EDA_Mobile Ad Anti-Fraud Algorithm Challenge_2019 iFLYTEK.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Mobile Ad Anti-Fraud Algorithm Challenge - Data Analysis (EDA)\n",
    "\n",
    "## 移动广告反欺诈算法挑战赛 数据分析与可视化 \n",
    "-2019 科大讯飞 A.I.开发者挑战赛\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "-- Date:   03 Sep 2019  \n",
    "-- Author:   Arianna.H"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Content:\n",
    "\n",
    "- 0.Set configuration  \n",
    "  \n",
    "  \n",
    "- 1.Explore on data  \n",
    "1.1 Read data  \n",
    "1.2 Shape and label distribution   \n",
    "1.3 Feature description  \n",
    "1.4 Identify missing value  \n",
    "1.5 First glimpse of distribution of the data  \n",
    "  \n",
    "  \n",
    "- 2.Data pre-process  \n",
    "-fill missing value with existing columns/external data  \n",
    "-clean punctuation/non-digit character/cases problem  \n",
    "-timestamp\\datatime\\day\\hour transformation  \n",
    "-feature engineering by combining columns  \n",
    "  \n",
    "  \n",
    "- 3.Data visualization  \n",
    "-bar chart  \n",
    "-line chart  \n",
    "-subplotgrid  \n",
    "  \n",
    "  \n",
    "- 4.Prepare dataset for modeling  \n",
    "4.1 Correlation test for feature selection  \n",
    "4.2 Label Encoding for non-numerical column  \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 0.Set Configuration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.core.interactiveshell import InteractiveShell\n",
    "InteractiveShell.ast_node_interactivity = \"all\"  #display multiple outputs in one cell\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import re  #Regular Expression\n",
    "from datetime import datetime  #timestamp\n",
    "\n",
    "import scipy.stats as stats  #for chi-square\n",
    "from scipy.stats import chi2_contingency  #for chi-square\n",
    "import seaborn as sns  #for ANOVA\n",
    "from scipy import stats  #for ANOVA\n",
    "import statsmodels  #for ANOVA\n",
    "import statsmodels.api as sm  #for ANOVA\n",
    "from statsmodels.formula.api import ols  #for ANOVA\n",
    "from sklearn.preprocessing import LabelEncoder #label encoder\n",
    "\n",
    "pd.set_option('display.max_columns', None)  #display all columns in dataframe\n",
    "pd.set_option('display.max_rows', None)  #display all columns in dataframe"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "____"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.Explore on dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.1 Read data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>sid</th>\n",
       "      <th>label</th>\n",
       "      <th>pkgname</th>\n",
       "      <th>ver</th>\n",
       "      <th>adunitshowid</th>\n",
       "      <th>mediashowid</th>\n",
       "      <th>apptype</th>\n",
       "      <th>nginxtime</th>\n",
       "      <th>ip</th>\n",
       "      <th>city</th>\n",
       "      <th>province</th>\n",
       "      <th>reqrealip</th>\n",
       "      <th>adidmd5</th>\n",
       "      <th>imeimd5</th>\n",
       "      <th>idfamd5</th>\n",
       "      <th>openudidmd5</th>\n",
       "      <th>macmd5</th>\n",
       "      <th>dvctype</th>\n",
       "      <th>model</th>\n",
       "      <th>make</th>\n",
       "      <th>ntt</th>\n",
       "      <th>carrier</th>\n",
       "      <th>os</th>\n",
       "      <th>osv</th>\n",
       "      <th>orientation</th>\n",
       "      <th>lan</th>\n",
       "      <th>h</th>\n",
       "      <th>w</th>\n",
       "      <th>ppi</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>d7460126-e071-4979-9ee8-42f72777a28a-156009070...</td>\n",
       "      <td>1</td>\n",
       "      <td>2d2664e827bcbb8b07100c7fbe072e9b</td>\n",
       "      <td>11.11.0</td>\n",
       "      <td>907d0f8c29663840491577a21c7b612a</td>\n",
       "      <td>ca64a500000d84c8fcb8a0587d0e1e0c</td>\n",
       "      <td>280.0</td>\n",
       "      <td>1.560091e+12</td>\n",
       "      <td>183.197.47.83</td>\n",
       "      <td>石家庄市</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>180.76.181.7</td>\n",
       "      <td>eb4fa0cf77cdd57a6993a4e42c8fa4cf</td>\n",
       "      <td>930b3a8ecff4f586c63fec5b96693f32</td>\n",
       "      <td>empty</td>\n",
       "      <td>empty</td>\n",
       "      <td>ab96caaba2b7973e63fb0d3d5c5bf8b0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>V1809A</td>\n",
       "      <td>vivo</td>\n",
       "      <td>2.0</td>\n",
       "      <td>46000.0</td>\n",
       "      <td>Android</td>\n",
       "      <td>8.1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>zh-CN</td>\n",
       "      <td>2340.0</td>\n",
       "      <td>1080.0</td>\n",
       "      <td>3.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>b660d559-db97-4b5f-9bd2-2450cb89ce77-156005074...</td>\n",
       "      <td>1</td>\n",
       "      <td>empty</td>\n",
       "      <td>NaN</td>\n",
       "      <td>10199dc8fea2e173525bc3151bd84312</td>\n",
       "      <td>3f2fc57a6e1f9c6fa4464c25cc1e88a3</td>\n",
       "      <td>319.0</td>\n",
       "      <td>1.560051e+12</td>\n",
       "      <td>106.34.14.149</td>\n",
       "      <td>开封市</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>47.95.111.138</td>\n",
       "      <td>a4ae6efcfaee62fb6da44a9c05753259</td>\n",
       "      <td>6e38deaa1f7b4118015c550bb21913d2</td>\n",
       "      <td>empty</td>\n",
       "      <td>a4ae6efcfaee62fb6da44a9c05753259</td>\n",
       "      <td>0f607264fc6318a92b9e13c65db7cd3c</td>\n",
       "      <td>2.0</td>\n",
       "      <td>BLA-AL00</td>\n",
       "      <td>HUAWEI</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>android</td>\n",
       "      <td>Android_9</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1080.0</td>\n",
       "      <td>2040.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>f49a740e-66c3-4605-9b67-4d3079fe69cb-156008914...</td>\n",
       "      <td>0</td>\n",
       "      <td>16b81f93f4b1a35cebbf15f07683f171</td>\n",
       "      <td>3.2.1.0524.1958</td>\n",
       "      <td>83f2ecfe65f936f5f2ed59f8e8ff1d01</td>\n",
       "      <td>eea7280e1a2313e4c2e89290b01d196c</td>\n",
       "      <td>273.0</td>\n",
       "      <td>1.560089e+12</td>\n",
       "      <td>223.104.16.151</td>\n",
       "      <td>长春市</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>39.96.159.155</td>\n",
       "      <td>3404f799628be2cf8fda7428aac2cca4</td>\n",
       "      <td>3fa42b06332c729f530b973973fabe3d</td>\n",
       "      <td>empty</td>\n",
       "      <td>empty</td>\n",
       "      <td>5ba1cf1efeee4f83e91f9c0adc3015e3</td>\n",
       "      <td>2.0</td>\n",
       "      <td>PBEM00</td>\n",
       "      <td>OPPO</td>\n",
       "      <td>2.0</td>\n",
       "      <td>46000.0</td>\n",
       "      <td>Android</td>\n",
       "      <td>8.1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>zh-CN</td>\n",
       "      <td>2196.0</td>\n",
       "      <td>1080.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 sid  label  \\\n",
       "0  d7460126-e071-4979-9ee8-42f72777a28a-156009070...      1   \n",
       "1  b660d559-db97-4b5f-9bd2-2450cb89ce77-156005074...      1   \n",
       "2  f49a740e-66c3-4605-9b67-4d3079fe69cb-156008914...      0   \n",
       "\n",
       "                            pkgname              ver  \\\n",
       "0  2d2664e827bcbb8b07100c7fbe072e9b          11.11.0   \n",
       "1                             empty              NaN   \n",
       "2  16b81f93f4b1a35cebbf15f07683f171  3.2.1.0524.1958   \n",
       "\n",
       "                       adunitshowid                       mediashowid  \\\n",
       "0  907d0f8c29663840491577a21c7b612a  ca64a500000d84c8fcb8a0587d0e1e0c   \n",
       "1  10199dc8fea2e173525bc3151bd84312  3f2fc57a6e1f9c6fa4464c25cc1e88a3   \n",
       "2  83f2ecfe65f936f5f2ed59f8e8ff1d01  eea7280e1a2313e4c2e89290b01d196c   \n",
       "\n",
       "   apptype     nginxtime              ip  city  province      reqrealip  \\\n",
       "0    280.0  1.560091e+12   183.197.47.83  石家庄市      -1.0   180.76.181.7   \n",
       "1    319.0  1.560051e+12   106.34.14.149   开封市      -1.0  47.95.111.138   \n",
       "2    273.0  1.560089e+12  223.104.16.151   长春市      -1.0  39.96.159.155   \n",
       "\n",
       "                            adidmd5                           imeimd5 idfamd5  \\\n",
       "0  eb4fa0cf77cdd57a6993a4e42c8fa4cf  930b3a8ecff4f586c63fec5b96693f32   empty   \n",
       "1  a4ae6efcfaee62fb6da44a9c05753259  6e38deaa1f7b4118015c550bb21913d2   empty   \n",
       "2  3404f799628be2cf8fda7428aac2cca4  3fa42b06332c729f530b973973fabe3d   empty   \n",
       "\n",
       "                        openudidmd5                            macmd5  \\\n",
       "0                             empty  ab96caaba2b7973e63fb0d3d5c5bf8b0   \n",
       "1  a4ae6efcfaee62fb6da44a9c05753259  0f607264fc6318a92b9e13c65db7cd3c   \n",
       "2                             empty  5ba1cf1efeee4f83e91f9c0adc3015e3   \n",
       "\n",
       "   dvctype     model    make  ntt  carrier       os        osv  orientation  \\\n",
       "0      2.0    V1809A    vivo  2.0  46000.0  Android      8.1.0          0.0   \n",
       "1      2.0  BLA-AL00  HUAWEI  5.0      0.0  android  Android_9          0.0   \n",
       "2      2.0    PBEM00    OPPO  2.0  46000.0  Android      8.1.0          1.0   \n",
       "\n",
       "     lan       h       w  ppi  \n",
       "0  zh-CN  2340.0  1080.0  3.0  \n",
       "1    NaN  1080.0  2040.0  0.0  \n",
       "2  zh-CN  2196.0  1080.0  0.0  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "traindata = pd.read_csv('round1_iflyad_anticheat_traindata.txt',sep = '\\t',encoding = 'utf-8')\n",
    "traindata.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 Shape and label distribution  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1000000, 29)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "0    517106\n",
       "1    482894\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "traindata.shape\n",
    "# label distribution\n",
    "traindata.label.value_counts(dropna=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Remarks:** This indicates that no label imbalance exsits in train data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.3 Features description\n",
    "\n",
    "* **label:**  \n",
    "基本数据\tlabel\tint\t是否作弊 (contains 1 and 0)    \n",
    "    \n",
    "\n",
    "* **features:**  \n",
    "基本数据\tsid\tstring\t样本id/请求会话sid (unique identification for each request) \n",
    "时间\tnginxtime\tbigint\t请求到达服务时间，单位ms (definition: request processing time in seconds with a milliseconds resolution; time elapsed between the first bytes were read from the client and the log write after the last bytes were sent to the client)  \n",
    "媒体信息\tadunitshowid\tstring\t对外广告位ID（MD5加密）    \n",
    "媒体信息\tmediashowid\tstring\t对外媒体ID（MD5加密）       \n",
    "媒体信息\tpkgname\tstring\t包名(MD5加密) \n",
    "媒体信息\tapptype\tint\tapp所属分类    \n",
    "媒体信息\tver\tstring\tapp版本    \n",
    "IP信息\tip\tstring\t客户端IP地址    \n",
    "IP信息\treqrealip\tstring\t请求的http协议头携带IP，有可能是下游服务器的ip  \n",
    "IP信息\tcity\tstring\t城市     \n",
    "IP信息\tprovince\tstring\t省份   \n",
    "设备信息\tmake\tstring\t厂商  \n",
    "设备信息\tmodel\tstring\t机型  \n",
    "设备信息\tadidmd5\tstring\tAdroid ID的MD5值  \n",
    "设备信息\timeimd5\tstring\timei的MD5值  \n",
    "设备信息\tidfamd5\tstring\tidfa的MD5值  \n",
    "设备信息\topenudidmd5\tstring\topenudid的MD5值  \n",
    "设备信息\tmacmd5\tstring\tmac的MD5值  \n",
    "设备信息\tos\tstring\t操作系统 : Android, iOS  \n",
    "设备信息\tosv\tstring\t操作系统版本  \n",
    "设备信息\tdvctype\tint\t设备类型 0 – 未知,1 – PC,2 – 手机, 3– 平板,4– 电视盒,5– 智能电视,6 – 可穿戴设备,7 – 智能家电,8 - 音箱,9 - 智能硬件    \n",
    "设备信息\tntt\tint\t网络类型 0-未知, 1-有线网, 2-WIFI, 3-蜂窝网络未知, 4-2G, 5-3G, 6–4G    \n",
    "设备信息\tcarrier\tstring\t运营商 0-未知, 46000-移动, 46001-联通, 46003-电信    \n",
    "设备信息\tlan\tstring\t语言  \n",
    "设备信息\torientation\tint\t横竖屏:0竖屏，1横屏   \n",
    "设备信息\th\tint\t设备高    \n",
    "设备信息\tw\tint\t设备宽  \n",
    "设备信息\tppi\tint\t屏幕密度  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.4 Identify missing value and 'empty' values  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "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>index</th>\n",
       "      <th>missing_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>lan</td>\n",
       "      <td>367527</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ver</td>\n",
       "      <td>321824</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>make</td>\n",
       "      <td>123839</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>city</td>\n",
       "      <td>15193</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>osv</td>\n",
       "      <td>12997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>model</td>\n",
       "      <td>4467</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    index  missing_count\n",
       "25    lan         367527\n",
       "3     ver         321824\n",
       "19   make         123839\n",
       "9    city          15193\n",
       "23    osv          12997\n",
       "18  model           4467"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## identify the columns with missing values \n",
    "def get_missing_columns():\n",
    "    missing_df = traindata.isna().sum(axis=0).to_frame()\n",
    "    missing_df = missing_df.reset_index().rename(columns = {0:\"missing_count\"})\n",
    "    missing_df = missing_df.sort_values(\"missing_count\", ascending = False)\n",
    "    missing_df = missing_df[missing_df[\"missing_count\"] > 0]\n",
    "    return missing_df\n",
    "\n",
    "get_missing_columns()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "## fill missing values\n",
    "def cols_to_fix():\n",
    "    for c in traindata.columns:\n",
    "        if traindata[c].dtype == \"object\":\n",
    "            traindata[c] = traindata[c].fillna(\"nan\")\n",
    "        else:\n",
    "            traindata[c] = traindata[c].fillna(0.0)\n",
    "\n",
    "cols_to_fix()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.5 First glimpse of distribution of the data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To analyze the dataset, we need to explore the correlation between the features and the label. Since many of the features are strings, we can check the correlation in an indiract way by looking into the distribution of label counts or label percentage of the features' unique values. \n",
    "\n",
    "Split train data into label_1 and label_0 to check if distributions of variables change due to diffenrent labels.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Android    0.501831\n",
       "android    0.498169\n",
       "Name: os, dtype: float64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "android    0.708437\n",
       "Android    0.291563\n",
       "Name: os, dtype: float64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label_1 = traindata.loc[traindata.label==1,:]\n",
    "label_0 = traindata.loc[traindata.label==0,:]\n",
    "label_1.os.value_counts(normalize=True,dropna=False).head()\n",
    "label_0.os.value_counts(normalize=True,dropna=False).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here, I clustered the features by degree of correlation between predictor variables and target variable:   \n",
    "  \n",
    "* **HIGH importance:**  \n",
    "'ver':  large percentage(1st: 40%), large distribution difference(1st: ±30%) => HIGH importance   \n",
    "'apptype':  large percentage(1st: 30%), large difference(1st: ±30%) => HIGH importance  \n",
    "'dvctype':  large percentage('2'takes up 90%), large difference(1st: ±20%) => HIGH importance  \n",
    "'make':  large percentage(1st: 30%), large difference(1st: ±30%) => HIGH importance  \n",
    "'os':  large percentage(1st: 70%), large difference(1st & 2nd: ±20%) => HIGH importance  \n",
    "'orientation':  large percentage(1st: 80%), large difference(1st & 2nd: ±20%) => HIGH importance  \n",
    "'lan':  large percentage(1st: 60%), large difference(1st: ±40%) => HIGH importance  \n",
    "'h':  large percentage(1st: 30%), large difference(1st: ±20%) => HIGH importance  \n",
    "'pkgname': large percetage('empty'takes up 30%), large difference(2nd: ±30%) => HIGH importance    \n",
    "'mediashowid': large percetage(1st: 30%), large difference(1st: ±30%) => HIGH importance    \n",
    "   \n",
    "   \n",
    "* **MEDIUM importance:**    \n",
    "'osv':  large percentage(1st: 30%), medium differences(1st & 2nd: ±10%) => MEDIUM importance  \n",
    "'reqrealip':  medium percentage(1st: 10%), medium difference(1st: ±10%) => MEDIUM importance    \n",
    "'adidmd5': large percetage('empty': 30%), medium difference(1st: ±10%) => MEDIUM importance    \n",
    "'macmd5': large percetage('empty': 30%), medium difference(1st: ±10%) => MEDIUM importance  \n",
    "'ntt':  large percetage(1st: 60%), small difference(3rd ±10%) => MEDIUM importance   \n",
    "'carrier':  large percetage(1st: 70%), small difference(2nd ±7%) => MEDIUM importance   \n",
    "'w': large percentage(1st: 40%),  medium difference(2nd ±20%) => MEDIUM importance  \n",
    "   \n",
    "   \n",
    "* **LOW importance:** *(consider to increase their correlations by feature engineering)*    \n",
    "'ip':  small percentage(1st: 0.3%), almost no difference(after top 5: ±0.2%) => LOW importance   \n",
    "'city':  small percentage(1st: 3%), almost no difference(4th: ±2%) => LOW importance   \n",
    "'province':  large percentage('-1' takes up 90%), almost no difference(after top 5: ±0.2%) => LOW importance   \n",
    "'model':  small percentage(1st: 2%), small difference(2rd ±0.8%) => LOW importance   \n",
    "'ppi':  large percentage(1st: 70%), almost on difference in top 5 values(after top 5: ±0.4%) => LOW importance  \n",
    "'adunitshowid': small percetage(1st: 8%), small difference(unique value vary) => LOW importance     \n",
    "'imeimd5': small percentage(1st: 1%) and difference(1st: ±1%) => LOW importance     \n",
    "'idfamd5', 'openudidmd5': large percentage('empty'takes up 90%) with very small difference(1st: ±5%,others: ±0.01%) => LOW importance     \n",
    "'nginxtime': timestamp(very sparse) => LOW importance     "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Data pre-process"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This dataset is a great example for you to practice your data cleaning skills, because it contains various forms of data, e.g. floating numbers, timestamp, strings, Chinese/English characters, alphabets in upper/lower cases, etc. Get your hands dirty ;)\n",
    "\n",
    "#### Data Cleaning:  \n",
    "For string data ['city','province' 'model', 'make', 'lan']: clean disordered characters/cases/punctuations and try to fill null values  \n",
    "For timeseries data ['nginxtime']: transform it into datatime form    \n",
    "For numeric data ['apptype','h','w','ppi']: clean non-digit and set null values to 0   \n",
    "For ip/version data ['ip', 'reqrealip', 'ver','osv']: clean punctuations  \n",
    "For id data ['pkgname', 'adunitshowid', 'mediashowid', 'adidmd5', 'imeimd5', 'idfamd5', 'openudidmd5', 'macmd5']: clean punctuations  and extact top 3 digits  \n",
    "\n",
    "\n",
    "#### Feature engineering:  \n",
    "['h_w','hw_ratio','inch']: use 'h','w','ppi' to calculate device size, ratio and inch  \n",
    "['apptype_top1', 'apptype_top2']: extract top 1, 2 digits, convert 'apptype_top2' to int  \n",
    "['osv_top1', 'osv_top2', 'osv_top3']:  extract top 1, 2, 3 digits, convert 'osv_top2', 'osv_top3' to int  \n",
    "['ip_top', 'reqrealip_top']: extract top 1 string after splitting and convert them to integer  \n",
    "['date','dayofweek','hour','hourtominute','hourtosecond']: extract different types of time values from datatime"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.1 fill na value of 'province' with external data containing city-province information using mapping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import external data\n",
    "province = pd.read_csv('province_city.csv',usecols=[0,1])\n",
    "province.columns = ['province', 'city']\n",
    "\n",
    "# create dictionary for mapping\n",
    "mydict = dict(zip(province.city, province.province))\n",
    "traindata.province = traindata.city.map(mydict)\n",
    "traindata.province = traindata.province.fillna(\"nan\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "广东省    0.089094\n",
       "江苏省    0.083905\n",
       "山东省    0.066762\n",
       "nan    0.056557\n",
       "河南省    0.055822\n",
       "Name: province, dtype: float64"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "广东省    0.090815\n",
       "江苏省    0.073573\n",
       "河南省    0.064776\n",
       "浙江省    0.057977\n",
       "nan    0.055428\n",
       "Name: province, dtype: float64"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label_1 = traindata.loc[traindata.label==1,:]\n",
    "label_0 = traindata.loc[traindata.label==0,:]\n",
    "label_1.province.value_counts(normalize=True,dropna=False).head()\n",
    "label_0.province.value_counts(normalize=True,dropna=False).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.2 fill null values of 'make' with corresponding values in 'model' column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata.make = traindata.make.str.upper()\n",
    "traindata.model = traindata.model.str.upper()\n",
    "traindata.make = np.where((traindata.model.str.contains('OPPO') | \n",
    "                            traindata.model.str.contains(r'P.+[00]$',regex=True)),'OPPO',traindata.make)\n",
    "traindata.make = np.where((traindata.model.str.contains('VIVO') |\n",
    "                            traindata.model.str.contains(r'^V.+[A-Z]$',regex=True)),'VIVO',traindata.make)\n",
    "traindata.make = np.where((traindata.model.str.contains('HUAWEI') | \n",
    "                             traindata.model.str.contains(r'-AL',regex=True) | \n",
    "                             traindata.model.str.contains(r'-TL',regex=True) |\n",
    "                             traindata.make.str.contains('华为')),'HUAWEI',traindata.make)\n",
    "traindata.make = np.where((traindata.make.str.contains('360手机') |\n",
    "                            traindata.model.str.contains(r'^\\d.+[-A].+\\d$',regex=True) |\n",
    "                            traindata.model.str.contains(r'^\\d.+[-M].+\\d$',regex=True)),'360',traindata.make)\n",
    "traindata.make = np.where((traindata.make.str.contains('金立') |\n",
    "                            traindata.model.str.contains('GIONEE') |\n",
    "                            traindata.model.str.contains('S9')),'GIONEE',traindata.make)\n",
    "traindata.make = np.where((traindata.model.str.contains('SM') |\n",
    "                             traindata.make.str.contains('三星')),'SAMSUNG',traindata.make)\n",
    "traindata.make = np.where((traindata.model.str.contains('MI') |\n",
    "                             traindata.make.str.contains('小米')),'XIAOMI',traindata.make)\n",
    "traindata.make = np.where(traindata.make.str.contains('荣耀'),'HONOR',traindata.make)\n",
    "traindata.make = np.where((traindata.model.str.contains('MEIZU') | \n",
    "                            traindata.model.str.contains('16TH') |\n",
    "                            traindata.model.str.contains(r'^M\\d+',regex=True) |\n",
    "                            traindata.make.str.contains('魅族')),'MEIZU',traindata.make)\n",
    "traindata.make = np.where(traindata.make.str.contains('努比亚'),'NUBIA',traindata.make)\n",
    "traindata.make = np.where((traindata.model.str.contains('COOLPAD') | \n",
    "                            traindata.make.str.contains('酷派')),'COOLPAD',traindata.make)\n",
    "traindata.make = np.where((traindata.model.str.contains(r'^O.+\\d$',regex=True) | \n",
    "                            traindata.make.str.contains('锤子')),'SMARTISAN',traindata.make)\n",
    "traindata.make = np.where((traindata.make.str.contains('百立丰') |\n",
    "                            traindata.model.str.contains('LEPHONE')),'LEPHONE',traindata.make)\n",
    "traindata.make = np.where(traindata.make.str.contains('中兴'),'ZTE',traindata.make)\n",
    "traindata.make = np.where(traindata.make.str.contains('联想'),'LENOVO',traindata.make)\n",
    "traindata.make = np.where((traindata.make.str.contains('海信') |\n",
    "                            traindata.model.str.contains('HISENSE')),'HISENSE',traindata.make)\n",
    "traindata.make = np.where(traindata.model.str.contains('YU FLY F9'),'YUPLY',traindata.make) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.3  clean 'lan' column: case, punctuation, character "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_lan(x):\n",
    "    if x ==\"zh_\":\n",
    "        return \"zh\"\n",
    "    elif x ==\"fr\":\n",
    "        return \"others\"\n",
    "    elif x ==\"kr\":\n",
    "        return \"others\"\n",
    "    elif x ==\"in_id\":\n",
    "        return \"others\"\n",
    "    elif x ==\"mi\":\n",
    "        return \"others\"\n",
    "    elif x ==\"zh_mo\":\n",
    "        return \"others\"\n",
    "    elif x ==\"de\":\n",
    "        return \"others\"\n",
    "    elif x ==\"th_th\":\n",
    "        return \"others\"\n",
    "    else:\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata.lan = traindata.lan.str.lower().str.replace('-','_')\n",
    "traindata.lan = np.where(traindata.lan.str.contains('cn'),'zh_cn',traindata.lan)\n",
    "traindata.lan = np.where(traindata.lan.str.contains('tw'),'zh_tw',traindata.lan)\n",
    "traindata.lan = np.where(traindata.lan.str.contains('hk'),'zh_hk',traindata.lan)\n",
    "traindata.lan = np.where(traindata.lan.str.contains('us'),'en_us',traindata.lan)\n",
    "traindata.lan = traindata.lan.apply(lambda x: get_lan(x))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.4 transform 'nginxtime' from timestamp to datatime, and create different time dimensions: date, day of week, hour, minute, hourtosecond"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata['nginxtime_clean'] = pd.to_datetime(traindata.nginxtime, unit='ms')\n",
    "traindata['date'] = traindata.nginxtime_clean.apply(lambda x: x.day)\n",
    "traindata['dayofweek'] = traindata.nginxtime_clean.apply(lambda x: x.isoweekday())\n",
    "traindata['hour'] = traindata.nginxtime_clean.apply(lambda x: x.hour)\n",
    "traindata['minute'] = traindata.nginxtime_clean.apply(lambda x: x.minute)\n",
    "traindata['hourtosecond'] = traindata.nginxtime_clean.apply(lambda x: x.strftime('%H%M%S')).astype(int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.5 Clean and extract first few digits of 'ip', 'reqrealip', apptype'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Suspicious ip:   \n",
    "127.0.0.1 本机地址-测试 => all label1  \n",
    "18.219.51.149 美国 哥伦比亚 => all label1  \n",
    "192.168.(0.100) 本地局域网 => 90% label1  \n",
    "10.(130.113.70) 本地局域网 => 90% label1  \n",
    "\n",
    "This indicates that the top one or two parts before the punctuation '.' of ip can represent specific locations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "#'ip'\n",
    "traindata['ip_top'] = np.where(traindata['ip'].str.contains(':'),'0.',traindata['ip'])\n",
    "traindata['ip_top'] = traindata['ip_top'].apply(lambda x: x.split('.')[0]).astype(int)\n",
    "\n",
    "#'reqrealip_top'\n",
    "traindata['reqrealip_top'] = traindata['reqrealip'].apply(lambda x: x.split('.')[0]).astype(int)\n",
    "\n",
    "#'apptype'\n",
    "traindata['apptype_top1'] = traindata['apptype'].apply(lambda x: str(x)[0:1]).apply(lambda x: np.where(x=='-',0,x)).astype(int)\n",
    "traindata['apptype_top2'] = traindata['apptype'].apply(lambda x: str(x)[0:2]).astype(int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.6 Use h, w, ppi to create new features 'h_w' (size), 'hw_ratio', 'inch'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata['h_w'] = round((traindata.h * traindata.w),1)\n",
    "traindata['hw_ratio'] = round(traindata.h / traindata.w,1)\n",
    "traindata['hw_ratio'] = traindata['hw_ratio'].fillna(0.0)\n",
    "traindata['inch'] = np.where(traindata.ppi==0.0, 0.0, round(np.sqrt(traindata.h**2 + traindata.w**2) / traindata.ppi,1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.7 remove non-digit characters in  'osv', 'ver', extract top few digits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# remove all non-digit characters\n",
    "def clean_punctuation(x):\n",
    "    x = re.sub(\"\\D\", \"\", x)\n",
    "    if x=='':\n",
    "        return '0'\n",
    "    else:\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "#'osv'\n",
    "traindata['osv_clean'] = traindata['osv'].astype(str).apply(lambda x: clean_punctuation(x))\n",
    "traindata['osv_top1'] = traindata['osv_clean'].apply(lambda x: x[0:1]).astype(int)\n",
    "traindata['osv_top2'] = traindata['osv_clean'].apply(lambda x: x[0:2]).astype(int)\n",
    "traindata['osv_top3'] = traindata['osv_clean'].apply(lambda x: x[0:3]).astype(int)\n",
    "\n",
    "#'ver' \n",
    "traindata['ver_clean'] = traindata['ver'].apply(lambda x: clean_punctuation(x))\n",
    "traindata['ver_top1'] = traindata['ver_clean'].apply(lambda x: x[0:1]).astype(int)\n",
    "traindata['ver_top2'] = traindata['ver_clean'].apply(lambda x: x[0:2]).astype(int)\n",
    "traindata['ver_top3'] = traindata['ver_clean'].apply(lambda x: x[0:3]).astype(int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.8 remove alphabets in 'id' like columns, extract top 3 digits"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* The columns that contain 'Empty' values:   \n",
    "pkgname    263239  \n",
    "adunitshowid    296  \n",
    "mediashowid    296  \n",
    "adidmd5    254158  \n",
    "imeimd5    6868  \n",
    "idfamd5    999641  \n",
    "openudidmd5    918400  \n",
    "macmd5    418236  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# id_column = ['pkgname','adunitshowid','mediashowid','adidmd5','imeimd5','openudidmd5','macmd5']\n",
    "# for i in id_column:\n",
    "#     traindata[i] = traindata[i].apply(lambda x: clean_punctuation(x)).apply(lambda x: x[0:3]).astype(int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Data Visualization\n",
    "have fun with plotting:D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "label_1 = traindata.loc[traindata.label==1,:]\n",
    "label_0 = traindata.loc[traindata.label==0,:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### make : Plot make distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt  #plot\n",
    "\n",
    "# '_ =' to ignore unwanted output\n",
    "fig = plt.figure()\n",
    "_ = fig.set(alpha=0.5)\n",
    "fig.set_size_inches(12,3)\n",
    "_ = plt.subplot2grid((1,2),(0,0))\n",
    "_ = label_1.make.value_counts().head(10).plot.bar(width=0.4, color='xkcd:lightblue')\n",
    "_ = plt.title(u\"make_label_1\")\n",
    "_ = plt.subplot2grid((1,2),(0,1))\n",
    "_ = label_0.make.value_counts().head(10).plot.bar(width=0.4, color='xkcd:lavender')\n",
    "_ = plt.title(u\"make_label_0\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### lan : Plot language distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x216 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure()\n",
    "_ = fig.set(alpha=0.5)\n",
    "fig.set_size_inches(10,3)\n",
    "_ = plt.subplot2grid((1,2),(0,0))\n",
    "_ = label_1.lan.value_counts().head().plot.bar(width=0.4, color='xkcd:lightblue')\n",
    "_ = plt.title(u\"lan_label_1\")\n",
    "_ = plt.subplot2grid((1,2),(0,1))\n",
    "_ = label_0.lan.value_counts().head().plot.bar(width=0.4, color='xkcd:lavender')\n",
    "_ = plt.title(u\"lan_label_0\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### nginxtime : Plot label distribution per day of a week and per hour"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "dayofweek_df = pd.DataFrame(columns = ['dayofweek','label_1', 'label_0', 'label_all', 'label_1_%','label_0_%'])\n",
    "dayofweek_df['dayofweek'] = traindata.dayofweek.sort_values().unique()\n",
    "dayofweek_df['label_1'] = traindata[traindata['label']==1].dayofweek.value_counts().sort_index().tolist()\n",
    "dayofweek_df['label_0'] = traindata[traindata['label']==0].dayofweek.value_counts().sort_index().tolist()\n",
    "dayofweek_df['label_all'] = dayofweek_df['label_1']+dayofweek_df['label_0']\n",
    "dayofweek_df['label_1_%'] = dayofweek_df['label_1']/dayofweek_df['label_all']\n",
    "dayofweek_df['label_0_%'] = dayofweek_df['label_0']/dayofweek_df['label_all']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "hour_df = pd.DataFrame(columns = ['hour','label_1', 'label_0', 'label_all', 'label_1_%','label_0_%'])\n",
    "hour_df['hour'] = traindata.hour.sort_values().unique()\n",
    "hour_df['label_1'] = traindata.loc[traindata['label']==1].hour.value_counts().sort_index().tolist()\n",
    "hour_df['label_0'] = traindata.loc[traindata['label']==0].hour.value_counts().sort_index().tolist()\n",
    "hour_df['label_all'] = hour_df['label_1']+hour_df['label_0']\n",
    "hour_df['label_1_%'] = hour_df['label_1']/hour_df['label_all']\n",
    "hour_df['label_0_%'] = hour_df['label_0']/hour_df['label_all']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = np.arange(len(dayofweek_df['dayofweek']))  # the label locations\n",
    "width = 0.35  # the width of the bars\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(6,4)\n",
    "_ = rects1 = ax.bar(x - width/2, dayofweek_df['label_1'], width, label='label_1', color='xkcd:lightblue')\n",
    "_ = rects2 = ax.bar(x + width/2, dayofweek_df['label_0'], width, label='label_0', color='xkcd:lavender')\n",
    "\n",
    "# Add some text for labels, title and custom x-axis tick labels, etc.\n",
    "_ = ax.set_ylabel('Counts')\n",
    "_ = ax.set_title('Distribution of label counts per dayofweek')\n",
    "_ = ax.set_xticks(x)\n",
    "_ = ax.set_xticklabels(dayofweek_df['dayofweek'])\n",
    "_ = ax.legend()\n",
    "\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1gAAAEYCAYAAABBWFftAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3dfbxcVXno8d8DBAPyEiDBBhIMSkSEQsQIWKhvKATLNXIvKhY1EWx6uYIv0BYEWlHhXmkVW1rRUsEERZCmckEvClGjtDW8JAgYQCDyeiBCSAiKgCT43D/2OnQ4nJdJsvfMOTm/7+czn7Nn7bX3s/aeyc48s9ZeE5mJJEmSJGnDbdLtBkiSJEnSxsIES5IkSZJqYoIlSZIkSTUxwZIkSZKkmphgSZIkSVJNTLAkSZIkqSYmWJI0TETEVyLir2va1y4R8WREbFqe/zgiPlzHvsv+vhcRs+ra3zrEPTMiHouIX/Wz7s0R0dPmfmZHxH+sZxvWe9vRZl1eE0naWGzW7QZI0mgQEfcBLwPWAs8BtwMXAedn5u8BMvN/rsO+PpyZPxioTmY+AGy1Ya1+Pt4ZwG6Z+f6W/R9Wx77XsR2TgZOAl2fmo52OP1xFxBTgXmBMZq7tbmskSfZgSVLn/LfM3Bp4OfA54GTggrqDRMTG+uXZy4GVJlfDUzffd1HxM42kYcGLkSR1WGY+kZlXAu8FZkXEXgARMTcizizL4yPiuxGxOiJWRcS/R8QmEfF1YBfgO2UI4F9FxJSIyIg4NiIeAH7UUtb6ofeVEXFDRDwREVdExPYl1ouGcUXEfRHxtoiYAZwKvLfEu6Wsf37IYWnX6RFxf0Q8GhEXRcS2ZV1vO2ZFxANleN9pA52biNi2bL+i7O/0sv+3AQuAnUo75g51niPilIj4ZUT8JiJuj4gjXlwl/rGcj19ExMF92nFBRCyPiIfK0MRNh4pZtj0oIn5aXrsHI2L2YMdW1p0REd9o2ccLXr9yvj8bEf9ZjueaiBhfql9b/q4u5+YNEbFbRPykHNtjEfGtAdraG2dORDxcjveklvWbtJzHlRFxWcv75kXvu0HOyUnlvbE8Ij7U5zxvyDk5KyL+E3gKeEU7r48kNc0ES5K6JDNvAHqAP+5n9Ull3QSqoYWnVpvkB4AHqHrDtsrMv23Z5k3AHsChA4T8IHAMsBPVUMVz22jj94H/DXyrxNunn2qzy+MtVB9ytwL+qU+dg4DdgYOBv4mIPQYI+Y/AtmU/bypt/lAZDnkY8HBpx+yh2g78kurcbgt8GvhGRExsWb8/cA8wHvgU8O3e5AGYR3WOdgNeCxwCDHkPW0TsAnyvHMcEYBpw82DH1sZx9PrTUn9HYHPgL0r5G8vfceXcLAI+C1wDbAdMKrEH8xZgKtVxnlISWoCPAu8q7d0JeBz4Up9th3rf/QHVce8MHAt8KSK2K+s29Jx8AJgDbA3cvw7bSVJjTLAkqbseBrbvp3wNMJHqfqM1mfnvmZlD7OuMzPxtZj49wPqvZ+bSzPwt8NfAe9rtlRnC0cA5mXlPZj4JfBI4qk/v2acz8+nMvAW4BXhRolba8l7gk5n5m8y8D/gC1YfodZaZ/5qZD2fm7zPzW8DdwH4tVR4F/r6c328BdwJ/EhEvo0rmPl7O56PAF4Gj2gh7NPCDzLyk7HdlZt5c07F9LTPvKq/vZVTJ20DWUA2p3Ckzn8nMoSbl+HQ51p8DXwPeV8r/HDgtM3sy83fAGcCRfV7bod53a4DPlPNxFfAksHtN52RuZt6WmWszc806bCdJjTHBkqTu2hlY1U/53wHLgGsi4p6IOKWNfT24DuvvB8ZQ9d5sqJ14Ye/B/VSTKL2spax11r+n6H8CjvFUPTN997Xz+jQqIj4YETeXoXqrgb144fE+1CdpvZ/qWF5OdW6Wt2z7z1Q9R0OZTNVz1lcdx9bOOez1V0AAN0TEbRFxzBD77vve2Kksvxy4vOU83EE1ScvLBti2Pyv7TL7R2/Y6zslQsSWp40ywJKlLIuL1VB8mX9S7UL7RPykzXwH8N+DElnuEBurJGqqHa3LL8i5UPQuPAb8Ftmxp16ZUw9va3e/DVB/EW/e9FnhkiO36eoz/6nlp3ddD67gfIuLlwL8AxwM7ZOY4YClV0tFr54hofb4L1bE8CPwOGJ+Z48pjm8zcs43QDwKv7Kd8qGN7wWtANayuXS96fTLzV5n5Z5m5E1Uv1HkRsdsg++j73ni4LD8IHNZyHsZl5tjMbH1Nhnp/DKSOc7K+sSWpMSZYktRhEbFNRBwOXAp8owzL6lvn8DJRQQC/puo1eK6sfoT1u6H//RHxmojYEvgMMD8znwPuAsZGxJ9ExBjgdOAlLds9AkyJgWdpuwT4RETsGhFb8V/3bK3TlOGlLZcBZ0XE1iVJOhH4xuBb9uulVB++VwCUiRX26lNnR+CjETEmIt5NdR/RVZm5nOr+pS+U12qTiHhlRLypjbgXA2+LiPdExGYRsUNETGvj2G4G3hjV75dtSzXMsl0rgN/T8p6IiHdHxKTy9PFyLp7rZ9tefx0RW0bEnlT3QPVOivGV0uaXl/1OiIiZ69C2ATV8TiSpa0ywJKlzvhMRv6HqFTgNOIeBb+ifCvyA6n6VRcB5mfnjsu7/AKeXYVt/McD2/fk6MJdqqNlYqgkMyMwngP8FfJWq9+C3VBNs9PrX8ndlRNzUz34vLPu+lur3mJ4BTliHdrU6ocS/h6pn75tl/+skM2+nup9nEVWC+IfAf/apdj3VeX4MOAs4MjNXlnUfpBq+djtVgjKf6p64oeI+ALyDapKSVVRJQu/9ZgMeW2YuoEpqbgWWAN9dh2N9qrT/P8t74gDg9cD1EfEkcCXwscy8d5Dd/IRqSOoPgc9n5jWl/B/K9teU9+51VJOD1KWRcyJJ3RRD3zMtSZI2RuGPFEtS7ezBkiRJkqSamGBJkiRJUk0cIihJkiRJNbEHS5IkSZJqstnQVTYu48ePzylTpnS7GZIkSZJGsCVLljyWmRP6lo+6BGvKlCksXry4282QJEmSNIJFxP39lTtEUJIkSZJqYoIlSZIkSTUxwZIkSZKkmoy6e7AkSZIkvdCaNWvo6enhmWee6XZThp2xY8cyadIkxowZ01Z9EyxJkiRplOvp6WHrrbdmypQpRES3mzNsZCYrV66kp6eHXXfdta1tHCIoSZIkjXLPPPMMO+ywg8lVHxHBDjvssE49eyZYkiRJkkyuBrCu58UES5IkSZJq4j1YkiRJkl7g9OufrnV/Z+6/Ra37G87swZIkSZLUdVtttdWg6++77z722muvddrn7NmzmT9//oDr/+mf/onddtuNiOCxxx5bp30PxB4sSZI2AtfNXd123QNmj2uwJZI0chx44IEcfvjhvPnNb65tn/ZgSZIkSRo2nnzySQ4++GD23Xdf/vAP/5Arrrji+XVr165l1qxZ7L333hx55JE89dRTACxZsoQ3velNvO51r+PQQw9l+fLlbcV67Wtfy5QpU2ptvwmWJEmSpGFj7NixXH755dx0000sXLiQk046icwE4M4772TOnDnceuutbLPNNpx33nmsWbOGE044gfnz57NkyRKOOeYYTjvttK613yGCkiRJkoaNzOTUU0/l2muvZZNNNuGhhx7ikUceAWDy5MkceOCBALz//e/n3HPPZcaMGSxdupS3v/3tADz33HNMnDixa+03wZIkSZI0bFx88cWsWLGCJUuWMGbMGKZMmfL8D/32/U2qiCAz2XPPPVm0aFE3mvsiJliSJEmSXqCb06o/8cQT7LjjjowZM4aFCxdy//33P7/ugQceYNGiRbzhDW/gkksu4aCDDmL33XdnxYoVz5evWbOGu+66iz333LMr7fceLEmSJEnDxtFHH83ixYuZPn06F198Ma9+9aufX7fHHnswb9489t57b1atWsVxxx3H5ptvzvz58zn55JPZZ599mDZtGj/96U/binXuuecyadIkenp62Hvvvfnwhz+8we2P3hvGRovp06fn4sWLu90MSZJq5TTtkjbEHXfcwR577NHtZgxb/Z2fiFiSmdP71rUHS5IkSZJq4j1YkiRJkjZqRxxxBPfee+8Lys4++2wOPfTQ2mOZYEmSNEydfv3Tbdc9vMF2SNJId/nll3csVmMJVkSMBa4FXlLizM/MT0XErsClwPbATcAHMvPZiHgJcBHwOmAl8N7MvK/s65PAscBzwEcz8+pSPgP4B2BT4KuZ+bmmjkeSpHa0ey/UxnIflPd+SdILNXkP1u+At2bmPsA0YEZEHACcDXwxM6cCj1MlTpS/j2fmbsAXSz0i4jXAUcCewAzgvIjYNCI2Bb4EHAa8BnhfqStJkiRJXdFYgpWVJ8vTMeWRwFuB+aV8HvCusjyzPKesPziqXxKbCVyamb/LzHuBZcB+5bEsM+/JzGepesVmNnU8kiRJkjSURu/BKr1MS4DdqHqbfgmszsy1pUoPsHNZ3hl4ECAz10bEE8AOpfy6lt22bvNgn/L9GzgMSZIkaVRZl+G/7RhNQ4QbnaY9M5/LzGnAJKoep/4m1+/9Ia4YYN26lr9IRMyJiMURsXjFihVDN1ySJElSR2211VaDrr/vvvvYa6+91mmfs2fPZv78+QOuv/fee9l///2ZOnUq733ve3n22WfXaf/96cjvYGXmauDHwAHAuIjo7TmbBDxclnuAyQBl/bbAqtbyPtsMVN5f/PMzc3pmTp8wYUIdhyRJkiRphDv55JP5xCc+wd133812223HBRdcsMH7bHIWwQnAmsxcHRFbAG+jmrhiIXAk1T1Ts4AryiZXlueLyvofZWZGxJXANyPiHGAnYCpwA1UP1tQyK+FDVBNh/GlTxyNJ0mjS7hTxTg8vqW5PPvkkM2fO5PHHH2fNmjWceeaZzJxZTbWwdu1aZs2axc9+9jNe9apXcdFFF7HllluyZMkSTjzxRJ588knGjx/P3LlzmThx4qBxMpMf/ehHfPOb3wRg1qxZnHHGGRx33HEb1P4me7AmAgsj4lbgRmBBZn4XOBk4MSKWUd1j1ZsmXgDsUMpPBE4ByMzbgMuA24HvAx8pQw/XAscDVwN3AJeVupIkSZJGqLFjx3L55Zdz0003sXDhQk466SQyqzuB7rzzTubMmcOtt97KNttsw3nnnceaNWs44YQTmD9/PkuWLOGYY47htNNOGzLOypUrGTduHJttVvU5TZo0iYceemiD299YD1Zm3gq8tp/ye6jux+pb/gzw7gH2dRZwVj/lVwFXbXBjJUmSJA0Lmcmpp57KtddeyyabbMJDDz3EI488AsDkyZM58MADAXj/+9/Pueeey4wZM1i6dClvf/vbAXjuueeG7L3qjdNXNYn5hml0FkFJkiRJWhcXX3wxK1asYMmSJYwZM4YpU6bwzDPPAC9OgCKCzGTPPfdk0aJF6xRn/PjxrF69mrVr17LZZpvR09PDTjvttMHtN8GSJEmS9ALdnFb9iSeeYMcdd2TMmDEsXLiQ+++///l1DzzwAIsWLeINb3gDl1xyCQcddBC77747K1aseL58zZo13HXXXey5556DxokI3vKWtzB//nyOOuoo5s2b9/y9XhuiI7MISpIkSVI7jj76aBYvXsz06dO5+OKLefWrX/38uj322IN58+ax9957s2rVKo477jg233xz5s+fz8knn8w+++zDtGnT+OlPf9pWrLPPPptzzjmH3XbbjZUrV3LsscducPvtwZIkSZLUdU8++SRQDd0baLjf7bff3m/5tGnTuPbaa19UPnfu3EFjvuIVr+CGG25Yt4YOwQRLkrRRum7u6rbqdXMYjCRp42OCJUlqTLtJDpjoSJKac8QRR3Dvvfe+oOzss8/m0EMPrT2WCZYkSZIkMrOWacqHo8svv3y9t+1vOvfBOMmFJEmSNMqNHTuWlStXrnMysbHLTFauXMnYsWPb3sYeLEmSJGmUmzRpEj09PaxYsaLbTRl2xo4dy6RJk9qub4IlSZIkjXJjxoxh11137XYzNgomWNIw4WQAkiRJI5/3YEmSJElSTezBkuTvBUlDOP36p9uue3iD7ZAkDX8mWJI0CjgEVZKkznCIoCRJkiTVxARLkiRJkmpigiVJkiRJNTHBkiRJkqSaOMmFJHWQMzZKkrRxswdLkiRJkmpigiVJkiRJNTHBkiRJkqSaeA+WpK7p1o/f+qO7kiSpKfZgSZIkSVJNTLAkSZIkqSYmWJIkSZJUk8buwYqIycBFwB8AvwfOz8x/iIgzgD8DVpSqp2bmVWWbTwLHAs8BH83Mq0v5DOAfgE2Br2bm50r5rsClwPbATcAHMvPZpo5Jo4P350iSJGl9NTnJxVrgpMy8KSK2BpZExIKy7ouZ+fnWyhHxGuAoYE9gJ+AHEfGqsvpLwNuBHuDGiLgyM28Hzi77ujQivkKVnH25wWOSJHXR6dc/3XbdwxtshyRJA2lsiGBmLs/Mm8ryb4A7gJ0H2WQmcGlm/i4z7wWWAfuVx7LMvKf0Tl0KzIyIAN4KzC/bzwPe1czRSJIkSdLQOnIPVkRMAV4LXF+Kjo+IWyPiwojYrpTtDDzYsllPKRuofAdgdWau7VPeX/w5EbE4IhavWLGivyqSJEmStMEaT7AiYivg34CPZ+avqYbwvRKYBiwHvtBbtZ/Ncz3KX1yYeX5mTs/M6RMmTFjHI5AkSZKk9jT6Q8MRMYYqubo4M78NkJmPtKz/F+C75WkPMLll80nAw2W5v/LHgHERsVnpxWqtL0mSJEkd1+QsggFcANyRmee0lE/MzOXl6RHA0rJ8JfDNiDiHapKLqcANVD1VU8uMgQ9RTYTxp5mZEbEQOJLqvqxZwBVNHY+0vtq9Kd8b8iVJkka+JnuwDgQ+APw8Im4uZacC74uIaVTD+e4D/hwgM2+LiMuA26lmIPxIZj4HEBHHA1dTTdN+YWbeVvZ3MnBpRJwJ/IwqoZM0SpnMSpKkbmsswcrM/6D/+6SuGmSbs4Cz+im/qr/tMvMeqlkGJalrnDpckiT1avQeLElSc0zsNNq1+8Pw/ii8pE4ywZI2Un74liRJ6jwTLA1bfjMpSZKkkcYES5K0zpxQRJKk/plgqaMctrbx8zWWtCG8hkga6UywNGr4jbskSZKatkm3GyBJkiRJGwsTLEmSJEmqiQmWJEmSJNXEBEuSJEmSamKCJUmSJEk1cRZBDckf/JUkSZLaY4I1Svk7I5IkSVL9HCIoSZIkSTUxwZIkSZKkmphgSZIkSVJNTLAkSZIkqSYmWJIkSZJUExMsSZIkSaqJCZYkSZIk1cQES5IkSZJq4g8Nd1m7P/h75v5bNNwSSZIkSRvKHixJkiRJqokJliRJkiTVxARLkiRJkmrSWIIVEZMjYmFE3BERt0XEx0r59hGxICLuLn+3K+UREedGxLKIuDUi9m3Z16xS/+6ImNVS/rqI+HnZ5tyIiKaOR5IkSZKG0uQkF2uBkzLzpojYGlgSEQuA2cAPM/NzEXEKcApwMnAYMLU89ge+DOwfEdsDnwKmA1n2c2VmPl7qzAGuA64CZgDfa/CYuua6uavbrnvA7HENtkSSJEnSQBrrwcrM5Zl5U1n+DXAHsDMwE5hXqs0D3lWWZwIXZeU6YFxETAQOBRZk5qqSVC0AZpR122TmosxM4KKWfUmSJElSx3XkHqyImAK8FrgeeFlmLocqCQN2LNV2Bh5s2aynlA1W3tNPuSRJkiR1ReMJVkRsBfwb8PHM/PVgVfspy/Uo768NcyJicUQsXrFixVBNliRJkqT10miCFRFjqJKrizPz26X4kTK8j/L30VLeA0xu2XwS8PAQ5ZP6KX+RzDw/M6dn5vQJEyZs2EFJkiRJ0gCanEUwgAuAOzLznJZVVwK9MwHOAq5oKf9gmU3wAOCJMoTwauCQiNiuzDh4CHB1WfebiDigxPpgy74kSZIkqeOanEXwQOADwM8j4uZSdirwOeCyiDgWeAB4d1l3FfAOYBnwFPAhgMxcFRGfBW4s9T6TmavK8nHAXGALqtkDN8oZBCVJkiSNDI0lWJn5H/R/nxTAwf3UT+AjA+zrQuDCfsoXA3ttQDMlSZIkqTZN9mBJkiRJGmVOv/7ptuueuf8WDbakO9b5HqxyL9TeTTRGkiRJkkaythKsiPhxRGwTEdsDtwBfi4hzhtpOkiRJkkaTdnuwti2/YfXfga9l5uuAtzXXLEmSJEkaedpNsDYrv1n1HuC7DbZHkiRJkkasdhOsT1P9HtWyzLwxIl4B3N1csyRJkiRp5Gl3FsHlmfn8xBaZeY/3YEmSJEnSC7Xbg/WPbZZJkiRJ0qg1aA9WRLwB+CNgQkSc2LJqG2DTJhsmSZIkSSPNUEMENwe2KvW2bin/NXBkU42SJEmSpJFo0AQrM38C/CQi5mbm/R1qkyRJkqRR4Lq5q9uue8DscQ22pD7tTnLxkog4H5jSuk1mvrWJRkmSJEnSSNRugvWvwFeArwLPNdccSZIkSRq52k2w1mbmlxttiSRJkiSNcO1O0/6diPhfETExIrbvfTTaMkmSJEkaYdrtwZpV/v5lS1kCr6i3OZIkSZI0crWVYGXmrk03RJIkSZJGurYSrIj4YH/lmXlRvc2RJEmSpJGr3SGCr29ZHgscDNwEmGBJkiRJUtHuEMETWp9HxLbA1xtpkSRJkiSNUO3OItjXU8DUOhsiSZIkSSNdu/dgfYdq1kCATYE9gMuaapQkSVInnX79023XPfyO37Vd94DZ49anOZJGsHbvwfp8y/Ja4P7M7GmgPZIkSZI0YrU1RDAzfwL8Atga2A54tslGSZIkSdJI1FaCFRHvAW4A3g28B7g+Io5ssmGSJEmSNNK0O0TwNOD1mfkoQERMAH4AzG+qYZIkSZLW37rcW3jm/ls02JLRpd1ZBDfpTa6KlUNtGxEXRsSjEbG0peyMiHgoIm4uj3e0rPtkRCyLiDsj4tCW8hmlbFlEnNJSvmtEXB8Rd0fEtyJi8zaPRZIkSZIa0W6C9f2IuDoiZkfEbOD/AVcNsc1cYEY/5V/MzGnlcRVARLwGOArYs2xzXkRsGhGbAl8CDgNeA7yv1AU4u+xrKvA4cGybxyJJkiRJjRiqF2q3iDgwM/8S+Gdgb2AfYBFw/mDbZua1wKo22zETuDQzf5eZ9wLLgP3KY1lm3pOZzwKXAjMjIoC38l9DFOcB72ozliRJkiQ1YqgerL8HfgOQmd/OzBMz8xNUvVd/v54xj4+IW8sQwu1K2c7Agy11ekrZQOU7AKszc22f8n5FxJyIWBwRi1esWLGezZYkSZKkwQ2VYE3JzFv7FmbmYmDKesT7MvBKYBqwHPhCKY9+6uZ6lPcrM8/PzOmZOX3ChAnr1mJJkiRJatNQswiOHWTdOk81kpmP9C5HxL8A3y1Pe4DJLVUnAQ+X5f7KHwPGRcRmpRertb4kSZIkdcVQPVg3RsSf9S2MiGOBJesaLCImtjw9AuidYfBK4KiIeElE7ApMpfrdrRuBqWXGwM2pJsK4MjMTWAj0/hbXLOCKdW2PJEmSJNVpqB6sjwOXR8TR/FdCNR3YnCpBGlBEXAK8GRgfET3Ap4A3R8Q0quF89wF/DpCZt0XEZcDtwFrgI5n5XNnP8cDVwKbAhZl5WwlxMnBpRJwJ/Ay4oM1jliRJkqRGDJpglSF9fxQRbwH2KsX/LzN/NNSOM/N9/RQPmARl5lnAWf2UX0U/U8Jn5j1UswxKkiRJ2gDXzV3dVr0DZo9ruCUj31A9WABk5kKqIXmSJEmSpAG0+0PDkiRJkqQhmGBJkiRJUk1MsCRJkiSpJiZYkiRJklQTEyxJkiRJqokJliRJkiTVxARLkiRJkmpigiVJkiRJNTHBkiRJkqSamGBJkiRJUk1MsCRJkiSpJiZYkiRJklQTEyxJkiRJqokJliRJkiTVxARLkiRJkmpigiVJkiRJNTHBkiRJkqSamGBJkiRJUk1MsCRJkiSpJiZYkiRJklQTEyxJkiRJqslm3W6AJEmShnbd3NVt1z1g9rgGWyJpMPZgSZIkSVJNTLAkSZIkqSYmWJIkSZJUk8YSrIi4MCIejYilLWXbR8SCiLi7/N2ulEdEnBsRyyLi1ojYt2WbWaX+3RExq6X8dRHx87LNuRERTR2LJEmSJLWjyR6sucCMPmWnAD/MzKnAD8tzgMOAqeUxB/gyVAkZ8Clgf2A/4FO9SVmpM6dlu76xJEmSJKmjGkuwMvNaYFWf4pnAvLI8D3hXS/lFWbkOGBcRE4FDgQWZuSozHwcWADPKum0yc1FmJnBRy74kSZIkqSs6PU37yzJzOUBmLo+IHUv5zsCDLfV6Stlg5T39lPcrIuZQ9Xaxyy67bOAhSJIk1ef0659uq97hDbdDUj2GyyQX/d0/letR3q/MPD8zp2fm9AkTJqxnEyVJkiRpcJ3uwXokIiaW3quJwKOlvAeY3FJvEvBwKX9zn/Ifl/JJ/dSXJEmShpV2eynP3H+LhluiTuh0D9aVQO9MgLOAK1rKP1hmEzwAeKIMJbwaOCQitiuTWxwCXF3W/SYiDiizB36wZV+SJEmS1BWN9WBFxCVUvU/jI6KHajbAzwGXRcSxwAPAu0v1q4B3AMuAp4APAWTmqoj4LHBjqfeZzOydOOM4qpkKtwC+Vx6SJEmS1DWNJViZ+b4BVh3cT90EPjLAfi4ELuynfDGw14a0UZIkSRourpu7uu26B8we12BLtCGGyyQXkiRJkjTimWBJkiRJUk1MsCRJkiSpJiZYkiRJklQTEyxJkiRJqokJliRJkiTVxARLkiRJkmpigiVJkiRJNTHBkiRJkqSabNbtBkiSJGl4u27u6rbqHTB7XMMtkYY/e7AkSZIkqSYmWJIkSZJUExMsSZIkSaqJCZYkSZIk1cQES5IkSZJqYoIlSZIkSTUxwZIkSZKkmphgSZIkSVJN/KFhSZIkqUW7P6wM/riyXswES5IkScNSu4mOSY6GExMsSZKkUej0659uu+7hDbZD2th4D5YkSZIk1cQES5IkSZJqYoIlSZIkSTXxHixJkiSNCu3ed+Y9Z9oQXenBioj7IuLnEXFzRCwuZdtHxIKIuLv83a6UR0ScGxHLIuLWiNi3ZT+zSv27I2JWN45FkiRJknp1c4jgWzJzWmZOL89PAX6YmXpZnb0AAAuuSURBVFOBH5bnAIcBU8tjDvBlqBIy4FPA/sB+wKd6kzJJkiRJ6obhdA/WTGBeWZ4HvKul/KKsXAeMi4iJwKHAgsxclZmPAwuAGZ1utCRJkiT16tY9WAlcExEJ/HNmng+8LDOXA2Tm8ojYsdTdGXiwZdueUjZQ+YtExByq3i922WWXOo9DkiRJ68Df39LGrlsJ1oGZ+XBJohZExC8GqRv9lOUg5S8urBK48wGmT5/ebx1JkiRJ2lBdGSKYmQ+Xv48Cl1PdQ/VIGfpH+ftoqd4DTG7ZfBLw8CDlkiRJktQVHU+wIuKlEbF17zJwCLAUuBLonQlwFnBFWb4S+GCZTfAA4IkylPBq4JCI2K5MbnFIKZMkSZKkrujGEMGXAZdHRG/8b2bm9yPiRuCyiDgWeAB4d6l/FfAOYBnwFPAhgMxcFRGfBW4s9T6Tmas6dxiSJEmS9EIdT7Ay8x5gn37KVwIH91OewEcG2NeFwIV1t1GSJEmS1sdwmqZdkiRJkkY0EyxJkiRJqokJliRJkiTVxARLkiRJkmpigiVJkiRJNTHBkiRJkqSamGBJkiRJUk1MsCRJkiSpJiZYkiRJklQTEyxJkiRJqokJliRJkiTVxARLkiRJkmpigiVJkiRJNTHBkiRJkqSamGBJkiRJUk1MsCRJkiSpJiZYkiRJklQTEyxJkiRJqokJliRJkiTVxARLkiRJkmpigiVJkiRJNTHBkiRJkqSamGBJkiRJUk1MsCRJkiSpJiZYkiRJklQTEyxJkiRJqsmIT7AiYkZE3BkRyyLilG63R5IkSdLoNaITrIjYFPgScBjwGuB9EfGa7rZKkiRJ0mg1ohMsYD9gWWbek5nPApcCM7vcJkmSJEmjVGRmt9uw3iLiSGBGZn64PP8AsH9mHt+n3hxgTnm6O3BnRxtaj/HAY6MstnE3/tijLW43Y4+2uN2MbdyNP/Zoi9vN2Mbd+GN385g31Mszc0Lfws260ZIaRT9lL8oYM/N84Pzmm9OciFicmdNHU2zjbvyxR1vcbsYebXG7Gdu4G3/s0Ra3m7GNu/HH7uYxN2WkDxHsASa3PJ8EPNyltkiSJEka5UZ6gnUjMDUido2IzYGjgCu73CZJkiRJo9SIHiKYmWsj4njgamBT4MLMvK3LzWpKN4c4diu2cTf+2KMtbjdjj7a43Yxt3I0/9miL283Yxt34Y4/o23j6M6InuZAkSZKk4WSkDxGUJEmSpGHDBEuSJEmSamKCNQJExIyIuDMilkXEKR2KeWFEPBoRSzsRryXu5IhYGBF3RMRtEfGxDsYeGxE3RMQtJfanOxW7xN80In4WEd/tYMz7IuLnEXFzRCzuVNwSe1xEzI+IX5TX+w0diLl7Odbex68j4uNNxy2xP1HeV0sj4pKIGNuhuB8rMW9r+lj7u25ExPYRsSAi7i5/t+tQ3HeXY/59RDQ2/e8Asf+uvK9vjYjLI2Jch+J+tsS8OSKuiYidOhG3Zd1fRERGxPhOxI2IMyLioZZ/z++oO+5AsUv5CeX/5tsi4m87ETcivtVyvPdFxM0dijstIq7r/b8iIvarO+4gsfeJiEXl/6rvRMQ2DcTt97NH09evQeI2ev0aJG4nrl0DxW78+tVRmeljGD+oJu/4JfAKYHPgFuA1HYj7RmBfYGmHj3cisG9Z3hq4qxPHW+IFsFVZHgNcDxzQwWM/Efgm8N0OxrwPGN/J17gl9jzgw2V5c2Bch+NvCvyK6kcCm461M3AvsEV5fhkwuwNx9wKWAltSTWr0A2Bqg/FedN0A/hY4pSyfApzdobh7UP2w/I+B6R0+5kOAzcry2R085m1alj8KfKUTcUv5ZKoJp+5v4poywPGeAfxFU6/tELHfUv49vaQ837FT57pl/ReAv+nQ8V4DHFaW3wH8uIPn+kbgTWX5GOCzDcTt97NH09evQeI2ev0aJG4nrl0DxW78+tXJhz1Yw99+wLLMvCcznwUuBWY2HTQzrwVWNR2nn7jLM/Omsvwb4A6qD6ediJ2Z+WR5OqY8OjILTERMAv4E+Gon4nVb+QbyjcAFAJn5bGau7nAzDgZ+mZn3dyjeZsAWEbEZVcLTid/s2wO4LjOfysy1wE+AI5oKNsB1YyZVMk35+65OxM3MOzLzzrpjtRn7mnK+Aa6j+o3GTsT9dcvTl9LA9WuQ/xu+CPxVEzGHiNu4AWIfB3wuM39X6jzaobgAREQA7wEu6VDcBHp7jraloevXALF3B64tywuA/9FA3IE+ezR6/RoobtPXr0HiduLaNVDsxq9fnWSCNfztDDzY8ryHDiUc3RYRU4DXUvUkdSrmpmXIxaPAgszsVOy/p/pw8vsOxeuVwDURsSQi5nQw7iuAFcDXohoW+dWIeGkH40P1u3m1fzjpT2Y+BHweeABYDjyRmdd0IPRS4I0RsUNEbEn1zfPkIbap28syczlU/7ECO3Y4frcdA3yvU8Ei4qyIeBA4GvibDsV8J/BQZt7SiXh9HF+GFV3YxPDTQbwK+OOIuD4ifhIRr+9gbIA/Bh7JzLs7FO/jwN+V99bngU92KC5U17F3luV30/A1rM9nj45dv7rxmWeIuI1fu/rG7sb1qykmWMNf9FM2orP6dkTEVsC/AR/v861GozLzucycRvWtzX4RsVfTMSPicODRzFzSdKx+HJiZ+wKHAR+JiDd2KO5mVMNAvpyZrwV+SzX8oiOi+mHydwL/2qF421F9E7orsBPw0oh4f9NxM/MOqmEeC4DvUw0xXjvoRqpNRJxGdb4v7lTMzDwtMyeXmMc3Ha8k7qfRnQ9DXwZeCUyj+uLiCx2MvRmwHXAA8JfAZaVXqVPeR4e+ICqOAz5R3lufoIw+6JBjqP5/WkI1pOzZpgJ167PHcIvbiWtXf7E7ff1qkgnW8NfDC7+tmURnhhZ1TUSMofpHd3FmfrsbbSjD1X4MzOhAuAOBd0bEfVRDQN8aEd/oQFwy8+Hy91HgcqohqZ3QA/S09BDOp0q4OuUw4KbMfKRD8d4G3JuZKzJzDfBt4I86ETgzL8jMfTPzjVRDbzr1jXevRyJiIkD5W/tQquEoImYBhwNHZ2Y3vhT7Jg0MperHK6m+OLilXMMmATdFxB80HTgzHylfiv0e+Bc6d/2C6hr27TK0/Aaq0Qe1T+7RnzLM+L8D3+pEvGIW1XULqi+mOnauM/MXmXlIZr6OKqn8ZRNxBvjs0fj1q1ufeQaK24lrVxvH3KnrV2NMsIa/G4GpEbFr+db9KODKLrepMeUbwAuAOzLznA7HntA7Y05EbEH1ofgXTcfNzE9m5qTMnEL1+v4oMxvv3YiIl0bE1r3LVDe3dmTWyMz8FfBgROxeig4Gbu9E7KLT3/4+ABwQEVuW9/jBVOPOGxcRO5a/u1B9KOvkcUN1vZpVlmcBV3Q4fsdFxAzgZOCdmflUB+NObXn6Tjpz/fp5Zu6YmVPKNayH6gb2XzUdu/eDb3EEHbp+Ff8XeGtpx6uoJup5rEOx3wb8IjN7OhQPqi9231SW30oHv6hpuYZtApwOfKWBGAN99mj0+tWtzzwDxe3EtWuQ2B2/fjWq76wXPobfg+q+ibuovrU5rUMxL6EacrGG6j/MYzsU9yCqIZC3AjeXxzs6FHtv4Gcl9lIamJ2pjTa8mQ7NIkh1H9Qt5XFbp95bLfGnAYvL+f6/wHYdirslsBLYtsPH+2mq/zCWAl+nzD7Wgbj/TpW83gIc3HCsF103gB2AH1J9IPshsH2H4h5Rln8HPAJc3cFjXkZ172zvNayJ2fz6i/tv5f11K/AdqhvHG4/bZ/19NDOLYH/H+3Xg5+V4rwQmdvA13hz4RjnfNwFv7dS5BuYC/7OJYx3keA8ClpTryPXA6zoY+2NUn4HuAj4HRANx+/3s0fT1a5C4jV6/BonbiWvXQLEbv3518hHlYCVJkiRJG8ghgpIkSZJUExMsSZIkSaqJCZYkSZIk1cQES5IkSZJqYoIlSZIkSTUxwZIkSZKkmphgSZIkSVJN/j+ttKPGcSjqeAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = np.arange(len(hour_df['hour']))  # the label locations\n",
    "width = 0.35  # the width of the bars\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(12,4)\n",
    "_ = rects1 = ax.bar(x - width/2, hour_df['label_1'], width, label='label_1', color='xkcd:lightblue')\n",
    "_ = rects2 = ax.bar(x + width/2, hour_df['label_0'], width, label='label_0', color='xkcd:lavender')\n",
    "\n",
    "# Add some text for labels, title and custom x-axis tick labels, etc.\n",
    "_ = ax.set_ylabel('Counts')\n",
    "_ = ax.set_title('Distribution of label counts per hour')\n",
    "_ = ax.set_xticks(x)\n",
    "_ = ax.set_xticklabels(hour_df['hour'])\n",
    "_ = ax.legend()\n",
    "\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "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 = plt.figure()\n",
    "_ = fig.set(alpha=0.5)\n",
    "fig.set_size_inches(15,4)\n",
    "\n",
    "_ = plt.subplot2grid((1,2),(0,0))\n",
    "_ = plt.plot( 'dayofweek', 'label_1_%', data=dayofweek_df, marker='o', markerfacecolor='xkcd:teal', markersize=4, color='xkcd:teal', linewidth=1.5)\n",
    "_ = plt.plot( 'dayofweek', 'label_0_%', data=dayofweek_df, marker='o', markerfacecolor='xkcd:goldenrod', markersize=4, color='xkcd:goldenrod', linewidth=1.5)\n",
    "_ = plt.legend()\n",
    "_ = plt.title(u\"Distribution of label percentages per dayofweek\")\n",
    "\n",
    "_ = plt.subplot2grid((1,2),(0,1))\n",
    "_ = plt.plot( 'hour', 'label_1_%', data=hour_df, marker='o', markerfacecolor='xkcd:azure', markersize=3, color='xkcd:azure', linewidth=1)\n",
    "_ = plt.plot( 'hour', 'label_0_%', data=hour_df, marker='o', markerfacecolor='xkcd:salmon', markersize=3, color='xkcd:salmon', linewidth=1)\n",
    "_ = plt.legend()\n",
    "_ = plt.xticks(np.arange(len(hour_df['hour'])))\n",
    "_ = plt.title(u\"Distribution of label percentages per hour\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Remarks:**\n",
    "\n",
    "The graphs are more intuitive to show the relationships. We can see obviously that there are several features contributing to label 0(if we assume that label 0 refers to fraud)\n",
    "\n",
    "The first two pictures show that fraud behaviour concentrates in OPPO and VIVO, meaning that phones made by these two manufactures are more easily to be made use of in terms of faking ip or others.\n",
    "\n",
    "The second two bar charts show that the suspicious behaviour usually coincides with not defined language, says 'nan', and in general the normal behaviour will display 'chinese' as the language.  \n",
    "\n",
    "Other graphs indicate that the probability of ad fraud changes with hour and day of a week."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Prepare Dataset for modeling\n",
    "### 4.1 Correlation significant test  for feature selection  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- #### Chi Square Analysis   \n",
    "\n",
    "Simple correlation test using DateFrame.corr() can only be used in tesing between continuous variables. Given that our label is nominal and so are many independent variables, we use Chi Square analysis to test the correlation between them: According to the P value results, 'sid' and 'idfamd5' are uncorrelated features which should be discarded."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Feature Selection    \n",
    "# traindata.columns\n",
    "# traindata.dtypes == 'object'\n",
    "\n",
    "nominalColumns = [ 'sid', 'pkgname', 'ver', 'adunitshowid', 'mediashowid',\n",
    "       'apptype', 'ip', 'city', 'province', 'reqrealip',\n",
    "       'adidmd5', 'imeimd5', 'idfamd5', 'openudidmd5', 'macmd5', 'dvctype',\n",
    "       'model', 'make', 'ntt', 'carrier', 'os', 'osv', 'orientation', 'lan',\n",
    "       'date', 'dayofweek', 'hour',\n",
    "       'minute', 'hourtosecond', 'ip_top', 'reqrealip_top', 'apptype_top1',\n",
    "       'apptype_top2', 'osv_clean', 'osv_top1',\n",
    "       'osv_top2', 'osv_top3', 'ver_clean', 'ver_top1', 'ver_top2',\n",
    "       'ver_top3']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Staring testing...\n",
      "sid - NOT important predictor. Chi-Statistic: 1000000.00, P-value: 0.4995  ==>  Discard\n",
      "pkgname - IMPORTANT predictor. Chi-Statistic: 407817.87, P-value: 0.0000\n",
      "ver - IMPORTANT predictor. Chi-Statistic: 395837.76, P-value: 0.0000\n",
      "adunitshowid - IMPORTANT predictor. Chi-Statistic: 552043.97, P-value: 0.0000\n",
      "mediashowid - IMPORTANT predictor. Chi-Statistic: 500428.72, P-value: 0.0000\n",
      "apptype - IMPORTANT predictor. Chi-Statistic: 362222.62, P-value: 0.0000\n",
      "ip - IMPORTANT predictor. Chi-Statistic: 787269.08, P-value: 0.0000\n",
      "city - IMPORTANT predictor. Chi-Statistic: 16142.63, P-value: 0.0000\n",
      "province - IMPORTANT predictor. Chi-Statistic: 8908.61, P-value: 0.0000\n",
      "reqrealip - IMPORTANT predictor. Chi-Statistic: 451010.96, P-value: 0.0000\n",
      "adidmd5 - IMPORTANT predictor. Chi-Statistic: 754220.07, P-value: 0.0000\n",
      "imeimd5 - IMPORTANT predictor. Chi-Statistic: 996880.35, P-value: 0.0000\n",
      "idfamd5 - NOT important predictor. Chi-Statistic: 369.80, P-value: 0.3224  ==>  Discard\n",
      "openudidmd5 - IMPORTANT predictor. Chi-Statistic: 80604.47, P-value: 0.0000\n",
      "macmd5 - IMPORTANT predictor. Chi-Statistic: 385148.88, P-value: 0.0000\n",
      "dvctype - IMPORTANT predictor. Chi-Statistic: 185741.11, P-value: 0.0000\n",
      "model - IMPORTANT predictor. Chi-Statistic: 408810.29, P-value: 0.0000\n",
      "make - IMPORTANT predictor. Chi-Statistic: 230965.12, P-value: 0.0000\n",
      "ntt - IMPORTANT predictor. Chi-Statistic: 62160.62, P-value: 0.0000\n",
      "carrier - IMPORTANT predictor. Chi-Statistic: 22695.02, P-value: 0.0000\n",
      "os - IMPORTANT predictor. Chi-Statistic: 46275.12, P-value: 0.0000\n",
      "osv - IMPORTANT predictor. Chi-Statistic: 348172.25, P-value: 0.0000\n",
      "orientation - IMPORTANT predictor. Chi-Statistic: 52635.50, P-value: 0.0000\n",
      "lan - IMPORTANT predictor. Chi-Statistic: 152273.38, P-value: 0.0000\n",
      "date - IMPORTANT predictor. Chi-Statistic: 377.35, P-value: 0.0000\n",
      "dayofweek - IMPORTANT predictor. Chi-Statistic: 372.23, P-value: 0.0000\n",
      "hour - IMPORTANT predictor. Chi-Statistic: 1319.54, P-value: 0.0000\n",
      "minute - IMPORTANT predictor. Chi-Statistic: 98.58, P-value: 0.0009\n",
      "hourtosecond - IMPORTANT predictor. Chi-Statistic: 87366.07, P-value: 0.0005\n",
      "ip_top - IMPORTANT predictor. Chi-Statistic: 4429.45, P-value: 0.0000\n",
      "reqrealip_top - IMPORTANT predictor. Chi-Statistic: 272469.30, P-value: 0.0000\n",
      "apptype_top1 - IMPORTANT predictor. Chi-Statistic: 59159.56, P-value: 0.0000\n",
      "apptype_top2 - IMPORTANT predictor. Chi-Statistic: 292939.44, P-value: 0.0000\n",
      "osv_clean - IMPORTANT predictor. Chi-Statistic: 346120.17, P-value: 0.0000\n",
      "osv_top1 - IMPORTANT predictor. Chi-Statistic: 263725.99, P-value: 0.0000\n",
      "osv_top2 - IMPORTANT predictor. Chi-Statistic: 335951.97, P-value: 0.0000\n",
      "osv_top3 - IMPORTANT predictor. Chi-Statistic: 346083.11, P-value: 0.0000\n",
      "ver_clean - IMPORTANT predictor. Chi-Statistic: 394238.19, P-value: 0.0000\n",
      "ver_top1 - IMPORTANT predictor. Chi-Statistic: 154722.26, P-value: 0.0000\n",
      "ver_top2 - IMPORTANT predictor. Chi-Statistic: 298351.51, P-value: 0.0000\n",
      "ver_top3 - IMPORTANT predictor. Chi-Statistic: 361931.94, P-value: 0.0000\n"
     ]
    }
   ],
   "source": [
    "## Introduce ChiSquare Class     \n",
    "\n",
    "print('Staring testing...')\n",
    "class ChiSquare:\n",
    "    def __init__(self, dataframe):\n",
    "        \n",
    "        self.df = dataframe\n",
    "        self.p = None #P-Value\n",
    "        self.chi2 = None  #Chi Test Statistic\n",
    "        self.dof = None\n",
    "        self.dfObserved = None\n",
    "        self.dfExpected = None\n",
    "        \n",
    "    def _print_chisquare_result(self, colX, alpha):\n",
    "        result = \"\"\n",
    "        if self.p<alpha:\n",
    "            result=\"{} - IMPORTANT predictor. Chi-Statistic: {:.2f}, P-value: {:.4f}\".format(colX,self.chi2,self.p) \n",
    "        else:\n",
    "            result=\"{} - NOT important predictor. Chi-Statistic: {:.2f}, P-value: {:.4f}  ==>  Discard\".format(colX,self.chi2,self.p)\n",
    "\n",
    "        print(result)\n",
    "        \n",
    "    def TestIndependence(self, colX, colY, alpha=0.05):\n",
    "        X = self.df[colX].astype(str)\n",
    "        Y = self.df[colY].astype(str)\n",
    "        \n",
    "        self.dfObserved =  pd.crosstab(Y,X)\n",
    "        chi2, p, dof, expected = stats.chi2_contingency(self.dfObserved.values)\n",
    "        self.p = p\n",
    "        self.chi2 = chi2\n",
    "        self.dof = dof \n",
    "        self.dfExpected = pd.DataFrame(expected, columns=self.dfObserved.columns, index = self.dfObserved.index)\n",
    "        self._print_chisquare_result(colX,alpha)\n",
    "        \n",
    "        \n",
    "## Initialize ChiSquare Class   \n",
    "   \n",
    "cT = ChiSquare(traindata)\n",
    "\n",
    "for var in nominalColumns:\n",
    "    cT.TestIndependence(colX=var, colY='label')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- #### ANOVA Analysis  \n",
    "\n",
    "ANOVA cannot prove causation. It can tell us whether the mean value of independent variables of each label group are different at a statistically significant level. In this case taking 'h' for an example, our null hypothesis is that all group mean height are the same.\n",
    "\n",
    "We start by visualizing the impact of each label on the height variable with a box plot usinf Seaborn data visualization package. Figure 1 tells us that label 0 seems to have more zero than label 1 and have larger standard deviations; The mean height of each lable is similar. \n",
    "\n",
    "Analysis of Variance or ANOVA tells us whether the differences in the variance between each label group are statistically significant when compared to the differences within each group.\n",
    "\n",
    "One of the core assumptions with ANOVA is that the dependent variable is normally distributed.   \n",
    "However, the distribution of h is not normal, so ANOVA is not appropriate to use in this situation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Feature Selection    \n",
    "continuousColumns = ['h', 'w', 'ppi', 'h_w', 'hw_ratio', 'inch']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set(font_scale=1.3)  #sets font size for chart scales\n",
    "ax = sns.boxplot(x = \"label\", y = \"h\", data = traindata, # calls the boxplot\n",
    "                 palette=\"Set1\", # sets color scheme \n",
    "                 linewidth = 1.5) # sets chart line width\n",
    "_ = ax.set_title(\"Figure 1. Boxplot of h by label Type\\n\", fontsize=16) # title \n",
    "_ = ax.set_ylabel(\"h\", rotation = 0, fontsize=15, labelpad=30) # y-axis label\n",
    "_ = ax.set_xlabel(\"label\", rotation = 0, fontsize=15) # x-axis label\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "_ = plt.hist(traindata['h']) # check for variable normality in h dependent variable. \n",
    "#Data looks not normally distributed.\n",
    "_ = plt.title(\"Figure 2. Distribution of h Variable\\n\")\n",
    "_ = plt.xlabel(\"h\")\n",
    "_ = plt.ylabel(\"Count\\n of labels\", rotation = 0, labelpad = 40)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NormaltestResult(statistic=1373260.835350655, pvalue=0.0)\n"
     ]
    }
   ],
   "source": [
    "# check p-score of normal distribution\n",
    "print(stats.normaltest(traindata['h'], axis=0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we create an Ordinary Least Squares (OLS) model as a precursor to the ANOVA.   \n",
    "\n",
    "Before the ANOVA, we check the OLS residuals are normally distributed. This is another important prior assumption:    \n",
    "The residuals look not normal at all. Thus, we cannot run ANOVA here.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Figure 3. Distribution of OLS Model Residuals')"
      ]
     },
     "execution_count": 109,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0, 'Residual Values')"
      ]
     },
     "execution_count": 109,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Count')"
      ]
     },
     "execution_count": 109,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "anova_mod = ols(\"label ~ h\", traindata).fit()\n",
    "resids = statsmodels.regression.linear_model.RegressionResults.resid(anova_mod) # grabs the residual values from the OLS model\n",
    "hist2 = plt.hist(resids, bins = 'auto', color='dodgerblue') # check for variable normality in weight loss dependent variable. \n",
    "#Data looks not normally distributed.\n",
    "_ = plt.title(\"Figure 3. Distribution of OLS Model Residuals\")\n",
    "_ = plt.xlabel(\"Residual Values\")\n",
    "_ = plt.ylabel(\"Count\", rotation = 0, labelpad = 40)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Call for ANOVA function if residuals are normally distributed.\n",
    "\n",
    "table = sm.stats.anova_lm(model, typ=2) # Type 2 ANOVA DataFrame\n",
    "print(table) \n",
    "\n",
    "# In a one-way ANOVA the null hypothesis is that the means of the Diet types are equal. We can reject that hypothesis\n",
    "# because P < 0.05 "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.2 Label Encoding for non-numerical columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['sid', 'label', 'pkgname', 'ver', 'adunitshowid', 'mediashowid',\n",
       "       'apptype', 'nginxtime', 'ip', 'city', 'province', 'reqrealip',\n",
       "       'adidmd5', 'imeimd5', 'idfamd5', 'openudidmd5', 'macmd5', 'dvctype',\n",
       "       'model', 'make', 'ntt', 'carrier', 'os', 'osv', 'orientation', 'lan',\n",
       "       'h', 'w', 'ppi', 'nginxtime_clean', 'date', 'dayofweek', 'hour',\n",
       "       'minute', 'hourtosecond', 'ip_top', 'reqrealip_top', 'apptype_top1',\n",
       "       'apptype_top2', 'h_w', 'hw_ratio', 'inch', 'osv_clean', 'osv_top1',\n",
       "       'osv_top2', 'osv_top3', 'ver_clean', 'ver_top1', 'ver_top2',\n",
       "       'ver_top3'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "traindata.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata_1 = traindata.loc[:,['label', 'pkgname', 'ver', 'adunitshowid', 'mediashowid',\n",
    "       'apptype', 'ip', 'city', 'province', 'reqrealip',\n",
    "       'adidmd5', 'imeimd5',  'openudidmd5', 'macmd5', 'dvctype',\n",
    "       'model', 'make', 'ntt', 'carrier', 'os', 'osv', 'orientation', 'lan',\n",
    "       'h', 'w', 'ppi', 'date', 'dayofweek', 'hour', 'minute', 'hourtosecond', \n",
    "       'ip_top', 'reqrealip_top', 'apptype_top1', 'apptype_top2', \n",
    "       'h_w', 'hw_ratio', 'inch', 'osv_clean',  'osv_top1', 'osv_top2', 'osv_top3',\n",
    "       'ver_clean', 'ver_top1', 'ver_top2', 'ver_top3']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Transform non-numerical data to numerical labels (0 to n_classes-1)\n",
    "le = LabelEncoder()\n",
    "for col in traindata_1.columns:\n",
    "    if traindata_1[col].dtype == \"object\":\n",
    "        traindata_1[col] = le.fit_transform(traindata_1[col])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Check new dataset with all numeric data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1000000, 50)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(1000000, 46)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>pkgname</th>\n",
       "      <th>ver</th>\n",
       "      <th>adunitshowid</th>\n",
       "      <th>mediashowid</th>\n",
       "      <th>apptype</th>\n",
       "      <th>ip</th>\n",
       "      <th>city</th>\n",
       "      <th>province</th>\n",
       "      <th>reqrealip</th>\n",
       "      <th>adidmd5</th>\n",
       "      <th>imeimd5</th>\n",
       "      <th>openudidmd5</th>\n",
       "      <th>macmd5</th>\n",
       "      <th>dvctype</th>\n",
       "      <th>model</th>\n",
       "      <th>make</th>\n",
       "      <th>ntt</th>\n",
       "      <th>carrier</th>\n",
       "      <th>os</th>\n",
       "      <th>osv</th>\n",
       "      <th>orientation</th>\n",
       "      <th>lan</th>\n",
       "      <th>h</th>\n",
       "      <th>w</th>\n",
       "      <th>ppi</th>\n",
       "      <th>date</th>\n",
       "      <th>dayofweek</th>\n",
       "      <th>hour</th>\n",
       "      <th>minute</th>\n",
       "      <th>hourtosecond</th>\n",
       "      <th>ip_top</th>\n",
       "      <th>reqrealip_top</th>\n",
       "      <th>apptype_top1</th>\n",
       "      <th>apptype_top2</th>\n",
       "      <th>h_w</th>\n",
       "      <th>hw_ratio</th>\n",
       "      <th>inch</th>\n",
       "      <th>osv_clean</th>\n",
       "      <th>osv_top1</th>\n",
       "      <th>osv_top2</th>\n",
       "      <th>osv_top3</th>\n",
       "      <th>ver_clean</th>\n",
       "      <th>ver_top1</th>\n",
       "      <th>ver_top2</th>\n",
       "      <th>ver_top3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>402</td>\n",
       "      <td>174</td>\n",
       "      <td>452</td>\n",
       "      <td>239</td>\n",
       "      <td>280.0</td>\n",
       "      <td>475455</td>\n",
       "      <td>229</td>\n",
       "      <td>17</td>\n",
       "      <td>5102</td>\n",
       "      <td>652267</td>\n",
       "      <td>535931</td>\n",
       "      <td>73758</td>\n",
       "      <td>199768</td>\n",
       "      <td>2.0</td>\n",
       "      <td>6202</td>\n",
       "      <td>816</td>\n",
       "      <td>2.0</td>\n",
       "      <td>46000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>136</td>\n",
       "      <td>0.0</td>\n",
       "      <td>9</td>\n",
       "      <td>2340.0</td>\n",
       "      <td>1080.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>9</td>\n",
       "      <td>7</td>\n",
       "      <td>14</td>\n",
       "      <td>31</td>\n",
       "      <td>143141</td>\n",
       "      <td>183</td>\n",
       "      <td>180</td>\n",
       "      <td>2</td>\n",
       "      <td>28</td>\n",
       "      <td>2527200.0</td>\n",
       "      <td>2.2</td>\n",
       "      <td>859.1</td>\n",
       "      <td>138</td>\n",
       "      <td>8</td>\n",
       "      <td>81</td>\n",
       "      <td>810</td>\n",
       "      <td>74</td>\n",
       "      <td>1</td>\n",
       "      <td>11</td>\n",
       "      <td>111</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2174</td>\n",
       "      <td>3180</td>\n",
       "      <td>45</td>\n",
       "      <td>61</td>\n",
       "      <td>319.0</td>\n",
       "      <td>41248</td>\n",
       "      <td>119</td>\n",
       "      <td>18</td>\n",
       "      <td>8127</td>\n",
       "      <td>456711</td>\n",
       "      <td>402019</td>\n",
       "      <td>50767</td>\n",
       "      <td>18010</td>\n",
       "      <td>2.0</td>\n",
       "      <td>602</td>\n",
       "      <td>332</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>175</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6</td>\n",
       "      <td>1080.0</td>\n",
       "      <td>2040.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>9</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>25</td>\n",
       "      <td>32540</td>\n",
       "      <td>106</td>\n",
       "      <td>47</td>\n",
       "      <td>3</td>\n",
       "      <td>31</td>\n",
       "      <td>2203200.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>140</td>\n",
       "      <td>9</td>\n",
       "      <td>9</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>230</td>\n",
       "      <td>626</td>\n",
       "      <td>422</td>\n",
       "      <td>282</td>\n",
       "      <td>273.0</td>\n",
       "      <td>563455</td>\n",
       "      <td>296</td>\n",
       "      <td>5</td>\n",
       "      <td>7875</td>\n",
       "      <td>144447</td>\n",
       "      <td>232089</td>\n",
       "      <td>73758</td>\n",
       "      <td>106563</td>\n",
       "      <td>2.0</td>\n",
       "      <td>4816</td>\n",
       "      <td>572</td>\n",
       "      <td>2.0</td>\n",
       "      <td>46000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>136</td>\n",
       "      <td>1.0</td>\n",
       "      <td>9</td>\n",
       "      <td>2196.0</td>\n",
       "      <td>1080.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>9</td>\n",
       "      <td>7</td>\n",
       "      <td>14</td>\n",
       "      <td>5</td>\n",
       "      <td>140545</td>\n",
       "      <td>223</td>\n",
       "      <td>39</td>\n",
       "      <td>2</td>\n",
       "      <td>27</td>\n",
       "      <td>2371680.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>138</td>\n",
       "      <td>8</td>\n",
       "      <td>81</td>\n",
       "      <td>810</td>\n",
       "      <td>1188</td>\n",
       "      <td>3</td>\n",
       "      <td>32</td>\n",
       "      <td>321</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   label  pkgname   ver  adunitshowid  mediashowid  apptype      ip  city  \\\n",
       "0      1      402   174           452          239    280.0  475455   229   \n",
       "1      1     2174  3180            45           61    319.0   41248   119   \n",
       "2      0      230   626           422          282    273.0  563455   296   \n",
       "\n",
       "   province  reqrealip  adidmd5  imeimd5  openudidmd5  macmd5  dvctype  model  \\\n",
       "0        17       5102   652267   535931        73758  199768      2.0   6202   \n",
       "1        18       8127   456711   402019        50767   18010      2.0    602   \n",
       "2         5       7875   144447   232089        73758  106563      2.0   4816   \n",
       "\n",
       "   make  ntt  carrier  os  osv  orientation  lan       h       w  ppi  date  \\\n",
       "0   816  2.0  46000.0   0  136          0.0    9  2340.0  1080.0  3.0     9   \n",
       "1   332  5.0      0.0   1  175          0.0    6  1080.0  2040.0  0.0     9   \n",
       "2   572  2.0  46000.0   0  136          1.0    9  2196.0  1080.0  0.0     9   \n",
       "\n",
       "   dayofweek  hour  minute  hourtosecond  ip_top  reqrealip_top  apptype_top1  \\\n",
       "0          7    14      31        143141     183            180             2   \n",
       "1          7     3      25         32540     106             47             3   \n",
       "2          7    14       5        140545     223             39             2   \n",
       "\n",
       "   apptype_top2        h_w  hw_ratio   inch  osv_clean  osv_top1  osv_top2  \\\n",
       "0            28  2527200.0       2.2  859.1        138         8        81   \n",
       "1            31  2203200.0       0.5    0.0        140         9         9   \n",
       "2            27  2371680.0       2.0    0.0        138         8        81   \n",
       "\n",
       "   osv_top3  ver_clean  ver_top1  ver_top2  ver_top3  \n",
       "0       810         74         1        11       111  \n",
       "1         9          0         0         0         0  \n",
       "2       810       1188         3        32       321  "
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "traindata.shape\n",
    "traindata_1.shape\n",
    "traindata_1.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata_1.to_csv('round1_iflyad_anticheat_train_clean.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# delete idle dataframe to save memory\n",
    "#del traindata"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---  End of EDA part  ---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "_____"
   ]
  }
 ],
 "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": "3.7.2"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}