LightGBM-poutcome.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 【第1回_Beginner限定コンペ】銀行の顧客ターゲティング\n",
    "\n",
    "顧客の属性情報などから定期預金キャンペーンの反応率を予測しよう。\n",
    "\n",
    "https://signate.jp/competitions/292\n",
    "\n",
    "LightGBM編\n",
    "\n",
    "前処理として、poutcome の unknown を予測してみる。\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ライブラリインポート"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "\n",
    "#import lightgbm as lgb\n",
    "from optuna.integration import lightgbm as lgb\n",
    "#import featuretools as ft\n",
    "\n",
    "#from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score, recall_score, precision_score, f1_score"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## データ読み込み・前処理"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = pd.read_csv('train.csv')\n",
    "test_df = pd.read_csv('test.csv')\n",
    "submit_df = pd.read_csv('submit_sample.csv', header=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(27100, 18)\n",
      "(18050, 17)\n"
     ]
    }
   ],
   "source": [
    "print(train_df.shape)\n",
    "print(test_df.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(45150, 18)\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 45150 entries, 0 to 18049\n",
      "Data columns (total 18 columns):\n",
      " #   Column     Non-Null Count  Dtype  \n",
      "---  ------     --------------  -----  \n",
      " 0   id         45150 non-null  int64  \n",
      " 1   age        45150 non-null  int64  \n",
      " 2   job        45150 non-null  object \n",
      " 3   marital    45150 non-null  object \n",
      " 4   education  45150 non-null  object \n",
      " 5   default    45150 non-null  object \n",
      " 6   balance    45150 non-null  int64  \n",
      " 7   housing    45150 non-null  object \n",
      " 8   loan       45150 non-null  object \n",
      " 9   contact    45150 non-null  object \n",
      " 10  day        45150 non-null  int64  \n",
      " 11  month      45150 non-null  object \n",
      " 12  duration   45150 non-null  int64  \n",
      " 13  campaign   45150 non-null  int64  \n",
      " 14  pdays      45150 non-null  int64  \n",
      " 15  previous   45150 non-null  int64  \n",
      " 16  poutcome   45150 non-null  object \n",
      " 17  y          27100 non-null  float64\n",
      "dtypes: float64(1), int64(8), object(9)\n",
      "memory usage: 6.5+ MB\n"
     ]
    },
    {
     "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>id</th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>balance</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>day</th>\n",
       "      <th>month</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>18045</th>\n",
       "      <td>18045</td>\n",
       "      <td>49</td>\n",
       "      <td>self-employed</td>\n",
       "      <td>married</td>\n",
       "      <td>tertiary</td>\n",
       "      <td>no</td>\n",
       "      <td>98357</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>6</td>\n",
       "      <td>jul</td>\n",
       "      <td>101</td>\n",
       "      <td>2</td>\n",
       "      <td>417</td>\n",
       "      <td>0</td>\n",
       "      <td>failure</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18046</th>\n",
       "      <td>18046</td>\n",
       "      <td>34</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>married</td>\n",
       "      <td>secondary</td>\n",
       "      <td>no</td>\n",
       "      <td>29621</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>12</td>\n",
       "      <td>may</td>\n",
       "      <td>345</td>\n",
       "      <td>1</td>\n",
       "      <td>815</td>\n",
       "      <td>0</td>\n",
       "      <td>unknown</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18047</th>\n",
       "      <td>18047</td>\n",
       "      <td>34</td>\n",
       "      <td>admin.</td>\n",
       "      <td>single</td>\n",
       "      <td>secondary</td>\n",
       "      <td>no</td>\n",
       "      <td>94260</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>unknown</td>\n",
       "      <td>16</td>\n",
       "      <td>may</td>\n",
       "      <td>121</td>\n",
       "      <td>2</td>\n",
       "      <td>370</td>\n",
       "      <td>0</td>\n",
       "      <td>unknown</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18048</th>\n",
       "      <td>18048</td>\n",
       "      <td>31</td>\n",
       "      <td>technician</td>\n",
       "      <td>single</td>\n",
       "      <td>secondary</td>\n",
       "      <td>no</td>\n",
       "      <td>65483</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>unknown</td>\n",
       "      <td>15</td>\n",
       "      <td>may</td>\n",
       "      <td>345</td>\n",
       "      <td>2</td>\n",
       "      <td>41</td>\n",
       "      <td>0</td>\n",
       "      <td>unknown</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18049</th>\n",
       "      <td>18049</td>\n",
       "      <td>30</td>\n",
       "      <td>retired</td>\n",
       "      <td>married</td>\n",
       "      <td>primary</td>\n",
       "      <td>no</td>\n",
       "      <td>6474</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>13</td>\n",
       "      <td>aug</td>\n",
       "      <td>158</td>\n",
       "      <td>1</td>\n",
       "      <td>699</td>\n",
       "      <td>0</td>\n",
       "      <td>unknown</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          id  age            job  marital  education default  balance housing  \\\n",
       "18045  18045   49  self-employed  married   tertiary      no    98357     yes   \n",
       "18046  18046   34    blue-collar  married  secondary      no    29621     yes   \n",
       "18047  18047   34         admin.   single  secondary      no    94260     yes   \n",
       "18048  18048   31     technician   single  secondary      no    65483     yes   \n",
       "18049  18049   30        retired  married    primary      no     6474      no   \n",
       "\n",
       "      loan   contact  day month  duration  campaign  pdays  previous poutcome  \\\n",
       "18045   no  cellular    6   jul       101         2    417         0  failure   \n",
       "18046   no  cellular   12   may       345         1    815         0  unknown   \n",
       "18047   no   unknown   16   may       121         2    370         0  unknown   \n",
       "18048   no   unknown   15   may       345         2     41         0  unknown   \n",
       "18049   no  cellular   13   aug       158         1    699         0  unknown   \n",
       "\n",
       "        y  \n",
       "18045 NaN  \n",
       "18046 NaN  \n",
       "18047 NaN  \n",
       "18048 NaN  \n",
       "18049 NaN  "
      ]
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "merge_df = pd.concat([train_df, test_df])\n",
    "\n",
    "print(merge_df.shape)\n",
    "merge_df.info()\n",
    "merge_df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_dict = {}\n",
    "for name, group in merge_df.groupby('poutcome'):\n",
    "    df_dict[name] = group\n",
    "    \n",
    "p_train_df = pd.concat([df_dict['success'], df_dict['failure'], df_dict['other']])\n",
    "p_test_df = df_dict['unknown']\n",
    "\n",
    "py = p_train_df.pop('y')\n",
    "pz = p_test_df.pop('y')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6625, 17)\n",
      "(38525, 17)\n"
     ]
    }
   ],
   "source": [
    "print(p_train_df.shape)\n",
    "print(p_test_df.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "データメモ\n",
    "\n",
    "- testデータの `job` に `unknown` が一人いる (trainデータにはいない)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [],
   "source": [
    "# one hot encoding\n",
    "#train_df = pd.get_dummies(train_df, columns=['job', 'marital', 'education', 'contact', 'poutcome'])\n",
    "#test_df = pd.get_dummies(test_df, columns=['job', 'marital', 'education', 'contact', 'poutcome'])\n",
    "#test_df.drop(['job_unknown'], axis=1, inplace=True)\n",
    "\n",
    "df_list = [p_train_df, p_test_df]\n",
    "\n",
    "#job_encoder = LabelEncoder()\n",
    "#job_encoder.fit(test_df['job'])\n",
    "\n",
    "for df in df_list:\n",
    "    #df['job'] = job_encoder.transform(df['job'])\n",
    "    df['job'] = df['job'].map({'unknown': 1, 'technician': 2, 'blue-collar': 3, 'services': 4, 'entrepreneur': 5, 'admin.': 6, 'management': 7, 'housemaid': 8, 'self-employed': 9, 'unemployed': 10, 'retired': 11, 'student': 12})\n",
    "\n",
    "    df['marital'] = df['marital'].map({'married': 2, 'divorced':1, 'single': 0})\n",
    "    #df.drop(['marital'], axis=1, inplace=True)\n",
    "\n",
    "    df['education'] = df['education'].map({'tertiary': 3, 'secondary': 2, 'primary': 1, 'unknown': 0})\n",
    "    #df.drop(['education'], axis=1, inplace=True)\n",
    "\n",
    "    #df['default'] = df['default'].map({'yes': 1, 'no': 0})\n",
    "    df.drop(['default'], axis=1, inplace=True)\n",
    "\n",
    "    df['housing'] = df['housing'].map({'yes': 1, 'no': 0})\n",
    "    #df.drop(['housing'], axis=1, inplace=True)\n",
    "    \n",
    "    df['loan'] = df['loan'].map({'yes': 1, 'no': 0})\n",
    "    #df.drop(['loan'], axis=1, inplace=True)\n",
    "\n",
    "    df['contact'] = df['contact'].map({'telephone': 2, 'cellular': 1, 'unknown': 0})\n",
    "    #df.drop(['contact'], axis=1, inplace=True)\n",
    "\n",
    "    df['poutcome'] = df['poutcome'].map({'success': 1, 'unknown': 2, 'failure': 0, 'other': 0})\n",
    "    \n",
    "    df['month'] = df['month'].map({'jan': 1, 'feb': 2, 'mar': 3, 'apr': 4, 'may': 5, 'jun': 6, 'jul': 7, 'aug': 8, 'sep': 9, 'oct': 10, 'nov': 11, 'dec': 12})\n",
    "    #df.drop(['day', 'month'], axis=1, inplace=True)\n",
    "    \n",
    "    # Feb 30 とかあって、正確には変換できない\n",
    "    # → データの Feb 30 を Mar 1 に変換した(他にも、 2/31, 6/31, 11/31)\n",
    "    #df['days'] = pd.to_datetime('2016/' + df['month'].astype(str).str.pad(2,fillchar='0') + '/' + df['day'].astype(str).str.pad(2,fillchar='0'), format='%Y/%m/%d').dt.strftime('%j')\n",
    "    df['days'] = df['month'] * 31 + df['day']\n",
    "    \n",
    "    df['duration'] = np.log(df['duration'] + 1)\n",
    "    \n",
    "    df.drop(['id'], axis=1, inplace=True)\n",
    "    \n",
    "# featuretools用\n",
    "#y = train_df.pop('y')\n",
    "#train_df['train'] = 1\n",
    "#test_df['train'] = 0\n",
    "#test_df['id'] = test_df['id'] + len(train_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 6625 entries, 8 to 17975\n",
      "Data columns (total 16 columns):\n",
      " #   Column     Non-Null Count  Dtype  \n",
      "---  ------     --------------  -----  \n",
      " 0   age        6625 non-null   int64  \n",
      " 1   job        6625 non-null   int64  \n",
      " 2   marital    6625 non-null   int64  \n",
      " 3   education  6625 non-null   int64  \n",
      " 4   balance    6625 non-null   int64  \n",
      " 5   housing    6625 non-null   int64  \n",
      " 6   loan       6625 non-null   int64  \n",
      " 7   contact    6625 non-null   int64  \n",
      " 8   day        6625 non-null   int64  \n",
      " 9   month      6625 non-null   int64  \n",
      " 10  duration   6625 non-null   float64\n",
      " 11  campaign   6625 non-null   int64  \n",
      " 12  pdays      6625 non-null   int64  \n",
      " 13  previous   6625 non-null   int64  \n",
      " 14  poutcome   6625 non-null   int64  \n",
      " 15  days       6625 non-null   int64  \n",
      "dtypes: float64(1), int64(15)\n",
      "memory usage: 879.9 KB\n"
     ]
    }
   ],
   "source": [
    "p_train_df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 38525 entries, 1 to 18049\n",
      "Data columns (total 16 columns):\n",
      " #   Column     Non-Null Count  Dtype  \n",
      "---  ------     --------------  -----  \n",
      " 0   age        38525 non-null  int64  \n",
      " 1   job        38525 non-null  int64  \n",
      " 2   marital    38525 non-null  int64  \n",
      " 3   education  38525 non-null  int64  \n",
      " 4   balance    38525 non-null  int64  \n",
      " 5   housing    38525 non-null  int64  \n",
      " 6   loan       38525 non-null  int64  \n",
      " 7   contact    38525 non-null  int64  \n",
      " 8   day        38525 non-null  int64  \n",
      " 9   month      38525 non-null  int64  \n",
      " 10  duration   38525 non-null  float64\n",
      " 11  campaign   38525 non-null  int64  \n",
      " 12  pdays      38525 non-null  int64  \n",
      " 13  previous   38525 non-null  int64  \n",
      " 14  poutcome   38525 non-null  int64  \n",
      " 15  days       38525 non-null  int64  \n",
      "dtypes: float64(1), int64(15)\n",
      "memory usage: 5.0 MB\n"
     ]
    }
   ],
   "source": [
    "p_test_df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "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>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>balance</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>day</th>\n",
       "      <th>month</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>days</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "      <td>6625.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>36.192906</td>\n",
       "      <td>5.418868</td>\n",
       "      <td>1.248151</td>\n",
       "      <td>2.110340</td>\n",
       "      <td>48274.351849</td>\n",
       "      <td>0.622491</td>\n",
       "      <td>0.067623</td>\n",
       "      <td>1.049962</td>\n",
       "      <td>14.413585</td>\n",
       "      <td>5.677736</td>\n",
       "      <td>4.935661</td>\n",
       "      <td>1.648151</td>\n",
       "      <td>436.444981</td>\n",
       "      <td>0.576453</td>\n",
       "      <td>190.423396</td>\n",
       "      <td>0.114717</td>\n",
       "      <td>-0.325132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>7.705336</td>\n",
       "      <td>2.794670</td>\n",
       "      <td>0.921597</td>\n",
       "      <td>0.761024</td>\n",
       "      <td>31589.561041</td>\n",
       "      <td>0.484801</td>\n",
       "      <td>0.251116</td>\n",
       "      <td>0.287224</td>\n",
       "      <td>8.443518</td>\n",
       "      <td>2.925374</td>\n",
       "      <td>0.836736</td>\n",
       "      <td>0.824669</td>\n",
       "      <td>251.081981</td>\n",
       "      <td>0.784582</td>\n",
       "      <td>91.229310</td>\n",
       "      <td>0.318704</td>\n",
       "      <td>0.601134</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>22.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-6847.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2.639057</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>34.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>32.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>20829.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>4.624973</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>222.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>139.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>33.000000</td>\n",
       "      <td>6.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>48720.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>4.624973</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>440.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>166.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>37.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>75366.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>5.068904</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>651.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>226.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>77.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>102117.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>31.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>7.277939</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>870.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>373.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               age          job      marital    education        balance  \\\n",
       "count  6625.000000  6625.000000  6625.000000  6625.000000    6625.000000   \n",
       "mean     36.192906     5.418868     1.248151     2.110340   48274.351849   \n",
       "std       7.705336     2.794670     0.921597     0.761024   31589.561041   \n",
       "min      22.000000     2.000000     0.000000     0.000000   -6847.000000   \n",
       "25%      32.000000     3.000000     0.000000     2.000000   20829.000000   \n",
       "50%      33.000000     6.000000     2.000000     2.000000   48720.000000   \n",
       "75%      37.000000     7.000000     2.000000     3.000000   75366.000000   \n",
       "max      77.000000    12.000000     2.000000     3.000000  102117.000000   \n",
       "\n",
       "           housing         loan      contact          day        month  \\\n",
       "count  6625.000000  6625.000000  6625.000000  6625.000000  6625.000000   \n",
       "mean      0.622491     0.067623     1.049962    14.413585     5.677736   \n",
       "std       0.484801     0.251116     0.287224     8.443518     2.925374   \n",
       "min       0.000000     0.000000     0.000000     1.000000     1.000000   \n",
       "25%       0.000000     0.000000     1.000000     7.000000     4.000000   \n",
       "50%       1.000000     0.000000     1.000000    12.000000     5.000000   \n",
       "75%       1.000000     0.000000     1.000000    20.000000     7.000000   \n",
       "max       1.000000     1.000000     2.000000    31.000000    12.000000   \n",
       "\n",
       "          duration     campaign        pdays     previous         days  \\\n",
       "count  6625.000000  6625.000000  6625.000000  6625.000000  6625.000000   \n",
       "mean      4.935661     1.648151   436.444981     0.576453   190.423396   \n",
       "std       0.836736     0.824669   251.081981     0.784582    91.229310   \n",
       "min       2.639057     1.000000    -1.000000     0.000000    34.000000   \n",
       "25%       4.624973     1.000000   222.000000     0.000000   139.000000   \n",
       "50%       4.624973     1.000000   440.000000     0.000000   166.000000   \n",
       "75%       5.068904     2.000000   651.000000     1.000000   226.000000   \n",
       "max       7.277939     4.000000   870.000000     3.000000   373.000000   \n",
       "\n",
       "          poutcome            y  \n",
       "count  6625.000000  6625.000000  \n",
       "mean      0.114717    -0.325132  \n",
       "std       0.318704     0.601134  \n",
       "min       0.000000    -1.000000  \n",
       "25%       0.000000    -1.000000  \n",
       "50%       0.000000     0.000000  \n",
       "75%       0.000000     0.000000  \n",
       "max       1.000000     1.000000  "
      ]
     },
     "execution_count": 159,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "p_train_df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "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>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>balance</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>day</th>\n",
       "      <th>month</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>days</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "      <td>38525.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>36.067229</td>\n",
       "      <td>5.091655</td>\n",
       "      <td>1.408021</td>\n",
       "      <td>2.031979</td>\n",
       "      <td>47335.473381</td>\n",
       "      <td>0.573991</td>\n",
       "      <td>0.138144</td>\n",
       "      <td>0.741077</td>\n",
       "      <td>17.092537</td>\n",
       "      <td>6.057313</td>\n",
       "      <td>5.189741</td>\n",
       "      <td>1.795094</td>\n",
       "      <td>433.529552</td>\n",
       "      <td>0.003738</td>\n",
       "      <td>204.869254</td>\n",
       "      <td>0.089896</td>\n",
       "      <td>-0.357794</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>7.857411</td>\n",
       "      <td>2.635810</td>\n",
       "      <td>0.862004</td>\n",
       "      <td>0.716055</td>\n",
       "      <td>31575.694717</td>\n",
       "      <td>0.494501</td>\n",
       "      <td>0.345056</td>\n",
       "      <td>0.515824</td>\n",
       "      <td>8.539274</td>\n",
       "      <td>1.953683</td>\n",
       "      <td>0.758862</td>\n",
       "      <td>0.965967</td>\n",
       "      <td>251.984413</td>\n",
       "      <td>0.079823</td>\n",
       "      <td>60.250935</td>\n",
       "      <td>0.080589</td>\n",
       "      <td>0.561274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>20.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-6846.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>34.000000</td>\n",
       "      <td>0.001354</td>\n",
       "      <td>-1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>31.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>19666.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>9.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>4.905275</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>215.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>171.000000</td>\n",
       "      <td>0.035111</td>\n",
       "      <td>-1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>33.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>47078.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>18.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.068904</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>433.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>182.000000</td>\n",
       "      <td>0.061891</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>37.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>74815.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>26.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>5.846439</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>652.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>234.000000</td>\n",
       "      <td>0.122435</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>90.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>102124.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>31.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>8.031710</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>870.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>373.000000</td>\n",
       "      <td>0.811811</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                age           job       marital     education        balance  \\\n",
       "count  38525.000000  38525.000000  38525.000000  38525.000000   38525.000000   \n",
       "mean      36.067229      5.091655      1.408021      2.031979   47335.473381   \n",
       "std        7.857411      2.635810      0.862004      0.716055   31575.694717   \n",
       "min       20.000000      1.000000      0.000000      0.000000   -6846.000000   \n",
       "25%       31.000000      3.000000      0.000000      2.000000   19666.000000   \n",
       "50%       33.000000      4.000000      2.000000      2.000000   47078.000000   \n",
       "75%       37.000000      7.000000      2.000000      2.000000   74815.000000   \n",
       "max       90.000000     12.000000      2.000000      3.000000  102124.000000   \n",
       "\n",
       "            housing          loan       contact           day         month  \\\n",
       "count  38525.000000  38525.000000  38525.000000  38525.000000  38525.000000   \n",
       "mean       0.573991      0.138144      0.741077     17.092537      6.057313   \n",
       "std        0.494501      0.345056      0.515824      8.539274      1.953683   \n",
       "min        0.000000      0.000000      0.000000      1.000000      1.000000   \n",
       "25%        0.000000      0.000000      0.000000      9.000000      5.000000   \n",
       "50%        1.000000      0.000000      1.000000     18.000000      5.000000   \n",
       "75%        1.000000      0.000000      1.000000     26.000000      7.000000   \n",
       "max        1.000000      1.000000      2.000000     31.000000     12.000000   \n",
       "\n",
       "           duration      campaign         pdays      previous          days  \\\n",
       "count  38525.000000  38525.000000  38525.000000  38525.000000  38525.000000   \n",
       "mean       5.189741      1.795094    433.529552      0.003738    204.869254   \n",
       "std        0.758862      0.965967    251.984413      0.079823     60.250935   \n",
       "min        0.000000      1.000000     -1.000000      0.000000     34.000000   \n",
       "25%        4.905275      1.000000    215.000000      0.000000    171.000000   \n",
       "50%        5.068904      1.000000    433.000000      0.000000    182.000000   \n",
       "75%        5.846439      2.000000    652.000000      0.000000    234.000000   \n",
       "max        8.031710      5.000000    870.000000      3.000000    373.000000   \n",
       "\n",
       "           poutcome             y  \n",
       "count  38525.000000  38525.000000  \n",
       "mean       0.089896     -0.357794  \n",
       "std        0.080589      0.561274  \n",
       "min        0.001354     -1.000000  \n",
       "25%        0.035111     -1.000000  \n",
       "50%        0.061891      0.000000  \n",
       "75%        0.122435      0.000000  \n",
       "max        0.811811      1.000000  "
      ]
     },
     "execution_count": 160,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "p_test_df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [],
   "source": [
    "#train_pg = sns.pairplot(train_df.drop(['month', 'day'], axis=1), hue='y')\n",
    "#train_pg = sns.pairplot(p_train_df, hue='poutcome')\n",
    "#train_pg.savefig('p_train_df.png')\n",
    "\n",
    "#test_pg = sns.pairplot(p_test_df)\n",
    "#test_pg.savefig('p_test_df.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 学習準備"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = p_train_df.pop('poutcome')\n",
    "z = p_test_df.pop('poutcome')\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(p_train_df, y, test_size=0.2, random_state=9)\n",
    "\n",
    "train_data = lgb.Dataset(data=X_train, label=y_train, free_raw_data=False)\n",
    "test_data = lgb.Dataset(data=X_test, label=y_test, free_raw_data=False)\n",
    "final_train_set = lgb.Dataset(data=p_train_df, label=y, free_raw_data=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ハイパーパラメータ"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [],
   "source": [
    "lgb_params = {\n",
    "    'boosting': 'dart',          # dart (drop out trees) often performs better\n",
    "    'application': 'binary',     # Binary classification\n",
    "    'metric': 'binary_logloss',  # Area under ROC curve as the evaulation metric\n",
    "    # 以下はoptunaでチューニングする\n",
    "    #'learning_rate': 0.05,       # Learning rate, controls size of a gradient descent step\n",
    "    #'min_data_in_leaf': 20,      # Data set is quite small so reduce this a bit\n",
    "    #'feature_fraction': 0.7,     # Proportion of features in each boost, controls overfitting\n",
    "    #'num_leaves': 41,            # Controls size of tree since LGBM uses leaf wise splits\n",
    "    #'drop_rate': 0.15\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 学習"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "feature_fraction, val_score: 0.271679:  14%|#4        | 1/7 [00:02<00:17,  2.87s/it][I 2020-08-12 07:52:04,204] Trial 0 finished with value: 0.2716790446553251 and parameters: {'feature_fraction': 0.6}. Best is trial 0 with value: 0.2716790446553251.\n",
      "feature_fraction, val_score: 0.271679:  29%|##8       | 2/7 [00:05<00:14,  2.90s/it][I 2020-08-12 07:52:07,179] Trial 1 finished with value: 0.2784328297421988 and parameters: {'feature_fraction': 0.8999999999999999}. Best is trial 0 with value: 0.2716790446553251.\n",
      "feature_fraction, val_score: 0.271679:  43%|####2     | 3/7 [00:08<00:11,  2.94s/it][I 2020-08-12 07:52:10,218] Trial 2 finished with value: 0.27887856985374326 and parameters: {'feature_fraction': 0.8}. Best is trial 0 with value: 0.2716790446553251.\n",
      "feature_fraction, val_score: 0.271679:  57%|#####7    | 4/7 [00:11<00:08,  2.96s/it][I 2020-08-12 07:52:13,222] Trial 3 finished with value: 0.2779634968585597 and parameters: {'feature_fraction': 0.7}. Best is trial 0 with value: 0.2716790446553251.\n",
      "feature_fraction, val_score: 0.271679:  71%|#######1  | 5/7 [00:15<00:06,  3.01s/it][I 2020-08-12 07:52:16,337] Trial 4 finished with value: 0.280625884561762 and parameters: {'feature_fraction': 1.0}. Best is trial 0 with value: 0.2716790446553251.\n",
      "feature_fraction, val_score: 0.262413:  86%|########5 | 6/7 [00:17<00:02,  2.96s/it][I 2020-08-12 07:52:19,183] Trial 5 finished with value: 0.262412574731139 and parameters: {'feature_fraction': 0.4}. Best is trial 5 with value: 0.262412574731139.\n",
      "feature_fraction, val_score: 0.262413: 100%|##########| 7/7 [00:20<00:00,  2.93s/it][I 2020-08-12 07:52:22,046] Trial 6 finished with value: 0.27586889709287155 and parameters: {'feature_fraction': 0.5}. Best is trial 5 with value: 0.262412574731139.\n",
      "feature_fraction, val_score: 0.262413: 100%|##########| 7/7 [00:20<00:00,  2.96s/it]\n",
      "num_leaves, val_score: 0.262413:   5%|5         | 1/20 [00:04<01:16,  4.05s/it][I 2020-08-12 07:52:26,102] Trial 7 finished with value: 0.3116783956204902 and parameters: {'num_leaves': 55}. Best is trial 7 with value: 0.3116783956204902.\n",
      "num_leaves, val_score: 0.237706:  10%|#         | 2/20 [00:05<00:59,  3.28s/it][I 2020-08-12 07:52:27,596] Trial 8 finished with value: 0.23770643797531005 and parameters: {'num_leaves': 8}. Best is trial 8 with value: 0.23770643797531005.\n",
      "num_leaves, val_score: 0.237706:  15%|#5        | 3/20 [00:14<01:24,  4.97s/it][I 2020-08-12 07:52:36,504] Trial 9 finished with value: 0.49961395474126236 and parameters: {'num_leaves': 250}. Best is trial 8 with value: 0.23770643797531005.\n",
      "num_leaves, val_score: 0.234727:  20%|##        | 4/20 [00:15<01:02,  3.89s/it][I 2020-08-12 07:52:37,872] Trial 10 finished with value: 0.23472670089892847 and parameters: {'num_leaves': 6}. Best is trial 10 with value: 0.23472670089892847.\n",
      "num_leaves, val_score: 0.234035:  25%|##5       | 5/20 [00:17<00:46,  3.12s/it][I 2020-08-12 07:52:39,209] Trial 11 finished with value: 0.23403479762828833 and parameters: {'num_leaves': 5}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  30%|###       | 6/20 [00:18<00:38,  2.72s/it][I 2020-08-12 07:52:41,003] Trial 12 finished with value: 0.24245107285605935 and parameters: {'num_leaves': 12}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  35%|###5      | 7/20 [00:25<00:49,  3.80s/it][I 2020-08-12 07:52:47,305] Trial 13 finished with value: 0.4034218852179036 and parameters: {'num_leaves': 101}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  40%|####      | 8/20 [00:34<01:04,  5.34s/it][I 2020-08-12 07:52:56,252] Trial 14 finished with value: 0.4940046641392044 and parameters: {'num_leaves': 203}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  45%|####5     | 9/20 [00:35<00:46,  4.26s/it][I 2020-08-12 07:52:57,982] Trial 15 finished with value: 0.24403425107078103 and parameters: {'num_leaves': 11}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  50%|#####     | 10/20 [00:40<00:44,  4.46s/it][I 2020-08-12 07:53:02,917] Trial 16 finished with value: 0.34557214397236147 and parameters: {'num_leaves': 72}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  55%|#####5    | 11/20 [00:49<00:50,  5.61s/it][I 2020-08-12 07:53:11,219] Trial 17 finished with value: 0.5096785198955895 and parameters: {'num_leaves': 169}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  60%|######    | 12/20 [00:52<00:39,  4.94s/it][I 2020-08-12 07:53:14,590] Trial 18 finished with value: 0.28116268463818667 and parameters: {'num_leaves': 41}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  65%|######5   | 13/20 [00:59<00:38,  5.56s/it][I 2020-08-12 07:53:21,598] Trial 19 finished with value: 0.4517321966288653 and parameters: {'num_leaves': 124}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  70%|#######   | 14/20 [01:00<00:25,  4.27s/it][I 2020-08-12 07:53:22,868] Trial 20 finished with value: 0.23403479762828833 and parameters: {'num_leaves': 5}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  75%|#######5  | 15/20 [01:02<00:16,  3.38s/it][I 2020-08-12 07:53:24,158] Trial 21 finished with value: 0.23403479762828833 and parameters: {'num_leaves': 5}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  80%|########  | 16/20 [01:05<00:13,  3.36s/it][I 2020-08-12 07:53:27,487] Trial 22 finished with value: 0.27955523979665836 and parameters: {'num_leaves': 40}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.234035:  85%|########5 | 17/20 [01:10<00:11,  3.96s/it][I 2020-08-12 07:53:32,845] Trial 23 finished with value: 0.36759508909965344 and parameters: {'num_leaves': 83}. Best is trial 11 with value: 0.23403479762828833.\n",
      "num_leaves, val_score: 0.233470:  90%|######### | 18/20 [01:11<00:06,  3.09s/it][I 2020-08-12 07:53:33,893] Trial 24 finished with value: 0.23346962042716426 and parameters: {'num_leaves': 3}. Best is trial 24 with value: 0.23346962042716426.\n",
      "num_leaves, val_score: 0.233470:  95%|#########5| 19/20 [01:14<00:03,  3.07s/it][I 2020-08-12 07:53:36,936] Trial 25 finished with value: 0.2662476207710006 and parameters: {'num_leaves': 34}. Best is trial 24 with value: 0.23346962042716426.\n",
      "num_leaves, val_score: 0.233470: 100%|##########| 20/20 [01:17<00:00,  2.92s/it][I 2020-08-12 07:53:39,480] Trial 26 finished with value: 0.25515513287602964 and parameters: {'num_leaves': 24}. Best is trial 24 with value: 0.23346962042716426.\n",
      "num_leaves, val_score: 0.233470: 100%|##########| 20/20 [01:17<00:00,  3.87s/it]\n",
      "bagging, val_score: 0.233470:  10%|#         | 1/10 [00:01<00:09,  1.04s/it][I 2020-08-12 07:53:40,525] Trial 27 finished with value: 0.23473648196727503 and parameters: {'bagging_fraction': 0.5768266405742613, 'bagging_freq': 6}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  20%|##        | 2/10 [00:02<00:08,  1.05s/it][I 2020-08-12 07:53:41,605] Trial 28 finished with value: 0.23585971495531635 and parameters: {'bagging_fraction': 0.9864747563791268, 'bagging_freq': 2}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  30%|###       | 3/10 [00:03<00:07,  1.08s/it][I 2020-08-12 07:53:42,762] Trial 29 finished with value: 0.23517056882120102 and parameters: {'bagging_fraction': 0.97769419963206, 'bagging_freq': 7}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  40%|####      | 4/10 [00:04<00:06,  1.09s/it][I 2020-08-12 07:53:43,860] Trial 30 finished with value: 0.23549470221657923 and parameters: {'bagging_fraction': 0.41158471994690005, 'bagging_freq': 1}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  50%|#####     | 5/10 [00:05<00:05,  1.08s/it][I 2020-08-12 07:53:44,917] Trial 31 finished with value: 0.2355109332773393 and parameters: {'bagging_fraction': 0.7776422954735288, 'bagging_freq': 4}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  60%|######    | 6/10 [00:06<00:04,  1.07s/it][I 2020-08-12 07:53:45,976] Trial 32 finished with value: 0.23509924675773433 and parameters: {'bagging_fraction': 0.7651617460120202, 'bagging_freq': 4}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  70%|#######   | 7/10 [00:07<00:03,  1.06s/it][I 2020-08-12 07:53:47,004] Trial 33 finished with value: 0.23711797336531862 and parameters: {'bagging_fraction': 0.42777383509432226, 'bagging_freq': 6}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  80%|########  | 8/10 [00:08<00:02,  1.06s/it][I 2020-08-12 07:53:48,051] Trial 34 finished with value: 0.23544528671145692 and parameters: {'bagging_fraction': 0.613783778895431, 'bagging_freq': 2}. Best is trial 27 with value: 0.23473648196727503.\n",
      "bagging, val_score: 0.233470:  90%|######### | 9/10 [00:09<00:01,  1.05s/it][I 2020-08-12 07:53:49,104] Trial 35 finished with value: 0.23371379112724663 and parameters: {'bagging_fraction': 0.867967782261439, 'bagging_freq': 3}. Best is trial 35 with value: 0.23371379112724663.\n",
      "bagging, val_score: 0.233470: 100%|##########| 10/10 [00:10<00:00,  1.05s/it][I 2020-08-12 07:53:50,156] Trial 36 finished with value: 0.23402985848995903 and parameters: {'bagging_fraction': 0.8643060771083824, 'bagging_freq': 3}. Best is trial 35 with value: 0.23371379112724663.\n",
      "bagging, val_score: 0.233470: 100%|##########| 10/10 [00:10<00:00,  1.07s/it]\n",
      "feature_fraction_stage2, val_score: 0.233470:  33%|###3      | 1/3 [00:01<00:02,  1.04s/it][I 2020-08-12 07:53:51,198] Trial 37 finished with value: 0.2334696204271643 and parameters: {'feature_fraction': 0.41600000000000004}. Best is trial 37 with value: 0.2334696204271643.\n",
      "feature_fraction_stage2, val_score: 0.233470:  67%|######6   | 2/3 [00:02<00:01,  1.04s/it][I 2020-08-12 07:53:52,237] Trial 38 finished with value: 0.2355182073646272 and parameters: {'feature_fraction': 0.44800000000000006}. Best is trial 37 with value: 0.2334696204271643.\n",
      "feature_fraction_stage2, val_score: 0.233470: 100%|##########| 3/3 [00:03<00:00,  1.05s/it][I 2020-08-12 07:53:53,304] Trial 39 finished with value: 0.2355182073646272 and parameters: {'feature_fraction': 0.48000000000000004}. Best is trial 37 with value: 0.2334696204271643.\n",
      "feature_fraction_stage2, val_score: 0.233470: 100%|##########| 3/3 [00:03<00:00,  1.05s/it]\n",
      "regularization_factors, val_score: 0.233470:   5%|5         | 1/20 [00:01<00:19,  1.04s/it][I 2020-08-12 07:53:54,345] Trial 40 finished with value: 0.23445886692161005 and parameters: {'lambda_l1': 1.377523554290962e-06, 'lambda_l2': 0.06653003706898261}. Best is trial 40 with value: 0.23445886692161005.\n",
      "regularization_factors, val_score: 0.233470:  10%|#         | 2/20 [00:02<00:18,  1.04s/it][I 2020-08-12 07:53:55,384] Trial 41 finished with value: 0.2346156193571462 and parameters: {'lambda_l1': 1.6616079179406236, 'lambda_l2': 1.9335302599902702e-08}. Best is trial 40 with value: 0.23445886692161005.\n",
      "regularization_factors, val_score: 0.233470:  15%|#5        | 3/20 [00:03<00:17,  1.04s/it][I 2020-08-12 07:53:56,421] Trial 42 finished with value: 0.2351783616925252 and parameters: {'lambda_l1': 2.040177150273966, 'lambda_l2': 1.063637333272805e-07}. Best is trial 40 with value: 0.23445886692161005.\n",
      "regularization_factors, val_score: 0.233470:  20%|##        | 4/20 [00:04<00:16,  1.05s/it][I 2020-08-12 07:53:57,496] Trial 43 finished with value: 0.3097533831315217 and parameters: {'lambda_l1': 1.3030465763816985e-08, 'lambda_l2': 5.478712531338058}. Best is trial 40 with value: 0.23445886692161005.\n",
      "regularization_factors, val_score: 0.233464:  25%|##5       | 5/20 [00:05<00:15,  1.04s/it][I 2020-08-12 07:53:58,527] Trial 44 finished with value: 0.23346396399346142 and parameters: {'lambda_l1': 0.0009346383597973713, 'lambda_l2': 5.2241992461232584e-05}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  30%|###       | 6/20 [00:06<00:14,  1.04s/it][I 2020-08-12 07:53:59,558] Trial 45 finished with value: 0.23346401235184067 and parameters: {'lambda_l1': 0.000977293661566761, 'lambda_l2': 7.489541173258877e-05}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  35%|###5      | 7/20 [00:07<00:13,  1.04s/it][I 2020-08-12 07:54:00,589] Trial 46 finished with value: 0.23347023985349072 and parameters: {'lambda_l1': 0.0007029993591925515, 'lambda_l2': 4.855533157817753e-05}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  40%|####      | 8/20 [00:08<00:12,  1.04s/it][I 2020-08-12 07:54:01,624] Trial 47 finished with value: 0.23347021336209128 and parameters: {'lambda_l1': 0.0006705184236154843, 'lambda_l2': 5.1648386261990354e-05}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  45%|####5     | 9/20 [00:09<00:11,  1.04s/it][I 2020-08-12 07:54:02,659] Trial 48 finished with value: 0.23346406344687493 and parameters: {'lambda_l1': 0.001050875377960507, 'lambda_l2': 4.4398662876708525e-05}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  50%|#####     | 10/20 [00:10<00:10,  1.03s/it][I 2020-08-12 07:54:03,688] Trial 49 finished with value: 0.2341098967664373 and parameters: {'lambda_l1': 0.004714146197350587, 'lambda_l2': 0.0005394441206963371}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  55%|#####5    | 11/20 [00:11<00:09,  1.04s/it][I 2020-08-12 07:54:04,735] Trial 50 finished with value: 0.234147606026695 and parameters: {'lambda_l1': 0.019282495558992963, 'lambda_l2': 2.9143473390586677e-06}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  60%|######    | 12/20 [00:12<00:08,  1.03s/it][I 2020-08-12 07:54:05,761] Trial 51 finished with value: 0.2334697099619106 and parameters: {'lambda_l1': 3.3007828088570755e-05, 'lambda_l2': 0.00015564721612083512}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  65%|######5   | 13/20 [00:13<00:07,  1.03s/it][I 2020-08-12 07:54:06,789] Trial 52 finished with value: 0.23347019385167928 and parameters: {'lambda_l1': 3.538165847235269e-06, 'lambda_l2': 0.0014463327256766165}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  70%|#######   | 14/20 [00:14<00:06,  1.03s/it][I 2020-08-12 07:54:07,822] Trial 53 finished with value: 0.23346963495557338 and parameters: {'lambda_l1': 1.122239111280117e-05, 'lambda_l2': 1.2528199778453384e-05}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  75%|#######5  | 15/20 [00:15<00:05,  1.05s/it][I 2020-08-12 07:54:08,923] Trial 54 finished with value: 0.23493361787632952 and parameters: {'lambda_l1': 0.030351280268777023, 'lambda_l2': 3.234071369184319e-06}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  80%|########  | 16/20 [00:16<00:04,  1.06s/it][I 2020-08-12 07:54:10,003] Trial 55 finished with value: 0.23346974567105283 and parameters: {'lambda_l1': 0.0001451052594168211, 'lambda_l2': 3.424200332044954e-06}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  85%|########5 | 17/20 [00:17<00:03,  1.06s/it][I 2020-08-12 07:54:11,057] Trial 56 finished with value: 0.23383260869277153 and parameters: {'lambda_l1': 7.044295955921658e-06, 'lambda_l2': 0.0024520539911639413}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  90%|######### | 18/20 [00:18<00:02,  1.06s/it][I 2020-08-12 07:54:12,119] Trial 57 finished with value: 0.23383397387173824 and parameters: {'lambda_l1': 0.0027458114718473463, 'lambda_l2': 1.1373198767722994e-05}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464:  95%|#########5| 19/20 [00:19<00:01,  1.05s/it][I 2020-08-12 07:54:13,151] Trial 58 finished with value: 0.2334697197836266 and parameters: {'lambda_l1': 0.00011616248123230667, 'lambda_l2': 3.836806440162599e-07}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464: 100%|##########| 20/20 [00:20<00:00,  1.05s/it][I 2020-08-12 07:54:14,206] Trial 59 finished with value: 0.23416756806233804 and parameters: {'lambda_l1': 0.22776108472401924, 'lambda_l2': 0.004364142732062296}. Best is trial 44 with value: 0.23346396399346142.\n",
      "regularization_factors, val_score: 0.233464: 100%|##########| 20/20 [00:20<00:00,  1.04s/it]\n",
      "min_data_in_leaf, val_score: 0.233126:  20%|##        | 1/5 [00:01<00:04,  1.05s/it][I 2020-08-12 07:54:15,262] Trial 60 finished with value: 0.2331255692653107 and parameters: {'min_child_samples': 25}. Best is trial 60 with value: 0.2331255692653107.\n",
      "min_data_in_leaf, val_score: 0.232747:  40%|####      | 2/5 [00:02<00:03,  1.05s/it][I 2020-08-12 07:54:16,299] Trial 61 finished with value: 0.23274690401613193 and parameters: {'min_child_samples': 10}. Best is trial 61 with value: 0.23274690401613193.\n",
      "min_data_in_leaf, val_score: 0.231889:  60%|######    | 3/5 [00:03<00:02,  1.04s/it][I 2020-08-12 07:54:17,338] Trial 62 finished with value: 0.2318894430775504 and parameters: {'min_child_samples': 5}. Best is trial 62 with value: 0.2318894430775504.\n",
      "min_data_in_leaf, val_score: 0.231889:  80%|########  | 4/5 [00:04<00:01,  1.04s/it][I 2020-08-12 07:54:18,379] Trial 63 finished with value: 0.23394454995741412 and parameters: {'min_child_samples': 50}. Best is trial 62 with value: 0.2318894430775504.\n",
      "min_data_in_leaf, val_score: 0.231889: 100%|##########| 5/5 [00:05<00:00,  1.05s/it][I 2020-08-12 07:54:19,434] Trial 64 finished with value: 0.24067594944099033 and parameters: {'min_child_samples': 100}. Best is trial 62 with value: 0.2318894430775504.\n",
      "min_data_in_leaf, val_score: 0.231889: 100%|##########| 5/5 [00:05<00:00,  1.04s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Result:\n",
      "{'boosting': 'dart', 'application': 'binary', 'metric': 'binary_logloss', 'lambda_l1': 0.0009346383597973713, 'lambda_l2': 5.2241992461232584e-05, 'num_leaves': 3, 'feature_fraction': 0.4, 'bagging_fraction': 1.0, 'bagging_freq': 0, 'min_child_samples': 5}\n",
      "defaultdict(<class 'collections.OrderedDict'>, {'Train': OrderedDict([('binary_logloss', 0.2369943166039877)]), 'Test': OrderedDict([('binary_logloss', 0.2318894430775504)])})\n"
     ]
    }
   ],
   "source": [
    "evaluation_results = {}\n",
    "\n",
    "clf = lgb.train(\n",
    "    train_set=train_data,\n",
    "    params=lgb_params,\n",
    "    valid_sets=[train_data, test_data], \n",
    "    valid_names=['Train', 'Test'],\n",
    "    evals_result=evaluation_results,\n",
    "    num_boost_round=1000,\n",
    "    verbose_eval=0\n",
    ")\n",
    "optimum_boost_rounds = clf.best_iteration\n",
    "\n",
    "print('Result:')\n",
    "print(clf.params)\n",
    "print(clf.best_score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 可視化"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axs = plt.subplots(1, 2, figsize=[15, 4])\n",
    "\n",
    "# Plot the log loss during training\n",
    "axs[0].plot(evaluation_results['Train']['binary_logloss'], label='Train')\n",
    "axs[0].plot(evaluation_results['Test']['binary_logloss'], label='Test')\n",
    "axs[0].set_ylabel('Log loss')\n",
    "axs[0].set_xlabel('Boosting round')\n",
    "axs[0].set_title('Training performance')\n",
    "axs[0].legend()\n",
    "\n",
    "# Plot feature importance\n",
    "importances = pd.DataFrame({'features': clf.feature_name(), \n",
    "                            'importance': clf.feature_importance()}).sort_values('importance', ascending=False)\n",
    "axs[1].bar(x=np.arange(len(importances)), height=importances['importance'])\n",
    "axs[1].set_xticks(np.arange(len(importances)))\n",
    "axs[1].set_xticklabels(importances['features'], rotation = 45)\n",
    "axs[1].set_ylabel('Feature importance (# times used to split)')\n",
    "axs[1].set_title('Feature importance')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "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>features</th>\n",
       "      <th>importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>marital</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>pdays</td>\n",
       "      <td>35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>contact</td>\n",
       "      <td>45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>month</td>\n",
       "      <td>46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>campaign</td>\n",
       "      <td>47</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>education</td>\n",
       "      <td>50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>loan</td>\n",
       "      <td>50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>balance</td>\n",
       "      <td>66</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>age</td>\n",
       "      <td>99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>job</td>\n",
       "      <td>106</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>housing</td>\n",
       "      <td>141</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>days</td>\n",
       "      <td>158</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>duration</td>\n",
       "      <td>325</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>previous</td>\n",
       "      <td>405</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>day</td>\n",
       "      <td>418</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     features  importance\n",
       "2     marital           9\n",
       "12      pdays          35\n",
       "7     contact          45\n",
       "9       month          46\n",
       "11   campaign          47\n",
       "3   education          50\n",
       "6        loan          50\n",
       "4     balance          66\n",
       "0         age          99\n",
       "1         job         106\n",
       "5     housing         141\n",
       "14       days         158\n",
       "10   duration         325\n",
       "13   previous         405\n",
       "8         day         418"
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "importances.sort_values('importance')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy score = \t 0.9154716981132075\n",
      "Precision score = \t 0.8125\n",
      "Recall score =   \t 0.2746478873239437\n",
      "F1 score =      \t 0.4105263157894737\n"
     ]
    }
   ],
   "source": [
    "preds = np.round(clf.predict(X_test))\n",
    "print('Accuracy score = \\t {}'.format(accuracy_score(y_test, preds)))\n",
    "print('Precision score = \\t {}'.format(precision_score(y_test, preds)))\n",
    "print('Recall score =   \\t {}'.format(recall_score(y_test, preds)))\n",
    "print('F1 score =      \\t {}'.format(f1_score(y_test, preds)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 推論"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_pred = clf.predict(p_test_df, num_iteration=optimum_boost_rounds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [],
   "source": [
    "p_test_df['poutcome'] = test_pred\n",
    "p_train_df['poutcome'] = y\n",
    "\n",
    "p_train_df['y'] = py\n",
    "p_test_df['y'] = pz\n",
    "\n",
    "p_train_df.fillna(-1, inplace=True)\n",
    "p_test_df.fillna(-1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_dict = {}\n",
    "for name, group in pd.concat([p_train_df, p_test_df]).groupby('y'):\n",
    "    df_dict[name] = group\n",
    "    \n",
    "y_train_df = pd.concat([df_dict[0], df_dict[1]])\n",
    "y_test_df = df_dict[-1]\n",
    "\n",
    "y_test_df.drop(['y'], axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 27100 entries, 8 to 27062\n",
      "Data columns (total 17 columns):\n",
      " #   Column     Non-Null Count  Dtype  \n",
      "---  ------     --------------  -----  \n",
      " 0   age        27100 non-null  int64  \n",
      " 1   job        27100 non-null  int64  \n",
      " 2   marital    27100 non-null  int64  \n",
      " 3   education  27100 non-null  int64  \n",
      " 4   balance    27100 non-null  int64  \n",
      " 5   housing    27100 non-null  int64  \n",
      " 6   loan       27100 non-null  int64  \n",
      " 7   contact    27100 non-null  int64  \n",
      " 8   day        27100 non-null  int64  \n",
      " 9   month      27100 non-null  int64  \n",
      " 10  duration   27100 non-null  float64\n",
      " 11  campaign   27100 non-null  int64  \n",
      " 12  pdays      27100 non-null  int64  \n",
      " 13  previous   27100 non-null  int64  \n",
      " 14  days       27100 non-null  int64  \n",
      " 15  poutcome   27100 non-null  float64\n",
      " 16  y          27100 non-null  float64\n",
      "dtypes: float64(3), int64(14)\n",
      "memory usage: 3.7 MB\n"
     ]
    }
   ],
   "source": [
    "y_train_df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 18050 entries, 0 to 18049\n",
      "Data columns (total 16 columns):\n",
      " #   Column     Non-Null Count  Dtype  \n",
      "---  ------     --------------  -----  \n",
      " 0   age        18050 non-null  int64  \n",
      " 1   job        18050 non-null  int64  \n",
      " 2   marital    18050 non-null  int64  \n",
      " 3   education  18050 non-null  int64  \n",
      " 4   balance    18050 non-null  int64  \n",
      " 5   housing    18050 non-null  int64  \n",
      " 6   loan       18050 non-null  int64  \n",
      " 7   contact    18050 non-null  int64  \n",
      " 8   day        18050 non-null  int64  \n",
      " 9   month      18050 non-null  int64  \n",
      " 10  duration   18050 non-null  float64\n",
      " 11  campaign   18050 non-null  int64  \n",
      " 12  pdays      18050 non-null  int64  \n",
      " 13  previous   18050 non-null  int64  \n",
      " 14  days       18050 non-null  int64  \n",
      " 15  poutcome   18050 non-null  float64\n",
      "dtypes: float64(2), int64(14)\n",
      "memory usage: 2.3 MB\n"
     ]
    }
   ],
   "source": [
    "y_test_df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "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>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>balance</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>day</th>\n",
       "      <th>month</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>days</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "      <td>27100.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>36.073284</td>\n",
       "      <td>5.152509</td>\n",
       "      <td>1.386162</td>\n",
       "      <td>2.046125</td>\n",
       "      <td>47682.901771</td>\n",
       "      <td>0.583727</td>\n",
       "      <td>0.127269</td>\n",
       "      <td>0.788007</td>\n",
       "      <td>16.700443</td>\n",
       "      <td>6.003542</td>\n",
       "      <td>5.154210</td>\n",
       "      <td>1.775830</td>\n",
       "      <td>432.482399</td>\n",
       "      <td>0.085720</td>\n",
       "      <td>202.810258</td>\n",
       "      <td>0.093498</td>\n",
       "      <td>0.077934</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>7.816417</td>\n",
       "      <td>2.669990</td>\n",
       "      <td>0.872384</td>\n",
       "      <td>0.727044</td>\n",
       "      <td>31650.760036</td>\n",
       "      <td>0.492949</td>\n",
       "      <td>0.333281</td>\n",
       "      <td>0.498535</td>\n",
       "      <td>8.576252</td>\n",
       "      <td>2.135158</td>\n",
       "      <td>0.772170</td>\n",
       "      <td>0.950045</td>\n",
       "      <td>252.150648</td>\n",
       "      <td>0.365889</td>\n",
       "      <td>66.135046</td>\n",
       "      <td>0.143450</td>\n",
       "      <td>0.268072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>22.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-6847.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>34.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>31.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>20015.750000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>4.804021</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>214.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>170.000000</td>\n",
       "      <td>0.025321</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>33.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>47624.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>17.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.068904</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>432.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>182.000000</td>\n",
       "      <td>0.052116</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>37.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>75330.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>26.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>5.846439</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>650.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>234.000000</td>\n",
       "      <td>0.115853</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>90.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>102121.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>31.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>8.031710</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>870.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>373.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                age           job       marital     education        balance  \\\n",
       "count  27100.000000  27100.000000  27100.000000  27100.000000   27100.000000   \n",
       "mean      36.073284      5.152509      1.386162      2.046125   47682.901771   \n",
       "std        7.816417      2.669990      0.872384      0.727044   31650.760036   \n",
       "min       22.000000      2.000000      0.000000      0.000000   -6847.000000   \n",
       "25%       31.000000      3.000000      0.000000      2.000000   20015.750000   \n",
       "50%       33.000000      5.000000      2.000000      2.000000   47624.000000   \n",
       "75%       37.000000      7.000000      2.000000      2.000000   75330.000000   \n",
       "max       90.000000     12.000000      2.000000      3.000000  102121.000000   \n",
       "\n",
       "            housing          loan       contact           day         month  \\\n",
       "count  27100.000000  27100.000000  27100.000000  27100.000000  27100.000000   \n",
       "mean       0.583727      0.127269      0.788007     16.700443      6.003542   \n",
       "std        0.492949      0.333281      0.498535      8.576252      2.135158   \n",
       "min        0.000000      0.000000      0.000000      1.000000      1.000000   \n",
       "25%        0.000000      0.000000      0.000000      8.000000      5.000000   \n",
       "50%        1.000000      0.000000      1.000000     17.000000      5.000000   \n",
       "75%        1.000000      0.000000      1.000000     26.000000      7.000000   \n",
       "max        1.000000      1.000000      2.000000     31.000000     12.000000   \n",
       "\n",
       "           duration      campaign         pdays      previous          days  \\\n",
       "count  27100.000000  27100.000000  27100.000000  27100.000000  27100.000000   \n",
       "mean       5.154210      1.775830    432.482399      0.085720    202.810258   \n",
       "std        0.772170      0.950045    252.150648      0.365889     66.135046   \n",
       "min        0.000000      1.000000     -1.000000      0.000000     34.000000   \n",
       "25%        4.804021      1.000000    214.000000      0.000000    170.000000   \n",
       "50%        5.068904      1.000000    432.000000      0.000000    182.000000   \n",
       "75%        5.846439      2.000000    650.000000      0.000000    234.000000   \n",
       "max        8.031710      5.000000    870.000000      3.000000    373.000000   \n",
       "\n",
       "           poutcome             y  \n",
       "count  27100.000000  27100.000000  \n",
       "mean       0.093498      0.077934  \n",
       "std        0.143450      0.268072  \n",
       "min        0.000000      0.000000  \n",
       "25%        0.025321      0.000000  \n",
       "50%        0.052116      0.000000  \n",
       "75%        0.115853      0.000000  \n",
       "max        1.000000      1.000000  "
      ]
     },
     "execution_count": 157,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "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>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>balance</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>day</th>\n",
       "      <th>month</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>days</th>\n",
       "      <th>poutcome</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "      <td>18050.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>36.104266</td>\n",
       "      <td>5.120388</td>\n",
       "      <td>1.382161</td>\n",
       "      <td>2.039501</td>\n",
       "      <td>47158.451801</td>\n",
       "      <td>0.577175</td>\n",
       "      <td>0.128587</td>\n",
       "      <td>0.783989</td>\n",
       "      <td>16.697950</td>\n",
       "      <td>5.998726</td>\n",
       "      <td>5.149830</td>\n",
       "      <td>1.770083</td>\n",
       "      <td>436.171801</td>\n",
       "      <td>0.090859</td>\n",
       "      <td>202.658449</td>\n",
       "      <td>0.093597</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>7.863805</td>\n",
       "      <td>2.650421</td>\n",
       "      <td>0.873514</td>\n",
       "      <td>0.717774</td>\n",
       "      <td>31469.525699</td>\n",
       "      <td>0.494022</td>\n",
       "      <td>0.334752</td>\n",
       "      <td>0.504881</td>\n",
       "      <td>8.580203</td>\n",
       "      <td>2.118452</td>\n",
       "      <td>0.781723</td>\n",
       "      <td>0.944872</td>\n",
       "      <td>251.392519</td>\n",
       "      <td>0.375801</td>\n",
       "      <td>65.583220</td>\n",
       "      <td>0.142955</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>20.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-6838.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>34.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>31.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>19654.500000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>4.804021</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>219.250000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>170.000000</td>\n",
       "      <td>0.025928</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>33.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>46886.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>17.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.068904</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>436.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>182.000000</td>\n",
       "      <td>0.052890</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>37.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>74178.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>26.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>5.846439</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>654.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>234.000000</td>\n",
       "      <td>0.115518</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>90.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>102124.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>31.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>8.031710</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>870.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>373.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                age           job       marital     education        balance  \\\n",
       "count  18050.000000  18050.000000  18050.000000  18050.000000   18050.000000   \n",
       "mean      36.104266      5.120388      1.382161      2.039501   47158.451801   \n",
       "std        7.863805      2.650421      0.873514      0.717774   31469.525699   \n",
       "min       20.000000      1.000000      0.000000      0.000000   -6838.000000   \n",
       "25%       31.000000      3.000000      0.000000      2.000000   19654.500000   \n",
       "50%       33.000000      5.000000      2.000000      2.000000   46886.000000   \n",
       "75%       37.000000      7.000000      2.000000      2.000000   74178.000000   \n",
       "max       90.000000     12.000000      2.000000      3.000000  102124.000000   \n",
       "\n",
       "            housing          loan       contact           day         month  \\\n",
       "count  18050.000000  18050.000000  18050.000000  18050.000000  18050.000000   \n",
       "mean       0.577175      0.128587      0.783989     16.697950      5.998726   \n",
       "std        0.494022      0.334752      0.504881      8.580203      2.118452   \n",
       "min        0.000000      0.000000      0.000000      1.000000      1.000000   \n",
       "25%        0.000000      0.000000      0.000000      8.000000      5.000000   \n",
       "50%        1.000000      0.000000      1.000000     17.000000      5.000000   \n",
       "75%        1.000000      0.000000      1.000000     26.000000      7.000000   \n",
       "max        1.000000      1.000000      2.000000     31.000000     12.000000   \n",
       "\n",
       "           duration      campaign         pdays      previous          days  \\\n",
       "count  18050.000000  18050.000000  18050.000000  18050.000000  18050.000000   \n",
       "mean       5.149830      1.770083    436.171801      0.090859    202.658449   \n",
       "std        0.781723      0.944872    251.392519      0.375801     65.583220   \n",
       "min        0.000000      1.000000     -1.000000      0.000000     34.000000   \n",
       "25%        4.804021      1.000000    219.250000      0.000000    170.000000   \n",
       "50%        5.068904      1.000000    436.000000      0.000000    182.000000   \n",
       "75%        5.846439      2.000000    654.000000      0.000000    234.000000   \n",
       "max        8.031710      4.000000    870.000000      3.000000    373.000000   \n",
       "\n",
       "           poutcome  \n",
       "count  18050.000000  \n",
       "mean       0.093597  \n",
       "std        0.142955  \n",
       "min        0.000000  \n",
       "25%        0.025928  \n",
       "50%        0.052890  \n",
       "75%        0.115518  \n",
       "max        1.000000  "
      ]
     },
     "execution_count": 158,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test_df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_df.to_csv('y_train.csv', index=False)\n",
    "y_test_df.to_csv('y_test.csv', index=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tensorflow",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}