            # Add the description to the sentence parts for dataset A\n",
    "            sentence_parts[0].append(description)\n",
    "\n",
    "            # === Dataset B ===\n",
    "            # Get the description for the current value of the variable\n",
    "            description = string_var_values[column_name].get(value, '')\n",
    "            # Add a comma to the end of the description\n",
    "            description += ','\n",
    "            # Add the description to the sentence parts for dataset B\n",
    "            sentence_parts[1].append(description)\n",
    "        \n",
    "        # Check if the current column is the target variable\n",
    "        elif column_name == '_MICHD':\n",
    "            # Add the target variable's value to the second column of both datasets\n",
    "            classification_value = value\n",
    "    \n",
    "    # === Dataset A ===\n",
    "    # Concatenate the sentence parts into a single sentence\n",
    "    sentence = ' '.join(sentence_parts[0])\n",
    "    \n",
    "    # Create a new row for dataset A\n",
    "    new_dataset_rows[0].append({'Sentence': sentence, 'Classification': classification_value})\n",
    "\n",
    "    # === Dataset B ===\n",
    "    # Concatenate the sentence parts into a single sentence \n",
    "    sentence = ' '.join(sentence_parts[1])\n",
    "    \n",
    "    # Create a new row for dataset B\n",
    "    new_dataset_rows[1].append({'Sentence': sentence, 'Classification': classification_value})\n",
    "\n",
    "# Create the new dataset from the list of rows\n",
    "text_data.append(pd.DataFrame(new_dataset_rows[0]))\n",
    "text_data.append(pd.DataFrame(new_dataset_rows[1]))\n",
    "\n",
    "# Display dataset A\n",
    "print(text_data[0])\n",
    "\n",
    "print()\n",
    "\n",
    "# Display dataset B\n",
    "print(text_data[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Capitalize the \"i\" in the beginning of each sentence, remove the comma from the end fo the sentence and replace it with a period.\n",
    "text_data[0]['Sentence'] = text_data[0]['Sentence'].str.rstrip(',')  # Remove comma from the end\n",
    "text_data[0]['Sentence'] = text_data[0]['Sentence'].str.capitalize()  # Capitalize \"I\" at the beginning\n",
    "text_data[0]['Sentence'] = text_data[0]['Sentence'].str.replace(r'\\bi\\b', 'I')  # Capitalize standalone \"i\"\n",
    "text_data[0]['Sentence'] = text_data[0]['Sentence'] + '.'  # Replace comma with a period at the end"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                Sentence  Classification\n",
      "0      I think my health is good or very good, my hea...               0\n",
      "1      I think my health is good or very good, my hea...               0\n",
      "2      I think my health is good or very good, my hea...               0\n",
      "3      I think my health is good or very good, my hea...               0\n",
      "4      I think my health is good or very good, my hea...               0\n",
      "...                                                  ...             ...\n",
      "27739  I think my health is good or very good, my hea...               1\n",
      "27740  I think my health is good or very good, my hea...               1\n",
      "27741  I think my health is good or very good, my hea...               1\n",
      "27742  I think my health is good or very good, my hea...               1\n",
      "27743  I think my health is good or very good, my hea...               1\n",
      "\n",
      "[27744 rows x 2 columns]\n"
     ]
    }
   ],
   "source": [
    "print(text_data[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "text_data[1]['Sentence'] = text_data[1]['Sentence'].str.rstrip(',')  # Remove comma from the end"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Based on the value (custom sentence, string/text), convert the \"_MICHD\" target variable values to 1 and 0\n",
    "text_data['Classification'] = text_data['Classification'].map({'Yes': 1, 'No': 0})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Classification\n",
       "0    13872\n",
       "1    13872\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text_data[0]['Classification'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Classification\n",
       "0    13872\n",
       "1    13872\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text_data[1]['Classification'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Sentence</th>\n",
       "      <th>Classification</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>I think my health is good or very good, my hea...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>I think my health is good or very good, my hea...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>I think my health is good or very good, my hea...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>I think my health is good or very good, my hea...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>I think my health is good or very good, my hea...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            Sentence  Classification\n",
       "0  I think my health is good or very good, my hea...               0\n",
       "1  I think my health is good or very good, my hea...               0\n",
       "2  I think my health is good or very good, my hea...               0\n",
       "3  I think my health is good or very good, my hea...               0\n",
       "4  I think my health is good or very good, my hea...               0"
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text_data[0].head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the new sentence dataset. This dataset will be used for sequence classification tasks.\n",
    "text_data[1].to_excel(\"Text_Sequence_Dataset.xlsx\",  index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Wordcloud"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function to generate word cloud for a specific class\n",
    "def generate_word_cloud(text):\n",
    "    wordcloud = WordCloud(width = 800, height = 800, \n",
    "                background_color ='white', \n",
    "                stopwords = None, \n",
    "                min_font_size = 10).generate(text)\n",
    "    return wordcloud"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Tokenize text and count word frequency for each class\n",
    "class_data = {}\n",
    "for class_name, group in text_data[1].groupby('Classification'):\n",
    "    text = ' '.join(group['Sentence'])\n",
    "    tokens = word_tokenize(text)\n",
    "    word_freq = ' '.join(tokens)\n",
    "    class_data[class_name] = word_freq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate word cloud for each class\n",
    "for class_name, text in class_data.items():\n",
    "    wordcloud = generate_word_cloud(text)\n",
    "    plt.figure(figsize = (8, 8), facecolor = None) \n",
    "    plt.imshow(wordcloud) \n",
    "    plt.axis(\"off\") \n",
    "    plt.tight_layout(pad = 0) \n",
    "    plt.title(f'Word Cloud for {class_name}')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classification using Text-based Classifiers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We can add the \"NOT_\" word infront of words that come after negation words\n",
    "def add_not_to_words(sentence):\n",
    "    words = sentence.split()\n",
    "    new_sentence = []\n",
    "\n",
    "    add_not = False\n",
    "\n",
    "    for word in words:\n",
    "        if add_not:\n",
    "            new_sentence.append(\"NOT_\" + word)\n",
    "            # Stop adding \"NOT_\" when a comma is encountered\n",
    "            if \",\" in word:\n",
    "                add_not = False\n",
    "        else:\n",
    "            new_sentence.append(word)\n",
    "\n",
    "        if word.lower() in ['not', \"didn't\", \"don't\", \"haven't\", \"never\"]:\n",
    "            add_not = True\n",
    "\n",
    "    return ' '.join(new_sentence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "text_dataset = text_data[1].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Apply the function to every sentence in the dataset (optional)\n",
    "text_dataset['Sentence'] = text_dataset['Sentence'].apply(add_not_to_words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "COST = [\n",
    "    [0, 1],\n",
    "    [1, 0]\n",
    "]\n",
    "\n",
    "text_based_classifiers = {}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Multinomial NB\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      "[[2740  728]\n",
      " [1139 2329]]\n",
      "\n",
      "Total Cost: 1867\n",
      "\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.71      0.79      0.75      3468\n",
      "           1       0.76      0.67      0.71      3468\n",
      "\n",
      "    accuracy                           0.73      6936\n",
      "   macro avg       0.73      0.73      0.73      6936\n",
      "weighted avg       0.73      0.73      0.73      6936\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Split the dataset into training and testing sets\n",
    "x_train, x_test, y_train, y_test = train_test_split(text_dataset['Sentence'], text_dataset['Classification'], test_size=0.25, random_state=SEED, stratify=text_dataset['Classification'])\n",
    "\n",
    "weights = np.zeros(y_train.shape[0])\n",
    "weights[np.where(y_train == 0)] = 1\n",
    "weights[np.where(y_train == 1)] = 1\n",
    "\n",
    "# Create a CountVectorizer to convert the text data into numerical features\n",
    "vectorizer = CountVectorizer()\n",
    "x_train_vectorized = vectorizer.fit_transform(x_train)\n",
    "x_test_vectorized = vectorizer.transform(x_test)\n",
    "\n",
    "# Define TF-IDF vectorizer\n",
    "#tfidf_vectorizer = TfidfVectorizer()\n",
    "\n",
    "# Train a Multinomial Naive Bayes classifier\n",
    "clf = MultinomialNB()\n",
    "\n",
    "# Create a pipeline with TF-IDF vectorizer and Multinomial NB classifier\n",
    "#pipeline = Pipeline([('tfidf', tfidf_vectorizer), ('clf', clf)])\n",
    "\n",
    "clf = CalibratedClassifierCV(clf, method='isotonic')\n",
    "clf.fit(x_train_vectorized, y_train, sample_weight=weights)\n",
    "\n",
    "\n",
    "text_based_classifiers['Multinomial NB'] = clf\n",
    "\n",
    "\n",
    "# Make predictions on the test set\n",
    "y_pred = clf.predict(x_test_vectorized)\n",
    "\n",
    "\n",
    "# Calculate the confusion matrix\n",
    "conf_matrix = confusion_matrix(y_test, y_pred)\n",
    "print(f'Confusion Matrix: \\n{conf_matrix}\\n')\n",
    "\n",
    "# Calculate the total cost\n",
    "total_cost = np.sum(conf_matrix * COST)\n",
    "print(f'Total Cost: {total_cost}\\n')\n",
    "\n",
    "\n",
    "# Evaluate the performance\n",
    "classification_rep = classification_report(y_test, y_pred)\n",
    "\n",
    "# Display the results\n",
    "print('Classification Report:')\n",
    "print(classification_rep)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Multi-variate Bernoulli NB\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      "[[2698  770]\n",
      " [1121 2347]]\n",
      "\n",
      "Total Cost: 1891\n",
      "\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.71      0.78      0.74      3468\n",
      "           1       0.75      0.68      0.71      3468\n",
      "\n",
      "    accuracy                           0.73      6936\n",
      "   macro avg       0.73      0.73      0.73      6936\n",
      "weighted avg       0.73      0.73      0.73      6936\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Split the dataset into training and testing sets\n",
    "x_train, x_test, y_train, y_test = train_test_split(text_dataset['Sentence'], text_dataset['Classification'], test_size=0.25, random_state=SEED, stratify=text_dataset['Classification'])\n",
    "\n",
    "weights = np.zeros(y_train.shape[0])\n",
    "weights[np.where(y_train == 0)] = 1\n",
    "weights[np.where(y_train == 1)] = 1\n",
    "\n",
    "# Create a CountVectorizer to convert the text data into numerical features\n",
    "vectorizer = CountVectorizer()\n",
    "x_train_vectorized = vectorizer.fit_transform(x_train)\n",
    "x_test_vectorized = vectorizer.transform(x_test)\n",
    "\n",
    "\n",
    "# Define TF-IDF vectorizer\n",
    "#tfidf_vectorizer = TfidfVectorizer()\n",
    "\n",
    "\n",
    "# Train a Multinomial Naive Bayes classifier\n",
    "clf = BernoulliNB()\n",
    "\n",
    "#pipeline = Pipeline([('tfidf', tfidf_vectorizer), ('clf', clf)])\n",
    "\n",
    "clf = CalibratedClassifierCV(clf, method='isotonic')\n",
    "clf.fit(x_train_vectorized, y_train, sample_weight=weights)\n",
    "\n",
    "\n",
    "text_based_classifiers['Multivariate Bernoulli NB'] = clf\n",
    "\n",
    "\n",
    "# Make predictions on the test set\n",
    "y_pred = clf.predict(x_test_vectorized)\n",
    "\n",
    "\n",
    "# Calculate the confusion matrix\n",
    "conf_matrix = confusion_matrix(y_test, y_pred)\n",
    "print(f'Confusion Matrix: \\n{conf_matrix}\\n')\n",
    "\n",
    "# Calculate the total cost\n",
    "total_cost = np.sum(conf_matrix * COST)\n",
    "print(f'Total Cost: {total_cost}\\n')\n",
    "\n",
    "\n",
    "# Evaluate the performance\n",
    "classification_rep = classification_report(y_test, y_pred, zero_division=0.0)\n",
    "\n",
    "# Display the results\n",
    "print('Classification Report:')\n",
    "print(classification_rep)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Binary Multinomail NB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a preprocessing function\n",
    "def preprocess_text(text):\n",
    "    return ' '.join(set(text.split()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      "[[2799  669]\n",
      " [1217 2251]]\n",
      "\n",
      "Total Cost: 1886\n",
      "\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.70      0.81      0.75      3468\n",
      "           1       0.77      0.65      0.70      3468\n",
      "\n",
      "    accuracy                           0.73      6936\n",
      "   macro avg       0.73      0.73      0.73      6936\n",
      "weighted avg       0.73      0.73      0.73      6936\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Split the dataset into training and testing sets\n",
    "x_train, x_test, y_train, y_test = train_test_split(text_dataset['Sentence'], text_dataset['Classification'], test_size=0.25, random_state=SEED, stratify=text_dataset['Classification'])\n",
    "\n",
    "weights = np.zeros(y_train.shape[0])\n",
    "weights[np.where(y_train == 0)] = 1\n",
    "weights[np.where(y_train == 1)] = 1\n",
    "\n",
    "# Create a CountVectorizer to convert the text data into numerical features\n",
    "vectorizer = CountVectorizer()\n",
    "x_train_vectorized = vectorizer.fit_transform(x_train.apply(lambda x: ' '.join(set(x.split()))))\n",
    "x_test_vectorized = vectorizer.transform(x_test.apply(lambda x: ' '.join(set(x.split()))))\n",
    "\n",
    "# Define TF-IDF vectorizer\n",
    "#tfidf_vectorizer = TfidfVectorizer(preprocessor=preprocess_text)\n",
    "\n",
    "# Train a Multinomial Naive Bayes classifier\n",
    "clf = MultinomialNB()\n",
    "\n",
    "#pipeline = Pipeline([('tfidf', tfidf_vectorizer), ('clf', clf)])\n",
    "\n",
    "clf = CalibratedClassifierCV(clf, method='isotonic')\n",
    "clf.fit(x_train_vectorized, y_train, sample_weight=weights)\n",
    "\n",
    "\n",
    "text_based_classifiers['Binary NB'] = clf\n",
    "\n",
    "# Make predictions on the test set\n",
    "y_pred = clf.predict(x_test_vectorized)\n",
    "\n",
    "\n",
    "# Calculate the confusion matrix\n",
    "conf_matrix = confusion_matrix(y_test, y_pred)\n",
    "print(f'Confusion Matrix: \\n{conf_matrix}\\n')\n",
    "\n",
    "# Calculate the total cost\n",
    "total_cost = np.sum(conf_matrix * COST)\n",
    "print(f'Total Cost: {total_cost}\\n')\n",
    "\n",
    "\n",
    "# Evaluate the performance\n",
    "classification_rep = classification_report(y_test, y_pred)\n",
    "\n",
    "# Display the results\n",
    "print('Classification Report:')\n",
    "print(classification_rep)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the dataset into training and testing sets\n",
    "x_train, x_test, y_train, y_test = train_test_split(text_dataset['Sentence'], text_dataset['Classification'], test_size=0.25, random_state=SEED, stratify=text_dataset['Classification'])\n",
    "\n",
    "weights = np.zeros(y_train.shape[0])\n",
    "weights[np.where(y_train == 0)] = 1\n",
    "weights[np.where(y_train == 1)] = 1\n",
    "\n",
    "# Create a CountVectorizer to convert the text data into numerical features\n",
    "#vectorizer = CountVectorizer()\n",
    "#x_train_vectorized = vectorizer.fit_transform(x_train)\n",
    "#x_test_vectorized = vectorizer.transform(x_test)\n",
    "\n",
    "tfidf_vectorizer = TfidfVectorizer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a custom scoring function using the cost matrix\n",
    "def custom_dt_score(y_true, y_pred):\n",
    "    cm = confusion_matrix(y_true, y_pred)\n",
    "    cost = np.sum(cm * COST)\n",
    "    return -cost  # Minimize cost, so negative of cost is returned"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make the custom scoring function into a scorer object\n",
    "custom_dt_scorer = make_scorer(custom_dt_score, greater_is_better = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_grid = {\n",
    "    'max_depth': [5, 10, 15, None],\n",
    "    'min_samples_split': [100, 200, 500],\n",
    "    'min_samples_leaf': [50, 100, 200],\n",
    "    'class_weight':[{0: 1, 1: 1}]\n",
    "}\n",
    "\n",
    "grid_search = GridSearchCV(DecisionTreeClassifier(random_state=SEED), param_grid, scoring='f1_weighted')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grid_search.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameters: {'class_weight': {0: 1, 1: 1}, 'max_depth': 10, 'min_samples_leaf': 200, 'min_samples_split': 100}\n"
     ]
    }
   ],
   "source": [
    "best_params = grid_search.best_params_\n",
    "print(\"Best parameters:\", best_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/georgematlis/AUTH/DWS/ADVANCED_TOPICS_IN_MACHINE_LEARNING/CDC-BRFSS/lib/python3.10/site-packages/sklearn/calibration.py:367: UserWarning: Since Pipeline does not appear to accept sample_weight, sample weights will only be used for the calibration itself. This can be caused by a limitation of the current scikit-learn API. See the following issue for more details: https://github.com/scikit-learn/scikit-learn/issues/21134. Be warned that the result of the calibration is likely to be incorrect.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      "[[2756  712]\n",
      " [1075 2393]]\n",
      "\n",
      "Total Cost: 980.75\n",
      "\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.72      0.79      0.76      3468\n",
      "           1       0.77      0.69      0.73      3468\n",
      "\n",
      "    accuracy                           0.74      6936\n",
      "   macro avg       0.75      0.74      0.74      6936\n",
      "weighted avg       0.75      0.74      0.74      6936\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Train a Decision Tree classifier\n",
    "dt = DecisionTreeClassifier(max_depth=15, min_samples_split=200, min_samples_leaf=200)\n",
    "\n",
    "pipeline = Pipeline([\n",
    "    ('tfidf', tfidf_vectorizer),\n",
    "    ('clf', dt)\n",
    "])\n",
    "\n",
    "clf = CalibratedClassifierCV(pipeline, method='isotonic')\n",
    "clf.fit(x_train_vectorized, y_train, sample_weight=weights)\n",
    "\n",
    "text_based_classifiers['DT'] = clf\n",
    "\n",
    "\n",
    "# Make predictions on the test set\n",
    "y_pred = clf.predict(x_test)\n",
    "\n",
    "\n",
    "# Calculate the confusion matrix\n",
    "conf_matrix = confusion_matrix(y_test, y_pred)\n",
    "print(f'Confusion Matrix: \\n{conf_matrix}\\n')\n",
    "\n",
    "# Calculate the total cost\n",
    "total_cost = np.sum(conf_matrix * COST)\n",
    "print(f'Total Cost: {total_cost}\\n')\n",
    "\n",
    "\n",
    "# Evaluate the performance\n",
    "classification_rep = classification_report(y_test, y_pred)\n",
    "\n",
    "# Display the results\n",
    "print('Classification Report:')\n",
    "print(classification_rep)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CatBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the dataset into training and testing sets\n",
    "x_train, x_test, y_train, y_test = train_test_split(text_dataset['Sentence'], text_dataset['Classification'], test_size=0.25, random_state=SEED, stratify=text_dataset['Classification'])\n",
    "\n",
    "weights = np.zeros(y_train.shape[0])\n",
    "weights[np.where(y_train == 0)] = 1\n",
    "weights[np.where(y_train == 1)] = 1\n",
    "\n",
    "# Create a CountVectorizer to convert the text data into numerical features\n",
    "vectorizer = CountVectorizer()\n",
    "x_train_vectorized = vectorizer.fit_transform(x_train)\n",
    "x_test_vectorized = vectorizer.transform(x_test)\n",
    "\n",
    "#tfidf_vectorizer = TfidfVectorizer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a custom scoring function using the cost matrix\n",
    "def custom_rf_score(y_true, y_pred):\n",
    "    cm = confusion_matrix(y_true, y_pred)\n",
    "    cost = np.sum(cm * COST)\n",
    "    return -cost  # Minimize cost, so negative of cost is returned"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make the custom scoring function into a scorer object\n",
    "custom_rf_scorer = make_scorer(custom_rf_score, greater_is_better = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "param_dist = {\n",
    "    'iterations': randint(100, 1000),\n",
    "    'learning_rate': uniform(0.01, 0.3),\n",
    "    'depth': randint(3, 10),\n",
    "}\n",
    "\n",
    "cat_clf = CatBoostClassifier(random_state=SEED)\n",
    "\n",
    "random_search = RandomizedSearchCV(cat_clf, param_grid, scoring='f1_weighted', n_jobs=4, n_iter=50, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_search.fit(x_train_vectorized, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best parameters: {'n_estimators': 300, 'max_depth': None}\n"
     ]
    }
   ],
   "source": [
    "best_params = random_search.best_params_\n",
    "print(\"Best parameters:\", best_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train a Decision Tree classifier\n",
    "cat_clf = CatBoostClassifier(random_state=SEED, n_estimators=300, max_depth=None)\n",
    "\n",
    "\n",
    "#pipeline = Pipeline([('tfidf', tfidf_vectorizer),('clf', rf)])\n",
    "\n",
    "clf = CalibratedClassifierCV(cat_clf, method='isotonic')\n",
    "clf.fit(x_train_vectorized, y_train, sample_weight=weights)\n",
    "\n",
    "text_based_classifiers['CatB'] = clf\n",
    "\n",
    "\n",
    "# Make predictions on the test set\n",
    "y_pred = clf.predict(x_test_vectorized)\n",
    "\n",
    "# Calculate the confusion matrix\n",
    "conf_matrix = confusion_matrix(y_test, y_pred)\n",
    "print(f'Confusion Matrix: \\n{conf_matrix}\\n')\n",
    "\n",
    "# Calculate the total cost\n",
    "total_cost = np.sum(conf_matrix * COST)\n",
    "print(f'Total Cost: {total_cost}\\n')\n",
    "\n",
    "\n",
    "# Evaluate the performance\n",
    "classification_rep = classification_report(y_test, y_pred)\n",
    "\n",
    "# Display the results\n",
    "print('Classification Report:')\n",
    "print(classification_rep)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ***Probability Calibration Plot***\n",
    "The probability calibration plot displays how confident the mehtods are for their predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA04AAAK9CAYAAAAT0TyCAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/H5lhTAAAACXBIWXMAAA9hAAAPYQGoP6dpAADHyklEQVR4nOzdd3QUVR/G8e/upldqCCVA6CC9F5FeBaT3Kog0pQmCgoCigL4CFor0jiK9iSIKSEcQBGkCoXcDCSE9O+8fK9EYIIkmbEKezzkcmbt3Zn4TRsiTe+eOyTAMAxEREREREXkss70LEBERERERSe0UnERERERERBKg4CQiIiIiIpIABScREREREZEEKDiJiIiIiIgkQMFJREREREQkAQpOIiIiIiIiCVBwEhERERERSYCCk4iIiIiISAIUnEREJEXUrFmTmjVrxm5fuHABk8nEggULYtu6d++Oh4fH0y/uH8aOHYvJZLJ3GU+UN29eunfvbu8yRETSLQUnEREB4Ny5c7z66qvky5cPFxcXvLy8qFatGp988glhYWH2Lu8/Cw0NZezYsWzfvt3epcRhMplif5nNZnLkyEH9+vWTrc5r164xduxYjhw5kizHExFJrxzsXYCIiNjfpk2baNOmDc7OznTt2pXixYsTGRnJrl27GDZsGL/99huzZs36T+fIkycPYWFhODo6JlPVSRMaGsq4ceMA4oyEAYwaNYoRI0bYoSqbevXq0bVrVwzDICAggOnTp1O7dm02bdpEo0aN/tOxr127xrhx48ibNy+lS5dOnoJFRNIhBScRkXQuICCA9u3bkydPHn744QeyZ88e+1n//v05e/YsmzZt+s/nMZlMuLi4/OfjPBQdHY3VasXJyek/H8vBwQEHB/v9k1ioUCE6d+4cu92iRQtKlizJ1KlT/3NwEhGR5KGpeiIi6dyHH35ISEgIc+fOjROaHipQoAADBw6M3Z4/fz61a9fGx8cHZ2dnihUrxowZMxI8z6OecXro/PnzNGjQAHd3d3LkyMG7776LYRjx9v3f//7H1KlTyZ8/P87Ozpw4cYLIyEjeeecdypUrh7e3N+7u7lSvXp0ff/wxzv5Zs2YFYNy4cbFT48aOHQs8+hmn6Oho3nvvvdhz5c2bl7feeouIiIg4/fLmzUuTJk3YtWsXFStWxMXFhXz58rFo0aIEvyaPU6JECbJkyUJAQMAT+50/f542bdqQKVMm3NzcqFy5cpyQu337dipUqABAjx49Yq/7UX8GIiLyZBpxEhFJ5zZs2EC+fPmoWrVqovrPmDGD5557jmbNmuHg4MCGDRvo168fVquV/v37J/n8MTExNGzYkMqVK/Phhx+yZcsWxowZQ3R0NO+++26cvvPnzyc8PJzevXvj7OxMpkyZCA4OZs6cOXTo0IFXXnmF+/fvM3fuXBo0aMCBAwcoXbo0WbNmZcaMGfTt25cWLVrQsmVLAEqWLPnYunr16sXChQtp3bo1Q4cOZf/+/UyYMIGTJ0+yZs2aOH3Pnj1L69at6dmzJ926dWPevHl0796dcuXK8dxzzyX5a3L37l3u3r1LgQIFHtvn5s2bVK1aldDQUF5//XUyZ87MwoULadasGStXrqRFixYULVqUd999l3feeYfevXtTvXp1gET/WYuIyN8YIiKSbgUFBRmA8dJLLyV6n9DQ0HhtDRo0MPLlyxenrUaNGkaNGjVitwMCAgzAmD9/fmxbt27dDMB47bXXYtusVqvx4osvGk5OTsbt27fj7Ovl5WXcunUrznmio6ONiIiIOG137941smXLZrz88suxbbdv3zYAY8yYMfHqHzNmjPH3fxKPHDliAEavXr3i9HvjjTcMwPjhhx9i2/LkyWMAxs6dO2Pbbt26ZTg7OxtDhw6Nd65/AoyePXsat2/fNm7dumXs37/fqFOnjgEYH3/8cZzzdOvWLXZ70KBBBmD89NNPsW337983/P39jbx58xoxMTGGYRjGwYMH433dRUQk6TRVT0QkHQsODgbA09Mz0fu4urrG/j4oKIg7d+5Qo0YNzp8/T1BQ0L+qY8CAAbG/N5lMDBgwgMjISL7//vs4/Vq1ahU75e4hi8US+5yT1WolMDCQ6Ohoypcvz+HDh/9VPZs3bwZgyJAhcdqHDh0KEO+Zr2LFisWO5gBkzZqVwoULc/78+USdb+7cuWTNmhUfHx8qVarE7t27GTJkCIMGDXpijRUrVuT555+PbfPw8KB3795cuHCBEydOJOrcIiKSOJqqJyKSjnl5eQFw//79RO+ze/duxowZw969ewkNDY3zWVBQEN7e3kmqwWw2ky9fvjhthQoVAmzPJv2dv7//I4+xcOFCPv74Y06dOkVUVFSC/RNy8eJFzGZzvKlyvr6+ZMiQgYsXL8Zpz507d7xjZMyYkbt37ybqfC+99BIDBgzAZDLh6enJc889h7u7e4I1VqpUKV570aJFYz8vXrx4os4vIiIJU3ASEUnHvLy8yJEjB8ePH09U/3PnzlGnTh2KFCnC5MmT8fPzw8nJic2bNzNlyhSsVmuK1vv30a6HlixZQvfu3WnevDnDhg3Dx8cHi8XChAkTOHfu3H86X2JfimuxWB7ZbvxtgYsnyZUrF3Xr1k10XSIi8vQpOImIpHNNmjRh1qxZ7N27lypVqjyx74YNG4iIiGD9+vVxRln+voJdUlmtVs6fPx87ygRw5swZwLZiXUJWrlxJvnz5WL16dZygM2bMmDj9EhuCwPbOKavVyu+//x47ggO2BRnu3btHnjx5En2slJInTx5Onz4dr/3UqVOxn0PSrltERB5PzziJiKRzw4cPx93dnV69enHz5s14n587d45PPvkE+Gtk5e8jKUFBQcyfP/8/1fD555/H/t4wDD7//HMcHR2pU6dOgvs+qqb9+/ezd+/eOP3c3NwAuHfvXoLHbNy4MQBTp06N0z558mQAXnzxxQSPkdIaN27MgQMH4lzngwcPmDVrFnnz5qVYsWIAsVP+EnPdIiLyeBpxEhFJ5/Lnz8+yZcto164dRYsWpWvXrhQvXpzIyEj27NnD119/Tffu3QGoX78+Tk5ONG3alFdffZWQkBBmz56Nj48P169f/1fnd3FxYcuWLXTr1o1KlSrxzTffsGnTJt566614C0E8SpMmTVi9ejUtWrTgxRdfJCAggJkzZ1KsWDFCQkJi+7m6ulKsWDG++uorChUqRKZMmShevPgjnwMqVaoU3bp1Y9asWdy7d48aNWpw4MABFi5cSPPmzalVq9a/utbkNGLECJYvX06jRo14/fXXyZQpEwsXLiQgIIBVq1ZhNtt+Npo/f34yZMjAzJkz8fT0xN3dnUqVKv3r579ERNIrjTiJiAjNmjXj119/pXXr1qxbt47+/fszYsQILly4wMcff8ynn34KQOHChVm5ciUmk4k33niDmTNn0rt37zgvyE0qi8XCli1buHHjBsOGDePgwYOMGTOG9957L1H7d+/enQ8++ICjR4/y+uuv8+2337JkyRLKly8fr++cOXPImTMngwcPpkOHDqxcufKxx50zZw7jxo3j4MGDDBo0iB9++IGRI0fy5Zdf/utrTU7ZsmVjz5491KtXj88++4yRI0fi5OTEhg0baNGiRWw/R0dHFi5ciMVioU+fPnTo0IEdO3bYsXIRkbTJZCT2yVUREREREZF0SiNOIiIiIiIiCVBwEhERERERSYCCk4iIiIiISAIUnERERERERBKg4CQiIiIiIpIABScREREREZEEpLsX4FqtVq5du4anpycmk8ne5YiIiIiIiJ0YhsH9+/fJkSNH7IvDHyfdBadr167h5+dn7zJERERERCSVuHz5Mrly5Xpin3QXnDw9PQHbF8fLy8vO1UBUVBTfffcd9evXx9HR0d7lSCqn+0WSSveMJJXuGUkq3TOSVKnpngkODsbPzy82IzxJugtOD6fneXl5pZrg5ObmhpeXl91vHEn9dL9IUumekaTSPSNJpXtGkio13jOJeYRHi0OIiIiIiIgkQMFJREREREQkAQpOIiIiIiIiCUh3zzglhmEYREdHExMTk+LnioqKwsHBgfDw8KdyPknb/uv9YrFYcHBw0FL8IiIiIkmk4PQPkZGRXL9+ndDQ0KdyPsMw8PX15fLly/pmVhKUHPeLm5sb2bNnx8nJKZmrExEREXl2KTj9jdVqJSAgAIvFQo4cOXByckrxMGO1WgkJCcHDwyPBl26J/Jf7xTAMIiMjuX37NgEBARQsWFD3nIiIiEgiKTj9TWRkJFarFT8/P9zc3J7KOa1WK5GRkbi4uOibWEnQf71fXF1dcXR05OLFi7HHEREREZGE6Tv1R1CAkWeZ7m8RERGRpNN3UCIiIiIiIglQcBIREREREUmAnnFKITFWgwMBgdy6H46PpwsV/TNhMWvVPBERERGRtEgjTilgy/HrPD/pBzrM3sfAL4/QYfY+np/0A1uOX0+xc3bv3h2TyYTJZMLR0RF/f3+GDx9OeHh4bJ+Hn//91/PPP59iNYmIiIiIPCs04pTMthy/Tt8lhzH+0X4jKJy+Sw4zo3NZGhbPniLnbtiwIfPnzycqKopDhw7RrVs3TCYTkyZNiu0zf/58GjZsGLutd/mIiIiIiCRMwSkBhmEQFhWTqL4xVoMx63+LF5oADMAEjF1/gmoFssRO27NarYRFxuAQGR1vtTNXR0uS3iPl7OyMr68vAH5+ftStW5etW7fGCU4ZMmSI7SMiIiIiIomj4JSAsKgYir3zbbIcywBuBIdTYux3iep/4t0GuDn9uz+i48ePs2fPHvLkyfOv9hcRERERkb8oOD1DNm7ciIeHB9HR0URERGA2m/n888/j9OnQoQMWiyV2e8mSJTRv3vwpVyoiIiIikrYoOCXA1dHCiXcbJKrvgYBAus8/mGC/BT0qUNE/E2Cbqnc/+D6eXp6PnKqXFLVq1WLGjBk8ePCAKVOm4ODgQKtWreL0mTJlCnXr1o3dzp49ZZ63EhERERF5lig4JcBkMiV6ulz1glnJ7u3CjaDwRz7nZAJ8vV2oXjBrnGecop0suDk5xAtOSeXu7k6BAgUAmDdvHqVKlWLu3Ln07Nkzto+vr29sHxERERERSRwtR56MLGYTY5oWA2wh6e8ebo9pWuypvM/JbDbz1ltvMWrUKMLCwlL8fCIiIiIizzIFp2TWsHh2ZnQui6+3S5x2X2+XFF2K/FHatGmDxWJh2rRpT+2cIiIiIiLPIk3VSwENi2enXjFfDgQEcut+OD6eLlT0z/RURpr+zsHBgQEDBvDhhx/St2/fp3puEREREZFniV1HnHbu3EnTpk3JkSMHJpOJtWvXJrjP9u3bKVu2LM7OzhQoUIAFCxakeJ3/hsVsokr+zLxUOidV8mdO8dC0YMGCR379RowYwa1bt3B3d8cwDK2gJyIiIiLyL9g1OD148IBSpUoleipZQEAAL774IrVq1eLIkSMMGjSIXr168e23yfOeJRERERERkUex61S9Ro0a0ahRo0T3nzlzJv7+/nz88ccAFC1alF27djFlyhQaNEjckuEiIiIiImI/MTEx9i7hX0lTzzjt3bs3zjuIABo0aMCgQYMeu09ERAQRERGx28HBwQBERUURFRUVp29UVBSGYWC1WrFarclX+BMYhhH736d1Tkm7kuN+sVqtGIZBVFRUnJchy7Pp4d9z//z7TuRxdM9IUumekaRYsGAB06dP580330wV90xSakhTwenGjRtky5YtTlu2bNkIDg4mLCwMV1fXePtMmDCBcePGxWv/7rvvcHNzi9Pm4OCAr68vISEhREZGJm/xCbh///5TPZ+kbf/lfomMjCQsLIydO3cSHR2djFVJarZ161Z7lyBpjO4ZSSrdM/IkUVFRzJ07ly1btgDw7bffPvJ796ctNDQ00X3TVHD6N0aOHMmQIUNit4ODg/Hz86N+/fp4eXnF6RseHs7ly5fx8PDAxcXln4dKEYZhcP/+fTw9PTGZnu6qe5L2JMf9Eh4ejqurKy+88MJTu8/FfqKioti6dSv16tXD0dHR3uVIGqB7RpJK94wk5Pr167Rv3569e/diMpkYNWoUZcqUSRX3zMPZaImRpoKTr68vN2/ejNN28+ZNvLy8HptYnZ2dcXZ2jtfu6OgY7w8qJiYGk8mE2WzGbH4662Y8nG718LwiT5Ic94vZbMZkMj3y/wF5dunPW5JK94wkle4ZeZwuXbqwd+9evL29Wbp0KfXr12fz5s2p4p5JyvnT1HfqVapUYdu2bXHatm7dSpUqVexUkYiIiIiIPMm0adOoVKkSBw8e5MUXX7R3Of+aXYNTSEgIR44c4ciRI4BtufEjR45w6dIlwDbNrmvXrrH9+/Tpw/nz5xk+fDinTp1i+vTprFixgsGDB9ujfBERERER+Yfw8HB+/PHH2O3ixYuzd+9eChYsaMeq/ju7Bqeff/6ZMmXKUKZMGQCGDBlCmTJleOeddwDbfMiHIQrA39+fTZs2sXXrVkqVKsXHH3/MnDlztBS5iIiIiEgqcPnyZV544QXq16/Prl27YtufhWf57fqMU82aNWOXV36UBQsWPHKfX375JQWrSibWGLi4B0Jugkc2yFMVzFr6WURERESeTTt27KBNmzbcvn2bTJkyPfVVqlNamnrGKc04sR6mFoeFTWBVT9t/pxa3taegGzdu8Nprr5EvXz6cnZ3x8/OjadOm8Z4Le5wFCxaQIUOGeO01a9bEZDLF/sqWLRtt2rTh4sWLyXwFIiIiIpLWGIbBp59+Sp06dbh9+zalSpXi559/pnbt2vYuLVkpOCW3E+thRVcIvha3Pfi6rT2FwtOFCxcoV64cP/zwAx999BHHjh1jy5Yt1KpVi/79+//n47/yyitcv36da9eusW7dOi5fvkznzp2ToXIRERERSavCwsLo1q0bAwcOJCYmhk6dOrFnzx78/f3tXVqyU3BKiGFA5IPE/QoPhm+GA4+afvhn25Y3bf3+vl9U6KOP94RpjP/Ur18/TCYTBw4coFWrVhQqVIjnnnuOIUOGsG/fPgAmT55MiRIlcHd3x8/Pj379+hESEgLA9u3b6dGjB0FBQbEjS2PHjo09vpubG76+vmTPnp3KlSszYMAADh8+/C+/qCIiIiLyLFi+fDmLFy/GYrEwZcoUFi9ejJubm73LShFp6j1OdhEVCh/kSKaDGbaRqIl+sS1mIMPjur91DZzcEzxqYGAgW7Zs4f3338fdPX7/h9PvzGYzn376Kf7+/pw/f55+/foxfPhwpk+fTtWqVZk6dSrvvPMOp0+fBsDDw+Ox51uxYgWVKlVKsDYREREReXb16NGDQ4cO0aZNG2rWrGnvclKURpyeAWfPnsUwDIoUKfLEfoMGDaJWrVrkzZuX2rVrM378eFasWAGAk5MT3t7emEwmfH198fX1jROcpk+fjoeHB+7u7mTOnJnTp08zb968FL0uEREREUldDMNgzpw5sbOWTCYT06ZNe+ZDE2jEKWGObraRn8S4uAeWtk64X6eVtlX2AKvVSvD9+3h5emI2/yPHOiZumPNJKxP+3ffff8+ECRM4deoUwcHBREdHEx4eTmhoaIJDqp06deLtt98G4ObNm3zwwQfUr1+fQ4cO4enpmajzi4iIiEja9eDBA15++WVWrFjBt99+y4oVK56JZcYTSyNOCTGZbNPlEvMrf23wygE87gYygVdOW7+/7+fo9ujjJfJGLFiwICaTiVOnTj22z4ULF2jSpAklS5Zk1apVHDp0iGnTpgEkaqlIb29vChQoQIECBahWrRpz587l999/56uvvkpUjSIiIiKSdp09e5bKlSuzYsUKHB0dn7kV8xJDwSk5mS3QcNKfG/8MPX9uN5yY7O9zypQpEw0aNGDatGk8ePAg3uf37t3j0KFDWK1WPv74YypXrkyhQoW4di3uSJqTkxMxMTGJOqfFYruGsLCw/34BIiIiIpJqffPNN1SoUIHjx4/j6+vLjz/+SN++fdPVaBMoOCW/Ys2g7SLwyh633SuHrb1YsxQ57bRp04iJiaFixYqsWrWK33//nZMnT/Lpp59SpUoVChQoQFRUFJ999hnnz59n8eLFzJw5M84x8ubNS0hICNu2bePOnTuEhobGfhYaGsqNGze4ceMGR48epW/fvri4uFC/fv0UuR4RERERsS+r1cr48eN58cUXuXfvHlWqVOHQoUNUq1bN3qXZhYJTSijWDAYdh24bodVc238HHUux0ASQL18+Dh8+TK1atRg6dCjFixenXr16bNu2jRkzZlCqVCkmT57MpEmTKF68OEuXLmXChAlxjlG1alX69OlDu3btyJo1Kx9++GHsZ7NnzyZ79uxkz56dWrVqcefOHTZv3kzhwoVT7JpERERExH4CAwOZNm0ahmHQp08ftm/fTo4cybXadNqjxSFSitkC/tWf6imzZ8/O559/zueff/7IzwcPHszgwYPjtHXp0iXO9owZM5gxY0actu3btydrnSIiIiKS+mXJkoVVq1Zx8uRJevbsae9y7E7BSUREREREAFi3bh1RUVG0bm1bKbpq1apUrVrVzlWlDgpOIiIiIiLPAmuM7fU4ITfBI5vt9TeJXJTMarUyduxY3nvvPdzc3ChRooQeyfgHBScRERERkbTuxHqMLW9iCv5r1WTDKwemhpMSfM7+7t27dO7cmc2bNwPwyiuvkC9fvhQtNy3S4hAiIiIiImnZifUYK7piBMd91YwRfA1jRVc4sf6xux4/fpwKFSqwefNmXFxcWLx4MVOnTsXR0TGlq05zFJxERERERNIqawxhG4ZhGEa8b+zNgGEYhG0YZpvG9w8rVqygUqVKnDt3jjx58rB79246d+78VMpOixScRERERETSqJgLu3ENu4H5Me+iNZvANewGMRd2x/vs4MGDhIaGUqdOHX7++WfKli2bwtWmbQpOIiIiIiJp1Lnz5/51vwkTJjBz5ky2bNlClixZkru0Z46Ck4iIiIhIGnXLyJDofr/88gsdO3YkMjISAAcHB1599VUcHLReXGIoOImIiIiIpFEOfuUJM5we+7nVgGtGZrYevU7VqlVZvnw548ePf4oVPjsUnCRBY8eOJVu2bJhMJtauXZti50np4ydk+/btmEwm7t27B8CCBQvIkCFD7Odjx46ldOnSdqktKf55HSIiIvKMCg2kzE+9cDVFYhhgGHE/thoQFWPQbKsv7w7rT3h4OI0aNWLw4MH2qTeNU3B6RnTv3h2TyYTJZMLJyYkCBQrw7rvvEh0d/Z+Oe/LkScaNG8cXX3zB9evXadSo0X+uNa0EkHbt2nHmzJmnci6FHREREUmSe5eJmlMf56v7CDbcmBzdmutkitPl+IMM+C/OxKG9ewF4++232bBhAxkzZrRHxWmeJjQ+Qxo2bMj8+fOJiIhg8+bN9O/fH0dHR0aOHJnkY8XExGAymTh3zvYg4UsvvYTJ9JjlWp5Rrq6uuLq6/qdjREZG4uT0+OFzERERkSS7cYyYxa1wfHCT60Ymhru8Q/2atWnzYzv8Qo7iwz1O3Qhhx+plRAbdwcPDg0WLFtGiRQt7V56macQpkR48ePDYX+Hh4YnuGxYWlqi+/4azszO+vr7kyZOHvn37UrduXdavt73wLCIigjfeeIOcOXPi7u5OpUqV2L59e+y+D6elrV+/nmLFiuHs7MzLL79M06ZNATCbzXGC05w5cyhatCguLi4UKVKE6dOnx6nlypUrdOjQgUyZMuHu7k758uXZv38/CxYsYNy4cRw9ejR2hGzBggXxrqV27doMGDAgTtvt27dxcnJi27Ztj/0abNiwgQoVKuDi4kKWLFni/AWxePFiypcvj6enJ76+vnTs2JFbt2499lj/nKr30BdffIGfnx9ubm60bduWoKCg2M+6d+9O8+bNef/998mRIweFCxdO8NwXLlygVq1aAGTMmBGTyUT37t0BsFqtTJgwAX9/f1xdXSlTpgzr1q2LU8/mzZspVKgQrq6u1KpViwsXLjz2mkRERCSNO/cj1rkNsTy4ySmrH6+7fcikvu3pUiUPO0fUY2DPl6nTth+9OnXClSgKFy7MgQMHFJqSgUacEsnDw+OxnzVu3JhNmzbFbvv4+BAaGvrIvjVq1IgTWPLly8edO3fi9TP+OUn1X3B1deWPP/4AYMCAAZw4cYIvv/ySHDlysGbNGho2bMixY8coWLAgAKGhoUyaNIk5c+aQOXNmsmfPTs2aNenRowfXr1+PPe7SpUt55513+PzzzylTpgy//PILr7zyCu7u7nTr1o2QkBBq1KhBzpw5Wb9+Pb6+vhw+fBir1Uq7du04fvw4W7Zs4fvvvwfA29s7Xu29evViwIABfPzxxzg7OwOwZMkScubMSe3atR95vZs2baJFixa8/fbbLFq0iMjISDZv3hz7eVRUFO+99x6FCxfm1q1bDBkyhO7du8fpk5CzZ8+yYsUKNmzYQHBwMD179qRfv34sXbo0ts+2bdvw8vJi69atiTq3n58fq1atolWrVpw+fRovL6/Yka4JEyawZMkSZs6cScGCBdm+fTuvvvoquXPnplatWly+fJmWLVvSv39/evfuzc8//8zQoUMTfT0iIiKShhz9CmNdP8zWaPbGFGNShlHMeqUOPl4ugO2dTVXyZ/6zc05yb9lC0aJFH/m9lvwLRjoTFBRkAEZQUFC8z8LCwowTJ04YYWFh8T4DHvurcePGcfq6ubk9tm+NGjXi9M2SJcsj+yVVt27djJdeeskwDMOwWq3G1q1bDWdnZ+ONN94wLl68aFgsFuPq1atx9qlTp44xcuRIwzAMY/78+QZgHDlyJE6fNWvWxKsnf/78xrJly+K0vffee0aVKlUMwzCML774wvD09DT++OOPR9Y6ZswYo1SpUvHaAWPNmjWGYdj+LDJmzGh89dVXsZ+XLFnSGDt27GO/BlWqVDE6der02M//6eDBgwZg3L9/3zAMw/jxxx8NwLh7965hGLavibe3d5y6LRaLceXKldi2b775xjCbzcb169cNw7D9OWTLls2IiIj4T+c2DMMIDw833NzcjD179sS2xcTEGF26dDHat29vGIZhjBw50ihWrFicY7/55pvxjvV3T7rP5dkTGRlprF271oiMjLR3KZJG6J6RpNI98xRYrYax82PDGONlGGO8jHWjGhhNp2wz7twPj+1y7do1o3r16sbWrVvtWGjipKZ75knZ4J804pRIISEhj/3MYrHE2X7S9C+zOe7syPPnzxMcHIyXl1e8z5Jq48aNeHh4EBUVhdVqpWPHjowdO5bt27cTExNDoUKF4vSPiIggc+bMsdtOTk6ULFnyied48OAB586do2fPnrzyyiux7dHR0bE/zThy5AhlypQhU6ZMjztMglxcXOjSpQvz5s2jbdu2HD58mOPHj8dOPXyUI0eOxKnpnw4dOsTYsWM5evQod+/exWq1AnDp0iWKFSuWqLpy585Nzpw5Y7erVKmC1Wrl9OnT+Pr6AlCiRIl4zzX9m3OfPXuW0NBQ6tWrF6c9MjKSMmXKALbFOypVqhTn8ypVqiTqWkRERCQNsMbAN8Ph4BwAvoh+kW+y9WFRz8pkcLN9v7Fnzx5at27N9evX6dOnD6dOndK7mVKAvqKJ5O7unmJ9Y2JicHd3/8/BqVatWsyYMQMnJydy5MgR+z9MSEgIFouFQ4cOxQt5f5+C6OrqmuACEA8D5OzZs+N9w/7w2P91QYWHevXqRenSpbly5Qrz58+ndu3a5MmT57H9n3TeBw8e0KBBAxo0aMDSpUvJmjUrly5dokGDBrEvgUsu//zz/7fnfvi13rRpU2xYs1qthISExAm8IiIi8oyKCoNVveDURqyGifeiO3MsV0cW9aiAl4sjhmEwc+ZMBg4cSFRUFM899xxr1qxRaEoh+qo+Q9zd3SlQoEC89jJlyhATE8OtW7eoXr36fzpHtmzZyJEjB+fPn6dTp06P7FOyZEnmzJlDYGDgI0ednJyciImJSfBcJUqUoHz58syePZtly5bx+eefP7F/yZIl2bZtGz169Ij32alTp/jjjz+YOHEifn5+APz8888J1vBPly5d4tq1a+TIkQOAffv2YTabYxeBeJTEnPvhCNXfvy4PF+m4dOkSNWrUAGzB6eEIJUDRokXjjcLt27cvydclIiIiqUxoICxrB1cOEGE4MiiqH/fyNmZht/K4OzsQHh5O//79mTdvHgCtW7dm/vz5T3wuX/4braqXDhQqVIhOnTrRtWtXVq9eTUBAAAcOHGDChAlxFrVIrHHjxjFhwgQ+/fRTzpw5w7Fjx5g/fz6TJ08GoEOHDvj6+tK8eXN2797N+fPnWbVqFXv/fIdA3rx5CQgI4MiRI9y5c4eIiIjHnqtXr15MnDgRwzASXA1mzJgxLF++nDFjxnDy5EmOHTvGpEmTANsUOycnJz777DPOnz/P+vXree+995J87S4uLnTr1o2jR4/y008/8frrr9O2bdvYaXqPkphz58mTB5PJxMaNG7l9+zYhISF4enryxhtvMHjwYBYuXMi5c+c4fPgws2bNYuHChQD06dOH33//nWHDhnH69GmWLVv2yFUKRUREJA25ewHm1oMrB7hnuNM5ciShBZowv0cF3J0duH//Pi+88ALz5s3DbDYzadIkVqxYodCUwhSc0on58+fTtWtXhg4dSuHChWnevDkHDx4kd+7cST5Wr169mDNnDvPnz6dEiRLUqFGDBQsW4O/vD9hGT7777jt8fHxo3LgxJUqUYOLEibFT+Vq1akXDhg2pVasWWbNmZfny5Y89V4cOHXBwcKBDhw64uLg8sa6aNWvy9ddfs379ekqXLk3t2rU5cOAAAFmzZmXBggV8/fXXFCtWjIkTJ/K///0vyddeoEABWrZsSePGjalfvz4lS5aMtxT7PyXm3Dlz5mTcuHGMGDGCbNmyxS7F/t577zF69GgmTJhA0aJFady4Md99913s1zp37tysWrWKtWvXUqpUKWbOnMkHH3yQ5OsSERGRVOLaEZhTD/44y1UjC60ix5KhaA1mdS2Hi6PteykPDw+KFStGpkyZ2LJlC8OHD09379u0B5NhJMO612lIcHAw3t7eBAUFxU53eig8PJyAgAD8/f0T/CY9ufx96tV/fcbpWXThwgXy58/PwYMHKVu2rL3LsbvkuF/scZ+L/URFRbF582YaN26Mo6OjvcuRNED3jCSV7plkdPZ7WNENIkM4Yc1D98jhVCxZjCntSuNgNhERERH7b3dYWBi3bt164vPfqVVqumeelA3+Sd+pS6oUFRXFjRs3GDVqFJUrV1ZoEhERkWfbL0ttzzRFhrDLWpy2kaOpXrYEn7QvQ1REOF26dKFVq1axK/O6urqmydCUlmlxCEmVdu/eTa1atShUqBArV660dzkiIiIiKcMwYOf/4MfxAKyOeZ43o3rTtlI+3nupOJcuXaRFixYcOXIEi8XC/v379eoRO1FwklSpZs2apLNZpCIiIpLexETD5qFwaAEA06Ob8WF0O16ulo/RTYqybds22rVrR2BgIFmzZmXFihUKTXak4CQiIiIi8rRFPoCVL8OZLRiYGB3VnSUx9ehXMz9v1C/E//73P0aMGIHVaqVChQqsWrUq9rUmYh8KTo+gkQ55lun+FhERsbMHd2BZW7h6iCiTM/0j+vGdtQJD6xXitToFGTJkCFOmTAHg5ZdfZtq0aVrQKRXQ4hB/83BVj9DQUDtXIpJyHt7f9l7FRkREJF0KPG97R9PVQ4RavGgXPpLvrBV4q3ERXqtTEIBOnTrh5eXFjBkzmDNnjkJTKqERp7+xWCxkyJCBW7duAeDm5pbia+JbrVYiIyMJDw/XcuSSoP9yvxiGQWhoKLdu3SJDhgyx79USERGRp+TqIVjaFkLvcNcpO63uD+W8kYN3X3qO+v5/haNy5cpx4cIFMmbMaMdi5Z8UnP7B19cXIDY8pTTDMAgLC8PV1VUvLpMEJcf9kiFDhtj7XERERJ6SM9/C190hKpQrLoVocW8Qd0wZmNDiOc5/v5R877/Pzp07KV++PIBCUyqk4PQPJpOJ7Nmz4+PjQ1RUVIqfLyoqip07d/LCCy9o6pQk6L/eL46OjhppEhERedoOLYSNg8GI4aRbBVoH9iHc7Mb7jfKxYuJg1q5dC8CGDRtig5OkPgpOj2GxWJ7KN5gWi4Xo6GhcXFwUnCRBul9ERETSEMOA7RNhx0QA9ng2oOvtTmB25M3KHrzXuwWnTp3CycmJGTNm8PLLL9u5YHkSBScRERERkeQWEwUbB8EvSwBY592JgTcb42Sx0Dn7Ld7o3J779++TK1cuVq1aRcWKFe1bryRIwUlEREREJDlFhNieZzq7FcNkZq73a4y/UQkXRzN9C4UzqGt3AF544QW+/vprfHx87FquJI6Ck4iIiIhIcgm5BUvbwPUjGA6uTPIYzswbhXF3sjC3ewUq5s3Itq+bki9fPj766CNNvU9DFJxERERERJLDnbOwpCXcu4jVNTPDnUex8kY2nEKuM/v1xlTOlxmAVatWKTClQXpxkIiIiIjIf3X5oO3FtvcuEpMhL684TmDljWyYAvZyZf4gvnh/JIZhAHoJfVqlEScRERERkf/i1GZY+TJEhxGVrTQdHwzhwC0TEXsXcXPXCgCuXr1KWFgYbm5udi5W/i2NOImIiIiI/FsH58BXnSA6jHD/ujQLGcm+y6HcWz0uNjQNGzaMLVu2KDSlcRpxEhERERFJKsOAH96Dnz4GIOS5Trx4rgW/nz5J4LoJRNy9gZubG/Pnz6dt27Z2LlaSg4KTiIiIiEhSREfChtfh6HIAAiu+wYtHqnDtbgiBa98n4t4t8ufPz5o1ayhRooSdi5Xkoql6IiIiIiKJFXEflrW1hSaThRu1Pqb+4SpcD46gYPaMLF20kKZNm3Lw4EGFpmeMRpxERERERBLj/g1Y2hpuHANHdy7UmU7TNZHcvnKIMhWrsrhnRbJ4ONOqaUN7VyopQMFJRERERCQht0/DktYQdAncs3KqzlyazDjKhS/fxRQdzrvD9pLFw9neVUoKUnASEREREXmSS/tgWTsIvweZ8nO05jwaj1rItc2fQ0w0BQsVwtNF72Z61ik4iYiIiIg8zon1sKoXxERArgrsKjeVpj3e5N7hzQA0adqMpUsW4+XlZedCJaVpcQgRERERkUfZ/wWs6GoLTYVfZFW+D6jXrK0tNJlMvDNmLOvWrlFoSic04iQiIiIi8ndWK3w/BvZ8atsu35MteYbySp8hhF85iZObB18tX0rzZs3sW6c8VQpOIiIiIiIPRUfAuv5w7Gvbdp13WOfRjiHLj+JZtQO+ThGsnDGJYkUK27dOeeoUnEREREREAMKD4MtOcOEnMDsQ3nAKryz7nZ+cD4PJQusK/nw4aQ0Ws8nelYodKDiJiIiIiARfsy03fus3cPLgcvWPqdl9AudP/opXhRb0HzmO8S8Vx6zQlG4pOImIiIhI+nbrJCxpBcFXwSMbO/KNoGmz/ty/F4jZxZOXmjTi/ebFMZkUmtIzBScRERERSb8u7IIvO0J4EEbmgnwa8iJD2vTGao3B0ScfQyZ9wYRudRSaRMFJRERERNKp46thzasQE0moTwV678zI0q/eBcC9WE3GfvgJQxuXUGgSQO9xEhEREZH0aO80WNkDYiKhaFMuVPuQr9dtApOZjHVe4ePps3njxZIKTRJLI04iIiIikn5YrfDdKNg3zbZd8VWs9T9g2caTeDcagtnFnUkDOtCtal67limpj4KTiIiIiKQPUeGwtg/8tgbDMPjfnVpUdG/CxjW/8fWhK7gXrsLEliVoVyG3vSuVVEjBSURERESefWF3be9ouribkCgHXv65EF9/tx63pbvJ1H0aTm6efNymFM3L5LR3pZJKKTiJiIiIyLMt6IptufHbpzgb4kbz9S789vs+zBYHXCq1w8nVg887lKFRiez2rlRSMQUnEREREXk2WGPg4h4IuQke2SBPVds7mpa2hvvX2XTZi04rgwgKvoGrd2a8mozAK89zzOhcljpFs9m7eknlFJxEREREJO07sR62vAnB1/5qc8uMERmKERXKmAMevP/tVQzDIEv+Ejg3HIZHxizM7lqe6gWz2q9uSTMUnEREREQkbTuxHlZ0BYw4zUboH5iA09acfBaQAcO4jnfZxrjVfgUPVxfmdq9A5XyZ7VKypD0KTiIiIiKSdlljbCNN/whNACbAMMDLHE6GxoNxLHwY96LVAehXq4BCkySJXoArIiIiImnXxT1xp+f9ae2pKF5eFwYY5DAFUtnlYmxoAliy7yIx1vhhS+RxFJxEREREJO0KuRlnM8ZqMPqHcFp8Fcb8I1Es+TUKAB/uxel3PSicAwGBT6tKeQZoqp6IiIiIpF0ef62GdzfMoNPqML45Gw3AoEpOtC/uCMAtMsTb9db98KdSojwbFJxEREREJO3KUxWcvTl+OZDmX4Zy7q6BiwPMbupC55JOWA24ZmTmgLVIvF19PF3sULCkVQpOIiIiIpJ2HVnKuqN36LQ6jAdRkMfbxJp2bpTJbuHhI0zjorpg/dsTKibA19uFiv6Z7FOzpEkKTiIiIiKSNh1bCetfJ6u7iUirmboFXVje3EwWN1tIukFm3o3qwrfWirG7mP7875imxbCYTY84qMijKTiJiIiISJpj/W095tW9AYOqLV5lR8/WXHf0ZeDa1WSOvEsuv7wUqdyAo1t+h6C/nmXy9XZhTNNiNCye3X7FS5qk4CQiIiIiacov62bSpfcAlrd0oUT9ztD4f/y+/xJjVv+GYRSleekcDG5TCkeLmRdL+XEgIJBb98Px8bRNz9NIk/wbCk4iIiIikmYsmfIOrwx/j/BoeGOPF1s+/4zPfjzH5K1nAOheNS/vNCmG+c9wZDGbqJJfL7qV/07BSURERERSvaioKIb17cYnc5cD0KiUL4u/O8y7m88wf/cFAAbWKcigugUxmTSiJMlPwUlEREREUrVbt27RtvmL7Nj7MwCjmhZk1JeHGLnxHKsPXwVsiz30qOZvzzLlGafgJCIiIiKp1oULF6herQpXrt3A0wkWvVKKhpN20P/r03x/8iYWs4mPWpekZdlc9i5VnnEKTiIiIiKSauVyi6KIezDumc2sfa0MOQd+S/elJ9h3PhAnBzPTO5albrFs9i5T0gEFJxERERFJVSIjIwFwCruNw7KWfNXCgkPWMkR3XU/HxSc5djUID2cH5nQrT+V8WvhBng4FJxERERFJNa5du0abNm0oWbQgM0odg3uXyJQzP9dbraTT4jOcv/2AzO5OLHy5IsVzetu7XElHFJxEREREJFXYvXs3rVu35saNG/z2y37e9nIll18eLrz4JR0WneN6UDg5vF1Y3KsS+bN62LtcSWfM9i5ARERERNI3wzCYMWMGtWrV4saNGxTP4c7PvVzJlSMHp+svpeWyS1wPCid/VndW9q2q0CR2oREnEREREbGb8PBw+vXrx/z58wFoUz4b8+qF4uGdmSO1FtJ5xQ1CIqIpkdObBT0qkNnD2c4VS3ql4CQiIiIidmEYBs2aNWPr1q2YzWYmtinCG4UvY3LJwL7n59F1zV0io61UzpeJ2V3L4+niaO+SJR3TVD0RERERsQuTycTAgQPJkiULW958gWFFrmBy8uDH8tPptCmMyGgrdYtmY0GPigpNYncKTiIiIiLy1BiGwcWLF2O3X2zUkPMfN6Ce02FwcGFzian02GYixmrQqmwuZnYui4ujxY4Vi9goOImIiIjIUxEaGkrnzp0pW7YsAQEBYLXC+tfxPL8Bw+zImoIT6LfHHYCez/vzUeuSOFj07aqkDnrGSURERERSXEBAAC1btuTIkSNYLBb27tmD/8npcGQJhsnMV3nGMOKXbAC8Ub8Q/WsVwGQy2blqkb8owouIiIhIitq6dSvly5fnyJEjZM2ale+3bqWjzxk4MAsDE0t8RzDiZD5MJniveXEG1C6o0CSpjoKTiIiIiKQIwzD48MMPadiwIYGBgZQvX55Dhw5R0/wz7J4KwKLMAxkdUBwHs4mp7UrTpXIe+xYt8hgKTiIiIiKSIr744gvefPNNrFYrPXr04KeffsLvynr4cTwACzx7M+ZqRVwczczuWp6XSue0c8Uij6fgJCIiIiIponv37lSpUoXp06czd+5cXI4vh2/fAmCxSyfG3q6Jp4sDi3tWolYRHztXK/JkWhxCRERERJLNvn37qFixImazGRcXF3766ScsFgsc/Qo2DgZgmWNLRt9rTBYPZxa9XJFiObzsXLVIwuw+4jRt2jTy5s2Li4sLlSpV4sCBA0/sP3XqVAoXLoyrqyt+fn4MHjyY8PDwp1StiIiIiDyK1Wrlvffeo2rVqowdOza23WKxwIl1sLYPYPC1uRFv3W9FroxurOxTRaFJ0gy7jjh99dVXDBkyhJkzZ1KpUiWmTp1KgwYNOH36ND4+8Ydrly1bxogRI5g3bx5Vq1blzJkzdO/eHZPJxOTJk+1wBSIiIiLy4MEDWrduzcaNGwG4c+cOhmHYVsY78x2s7AmGlbXUYnhoJwr6eLK4ZyV8vV3sXLlI4tl1xGny5Mm88sor9OjRg2LFijFz5kzc3NyYN2/eI/vv2bOHatWq0bFjR/LmzUv9+vXp0KFDgqNUIiIiIpIyTp48yfDhw9m4cSNOTk7MnTuX6dOn20JTwE5Y0QWsUWw2qjAkvCcl/TKx4tUqCk2S5thtxCkyMpJDhw4xcuTI2Daz2UzdunXZu3fvI/epWrUqS5Ys4cCBA1SsWJHz58+zefNmunTp8tjzREREEBEREbsdHBwMQFRUFFFRUcl0Nf/ewxpSQy2S+ul+kaTSPSNJpXtGkmLt2rW8/PLLhISEkDNnTlasWEGFChWIiorCdOUglmXtMUWHs81altcj+1I5XxamdyyNu5NJ91g6lpr+nklKDSbDMIwUrOWxrl27Rs6cOdmzZw9VqlSJbR8+fDg7duxg//79j9zv008/5Y033sAwDKKjo+nTpw8zZsx47HnGjh3LuHHj4rUvW7YMNze3/34hIiIiIunQ3bt3efXVV4mMjOS5555j2LBhZMiQAQDv0AtUOzsRx5hQfoopTq+oNyicyYFuBa042P0Je5G/hIaG0rFjR4KCgvDyevLzdmlqVb3t27fzwQcfMH36dCpVqsTZs2cZOHAg7733HqNHj37kPiNHjmTIkCGx28HBwfj5+VG/fv0EvzhPQ1RUFFu3bqVevXo4OjrauxxJ5XS/SFLpnpGk0j0jSRETE8Phw4epXbs2jRo1st0zt0/jsGQwpphQDlgL0ztqCM3K+fNu06I4WJSaJHX9PfNwNlpi2C04ZcmSBYvFws2bN+O037x5E19f30fuM3r0aLp06UKvXr0AKFGiBA8ePKB37968/fbbmM3x/2d0dnbG2dk5Xrujo6Pd/6D+LrXVI6mb7hdJKt0zklS6Z+RRjh07RlRUFGXLlgWgZ8+edO3alc2bN9vumeBLGMtaYQr9g6PWfLwcOYyuLxRjRKMitmeeRP4mNfw9k5Tz2y32Ozk5Ua5cObZt2xbbZrVa2bZtW5ype38XGhoaLxxZLBYA7DTjUERERCRdWLFiBZUrV6Z58+bcunUrfoegKxiLmmEKucFJqx9dI0fQv2FZRjYuqtAkzwS7TtUbMmQI3bp1o3z58lSsWJGpU6fy4MEDevToAUDXrl3JmTMnEyZMAKBp06ZMnjyZMmXKxE7VGz16NE2bNo0NUCIiIiKSfKKjo3nrrbf46KOPANtiXf/8vss56h6WpS0wBV3hnDU7XaPe4s0WVehYKbc9ShZJEXYNTu3ateP27du888473Lhxg9KlS7NlyxayZcsGwKVLl+KMMI0aNQqTycSoUaO4evUqWbNmpWnTprz//vv2ugQRERGRZ9Yff/xB+/bt+f777wF48803ef/99+MGp9A/qPL7JMwRV7lszUr3mLcZ06EmTUrmsFPVIinD7otDDBgwgAEDBjzys+3bt8fZdnBwYMyYMYwZM+YpVCYiIiKSfv3yyy+0bNmSCxcu4O7uzrx582jbtm3cTuFBmJa2xjviKjeMjLxsjGZ8t4bUKJTVPkWLpCC7BycRERERSX0mTpzIhQsXKFCgAGvWrKF48eJxO0SEELmoFU63jnHH8KK36R0m9mpGuTwZ7VOwSApTcBIRERGReGbNmkXmzJn54IMPYt/PFCsqnPDF7XC5dpAgw40+1hF80Ls5xf0UmuTZpcX0RURERISbN2/y0Ucfxa5U7O3tzfTp0+OHpuhIQhZ3xOXKLkIMF4a7jKFR8VwU9vV8+kWLPEUKTiIiIiLp3IEDByhXrhzDhw9n+vTpj+8YE03gku54XNpGmOHEOM93eOfVzmRxeXq1itiLgpOIiIhIOjZ37lyqV6/O1atXKVKkCHXq1Hl0R6uVm0t6kenCJiINCx9lGM2ofr3x8XR+ugWL2ImCk4iIiEg6FBkZSd++fenVqxeRkZE0b96c/fv3U6RIkfidDYOLi/uSLWAN0YaZ6VlH80b/fni7OT79wkXsRMFJREREJJ25du0atWrVYubMmZhMJsaPH8+qVavw8vKK39kwOLloIHkCvsRqmFiSfST9+gzEzUlrjEn6ojteREREJJ05ffo0+/btw9vbm2XLltG4cePH9j24YDgVLi4EYK3fcLq8PAyL2fS0ShVJNRScRERERNKZWrVqsWDBAqpUqUKBAgUe2ccwDH6cN4ral2cBsC3vYFp0G4nJpNAk6ZOm6omIiIg848LDw3nttdc4ffp0bFuXLl0eG5pirAZrZ71L7cufA3Aw3wDqdB+r0CTpmkacRERERJ5hly9fplWrVhw8eJDt27dz5MgRLBbLY/tHRMfw5axJdLs1GYAT+V+hQpf3n1a5IqmWRpxEREREnlE7duygXLlyHDx4kEyZMjF58uQnhqYHEdHMmv4xnW9+CMCFAl0o1vmjp1WuSKqm4CQiIiLyjDEMg08++YQ6depw+/ZtSpcuzaFDh6hXr95j97n7IJIp0z6lzx8TsZgMbhRoS95On4Gm54kACk4iIiIiz5SwsDC6dOnCoEGDiImJoXPnzuzevZu8efM+dp8bQeFMmDaDYUEf4GiKITD/S/h2nKnQJPI3esZJRERE5BlisVi4cOECFouFyZMn89prrz1xUYcLdx4wadYCPo54H2dTFCH+DcjUcS6YHz+lTyQ9UnASEREReYY4OTmxcuVKzpw5wwsvvPDEvieuBTNh7jKmRb+HmymCsNw18ei0GCyOT6lakbRDwUlEREQkDTMMg48++og7d+7w4Ye2RR18fX3x9fV94n4/Xwjk/QWrmGe8h5cpjMhcVXDtvBwcnJ9G2SJpjoKTiIiISBoVEhLCyy+/zNdffw1Ay5YtqVy5coL7/Xj6FpOWbGCx+T0ymkKIzl4Wpy5fg5NbSpcskmYpOImIiIikQb///jstWrTgt99+w9HRkU8//ZRKlSoluN+6I1eZvOJ7ljuOJ6spCKtPcRy6rgZnz6dQtUjapeAkIiIiksZs2rSJTp06ERQUhK+vL6tWraJq1aoJ7rd430U+X/cTKxzHk8MUiJG5EOaua8E1Y8oXLZLGaTlyERERkTTko48+omnTpgQFBVG1alUOHz6cYGgyDIPPf/idKWv3sMTxA/KYb2FkyIOp2zrwyPqUKhdJ2xScRERERNIQf39/DMOgb9++/Pjjj2TPnv2J/a1Wg/GbTjLru8MsdppAQfNVDK+cmLqtB68cT6lqkbRPU/VEREREUrno6GgcHGzftrVu3ZqDBw9Svnz5hPeLsfLmqmNsOfw7i50+5DnzRXDPiqnrOsiYN4WrFnm2aMRJREREJBVbs2YNxYoV4+rVq7FtiQlN4VEx9F16mI2HzzPX6WPKms+CSwboshayFEy5gkWeUQpOIiIiIqlQTEwMo0aNomXLlvz+++989NFHid73fngUPeYfZMeJK8x2mkJl8wlw8oQuq8G3eApWLfLs0lQ9ERERkVTm7t27dOrUiW+++QaAQYMGxb7cNiF/hETQff5BTlwNZKbzNF4wHQUHV+i0AnKWS8myRZ5pCk4iIiIiqcixY8do0aIF586dw9XVldmzZ9OpU6dE7XvtXhhd5u4n4PZ9PnedRT3jAFicoMMyyJPwcuUi8ngKTiIiIiKpxK5du2jQoAGhoaHkzZuXNWvWULp06UTte+52CF3m7OdaUBifuC2ksfUnMDtAm4WQv3bKFi6SDig4iYiIiKQSZcqUIV++fPj6+vLll1+SOXPmRO13/GoQ3eYd4I8HEfzP80teitoKmKDlLCjSOGWLFkknFJxERERE7OjevXt4e3tjMplwd3fn+++/J3PmzLHLj/9TjNXgQEAgt+6H4+PpgmEY9F58iJCIaCZmXE/rsA22ji99DsVbPcUrEXm2KTiJiIiI2Mnhw4dp2bIlffr0YcSIEQBky5btsf23HL/OuA0nuB4UHu+zD7Jupf39r2wbjT6CMp1TpGaR9ErLkYuIiIjYweLFi6lWrRoXL15k/vz5hIfHD0N/t+X4dfouOfzI0NTN8i0d78+3bdQdC5V6p0DFIumbgpOIiIjIUxQVFcXAgQPp2rUr4eHhNG7cmH379uHi4vLYfWKsBuM2nMB4xGdtLNsZ57gQAGv1YfD84JQpXCSd01Q9ERERkafk5s2btG3blp07dwIwevRoxo4di9n85J9lHwgI5GZQKJXNp/DhHrfIwAFrEZqY9zHJYTYAc6Ib8VyePlRJ8asQSZ8UnERERESegoiICKpWrcr58+fx9PRk0aJFNG/ePMH9omOsXNi1nF3OH5PDFBjbHmh44M0DzCaDpdF1GB/dmU9CIlLwCkTSNwUnERERkafA2dmZYcOGMXXqVNauXUuRIkWe2P9BRDRfHbzM2R3LGB/5YbzPMxKCyQT7YoowKroHYMLH8/HT/UTkv1FwEhEREUkhERERXL9+nbx58wLw6quv0q1bN1xdXR+7z63gcBbsucCSfRcJCY9kl/McMMV/MN1kAsOA3OZbmAEfbxcq+mdKsWsRSe8UnERERERSwLVr12jdujU3b97k559/JmPGjJhMpseGprO37jNr53nW/nKNyBgrAM0zXCBHeOAj+4MtPOUgkArmU3Rv2hmL2ZQi1yIiCk4iIiIiyW737t20bt2aGzdukCFDBk6fPk3lypXj9TMMg/0BgczeeZ5tp27FtpfLk5HeL+SjXkworE74fCOfz0Cp4tmT8xJE5B8UnERERESSiWEYzJgxg4EDBxIdHU3x4sVZs2YNBQoUiNMvxmqw5fgNZu08x9ErQYBt9Kh+sWz0fiEf5fL8OeUuwDdR5y1V9MnPS4nIf6fgJCIiIpIMwsPD6devH/Pn215E27ZtW+bOnYuHh0dsn9DIaFYeusKcnwK4FBgKgLODmdblctHzeX/yZf2rL9YYuPpzAmc1gVcOyFM1uS9HRP5BwUlEREQkGYwcOZL58+djNpuZOHEib7zxBiaT7ZmjOyERLNpzgUX7LnIvNAqAjG6OdKmSl65V8pDFwznuwe5egDV94dKevzWaIM4rcP98nqnhRDBbUuqyRORPCk4iIiIiyWDUqFH89NNPTJw4kbp16wJw/nYIc3YFsOrQFSKibQs+5M7kRq/q/rQp54er0z8Cj2HAL0tgywiIDAFHd2j4AbhmtLUFX/urr1cOW2gq1uxpXaJIuqbgJCIiIvIvGIbBDz/8QJ06dQDInDkzBw8exGQycehiIF/sOM/Wkzcx/hwkKuWXgVdfyEeD53wfvfpdyC3YMBBOb7Zt+1WGFjMhk79tu0gTuLgHQm6CRzbb9DyNNIk8NQpOIiIiIkkUGhrKK6+8wrJly5g9eza9evUixmqw9cRNZu08x+FL92L71i3qwyvV81HRP1Ps1L14Tm60habQO2B2hNpvQ9XX4wYjswX8q6fshYnIYyk4iYiIiCRBQEAALVu25MiRI1gsFkJCw1i6/yJzfgog4M4DAJwsZlqUyckrL/hTwMfz8QcLD7ZNwTuy1Lbt8xy0/AJ8SzyFKxGRpFBwEhEREUmkrVu30r59ewIDA8mSJSvtR0xhcZAPf6w5DoCXiwNdquShW9W8+Hi6PPlgAT/B2n4QdAkwQbXXodbb4OD85P1ExC4UnEREREQSYBgGH330ESNHjsRqtZKjYHFcGg5jw+0MQCQ5M7jS83l/2lXww905gW+vosLhh/dg7zTAgAx5bM8yaUlxkVRNwUlEREQkAYcOHWLEiBEYhoFHyfo41OtDjIMTxXN60fuF/DQu7ouDxZzwga4fhdWvwu2Ttu0yXaDhBHB+wnQ+EUkVFJxEREREHsNqNfjh1C1mHYrC+4WumF088CjVkFpFfOhdPR9V8md+/IIPfxcTDbunwvaJYI0C96zQ7DMo3CjFr0FEkoeCk4iIiMg/RETHMGbaEn665czVGNtoUJZqbWlWKie9X8hHYd8kjBD9cQ7W9IErB2zbRZpA00/APUsKVC4iKUXBSURERORPQaFRLNobwMQJH3Bt2yIcffwp2HMynasXokdVf3y9E1jw4e8MA36eB9+NgqhQcPKExh9CqQ6QmFEqEUlVFJxEREQk3btyN5S5uwJYtusUl9f8j7Df9wFQrWpVVo6oS2Yv96Qd8P4NWDcAzm61beetDs2nQ4bcyVy5iDwtCk4iIiKSbh2/GsQXO8+z+dh1wm9d4taa94kOvIKjkzPTp0+jV8+eST/ob2tg42AIuwsWZ6g7Bir1BXMiFo8QkVRLwUlERETSFcMw2H7mNrN3nmfPuT8ACD2zh7ubpxIdEUquXLlYvXo1FSpUSNqBw+7C5uFwbIVt27cktJwFPkWT+QpExB4UnERERCRdiIy2sv7oNWbvPM/pm/cBsJhNvFjchx+/3cLtiFBq1KjBihUr8PHxSdrBz/1oe5nt/WtgMsPzQ6DGm+DglAJXIiL2oOAkIiIiz7Tg8CiW77/E/N0XuBEcDoC7k4UOFXPT43l/cmZw5XL1tXzxxReMGTMGR0fHxB88MhS+HwsHvrBtZ8oHLWaBXxJHq0Qk1VNwEhERkWfStXthzN8dwPIDlwmJiAbAx9OZHtX8KeUexKF928jZ5DUA/Pz8GD9+fNJOcPWQ7WW2f/xu2y7fE+q/B05JXEhCRNIEBScRERF5ppy8HszsnedZf/Qa0VYDgELZPHilej6alc7ButWrqNejB6GhoeTLl48XX3wxaSeIiYKd/4OdH4ERAx6+8NI0KFg3Ba5GRFILBScRERFJ8wzDYPfZP/hi5zl++v1ObHvlfJl49YX81CyclZiYGEaOHMH//vc/AOrWrUvlypWTdqLbZ2BNb7j2i237uRbw4mRwy5RclyIiqZSCk4iIiKRZUTFWNv16nVk7z3PiejAAZhM0LpGd3i/ko2SuDADcuXOH9u3bs23bNgDefPNN3n//fSwWS+JOZLXCwdmw9R2IDgcXb1tgKtE6JS5LRFIhBScRERFJdWKsBgcCArl1PxwfTxcq+mfCYjbFfh4SEc2XB2wLPly9FwaAq6OFdhX86Pm8P36Z3GL7Hj58mJYtW3Lx4kXc3d2ZP38+bdq0SXwxQVdgXX84v922na+WbWqed87kuFQRSSMUnERERCRV2XL8OuM2nOB6UHhsW3ZvF8Y0LUaZ3BmZv/sCS/df5H64bcGHLB5OdK+al86V85DBLf7y37/99hsXL16kQIECrFmzhuLFiyeuEMOAY1/DpjcgIggcXG2LP5TvqZfZiqRDCk4iIiKSamw5fp2+Sw5j/KP9elA4fZYcxmKGGKutLV9Wd16pno8WZXLi4vj4KXddunQhIiKC1q1bkyFDhsQVEhoIGwfDibW27ZzloMUXkKVgUi9JRJ4RCk4iIiKSKsRYDcZtOBEvNMXtA+XzZODVGgWoU8QH89+m7z108+ZNBg0axCeffBL7IttevXolvpDft9qm5oXcBJPF9iLb6kPBom+bRNIz/Q0gIiIiqcKBgMA40/MeZ2j9IlTJn/mRn+3fv59WrVpx9epVHjx4wPr16xNfQEQIfDcKDs23bWcpZBtlylk28ccQkWeWgpOIiIikCrfuJxyantRvzpw59O/fn8jISIoUKcKHH36Y+JNfPgCre8PdANt2pb5Qdww4uib+GCLyTFNwEhERkVTBx9PlX/WLiIjg9ddfZ9asWQA0b96chQsX4uXllfDBoiNhx0TYNQUMK3jlhObTIV/NpJYvIs84BScRERFJFbJ6OmM2gfUxDzmZAF9v29LkD924cYMWLVqwb98+TCYT7733HiNHjsScmFXvbp6wvcz2xjHbdsl20OhDcM3wn69FRJ49Ck4iIiJid2dv3afj7P1PDE0AY5oWi/M+J1dXVwIDA8mQIQPLli2jUaNGCZ/MaoV902DbexATAa6ZoMkUeK75f74OEXl2KTiJiIiIXZ2+cZ9Oc/ZxJySSwtk8efl5f6Z+fybOQhG+f77HqWHx7BiGLV2ZTCa8vb1Zt24dDg4OFChQIOGT3b0Ia/vBxV227YL1odln4OmbEpcmIs8QBScRERGxm9+uBdF5zn7uhkZRLLsXS3pVIpO7E63L5eJAQCC37ofj42mbnmcxmwgPD6dv376ULVuW1157DYAiRYokfCLDgCPL4Js3IfI+OLpDg/ehXHcwxV/SXETknxScRERExC5+vXKPLnMPEBQWRalc3ix6uRLebo4AWMymeEuOX7p0iVatWvHzzz+zfPlyWrduTfbs2RM+Ucht2DgITm20bftVghYzIVO+ZL4iEXmWKTiJiIjIU3fo4l26zzvA/YhoyuXJyPweFfBycXxs/x9//JG2bdty584dMmXKxFdffZW40HRqM2x4HR7cBrMj1HoLqg0EsyUZr0ZE0gMFJxEREXmq9p//g5cXHORBZAwV/TMxr3sFPJwf/S2JYRhMnTqVYcOGERMTQ+nSpVmzZg158+Z98knCg+HbkfDLEtu2TzHby2yzl0zeixGRdEPBSURERJ6a3Wfv0Gvhz4RFxVCtQGZmdy2Pm9PjQ1OPHj1YuHAhAJ07d+aLL77Azc3tySe5sBvW9oF7lwATVB0AtUaBY+LeEyUi8igKTiIiIvJU7Dhzm96LfiYi2kqNQln5oks5XBwfP2XOZDJRokQJLBYLU6ZMYcCAAZietJBDVDj8OB72fA4YkCE3NJ8Jeasl/8WISLqj4CQiIiIpbtvJm/RdcpjIGCt1i/owrVNZnB0eHZoiIiJwdnYGYMiQITRs2JDnnnvuySe4cQxW94ZbJ2zbZTpDgwng4pWclyEi6VgiXqstIiIi8u9tOX6DPksOERljpVFxX6Z3KvfI0GQYBpMmTaJ8+fLcv38fsI06PTE0WWPgp8kwq5YtNLllgfbL4KVpCk0ikqw04iQiIiIpZsPRawz66ggxVoOmpXIwpW0pHCzxf24bEhJCjx49WLlyJQDLli3j1VdftX1ojYGLeyDkJnhkgzxVbaviBZ6HNX3g8n5bvyJNoMlU8Mj6lK5ORNITBScRERFJEasPX+GNr49iNaBl2Zx81LoUFnP8Z5R+//13WrRowW+//YajoyOfffYZvXv3tn14Yj1seROCr/21g1cOKNgAfl0BUQ/AyRMaTYLSHfUyWxFJMQpOIiIikuxWHLzMm6t/xTCgfQU/PmhRAvMjQtOmTZvo1KkTQUFBZM+enVWrVlGlShXbhyfWw4qugBF3p+BrcGi+7fd5nofm0yFjnpS9IBFJ9xScREREJFkt3neR0WuPA9Clch7GNXvukaFp8eLFdOvWDcMwqFq1KitXrvzrpbbWGNtI0z9D0985e0OXteDw+BfniogkFy0OISIiIslm3q6A2ND0cjV/3n3p0aEJoF69emTPnp1+/frx448//hWawPZM09+n5z1KRBBc3pdcpYuIPJFGnERERCRZfLHjHBO+OQVAnxr5ebNh4XjvXbp9+zZZs9oWb/D19eXo0aNkyZIl/sFCbibupIntJyLyH2nESURERP6zz7b9HhuaXq9T8JGhac2aNeTPn59ly5bFtj0yNF0+AHunJe7EHtn+dc0iIkmh4CQiIiL/mmEYTP7uNB9vPQPAG/ULMaReoTihKSYmhlGjRtGyZUvu37/P0qVLMYxHPLt0cS8segnm1oNrhxM4swm8ctqWJhcReQo0VU9ERET+FcMwmLjlFF/sOA/AW42L0PuF/HH63L17l06dOvHNN98AMHjwYD788MO4o1EXdsGOSRCw07ZtdoBSHSB7adj8xsOz/e2of+7bcKLtfU4iIk+BgpOIiIgkmWEYvLvxBPN3XwBgTNNi9KjmH6fPsWPHaNGiBefOncPV1ZU5c+bQsWPHhweACz/B9klwcZetzewIZTrB80P+Wl7cw+fR73FqOBGKNUvhqxQR+YuCk4iIiCSJ1WrwzvrjLNl3CYDxzYvTuXLc9yhdv36dKlWq8ODBA/LmzcuaNWsoXbq0LTCd324bYbq019bZ4gRlusDzgyGDX9yTFWsGRV60rbIXctP2TFOeqhppEpGnTsFJREREEs1qNRi5+hhf/XwZkwkmtSxJ2wp+8fplz56d1157jUOHDrF8+XIyZ8oEZ7+HHR/C5f22ThZnKNcNqg0C75yPP6nZAv7VU+aCREQSScFJREREEiXGajBs5VFWH76K2QQfty1FizK5Yj+/c+cOUVFRse9jGj9+PBgGlvM/wOpJcPVnW0cHFyjXHaoNtE27ExFJAxScREREJEFRMVaGrDjKhqPXsJhNTG1Xmqal/go9hw8fpmXLlvj6+rJjxw6cnZywnP3ONiXv2i+2Tg6uUP5lqPY6ePra6UpERP6dJAenLVu24OHhwfPPPw/AtGnTmD17NsWKFWPatGlkzJgx2YsUERER+4mMtvL68l/Y8tsNHC0mPutQlobF/wo+ixcvpnfv3oSHh+Po6MiNXcvIc3Y+XD9q6+DoBhV6QtXXbYs9iIikQUl+j9OwYcMIDg4GbKvlDB06lMaNGxMQEMCQIUOSvUARERGxn4joGPotPcSW327gZDEzs3O52NAUFRXFwIED6dq1K+Hh4TR+oRwH+2Qmz0+DbKHJ0d32/NLAX6H+eIUmEUnTkjziFBAQQLFixQBYtWoVTZo04YMPPuDw4cM0btw42QsUERER+wiPiqH34kPsPHMbZwczs7qWp0ahrADcvHmTtm3bsnOn7d1L7zTKxZgKZzDfN4GTB1TsDVUGgHtme16CiEiySXJwcnJyIjQ0FIDvv/+erl27ApApU6bYkSgRERFJ20Ijo+m77Ch7zv2Bq6OFud3KU7VAltjPe3Tvzs6dO/F0MbP4JWdeKhIMzt5Q6VWo3A/cMtmxehGR5Jfk4PT8888zZMgQqlWrxoEDB/jqq68AOHPmDLly5UpgbxEREUntwmOg56LD/HzxHu5OFub3qEhF/z+DkDUGjq/i04pX6HrcwryXXCiSMxNU7guV+4CrnnUWkWdTkoPT559/Tr9+/Vi5ciUzZswgZ07bexe++eYbGjZsmOwFioiIyNNzPzyKGScsXAi5h6ezAwterki5PBmJCH3A9oXjaRD9HfxxlgIm2N0vB6aqA2yjTC7e9i5dRCRFJTk45c6dm40bN8ZrnzJlSrIUJCIiIvYRFBpF9wWHuBBiwtvVgcU9K1EyuzvXvvucVq++yf4LoXzTyY0GJbJClQGYKvYGFy97ly0i8lQkeVU9gHPnzjFq1Cg6dOjArVu3ANuI02+//ZasxYmIiMjTEfggkg6z9/Hr1WDcHQwWdStFyVvr2TWkKGVbvs6+C6F4u5qhTGcYdAxeeEOhSUTSlSQHpx07dlCiRAn279/P6tWrCQkJAeDo0aOMGTMmyQVMmzaNvHnz4uLiQqVKlThw4MAT+9+7d4/+/fuTPXt2nJ2dKVSoEJs3b07yeUVERMTm9v0IOszax4nrwfi6m5iW83tKrK7D9NGvUuvzc9x8YFAif3Z+/vkXGgz9Apw97V2yiMhTl+TgNGLECMaPH8/WrVtxcnKKba9duzb79u1L0rG++uorhgwZwpgxYzh8+DClSpWiQYMGsaNY/xQZGUm9evW4cOECK1eu5PTp08yePTv2OSsRERFJmpvB4bSftZeAm4H0dfuRn1yGUunKPHou/p3+m8OJtkK7Nq3Ye/R38hcrae9yRUTsJsnPOB07doxly5bFa/fx8eHOnTtJOtbkyZN55ZVX6NGjBwAzZ85k06ZNzJs3jxEjRsTrP2/ePAIDA9mzZw+Ojo4A5M2bN6mXICIiIsC1e2F0n7WTakGbGOCyAR9rIDyAL886seDIfcxmM5MmTWLo0KGYTCZ7lysiYldJDk4ZMmTg+vXr+Pv7x2n/5ZdfkjTyExkZyaFDhxg5cmRsm9lspm7duuzdu/eR+6xfv54qVarQv39/1q1bR9asWenYsSNvvvkmFovlkftEREQQERERu/3wXVNRUVFERUUlut6U8rCG1FCLpH66XySpdM/I41y9Hcja+ZNYFLkaX8e7ABie2YmqNADvkj687rubRo0aUadOHaKjo+1craRm+ntGkio13TNJqSHJwal9+/a8+eabfP3115hMJqxWK7t37+aNN96IfRluYty5c4eYmBiyZcsWpz1btmycOnXqkfucP3+eH374gU6dOrF582bOnj1Lv379iIqKeuzzVRMmTGDcuHHx2r/77jvc3NwSXW9K27p1q71LkDRE94skle4ZechijSDL9R/Jf3Mzg0z3wAQhloxM+t2fonW74v5HJrDYpuBHREToOWJJNP09I0mVGu6Z0NDQRPc1GYZhJOXgkZGR9O/fnwULFhATE4ODgwMxMTF07NiRBQsWPHbk55+uXbtGzpw52bNnD1WqVIltHz58ODt27GD//v3x9ilUqBDh4eEEBATEnmfy5Ml89NFHXL9+/ZHnedSIk5+fH3fu3MHLy/6rAUVFRbF161bq1asXO/1Q5HF0v0hS6Z6RWJEPMB+ej7HncxzCbFPrb5qyEl35NYbO281XK76mQYMGrFy5km3btumekUTT3zOSVKnpngkODiZLliwEBQUlmA2SPOLk5OTE7NmzGT16NMePHyckJIQyZcpQsGDBJB0nS5YsWCwWbt68Gaf95s2b+Pr6PnKf7Nmz4+joGCecFS1alBs3bhAZGRlnsYqHnJ2dcXZ2jtfu6Oho9z+ov0tt9UjqpvtFkkr3TDoWEQIHZ8OezyD0DwAuW7PytXs7qjdoR69uXfj111+xWCy8+OKLsf+W6p6RpNI9I0mVGu6ZpJw/ycFp165dPP/88+TOnZvcuXMndfdYTk5OlCtXjm3bttG8eXMArFYr27ZtY8CAAY/cp1q1aixbtgyr1YrZbFsQ8MyZM2TPnv2RoUlERCTdCg/+MzB9DmGBAFzGl0+jmnEya2O6+z+gQe2a3L17Fx8fH1asWEGNGjVSxTMHIiKpUZKXI69duzb+/v689dZbnDhx4j+dfMiQIcyePZuFCxdy8uRJ+vbty4MHD2JX2evatWucxSP69u1LYGAgAwcO5MyZM2zatIkPPviA/v37/6c6REREnhnhQbDjI5haAra9C2GBRHj58zYDqBn+ESezNaP8/b20a/kSd+/epWLFihw6dIgaNWrYu3IRkVQtySNO165d48svv2T58uVMnDiRkiVL0qlTJzp06ECuXLmSdKx27dpx+/Zt3nnnHW7cuEHp0qXZsmVL7IIRly5dih1ZAvDz8+Pbb79l8ODBlCxZkpw5czJw4EDefPPNpF6GiIjIsyXsHuyfCfum28ITQOaCXHiuH813+nIv3KBM7gx80rIwNap0xWq10rNnTz7//HNcXFzsWrqISFqQ5OCUJUsWBgwYwIABAwgICGDZsmUsXLiQkSNH8sILL/DDDz8k6XgPj/Uo27dvj9dWpUqVJL9oV0RE5JkVGgj7ZthCU4TtlRtkKQw1hnPQvQY9Fh4mJCKaCnkzMr9HRTycHVi9ejUHDhygd+/eej+TiEgiJTk4/Z2/vz8jRoygVKlSjB49mh07diRXXSIiIvIkoYGw93PYPwsi79vashaFGsOh2EvsDbhHzwUHCY2MIc+DU9R3yoSHc1UAypQpQ5kyZexYvIhI2vOvg9Pu3btZunQpK1euJDw8nJdeeokJEyYkZ20iIiLyTw/u2FbIOzgHIkNsbdmK2wJTkaZgNvPT77d5ZdHPhEVGk+H0Bn5aP4d9jo5ULFeWUqVK2bd+EZE0KsnBaeTIkXz55Zdcu3aNevXq8cknn/DSSy+lqpfJioiIPHNCbsOeT+HgXIh6YGvzLQE13oTCL8KfzwT/eOoWry45RPiD+7B9Gr8e2QlAr169KFq0qL2qFxFJ85IcnHbu3MmwYcNo27YtWbJkSYmaRERE5KH7N/8KTNFhtrbspf8MTI3gb88offvbDQYsO0zozUs82DSBoBuXcHZ2ZsaMGbEr1oqIyL+T5OC0e/fulKhDRERE/i74Ouz+BA7Nh+hwW1uOslBzBBSsHycwAWz69ToDv/yF4FO7uffNVKLCQ8mVKxerV6+mQoUKdrgAEZFnS6KC0/r162nUqBGOjo6sX7/+iX2bNWuWLIWJiIg8k6wxcHEPhNwEj2yQpyqYLX99HnQVdk+FQwshJsLWlqsC1BgBBerEC0wAa3+5ypAVR7AaUMDxHnvDQ6lRowYrVqzAx8fn6VyXiMgzLlHBqXnz5ty4cQMfHx+aN2/+2H4mk4mYmJjkqk1EROTZcmI9bHkTgq/91eaVAxpOghxlYNcU+GUxxETaPvOrDDXfhHy1HhmYAL7++TLDV/2KYUCbcrn4YPynLKpflq5du+Lo6PgULkpEJH1IVHCyWq2P/L2IiIgk0on1sKIrYMRtD74OK7qAyQLGnz98zF3VFpj8azw2MAEs23+JN2ZtIGjXMvqPm8LEViUxm0307Nkz5a5DRCSdMid1h0WLFhERERGvPTIykkWLFiVLUSIiIs8Ua4xtpOmfoQn+ajNiIM/z0G0jvPwN5Kv5xNC0cM8FBn4wjRuLhxJ6Zg9R+5djNutltiIiKSXJwalHjx4EBQXFa79//75W7BEREXmUi3viTs97nJojwL96gt1m/nCG1wcP5c76DzGiIqhXrx6jRo1KhkJFRORxkhycDMPA9IifgF25cgVvb+9kKUpEROSZEnIz2fpNWnuAwS+3JfjAagCGDx/ON998Q+bMmf9LhSIikoBEL0depkwZTCYTJpOJOnXq4ODw164xMTEEBATQsGHDFClSREQkTfPI9p/7GYbB8NkbmTKsFzHBt3BycWPJogW0adMmmYoUEZEnSXRweria3pEjR2jQoAEeHh6xnzk5OZE3b15atWqV7AWKiIikeTnKgMX5r+XF4zHZVtfLU/WRnxqGwUffnmb5kT8woiPImjMPP2zZSPHixVOuZhERiSPRwWnMmDEA5M2bl3bt2uHi4pJiRYmIiDwzIkPhq85PDk0ADSfGfZ/Tn6xWKx9sPsWcXQE4eGXlrU8WMbT1C2TIkCHFShYRkfgSHZwe6tatW0rUISIi8uyJDIXl7SFgBzi6w/OD4ND8R7zHaSIUi/8C+evXb1ClXlNCCzXArVAV3nvpObpUyfvUyhcRkb8kKjhlypSJM2fOkCVLFjJmzPjIxSEeCgwMTLbiRERE0qzIB7CsHVz4CZw8oNNKyFMFqg+1rbIXctP2TFOeqo8cadq7dx8NmrzE/cBbWC4F8PGQbgpNIiJ2lKjgNGXKFDw9PWN//6TgJCIiku5FhNhC08Vd4OQJnVdB7kq2z8yWBJccnzVrNv369ycmOgrHzLn43xdL6PJ8wadQuIiIPE6igtPfp+d17949pWoRERFJ+yJCYGkbuLQHnL2g82rwq5C4XSMiGPDaa8yZPRsAt0JVmDN3Hh2eL5KSFYuISCIk+T1Ohw8f5tixY7Hb69ato3nz5rz11ltERkYma3EiIiJpSsR9WNLqr9DUZU2iQ1N4eDg1atT8MzSZyFSjKytWfK3QJCKSSiQ5OL366qucOXMGgPPnz9OuXTvc3Nz4+uuvGT58eLIXKCIikiaEB9tC0+V94OwNXddCrvKJ3t3k4EiIVx7Mzu7kaDeWL6d/yIulcqZcvSIikiRJXlXvzJkzlC5dGoCvv/6aGjVqsGzZMnbv3k379u2ZOnVqMpcoIiKSyoUH2ULTlYPg4g1d1kLOso/tHmM1OBAQyM3gMLwdDSoWzM6AZYe5X6oDuQs3ZP6AxtQs7PP06hcRkQQlOTgZhoHVagXg+++/p0mTJgD4+flx586d5K1OREQktQu7B0tawtVD4JIBuq6DHKUf233L8euM23CCa38E88e304kOuoFfxw+IwoyrixNzX21CtQJZnlb1IiKSSEkOTuXLl2f8+PHUrVuXHTt2MGPGDAACAgLIli1bshcoIiKSaoXdg8Ut4NphcM1oC03ZSz22+5bj1+m75DBRwbe4vWYCkTd+B5OZ4IvHcM1Tin41Cyg0iYikUkl+xmnq1KkcPnyYAQMG8Pbbb1OgQAEAVq5cSdWqVZO9QBERkVQpNBAWvfRnaMoE3TY8MTTFWA3GbThB2MVfub5gEJE3fsfs6oVP23dxzWPbb/mBS8RYjad1BSIikgRJHnEqWbJknFX1Hvroo4+wWOK/wE9EROSZ8zA03fgV3DLbQlO25564y/7zf3D6+y+5++M8MKw4ZctP1hZv4eD912yN60HhHAgIpEr+zCl9BSIikkRJDk4PHTp0iJMnTwJQrFgxypZ9/EOwIiIiz4zQQFjUDG4cA7csf4amYgnuNnnCOO7+MAcA9+K1yVS/P2ZH53j9bt0PT/aSRUTkv0tycLp16xbt2rVjx44dZMiQAYB79+5Rq1YtvvzyS7JmzZrcNYqIiKQOD/6whaabx8HdxxaafBL3nqXmrduzZtkCvJ/vhGfZJphMpkf28/F0Sc6KRUQkmST5GafXXnuNkJAQfvvtNwIDAwkMDOT48eMEBwfz+uuvp0SNIiIi9vfgDixsagtNHtmg+8YEQ9OVK1cACHwQyZarFnL2mYtXuaaPDE0mILu3CxX9M6VE9SIi8h8lOTht2bKF6dOnU7Ro0di2YsWKMW3aNL755ptkLU5ERCRVCLkNC5rArd/Awxe6b4KshR/b3TAMJkyYQP78+Zn/9SaafraLXWf/wMXNE7CFpL97uD2maTEs5kePRImIiH0lOThZrVYcHR3jtTs6Osa+30lEROSZEXILFjaB2yfBM7stNGUp+Nju9+/fp02bNrz11ltERkYydMoCrt4Lwz+LOxtee56Zncvi6x13Op6vtwszOpelYfHsKX01IiLyLyX5GafatWszcOBAli9fTo4cOQC4evUqgwcPpk6dOsleoIiIiN3cv2GbnnfnDHjmsE3Py5z/sd3PnDlDixYtOHHiBGYHBzLU6YNn6YbULZqNye1K4eXiSGFfT+oV8+VAQCC37ofj42mbnqeRJhGR1C3Jwenzzz+nWbNm5M2bFz8/PwAuX75M8eLFWbJkSbIXKCIiYhfB120jTX+cBa9c0H0DZMr32O4bN26kU6dOBAcH4+yVmYzNRuCSqyhD6xWiX80CmP8WjCxmk5YcFxFJY5IcnPz8/Dh8+DDff/89p06dAqBo0aLUrVs32YsTERGxi+BrtmeaAs+Bt59t9bxM/o/tvm/fPpo2bQqAe+7nyND0TbL4ZOOT9mWoUUirzYqIPAv+1XucTCYT9erVo169esldj4iIiH0FXbWNNAWeB+/ctul5GfM8cZeKFStSvnYTTt8zyFi7F8X9MjOzczn8Mrk9paJFRCSlJXlxCIBt27bRpEkT8ufPT/78+WnSpAnff/99ctcmIiLydAVdgQUv2kJThtzQY9NjQ9OpU6cIDg4mJCKaAct/4Va5V8hUry9tKvqzqm9VhSYRkWdMkoPT9OnTadiwIZ6engwcOJCBAwfi5eVF48aNmTZtWkrUKCIikvLuXYL5jeFuAGTMC90328LTI6xevZoKFSrQukNnXvrsJzYfu4GTowPjmxfno9YlcXG0PN3aRUQkxSV5qt4HH3zAlClTGDBgQGzb66+/TrVq1fjggw/o379/shYoIiKS4u5etE3Pu3cJMvrbpud554rXLSYmhtGjRzNhwgQAdv92kcyF7pAjayamdy5L2dwZn3blIiLylCR5xOnevXs0bNgwXnv9+vUJCgpKlqJERESemrsXbNPz7l2CTPlt72l6RGgKDAzkxRdfjA1NnuVfIkvb96haNBcbXnteoUlE5BmX5ODUrFkz1qxZE6993bp1NGnSJFmKEhEReSoCA2D+ixB0GTIX+DM05YzX7ddff6VChQp8++23WJycydL0DTLVeYXeNQqwpGclsno626F4ERF5mpI8Va9YsWK8//77bN++nSpVqgC2ZVh3797N0KFD+fTTT2P7vv7668lXqYiISHL645zt5bbBVyFzQdv0PE/feN1iYmJo06YN58+fxzmjL5mav0WGXAX5sHVJmpTMYYfCRUTEHpIcnObOnUvGjBk5ceIEJ06ciG3PkCEDc+fOjd02mUwKTiIikjr9cc72nqb71yBLYdt7mjyzPbKrxWKh8/BJTJo4gYyNB1HQLzszu5SjUDbPp1y0iIjYU5KDU0BAQErUISIi8nTc+d020nT/OmQtYgtNHj5xu9y5wy+//EL1mrUZu/43vvzdkSyt3qF+sWz8r20pvFwc7VS8iIjYy796Aa6IiEiadPuMbfW8kJvgUwy6rgePrHG6HD58mBYtWnDr9m0qvj6di2TFbIKh9QvTt0Z+zGaTnYoXERF7+lcvwBUREUlzbp+2rZ4XchOyFf9zpCluaFq0aBHVqlXj0qVLWF0z8vvtUDK6ObLw5Yr0r1VAoUlEJB3TiJOIiDz7bp20Tc97cBuylYCu68A9c+zHUVFRDB06lM8++wwA1/wVyNJkKKXy52RG57Lkyuhmr8pFRCSVUHASEZFn280TttAUegd8S9pCk1umvz6+eZM2bdrw008/AeBdtQPez3egXYXcvPtScVwcLfaqXEREUpFETdVr2bIlwcHBgG0aQ0RERIoWJSIikixuHLc90xR6B7KXiheawLZa7E8//YTF2Y2sLUfjU7MLE1qWYlKrkgpNIiISK1HBaePGjTx48ACAHj16EBQUlKJFiYiI/Gc3jv050vQH5CjzyNAEUKJxVzKWb0K2Lh+Tv3wNVvSpQsdKuTGZ9DyTiIj8JVFT9YoUKcLIkSOpVasWhmGwYsUKvLy8Htm3a9euyVqgiIhIkl0/CotegrC7kLMcdF4NrhkAiIiIYOrUqQx47XU+3X6BL3aex6tOH6rky8xnHcuQxcPZvrWLiEiqlKjgNHPmTIYMGcKmTZswmUyMGjXqkT+JM5lMCk4iImJf136BRc0h/B7kLA9dVoOLt+2ja9do1aoV+/btY/amvURXfQWAV1/Ix7AGhXGwaLFZERF5tEQFp6pVq7Jv3z4AzGYzZ86cwcfHJ4G9REREnrKrh2FxcwgPglwVofMqcLHNkNi1axetW7fm5s2bWFw8uJ+tDFmcLHzUphSNS2S3b90iIpLqJXlVvYCAALJmzZpwRxERkafpyiFY3AIigsCvMnReCc6eGIbBjBkzGDhwINHR0ThlzUuWFm9TuFABZnUpRwEfT3tXLiIiaUCSg1OePHm4d+8ec+fO5eTJkwAUK1aMnj174u3tnewFioiIJOjKz3+GpmDIXQU6fQ3OnoSFhdGvXz8WLFgAgFuR6mRuNJBGZfLwvzal8HRxtG/dIiKSZiR5MvfPP/9M/vz5mTJlCoGBgQQGBjJlyhTy58/P4cOHU6JGERGRx7t8wPZMU0Qw5KkGnWwjTQA3btxgzZq1YDKToebL+Lw0nJHNSjOzczmFJhERSZIkjzgNHjyYZs2aMXv2bBwcbLtHR0fTq1cvBg0axM6dO5O9SBERkUe6tA+WtILIEMhbHTp+BU7usR9fifYgc7PhOEdayVGsAp91KMvzBbPYsWAREUmrkhycfv755zihCcDBwYHhw4dTvnz5ZC1ORETksS7uhaWtbaHJ/wXo8BWGoytTp0yhcOEiBLgW5OPvTmPNUZJKubyZ0bkcOTO42rtqERFJo5IcnLy8vLh06RJFihSJ03758mU8PfWArYiIPAUXdsPSNhD1APLVhPbLCY2GXt07sXz5cpzcPPHpMR2LR0baV/BjbLPncHG02LtqERFJw5IcnNq1a0fPnj353//+R9WqVQHYvXs3w4YNo0OHDsleoIiISBwBP8GythAVCvlrQ/tlnL98nRYtWvDrr79iMltwr9oRF69MvNe8OO0r5rZ3xSIi8gxIcnD63//+F/ui2+joaAAcHR3p27cvEydOTPYCRUREYp3fAcvaQXQYFKgL7Zby3Y87ad++PXfv3sXinoEsL40gX/HyzOhcjlJ+GexdsYiIPCOSHJycnJz45JNPmDBhAufOnQMgf/78uLm5JXtxIiIisc79CMvbQ3Q4FKyP0WYRkz6eyttvv43VasUpeyGyNn+LGmUL82n7MmT2cLZ3xSIi8gxJcnB6yM3NjRIlSiRnLSIiIo92dht82dEWmgo1hLaLwOLE8ZOnsVqteJSsT6Z6felbpwhv1C+EgyXJb9sQERF5on8dnERERJ6Ks9/D8o4QEwGFG0ObBeDgzC+X7nImX1uyvOSLT8kafNy2FA2LZ7d3tSIi8ozSj+RERCT1+n3rX6GpSBM2urWlfeduLNx9nnZf7OVWmJWS1RuwbsDzCk0iIpKiNOIkIiKp0+ktsKILxERiLdyE8acKMGZcSwC23s2MZ5nGNCruy0dtSuHhrH/OREQkZelfGhERSX1OfwNfdQFrFEF5G9F1eTDrN7wLgGfZF/EqVY8RjYrQ+4V8mEwmOxcrIiLpwb8KTr///js//vgjt27dwmq1xvnsnXfeSZbCREQknTq1CVZ0A2sUJ7xr0WLCYc6c+R2TgyOZ6vcnd+XGfN6hDFULZLF3pSIiko4kOTjNnj2bvn37kiVLFnx9feP8pM9kMik4iYjIv3dyA3zdHazRbI6sSLt3thESEoLFKytZm79FxYoVmNGpLDkyuNq7UhERSWeSHJzGjx/P+++/z5tvvpkS9YiISHp1Yh2sfBms0VCiDZ45ehH2cU2cc5cg60sj6FyzBGObFcPZwWLvSkVEJB1KcnC6e/cubdq0SYlaREQkvTq+Glb1IiYmGkvpDpyuPInRS38ha/uJeOQswPstStG2gp+9qxQRkXQsycuRt2nThu+++y4lahERkfTo2EpY1Yuj1yMpMd+Bjx68SPMZ+7jwRyj5ipZkVb/nFZpERMTukjziVKBAAUaPHs2+ffsoUaIEjo6OcT5//fXXk604ERF5xv36NazpzfJfI+i5MYqwyAeMfustsnWcSPWCWfm0QxkyuTvZu0oREZGkB6dZs2bh4eHBjh072LFjR5zPTCaTgpOIiCTO0a+IXvUqb24NY/K+SABc8pYhS7Ph9K9VgKH1C2Mxa6lxERFJHZIcnAICAlKiDhERSU+OLOf20ldpvyqUHwJiAPCq3JpcdXvwcbuyNCzua+cCRURE4vpPL8A1DANALx8UEZHE+2UpVxf3peq8EC4FGZgcXcjceBClX2jIzC7lyJ/Vw94VioiIxJPkxSEAFi1aRIkSJXB1dcXV1ZWSJUuyePHi5K5NRETSMmsMBPyE9dev+W33Jtb9comz387AWNef7J6Q2y8XDhlz4NvlY9q2acPa/tUUmkREJNVK8ojT5MmTGT16NAMGDKBatWoA7Nq1iz59+nDnzh0GDx6c7EWKiEgac2I9bHkTgq9hBp4DfKLd8DAeYHI0sdKhCRfrNienCd5uXp5e1f01e0FERFK1JAenzz77jBkzZtC1a9fYtmbNmvHcc88xduxYBScRkfTuxHpY0RUDg4dR6EaIlbZf38bP20SPZlUYFt4RLy9HZnctT5X8me1aroiISGIkOThdv36dqlWrxmuvWrUq169fT5aiREQkjbLGwJY344SmfVeiabUijGv3DbycoX/QeczuBm5OFir6Z7JruSIiIomV5GecChQowIoVK+K1f/XVVxQsWDBZihIRkTTq4h4IvhYbmmYfiqTGglCu3TcoksXMgV7uVM18n4rmU9y6H8GBgEC7lisiIpJYSR5xGjduHO3atWPnzp2xzzjt3r2bbdu2PTJQiYhIOhJyE4CIaIPXvwln1uEoAFoUcWBBc1e8nG2Ryod7ANy6H26XMkVERJIqycGpVatW7N+/nylTprB27VoAihYtyoEDByhTpkxy1yciImmJRzYA2q4MY/3paEzA+NrOjHjeCfPfFn+4RQYAfDxd7FCkiIhI0v2r9ziVK1eOJUuWJHctIiKS1uUsDw6uDKkczd7LMSxs7kKjgo6xH1sNuEFmDlqLkN3bRc84iYhImpGo4BQcHIyXl1fs75/kYT8REUk/DMMg4NRx8h4cjTk6jBfyOHD+dQ88nP8aZbLa3pnOu1FdsGJmTNNiWMxaglxERNKGRC0OkTFjRm7dugVAhgwZyJgxY7xfD9tFRCR9CQsLo0eXjpQqU4bT+7/ngeHMJ9EtMTyzxel3g8z0jRrEUc8XmNG5LA2LZ7dTxSIiIkmXqBGnH374gUyZbNMpfvzxxxQtSERE0o5Lly7R8qUmHDpyDLMJvr/iyE+Z3qFH+1Z4PucDF/dgvX+Dk/fdOOtWgu5e7lT0z6SRJhERSXMSFZxq1KgR+3t/f3/8/PziveHdMAwuX76cvNWJiEiq9eOPP9K2TSvu/HGXzK4mPm+VjXVFJvJO9+YUz+lt6+RfHTPw3J+/RERE0qokv8fJ39+f27dvx2sPDAzE398/WYoSEZHUyzAMJk+eTL169bjzx13K+Jr5ulc+vq84k09eb/dXaBIREXmGJHlVPcMw4o02AYSEhODiomVlRUSedYsXL2bo0KEAdCnpSN/Gz7GpxFQ+b/M8Lo4WO1cnIiKSMhIdnIYMGQKAyWRi9OjRuLm5xX4WExPD/v37KV26dLIXKCIiqUvTQmZq+zvQvLADJcqX47daM3mv1nOP/KGaiIjIsyLRwemXX34BbCNOx44dw8nJKfYzJycnSpUqxRtvvJH8FYqIiN3t2bOHChUqcG/HdLLueoetXVzZzPPEtP6CXiX87F2eiIhIikt0cHq4ml6PHj345JNP9L4mEZF0wDAMJk6cyNtvv033RuWYV+EMAKscmlCy1wwKZ9fzTCIikj4k+RmnqVOnEh0dHa89MDAQBwcHBSoRkWfE/fv36dGjB6tWrQLA4cZRrIYLX3l2p/6rH5LZU8+1iohI+pHkVfXat2/Pl19+Ga99xYoVtG/fPlmKEhER+zpz5gyVK1dm1apVOFhMfNHEhRlN3FibazitBk1VaBIRkXQnycFp//791KpVK157zZo12b9/f7IUJSIi9rNx40YqVKjAiRMn8PF0Ymc3V7qVdWN7yY9o0ettnByS/E+HiIhImpfkqXoRERGPnKoXFRVFWFhYshQlIiL2cffuXTp16kRwcDDlcruzsY0JTw83TtWeTZ0Xmtm7PBEREbtJ8o8NK1asyKxZs+K1z5w5k3LlyiVLUSIiYh8ZM2Zk2Nuj6FIhA3u6mnH2yEhg6zWUUmgSEZF0LskjTuPHj6du3bocPXqUOnXqALBt2zYOHjzId999l+wFiohIyjpx4gT37t2jSpUqfLnpW1558DnZGlu5bfHFucc6/HIVtXeJIiIidpfkEadq1aqxd+9e/Pz8WLFiBRs2bKBAgQL8+uuvVK9ePSVqFBGRFLJ69WoqVapE8+bNGfvxVBof7EE20z1uuuQjw2s/4qXQJCIiAvyLESeA0qVLs3Tp0uSuRUREnpKYmBhGjx7NhAkTAMjq60vPG+Px9ojmVoYyZHt1DbhmtHOVIiIiqce/Ck4PhYeHExkZGadN73ESEUndAgMD6dixI99++y0AVapW5ofaJ3GxGATmrIVPt2Xg5GbnKkVERFKXJE/VCw0NZcCAAfj4+ODu7k7GjBnj/BIRkdTr119/pUKFCnz77bc4ubjQrkU99tQ7gYvFIKRwazK9/LVCk4iIyCMkOTgNGzaMH374gRkzZuDs7MycOXMYN24cOXLkYNGiRf+qiGnTppE3b15cXFyoVKkSBw4cSNR+X375JSaTiebNm/+r84qIpDeffPIJ58+fJ5NvLoZ2rsOXJW3v34uo0BePdrPB4mjnCkVERFKnJE/V27BhA4sWLaJmzZr06NGD6tWrU6BAAfLkycPSpUvp1KlTko731VdfMWTIEGbOnEmlSpWY+v/27js6qmpv4/h3ZtJJAgRIQiBUaUGKgGAookhTpDeRLqBcQBFQARUBG6CgWBCklyu9vYBIEeGCdEFQBKJ0kIQOKaRMMuf9IxINAZJBkpkkz2ct1s3s2efM70z2yp3Hvc+eiRNp0qQJYWFh+Pv73/W4U6dO8dprr2lDChERO4yb8Cl7z0YzsFIkvXy2AWB7ahTudV8Fk8mxxYmIiDgxu2ecrl69SqlSpYDk+5muXr0KQN26ddm6davdBXzyySf06dOHnj17EhISwpQpU/Dy8mLmzJl3PSYpKYnOnTszevTolFpERCStS5cusWjRImw2G+evx9Jr9l4+rnWZXj47sGGGFl9grjdIoUlERCQdds84lSpVipMnT1KsWDHKly/P4sWLqVmzJqtXryZfvnx2nSshIYF9+/YxfPjwlDaz2UzDhg3ZuXPnXY9799138ff3p1evXmzbtu2erxEfH098fHzK48jISACsVitWq9WuejPDrRqcoRZxfhovYo/9+/fTvn17zp49S97CJTicrybjrGN5zHIEm9kNW5vpJJV7BjSe5B/0d0bspTEj9nKmMWNPDXYHp549e3Lw4EHq16/PsGHDaN68OV9++SVWq5VPPvnErnNdvnyZpKQkAgICUrUHBARw9OjROx7z448/MmPGDA4cOJCh1xgzZgyjR49O075hwwa8vJznBuiNGzc6ugTJRjReJD237kW1Wq34+Qfx4w1fFuV5hxDzaRLMnuwpNYgrx4Hjax1dqjgp/Z0Re2nMiL2cYczcvHkzw33tDk6DBg1K+blhw4YcPXqUffv28dBDD1G5cmV7T2eXqKgounbtyrRp0yhYsGCGjhk+fDiDBw9OeRwZGUlwcDCNGzd2iq3TrVYrGzdupFGjRri66qZsuTeNF0mP1Wrl9ddf56uvvgKg3KOPk79Oexb4fE4J8wVsXoUwdVpMrcBKDq5UnJX+zoi9NGbEXs40Zm6tRssIu4KT1WqladOmTJkyhTJlygBQvHhxihcvbl+FfylYsCAWi4ULFy6kar9w4QKBgYFp+h8/fpxTp07RvHnzlDabzQaAi4sLYWFhlC5dOtUx7u7uuLu7pzmXq6urw39R/+Rs9Yhz03iRO4mIiKBDhw4pS5gfadkb//JVmOv+Mf6m6xj5S2DuugKzn+4NlfTp74zYS2NG7OUMY8ae17drcwhXV1d++eUXuwu6Gzc3N6pXr86mTZtS2mw2G5s2bSI0NDRN//Lly/Prr79y4MCBlH8tWrTgySef5MCBAwQHBz+w2kREspvTp0+za9cufHx8qfrChzxUoRxL3N9PDk3+D2N6YT0oNImIiNwXu5fqdenShRkzZjB27NgHUsDgwYPp3r07NWrUoGbNmkycOJGYmBh69uwJQLdu3ShSpAhjxozBw8ODhx9+ONXxtzakuL1dRCS3qVWrFqM+mcK83w1qeJ/jS7cJuGPlcp5y5O36f7j6ZGyJs4iIiKRld3BKTExk5syZfP/991SvXp08efKket7eDSI6duzIpUuXeOedd4iIiKBq1aqsW7cuZcOIM2fOYDbbvWu6iEiOFx8fz5AhQ+jduzdVq1Zl0d4zzAoPpJXPZsa6TseCDVuZpuz0akdTj7yOLldERCRbszs4HTp0iGrVqgHw+++/p3rOdJ/fAzJgwAAGDBhwx+e2bNlyz2Nnz559X68pIpKd/fnnn7Rt25bdu3ezbt06nv9oKXP3/MlLltUMd12Q3KlqF5KeHo9t3QbHFisiIpIDZDg4nThxgpIlS7J58+bMrEdERNKxbds22rdvz4ULF8ibLx/Fm7/MvD1nGe6ygJdcvk3uVGcgNBwNiYmOLVZERCSHyPAauDJlynDp0qWUxx07dkyzG56IiGQewzD48ssvadCgARcuXKBcSEUq9P2S0+4l+MRt6t+hqdF70OhduM9VACIiIpJWhoOTYRipHq9du5aYmJgHXpCIiKQVHx9Pz549efnll0lMTOSpZq1xbfUh1wxPZnt9TmvzVjBZoNVkqPOKo8sVERHJcey+x0lERLKei4sLERERmM1mOvQbxm7v2ngnRbPc5zNCrL+Biwe0nw3lnnZ0qSIiIjlShoOTyWRKs/nD/W4GISIiGWMYBiaTCYvFwpx5/+W1r1exLToAf+MaK/NOICj+BLjnhecXQfG0338nIiIiD0aGg5NhGPTo0QN3d3cA4uLi6Nu3b5rtyJcvX/5gKxQRyYUMw+DTTz/ljz/+YPLkyVyLSeDVFcfYGR1ASXM4/+c7Ad+48+AdCF2WQaC+y05ERCQzZTg4de/ePdXjLl26PPBiREQEYmJi6N27NwsXLgSgdsNmTD+Rh9NXblLD7TTfeI7HPe4K+JWCrisgfwnHFiwiIpILZDg4zZo1KzPrEBERkr/6oXXr1vzyyy+4uLjwn2HvMu6gC9EJN2nue4yJxkdY4qMhsDJ0WQ7ehRxdsoiISK6Q4V31REQkc61fv54aNWrwyy+/4O/vz6BP/8uaxMpEJyQxIPAwnye9j8UaDSXqQY9vFZpERESykHbVExFxAp9//jmvvvoqhmFQs2ZNHnnhPRaftAIwodTPtAmfgMmwQYXm0GY6uHo4uGIREZHcRcFJRMQJlC1bFoAu3XsSXa0b607GYDYZLA7ZSY3jXyZ3qtYdnv0UzBYHVioiIpI7KTiJiDiI1WrF1dUVgKZNm7J0/VYm7Evgz/Mx+HqYWVtuHUXDZid3rvcaNHgb9DUQIiIiDqF7nEREHGDNmjWULVuW48ePA7D+twje3hbNn9djKe3nxvYyi/8OTU3HwlMjFJpEREQcSMFJRCQL2Ww2Ro8eTfPmzTl16hRjx45l0uZjvDRvHzcTkmhQKg/rAifj88dyMLtAm2nw2H8cXbaIiEiup6V6IiJZ5MaNG3Tt2pXVq1cD8NJ/+uFSuzsfrw9LfvxofoZefQfzib3g4gkd50GZRo4sWURERP6i4CQikgUOHz5Mq1at+OOPP3B3d+ejT79gky2Eg4cuYTGb+KhxAdr+9gpcOgoe+aDzEgiu6eiyRURE5C8KTiIimWzv3r00aNCA6OhogoOD+WjyXCYeTCIi8jp5PV2Z9Ww+qm19AW6cBZ8g6Loc/Cs4umwRERH5BwUnEZFMVrlyZUJCQvDy8qL3yM8ZtekccVYbD/l7M7eJC0HfPgc3r0CBh6DrCshXzNEli4iIyG0UnEREMsH169fx8fHBYrHg7u7O6tVrmLv/Em+tOwnAE+UKMSk0kjzLOoE1BoIegc5LIU9BB1cuIiIid6Jd9UREHrCDBw9SrVo1RowYAcDNhERGbTzDl1uSQ1OfeiWZWeMceRY/lxyaSj0B3VcrNImIiDgxBScRkQdowYIFhIaGcvLkSRYtWsTvZy/SfspO1v4agavFxEdtK/OW/w7My14AmxVCWsHzi8Hdx9Gli4iIyD0oOImIPACJiYkMGTKE559/ntjYWBo3bsz0FRt5fs4v/HY+Er88bszvXYsOMd/At0MAA2r0gnYzwcXd0eWLiIhIOnSPk4jIvxRx4SLPtm7Hvp3bABg6bBjV2/TlpcWHSUi0UT7Qh2ldqxG8axTsnZZ8UP1h8MQwMJkcV7iIiIhkmIKTiMi/8O2Bs7RrXJe4S2cwuXlS4JlXWe36OAuXHQKgYYUAJrYPwfvb/vDbcsAEz3wMNfs4tnARERGxi4KTiMh9WnconAELf8H7sQ4kbl9AodZv4VawGDEJSQA0qRjA5PblMC9+Hk5sBrMrtJ4Cldo5uHIRERGxl4KTiIidrFYrJ06eYvTqMxhAnpAn8CpbG5OLW6p+p8+exTT3dTi/H1zzQMd58NBTjilaRERE/hUFJxERO0RERNChQweO/n4M93YfYfHOD4DFxYWa5sP4c52L5OOcrSCT4sZhOh8Onn7J39FUtLqDqxcREZH7peAkIpJBu3btom3btpw/fx4vbx/M1/7E4p2fJuY9jHSdS5DpakrfJMOExWRw0zMQrxdWQaFyDqxcRERE/i1tRy4ikgFTp06lfv36nD9/ngoVKjBzxUY8gh+miXkPk10nEsjVVP0tJgPDgPDKLys0iYiI5AAKTiIi9xAfH8+LL77ISy+9REJCAm3atGHXrl2cTsyLGRsjXecCYL7DruKGCUod+QpsSVlctYiIiDxoCk4iIvfw7rvvMm3aNEwmEx9++CHzFy5m1Hcn+GrLcWqajxJkunrH0ATJf2BNkX/C6R1ZWrOIiIg8eApOIiL3MHToUGrXrs3atWvpP+g1Xpizl+U//4nFbGJgTZ+MnST6QuYWKSIiIplOwUlE5B8Mw2D9+vUYhgGAr68vP/74I1VC69Nhyk62H7uCl5uFWZ3KERqzKWMn9Q7IxIpFREQkKyg4iYj8JTY2lh49etC0aVM+++yzlPawC1G0nrSDoxFRFPR2Z23TGB7f8Cz8sSGdM5rAtwgUr525hYuIiEim03bkIiLAmTNnaN26Nfv378ds/vu/KW0/dpm+8/YRFZ/IIwWS+G/R/5Jnw4rkJ/1KQZVOsPnDv3ob/zjjXzc+NR0LZkuWXIOIiIhkHgUnEcn1Nm/eTIcOHbh8+TIFCxZk0aJFNGjQgOX7zzF02S9Yk2wMDPiNgdapmMMug8kMof3hiTfBzQsKlYd1QyHy/N8n9Q1KDk0hLRx3YSIiIvLAKDiJSK5lGAaffvopb7zxBklJSVSvXp1ly5ZRrFgxJm0+xsfrwyjENaYVWEDVGz8mH1SoArScBEWr/32ikBZQvlny7nnRF5LvaSpeWzNNIiIiOYiCk4jkWocPH04JTd27d2fy5Mm4urnz5opDLNhzmvaW//GuxwI8Y6LA7AL1hiT/c3FPezKzBUrWy/qLEBERkSyh4CQiuVbFihWZOHEiJpOJfv36cTMhiRfn7SPs6G/MdZvO4+ZfIQkoXDV5linwYUeXLCIiIg6i4CQiucr69esJDg4mJCQEgAEDBgBwKSqe3rN3UzliGV+4LySPKQ4s7vDkmxA6ACz6cykiIpKb6ZOAiOQKhmEwZswY3n77bcqUKcOePXvImzcvAMcvRTNixgrevPkltVyPJh9QLBRafAkFH3Jg1SIiIuIsFJxEJMeLioqiR48eLF++HIAnnngCDw8PAH46cZFt80Yz07YID7MVm4sX5kaj4dHeYNZX3YmIiEgyBScRydHCwsJo3bo1R44cwc3NjS+//JI+ffoAsO3HLeTbOIhBphNggoTi9XFr9QXkL+7gqkVERMTZKDiJSI61evVqunTpQmRkJEFBQSxbtozHHnsMEhPY/83b1DoxHTdTEjFmb1ye/hD3Gt3AZHJ02SIiIuKEFJxEJEcyDINPPvmEyMhI6taty5IlSwgMDMR2dh+X5vehWuxxMMFh33qU6zUVS94gR5csIiIiTkwL+EUkRzKZTCxatIi33nqLTZs2EVggL4nr3oYZDQmIPc4Vw4cNIWOo8OoqhSYRERFJl4KTiOQYhw8fZty4cSmP/f39ef/993EL/4mkr2rjsusLzNhYbavNnme+o3GHfpi0AYSIiIhkgJbqiUiOsGzZMnr06EF0dDQlSpSgY8eOEB8F34+GvdOwABFGft439aFz976Eli7g6JJFREQkG1FwEpFsLSkpiREjRjBmzBggeavxBg0awLFNsHog3DgLwILEJ5nl9QJf9nqSsgE+jixZREREsiEFJxHJtq5evcrzzz/P+vXrARg8eDDjRg3DZdMoOPBfAM4ZhRhq7c3VgDrM6/koAb4eDqxYREREsisFJxHJlg4ePEjr1q05efIknp6eTJ8+necf8YWv60D0BQxMzE5szMeJHalepiiLO1fDx8PV0WWLiIhINqXgJCLZ0smTJzl58iQlS5ZkxTczqHJuDixaAcAVj+K8eKMH+4xytKtelDFtKuFq0SYQIiIicv8UnEQkW2rVqhVz58yhWckk/La+ALFXMUwWvvd7jgF/NiIeNwY+VYZXG5bBpC+1FRERkX9J/wlWRLKFS5cu0aFDB86eTd7sgcjzdHX5Fr9Nr0LsVZIKVeTtQp/R589mJJrdGde2EoMalVVoEhERkQdCM04i4vT27dtH69atOXv2LJcvX+aHj7vDhrchPhLMrkTVGkSnw49x6EIcedwsTOpcjSfK+Tu6bBEREclBNOMkIk5t9uzZ1KlTh7Nnz1KmdAm+aGjA6leSQ1OR6hxv+x2N9iWHpkI+7ix6KVShSURERB44zTiJiFNKSEhg8ODBTJo0CYBn6zzMvCcuki9hP7h4QoO32V6wPX2/OUBUfCIP+Xszu+ejFM3v5eDKRUREJCdScBIRp3Pp0iXatGnDjz/+CMDI5qV455HTmE0mKF4XWnzO8tPuvDF7H4k2g5ol/ZjWtQZ5vbTduIiIiGQOBScRcTp58uQhKioS3zwezGvpRosyl8HNFxqNxqjeg0lbTjB+w0EAnq1cmAkdquDuYnFw1SIiIpKTKTiJiNMwDAOTyYTXjWOsaOdGQoSFcgWBhxpB84kkegcxYuVvLNhzBoCXHi/F0KblMZu1c56IiIhkLgUnEXG4+Ph4Xn75ZYoHF+Gt2mbYPpGStkQoWgCeHgeVOxKTkMTL8/bxw9GLmEwwukVFuoWWcHTpIiIikksoOImIQ/3555+0bduW3bt342I28fyAPJTMb4YKLeCZ8eATwKWoeF6YvZdf/7yBu4uZzzs9QpOKgY4uXURERHIRBScRcZht27bRvn07Lly4SH4PEwvaelKyaCA0Gw8hLQE4fimaHrP2cPZqLH553JjevQbViuV3cOUiIiKS2yg4iUiWMwyDSZMmMWjQqyQmJlE5wMyKjl6UeqIzNPkQvPwA+OnUVXrP/YnrN60UL+DFnJ41KVEwj4OrFxERkdxIwUlEslzfPi8wdcZsADo97MK050uTp+3nULZxSp+1v4bz6qIDJCTaqBKcj5nda1DA291BFYuIiEhup+AkIlnr9w08GrmOGSb4uJE7rw7oi6nRu+Dhm9Jl+rYTfLD2CIYBDSsE8EWnR/B003bjIiIi4jgKTiKSJeKunsdjy2j4ZSG9Q6DusHKU7zMNStZL6WOzGbz/7RFmbj8JQNfHijOqRUUs2m5cREREHMzs6AJEJGczDIMJr/ekUtkSXNk1H0xmCB1A+VH7U4WmOGsS/efvTwlNw54uz7stFZpERETEOWjGSUQyTUzECXq3bcDCHacBmHu8MIM+WwZFa6Tqdy0mgd5zf2Lf6Wu4Wcx83L4yLasWcUTJIiIiInek4CQiD55hcOLbz2jd5w1+ibDiYoaJ/Z+l34Ql4OqRquuZKzfpMWsPJy7H4OvhwtRuNXisVAEHFS4iIiJyZwpOIvJg3TjH+g+eo9MX27kWBwG+riyZM5V6rXqk6Xrw7HV6zdnL5egEgvJ6MPuFmpQN8Mn6mkVERETSoeAkIvaxJcHpHRB9AbwDoHhtMFvAZoN9s1jyyRt0XHgVA6gVUoxl322lSLHiaU6z6cgFBsz/mVhrEiGFfZnV81ECfD3Svp6IiIiIE1BwEpGMO7wK1g2FyPN/t/kGQZ1BcPj/4PSPPFXMoGQhTxo2bcHn0+bg7p72u5e+2X2aESsPYTOgXpmCTO5SHW93/TkSERER56VPKiKSMYdXweJugJG6PfI8EUuGEOhtBlcv/JqOZO/rbfErWCjNKQzD4OP1YXy15TgA7asX5cM2lXC1aINPERERcW4KTiKSPltS8kzT7aEJWB1mpcuKWD5u4s2LM7ZBwYfwu8MpEhJtvLH0ICsPJM9WvdqwDAOfKoPJpO3GRURExPnpP/OKSPpO70i9PA+wGQajtsTRYmEskfGw+NdYjNv63BIZZ6XHrD2sPHAei9nER20r82rDsgpNIiIikm1oxklE0hd9IdXD63EGXVfEsub3RABerunGhMbumGIupjn0/PVYes7aS9iFKPK4WfiqS3Xql027jE9ERETEmSk4iUj6vANSfvztYhKtF8Xyx1Ub7hb4+lkPuld1S9MP4Eh4JD1n7SUiMg5/H3dm9niUh4vkzcrKRURERB4IBScRSV+xUHDz5sr1SGrPjCEyHorlNbG8gxfVgyyAKXl3veK1Uw758Y/L9P3vPqLjE3nI35vZPR+laH4vx12DiIiIyL+g4CQi92YY8P1ISIimgJeZYXXc2XgikUXtPCmUxwz8dZ9S07HJ3+cELNt3jqHLfiHRZlCrpB9Tu9Ygr5er465BRERE5F9ScBKRuzMMri5+haidsymezwzVujPMZwOv3ziPi/mvwOQblByaQlpgGAaTNh9j/IbfAWheJYjx7Svj7mJx4EWIiIiI/HsKTiJyZzYbB7/sQuuRC/F2M7Fz2VfkqfsSJlsSLqd3JG8Y4R2QvDzPbCExycaI/zvEgj1nAXipfimGNimP2ayd80RERCT7U3ASkbRsScx/rQm9v9hEbCKULFKQ8MCneAjAbCGpeF32nLzKxRtx+J+8TsUgXwYu/JnNYZcwm2BUi4p0Cy3h4IsQEREReXAUnEQklcT4OIa2f5RPVh8CoHHtKixY/QN+fslfa7vuUDijVx8m/EZcyjGuZhNWm4GHq5nPn3uExhUDHVK7iIiISGZRcBKRFJciztOxYXU2/xYBwPA+bXlv8iIsluR7lNYdCuc//92PcdtxVltyyysNyig0iYiISI5kdnQBIuIkEuPp1/IxNv8WQR43WPrZm3w4dWlKaEqyGYxefThNaPqnebtOk2S7Vw8RERGR7EnBSUTAGgeLujCxzjXqFXdl94pptH3lg1Rd9py8mmp53p2E34hjz8mrmVmpiIiIiENoqZ5ILpaQkMD6NatoHv1fOLGZIn552LppFZR+Mk3fi1H3Dk329hMRERHJTjTjJJJLRURE8NSTT9CibXuWrdkArnmgy9I7hqb4xCR+OHIxQ+f19/F4wJWKiIiIOJ5mnERyoZ07d9KubVvOh4fj6w7uHp7QdTkUeyxN3wNnr/P6koP8cTH6nuc0AYF5PahZ0i+TqhYRERFxHM04ieQyU6dOpX79+pwPD6dCQTN7+wfy7Icb0oSmOGsSY747QpuvtvPHxWgKervx0uMlMZEckv7p1uORzUOw6AtvRUREJAfSjJNILhEfH8+AAQOYPn06AG0quDD7ucL49F4FQVVT9d13+hpvLD3I8UsxALSsGsSo5hXJn8eNR4rlT/M9ToF5PRjZPISmDxfOsusRERERyUoKTiK5xPfff8/06dMxmeCDJ90Z1jgIU4/VEFAxpU+cNYkJG8KY/uNJDAMK+bjzQauHU303U9OHC9MoJJA9J69yMSoOf5/k5XmaaRIREZGcTMFJJJdo9nh13nk6iFC/6zStWgS6rQL/8inP/3TqKm8s/YUTl5NnmdpUK8I7z4aQz8stzbksZhOhpQtkWe0iIiIijqbgJJJDGYbBtGnTaNGiBYFeNpjTnNE1o8GnGHRfDQUfAiA2IYmP14cxa0fyLFOArzsftq7EUxUCHHwFIiIiIs5DwUkkB4qNjaVv377MnTuXebOm80P7OFwjT0PeYtB9FfiVBGD3iSu8sewXTl+5CUC76kUZ0SyEvF6ujixfRERExOkoOInkMKdPn6ZNmzbs378fs9lMa/8zuNy4mRyWuq+GfMW4mZDIR+vCmL3jFACBvh6MaVuJJ8v5O7Z4ERERESel4CSSg/zwww906NCBK1euULBAfha186ZB4A0oUCY5NPkGsfP4FYYu+4UzV5NnmZ57NJg3m1XA10OzTCIiIiJ3o+AkkgMYhsEnn3zCG2+8gc1mo1rlEJY/e5PirlehUHnotooYtwKMXXmIebtOAxCU14MxbStTv2whB1cvIiIi4vwUnERygJs3bzJ9+nRsNhvd2rdgSpVf8LRehYCHodv/sT0chi7byrlrsQA8X6sYw58uj49mmUREREQyRMFJJAfIkycPK1euZNOyOfzH/A2muOtQuApRHZYwZkM483efAaBIPk/Gta1M3TIFHVuwiIiISDZjdnQBAJMmTaJEiRJ4eHhQq1Yt9uzZc9e+06ZNo169euTPn5/8+fPTsGHDe/YXyanWrVvH5MmTUx6XyxNNP/N/k0NTkRpsrzOLpl//lhKauj5WnPWDHldoEhEREbkPDg9OixYtYvDgwYwcOZL9+/dTpUoVmjRpwsWLF+/Yf8uWLXTq1InNmzezc+dOgoODady4MX/++WcWVy7iGIZhMHbsWJ555hlefvlldu7cCWd2wdyWEHeDxKKPMTLv+3T+71H+vB5LsJ8n8/vU4r1WD+PtrklmERERkfvh8E9Rn3zyCX369KFnz54ATJkyhW+//ZaZM2cybNiwNP2/+eabVI+nT5/OsmXL2LRpE926dcuSmkUcJSoqinHjxrFr1y4AevfuTbX8N2FeV7DGcM3/MdpefJkTx64C0KN2CV5vUo48CkwiIiIi/4pDP00lJCSwb98+hg8fntJmNptp2LBh8n9Fz4CbN29itVrx8/O74/Px8fHEx8enPI6MjATAarVitVr/RfUPxq0anKEWcW5hYWG0b9+eo0eP4ubmxueff06vJ0tjWdwZEmMJy1ODlmdeIg4Txfw8GdO6IjVL+AGGxlcupr8xYi+NGbGXxozYy5nGjD01ODQ4Xb58maSkJAICAlK1BwQEcPTo0QydY+jQoQQFBdGwYcM7Pj9mzBhGjx6dpn3Dhg14eXnZX3Qm2bhxo6NLECe2Z88eJk6cyM2bNylQoABDhw6lsucFTAvewmRY2WpUpc+VASTgSv3CNp4NjuLy4V2sPezoysVZ6G+M2EtjRuylMSP2coYxc/PmzQz3zdbrd8aOHcvChQvZsmULHh4ed+wzfPhwBg8enPI4MjIy5b4oX1/frCr1rqxWKxs3bqRRo0a4umpraLmz48ePc/PmTerUqUOfPn14rqoPbv/3IibDyoak6gywvkJQgbyMbV2R6sXzO7pccSL6GyP20pgRe2nMiL2caczcWo2WEQ4NTgULFsRisXDhwoVU7RcuXCAwMPCex44fP56xY8fy/fffU7ly5bv2c3d3x93dPU27q6urw39R/+Rs9YhzGTRoEIUKFaJt27b8tvgDXFdOwWQksiapFoMS+9OjXhkGNyqHp5vF0aWKk9LfGLGXxozYS2NG7OUMY8ae13fornpubm5Ur16dTZs2pbTZbDY2bdpEaGjoXY/76KOPeO+991i3bh01atTIilJFstRvv/1G8+bNuXHjBgAmk4muXbti/XUZ1U5NwmwksiKpDhPzDmVh38d5q1mIQpOIiIhIJnL4duSDBw9m2rRpzJkzhyNHjvCf//yHmJiYlF32unXrlmrziHHjxjFixAhmzpxJiRIliIiIICIigujoaEddgsgDtXTpUmrVqsWaNWt44403Utp/WTMZn+9exoLB0qTHCQv9iDUDn9DSPBEREZEs4PB7nDp27MilS5d45513iIiIoGrVqqxbty5lw4gzZ85gNv+d7yZPnkxCQgLt2rVLdZ6RI0cyatSorCxd5IFKSkri7bffZuzYsQA0aNCA999/nyvR8ayfO47nLnyC2WSwnAYU7zGFdqUD0jmjiIiIiDwoDg9OAAMGDGDAgAF3fG7Lli2pHp86dSrzCxLJYlevXqVTp05s2LABgNdee40xY8aw4cglflk+nGHGDDDBTwHtsQU8S9Vid95+X0REREQyh1MEJ5Hc7MiRIzRr1oyTJ0/i5eXFjBkzaPhsawYu+oXAw9MZ4Zr8pc+XKr1IleYf8Od33zm4YhEREZHcx+H3OInkdgUKFMBqtVKqVCl27NhBngqP0/jTrRQ/PCUlNCXWGUKhNh+ByeTgakVERERyJ804iTiAzWZLuXfP39+f7777DlcfPz7Zep71v+1noGU5g1yXJXd+8i1c6r9xj7OJiIiISGbTjJNIFrt06RINGzZk3rx5ABiGwfHE/HScc4j1v0Uw1HXx36Gp4ShQaBIRERFxOM04iWShffv20bp1a86ePcuhQ4eo2/AZPtx4ko2HLwAGn+RdQpv4/0vu3ORDCO3v0HpFREREJJmCk0gWmT17Nn379iU+Pp6yZcvS74PJtJq6jxuxVtwsBkuK/R9Vwlcmd35mPNTs49B6RURERORvCk4imSwhIYHBgwczadIkABo3bUah5kP49KebAFQK8mZewELyHVkMmKD5Z1C9uwMrFhEREZHbKTiJZCKr1UrDhg3Ztm0bAB1eHMxvgQ0JO3MTN4uZV58qRd/rn2L+ZQGYzNDyK6jaycFVi4iIiMjttDmESCZydXWlfv36+Pj40uCV8ezO34DoeBtViuZlzYDH6Hft479CkwXaTFNoEhEREXFSmnESyQTR0dF4e3tjGAYhzV6gcGQZjnsWwM3FzOBGZekdWhSXlX3g8P+B2QXazYSQlo4uW0RERETuQsFJ5AGKj49nwIABHDhwgIWrNzDq29/Z9sdl8CxA1eB8jG9fmYf83GBJDwhbCxY36DAXyj3t6NJFRERE5B4UnEQekHPnztGuXTt2796NyWSiweAvMRWrhruLmSGNy9KrbiksSXGwsDMc2wgWd3huPpRp6OjSRURERCQdCk4iD8DWrVtp3749Fy9exM3Lh3zNXsNUrBrVi+fno3aVKV3IGxJuwsJOcGILuHjC8wuh1BOOLl1EREREMkDBSeRfMAyDL774giFDhpCYmIi7f0kKtH4Ln0JBvN6kPD1ql8BiNkF8NMzvCKd/BNc80HkxlKjr6PJFREREJIMUnET+hffee4+RI0cC4FWhPgWavsxjZYMY164yJQvmSe4UdwO+aQ9nd4O7L3ReCsVqObBqEREREbGXgpNIBiTZDPacvMrFqDj8fTyoWdIPE+BZ4QksefLhW6sd/qFtGPZ0ebqFlsBsNiUfGHsN5rWB8/vBIy90XQFFqjv0WkRERETEfgpOIulYdyic0asPE34jDoDEyIsUDgomr5crxy5GE9RnKqHli/JRu8oUL5Dn7wNjrsC8VhDxC3j6QbeVULiKQ65BRERERP4dBSeRe1h3KJz//Hc/Bsn3M0XuWcH1rXOwtnkbz9KP4uZiZkTLR+lcq/jfs0wA0Zdgbku4+BvkKQTd/g8CKjrsOkRERETk3zE7ugARZ5VkMxi9+jAGYEuI4/Kqj7i+ZSbYkog9uR+AfJ6uPH97aIqKgNnNkkOTdyD0+FahSURERCSb04yTyG0i46z8eu4Gqw+eJ/xGHNZr4Vxa8QHWS6fAbMHvqT54P9IMgItR8ew5eZXQ0gWSD77xJ8xpDlePg28R6L4aCpR23MWIiIiIyAOh4CS5WnxiEkfDozh47joHzl7n4NnrHL8Uk/J87Il9XF71Ebb4GMx58lGo5TA8gh9OdY6LUcn3PnH9THJounYK8haDHqshf4msuxgRERERyTQKTpJ1bElwegdEXwDvACheG8yWrHt5m8GJyzEcPHudg+eSQ9KR8CgSkmxp+gb7eeIVdY4NS0YBBm6Fy1Go9XBcfAqm6evv4wFXT8CcFnDjLOQvmTzTlC848y9KRERERLKEgpNkjcOrYN1QiDz/d5tvEDQdByEtMuUlI27EJc8i/RWSfj13g6j4xDT98nu5UiU4H1WK5qNqcD4qF81LAW93kmwGQesXczMhCb+GL2FycU11nAkIzOtBTZ8rMKslRJ2HAmWg+6rkaxMRERGRHEPBSTLf4VWwuBtgpG6PDE9u7zD3X4enG7HJ9yXdCkkHz13nQmR8mn4ermYqFclLlaL5qBKcHJSK5vfEZEre3CEsLAziLeDtjsVsYua0KQxYcBBuq/7WVhAf1XfDMqcZxFyEQuWh2yrwCfhX1yIiIiIizkfBSTKXLSl5pun20ASAgYGJ2NWvc9D1MWqWLoTln7vT3UWcNYkj4ZH8cu4GB89e58C565z4x31Jt1jMJsoG+FA1+O+gVMbfGxfLnTeTXLVqFV27dqVWrVp89913WCwWmlUpisVi4b1VvxIcfRB/rnORfJz1rsK4x92o+2MPuHkZAiolf09TnrRL+UREREQk+1Nwksx1ekfq5Xm3MWHgFRtBwLy67LUUpljRIgQVLgJeBcDTD5tnfsITvDhyw5UDV8zsvgAHIuKxJqU9VzE/r7+W3OWlSnA+Kgb54uWW/hC32WyMHj2ad999F4C4uDiioqLIly8fAE3Ne2niMRRTwt/XYbgWwrT1JiTEQOGq0HUFePnZ9daIiIiISPah4CSZK/pChrqVMl+glHEBzh6As3+3m4Eif/1r+FdbvIsr1119iHfNh8nLD3ffQvj6BeDhWzA5cHn5QYIfXCwAXvmT29x9wZR2Nuv69et07dqVNWvWAPDyyy8zYcIEXF3/up/pr2WGpttmzEwxl5J/KFA6+cttPfNl9B0RERERkWxIwUkyl3fG7vcZa+3IRSM/+U1R5DdFk5/o5J+Jxs8cTSFLNL5GFC6GFXeTlQCugvUq3DgBN0gVtu7I7AKe+VNmsvDy47fLBq3HruePP6/i4e7K16NeptvzHeDGqeQ+7j73WGb4l4TY5H4iIiIikqMpOEnmKl47eYe5yHDuFEBsBkRQgKlJzbHx971HDcr706RiAKWC81G60F/3JRlG8tK4m1cg9mry/9689o+fr97287Xkn603wZYIMZeS/wE2w6Dz1zH8ccFGsbwmVnR0o1rcdJg53b7rizqfvByxZL1/8y6JiIiIiJNTcJLMZbYkbzm+uBvJe9H9HZ5sf/042to1VWgCaFk1iJZVi6Q+l8kE7t7J//IXz3gN1th/hKrkYGWOvcrc4ocZMW01M14KpaBLTOo+8ZEZP38GlyOKiIiISPal4CSZL6RF8pbjt32PUwQFGG3tynpbzTSH+Pt4PLjXd/WEvEW4kujBriNnaNasDQCVH4X/e+GTOx+TZIWwdbC4S/rnz+ByRBERERHJvhScJGuEtIDyzUg6tZ3R83/g95t52GMrn2amKeVLZUs+2B3qDh48SOvWrTl37hz/+9//CA0NvfcBFlco/8w9lxmCKfn54rUfaK0iIiIi4nzu/IU2IpnBbMFS6nFqt3qJ3bYQjDuEJoCRzUMy9H1OGTV//nxCQ0M5efIkwcHB5MmTJ8P10nTcbdWR+nHTscn9RERERCRHU3CSLNf04cJM7lKNwLypl+MF5vVgcpdqNH248AN5ncTERIYMGULnzp2JjY2ladOm7N27l8qVK2f8JLeWGfreVpNvUHJ7SIsHUquIiIiIODct1ROHaPpwYRqFBLLn5FUuRsXh75O8PO9BzTRdunSJjh07snnzZgDefPNN3n33XSyW+5gd+muZIad3JG8E4R2QvDxPM00iIiIiuYaCkziMxWwitHSBTDn3ggUL2Lx5M97e3sydO5fWrVv/uxOaLdpyXERERCQXU3CSHOnll1/mzJkz9OrViwoVKji6HBERERHJ5nSPk+QICQkJfPjhh8TExABgMpkYP368QpOIiIiIPBCacZJsLyIigvbt2/Pjjz9y6NAh5s+f7+iSRERERCSHUXCSbG3Xrl20bduW8+fPkzdvXjp37uzokkREREQkB9JSPcm2pk6dyuOPP8758+cJCQlh7969NGvWzNFliYiIiEgOpOAk2U58fDx9+vThpZdewmq10rZtW3bt2kWZMmUcXZqIiIiI5FAKTpLtXLlyhdWrV2MymRgzZgxLlizBx8fH0WWJiIiISA6me5wk2wkKCmLp0qXExMTQpEkTR5cjIiIiIrmAgpM4PcMw+PLLLwkKCqJt27YA1K1b18FViYiIiEhuouAkTi02Npa+ffsyd+5c8uTJw6OPPkqxYsUcXZaIiIiI5DIKTuK0Tp8+TZs2bdi/fz8Wi4X33nuP4OBgR5clIiIiIrmQgpM4pU2bNtGxY0euXLlCwYIFWbx4MU8++aSjyxIRERGRXEq76onTGT9+PI0bN+bKlStUr16dffv2KTSJiIiIiEMpOInTCQ8Px2az0aNHD7Zt26Z7mkRERETE4bRUT5zOuHHjqF27Nm3atMFkMjm6HBERERERzTiJ43333Xe0aNGChIQEAFxcXGjbtq1Ck4iIiIg4DQUncRibzcYHH3xAs2bNWL16NZ9//rmjSxIRERERuSMt1ROHiIyMpHv37qxcuRKAl156iZdfftmxRYmIiIiI3IWCk2S5sLAwWrVqxdGjR3Fzc2PSpEn07t3b0WWJiIiIiNyVgpNkqY0bN9K2bVuioqIoUqQIy5Yto1atWo4uS0RERETknnSPk2SpYsWKYTKZqFevHvv27VNoEhEREZFsQTNOkukSExNxcUkeauXKleN///sfFStWxNXV1cGViYiIiIhkjGacJFP99ttvVKpUiR9++CGlrWrVqgpNIiIiIpKtKDhJplm6dCm1atXi6NGjvPHGGxiG4eiSRERERETui4KTPHBJSUkMGzaM9u3bExMTQ4MGDfjuu+/0hbYiIiIikm3pHid5oK5cucLzzz/Phg0bABgyZAhjx45NucdJRERERCQ70qdZeWAuXbpEzZo1OXXqFF5eXsyYMYPnnnvO0WWJiIiIiPxrCk7ywBQsWJC6detiNptZsWIFlStXdnRJIiIiIiIPhIKT/CtWq5X4+Hi8vb0xmUx8/fXXxMXF4efn5+jSREREREQeGG0OIfft4sWLNGrUiOeffx6bzQaAl5eXQpOIiIiI5DiacZL7snfvXtq0acO5c+fw9vbm6NGjhISEOLosEREREZFMoRknsdusWbOoV68e586do2zZsuzevVuhSURERERyNAUnybCEhAT69+/PCy+8QHx8PC1atGDPnj0KTSIiIiKS4yk4SYZ169aNr776CpPJxLvvvsuKFSvImzevo8sSEREREcl0Ck6SYUOGDCEgIIDVq1czYsQIzGYNHxERERHJHbQ5hNyVYRgcO3aMMmXKAPDoo49y8uRJPD09HVyZiIiIiEjW0pSB3FFcXBx9+vShcuXK7N+/P6VdoUlEREREciMFJ0nj3Llz1K9fnxkzZpCQkMDevXsdXZKIiIiIiENpqZ6ksnXrVtq3b8/Fixfx8/Nj4cKFNGrUyNFliYiIiIg4lGacBEi+n+mLL77gqaee4uLFi1SpUoWffvpJoUlEREREBAUn+cvy5ct55ZVXSExMpHPnzuzYsYOSJUs6uiwREREREaegpXoCQKtWrWjRogVPPvkkAwcOxGQyObokERERERGnoeCUi+3YsYNq1arh4eGBxWJh5cqVCkwiIiIiInegpXq5kGEYjB8/nnr16tGvXz8MwwBQaBIRERERuQvNOOUyMTEx9OrVi0WLFqW0JSUl4eKioSAiIiIicjf6tJyLHD9+nNatW/Prr7/i4uLC559/Tt++fTXTJCIiIiKSDgWnXGLdunV06tSJ69evExgYyJIlS6hbt66jyxIRERERyRYUnHKB6OhounXrxvXr1wkNDWXp0qUEBQU5uiwRERERkWxDm0PkAt7e3nzzzTf07duXzZs3KzSJiIiIiNhJM045VFhYGH/++ScNGjQAoFGjRjRq1MjBVYmIiIiIZE+accqBVq1aRc2aNWnTpg1//PGHo8sREREREcn2FJxyEJvNxjvvvEPLli2JjIykcuXK+Pr6OrosEREREZFsT0v1cojr16/TpUsXvv32WwBeeeUVxo8fj6urq4MrExERERHJ/hSccoBDhw7RunVrjh07hoeHB1OnTqVr166OLktEREREJMdQcMoBpk2bxrFjxyhWrBgrVqygWrVqji5JRERERCRHUXDKAT766CNcXFwYPnw4BQsWdHQ5IiIiIiI5jjaHyIauXLnC22+/TWJiIgDu7u5MmDBBoUlEREREJJNoximbOXDgAK1bt+bUqVMYhsEHH3zg6JJERERERHI8zThlI9988w21a9fm1KlTlCpVio4dOzq6JBERERGRXEHBKRuwWq0MGjSILl26EBsbS9OmTdm7dy+VK1d2dGkiIiIiIrmCgpOTu3jxIo0aNWLixIkAvPnmm6xZswY/Pz/HFiYiIiIikovoHicnd+HCBfbu3Yu3tzdz5syhTZs2ji5JRERERCTXUXBycpUqVWLRokWUKlWKkJAQR5cjIiIiIpIraamek0lISOCVV15hx44dKW3PPvusQpOIiIiIiAMpODmR8PBwGjRowBdffEH79u25efOmo0sSERERERGcJDhNmjSJEiVK4OHhQa1atdizZ889+y9ZsoTy5cvj4eFBpUqVWLt2bRZVmnl27txJ9erV2b59O3nz5mXq1Kl4eXk5uiwREREREcEJgtOiRYsYPHgwI0eOZP/+/VSpUoUmTZpw8eLFO/bfsWMHnTp1olevXvz888+0atWKVq1acejQoSyu/MEwDIN169bRsGFDwsPDCQkJYe/evTRr1szRpYmIiIiIyF8cHpw++eQT+vTpQ8+ePQkJCWHKlCl4eXkxc+bMO/b/7LPPaNq0Ka+//joVKlTgvffeo1q1anz55ZdZXPm/Z7Va6du3L1OmTMFqtdKuXTt2795NmTJlHF2aiIiIiIj8g0N31UtISGDfvn0MHz48pc1sNtOwYUN27tx5x2N27tzJ4MGDU7U1adKElStX3rF/fHw88fHxKY8jIyOB5NBitVr/5RX8O4ZhcO3aNcxmM6NGjWLo0KGYTCaH1yXO69bY0BiRjNKYEXtpzIi9NGbEXs40ZuypwaHB6fLlyyQlJREQEJCqPSAggKNHj97xmIiIiDv2j4iIuGP/MWPGMHr06DTtGzZscIp7iDp27Ej16tWpVKkS3333naPLkWxi48aNji5BshmNGbGXxozYS2NG7OUMY8aezdhy/Pc4DR8+PNUMVWRkJMHBwTRu3BhfX18HVpbMarWyceNGGjVqhKurq6PLESen8SL20pgRe2nMiL00ZsRezjRmbq1GywiHBqeCBQtisVi4cOFCqvYLFy4QGBh4x2MCAwPt6u/u7o67u3uadldXV4f/ov7J2eoR56bxIvbSmBF7acyIvTRmxF7OMGbseX2Hbg7h5uZG9erV2bRpU0qbzWZj06ZNhIaG3vGY0NDQVP0heZrvbv1FRERERET+LYcv1Rs8eDDdu3enRo0a1KxZk4kTJxITE0PPnj0B6NatG0WKFGHMmDEADBw4kPr16zNhwgSaNWvGwoUL+emnn5g6daojL0NERERERHIwhwenjh07cunSJd555x0iIiKoWrUq69atS9kA4syZM5jNf0+M1a5dm/nz5/P222/z5ptvUqZMGVauXMnDDz/sqEsQEREREZEczuHBCWDAgAEMGDDgjs9t2bIlTVv79u1p3759JlclIiIiIiKSzOFfgCsiIiIiIuLsFJxERERERETSoeAkIiIiIiKSDgUnERERERGRdCg4iYiIiIiIpEPBSUREREREJB0KTiIiIiIiIulQcBIREREREUmHgpOIiIiIiEg6FJxERERERETSoeAkIiIiIiKSDgUnERERERGRdCg4iYiIiIiIpEPBSUREREREJB0KTiIiIiIiIulQcBIREREREUmHgpOIiIiIiEg6FJxERERERETSoeAkIiIiIiKSDgUnERERERGRdLg4uoCsZhgGAJGRkQ6uJJnVauXmzZtERkbi6urq6HLEyWm8iL00ZsReGjNiL40ZsZczjZlbmeBWRriXXBecoqKiAAgODnZwJSIiIiIi4gyioqLImzfvPfuYjIzEqxzEZrNx/vx5fHx8MJlMji6HyMhIgoODOXv2LL6+vo4uR5ycxovYS2NG7KUxI/bSmBF7OdOYMQyDqKgogoKCMJvvfRdTrptxMpvNFC1a1NFlpOHr6+vwgSPZh8aL2EtjRuylMSP20pgReznLmElvpukWbQ4hIiIiIiKSDgUnERERERGRdCg4OZi7uzsjR47E3d3d0aVINqDxIvbSmBF7acyIvTRmxF7Zdczkus0hRERERERE7KUZJxERERERkXQoOImIiIiIiKRDwUlERERERCQdCk4iIiIiIiLpUHDKZJMmTaJEiRJ4eHhQq1Yt9uzZc8/+S5YsoXz58nh4eFCpUiXWrl2bRZWKs7BnzEybNo169eqRP39+8ufPT8OGDdMdY5Lz2Pt35paFCxdiMplo1apV5hYoTsfeMXP9+nX69+9P4cKFcXd3p2zZsvr/p1zG3jEzceJEypUrh6enJ8HBwQwaNIi4uLgsqlYcbevWrTRv3pygoCBMJhMrV65M95gtW7ZQrVo13N3deeihh5g9e3am12kvBadMtGjRIgYPHszIkSPZv38/VapUoUmTJly8ePGO/Xfs2EGnTp3o1asXP//8M61ataJVq1YcOnQoiysXR7F3zGzZsoVOnTqxefNmdu7cSXBwMI0bN+bPP//M4srFUewdM7ecOnWK1157jXr16mVRpeIs7B0zCQkJNGrUiFOnTrF06VLCwsKYNm0aRYoUyeLKxVHsHTPz589n2LBhjBw5kiNHjjBjxgwWLVrEm2++mcWVi6PExMRQpUoVJk2alKH+J0+epFmzZjz55JMcOHCAV199ld69e7N+/fpMrtROhmSamjVrGv379095nJSUZAQFBRljxoy5Y/8OHToYzZo1S9VWq1Yt46WXXsrUOsV52DtmbpeYmGj4+PgYc+bMyawSxcncz5hJTEw0ateubUyfPt3o3r270bJlyyyoVJyFvWNm8uTJRqlSpYyEhISsKlGcjL1jpn///kaDBg1StQ0ePNioU6dOptYpzgkwVqxYcc8+b7zxhlGxYsVUbR07djSaNGmSiZXZTzNOmSQhIYF9+/bRsGHDlDaz2UzDhg3ZuXPnHY/ZuXNnqv4ATZo0uWt/yVnuZ8zc7ubNm1itVvz8/DKrTHEi9ztm3n33Xfz9/enVq1dWlClO5H7GzKpVqwgNDaV///4EBATw8MMP8+GHH5KUlJRVZYsD3c+YqV27Nvv27UtZznfixAnWrl3LM888kyU1S/aTXT4Duzi6gJzq8uXLJCUlERAQkKo9ICCAo0eP3vGYiIiIO/aPiIjItDrFedzPmLnd0KFDCQoKSvPHR3Km+xkzP/74IzNmzODAgQNZUKE4m/sZMydOnOCHH36gc+fOrF27lmPHjtGvXz+sVisjR47MirLFge5nzDz//PNcvnyZunXrYhgGiYmJ9O3bV0v15K7u9hk4MjKS2NhYPD09HVRZappxEskhxo4dy8KFC1mxYgUeHh6OLkecUFRUFF27dmXatGkULFjQ0eVINmGz2fD392fq1KlUr16djh078tZbbzFlyhRHlyZOasuWLXz44Yd89dVX7N+/n+XLl/Ptt9/y3nvvObo0kX9FM06ZpGDBglgsFi5cuJCq/cKFCwQGBt7xmMDAQLv6S85yP2PmlvHjxzN27Fi+//57KleunJllihOxd8wcP36cU6dO0bx585Q2m80GgIuLC2FhYZQuXTpzixaHup+/M4ULF8bV1RWLxZLSVqFCBSIiIkhISMDNzS1TaxbHup8xM2LECLp27Urv3r0BqFSpEjExMbz44ou89dZbmM367/aS2t0+A/v6+jrNbBNoxinTuLm5Ub16dTZt2pTSZrPZ2LRpE6GhoXc8JjQ0NFV/gI0bN961v+Qs9zNmAD766CPee+891q1bR40aNbKiVHES9o6Z8uXL8+uvv3LgwIGUfy1atEjZxSg4ODgryxcHuJ+/M3Xq1OHYsWMpIRvg999/p3DhwgpNucD9jJmbN2+mCUe3grdhGJlXrGRb2eYzsKN3p8jJFi5caLi7uxuzZ882Dh8+bLz44otGvnz5jIiICMMwDKNr167GsGHDUvpv377dcHFxMcaPH28cOXLEGDlypOHq6mr8+uuvjroEyWL2jpmxY8cabm5uxtKlS43w8PCUf1FRUY66BMli9o6Z22lXvdzH3jFz5swZw8fHxxgwYIARFhZmrFmzxvD39zfef/99R12CZDF7x8zIkSMNHx8fY8GCBcaJEyeMDRs2GKVLlzY6dOjgqEuQLBYVFWX8/PPPxs8//2wAxieffGL8/PPPxunTpw3DMIxhw4YZXbt2Tel/4sQJw8vLy3j99deNI0eOGJMmTTIsFouxbt06R13CHSk4ZbIvvvjCKFasmOHm5mbUrFnT2LVrV8pz9evXN7p3756q/+LFi42yZcsabm5uRsWKFY1vv/02iysWR7NnzBQvXtwA0vwbOXJk1hcuDmPv35l/UnDKnewdMzt27DBq1apluLu7G6VKlTI++OADIzExMYurFkeyZ8xYrVZj1KhRRunSpQ0PDw8jODjY6Nevn3Ht2rWsL1wcYvPmzXf8fHJrnHTv3t2oX79+mmOqVq1quLm5GaVKlTJmzZqV5XWnx2QYmjMVERERERG5F93jJCIiIiIikg4FJxERERERkXQoOImIiIiIiKRDwUlERERERCQdCk4iIiIiIiLpUHASERERERFJh4KTiIiIiIhIOhScRERERERE0qHgJCIiTq1Hjx60atUq5fETTzzBq6++muV1bNmyBZPJxPXr17P8tU+dOoXJZOLAgQP/6jy3v5d3cvv7W6JECSZOnJjy2GQysXLlyn9Vh4hIdqTgJCLiJHr06IHJZKJv375pnuvfvz8mk4kePXpkfWFOZvny5bz33nsZ6uvIsJNdpff+hoeH8/TTTwMPLtCJiGQHCk4iIk4kODiYhQsXEhsbm9IWFxfH/PnzKVasmAMr+3cSEhIe2Ln8/Pzw8fF5YOdzNKvV6ugSUknv/Q0MDMTd3T0LKxIRcQ4KTiIiTqRatWoEBwezfPnylLbly5dTrFgxHnnkkVR9bTYbY8aMoWTJknh6elKlShWWLl2a8nxSUhK9evVKeb5cuXJ89tlnqc5xa+nW+PHjKVy4MAUKFKB///73/DA/atQoqlatytdff01wcDBeXl506NCBGzdupDnvBx98QFBQEOXKlQPg7NmzdOjQgXz58uHn50fLli05depUqpoHDx5Mvnz5KFCgAG+88QaGYaR6/duXksXHxzN06FCCg4Nxd3fnoYceYsaMGZw6dYonn3wSgPz586easUvvvQNYu3YtZcuWxdPTkyeffDJVnXdjMpmYPHkyTz/9NJ6enpQqVSrVeW/N0CxatIj69evj4eHBN998g81m491336Vo0aK4u7tTtWpV1q1bl+b8R48epXbt2nh4ePDwww/zv//9L9V7l97v+5bRo0dTqFAhfH196du3b6pgm95SyH8u1StZsiQAjzzyCCaTiSeeeIKtW7fi6upKREREquNeffVV6tWrl+57KCLirBScRESczAsvvMCsWbNSHs+cOZOePXum6TdmzBjmzp3LlClT+O233xg0aBBdunRJ+TBts9koWrQoS5Ys4fDhw7zzzju8+eabLF68ONV5Nm/ezPHjx9m8eTNz5sxh9uzZzJ49+541Hjt2jMWLF7N69WrWrVvHzz//TL9+/VL12bRpE2FhYWzcuJE1a9ZgtVpp0qQJPj4+bNu2je3bt+Pt7U3Tpk1TPrhPmDCB2bNnM3PmTH788UeuXr3KihUr7llLt27dWLBgAZ9//jlHjhzh66+/xtvbm+DgYJYtWwZAWFgY4eHhKUEivffu7NmztGnThubNm3PgwAF69+7NsGHD7lnHLSNGjKBt27YcPHiQzp0789xzz3HkyJFUfYYNG8bAgQM5cuQITZo04bPPPmPChAmMHz+eX375hSZNmtCiRQv++OOPVMe9/vrrDBkyhJ9//pnQ0FCaN2/OlStXgIz/vjdt2sSRI0fYsmULCxYsYPny5YwePTpD13a7PXv2APD9998THh7O8uXLefzxxylVqhTz5s1L6We1Wvnmm2944YUX7ut1REScgiEiIk6he/fuRsuWLY2LFy8a7u7uxqlTp4xTp04ZHh4exqVLl4yWLVsa3bt3NwzDMOLi4gwvLy9jx44dqc7Rq1cvo1OnTnd9jf79+xtt27ZN9ZrFixc3EhMTU9rat29vdOzY8a7nGDlypGGxWIxz586ltH333XeG2Ww2wsPDU84bEBBgxMfHp/SZN2+eUa5cOcNms6W0xcfHG56ensb69esNwzCMwoULGx999FHK81ar1ShatKjRsmXLlLb69esbAwcONAzDMMLCwgzA2Lhx4x1r3bx5swEY165dS2nLyHs3fPhwIyQkJNXzQ4cOTXOu2wFG3759U7XVqlXL+M9//mMYhmGcPHnSAIyJEyem6hMUFGR88MEHqdoeffRRo1+/fqmOGzt2bMrzt96bcePG3bWeO/2+/fz8jJiYmJS2yZMnG97e3kZSUpJhGKnfX8MwjOLFixuffvppqmtcsWJFqrp+/vnnVK87btw4o0KFCimPly1bZnh7exvR0dF3rVVExNm5OC6yiYjInRQqVIhmzZoxe/ZsDMOgWbNmFCxYMFWfY8eOcfPmTRo1apSqPSEhIdWSvkmTJjFz5kzOnDlDbGwsCQkJVK1aNdUxFStWxGKxpDwuXLgwv/766z1rLFasGEWKFEl5HBoais1mIywsjMDAQAAqVaqEm5tbSp+DBw9y7NixNPfPxMXFcfz4cW7cuEF4eDi1atVKec7FxYUaNWqkWa53y4EDB7BYLNSvX/+e9f5TRt67I0eOpKrj1jVmxO39QkND02yeUKNGjZSfIyMjOX/+PHXq1EnVp06dOhw8ePCu57713vxzNisjv+8qVarg5eWV6pzR0dGcPXuW4sWLZ+ga09OjRw/efvttdu3axWOPPcbs2bPp0KEDefLkeSDnFxFxBAUnEREn9MILLzBgwAAg+cPw7aKjowH49ttvUwUYIOXG/YULF/Laa68xYcIEQkND8fHx4eOPP2b37t2p+ru6uqZ6bDKZsNls//oabv+QHB0dTfXq1fnmm2/S9C1UqNB9vYanp6fdx2TkvctsmREgMvr7zgr+/v40b96cWbNmUbJkSb777ju2bNmS5XWIiDxICk4iIk7o1n0/JpOJJk2apHk+JCQEd3d3zpw5c9fZlu3bt1O7du1U9x4dP378gdR35swZzp8/T1BQEAC7du3CbDanbAJxJ9WqVWPRokX4+/vj6+t7xz6FCxdm9+7dPP744wAkJiayb98+qlWrdsf+lSpVwmaz8b///Y+GDRumef7WjFdSUlJKW0beuwoVKrBq1apUbbt27brrtd3er1u3bqke376xxz/5+voSFBTE9u3bU9Wzfft2atasmebct783twJ2Rn/fBw8eJDY2NiV07tq1K+WeMHvd6f29pXfv3nTq1ImiRYtSunTpNDNqIiLZjTaHEBFxQhaLhSNHjnD48OFUy+hu8fHx4bXXXmPQoEHMmTOH48ePs3//fr744gvmzJkDQJkyZfjpp59Yv349v//+OyNGjGDv3r0PpD4PDw+6d+/OwYMH2bZtG6+88godOnRIWaZ3J507d6ZgwYK0bNmSbdu2cfLkSbZs2cIrr7zCuXPnABg4cCBjx45l5cqVHD16lH79+t3zO5hKlChB9+7deeGFF1i5cmXKOW9tiFC8eHFMJhNr1qzh0qVLREdHZ+i969u3L3/88Qevv/46YWFhzJ8/P90NM25ZsmQJM2fO5Pfff2fkyJHs2bMnJdzczeuvv864ceNYtGgRYWFhDBs2jAMHDjBw4MBU/SZNmsSKFSs4evQo/fv359q1aykbLmT0952QkECvXr04fPgwa9euZeTIkQwYMACz2f6PBP7+/nh6erJu3TouXLiQamfFJk2a4Ovry/vvv3/HzU1ERLIbBScRESfl6+t715kZgPfee48RI0YwZswYKlSoQNOmTfn2229Ttoh+6aWXaNOmDR07dqRWrVpcuXIlzc539+uhhx6iTZs2PPPMMzRu3JjKlSvz1Vdf3fMYLy8vtm7dSrFixWjTpg0VKlSgV69exMXFpVznkCFD6Nq1K927d09Zbta6det7nnfy5Mm0a9eOfv36Ub58efr06UNMTAwARYoUYfTo0QwbNoyAgICUAJPee1esWDGWLVvGypUrqVKlClOmTOHDDz/M0HszevRoFi5cSOXKlZk7dy4LFiwgJCTknse88sorDB48mCFDhlCpUiXWrVvHqlWrKFOmTKp+Y8eOZezYsVSpUoUff/yRVatWpdz/ltHf91NPPUWZMmV4/PHH6dixIy1atGDUqFEZurbbubi48Pnnn/P1118TFBREy5YtU54zm8306NGDpKSkVDNwIiLZlcm42x23IiIidzBq1ChWrlyZZsMDSb4/bMWKFbRq1crRpTiFXr16cenSpTTLHkVEsiPd4yQiIiIP1I0bN/j111+ZP3++QpOI5BgKTiIiIvJAtWzZkj179tC3b980276LiGRXWqonIiIiIiKSDm0OISIiIiIikg4FJxERERERkXQoOImIiIiIiKRDwUlERERERCQdCk4iIiIiIiLpUHASERERERFJh4KTiIiIiIhIOhScRERERERE0vH/1JfVMYPj6iMAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 1000x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Calibration plot\n",
    "plt.figure(figsize=(10, 8))\n",
    "\n",
    "for name, clf in text_based_classifiers.items():\n",
    "\n",
    "    # Get calibrated probabilities\n",
    "    prob_pos = clf.predict_proba(x_test_vectorized)[:, 1]\n",
    "    fraction_of_positives, mean_predicted_value = calibration_curve(y_test, prob_pos, n_bins=10)\n",
    "\n",
    "    # Plot calibration curve\n",
    "    plt.plot(mean_predicted_value, fraction_of_positives, marker='o', label=name)\n",
    "\n",
    "# Plot reference line\n",
    "plt.plot([0, 1], [0, 1], linestyle='--', color='black', label='Perfectly calibrated')\n",
    "\n",
    "# Add labels and legend\n",
    "plt.xlabel('Mean predicted probability')\n",
    "plt.ylabel('Fraction of positives')\n",
    "plt.title('Calibration Plot')\n",
    "plt.legend()\n",
    "plt.grid(True)\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "CDC-BRFSS",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}