cleaner all together.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.insert(0, \"./data/\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "from sklearn.metrics import confusion_matrix, precision_score, recall_score\n",
    "\n",
    "import roc\n",
    "from roc import plot_roc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "churn_raw = pd.read_csv(\"./data/churn.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'churn_raw' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1-cbc5611b45a2>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     14\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     15\u001b[0m \u001b[0;31m# use the function\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 16\u001b[0;31m \u001b[0mactive_flag\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchurn_raw\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'last_trip_date'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     17\u001b[0m \u001b[0;31m#churn_raw.head()\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'churn_raw' is not defined"
     ]
    }
   ],
   "source": [
    "from datetime import datetime\n",
    "def active_flag(df, last_trip_date_columns):\n",
    "    # calucate preceding data\n",
    "    df['pulled_data'] = datetime.strptime('2014-7-1', \"%Y-%m-%d\") # temp pulled_data columns\n",
    "    df['preceding_data'] = df['pulled_data'] - pd.to_datetime(df[last_trip_date_columns])\n",
    "    df['preceding_data'] = df['preceding_data'].astype(str).str[:-24].astype(int)\n",
    "    \n",
    "    \n",
    "    # if active or not\n",
    "    df['active_flag'] = np.where(df['preceding_data'] <= 30, 1, 0)\n",
    "    \n",
    "    # drop temp columns\n",
    "    df.drop('pulled_data',axis=1)\n",
    "\n",
    "# use the function\n",
    "active_flag(churn_raw,'last_trip_date')\n",
    "#churn_raw.head()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 50000 entries, 0 to 49999\n",
      "Data columns (total 15 columns):\n",
      "avg_dist                  50000 non-null float64\n",
      "avg_rating_by_driver      49799 non-null float64\n",
      "avg_rating_of_driver      41878 non-null float64\n",
      "avg_surge                 50000 non-null float64\n",
      "city                      50000 non-null object\n",
      "last_trip_date            50000 non-null object\n",
      "phone                     49604 non-null object\n",
      "signup_date               50000 non-null object\n",
      "surge_pct                 50000 non-null float64\n",
      "trips_in_first_30_days    50000 non-null int64\n",
      "luxury_car_user           50000 non-null bool\n",
      "weekday_pct               50000 non-null float64\n",
      "pulled_data               50000 non-null datetime64[ns]\n",
      "preceding_data            50000 non-null int64\n",
      "active_flag               50000 non-null int64\n",
      "dtypes: bool(1), datetime64[ns](1), float64(6), int64(3), object(4)\n",
      "memory usage: 5.4+ MB\n"
     ]
    }
   ],
   "source": [
    "churn=churn_raw.copy()\n",
    "churn.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# EDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f7375868c18>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x1080 with 5 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "sns.set_style('darkgrid')\n",
    "sns.set_palette('muted')\n",
    "\n",
    "fig, axs = plt.subplots(5, 1, figsize = (20,15))\n",
    "sns.countplot(x = 'city',data = churn, ax=axs[0])\n",
    "sns.countplot(x = 'phone', data = churn, ax=axs[1])\n",
    "sns.countplot(x = 'luxury_car_user',data = churn, ax=axs[2])\n",
    "sns.countplot(x = 'avg_rating_of_driver',data = churn, ax=axs[3])\n",
    "sns.countplot(x = 'avg_rating_by_driver',data = churn, ax=axs[4])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Winterfell        23336\n",
       " Astapor           16534\n",
       " King's Landing    10130\n",
       " Name: city, dtype: int64, 50000)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "churn.city.value_counts(), churn.city.size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "False    31146\n",
       "True     18854\n",
       "Name: luxury_car_user, dtype: int64"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "churn.luxury_car_user.value_counts() # 5000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "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></th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>last_trip_date</th>\n",
       "      <th>phone</th>\n",
       "      <th>signup_date</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>trips_in_first_30_days</th>\n",
       "      <th>luxury_car_user</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>pulled_data</th>\n",
       "      <th>preceding_data</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>city</th>\n",
       "      <th>active_flag</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">Astapor</th>\n",
       "      <th>0</th>\n",
       "      <td>12306</td>\n",
       "      <td>12251</td>\n",
       "      <td>9790</td>\n",
       "      <td>12306</td>\n",
       "      <td>12306</td>\n",
       "      <td>12215</td>\n",
       "      <td>12306</td>\n",
       "      <td>12306</td>\n",
       "      <td>12306</td>\n",
       "      <td>12306</td>\n",
       "      <td>12306</td>\n",
       "      <td>12306</td>\n",
       "      <td>12306</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4228</td>\n",
       "      <td>4225</td>\n",
       "      <td>3857</td>\n",
       "      <td>4228</td>\n",
       "      <td>4228</td>\n",
       "      <td>4198</td>\n",
       "      <td>4228</td>\n",
       "      <td>4228</td>\n",
       "      <td>4228</td>\n",
       "      <td>4228</td>\n",
       "      <td>4228</td>\n",
       "      <td>4228</td>\n",
       "      <td>4228</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">King's Landing</th>\n",
       "      <th>0</th>\n",
       "      <td>3767</td>\n",
       "      <td>3751</td>\n",
       "      <td>3066</td>\n",
       "      <td>3767</td>\n",
       "      <td>3767</td>\n",
       "      <td>3741</td>\n",
       "      <td>3767</td>\n",
       "      <td>3767</td>\n",
       "      <td>3767</td>\n",
       "      <td>3767</td>\n",
       "      <td>3767</td>\n",
       "      <td>3767</td>\n",
       "      <td>3767</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6363</td>\n",
       "      <td>6339</td>\n",
       "      <td>5781</td>\n",
       "      <td>6363</td>\n",
       "      <td>6363</td>\n",
       "      <td>6325</td>\n",
       "      <td>6363</td>\n",
       "      <td>6363</td>\n",
       "      <td>6363</td>\n",
       "      <td>6363</td>\n",
       "      <td>6363</td>\n",
       "      <td>6363</td>\n",
       "      <td>6363</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">Winterfell</th>\n",
       "      <th>0</th>\n",
       "      <td>15123</td>\n",
       "      <td>15029</td>\n",
       "      <td>11850</td>\n",
       "      <td>15123</td>\n",
       "      <td>15123</td>\n",
       "      <td>14977</td>\n",
       "      <td>15123</td>\n",
       "      <td>15123</td>\n",
       "      <td>15123</td>\n",
       "      <td>15123</td>\n",
       "      <td>15123</td>\n",
       "      <td>15123</td>\n",
       "      <td>15123</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8213</td>\n",
       "      <td>8204</td>\n",
       "      <td>7534</td>\n",
       "      <td>8213</td>\n",
       "      <td>8213</td>\n",
       "      <td>8148</td>\n",
       "      <td>8213</td>\n",
       "      <td>8213</td>\n",
       "      <td>8213</td>\n",
       "      <td>8213</td>\n",
       "      <td>8213</td>\n",
       "      <td>8213</td>\n",
       "      <td>8213</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            avg_dist  avg_rating_by_driver  \\\n",
       "city           active_flag                                   \n",
       "Astapor        0               12306                 12251   \n",
       "               1                4228                  4225   \n",
       "King's Landing 0                3767                  3751   \n",
       "               1                6363                  6339   \n",
       "Winterfell     0               15123                 15029   \n",
       "               1                8213                  8204   \n",
       "\n",
       "                            avg_rating_of_driver  avg_surge  last_trip_date  \\\n",
       "city           active_flag                                                    \n",
       "Astapor        0                            9790      12306           12306   \n",
       "               1                            3857       4228            4228   \n",
       "King's Landing 0                            3066       3767            3767   \n",
       "               1                            5781       6363            6363   \n",
       "Winterfell     0                           11850      15123           15123   \n",
       "               1                            7534       8213            8213   \n",
       "\n",
       "                            phone  signup_date  surge_pct  \\\n",
       "city           active_flag                                  \n",
       "Astapor        0            12215        12306      12306   \n",
       "               1             4198         4228       4228   \n",
       "King's Landing 0             3741         3767       3767   \n",
       "               1             6325         6363       6363   \n",
       "Winterfell     0            14977        15123      15123   \n",
       "               1             8148         8213       8213   \n",
       "\n",
       "                            trips_in_first_30_days  luxury_car_user  \\\n",
       "city           active_flag                                            \n",
       "Astapor        0                             12306            12306   \n",
       "               1                              4228             4228   \n",
       "King's Landing 0                              3767             3767   \n",
       "               1                              6363             6363   \n",
       "Winterfell     0                             15123            15123   \n",
       "               1                              8213             8213   \n",
       "\n",
       "                            weekday_pct  pulled_data  preceding_data  \n",
       "city           active_flag                                            \n",
       "Astapor        0                  12306        12306           12306  \n",
       "               1                   4228         4228            4228  \n",
       "King's Landing 0                   3767         3767            3767  \n",
       "               1                   6363         6363            6363  \n",
       "Winterfell     0                  15123        15123           15123  \n",
       "               1                   8213         8213            8213  "
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "churn.groupby(['city','active_flag']).count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "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></th>\n",
       "      <th>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>city</th>\n",
       "      <th>last_trip_date</th>\n",
       "      <th>phone</th>\n",
       "      <th>signup_date</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>trips_in_first_30_days</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>pulled_data</th>\n",
       "      <th>preceding_data</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>luxury_car_user</th>\n",
       "      <th>active_flag</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">False</th>\n",
       "      <th>0</th>\n",
       "      <td>21839</td>\n",
       "      <td>21696</td>\n",
       "      <td>17371</td>\n",
       "      <td>21839</td>\n",
       "      <td>21839</td>\n",
       "      <td>21839</td>\n",
       "      <td>21731</td>\n",
       "      <td>21839</td>\n",
       "      <td>21839</td>\n",
       "      <td>21839</td>\n",
       "      <td>21839</td>\n",
       "      <td>21839</td>\n",
       "      <td>21839</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>9307</td>\n",
       "      <td>9283</td>\n",
       "      <td>8388</td>\n",
       "      <td>9307</td>\n",
       "      <td>9307</td>\n",
       "      <td>9307</td>\n",
       "      <td>9261</td>\n",
       "      <td>9307</td>\n",
       "      <td>9307</td>\n",
       "      <td>9307</td>\n",
       "      <td>9307</td>\n",
       "      <td>9307</td>\n",
       "      <td>9307</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">True</th>\n",
       "      <th>0</th>\n",
       "      <td>9357</td>\n",
       "      <td>9335</td>\n",
       "      <td>7335</td>\n",
       "      <td>9357</td>\n",
       "      <td>9357</td>\n",
       "      <td>9357</td>\n",
       "      <td>9202</td>\n",
       "      <td>9357</td>\n",
       "      <td>9357</td>\n",
       "      <td>9357</td>\n",
       "      <td>9357</td>\n",
       "      <td>9357</td>\n",
       "      <td>9357</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>9497</td>\n",
       "      <td>9485</td>\n",
       "      <td>8784</td>\n",
       "      <td>9497</td>\n",
       "      <td>9497</td>\n",
       "      <td>9497</td>\n",
       "      <td>9410</td>\n",
       "      <td>9497</td>\n",
       "      <td>9497</td>\n",
       "      <td>9497</td>\n",
       "      <td>9497</td>\n",
       "      <td>9497</td>\n",
       "      <td>9497</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                             avg_dist  avg_rating_by_driver  \\\n",
       "luxury_car_user active_flag                                   \n",
       "False           0               21839                 21696   \n",
       "                1                9307                  9283   \n",
       "True            0                9357                  9335   \n",
       "                1                9497                  9485   \n",
       "\n",
       "                             avg_rating_of_driver  avg_surge   city  \\\n",
       "luxury_car_user active_flag                                           \n",
       "False           0                           17371      21839  21839   \n",
       "                1                            8388       9307   9307   \n",
       "True            0                            7335       9357   9357   \n",
       "                1                            8784       9497   9497   \n",
       "\n",
       "                             last_trip_date  phone  signup_date  surge_pct  \\\n",
       "luxury_car_user active_flag                                                  \n",
       "False           0                     21839  21731        21839      21839   \n",
       "                1                      9307   9261         9307       9307   \n",
       "True            0                      9357   9202         9357       9357   \n",
       "                1                      9497   9410         9497       9497   \n",
       "\n",
       "                             trips_in_first_30_days  weekday_pct  pulled_data  \\\n",
       "luxury_car_user active_flag                                                     \n",
       "False           0                             21839        21839        21839   \n",
       "                1                              9307         9307         9307   \n",
       "True            0                              9357         9357         9357   \n",
       "                1                              9497         9497         9497   \n",
       "\n",
       "                             preceding_data  \n",
       "luxury_car_user active_flag                  \n",
       "False           0                     21839  \n",
       "                1                      9307  \n",
       "True            0                      9357  \n",
       "                1                      9497  "
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "churn.groupby(['luxury_car_user','active_flag']).count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# plt.scatter(churn.luxury_car_user, churn.active_flag)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# clean data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "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>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>city</th>\n",
       "      <th>last_trip_date</th>\n",
       "      <th>phone</th>\n",
       "      <th>signup_date</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>trips_in_first_30_days</th>\n",
       "      <th>luxury_car_user</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>pulled_data</th>\n",
       "      <th>preceding_data</th>\n",
       "      <th>active_flag</th>\n",
       "      <th>signup_period</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3.67</td>\n",
       "      <td>5.0</td>\n",
       "      <td>4.7</td>\n",
       "      <td>1.10</td>\n",
       "      <td>King's Landing</td>\n",
       "      <td>2014-06-17</td>\n",
       "      <td>iPhone</td>\n",
       "      <td>2014-01-25</td>\n",
       "      <td>15.4</td>\n",
       "      <td>4</td>\n",
       "      <td>True</td>\n",
       "      <td>46.2</td>\n",
       "      <td>2014-07-01</td>\n",
       "      <td>14</td>\n",
       "      <td>1</td>\n",
       "      <td>157</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8.26</td>\n",
       "      <td>5.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1.00</td>\n",
       "      <td>Astapor</td>\n",
       "      <td>2014-05-05</td>\n",
       "      <td>Android</td>\n",
       "      <td>2014-01-29</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>50.0</td>\n",
       "      <td>2014-07-01</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>153</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.77</td>\n",
       "      <td>5.0</td>\n",
       "      <td>4.3</td>\n",
       "      <td>1.00</td>\n",
       "      <td>Astapor</td>\n",
       "      <td>2014-01-07</td>\n",
       "      <td>iPhone</td>\n",
       "      <td>2014-01-06</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>100.0</td>\n",
       "      <td>2014-07-01</td>\n",
       "      <td>175</td>\n",
       "      <td>0</td>\n",
       "      <td>176</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2.36</td>\n",
       "      <td>4.9</td>\n",
       "      <td>4.6</td>\n",
       "      <td>1.14</td>\n",
       "      <td>King's Landing</td>\n",
       "      <td>2014-06-29</td>\n",
       "      <td>iPhone</td>\n",
       "      <td>2014-01-10</td>\n",
       "      <td>20.0</td>\n",
       "      <td>9</td>\n",
       "      <td>True</td>\n",
       "      <td>80.0</td>\n",
       "      <td>2014-07-01</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>172</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3.13</td>\n",
       "      <td>4.9</td>\n",
       "      <td>4.4</td>\n",
       "      <td>1.19</td>\n",
       "      <td>Winterfell</td>\n",
       "      <td>2014-03-15</td>\n",
       "      <td>Android</td>\n",
       "      <td>2014-01-27</td>\n",
       "      <td>11.8</td>\n",
       "      <td>14</td>\n",
       "      <td>False</td>\n",
       "      <td>82.4</td>\n",
       "      <td>2014-07-01</td>\n",
       "      <td>108</td>\n",
       "      <td>0</td>\n",
       "      <td>155</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   avg_dist  avg_rating_by_driver  avg_rating_of_driver  avg_surge  \\\n",
       "0      3.67                   5.0                   4.7       1.10   \n",
       "1      8.26                   5.0                   5.0       1.00   \n",
       "2      0.77                   5.0                   4.3       1.00   \n",
       "3      2.36                   4.9                   4.6       1.14   \n",
       "4      3.13                   4.9                   4.4       1.19   \n",
       "\n",
       "             city last_trip_date    phone signup_date  surge_pct  \\\n",
       "0  King's Landing     2014-06-17   iPhone  2014-01-25       15.4   \n",
       "1         Astapor     2014-05-05  Android  2014-01-29        0.0   \n",
       "2         Astapor     2014-01-07   iPhone  2014-01-06        0.0   \n",
       "3  King's Landing     2014-06-29   iPhone  2014-01-10       20.0   \n",
       "4      Winterfell     2014-03-15  Android  2014-01-27       11.8   \n",
       "\n",
       "   trips_in_first_30_days  luxury_car_user  weekday_pct pulled_data  \\\n",
       "0                       4             True         46.2  2014-07-01   \n",
       "1                       0            False         50.0  2014-07-01   \n",
       "2                       3            False        100.0  2014-07-01   \n",
       "3                       9             True         80.0  2014-07-01   \n",
       "4                      14            False         82.4  2014-07-01   \n",
       "\n",
       "   preceding_data  active_flag  signup_period  \n",
       "0              14            1            157  \n",
       "1              57            0            153  \n",
       "2             175            0            176  \n",
       "3               2            1            172  \n",
       "4             108            0            155  "
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# creat signup_period - how long be user\n",
    "\n",
    "def signup_period(df, signup_date):\n",
    "    # calucate preceding data\n",
    "    df['signup_period'] = df['pulled_data'] - pd.to_datetime(df[signup_date])\n",
    "    df['signup_period'] = df['signup_period'].astype(str).str[:-24].astype(int)\n",
    "\n",
    "# use the function\n",
    "signup_period(churn_raw,'signup_date')\n",
    "churn_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(151, 181)"
      ]
     },
     "execution_count": 132,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "min(churn_raw.signup_period), max(churn_raw.signup_period)"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "1. drop unimportant columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop drop columns\n",
    "def drop_data(df, drop_columns, axis=1):\n",
    "    return df.drop(drop_columns,axis)\n",
    "\n",
    "drop_columns = ['pulled_data', 'last_trip_date','signup_date','phone']\n",
    "churn_raw = drop_data(churn_raw,drop_columns)"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "2. classifier 'city'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [],
   "source": [
    "#King's Landing -1 , Astapor -2, Winterfell-3\n",
    "churn_raw.replace(\"King's Landing\", 1, inplace=True)\n",
    "churn_raw.replace(\"Astapor\", 2, inplace=True)\n",
    "churn_raw.replace(\"Winterfell\", 3, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Final cleaned dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 50000 entries, 0 to 49999\n",
      "Data columns (total 12 columns):\n",
      "avg_dist                  50000 non-null float64\n",
      "avg_rating_by_driver      49799 non-null float64\n",
      "avg_rating_of_driver      41878 non-null float64\n",
      "avg_surge                 50000 non-null float64\n",
      "city                      50000 non-null int64\n",
      "surge_pct                 50000 non-null float64\n",
      "trips_in_first_30_days    50000 non-null int64\n",
      "luxury_car_user           50000 non-null bool\n",
      "weekday_pct               50000 non-null float64\n",
      "preceding_data            50000 non-null int64\n",
      "active_flag               50000 non-null int64\n",
      "signup_period             50000 non-null int64\n",
      "dtypes: bool(1), float64(6), int64(5)\n",
      "memory usage: 4.2 MB\n"
     ]
    }
   ],
   "source": [
    "# Final cleaned dataset\n",
    "churn = churn_raw.copy()\n",
    "churn.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "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>avg_dist</th>\n",
       "      <th>avg_rating_by_driver</th>\n",
       "      <th>avg_rating_of_driver</th>\n",
       "      <th>avg_surge</th>\n",
       "      <th>city</th>\n",
       "      <th>surge_pct</th>\n",
       "      <th>trips_in_first_30_days</th>\n",
       "      <th>luxury_car_user</th>\n",
       "      <th>weekday_pct</th>\n",
       "      <th>preceding_data</th>\n",
       "      <th>active_flag</th>\n",
       "      <th>signup_period</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3.67</td>\n",
       "      <td>5.0</td>\n",
       "      <td>4.7</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1</td>\n",
       "      <td>15.4</td>\n",
       "      <td>4</td>\n",
       "      <td>True</td>\n",
       "      <td>46.2</td>\n",
       "      <td>14</td>\n",
       "      <td>1</td>\n",
       "      <td>157</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8.26</td>\n",
       "      <td>5.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>50.0</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>153</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.77</td>\n",
       "      <td>5.0</td>\n",
       "      <td>4.3</td>\n",
       "      <td>1.00</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>100.0</td>\n",
       "      <td>175</td>\n",
       "      <td>0</td>\n",
       "      <td>176</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2.36</td>\n",
       "      <td>4.9</td>\n",
       "      <td>4.6</td>\n",
       "      <td>1.14</td>\n",
       "      <td>1</td>\n",
       "      <td>20.0</td>\n",
       "      <td>9</td>\n",
       "      <td>True</td>\n",
       "      <td>80.0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>172</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3.13</td>\n",
       "      <td>4.9</td>\n",
       "      <td>4.4</td>\n",
       "      <td>1.19</td>\n",
       "      <td>3</td>\n",
       "      <td>11.8</td>\n",
       "      <td>14</td>\n",
       "      <td>False</td>\n",
       "      <td>82.4</td>\n",
       "      <td>108</td>\n",
       "      <td>0</td>\n",
       "      <td>155</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   avg_dist  avg_rating_by_driver  avg_rating_of_driver  avg_surge  city  \\\n",
       "0      3.67                   5.0                   4.7       1.10     1   \n",
       "1      8.26                   5.0                   5.0       1.00     2   \n",
       "2      0.77                   5.0                   4.3       1.00     2   \n",
       "3      2.36                   4.9                   4.6       1.14     1   \n",
       "4      3.13                   4.9                   4.4       1.19     3   \n",
       "\n",
       "   surge_pct  trips_in_first_30_days  luxury_car_user  weekday_pct  \\\n",
       "0       15.4                       4             True         46.2   \n",
       "1        0.0                       0            False         50.0   \n",
       "2        0.0                       3            False        100.0   \n",
       "3       20.0                       9             True         80.0   \n",
       "4       11.8                      14            False         82.4   \n",
       "\n",
       "   preceding_data  active_flag  signup_period  \n",
       "0              14            1            157  \n",
       "1              57            0            153  \n",
       "2             175            0            176  \n",
       "3               2            1            172  \n",
       "4             108            0            155  "
      ]
     },
     "execution_count": 151,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# True or False to 1 or 0\n",
    "churn= churn.fillna(churn.mean())\n",
    "churn.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "avg_dist                  False\n",
       "avg_rating_by_driver      False\n",
       "avg_rating_of_driver      False\n",
       "avg_surge                 False\n",
       "city                      False\n",
       "surge_pct                 False\n",
       "trips_in_first_30_days    False\n",
       "luxury_car_user           False\n",
       "weekday_pct               False\n",
       "preceding_data            False\n",
       "active_flag               False\n",
       "signup_period             False\n",
       "dtype: bool"
      ]
     },
     "execution_count": 152,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "churn.isnull().any()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "data = churn.copy()\n",
    "y = data.active_flag\n",
    "X = data.replace(True, 1)\n",
    "X = data.replace(False, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.svm import SVC\n",
    "from sklearn.ensemble import RandomForestClassifier as RF\n",
    "from sklearn.neighbors import KNeighborsClassifier as KNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Predictive model establishment\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n",
    "\n",
    "def run_cv(X_train, X_test, y_train, y_test, clf_class,**kwargs):\n",
    "    clf=clf_class(**kwargs)\n",
    "    clf.fit(X_train,y_train)\n",
    "    y_pred = clf.predict(X_test)\n",
    "    return y_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [],
   "source": [
    "# accuracy evaluation\n",
    "def accuracy(y_true,y_pred):\n",
    "    return np.mean(y_true==y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Support vector machines:\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lulu/anaconda3/lib/python3.7/site-packages/sklearn/svm/base.py:196: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.\n",
      "  \"avoid this warning.\", FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "print(\"Support vector machines:\")\n",
    "print(\"%.3f\"%accuracy(y_test,run_cv(X_train, X_test, y_train, y_test, SVC)))\n",
    "print(\"Random forest:\")\n",
    "print(\"%.3f\"%accuracy(y_test,run_cv(X_train, X_test, y_train, y_test, RF)))\n",
    "print(\"K-nearest-neighbors\")\n",
    "print(\"%.3f\"%accuracy(y_test,run_cv(X_train, X_test, y_train, y_test, KNN)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TFPN评估法\n",
    "def TFPN(y_true,y_pred):\n",
    "    Pall=y[y_true==1]\n",
    "    Nall=y[y_true==0]\n",
    "    TP=pd.Series(np.sum((y_true==1)&(y_pred==1))/len(Pall))\n",
    "    FP=pd.Series(np.sum((y_true==1)&(y_pred==0))/len(Pall))\n",
    "    TN=pd.Series(np.sum((y_true==0)&(y_pred==0))/len(Nall))\n",
    "    FN=pd.Series(np.sum((y_true==0)&(y_pred==1))/len(Nall))\n",
    "    AC=pd.Series(np.mean(y_true==y_pred))\n",
    "    tfpn=pd.concat([TP,FP,TN,FN,AC],axis=1)\n",
    "    tfpn.columns=['TP','FP','TN','FN','AC']\n",
    "    return tfpn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Support vector machines:\n"
     ]
    },
    {
     "ename": "TypeError",
     "evalue": "'KFold' object is not iterable",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-68-8074bfd6eaa9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Support vector machines:\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mTFPN\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mrun_cv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mSVC\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Random forest:\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mTFPN\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mrun_cv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mRF\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"K-nearest-neighbors\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-55-929c67a9fc1f>\u001b[0m in \u001b[0;36mrun_cv\u001b[0;34m(X, y, clf_class, **kwargs)\u001b[0m\n\u001b[1;32m      5\u001b[0m     \u001b[0my_pred\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcopy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m     \u001b[0;32mfor\u001b[0m \u001b[0mtrain_index\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mtest_index\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mkf\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m         \u001b[0mX_train\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mX_test\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtrain_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtest_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m         \u001b[0my_train\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtrain_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mTypeError\u001b[0m: 'KFold' object is not iterable"
     ]
    }
   ],
   "source": [
    "print(\"Support vector machines:\")\n",
    "print(TFPN(y,run_cv(X,y,SVC)))\n",
    "print(\"Random forest:\")\n",
    "print(TFPN(y,run_cv(X,y,RF)))\n",
    "print(\"K-nearest-neighbors\")\n",
    "print(TFPN(y,run_cv(X,y,KNN)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Receiver Operating Characteristic (ROC) curves are a data scientist's best friend and are always on top of their toolbox"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "离散型数据可分为两种: 一种是定序, 一种是定类.\n",
    "\n",
    "1) 定序\n",
    "\n",
    "薪资水平其含有顺序意义, 因此将其字符型转化为数值型\n",
    "df['salary'] = df.salary.map({\"low\": 0, \"medium\": 1, \"high\": 2})\n",
    "df.salary.unique()\n",
    "\n",
    "2) 定类\n",
    "\n",
    "岗位是定类型变量, 对其进行one-hot编码, 这里直接利用pandas的get_dummies方法.\n",
    "\n",
    "df_one_hot = pd.get_dummies(df, prefix=\"dep\")\n",
    "df_one_hot.shape\n",
    "\n",
    "2. 连续型数据处理\n",
    "\n",
    "逻辑回归模型能够适应连续型变量, 因此可以不用进行离散化处理, 又由于多个特征之间差异差异较大会造成梯度下降算法收敛速度变慢, 故进行归一化处理\n",
    "#采用max-min归一化方法\n",
    "hours = df_one_hot['hours']\n",
    "df_one_hot['hours'] = df_one_hot.hours.apply(lambda x: (x-hours.min()) / (hours.max()-hours.min()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "两个变量均是连续型且具有线性关系, 则可以使用皮尔逊相关系数, 否则使用斯皮尔曼相关系数, 这里采用斯皮尔曼相关系数\n",
    "#计算相关系数\n",
    "correlation = df_one_hot.corr(method = \"spearman\")\n",
    "plt.figure(figsize=(18, 10))\n",
    "#绘制热力图\n",
    "sns.heatmap(correlation, linewidths=0.2, vmax=1, vmin=-1, linecolor='w',fmt='.2f',\n",
    "            annot=True,annot_kws={'size':10},square=True)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}