project_bank_client.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "project_bank_client.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "mount_file_id": "1oAn50lfTKYwmvlZ2VhAVp7nzpH6kVcjU",
      "authorship_tag": "ABX9TyOQe5yA/QU3yU2AWad+RMfo",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/KonstantinSV/project_bank_client/blob/main/project_bank_client.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gW6dKKz4J6q7"
      },
      "source": [
        " **Проект: Cнижение клиентооборота**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DwvV33sTKbEf"
      },
      "source": [
        "Цель - обучить модель предсказывающую уйдёт клиент из банка в ближайшее время или нет."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X5c1eZ_eKrc2"
      },
      "source": [
        "Данные для обучения:\r\n",
        "\r\n",
        "Датасет на Kaggle - https://www.kaggle.com/barelydedicated/bank-customer-churn-modeling"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9uCCIi9WQ_2y"
      },
      "source": [
        "# Предварительная обработка данных и исследование задачи"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Yj2HTDI3QY2f"
      },
      "source": [
        "import pandas as pd\r\n",
        "import matplotlib.pyplot as plt\r\n",
        "import numpy as np\r\n",
        "import sklearn as sk\r\n",
        "from sklearn.tree import DecisionTreeClassifier\r\n",
        "from sklearn.ensemble import RandomForestClassifier\r\n",
        "from sklearn.linear_model import LogisticRegression\r\n",
        "from sklearn.model_selection import train_test_split\r\n",
        "from sklearn.preprocessing import StandardScaler\r\n",
        "from sklearn.metrics import f1_score\r\n",
        "from sklearn.utils import shuffle\r\n",
        "from sklearn.metrics import roc_curve\r\n",
        "from sklearn.metrics import roc_auc_score"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xn-o_m5kRJeA"
      },
      "source": [
        "df=pd.read_csv('/content/drive/MyDrive/Colab Notebooks/projects/project_bank_client/Churn_Modelling.csv')"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sKmjGcx-MDAc",
        "outputId": "731497f7-84eb-4593-be81-2d346cbf385b"
      },
      "source": [
        "# Размеры DataFrame\r\n",
        "df.shape"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(10000, 14)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 426
        },
        "id": "i56fJLI3Od8A",
        "outputId": "e452a535-a3fb-4488-c519-c16886f636f0"
      },
      "source": [
        "df"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>RowNumber</th>\n",
              "      <th>CustomerId</th>\n",
              "      <th>Surname</th>\n",
              "      <th>CreditScore</th>\n",
              "      <th>Geography</th>\n",
              "      <th>Gender</th>\n",
              "      <th>Age</th>\n",
              "      <th>Tenure</th>\n",
              "      <th>Balance</th>\n",
              "      <th>NumOfProducts</th>\n",
              "      <th>HasCrCard</th>\n",
              "      <th>IsActiveMember</th>\n",
              "      <th>EstimatedSalary</th>\n",
              "      <th>Exited</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>15634602</td>\n",
              "      <td>Hargrave</td>\n",
              "      <td>619</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>2</td>\n",
              "      <td>0.00</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>101348.88</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2</td>\n",
              "      <td>15647311</td>\n",
              "      <td>Hill</td>\n",
              "      <td>608</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>41</td>\n",
              "      <td>1</td>\n",
              "      <td>83807.86</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>112542.58</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>3</td>\n",
              "      <td>15619304</td>\n",
              "      <td>Onio</td>\n",
              "      <td>502</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>8</td>\n",
              "      <td>159660.80</td>\n",
              "      <td>3</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>113931.57</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>4</td>\n",
              "      <td>15701354</td>\n",
              "      <td>Boni</td>\n",
              "      <td>699</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>39</td>\n",
              "      <td>1</td>\n",
              "      <td>0.00</td>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>93826.63</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>5</td>\n",
              "      <td>15737888</td>\n",
              "      <td>Mitchell</td>\n",
              "      <td>850</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>43</td>\n",
              "      <td>2</td>\n",
              "      <td>125510.82</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>79084.10</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9995</th>\n",
              "      <td>9996</td>\n",
              "      <td>15606229</td>\n",
              "      <td>Obijiaku</td>\n",
              "      <td>771</td>\n",
              "      <td>France</td>\n",
              "      <td>Male</td>\n",
              "      <td>39</td>\n",
              "      <td>5</td>\n",
              "      <td>0.00</td>\n",
              "      <td>2</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>96270.64</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9996</th>\n",
              "      <td>9997</td>\n",
              "      <td>15569892</td>\n",
              "      <td>Johnstone</td>\n",
              "      <td>516</td>\n",
              "      <td>France</td>\n",
              "      <td>Male</td>\n",
              "      <td>35</td>\n",
              "      <td>10</td>\n",
              "      <td>57369.61</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>101699.77</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9997</th>\n",
              "      <td>9998</td>\n",
              "      <td>15584532</td>\n",
              "      <td>Liu</td>\n",
              "      <td>709</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>36</td>\n",
              "      <td>7</td>\n",
              "      <td>0.00</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>42085.58</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9998</th>\n",
              "      <td>9999</td>\n",
              "      <td>15682355</td>\n",
              "      <td>Sabbatini</td>\n",
              "      <td>772</td>\n",
              "      <td>Germany</td>\n",
              "      <td>Male</td>\n",
              "      <td>42</td>\n",
              "      <td>3</td>\n",
              "      <td>75075.31</td>\n",
              "      <td>2</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>92888.52</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9999</th>\n",
              "      <td>10000</td>\n",
              "      <td>15628319</td>\n",
              "      <td>Walker</td>\n",
              "      <td>792</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>28</td>\n",
              "      <td>4</td>\n",
              "      <td>130142.79</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>38190.78</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>10000 rows × 14 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "      RowNumber  CustomerId    Surname  ...  IsActiveMember EstimatedSalary Exited\n",
              "0             1    15634602   Hargrave  ...               1       101348.88      1\n",
              "1             2    15647311       Hill  ...               1       112542.58      0\n",
              "2             3    15619304       Onio  ...               0       113931.57      1\n",
              "3             4    15701354       Boni  ...               0        93826.63      0\n",
              "4             5    15737888   Mitchell  ...               1        79084.10      0\n",
              "...         ...         ...        ...  ...             ...             ...    ...\n",
              "9995       9996    15606229   Obijiaku  ...               0        96270.64      0\n",
              "9996       9997    15569892  Johnstone  ...               1       101699.77      0\n",
              "9997       9998    15584532        Liu  ...               1        42085.58      1\n",
              "9998       9999    15682355  Sabbatini  ...               0        92888.52      1\n",
              "9999      10000    15628319     Walker  ...               0        38190.78      0\n",
              "\n",
              "[10000 rows x 14 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zfqtf7wsRKAI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f91e8f0c-91ee-47f9-b22b-e3861296759e"
      },
      "source": [
        "df.info()"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "RangeIndex: 10000 entries, 0 to 9999\n",
            "Data columns (total 14 columns):\n",
            " #   Column           Non-Null Count  Dtype  \n",
            "---  ------           --------------  -----  \n",
            " 0   RowNumber        10000 non-null  int64  \n",
            " 1   CustomerId       10000 non-null  int64  \n",
            " 2   Surname          10000 non-null  object \n",
            " 3   CreditScore      10000 non-null  int64  \n",
            " 4   Geography        10000 non-null  object \n",
            " 5   Gender           10000 non-null  object \n",
            " 6   Age              10000 non-null  int64  \n",
            " 7   Tenure           10000 non-null  int64  \n",
            " 8   Balance          10000 non-null  float64\n",
            " 9   NumOfProducts    10000 non-null  int64  \n",
            " 10  HasCrCard        10000 non-null  int64  \n",
            " 11  IsActiveMember   10000 non-null  int64  \n",
            " 12  EstimatedSalary  10000 non-null  float64\n",
            " 13  Exited           10000 non-null  int64  \n",
            "dtypes: float64(2), int64(9), object(3)\n",
            "memory usage: 1.1+ MB\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_h43uCoQRN04"
      },
      "source": [
        "В данном датасете 10000 объектов, 13 признаков и один целевой."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "opW5AwRBRF_l",
        "outputId": "23ecc15e-5c88-48d2-8acf-83e4fc0dda7a"
      },
      "source": [
        "# Проверка df на наличие пропущенных значений.\r\n",
        "df.isna().sum()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "RowNumber          0\n",
              "CustomerId         0\n",
              "Surname            0\n",
              "CreditScore        0\n",
              "Geography          0\n",
              "Gender             0\n",
              "Age                0\n",
              "Tenure             0\n",
              "Balance            0\n",
              "NumOfProducts      0\n",
              "HasCrCard          0\n",
              "IsActiveMember     0\n",
              "EstimatedSalary    0\n",
              "Exited             0\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_o8prN-gdFul"
      },
      "source": [
        "Проверка показала, что в df нет пропущенных значений."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KPJGzXS1RnTP"
      },
      "source": [
        "# Признаки RowNumber, EstimatedSalary, CustomerId, Surname являются неинформативными, их следует удалить.\r\n",
        "del_col=['RowNumber','EstimatedSalary','CustomerId','Surname']\r\n",
        "df=df.drop(del_col,axis=1)"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1qyWrMw0WeIi",
        "outputId": "dc07da79-fb6c-4353-8112-b6390646d9de"
      },
      "source": [
        "df.info()"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "RangeIndex: 10000 entries, 0 to 9999\n",
            "Data columns (total 10 columns):\n",
            " #   Column          Non-Null Count  Dtype  \n",
            "---  ------          --------------  -----  \n",
            " 0   CreditScore     10000 non-null  int64  \n",
            " 1   Geography       10000 non-null  object \n",
            " 2   Gender          10000 non-null  object \n",
            " 3   Age             10000 non-null  int64  \n",
            " 4   Tenure          10000 non-null  int64  \n",
            " 5   Balance         10000 non-null  float64\n",
            " 6   NumOfProducts   10000 non-null  int64  \n",
            " 7   HasCrCard       10000 non-null  int64  \n",
            " 8   IsActiveMember  10000 non-null  int64  \n",
            " 9   Exited          10000 non-null  int64  \n",
            "dtypes: float64(1), int64(7), object(2)\n",
            "memory usage: 781.4+ KB\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 406
        },
        "id": "pnpJiUnvmPcf",
        "outputId": "a9c82ad8-1eb1-4b99-81aa-56eb158e6215"
      },
      "source": [
        "df"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>CreditScore</th>\n",
              "      <th>Geography</th>\n",
              "      <th>Gender</th>\n",
              "      <th>Age</th>\n",
              "      <th>Tenure</th>\n",
              "      <th>Balance</th>\n",
              "      <th>NumOfProducts</th>\n",
              "      <th>HasCrCard</th>\n",
              "      <th>IsActiveMember</th>\n",
              "      <th>Exited</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>619</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>2</td>\n",
              "      <td>0.00</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>608</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>41</td>\n",
              "      <td>1</td>\n",
              "      <td>83807.86</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>502</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>8</td>\n",
              "      <td>159660.80</td>\n",
              "      <td>3</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>699</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>39</td>\n",
              "      <td>1</td>\n",
              "      <td>0.00</td>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>850</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>43</td>\n",
              "      <td>2</td>\n",
              "      <td>125510.82</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9995</th>\n",
              "      <td>771</td>\n",
              "      <td>France</td>\n",
              "      <td>Male</td>\n",
              "      <td>39</td>\n",
              "      <td>5</td>\n",
              "      <td>0.00</td>\n",
              "      <td>2</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9996</th>\n",
              "      <td>516</td>\n",
              "      <td>France</td>\n",
              "      <td>Male</td>\n",
              "      <td>35</td>\n",
              "      <td>10</td>\n",
              "      <td>57369.61</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9997</th>\n",
              "      <td>709</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>36</td>\n",
              "      <td>7</td>\n",
              "      <td>0.00</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9998</th>\n",
              "      <td>772</td>\n",
              "      <td>Germany</td>\n",
              "      <td>Male</td>\n",
              "      <td>42</td>\n",
              "      <td>3</td>\n",
              "      <td>75075.31</td>\n",
              "      <td>2</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9999</th>\n",
              "      <td>792</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>28</td>\n",
              "      <td>4</td>\n",
              "      <td>130142.79</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>10000 rows × 10 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "      CreditScore Geography  Gender  ...  HasCrCard  IsActiveMember  Exited\n",
              "0             619    France  Female  ...          1               1       1\n",
              "1             608     Spain  Female  ...          0               1       0\n",
              "2             502    France  Female  ...          1               0       1\n",
              "3             699    France  Female  ...          0               0       0\n",
              "4             850     Spain  Female  ...          1               1       0\n",
              "...           ...       ...     ...  ...        ...             ...     ...\n",
              "9995          771    France    Male  ...          1               0       0\n",
              "9996          516    France    Male  ...          1               1       0\n",
              "9997          709    France  Female  ...          0               1       1\n",
              "9998          772   Germany    Male  ...          1               0       1\n",
              "9999          792    France  Female  ...          1               0       0\n",
              "\n",
              "[10000 rows x 10 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8iiZp7TQRvGg",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 403
        },
        "outputId": "3b35f8e1-75ee-4e10-8e2c-f8c997b3f71f"
      },
      "source": [
        "# Гистограмма значений ушедших и оставшихся клиентов\r\n",
        "df['Exited'].value_counts().plot(kind='bar',grid=True,figsize=(10,6),title='0 - оставшиеся, 1 - ушедшие клиенты')"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7feec8b8e5c0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_jcI0ldxR1Fw"
      },
      "source": [
        "Соотношение оттока и оставшихся 1:4. Вывод: такое соотношение говорит, что присутствует дисбаланс в данных."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 605
        },
        "id": "WFLwOI72VusM",
        "outputId": "cf926713-2e65-4041-9575-dc78a34fefd7"
      },
      "source": [
        "# Графики показывают зависимость целевого признака от остальных.\r\n",
        "fig, axes = plt.subplots(nrows=3, ncols=3, figsize=(15, 10)) # fig = plt.figure(), axes = fig.add_subplot()\r\n",
        "for idx, feature in enumerate(df.columns[:-1]):\r\n",
        "   df.plot(feature, \"Exited\", subplots=True, kind=\"scatter\", ax=axes[idx // 3, idx % 3])"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x720 with 9 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K7cDn5EkezgS"
      },
      "source": [
        "Ярко выраженной линейной зависимости не наблюдается"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "id": "ZxCimbeSiqEb",
        "outputId": "2b84ae94-9d19-47dc-e408-18fcd4a1c2d5"
      },
      "source": [
        "# Корреляция  Пирсона\r\n",
        "df.corr()"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>CreditScore</th>\n",
              "      <th>Age</th>\n",
              "      <th>Tenure</th>\n",
              "      <th>Balance</th>\n",
              "      <th>NumOfProducts</th>\n",
              "      <th>HasCrCard</th>\n",
              "      <th>IsActiveMember</th>\n",
              "      <th>Exited</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>CreditScore</th>\n",
              "      <td>1.000000</td>\n",
              "      <td>-0.003965</td>\n",
              "      <td>0.000842</td>\n",
              "      <td>0.006268</td>\n",
              "      <td>0.012238</td>\n",
              "      <td>-0.005458</td>\n",
              "      <td>0.025651</td>\n",
              "      <td>-0.027094</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Age</th>\n",
              "      <td>-0.003965</td>\n",
              "      <td>1.000000</td>\n",
              "      <td>-0.009997</td>\n",
              "      <td>0.028308</td>\n",
              "      <td>-0.030680</td>\n",
              "      <td>-0.011721</td>\n",
              "      <td>0.085472</td>\n",
              "      <td>0.285323</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Tenure</th>\n",
              "      <td>0.000842</td>\n",
              "      <td>-0.009997</td>\n",
              "      <td>1.000000</td>\n",
              "      <td>-0.012254</td>\n",
              "      <td>0.013444</td>\n",
              "      <td>0.022583</td>\n",
              "      <td>-0.028362</td>\n",
              "      <td>-0.014001</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Balance</th>\n",
              "      <td>0.006268</td>\n",
              "      <td>0.028308</td>\n",
              "      <td>-0.012254</td>\n",
              "      <td>1.000000</td>\n",
              "      <td>-0.304180</td>\n",
              "      <td>-0.014858</td>\n",
              "      <td>-0.010084</td>\n",
              "      <td>0.118533</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>NumOfProducts</th>\n",
              "      <td>0.012238</td>\n",
              "      <td>-0.030680</td>\n",
              "      <td>0.013444</td>\n",
              "      <td>-0.304180</td>\n",
              "      <td>1.000000</td>\n",
              "      <td>0.003183</td>\n",
              "      <td>0.009612</td>\n",
              "      <td>-0.047820</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>HasCrCard</th>\n",
              "      <td>-0.005458</td>\n",
              "      <td>-0.011721</td>\n",
              "      <td>0.022583</td>\n",
              "      <td>-0.014858</td>\n",
              "      <td>0.003183</td>\n",
              "      <td>1.000000</td>\n",
              "      <td>-0.011866</td>\n",
              "      <td>-0.007138</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>IsActiveMember</th>\n",
              "      <td>0.025651</td>\n",
              "      <td>0.085472</td>\n",
              "      <td>-0.028362</td>\n",
              "      <td>-0.010084</td>\n",
              "      <td>0.009612</td>\n",
              "      <td>-0.011866</td>\n",
              "      <td>1.000000</td>\n",
              "      <td>-0.156128</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Exited</th>\n",
              "      <td>-0.027094</td>\n",
              "      <td>0.285323</td>\n",
              "      <td>-0.014001</td>\n",
              "      <td>0.118533</td>\n",
              "      <td>-0.047820</td>\n",
              "      <td>-0.007138</td>\n",
              "      <td>-0.156128</td>\n",
              "      <td>1.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                CreditScore       Age  ...  IsActiveMember    Exited\n",
              "CreditScore        1.000000 -0.003965  ...        0.025651 -0.027094\n",
              "Age               -0.003965  1.000000  ...        0.085472  0.285323\n",
              "Tenure             0.000842 -0.009997  ...       -0.028362 -0.014001\n",
              "Balance            0.006268  0.028308  ...       -0.010084  0.118533\n",
              "NumOfProducts      0.012238 -0.030680  ...        0.009612 -0.047820\n",
              "HasCrCard         -0.005458 -0.011721  ...       -0.011866 -0.007138\n",
              "IsActiveMember     0.025651  0.085472  ...        1.000000 -0.156128\n",
              "Exited            -0.027094  0.285323  ...       -0.156128  1.000000\n",
              "\n",
              "[8 rows x 8 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u0aA92x5SI7u"
      },
      "source": [
        "У признаков наблюдается слабая корреляция между собой и с целевым признаком."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sjwTmT8SSMKO"
      },
      "source": [
        "**Преобразование данных**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4A9nCXbaSR9H"
      },
      "source": [
        "# Кодирование категориальных признаков методом \"One hot encoder\"\r\n",
        "df = pd.get_dummies(df, drop_first=True)\r\n",
        "\r\n",
        "# Выделение целевого признака и остальные признаков\r\n",
        "target=df['Exited'] \r\n",
        "features=df.drop(['Exited'], axis=1)"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "E1Gqu_mMSVvm"
      },
      "source": [
        "# Разделение данных на обучающую и проверочную выборки\r\n",
        "train_features, valid_features, train_target, valid_target=train_test_split(features, target, test_size=0.20, random_state=2020)"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gu3uxjWcSXuW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2c94b66f-bb4c-4886-d353-70d6d0c530b3"
      },
      "source": [
        "# Стандартизация обучающей и проверочной выборки\r\n",
        "numeric=['CreditScore','Age','Tenure','Balance','NumOfProducts','HasCrCard','IsActiveMember']\r\n",
        "scaler=StandardScaler()\r\n",
        "scaler.fit(train_features[numeric])\r\n",
        "train_features[numeric]=scaler.transform(train_features[numeric])\r\n",
        "valid_features[numeric]=scaler.transform(valid_features[numeric])"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:5: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  \"\"\"\n",
            "/usr/local/lib/python3.6/dist-packages/pandas/core/indexing.py:1734: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value[:, i].tolist())\n",
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:6: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  \n",
            "/usr/local/lib/python3.6/dist-packages/pandas/core/indexing.py:1734: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value[:, i].tolist())\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8Pmd2g8ksT0O"
      },
      "source": [
        "# Обучение моделей на несбалансированных данных"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "495_bEn2rlld"
      },
      "source": [
        "**Decision tree**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rkzBMS9vsfAh",
        "outputId": "455044ac-e363-41a7-9d46-b72945718be3"
      },
      "source": [
        "# Цикл ищет лучший гиперпораметр max_depth для Decision tree\r\n",
        "for depth in range(1,10,1):\r\n",
        "    model_tree=DecisionTreeClassifier(max_depth=depth,random_state=2020).fit(train_features,train_target)\r\n",
        "    prediction=model_tree.predict(valid_features)\r\n",
        "    print('max_depth:',depth,'F1:',f1_score(valid_target,prediction))"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "max_depth: 1 F1: 0.0\n",
            "max_depth: 2 F1: 0.5321888412017167\n",
            "max_depth: 3 F1: 0.40441176470588236\n",
            "max_depth: 4 F1: 0.5366614664586583\n",
            "max_depth: 5 F1: 0.5440251572327044\n",
            "max_depth: 6 F1: 0.5637795275590551\n",
            "max_depth: 7 F1: 0.5765230312035661\n",
            "max_depth: 8 F1: 0.573134328358209\n",
            "max_depth: 9 F1: 0.5760233918128655\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PgB5q3Bd2CG_"
      },
      "source": [
        "Лучшее значение max_depth=7, при котором F1=0.576"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "biQgDiAZSqCo"
      },
      "source": [
        "**Random forest**\r\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "E-HdA_MazBkg",
        "outputId": "3aef1a11-436a-4d09-f9e0-baeeb17854ce"
      },
      "source": [
        "# Цикл ищет лучший гиперпораметр max_depth для Random forest\r\n",
        "for max_depth in range(1,20,1):\r\n",
        "    model_forest=RandomForestClassifier(max_depth=max_depth,n_estimators=50,random_state=2020).fit(train_features,train_target)\r\n",
        "    prediction=model_forest.predict(valid_features)\r\n",
        "    print('max_depth:',max_depth,'F1:',f1_score(valid_target,prediction))"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "max_depth: 1 F1: 0.0\n",
            "max_depth: 2 F1: 0.17787418655097614\n",
            "max_depth: 3 F1: 0.189247311827957\n",
            "max_depth: 4 F1: 0.38649155722326456\n",
            "max_depth: 5 F1: 0.48263888888888895\n",
            "max_depth: 6 F1: 0.5409015025041736\n",
            "max_depth: 7 F1: 0.5691318327974276\n",
            "max_depth: 8 F1: 0.5786163522012578\n",
            "max_depth: 9 F1: 0.5882352941176471\n",
            "max_depth: 10 F1: 0.5929339477726574\n",
            "max_depth: 11 F1: 0.5987841945288754\n",
            "max_depth: 12 F1: 0.605421686746988\n",
            "max_depth: 13 F1: 0.5901639344262296\n",
            "max_depth: 14 F1: 0.5961251862891207\n",
            "max_depth: 15 F1: 0.6032210834553441\n",
            "max_depth: 16 F1: 0.5929203539823009\n",
            "max_depth: 17 F1: 0.6028571428571429\n",
            "max_depth: 18 F1: 0.5901162790697674\n",
            "max_depth: 19 F1: 0.5904486251808974\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IitbEUok2asP"
      },
      "source": [
        "Лучшее значение max_depth=12, при котором F1=0.605"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "apE-cVMASvTO",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a88bef70-3edf-4856-e12e-7988e89da95b"
      },
      "source": [
        "# Цикл ищет лучший гиперпораметр n_estimators (количество деревьев)\r\n",
        "for estim in range(1,70,2):\r\n",
        "    model_forest=RandomForestClassifier(max_depth=12,n_estimators=estim,random_state=2020).fit(train_features, train_target)\r\n",
        "    prediction=model_forest.predict(valid_features)\r\n",
        "    print('estim',estim,'F1:',f1_score(valid_target, prediction))"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "estim 1 F1: 0.5086551264980027\n",
            "estim 3 F1: 0.5582047685834501\n",
            "estim 5 F1: 0.5428571428571428\n",
            "estim 7 F1: 0.561863173216885\n",
            "estim 9 F1: 0.5776487663280115\n",
            "estim 11 F1: 0.577319587628866\n",
            "estim 13 F1: 0.5781710914454278\n",
            "estim 15 F1: 0.5852941176470587\n",
            "estim 17 F1: 0.5816023738872403\n",
            "estim 19 F1: 0.573134328358209\n",
            "estim 21 F1: 0.5705705705705706\n",
            "estim 23 F1: 0.5684210526315789\n",
            "estim 25 F1: 0.5827010622154781\n",
            "estim 27 F1: 0.5866261398176292\n",
            "estim 29 F1: 0.5878787878787879\n",
            "estim 31 F1: 0.5912518853695324\n",
            "estim 33 F1: 0.5878787878787879\n",
            "estim 35 F1: 0.5891238670694865\n",
            "estim 37 F1: 0.5869894099848714\n",
            "estim 39 F1: 0.5933734939759037\n",
            "estim 41 F1: 0.5969696969696969\n",
            "estim 43 F1: 0.5951661631419939\n",
            "estim 45 F1: 0.5930408472012103\n",
            "estim 47 F1: 0.6012084592145014\n",
            "estim 49 F1: 0.6033182503770739\n",
            "estim 51 F1: 0.6045112781954888\n",
            "estim 53 F1: 0.6066066066066066\n",
            "estim 55 F1: 0.6066066066066066\n",
            "estim 57 F1: 0.6077844311377245\n",
            "estim 59 F1: 0.5988023952095808\n",
            "estim 61 F1: 0.5945945945945946\n",
            "estim 63 F1: 0.600896860986547\n",
            "estim 65 F1: 0.6047904191616768\n",
            "estim 67 F1: 0.6059701492537314\n",
            "estim 69 F1: 0.6047904191616768\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MZWURgIN221l"
      },
      "source": [
        "Лучшее значение n_estimators=57 при котором F1=0.607"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xvoIgcJ8S0i1"
      },
      "source": [
        "**Logistic regression**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0YfJCNTNS2Ul",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8d3cc02e-7fab-4ba9-865a-853c9767add9"
      },
      "source": [
        "model_logistic=LogisticRegression().fit(train_features,train_target)\r\n",
        "prediction=model_logistic.predict(valid_features)\r\n",
        "print('F1:',f1_score(valid_target,prediction))"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "F1: 0.33391304347826095\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "__swCJQeS4DN"
      },
      "source": [
        "Для Logistic regression F1=0.333"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ncruv9cpVLhJ",
        "outputId": "41e580ab-3b29-4124-af09-f1516e2791da"
      },
      "source": [
        "# Обучение модели Random forest с параметрами показавшими лучшую F-метрику (max_depth=9 и n_estimators=41)\r\n",
        "model_forest = RandomForestClassifier(max_depth=12, n_estimators=57, random_state=2020).fit(train_features, train_target)\r\n",
        "prediction = model_forest.predict(valid_features)\r\n",
        "print(f1_score(valid_target, prediction))"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.6077844311377245\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yj3hoIzAS5cm"
      },
      "source": [
        "**Вывод:**\r\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TRSQNWFjS8e9"
      },
      "source": [
        "Лучшее значение F1 меры показала модель Random forest. При значениях max_depth=12 и n_estimators=57 F1 мера =0.607"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rGRZn1m9S_ZZ"
      },
      "source": [
        "# Обучение моделей на сбалансированных данных"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "npd56bImPvyd"
      },
      "source": [
        "**Балансировка данных**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_S-rBaqwTF6-"
      },
      "source": [
        "# Функция увеличивает количество ушедших клиентов (target==1)\r\n",
        "def upsampling (features,target,repeat):\r\n",
        "    #Разделяем выборку по значениям целевой функции\r\n",
        "    target_one=target[target==1]\r\n",
        "    target_null=target[target==0]\r\n",
        "    features_one=features[target==1]\r\n",
        "    features_null=features[target==0]\r\n",
        "    \r\n",
        "    #Увеличиваем и соединяем обратно\r\n",
        "    upsampling_features=pd.concat([features_null]+[features_one]*repeat)\r\n",
        "    upsampling_target=pd.concat([target_null]+[target_one]*repeat)\r\n",
        "    \r\n",
        "    # Перемешиваем\r\n",
        "    upsampling_features, upsampling_target = shuffle(upsampling_features, upsampling_target, random_state=2020)\r\n",
        "    \r\n",
        "    return upsampling_features,upsampling_target"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cIGxeSKcTHZl"
      },
      "source": [
        "features_balance_up, target_balance_up = upsampling(train_features, train_target,4)"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Bub5YxDwTXc1"
      },
      "source": [
        "**Обучение модели Decision tree на сбалансированных данных**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KBoRu0eMTY_c",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b451b8c5-4296-4cfe-b941-154b36bd23d1"
      },
      "source": [
        "%%time\r\n",
        "for depth in range(1,20,1):\r\n",
        "    model_tree = DecisionTreeClassifier(class_weight='balanced', max_depth=depth, random_state=2020).fit(features_balance_up, target_balance_up)\r\n",
        "    prediction = model_tree.predict(valid_features)\r\n",
        "    print('max_depth:',depth,'F1:',f1_score(valid_target, prediction))"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "max_depth: 1 F1: 0.5061611374407584\n",
            "max_depth: 2 F1: 0.531104921077066\n",
            "max_depth: 3 F1: 0.531104921077066\n",
            "max_depth: 4 F1: 0.5693822739480752\n",
            "max_depth: 5 F1: 0.5862361937128292\n",
            "max_depth: 6 F1: 0.6001829826166515\n",
            "max_depth: 7 F1: 0.5838509316770186\n",
            "max_depth: 8 F1: 0.5944272445820434\n",
            "max_depth: 9 F1: 0.577570093457944\n",
            "max_depth: 10 F1: 0.5793650793650794\n",
            "max_depth: 11 F1: 0.563297350343474\n",
            "max_depth: 12 F1: 0.5537190082644629\n",
            "max_depth: 13 F1: 0.5424083769633508\n",
            "max_depth: 14 F1: 0.5286343612334802\n",
            "max_depth: 15 F1: 0.5062146892655367\n",
            "max_depth: 16 F1: 0.5106861642294713\n",
            "max_depth: 17 F1: 0.5086306098964327\n",
            "max_depth: 18 F1: 0.5\n",
            "max_depth: 19 F1: 0.5042016806722689\n",
            "CPU times: user 749 ms, sys: 8.05 ms, total: 757 ms\n",
            "Wall time: 744 ms\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6RO9vl5gSQCf"
      },
      "source": [
        "Лучшее значение max_depth=6 при котором F1=0.600."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FizqDkfvTa-0"
      },
      "source": [
        "**Обучение модели Random forest на сбалансированных данных**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JiEMtR6XTdk8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0801b858-db5a-4948-c441-55b7946fcffb"
      },
      "source": [
        "%%time\r\n",
        "for max_depth in range(1,20,1):\r\n",
        "    model_forest=RandomForestClassifier(class_weight='balanced',max_depth=max_depth,n_estimators=50,random_state=2020).fit(features_balance_up,target_balance_up)\r\n",
        "    prediction=model_forest.predict(valid_features)\r\n",
        "    print('max_depth:',max_depth,'F1:',f1_score(valid_target,prediction))"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "max_depth: 1 F1: 0.534466477809254\n",
            "max_depth: 2 F1: 0.5372290292177192\n",
            "max_depth: 3 F1: 0.5744274809160306\n",
            "max_depth: 4 F1: 0.6055226824457594\n",
            "max_depth: 5 F1: 0.6149253731343283\n",
            "max_depth: 6 F1: 0.6120434353405726\n",
            "max_depth: 7 F1: 0.6170634920634921\n",
            "max_depth: 8 F1: 0.6195761856710393\n",
            "max_depth: 9 F1: 0.610134436401241\n",
            "max_depth: 10 F1: 0.6190975865687304\n",
            "max_depth: 11 F1: 0.6233766233766233\n",
            "max_depth: 12 F1: 0.6153846153846153\n",
            "max_depth: 13 F1: 0.6129398410896708\n",
            "max_depth: 14 F1: 0.6132075471698114\n",
            "max_depth: 15 F1: 0.6275946275946276\n",
            "max_depth: 16 F1: 0.6140567200986435\n",
            "max_depth: 17 F1: 0.6284289276807979\n",
            "max_depth: 18 F1: 0.6192893401015228\n",
            "max_depth: 19 F1: 0.6126126126126126\n",
            "CPU times: user 9.47 s, sys: 47.7 ms, total: 9.52 s\n",
            "Wall time: 9.51 s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vv64zkSmS2_9"
      },
      "source": [
        "Лучшее значение max_depth=17 при котором F1=0.6284."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SdmSpjwcTfW8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1fca2ebf-6b3f-4464-9e8f-5ae4be13644e"
      },
      "source": [
        "%%time\r\n",
        "for estim in range(1,55,1):\r\n",
        "    model_forest = RandomForestClassifier(class_weight='balanced', max_depth=17, n_estimators=estim, random_state=2020).fit(features_balance_up, target_balance_up)\r\n",
        "    prediction = model_forest.predict(valid_features)\r\n",
        "    print('estim',estim,'F1:', f1_score(valid_target, prediction))"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "estim 1 F1: 0.5053763440860215\n",
            "estim 2 F1: 0.4903581267217631\n",
            "estim 3 F1: 0.5635103926096998\n",
            "estim 4 F1: 0.5540897097625329\n",
            "estim 5 F1: 0.5876777251184834\n",
            "estim 6 F1: 0.5751295336787565\n",
            "estim 7 F1: 0.5988023952095807\n",
            "estim 8 F1: 0.5982028241335045\n",
            "estim 9 F1: 0.6189889025893959\n",
            "estim 10 F1: 0.6089171974522293\n",
            "estim 11 F1: 0.6195786864931846\n",
            "estim 12 F1: 0.6116129032258064\n",
            "estim 13 F1: 0.6142322097378277\n",
            "estim 14 F1: 0.6141935483870967\n",
            "estim 15 F1: 0.6274509803921569\n",
            "estim 16 F1: 0.6232790988735921\n",
            "estim 17 F1: 0.6279357231149568\n",
            "estim 18 F1: 0.6188679245283019\n",
            "estim 19 F1: 0.6191646191646193\n",
            "estim 20 F1: 0.6206030150753769\n",
            "estim 21 F1: 0.6280788177339901\n",
            "estim 22 F1: 0.6136363636363638\n",
            "estim 23 F1: 0.6182266009852218\n",
            "estim 24 F1: 0.6248431618569635\n",
            "estim 25 F1: 0.6279357231149568\n",
            "estim 26 F1: 0.628140703517588\n",
            "estim 27 F1: 0.6328800988875154\n",
            "estim 28 F1: 0.6368159203980099\n",
            "estim 29 F1: 0.6305418719211823\n",
            "estim 30 F1: 0.6293532338308458\n",
            "estim 31 F1: 0.6354679802955664\n",
            "estim 32 F1: 0.6360248447204969\n",
            "estim 33 F1: 0.6378244746600742\n",
            "estim 34 F1: 0.6360248447204969\n",
            "estim 35 F1: 0.6335403726708074\n",
            "estim 36 F1: 0.6332916145181476\n",
            "estim 37 F1: 0.6342072409488141\n",
            "estim 38 F1: 0.6273525721455457\n",
            "estim 39 F1: 0.6323713927227103\n",
            "estim 40 F1: 0.6229508196721311\n",
            "estim 41 F1: 0.6290726817042607\n",
            "estim 42 F1: 0.6254728877679697\n",
            "estim 43 F1: 0.6259351620947631\n",
            "estim 44 F1: 0.6322418136020151\n",
            "estim 45 F1: 0.6343283582089552\n",
            "estim 46 F1: 0.628140703517588\n",
            "estim 47 F1: 0.635\n",
            "estim 48 F1: 0.6272040302267003\n",
            "estim 49 F1: 0.6393034825870646\n",
            "estim 50 F1: 0.6284289276807979\n",
            "estim 51 F1: 0.6284289276807979\n",
            "estim 52 F1: 0.6292134831460675\n",
            "estim 53 F1: 0.6327543424317618\n",
            "estim 54 F1: 0.6334164588528679\n",
            "CPU times: user 20.7 s, sys: 78.8 ms, total: 20.8 s\n",
            "Wall time: 20.7 s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6_eX2ftLTkU-"
      },
      "source": [
        "Лучшее значение n_estimators=49 при котором F1=0.639"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "trAgOI83Tj2c"
      },
      "source": [
        "**Обучение модели Logistic regression на сбалансированных данных**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aMyHZizKTmhU",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "44b75f59-107f-4a6d-e488-426c522d7627"
      },
      "source": [
        "%%time\r\n",
        "model_logistic=LogisticRegression(class_weight='balanced',solver='liblinear').fit(features_balance_up,target_balance_up)\r\n",
        "prediction=model_logistic.predict(valid_features)\r\n",
        "print('F1:',f1_score(valid_target,prediction))"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "F1: 0.5017182130584192\n",
            "CPU times: user 36.9 ms, sys: 1.03 ms, total: 38 ms\n",
            "Wall time: 37.4 ms\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WDjzsaugTiGE",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3b2f0c13-5b11-4fff-f2da-75846ee74a71"
      },
      "source": [
        "# Обучение модели с параметрами показавшими лучшую F-метрику (max_depth=17 и n_estimators=49)\r\n",
        "model_forest = RandomForestClassifier(class_weight='balanced', max_depth=17, n_estimators=49, random_state=2020).fit(features_balance_up, target_balance_up)\r\n",
        "prediction = model_forest.predict(valid_features)\r\n",
        "print(f1_score(valid_target, prediction))"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.6393034825870646\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0rmc5ZqcToRG"
      },
      "source": [
        "Вывод: Мера F1 немного улучшилась. На несбалансированных данных F1=0.634 на сбалансированных F1=0.639"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "--DYZFBxuogx"
      },
      "source": [
        "# Metric"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HNcaapS1TqBk"
      },
      "source": [
        "**Метрика AUC-ROC**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ncQwxjA-yyZ3"
      },
      "source": [
        "AUC-ROC для модели Decision tree"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OiN4M28dTuZ8"
      },
      "source": [
        "probabilities_tree=model_tree.predict_proba(valid_features)\r\n",
        "probabilities_one_valid_tree=probabilities_tree[:,1]\r\n",
        "fpr_tree,tpr_tree,thresholds=roc_curve(valid_target,probabilities_one_valid_tree)"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LYKpw50GTvyD",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2ff3471d-4b3e-4c29-8049-f0b4acca8395"
      },
      "source": [
        "auc_roc_tree=roc_auc_score(valid_target,probabilities_one_valid_tree)\r\n",
        "auc_roc_tree"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.683861457432307"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "t9cztJBvy9AF"
      },
      "source": [
        "AUC-ROC для модели Random forest"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kn-DR2mrTyuj"
      },
      "source": [
        "probabilities_forest=model_forest.predict_proba(valid_features)\r\n",
        "probabilities_one_valid_forest=probabilities_forest[:,1]\r\n",
        "fpr_forest,tpr_forest,thresholds=roc_curve(valid_target,probabilities_one_valid_forest)"
      ],
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jx-Z-ucHT0Us",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c7652317-fede-4b1d-99db-a0b18b0e4e51"
      },
      "source": [
        "auc_roc_forest=roc_auc_score(valid_target,probabilities_one_valid_forest)\r\n",
        "auc_roc_forest"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8536476566144203"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0nzzzv4oTse0"
      },
      "source": [
        "AUC-ROC для модели Logistic regression\r\n",
        "\r\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "irzpNBFeT3Fk"
      },
      "source": [
        "probabilities_reg = model_logistic.predict_proba(valid_features)\r\n",
        "probabilities_one_valid_reg = probabilities_reg[:,1]\r\n",
        "fpr_log, tpr_log, thresholds = roc_curve(valid_target, probabilities_one_valid_reg)"
      ],
      "execution_count": 44,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vTM0jremT4O8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d8fa743f-5639-450c-fa44-b61a0a9bd85b"
      },
      "source": [
        "auc_roc_reg=roc_auc_score(valid_target,probabilities_one_valid_reg)\r\n",
        "auc_roc_reg"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.7717903082397022"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-sEib9OtT7Qs",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 475
        },
        "outputId": "42b04295-bdcf-4b91-d688-72b0b06fcc24"
      },
      "source": [
        "plt.figure()\r\n",
        "\r\n",
        "plt.figure(figsize=(10,7))\r\n",
        "plt.plot([0, 1], [0, 1], linestyle='--')\r\n",
        "plt.xlim([0.0, 1.0])\r\n",
        "plt.ylim([0.0, 1.0])\r\n",
        "plt.plot(fpr_log,tpr_log)\r\n",
        "plt.plot(fpr_tree,tpr_tree)\r\n",
        "plt.plot(fpr_forest,tpr_forest)\r\n",
        "plt.xlabel('False Positive Rate')\r\n",
        "plt.ylabel('True Positive Rate')\r\n",
        "plt.title('ROC-кривая')\r\n",
        "plt.legend(('Случайная модель','Случайный лес','Логистическая регрессия','Дерево решения'),\r\n",
        "           loc= 'upper left') \r\n",
        "plt.show()"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 0 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yikaM8WZT9Y6"
      },
      "source": [
        "Значение площади под кривой. На сколько наша модель отличается от случайной\r\n",
        "\r\n",
        "Лучшее значение метрики auc roc=0.853 у модели Random forest"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6P8eRhZNT-0t"
      },
      "source": [
        "**Общий вывод:**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AxgE8akjUDtb"
      },
      "source": [
        "Рассматривая данные с учетом дисбаланса + переведенные категориальные + стандартизированные данные и без учета дисбаланса тремя моделями можно сделать вывод, что самая лучшая модель оказалась Random forest. При лучшем значении max_depth=17, n_estimators=49, мы имеем F1 меру =0.639 на тестовой выборке. "
      ]
    }
  ]
}