pred_lgbm_sklearn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "5a467dec-ab65-499b-8fe3-4ce3ec7126aa",
   "metadata": {},
   "source": [
    "## ライブラリのインポート"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1220,
   "id": "38398f9d-ae81-4b00-9a43-44dfae30f013",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import lightgbm as lgb\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b9b5f1e0-5c7e-4651-a109-03497e6f2adc",
   "metadata": {},
   "source": [
    "## データの読み込み"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1221,
   "id": "7937daf2-f9fc-4c2f-95c7-4bbc6337fd18",
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('input/train.csv')\n",
    "test = pd.read_csv('input/test.csv')\n",
    "df = pd.concat([train, test], axis=0, ignore_index=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "12b2a991-0703-468d-aee2-2d2413161997",
   "metadata": {},
   "source": [
    "## 前処理"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1222,
   "id": "3d1db6f2-bab0-44b3-bb85-dbd7ada23cff",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['datetime', 'season', 'holiday', 'workingday', 'weather', 'temp',\n",
       "       'atemp', 'humidity', 'windspeed', 'casual', 'registered', 'count',\n",
       "       'year', 'month', 'day', 'weekday', 'hour', 'moving_ave_3',\n",
       "       'moving_ave_5', 'moving_ave_7', 'moving_ave_14'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 1222,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['datetime'] = pd.to_datetime(df['datetime'])\n",
    "df['year'] = pd.to_datetime(df['datetime']).dt.strftime('%Y')\n",
    "df['month'] = pd.to_datetime(df['datetime']).dt.strftime('%m')\n",
    "df['day'] = pd.to_datetime(df['datetime']).dt.strftime('%d')\n",
    "df['weekday'] = pd.to_datetime(df['datetime']).dt.strftime('%a')\n",
    "df['hour'] = pd.to_datetime(df['datetime']).dt.strftime('%H')\n",
    "\n",
    "df = df.sort_values('datetime')\n",
    "df['moving_ave_3'] = df['count'].rolling(window=3, center=True).mean() #3日間移動平均\n",
    "df['moving_ave_5'] = df['count'].rolling(window=5, center=True).mean() # 5日間移動平均\n",
    "df['moving_ave_7'] = df['count'].rolling(window= 7, center=True).mean() # 7日間移動平均\n",
    "df['moving_ave_14'] = df['count'].rolling(window=3, center=True).mean() #14日間移動平均\n",
    "# df.dropna(inplace=True)\n",
    "\n",
    "df.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1223,
   "id": "faf4f644-f250-45a7-9ab7-e8728b5a0201",
   "metadata": {},
   "outputs": [],
   "source": [
    "# df.iloc[400:450,:]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ee702cdc-5e04-4618-b6ba-3d0752cb6132",
   "metadata": {},
   "source": [
    "## カラムの選択"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1224,
   "id": "576168f3-e2a8-47a3-83fe-e64a5533971e",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df[\n",
    "    [\n",
    "        # 'datetime',\n",
    "        # 'season',\n",
    "        'temp',    \n",
    "        # 'humidity',\n",
    "        # 'windspeed',\n",
    "        'holiday',\n",
    "        'workingday',\n",
    "        'weather',\n",
    "        # 'month',\n",
    "        'weekday',\n",
    "        'hour',\n",
    "        'count',\n",
    "        'moving_ave_3',\n",
    "        # 'moving_ave_7',\n",
    "        # 'moving_ave_14',\n",
    "    ]\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1225,
   "id": "455a44c0-9931-44d6-a54b-6463ead0eb15",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.get_dummies(df, columns=[\n",
    "    # 'season',\n",
    "    # 'month',\n",
    "    'weekday',\n",
    "    'hour',\n",
    "    'weather',\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aaeea3f2-a873-4739-9907-4d6f69d91635",
   "metadata": {},
   "source": [
    "## 学習データの準備"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1226,
   "id": "46441ed6-b037-4119-bb50-91b441035ba7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['temp', 'holiday', 'workingday', 'count', 'moving_ave_3', 'weekday_Fri',\n",
       "       'weekday_Mon', 'weekday_Sat', 'weekday_Sun', 'weekday_Thu',\n",
       "       'weekday_Tue', 'weekday_Wed', 'hour_00', 'hour_01', 'hour_02',\n",
       "       'hour_03', 'hour_04', 'hour_05', 'hour_06', 'hour_07', 'hour_08',\n",
       "       'hour_09', 'hour_10', 'hour_11', 'hour_12', 'hour_13', 'hour_14',\n",
       "       'hour_15', 'hour_16', 'hour_17', 'hour_18', 'hour_19', 'hour_20',\n",
       "       'hour_21', 'hour_22', 'hour_23', 'weather_1', 'weather_2', 'weather_3',\n",
       "       'weather_4'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 1226,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1227,
   "id": "1b9b546c-01ee-42dc-be17-c1b1926895fd",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1228,
   "id": "e8547a61-3a8d-4c73-ba10-9a752aa4c1db",
   "metadata": {},
   "outputs": [],
   "source": [
    "train = df.iloc[:len(train),:].copy()\n",
    "train['count'] = train['count'].astype('int64')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1229,
   "id": "c645a903-0143-4122-a11e-c3958d3ce0a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "test = df.iloc[len(train):,:].copy()\n",
    "test.drop(['count'], axis=1, inplace=True)\n",
    "test = test.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1230,
   "id": "fcf6db28-80d2-4107-94e5-261d8d8de848",
   "metadata": {},
   "outputs": [],
   "source": [
    "y = train.loc[:,[\"count\"]].values\n",
    "train.drop(['count'], axis=1, inplace=True)\n",
    "X = train.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1231,
   "id": "238388af-470f-4828-9653-f11ff8108f80",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=0)\n",
    "y_train = y_train.ravel()   # 1    2d array -> 1d array \n",
    "y_test  = y_test.ravel()    # 2    2d array -> 1d array "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d25b792f-9afb-4a71-8f58-afb3eeac684e",
   "metadata": {},
   "source": [
    "## モデル構築"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1232,
   "id": "16c41721-9566-494e-b3e5-1ff500a7de98",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = lgb.LGBMRegressor(random_state=0)\n",
    "model.fit(X_train, y_train)\n",
    "y_pred = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1233,
   "id": "a852f4ea-e70d-4bc7-a8f9-71d37b081fa9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# スコアを表示。小さいほどよい\n",
    "def rmsle(real, predicted):\n",
    "    sum=0.0\n",
    "    for x in range(len(predicted)):\n",
    "        if predicted[x]<0 or real[x]<0: #check for negative values\n",
    "            continue\n",
    "        p = np.log(predicted[x]+1)\n",
    "        r = np.log(real[x]+1)\n",
    "        sum = sum + (p - r)**2\n",
    "    return (sum/len(predicted))**0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1234,
   "id": "41e0a042-7c18-4638-ab10-33ba55530671",
   "metadata": {},
   "outputs": [],
   "source": [
    "# from sklearn.linear_model import Lasso                         # LASSO回帰のクラスを読み込み\n",
    "# model = Lasso(alpha=1)                                # LASSO回帰のインスタンスを生成\n",
    "# model = model_lasso.fit(X_train, y_train)                                      # データを学習させる\n",
    "# y_pred = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1235,
   "id": "7b6f9d2b-0aea-41b0-840c-7659edf895aa",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEXCAYAAABCjVgAAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAAA6mklEQVR4nO3dd5gUVdbA4d+ZoYHBwICyioMIa0BFBATToitGDCiIIiKymBMqugiCy6fgGkBUwJwzIoo6BF3RFdxdXVHAIYiKoksaRVEYVBhgwvn+qOqip6e7pzpN6vM+zzz0ra6qvt22dbpuOFdUFWOMMQYgq6YrYIwxpvawoGCMMcZjQcEYY4zHgoIxxhiPBQVjjDEeCwrGGGM8FhRMvSIit4jIUzVdDwMislJETqrpepj4WFAwtYaIfCAilyVzDlW9S1WTOkd1E5E2IqIi0iAF53pORO5IRb1MZrKgYOqMVFw0TWLss88cFhRMyonIMBF5PWzbAyIyKcYxdwLHAg+JyO8i8pC7XUVksIh8A3zjbpskImtE5FcRWSgix4acZ7SIvOQ+Dv4CHyQiq0XkZxH5W5TXP1JE1olIdsi2s0Vkifv4CBFZ4L7mjyJyf5TzfC4iZ4aUA+7rdo7xkf3b/bfIfe9Hu8deIiJfishGEZktIvu420VEJojIT259lorIISJyBTAAGO6eZ2aM1ww274wUkS/c13hWRBq7z3UXkbUicrOIrAOeFZEsERkhIt+KyC8i8qqINA8530ARWeU+F/FzNnWAqtqf/aX0D2gJbAZy3XID4CegSxXHfQBcFrZNgfeA5kCOu+1CYDf3vEOBdUBj97nRwEvu4zbu8U8COUBHYBtwUJTX/xY4OaT8GjDCffwxMNB9vDNwVJRzDAemhpR7AUureN/BejYIO24FcJD7PkcB/3Wf6wEsBHIBcfdp6T73HHCHz/9OK4HPgb3dz/ej4LFAd6AUGAc0cj+/IcA8oJW77XFgirv/wcDvwJ/d5+53jz+ppr+P9hffn90pmJRT1R9wfv32dTedCvysqgsTPOXdqrpBVYvd87+kqr+oaqmq3odzEWoX4/gxqlqsqouBxTjBIZIpQH8AEdkFON3dBlAC7Cciu6vq76o6L8o5XgJOF5Fd3fJA4EWf7zPUVTjv+0tVLQXuAjq5dwslwC7AgYC4+/yQwGsAPKSqa1R1A3An7vt3lQO3qeo297O/Cvibqq5V1W04Afhct2npXGCWqv7bfe7/3ONNHWNBwaTL8zi/6HH/TeTCGLQmtCAiN7nNKptEpAhoCuwe4/h1IY+34PzSj+RloI+INAL6AJ+p6ir3uUuBA4CvRGS+iPSMdAJV/R7nF/c5IpILnAZMjvXmotgHmCQiRe573IBzV5CnqnOAh4CHgZ9E5ImQIBSv0M92FbBXSHm9qm4Nq9ObIXX6EigD9nCP886lqpuBXxKsk6lBFhRMuuQDh4rIIUBP/F0Yo6Xs9ba7/QfDgfOAZqqaC2zCuWAmRVW/wLkwngZcgBMkgs99o6r9gT/gNKlME5GdopwqGBD7Ah+ramFVLx1h2xrgSlXNDfnLUdX/uvV5QFW74DTbHAAMi3GuWPYOedwa+D5GvdYAp4XVqbH7/n4IPZeINMFp4jN1jAUFkxbuL8xpOBfWT1V1tY/DfgT+WMU+u+C0Va8HGojIrUCiv5IjeRmn7fzPOH0KAIjIhSLSQlXLgSJ3c7TmkXzgMPc8L/h4zfXuuULf+2PASBFp775+UxHp6z4+3O0YD+D03WwNqYufzzDUYBFp5XYY/w2YGmPfx4A7Qzq8W4hIL/e5aUBPETlGRBoCt2PXlzrJ/qOZdHoe6ID/pqNJOG3UG0XkgSj7zAbeAb7G+VW/lbDmpSRNAY4D5qjqzyHbTwWWicjvbj3PD/ZxhHO3vw60Bd6o6gVVdQtOe/5HbtPMUar6Js4dySsi8itOh/Bp7iG74nSeb8T5DH4BxrvPPQ0c7J4n38f7fRl4F/gOp6M91hyHScAM4F0R+Q2n0/lI9z0sAwa75/vBrdtaH69vahlRtUV2THqISGvgK2BPVf21putTndw7mANU9cIqd64hIrISZ7TXP2u6Lqb2sAkpJi1EJAv4K/BKBgaE5jgd0wNrui7GxMuaj0zKuR2wvwInA7eFPfd7lL9jI56sjhGRy3Gas/6hqv8O2T4gyvtelsa6tI7xebdO1+uaus2aj4wxxnjsTsEYY4ynzvcp7L777tqmTZuaroYxxtQpCxcu/FlVW4Rvr/NBoU2bNixYsKCmq2GMMXWKiKyKtN2aj4wxxngsKBhjjPFYUDDGGOOxoGCMMcZjQcEYY4ynzo8+MsaYTJNfUMj42cv5vqiYvXJzGNajHb0756Xk3BYUjDGmDhmVv5TJ81Z7i10UFhUz8o2lACkJDNZ8ZIwxdUR+QWGFgBBUXFLG+NnLU/IaFhSMMaaOGD97edSl9b4viri8R9wsKBhjTB3xfVExDUtL+PShgawc15Pu3873ntsrNyclr2FBwRhj6oiLv/uQr+87mz9s3gjAglbtAWeB8mE92qXkNayj2RhjartNmyA3l1vd4qwDj+Xas4aDCAIMOKq1jT4yxpiMMG4cjBjhFd/L/w93f7kdScNwVLCgYIwxtVNhIbRqtaM8bBjccw8nAyf3St/LWlAwxpgaED4B7fgDW/DWkh/YuKWEv7/7CAML3t6x848/wh/+UC31sqBgjDEJSnRmcX5BISPfWEpxSRngTEB7ad5q9l+/ioJnBnv7jT7pSvKP6cOm++d7gWPuV+vTMpM5yIKCMcYkINKFPdbM4tAAkiVCmYbMOFDlhVdv5c8rCwAoycqm45BX2NIwB4pLvPO/NG+1d0iqZzIH2ZBUY4xJwPjZy72AEFRcUsboGcsq7Tsqfyk3Tl1EYVExChUCwuFrPmflPWd6AeHqXiPYf9h0JyBUIZUzmYPsTsEYYxIQbQZxUXEJ+QWF3q/3aKkpGpSVMvuZwey7oRCAb5vn0eOShynNju+ynKqZzEF2p2CMMQmINYM49Nd7pNQUpy7/iBX39vYCQt8LxnLi5Y/HHRCqqkci7E7BGGMSMKxHO26Yuijic6G/3kMfN9lezOJJ5xMod5qd/vXHLgw6dzQiklAdcgLZKZvJHGR3CsYYk4DenfNo1iQQ8bnQX+/Bx4MWzuSLCX29gHDKJQ+zcdp0Vo7ryYR+ncjLzUGAvNwcJvbrxMR+nQhkRw8WAtzdp4ONPjLGmNritjPbVxiBBJV/vf/t8N04/aROXvmlTqfxfz0GV0hN0btzXtSL++gZyyhyRyAFBbKE8X07pjwggAUFY4xJWPCiHHWuwvDhnD5+vLf/0Vc/R1brvZngc35BMFikc6W1cKIaLTt33dC1a1ddsGBBTVfDGGN2WLEC9t9/R/muu2DkyJqrTwQislBVu4ZvtzsFY4xJFVXo1w9ee23Hto0bITe3xqoUL+toNsaYVJg/H7KydgSE5593gkQdCghgQcEYY5JTVgaHHw5HHAHAz7s0p93QN+n2fSvyCwpruHLxs+YjY0zGSqYDN7+gkA8nvcC9z9/ibbu0/x2837oTkL7cROmW9qAgIjcClwEKLAUuBloCrwC7AQuBgaq6XUQaAS8AXYBfgH6qujLddTTGZJ54E9qFmjHvW7qf0Jnexb8BsCDvIM4bMI5yqdj4EsxNFCtBXmFRMdlugry8NI8s8iOtzUcikgdcD3RV1UOAbOB8YBwwQVX3AzYCl7qHXApsdLdPcPczxpiUi5bQrsoEc888w1lH70euGxDOGDSRcy8cXykgBEXKTRQMSIXuc8EEecHAVJPNTtXRp9AAyBGRBkAT4AfgBGCa+/zzQG/3cS+3jPv8iZLo/G9jjIkhWiK5qAnmNm4EEbjU+Q2bf/BxtLl5Fsv23C/m60TKTRQpIAWlI/NpPNIaFFS1ELgXWI0TDDbhNBcVqWqpu9taIHivlAescY8tdfffLfy8InKFiCwQkQXr169P51swxtRT0RLJKdBt7JyKv9b//ndo3twr9h32IjecOazSseG/YKPlJqoqs2mqM5/GI93NR81wfv23BfYCdgJOTfa8qvqEqnZV1a4tWrRI9nTGmAw0rEc7cgLZEZ8LNuPM/sd85+7g1ludJ265BVQZ0P/4SsfmBLIZcFTrCjmMouUmqiqzaaozn8Yj3R3NJwH/U9X1ACLyBtANyBWRBu7dQCsgGJILgb2BtW5zU1OcDmdjjEmp0BQVhRF+md86YwI97nh3x4b162H33Ssdm8jIpWE92lXKmRSUjsyn8Uh3UFgNHCUiTYBi4ERgATAXOBdnBNIgYLq7/wy3/LH7/Byt63k4jDG1VjC3UNsRb3lrHrRbv5LZz1zr7TP6tGvpdOcIersBIfzYRF8XqJWjj9IaFFT1ExGZBnwGlAIFwBPAW8ArInKHu+1p95CngRdFZAWwAWekkjHGpNVeuTkUbtzC5Kl/o9uqJQBsDjSmy3UvsTXQmLwow0qTkUxQSae0z1NQ1duA28I2fwccEWHfrUDfdNfJGFN/JJtBNL+gkIOWL+Sj54Z72y7vM4r39j/KK0dqXqqvbEazMabOijUBDapu85/+6Uq6nnIUvTf9CMBXu+/DGRc/QFlWxU5kcV8r/PjqTGldXSwoGGPqrGgT0EbPWMa20vLYs5WnTqXX+TtaqPsMGM9nrQ6K+DrqvlboBT+ZGdG1mQUFY0ydFW08f/hKZeAEi6GvLqbB5t/peeyB3vb39juSy/uMcoaexvFasWZEW1AwxphqFGy2iXdo4kWfvknPsU955d6Dn2TRzi19HRs+dyDuGdF1hAUFY0ydEt5sEy4nkE3jQBYbt+y4W2jx+0bmPzzQK798ZG+aPPoQy6YthrKKoSVLIFuEknKtcM7wuQN75eZE7ICuyYlnqWDrKRhj6pRYeYOCs4hvO7O9N+N41PtPVggIhw9+gVu6X8b42cspKat8r9E0J8D4vh2rnJkcaUZ0TU88SwW7UzDG1CnRmmcE+GjECV55p1XfcfLZf/bKdxx/CU8d0afK8xRtKfE1hyDZWc21lQUFY4xvtWEIZpXNNqpw9tmcPH2699whN7zK742aeOXcnAA7NWqQdPNPbZ2AlgxrPjLG+BK6BoBSc7n/YzXb/Ov5Gc46yW5AePWvY9l/5FsVAkIgSxh9Vvt62/yTLLtTMMb4UluGYEZstjlpP44/7ySarvgKgLW7tuD4K56gQePG9OuSx9yv1ke9uxkzc5nXKd2ogf1OtqBgjPGlpodgRm26mjEDDt/H26//+Xfy8T4dASgpKWPKJ2soV43a3LW1pNx7XFRcUi8moCXDwqIxxpdobe3VMQQzUtPVmKnzKdlpZ+jVC4CPW3eg7fAZXkAIKlON2tyV8JKc9ZgFBWOMLzXZBh9+8b5g0T8oGHc2gS2bnQ2LFnHT1RPRKOskB4Vf8Gv67qc2sqBgjPGld+c87u7TwdfKYqkWvEjnFv/KynE9uWv2wwC82uEkZ7RRx44cf6C/VRhDL/g1efdTW1mfgjHGt3QOwQzvMzj+wBZeB3GWCEP+9QLXfzzV27/bVc9Q2PQPTBo7h2E92jH3K3/rtYde8COtgJbpI5AsKBhj0sbvvIZIGUdfmrcagFabfuTDxy719p3YrT8TjxnglYN9BdFmOYcKv+DX1wloybCgYIxJSrQLfzyppaOlrrh/1n30WTbXK3e8fgqbcnaptF9xSZm3pGU00Za6rI8T0JJhQcEYk7BoF/4FqzYw5ZM1lS7S0eY1hHfsHvzjd7z93PVe+eZTr2Nqxx4x6xIrIDRrEqiQAsNEZ0HBGJOwaEM6J89bHTWtdTAA5BcUVpg4BiBaztSXR3DE2i8A2NRoJ44Y/ALbAo2qrEtelPQX4OQzMv5YUDDGJCza0M1Y6xwo0GnMu/y2rZSykPTU3VYuYvLUUV75knNuZc5+lZZyj2pYj3aMn728Xqazrk4WFIwxCYuWnK4qoSujNSwt4d+PX8qev28AYOke+9LrL/dTHrZOclWCTVI2mig5Nk/BGJOwSBPa4tF72Vy+vu9sLyD0GngfZ140Ke6AkOfeCdTkXIr6wu4UjDEJC15sh766OGZHb7hdtm1m6cR+Xvmtdt0Y3GtEleskR5ITyOb4A1vQbewcG1aaAnanYIxJSu/OeZTHERCumjetQkDofvnjDO49Mq6AEHoncE6XPF5fWFjjKb3rC7tTMMbElF9QyOgZy7x+gGZNApxxaMsK6ahzmwQqjCKKZI/ffuaTRy7yyk8e3ps7T7gs7vrk5lQcXtpt7JxakdK7vrCgYIyJKr+gkGGvLa6wiP3GLSXebGNwfpkHsoTsLKkwmijU7e8+yl8K3vLKXa99kZ93ahZ3fYIL5ISypHapZUHBGBPV+NnLKwSEaErKlUiNP/v+vIb3n77aK4858XKe7dorobpkizC+b8dKv/6rXJ7TxMWCgjEmqnh+bVcIHao8M20MJ3y3AIAyyaLDDVPZ0jDxC3W5asTmIEtql1oWFIzJYFUlrItnHoLgBIaua5cxbfLN3vZreo3g7QOPSbqu0X75W1K71LKgYEyG8pOwbliPdpX6FCIJZAk5Wcobj17FfhvWAvC/Zi05+dJHKc2OfJkJBpFwTQJZKBLXL39Lapc6NiTVmAzlZynK3p3zGN+3I7k5AW9bk0AW2VkVexBOWv4RS+7s6QWEfv3v5vgrnow7IABsKSm3CWg1yO4UjMlQ0ZqFwvsRwn+Fdxs7xzs2Z/tWFj1wPo3KSgH4d5vO/OW826uccxDrviNbxH751yALCsZkoPyCwqi/1vfKzYnZ1xAMGgM/m8Xf33vMO+6USx7i6xZtkq5bPDOjTepZUDAmA42fvTxiQBDg+ANbxOxrOLjBNt668xzvmJc79uCWU6+r8jVjNRmFyrOhpDUq7UFBRHKBp4BDcL4TlwDLgalAG2AlcJ6qbhQRASYBpwNbgItU9bN019GYTBMr5XW0xXFGz1jGbzcO461/TfG2H331s/ywa4sqX89vQLChpDWvOjqaJwHvqOqBQEfgS2AE8L6q7g+875YBTgP2d/+uAB6thvoZk3GiDe8UIjff7F20jkWjezDQDQjjjx1Im5tn+QoIAA2ypUJndahsEetQrkXSeqcgIk2BPwMXAajqdmC7iPQCuru7PQ98ANwM9AJeUFUF5olIroi0VNUf0llPY+qi0Hb/pjkBRJwVxvyM04804Qsi/JpX5YGZ4znry397mw4d8gq/Nt45rrqWlCkizp1A+FDTdAaCquZhmMrS3XzUFlgPPCsiHYGFwBBgj5AL/TpgD/dxHrAm5Pi17rYKQUFErsC5k6B169Zpq7wxtVX4HIPQRWvC+wCCF8bComJvcfu83BwOa92Uj77dEPU1OvzwDTNfuNErDz39Rl7vcGLCdS7aUsKEfp2q7SLtZx6GqSzdQaEBcBhwnap+IiKT2NFUBICqqojENdxAVZ8AngDo2rWrDVUwGSfSHINQofMNhk1bTEmZ879JsGmosKg46pBU0XLefHEonX74BoD1TXI55upn2Nagoa+6BQNPuL1yc6p1qGmseRgWFKJLd1BYC6xV1U/c8jScoPBjsFlIRFoCP7nPFwJ7hxzfyt1mTMYLbQrx80vo+6Jixsxc5gUEP477biHPv3abVx7Udwz/+mMX38cL0P/IvXl9YWGFC3JwVFN1suypiUlrR7OqrgPWiEhwOMGJwBfADGCQu20QMN19PAP4iziOAjZZf4IxO5pCCn0GBHD6B6pa4yCoUel2Fj5wgRcQClq2o+3wGXEHhAFHteaO3h04p0vFX+IKTJ2/ploXvonWmW7ZU2OrjnkK1wGTRaQh8B1wMU4welVELgVWAee5+76NMxx1Bc6Q1IuroX7G1HpVNRcl49yl/+Tetyd65Z6DJvL5nvvFdY68sP6Bt5ZU/i1XUqaMmbms2ppuLHtqYtIeFFR1EdA1wlOVeqzcUUeD010nY+qadDR57Lr1d5ZMOt8rTz/oOIaceVPc6yQLVFgJDaLfofi9c0kFy56aGJvRbEwdEE8Kaz8G/3cqw/7zolc+9sqnWJO7Z0Lnqs3NMZZDKX4xg4KIzCTGRERVPSvlNTLGVBKpKcTvLOFQLX9dz8eP7miVffiovow/blCMI2KL1hyTmxOoMEw2dLup3aq6U7jX/bcPsCfwklvuD/yYrkoZYyoKbQoJnW8Qj7veeZALFs/2yoddN5kNTZomVa9oE89Gn9W+0joMkdZXNrVPzKCgqv8CEJH7VDW0X2CmiCxIa82MyWCxZuJGmokcS7v1K5n9zLVeedTJV/PSYWckXcc8d95BJNaeX3eJ+vi1ISJfAmeo6nduuS3wtqoelOb6Valr1666YIHFJ1N7xZtqIXwmLkAgW9ipYYOITTJRqfLS1FEcs2oxAFsbNKTzdS9T3LBxwu8ltD7jz+1oF/k6TEQWhv3YB/x3NN8IfCAi3+E0Ze4DXJnC+hlTL8WbaiG/oJChry6u1DRUUqZxBYQjVy9l6pSRXvnKs29h9gF/irv+ebk5HH9gCybPW+31X+zUMJs7z7bEdfWVr6Cgqu+IyP7Age6mr1R1W/qqZUz9EE+qhWAASWaRmQZlpbz/1FXsU7QOgK93a81plzxIWVZ2lcdGSlZ3/IEteH1hYYUO7SqWazZ1nK8ZzSLSBBgGXKuqi4HWItIzrTUzph6IJ9VCshPUzvjyP6y4t7cXEM4ZcA+nXPaIr4AQTFsdvi7y3K/WV7mOs6lf/DYfPYuT4fRot1wIvAbMSkeljKkvos0viDS2P9EJajtt28Kyied55X/ueziXnXOr70lowWGlkcb03zh1UcRjLH9Q/eU399G+qnoPUAKgqltw+haMMTEM69GOnEDFX+rRxvYnMgnskvnTKwSEEy99lMvOvc13QGjWJBBzPQPLH5R5/AaF7SKSgztXRkT2BaxPwZgq9O6cF7FZJtJFOFIAiabF7xtZOa4nt855EoDnDzuDNjfP4tvd967iyIq2lpTHfD6eoGbqB79DUk8GRgEHA+8C3XDWT/4grbXzwYakmvpkwJMfx1z4BmDU+09y2YLpXvmIa57np112S/g183JzKuUuCmWrl9VPCQ9JFZEsoBnOrOajcJqNhqjqzymvpTH1kN+L6qj8pTEDQtsNhcx9csdI8Lu7X8TjR56bdP2q6h+w/EGZpcqgoKrlIjJcVV8F3qqGOhlTb0Sap3DD1EWMnrGM0We1r3CxnTxvdeSTqPJo/t2c9vV/vU0dbpjKb412SkkdrX/AhPI7+uifInITMBXYHNyoqrHvc42pJWqqCSTaMNOi4hJumLqIW95YQnFJOU1zAhGT23X6fjn5Lw71ykN6DmV6++Ojvl5uToDRZ7XnhiijhsJZ/4AJ5zco9HP/DV3rQIE/prY6xqReTS7gXlXTzBa3ozd8tnJWeRkzn7+R9j9955xnl9057sonKcmOnWW0qLiEMTOX+apb+MI4xoD/Gc1t010RY9KlJhZwD96ZJDL594QVn/LM67d75QH97uCjNp18HZst4mshm0gL4xgDPoOCiDQGrgGOwblD+A/wmKpuTWPdjEmJ6l7APb+gsFLaaD8alWxj/kMD2XX7FgA+2fsQzu9/Fyr+l1L3myLD+hFMNH6bj14AfgMedMsXAC8CfdNRKWNSKZ5Zxakw8o0lcQeE8xe9w9jZD3nl0y96gC/2iL91Ni83h83bSmMmz7N+BBOL36BwiKoeHFKeKyJfpKNCxqRaOhZwDzYPhS54E8woWlzFhLBQucW/suiBC7zy6+2PZ2jPoTGOiE7Ae0/R7lSsH8FUxW9Q+ExEjlLVeQAiciRgM8ZMnZDqBV/CO66DTTaFRcW8FG1YaQQ3/mcyQ/47xSsfc9XTrG26R0J1EmDAUa0rvCebcGYSEc8iO+2A4De+NbAcKAVUVQ9NWw2rYDOaTXXrNnZOxOYov/I2/cRHj13ilSf96XwmHHth3Odp1iRA0ZYSu+ibhCS7yM6pVZy8mapuTKhmxtQxyQSE8W9NpO/n//TKna5/maKcXRM619aSMv43NvllNY0J5XdI6qpYz4vIZ8BhKamRMTXA7+S2/ILChM5/0E/f8Y9nr/fKI3tcy5ROMX9rVSmevgtj/PJ7p1AVS6Nt6qxIk9tunLqIG6YuqtAxOyp/aVx9BgCi5bwy5RaOXPM5AL82bMLh177ItkCjlL8PY1IhVUHBFugzddboGcsqTW4LfqFDA0S8/rRyES9PHeWVLz3n/3h/vyOTqKkx6ZeqoGBMnZNfUMjoGctijumH+H/xBMpK+Pdjl9Hy918A+OIPbek5aCLlPpbFjMdODVN7PmPAmo9MhgpvMkqVs774gAdm3uuVz77wXgryDqy0X25OgBa7NOSbnzZXes6vLdtTW3djwOfKayLyYhXbTkxZjYypBtGylyZq521bWDmupxcQZu9/FG2Gz4wYEAA2FZfw3l+7023f5gm/pqWqMOngN6lK+9CCiGQDXYJlS6Ft6ppU5j264pPX+TxkneTjL3+cK/uMirlOcvCCPvnyo8lL4OJuqSpMusRsPhKRkcAtQI6I/BrcDGwHnkhz3YyJm9+hpdHyIcXjD7/9wqePDPLKT3ftxd9PvLzK4wSnA7vb2DkM69HOV4AKZAk7N25gk9VM2vmd0Xy3qo6shvrEzWY0m6BI/QQ5gWzu7tOh0gU02T6F0e89xkWfzfLKhw9+kfU7N4v7PDmBbBoHsmKmuw4unGNBwKRSsjOaZ4nITqq6WUQuxJmoNqmqSW3GVKdo6yaMfGNJ1LuHeIea7vvLGt5/6mqv/PcTLuPpw3snXOfikjIaNcgikC2UlEX+gbattJwFqzZYLiNTLfzeKSwBOgKHAs8BTwHnqepxaa2dD3anYMD55e/3Ai84w0xzAln+ZwWr8vTrt3Pit/O9Te1veJXNjZrEXddI9ZnQrxNjZi6LescQrHNQtDsgY/yKdqfgt6O5VJ3o0Qt4SFUfBnZJZQWNSVSwKciv4MXVb0DosvYLVt5zphcQrj1rOG1unpWSgABO/0bvznkU3HpK1LHd4T/dgivHxSu/oJBuY+fQdsRbdBs7J+G0Hab+8tt89Jvb6TwQOFZEsoDYi8WGcEcrLQAKVbWniLQFXgF2AxYCA1V1u4g0wlnQpwvwC9BPVVf6fjcmI6V6eGlQdnkZ/3jmOg74xUltsbrpHpxw+eOUZsc/vSeY2vr1hYUx13WIpwM83hFUNblWtak7/N4p9AO2AZeo6jqgFTA+jtcZAnwZUh4HTFDV/YCNwKXu9kuBje72Ce5+xsSUjmU1T/n6Y74d38sLCOf3v4s/X/V0QgEBnOahO3p34O4+HcjLzUFwFrwJbwIa1qMdOYGKM5Wj3T3EO08h1lrVxgT5zZK6TkReB/Z3N/0MvOnnWBFpBZwB3An8VUQEOAFnSU+A54HRwKM4zVOj3e3TgIdERNRPx4fJWKkYXhqUs30rBQ9eQOPS7QB8uE9HLux3R8w5B1Vp1iTgXfh7d86L+as80oJAxx/Yoso7DD+qe61qUzf5CgoicjlwBdAc2BfIAx7D30zmicBwdvRB7AYUqWqpW17rng/33zUAqloqIpvc/X8Oq88Vbn1o3bq1n7dg6rFIy20m4sLP3uKO9x71yj0ueYjlLdokWTu47cz2Ve8UIlLg6LpP86RHH1X3WtWmbvJ7LzwYOAL4BEBVvxGRP1R1kIj0BH5S1YUi0j3RSoZT1SdwJ8917drV7iLqGb8T0EL3TSYgNN+yic8eHOCVpxx6CiNPuz7GEfFJRXt9VXcYfqRjrWpT//gNCtvcjmAARKQB/pJHdgPOEpHTgcbArsAkIFdEGrh3C62A4BCIQmBvYK37Gk1xOpxNhoinMzS/oJBh0xZHHd/vx7B/Pc/gea955aOvfpYfdm2R8PnCJZLCIl1SvVa1qZ/8BoV/iUgw3cXJwDXAzKoOcmdBjwRw7xRuUtUBIvIacC7OCKRBwHT3kBlu+WP3+TnWn5AZgr/4IzVvBDtDQy9e+QWF/PXVRZQn+O3Yu2gd/3n8Mq983zEDeLBb/8ROFkVt/BWeijsOU7/5DQojcEYGLQWuBN5W1SeTeN2bgVdE5A6gAHja3f408KKIrAA2AOcn8RqmjvCTciK0MzS4f6IBYdKM8fT68l9e+dAhr/Br450TO1mIQLawU8MGbCq2/ESm7vIbFK5T1UmAFwhEZIi7zRdV/QD4wH38HU4fRfg+W4G+fs9p6gc/fQKhnaGJ9iEcsm4Fs56/wSsPO+16Xjv0lLjPE4nlJzL1hd+gMAinLyDURRG2GRO3qoZEhjbD5BcUxj38VLSc118axmHfO+PxN+TsytHXPMe2Bg0TqzDOCFVVKqzhbEx9UFXq7P448wnaisiMkKd2wWneMSZpseYZhF50401nAfDn7xbywmu3eeWLzh3NB/tWSvcSN9UdwcoCgqlPqrpT+C/wA7A7cF/I9t+AJemqlMks0YZK3t2nA+A0F8WbzbRhaQkfPXoxLbYUAbCo5f70ufDelK6THKkD3Ji6LmZQcFNjrwKOjrWfiHysqjH3MSZc6HyE3CYBGjXIqtBJCyQ0Ke2cpe9z39sTvPJZf7mfJS0PSGndg2w2sKlvEkvkUlnjFJ3H1BHxTDCLdnzoBX/jlhJyAtlM6NfJO0+nMe/GFRB23fo7SybtGLA288Bjue6s4UmlqKiKzQY29U2qgoLNJcggqci2GSs5W7D/oKg4+mpk4a75+FWG//sFr3zcFU+wqtlevo9PRG2ch2BMslIVFEwGqeqCHkn4nUW0juXComLajHjLd132/PVn5j16kVd+7MhzGNv9Yt/HJ8pGHZn6ym9CvOuAl1R1Y7RdUlclU9vFm20zPB1FqjKa3jn7IQYsescrd7n2JX7ZKTcl544mJ5DFl38/La2vYUxN8nunsAcwX0Q+A54BZoelnxiY8pqZWivebJtjZi5LKj9RuP3Xr+K9ZwZ75VtPupIXupyZsvPHstXv8p3G1FG+FtlR1VE4ayk8jTNp7RsRuUtE9nWf/zxtNTS1TqSFYGK1r0dbdzhuqrww9f+8gLAtuwEH3Tgt6YAQz22udSyb+s53n4KqqoisA9YBpUAzYJqIvKeqw9NVQVP71ES2zSPWfM6rL4/wylf1Hsk77bql5NyK00dQWFRMtghlquTmBNi8vbTCHY51LJtM4LdPYQjwF5zFbp4ChqlqibtW8zc4i+iYDBIp22a0Yaq5OYG4RhKFalBWyntPX03bjT8AsKJ5K3pc+jBlKZyElpebw0cjTqi0Pdlht8bURX7vFJoDfdzJbB5VLXcX0jEZJvyCGb5kZOgw1fZ77cJH38afFeX0rz7kkeljvfK5A8axoFV8q5iFygLCewQC2RL117+lmTaZSOr6cgVdu3bVBQsW1HQ1MkqkVNdC5MkqzZoEKNpSEtdElibbi1k6sR/Z6lzCP2jbhYv6jk5qElqWwAVHtmbW4h+8u5ZmTQLcdqZlNjWZSUQWqmqlRGA2T8HELdI8hWgX/Xg7mS9aMIPR7z/hlU+69BFW7J78OtzlCnO/Ws+i21KTKtuY+sqCQj1Q3W3f6cj3s/vmjSx4aMfI5hc7n87/nXJNSl/D8hQZUzULCnVcKlJOxCvWjOREjJz7DFd++oZXPvKa5/hxl91Tdv4gG05qTNV8zVMwtVeslBPpcvyBqVnYvs2GQlaO6+kFhHHHDaLNzbPSEhBsOKkx/tidQh0Xb8qJZOQXFDJ6xrKEh5d6VHkk/25O//q/3qYON0zlt0Y7JVnDyGypTGP8s6BQx8WbciJR+QWFDHttMSXlyY1W6/j9cqa/ONQr33jGX3nzkMpzBFJpW6mlpjDGL2s+quPiTTmRqJFvLEkqIGSVlzHruSFeQFi3c3MOGPpmSgJCt32bk5ebgwDZEYatprs5zZj6xO4U6rh0ppwIjmpKtlP5+G/n8+y0MV554Hm385+2hyVbPc9nqzdxd58O9O6cR9soabdt5JEx/lhQqAfSMfM20gS1eDUq2canD/+Fpts2A7Ag7yD6DhiHSmpvUEPXcqiu5jRj6itrPjIRRRrVFI/zFr/L8vvP8QLCGYMmcu6F41MeEIKCdwLV1ZxmTH1ldwomokSbW5oW/8biB/p75TcP7s6NZ96UqmpFFbwTqIkMrsbUJxYU6qFkZjgHj02kS3nIhy9z40cve+VjrnyKtbl7JnCm+ITfCVgiO2MSZ0GhHggNAk3D1gGINcM5/LjtpWVsSWBlsbxNP/HRY5d45QeP7sd9f66exfgEOKeLBQFjUsWCQh0X3iEcaWJZaEdsPMf5sWjS+eRu/d0rd75uMhubNE3oXIlQnER3xpjUsI7mOs5vh3BoH0F+QSFDX12cVEfysf/7jJXjenoB4fEj+tDm5lnVGhCCbLipMaljdwp1nN8LYrAjdlT+Ul6atzrh1xMt53/3nFVhWzpTVGQBTd01GbLcpTLD2XBTY1LHgkId5ydjabAjNr+gMKmAcN7id7nnnQe88t3dL+LxI89N+HyhcnMCiDjrLwTXSc4L6ySPNHfChpsak1oWFOq4YT3aVbpQBrKEnRs3oGhLSYXRR53GvJvQazQq2cby+8+psG2/m/IpzU7N1yf84h+NDTc1Jv1sOc56INoQ1PDtiaSruOnfL3Dtx6965cFn3cxbBx2byuoDzi/+YKoKY0z6RVuO04JCPZVsVtNmWzZR8OCACtvaDJ+Z1DrJwSNFnOUxw+Xl5vDRiPRmTDXGOKIFhbSOPhKRvUVkroh8ISLLRGSIu725iLwnIt+4/zZzt4uIPCAiK0RkiYikLmtahrkliaymj755V4WAcM6Ae2hz86ykAkIgS5jQrxMT+nWKGBDARhEZUxuku0+hFBiqqp+JyC7AQhF5D7gIeF9Vx4rICGAEcDNwGrC/+3ck8Kj7b8YLbQrKbRJAFTYVl0RsVx+VvzShSWhtNhTywZNXeuVvm+dx4uWPp6T+4/t2pHfnPLqNnRN1HxtFZEzNS2tQUNUfgB/cx7+JyJdAHtAL6O7u9jzwAU5Q6AW8oE6b1jwRyRWRlu55Mlb4qJuNW3ZMNCssKubGqYtYsGoDd/TukPCQ0zlPXMEfN37vlbtf/jgrm6emfT8vN8cLWrHuBmwUkTE1r9pGH4lIG6Az8AmwR8iFfh2wh/s4D1gTcthad1tGB4WqJqgpMNkNBPEGhC5rv+D1ycO98tsH/Ilrzr4loXpGEj5kNFqHd25OwDqZjakFqiUoiMjOwOvADar6q4S0Tauqikhcjd8icgVwBUDr1q1TWdVayU9buwIvfxJHQFBl5T1nVtiU6hQV2SKVRhRFGkKbE8hm9FntU/a6xpjEpT3NhYgEcALCZFV9w938o4i0dJ9vCfzkbi8E9g45vJW7rQJVfUJVu6pq1xYtWqSv8rWE37Z2v/3KZ3z5nwoB4aGjz0tLiopy1Uq//nt3zuPuPh285TPzcnNsKKoxtUhah6SKc0vwPLBBVW8I2T4e+CWko7m5qg4XkTOAa4HTcTqYH1DVI2K9RiYMSU3FKmgADcpKWXFv7wrb2v31dbYFGiV13lj8TkwzxlSvaENS09181A0YCCwVkUXutluAscCrInIpsAo4z33ubZyAsALYAlyc5vrVCcEL6piZyyp0Msfjyk+mMfKD57zy8FOv59WOp6SiejRrEmBrSXnEoBUrdbcxpvZJ9+ijD9kxZynciRH2V2BwOutUl21NYJjpLts2s3Rivwrb2g6fkbJlMXMC2dx2ptMfMH728oidyJFSdxtjaifLfVSLjcpfypRP1kTMDOrH2H88wPlLduQ7Gnje7fynbWrmAwpUmiPRu3MebUe8FXHVNpuYZkzdYEGhlkomxfVev/7Efx/dsRLaxsa70HnIlFRVLWY6imhDTm1imjF1gwWFWmrKJ2uq3imCN14cymHfL/fKPS55iOUt2qSoVlWnqo425NQmphlTN1hQSLFoGUvjFW+TUfsfv+Wt54Z45Y9bd6B//7vjft1wWUDL3Bzf78fSWxtTt1lQSKHwoaOhI2/A34Uyv6CQW95YEtfrfnH/OTQp2eaVj7zmOX7cZfcE30VF2dkS90W9d+c8CwLG1FGWOjuFuo2dEzWFw7bS8kpNKud0yWPuV+v5vqiYpjkBSsrK2bzd/1yE7t/O57lpY7zyC53P4NZTrk7uTURgKa2NqX9qap5CRok2wqaouPLcguKSMibPW+2N1Im0TzRZ5WV8N75XhW3tb3iVzY2a+D5HPGzkkDGZI+1pLjJJvCNsErlHG1DwdoWAcPsJl9Pm5lkpCQh5UepvI4eMyRwWFFJoWI925ASyK2zLCWTTrEkg6XM3LtnKynE9ufPdR7xt+w6bzjOH94pxVHyi1d9GDhmTOaz5KIWijbxZsGpDwnMOAG6Z8zRXzH/TK1959i3MPuBPSdc3VGjqahs5ZEzmsqCQYpFG3oyfvTzK3rHtvnkjCx4aWGFbsuskRxLIEi91tY0cMiazWVCoBpFGJFXlqWljOOnb+V6598D7WLRXfM04AmSJRJzzkC1CuardDRhjKrCgUA0E/53K+/6yhvef2jGs9MsWbTjtkocSel0F7juvY8QZxraGgTEmEgsKKRRpNvOCVRt8B4QPH72YVr+u98rHXvkUa3L3TLg+2SLWT2CMiYsFhRSJNJv5hqmLfB175OqlTJ0ycse5Dj6OG84clnSdgs1G1k9gjPHLgkISQu8MorXdxxRhneSO109hU84uKalftHkHxhgTjQUFHyI1CwEV7gziDQi9ls1l0qz7vPKEbhcw6ZgLEqpfw2xBFUpCFmm2+QXGmERYUKhCpGahYdMWU1qmCc1Iblhawtf3nV1h2wFD32R7g8QnuGVnZVXIo2T9BsaYRFlQqML42csrrT1cUpZYEsHrPprC0A8ne+Ubz/grbx6SfKK54pIy5n613pLWGWOSZkGhCqlIBrfr1t9ZMun8CttSPQnNktYZY1LBgkIVoi0v6df9s+6jz7K5Xrn/+Xfx8T6HpqJqFVjSOmNMKlhCvCoM69GORH7Ptypax8pxPb2A8MPOu9Hm5llxBQQBcgJV/yeyTmVjTKrYnYIr2jKavTvn+Z5vEPTWs9fT/qfvvPJJlz7Cit1bJ1Svu/scyrDXFlcYWRQqzzqVjTEpZEGB2Mto9u6cR04gi+KS8irP0/H75Ux/cahXnvvHLlzcd0yMI2LbKzfHu9iPnrHMW4inWZMAt53Z3gKBMSblLCgQeYRRcUkZ42cvZ8GqDVUHhAiT0A4f/CLrd26WcJ0CWeI1CdmMZGNMdbE+BaJnMf2+qJjJVayDcPI38yoEhKe79qLNzbOSCgi5OQHG9+1ogcAYU+0y/k4hv6AwahbTWLMRssvL+DZsneSDbpxGccPGCddFgP+NPSPh440xJlkZf6cwfvbyuGcmX7RgRoWA8H8nX0Wbm2clFRDAhpUaY2pext8pxDPpq8n2Yr6Y0LfCtj8Om055VnaUIyoL3pWE353YsFJjTG2Q8UHB7+S02/75OBcvnOmVLznnVubsd0RcrxU6fDTaEFhjjKlJGR8U2uwWOyj84bdf+PSRQV55W3aAdje96fv8gWxh/LmVO41tRJExpjbK6KAwKn8pH327Ierz1380hb+GJLDrOWgin++5n+/z23wCY0xdk7FBYVT+Ul6KMty05a/r+fjRi73yopYH0Psv9/s6r80wNsbUZRkZFPILCqMGhLv/8QD9l7zrlQ+7bjIbmjSt8pw5gWzu7tPBgoExpk7LyKAQKZdRu/Urmf3MtV551MlX89Jh/uYM2N2BMaa+qHVBQUROBSYB2cBTqjo2ledvM+KtihtUmTz1b3RbtQSArQ0a0vm6l33NOWjWJEDBraeksnrGGFOjalVQEJFs4GHgZGAtMF9EZqjqF+l4vaNWL+GVKbd45SvPvoXZB/zJ17E5gWxuO7N9OqpljDE1plYFBeAIYIWqfgcgIq8AvYCUB4XQpTG/3q01p13yIGVVTEJr4mZLtXkFxpj6qrYFhTxgTUh5LXBk+E4icgVwBUDr1omtUxB0zoB7WNjq4Cr3m9ivkwUBY0y9V9uCgi+q+gTwBEDXrl3jTV0EwIPd+vNgt/4x97EEdcaYTFPbgkIhsHdIuZW7LWV2bZTNr9vKqtzP7gyMMZmotgWF+cD+ItIWJxicD1yQyhdYMuZUDr3tnYiBYaXdFRhjMlytCgqqWioi1wKzcYakPqOqy1L9OkvGnJrqUxpjTL1Qq4ICgKq+Dbxd0/UwxphMlPGL7BhjjNnBgoIxxhiPBQVjjDEeCwrGGGM8oprQ3K9aQ0TWA6sSPHx34OcUVqcus8+iIvs8KrLPY4f68lnso6otwjfW+aCQDBFZoKpda7oetYF9FhXZ51GRfR471PfPwpqPjDHGeCwoGGOM8WR6UHiipitQi9hnUZF9HhXZ57FDvf4sMrpPwRhjTEWZfqdgjDEmhAUFY4wxnowMCiJyqogsF5EVIjKiputTHURkbxGZKyJfiMgyERnibm8uIu+JyDfuv83c7SIiD7if0RIROaxm30HqiUi2iBSIyCy33FZEPnHf81QRaehub+SWV7jPt6nRiqeBiOSKyDQR+UpEvhSRozP8u3Gj+//J5yIyRUQaZ8r3I+OCgohkAw8DpwEHA/1FpOr1OOu+UmCoqh4MHAUMdt/3COB9Vd0feN8tg/P57O/+XQE8Wv1VTrshwJch5XHABFXdD9gIXOpuvxTY6G6f4O5X30wC3lHVA4GOOJ9LRn43RCQPuB7oqqqH4KTxP59M+X6oakb9AUcDs0PKI4GRNV2vGvgcpgMnA8uBlu62lsBy9/HjQP+Q/b396sMfzqp+7wMnALNwVl/9GWgQ/j3BWd/jaPdxA3c/qen3kMLPoinwv/D3lMHfjeBa8c3d/96zgB6Z8v3IuDsFdvwHD1rrbssY7u1tZ+ATYA9V/cF9ah2wh/u4vn9OE4HhQLlb3g0oUtVStxz6fr3Pwn1+k7t/fdEWWA886zanPSUiO5Gh3w1VLQTuBVYDP+D8915Ihnw/MjEoZDQR2Rl4HbhBVX8NfU6dnzr1foyyiPQEflLVhTVdl1qiAXAY8KiqdgY2s6OpCMic7waA23fSCydY7gXsBGTMco2ZGBQKgb1Dyq3cbfWeiARwAsJkVX3D3fyjiLR0n28J/ORur8+fUzfgLBFZCbyC04Q0CcgVkeBqhKHv1/ss3OebAr9UZ4XTbC2wVlU/ccvTcIJEJn43AE4C/qeq61W1BHgD5zuTEd+PTAwK84H93ZEEDXE6kGbUcJ3STkQEeBr4UlXvD3lqBjDIfTwIp68huP0v7kiTo4BNIU0JdZqqjlTVVqraBue//xxVHQDMBc51dwv/LIKf0bnu/vXmV7OqrgPWiEg7d9OJwBdk4HfDtRo4SkSauP/fBD+PzPh+1HSnRk38AacDXwPfAn+r6fpU03s+Buf2fwmwyP07Haft833gG+CfQHN3f8EZpfUtsBRnJEaNv480fC7dgVnu4z8CnwIrgNeARu72xm55hfv8H2u63mn4HDoBC9zvRz7QLJO/G8AY4Cvgc+BFoFGmfD8szYUxxhhPJjYfGWOMicKCgjHGGI8FBWOMMR4LCsYYYzwWFIwxxngsKBhjjPFYUDAmSSLSRkQuSPDY/6a6PsYkw4KCMclrA0QMCiFpESJS1T+lo0LGJMomrxkThYjcDmxQ1Ylu+U6cRHqTwvabBxyEk376eZxc+32AnXFy8Z+BkxKhGRAARqnqdPfY31V1ZxHpDozGSbt8CE5WzgvV/gc11cyCgjFRuCnG31DVw0QkCyfdwxGq+kvYft2Bm1S1p1u+CLgDOFRVN7h3C01U9VcR2R2YB+yvqhoWFKYD7YHvgY+AYar6YfrfqTE7xLy1NSaTqepKEflFRDrjrCVQEB4QYnhPVTe4jwW4S0T+jLN+Q557vnVhx3yqqmsBRGQRTrOUBQVTrSwoGBPbU8BFwJ7AM3Ectznk8QCgBdBFVUvclN2NIxyzLeRxGfb/p6kB1tFsTGxv4iywcjjOsouR/AbsEuMcTXH6IkpE5Hhgn9RW0ZjUsV8ixsSgqttFZC7OUoxlUXZbApSJyGLgOZyO5lCTgZkishQnPfVX6aqvMcmyjmZjYnA7mD8D+qrqNzVdH2PSzZqPjIlCRA7GWTjlfQsIJlPYnYIxPolIB5xVuEJtU9Uja6I+xqSDBQVjjDEeaz4yxhjjsaBgjDHGY0HBGGOMx4KCMcYYz/8DlI9rnLkVbWUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 散布図を描画(真値 vs 予測値)\n",
    "plt.plot(y_test, y_test, color = 'red', label = 'x=y') # 直線y = x (真値と予測値が同じ場合は直線状に点がプロットされる)\n",
    "plt.scatter(y_test, y_pred) # 散布図のプロット\n",
    "plt.xlabel('y_train') # x軸ラベル\n",
    "plt.ylabel('y_test_pred') # y軸ラベル\n",
    "plt.title('y_train vs y_test_pred') # グラフタイトル\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1236,
   "id": "edd8ab3f-d236-438f-ba73-5794d14da932",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test score:  0.1990635935326299\n",
      "train score:  0.186714685386302\n"
     ]
    }
   ],
   "source": [
    "print(\"test score: \",rmsle(y_test, model.predict(X_test)))\n",
    "print(\"train score: \",rmsle(y_train, model.predict(X_train)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1237,
   "id": "d42978e2-093f-446f-898c-c9a3a7ceac26",
   "metadata": {},
   "outputs": [],
   "source": [
    "# y_t = pd.DataFrame(model.predict(X_train))\n",
    "# y_t.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1238,
   "id": "d889d55b-b807-4f4a-815e-5e18d77ae3d2",
   "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>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>2722.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>194.157506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>179.912019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>2.045482</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>44.826894</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>151.370779</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>287.455982</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>907.515186</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 0\n",
       "count  2722.000000\n",
       "mean    194.157506\n",
       "std     179.912019\n",
       "min       2.045482\n",
       "25%      44.826894\n",
       "50%     151.370779\n",
       "75%     287.455982\n",
       "max     907.515186"
      ]
     },
     "execution_count": 1238,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_p = pd.DataFrame(model.predict(X_test))\n",
    "y_p.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1239,
   "id": "24e6533e-92c5-40b9-9c95-649f4795a884",
   "metadata": {},
   "outputs": [],
   "source": [
    "# y_p"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1240,
   "id": "6593b135-832d-4d6b-a962-e71e3d0c3f8f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 1240,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_p.loc[y_p[0]<0].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1241,
   "id": "fb060501-9591-4f15-a1ea-3bdb55986079",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_pred = model.predict(test)\n",
    "test_pred = pd.DataFrame(test_pred, columns=['count'])\n",
    "test_pred['count'] = test_pred['count'].astype('int64')\n",
    "\n",
    "submission  = pd.read_csv('input/sampleSubmission.csv')\n",
    "submission = pd.DataFrame(submission['datetime'])\n",
    "submission = pd.concat([submission, test_pred], axis=1)\n",
    "\n",
    "# submission['count'].sum()/submission['count'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1242,
   "id": "c0f457c3-a56b-4844-a781-3110775fa180",
   "metadata": {},
   "outputs": [],
   "source": [
    "# df['count'].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1243,
   "id": "0dc0a1c7-a4fe-4ca6-bee9-d779b0b03e13",
   "metadata": {},
   "outputs": [],
   "source": [
    "submission.loc[submission['count']<0, 'count'].count()\n",
    "submission.loc[submission['count']<0, 'count'] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1244,
   "id": "40a5450d-f2a9-449c-9f4a-6d27a6e070c2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# submission.loc[submission['count']<0, 'count'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1245,
   "id": "81687ca7-00c7-4e59-83b9-5717ce8db67e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# submission.head(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1246,
   "id": "0a59f49d-928d-419a-86f1-30f3d4583416",
   "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>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>6493.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>64.574311</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>16.501038</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>21.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>56.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>63.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>76.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>109.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             count\n",
       "count  6493.000000\n",
       "mean     64.574311\n",
       "std      16.501038\n",
       "min      21.000000\n",
       "25%      56.000000\n",
       "50%      63.000000\n",
       "75%      76.000000\n",
       "max     109.000000"
      ]
     },
     "execution_count": 1246,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1247,
   "id": "ea8aeaec-40cf-4d43-a8e7-5fefa99e2607",
   "metadata": {},
   "outputs": [],
   "source": [
    "submission.to_csv('output/submission.csv',index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8b80fa51-408d-4bf2-8416-6e5847737a2c",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}