chusai_month_baseline_pretreatment.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from pandas import Series, DataFrame\n",
    "import numpy as np\n",
    "from scipy.stats import mode\n",
    "import csv\n",
    "import matplotlib.dates\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.pylab as pylab\n",
    "import seaborn as sns\n",
    "from datetime import *\n",
    "from datetime import datetime\n",
    "import matplotlib.pylab as pylab\n",
    "from pylab import *  \n",
    "mpl.rcParams['font.sans-serif'] = ['SimHei']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 98924 entries, 0 to 98923\n",
      "Data columns (total 4 columns):\n",
      "user_id       98924 non-null int64\n",
      "age           98924 non-null int64\n",
      "sex           98924 non-null int64\n",
      "user_lv_cd    98924 non-null int64\n",
      "dtypes: int64(4)\n",
      "memory usage: 3.0 MB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "#首先导入用户基本信息，作为每个月的样本，构造训练集和测试集\n",
    "#用户基本信息表--jdata_user_basic_info\n",
    "jdata_user_basic_info_data = pd.read_csv('jdata_user_basic_info.csv')\n",
    "print(jdata_user_basic_info_data.info())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 792723 entries, 0 to 792722\n",
      "Data columns (total 6 columns):\n",
      "user_id      792723 non-null int64\n",
      "sku_id       792723 non-null int64\n",
      "o_id         792723 non-null int64\n",
      "o_date       792723 non-null object\n",
      "o_area       792723 non-null int64\n",
      "o_sku_num    792723 non-null int64\n",
      "dtypes: int64(5), object(1)\n",
      "memory usage: 36.3+ MB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "#用户订单表--jdata_user_order\n",
    "jdata_user_order_data = pd.read_csv('jdata_user_order.csv')\n",
    "print(jdata_user_order_data.info())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 6944141 entries, 0 to 6944140\n",
      "Data columns (total 5 columns):\n",
      "user_id    int64\n",
      "sku_id     int64\n",
      "a_date     object\n",
      "a_num      int64\n",
      "a_type     int64\n",
      "dtypes: int64(4), object(1)\n",
      "memory usage: 264.9+ MB\n",
      "None\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 99412 entries, 0 to 99411\n",
      "Data columns (total 6 columns):\n",
      "sku_id    99412 non-null int64\n",
      "price     99412 non-null float64\n",
      "cate      99412 non-null int64\n",
      "para_1    99412 non-null float64\n",
      "para_2    99412 non-null int64\n",
      "para_3    99412 non-null int64\n",
      "dtypes: float64(2), int64(4)\n",
      "memory usage: 4.6 MB\n",
      "None\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 224284 entries, 0 to 224283\n",
      "Data columns (total 4 columns):\n",
      "user_id              224284 non-null int64\n",
      "comment_create_tm    224284 non-null object\n",
      "o_id                 224284 non-null int64\n",
      "score_level          224284 non-null int64\n",
      "dtypes: int64(3), object(1)\n",
      "memory usage: 6.8+ MB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "#导入剩余的3张表\n",
    "#用户行为表--jdata_user_action\n",
    "jdata_user_action_data = pd.read_csv('jdata_user_action.csv')\n",
    "print(jdata_user_action_data.info())\n",
    "\n",
    "#SKU基本信息表--jdata_sku_basic_info\n",
    "jdata_sku_basic_info_data = pd.read_csv('jdata_sku_basic_info.csv')\n",
    "print(jdata_sku_basic_info_data.info())\n",
    "\n",
    "#评论分数数据表--jdata_user_comment_score\n",
    "jdata_user_comment_score_data = pd.read_csv('jdata_user_comment_score.csv')\n",
    "print(jdata_user_comment_score_data.info())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#定义时间处理函数\n",
    "def dealDateColumns(df, columnName):\n",
    "    df[columnName] = pd.to_datetime(df[columnName])\n",
    "    df['year'] = df[columnName].map(lambda x: x.year)\n",
    "    df['month'] = df[columnName].map(lambda x: x.month)\n",
    "    df['day'] = df[columnName].map(lambda x: x.day)\n",
    "    return df\n",
    "\n",
    "jdata_user_order_data = dealDateColumns(jdata_user_order_data, 'o_date')\n",
    "jdata_user_action_data = dealDateColumns(jdata_user_action_data, 'a_date')\n",
    "jdata_user_comment_score_data = dealDateColumns(jdata_user_comment_score_data, 'comment_create_tm')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "792723\n",
      "550378\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lab-zhao.yinhu/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:45: FutureWarning: by argument to sort_index is deprecated, pls use .sort_values(by=...)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2017 4 53553\n",
      "2017 3 52574\n",
      "2017 2 47684\n",
      "2017 1 31030\n",
      "2016 12 28760\n",
      "2016 11 31341\n",
      "    user_id  age  sex  user_lv_cd  month  year  is_order  order_day\n",
      "0         1    3    2           3      4  2017         1       14.0\n",
      "2         2    3    2           3      4  2017         1       24.0\n",
      "3         3    3    0           5      4  2017         1       26.0\n",
      "4         4    3    2           3      4  2017         1       22.0\n",
      "5         5    4    2           2      4  2017         1       11.0\n",
      "6         6   -1    2           2      4  2017         1       12.0\n",
      "7         7    3    2           4      4  2017         1        7.0\n",
      "8         8    2    2           2      4  2017         0       -1.0\n",
      "9         9    3    2           3      4  2017         0       -1.0\n",
      "10       10    4    0           4      4  2017         0       -1.0\n"
     ]
    }
   ],
   "source": [
    "#首先过滤出目标类目订单记录用于训练集打标签\n",
    "jdata_user_order_data = pd.merge(jdata_user_order_data, jdata_sku_basic_info_data[['sku_id', 'cate']], on=['sku_id'], how='left')\n",
    "jdata_user_action_data = pd.merge(jdata_user_action_data, jdata_sku_basic_info_data[['sku_id', 'cate']], on=['sku_id'], how='left')\n",
    "jdata_user_order_data_target = jdata_user_order_data[(jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30)]\n",
    "print(len(jdata_user_order_data))\n",
    "print(len(jdata_user_order_data_target))\n",
    "\n",
    "\n",
    "# #首先尝试对每个月构造训练集，并为每条样本打上标签\n",
    "# def getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, month, year, end_time):\n",
    "#     user_set = set(jdata_user_order_data_target['user_id'][(jdata_user_order_data.o_date < end_time)])\n",
    "#     jdata_train_df_1 = pd.DataFrame(columns=['user_id', 'age', 'sex', 'user_lv_cd'])\n",
    "#     for user_id, age, sex, user_lv_cd in jdata_user_basic_info_data[['user_id', 'age', 'sex', 'user_lv_cd']].values:\n",
    "#         if user_id in user_set:\n",
    "#             insertRow = pd.DataFrame([[user_id, age, sex, user_lv_cd]], columns=['user_id', 'age', 'sex', 'user_lv_cd'])\n",
    "#             jdata_train_df_1 = jdata_train_df_1.append(insertRow)\n",
    "#     print(len(jdata_train_df_1))\n",
    "#     jdata_train_df_1['month'] = month\n",
    "#     jdata_train_df_1['year'] = year\n",
    "#     jdata_train_df_1 = pd.merge(jdata_train_df_1, jdata_user_order_data_target[['user_id', 'year', 'month', 'day', 'o_date']], on=['user_id', 'year', 'month'], how='left')\n",
    "#     jdata_train_df_1 = jdata_train_df_1.sort_index(by=['user_id', 'o_date'], ascending=True)\n",
    "#     jdata_train_df_1 = jdata_train_df_1.drop_duplicates(['user_id'])\n",
    "#     jdata_train_df_1['o_date'] = jdata_train_df_1['o_date'].fillna(pd.to_datetime('2000-01-01'))\n",
    "#     jdata_train_df_1['is_order'] = jdata_train_df_1['day'].map(lambda x: 0 if math.isnan(x) else 1)\n",
    "#     jdata_train_df_1['order_day'] = jdata_train_df_1['day'].map(lambda x: -1 if math.isnan(x) else x)\n",
    "#     del jdata_train_df_1['day']\n",
    "#     del jdata_train_df_1['o_date']\n",
    "#     print(year, month, len(jdata_train_df_1[jdata_train_df_1.is_order == 1]))\n",
    "# #     jdata_train_df = pd.concat([jdata_train_df, jdata_train_df_1])\n",
    "#     return jdata_train_df_1\n",
    "\n",
    "# jdata_train_df_201704 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 4, 2017, datetime.datetime(2017, 4, 1))\n",
    "# jdata_train_df_201703 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 3, 2017, datetime.datetime(2017, 3, 1))\n",
    "# jdata_train_df_201702 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 2, 2017, datetime.datetime(2017, 2, 1))\n",
    "# jdata_train_df_201701 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 1, 2017, datetime.datetime(2017, 1, 1))\n",
    "# jdata_train_df_201612 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 12, 2016, datetime.datetime(2016, 12, 1))\n",
    "# jdata_train_df_201611 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 11, 2016, datetime.datetime(2016, 11, 1))\n",
    "# print(jdata_train_df_201704.head(10))\n",
    "\n",
    "def getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, month, year):\n",
    "    jdata_train_df_1 = jdata_user_basic_info_data.copy()\n",
    "    jdata_train_df_1['month'] = month\n",
    "    jdata_train_df_1['year'] = year\n",
    "    jdata_train_df_1 = pd.merge(jdata_train_df_1, jdata_user_order_data_target[['user_id', 'year', 'month', 'day', 'o_date']], on=['user_id', 'year', 'month'], how='left')\n",
    "    jdata_train_df_1 = jdata_train_df_1.sort_index(by=['user_id', 'o_date'], ascending=True)\n",
    "    jdata_train_df_1 = jdata_train_df_1.drop_duplicates(['user_id'])\n",
    "    jdata_train_df_1['o_date'] = jdata_train_df_1['o_date'].fillna(pd.to_datetime('2000-01-01'))\n",
    "    jdata_train_df_1['is_order'] = jdata_train_df_1['day'].map(lambda x: 0 if math.isnan(x) else 1)\n",
    "    jdata_train_df_1['order_day'] = jdata_train_df_1['day'].map(lambda x: -1 if math.isnan(x) else x)\n",
    "    del jdata_train_df_1['day']\n",
    "    del jdata_train_df_1['o_date']\n",
    "    print(year, month, len(jdata_train_df_1[jdata_train_df_1.is_order == 1]))\n",
    "#     jdata_train_df = pd.concat([jdata_train_df, jdata_train_df_1])\n",
    "    return jdata_train_df_1\n",
    "\n",
    "jdata_train_df_201704 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 4, 2017)\n",
    "jdata_train_df_201703 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 3, 2017)\n",
    "jdata_train_df_201702 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 2, 2017)\n",
    "jdata_train_df_201701 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 1, 2017)\n",
    "jdata_train_df_201612 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 12, 2016)\n",
    "jdata_train_df_201611 = getTrainDfMonthly(jdata_user_order_data_target, jdata_user_basic_info_data, 11, 2016)\n",
    "print(jdata_train_df_201704.head(10))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "98924\n",
      "   user_id  age  sex  user_lv_cd  year  month\n",
      "0        1    3    2           3  2017      5\n",
      "1        2    3    2           3  2017      5\n",
      "2        3    3    0           5  2017      5\n",
      "3        4    3    2           3  2017      5\n",
      "4        5    4    2           2  2017      5\n",
      "5        6   -1    2           2  2017      5\n",
      "6        7    3    2           4  2017      5\n",
      "7        8    2    2           2  2017      5\n",
      "8        9    3    2           3  2017      5\n",
      "9       10    4    0           4  2017      5\n"
     ]
    }
   ],
   "source": [
    "#构造测试集\n",
    "jdata_test_df = jdata_user_basic_info_data.copy()\n",
    "jdata_test_df['year'] = 2017\n",
    "jdata_test_df['month'] = 5\n",
    "print(len(jdata_test_df))\n",
    "print(jdata_test_df.head(10))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 为训练集和测试集定义滑窗区间\n",
    "FeatureMonthList_201704 = [(datetime.datetime(2017, 3, 1), datetime.datetime(2017, 4, 1), 1), \n",
    "                          (datetime.datetime(2017, 3, 15), datetime.datetime(2017, 4, 1), 0.5),\n",
    "                          (datetime.datetime(2017, 2, 1), datetime.datetime(2017, 3, 1), 'one'),\n",
    "                          (datetime.datetime(2017, 1, 1), datetime.datetime(2017, 4, 1), 3),\n",
    "                          (datetime.datetime(2016, 10, 1), datetime.datetime(2017, 4, 1), 6)]\n",
    "\n",
    "FeatureMonthList_201703 = [(datetime.datetime(2017, 2, 1), datetime.datetime(2017, 3, 1), 1),\n",
    "                          (datetime.datetime(2017, 2, 15), datetime.datetime(2017, 3, 1), 0.5),\n",
    "                          (datetime.datetime(2017, 1, 1), datetime.datetime(2017, 2, 1), 'one'),\n",
    "                          (datetime.datetime(2016, 12, 1), datetime.datetime(2017, 3, 1), 3),\n",
    "                          (datetime.datetime(2016, 9, 1), datetime.datetime(2017, 3, 1), 6)]\n",
    "\n",
    "FeatureMonthList_201702 = [(datetime.datetime(2017, 1, 1), datetime.datetime(2017, 2, 1), 1), \n",
    "                          (datetime.datetime(2017, 1, 15), datetime.datetime(2017, 2, 1), 0.5),\n",
    "                          (datetime.datetime(2016, 12, 1), datetime.datetime(2017, 1, 1), 'one'),\n",
    "                          (datetime.datetime(2016, 11, 1), datetime.datetime(2017, 2, 1), 3),\n",
    "                          (datetime.datetime(2016, 8, 1), datetime.datetime(2017, 2, 1), 6)]\n",
    "\n",
    "FeatureMonthList_201701 = [(datetime.datetime(2016, 12, 1), datetime.datetime(2017, 1, 1), 1), \n",
    "                          (datetime.datetime(2016, 12, 15), datetime.datetime(2017, 1, 1), 0.5),\n",
    "                          (datetime.datetime(2016, 11, 1), datetime.datetime(2016, 12, 1), 'one'),\n",
    "                          (datetime.datetime(2016, 10, 1), datetime.datetime(2017, 1, 1), 3),\n",
    "                          (datetime.datetime(2016, 7, 1), datetime.datetime(2017, 1, 1), 6)]\n",
    "\n",
    "FeatureMonthList_201612 = [(datetime.datetime(2016, 11, 1), datetime.datetime(2016, 12, 1), 1), \n",
    "                           (datetime.datetime(2016, 11, 15), datetime.datetime(2016, 12, 1), 0.5),\n",
    "                          (datetime.datetime(2016, 10, 1), datetime.datetime(2016, 11, 1), 'one'),\n",
    "                          (datetime.datetime(2016, 9, 1), datetime.datetime(2016, 12, 1), 3),\n",
    "                          (datetime.datetime(2016, 6, 1), datetime.datetime(2016, 12, 1), 6)]\n",
    "\n",
    "FeatureMonthList_201611 = [(datetime.datetime(2016, 10, 1), datetime.datetime(2016, 11, 1), 1), \n",
    "                          (datetime.datetime(2016, 10, 15), datetime.datetime(2016, 11, 1), 0.5),\n",
    "                          (datetime.datetime(2016, 9, 1), datetime.datetime(2016, 10, 1), 'one'),\n",
    "                          (datetime.datetime(2016, 8, 1), datetime.datetime(2016, 11, 1), 3),\n",
    "                          (datetime.datetime(2016, 5, 1), datetime.datetime(2016, 11, 1), 6)]\n",
    "\n",
    "FeatureMonthList_201705 = [(datetime.datetime(2017, 4, 1), datetime.datetime(2017, 5, 1), 1), \n",
    "                          (datetime.datetime(2017, 4, 15), datetime.datetime(2017, 5, 1), 0.5),\n",
    "                          (datetime.datetime(2017, 3, 1), datetime.datetime(2017, 4, 1), 'one'),\n",
    "                          (datetime.datetime(2017, 2, 1), datetime.datetime(2017, 5, 1), 3),\n",
    "                          (datetime.datetime(2016, 11, 1), datetime.datetime(2017, 5, 1), 6)]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lab-zhao.yinhu/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:25: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n"
     ]
    }
   ],
   "source": [
    "#　添加一个冷启动特征，用户是否是上个月第一次购买特征\n",
    "def getUserIsLastMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate):\n",
    "    jdata_df_copy = jdata_df.copy()\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "            jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "            jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "        jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_one_pivot_table = pd.pivot_table(jdata_user_order_data_one, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_one_pivot_table.rename(columns={'o_id':'one_count'}, inplace=True)\n",
    "    jdata_user_order_data_one_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_pre_pivot_table = pd.pivot_table(jdata_user_order_data_pre, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_pre_pivot_table.rename(columns={'o_id':'pre_count'}, inplace=True)\n",
    "    jdata_user_order_data_pre_pivot_table.reset_index(inplace=True)\n",
    "    jdata_df_copy = pd.merge(jdata_df_copy, jdata_user_order_data_one_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_df_copy = pd.merge(jdata_df_copy, jdata_user_order_data_pre_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_df_copy['one_count'] = jdata_df_copy['one_count'].fillna(-1)\n",
    "    jdata_df_copy['pre_count'] = jdata_df_copy['pre_count'].fillna(-1)\n",
    "    jdata_df_copy['user_cate' + str(cate) + '_isLastMonthFirstBuy'] = 0\n",
    "    jdata_df_copy['user_cate' + str(cate) + '_isLastMonthFirstBuy'][(jdata_df_copy.one_count != -1) & (jdata_df_copy.pre_count == -1)] = 1\n",
    "    jdata_df = pd.merge(jdata_df, jdata_df_copy[['user_id', 'user_cate' + str(cate) + '_isLastMonthFirstBuy']], on=['user_id'], how='left')\n",
    "    return jdata_df\n",
    "\n",
    "def getUserIsLastMonthFirstBuy(jdata_df, jdata_user_order_data, start_time, end_time):\n",
    "    jdata_df = getUserIsLastMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, -1)\n",
    "    jdata_df = getUserIsLastMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, -2)\n",
    "    jdata_df = getUserIsLastMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, 101)\n",
    "    jdata_df = getUserIsLastMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, 30)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserIsLastMonthFirstBuy(jdata_train_df_201704, jdata_user_order_data, datetime.datetime(2017, 3, 1), datetime.datetime(2017, 4, 1))\n",
    "jdata_train_df_201703 = getUserIsLastMonthFirstBuy(jdata_train_df_201703, jdata_user_order_data, datetime.datetime(2017, 2, 1), datetime.datetime(2017, 3, 1))\n",
    "jdata_train_df_201702 = getUserIsLastMonthFirstBuy(jdata_train_df_201702, jdata_user_order_data, datetime.datetime(2017, 1, 1), datetime.datetime(2017, 2, 1))\n",
    "jdata_train_df_201701 = getUserIsLastMonthFirstBuy(jdata_train_df_201701, jdata_user_order_data, datetime.datetime(2016, 12, 1), datetime.datetime(2017, 1, 1))\n",
    "jdata_train_df_201612 = getUserIsLastMonthFirstBuy(jdata_train_df_201612, jdata_user_order_data, datetime.datetime(2016, 11, 1), datetime.datetime(2016, 12, 1))\n",
    "jdata_train_df_201611 = getUserIsLastMonthFirstBuy(jdata_train_df_201611, jdata_user_order_data, datetime.datetime(2016, 10, 1), datetime.datetime(2016, 11, 1))\n",
    "jdata_test_df = getUserIsLastMonthFirstBuy(jdata_test_df, jdata_user_order_data, datetime.datetime(2017, 4, 1), datetime.datetime(2017, 5, 1))\n",
    "\n",
    "\n",
    "# set1 = set(jdata_user_order_data['user_id'][(jdata_user_order_data.o_date < datetime.datetime(2017, 4, 1)) & (jdata_user_order_data.o_date >= datetime.datetime(2017, 3, 1)) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))])\n",
    "# set2 = set(jdata_user_order_data['user_id'][(jdata_user_order_data.o_date < datetime.datetime(2017, 3, 1)) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))])\n",
    "# print(len(set1))\n",
    "# print(len(set2))\n",
    "# print(len(set1 & set2))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lab-zhao.yinhu/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:25: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n"
     ]
    }
   ],
   "source": [
    "#　添加一个冷启动特征，用户是否是上2个月第一次购买特征\n",
    "def getUserIsLastTwoMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate):\n",
    "    jdata_df_copy = jdata_df.copy()\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "            jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "            jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "        jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_one_pivot_table = pd.pivot_table(jdata_user_order_data_one, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_one_pivot_table.rename(columns={'o_id':'one_count'}, inplace=True)\n",
    "    jdata_user_order_data_one_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_pre_pivot_table = pd.pivot_table(jdata_user_order_data_pre, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_pre_pivot_table.rename(columns={'o_id':'pre_count'}, inplace=True)\n",
    "    jdata_user_order_data_pre_pivot_table.reset_index(inplace=True)\n",
    "    jdata_df_copy = pd.merge(jdata_df_copy, jdata_user_order_data_one_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_df_copy = pd.merge(jdata_df_copy, jdata_user_order_data_pre_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_df_copy['one_count'] = jdata_df_copy['one_count'].fillna(-1)\n",
    "    jdata_df_copy['pre_count'] = jdata_df_copy['pre_count'].fillna(-1)\n",
    "    jdata_df_copy['user_cate' + str(cate) + '_isLastTwoMonthFirstBuy'] = 0\n",
    "    jdata_df_copy['user_cate' + str(cate) + '_isLastTwoMonthFirstBuy'][(jdata_df_copy.one_count != -1) & (jdata_df_copy.pre_count == -1)] = 1\n",
    "    jdata_df = pd.merge(jdata_df, jdata_df_copy[['user_id', 'user_cate' + str(cate) + '_isLastTwoMonthFirstBuy']], on=['user_id'], how='left')\n",
    "    return jdata_df\n",
    "\n",
    "def getUserIsLastTwoMonthFirstBuy(jdata_df, jdata_user_order_data, start_time, end_time):\n",
    "    jdata_df = getUserIsLastTwoMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, -1)\n",
    "    jdata_df = getUserIsLastTwoMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, -2)\n",
    "    jdata_df = getUserIsLastTwoMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, 101)\n",
    "    jdata_df = getUserIsLastTwoMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, 30)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserIsLastTwoMonthFirstBuy(jdata_train_df_201704, jdata_user_order_data, datetime.datetime(2017, 2, 1), datetime.datetime(2017, 3, 1))\n",
    "jdata_train_df_201703 = getUserIsLastTwoMonthFirstBuy(jdata_train_df_201703, jdata_user_order_data, datetime.datetime(2017, 1, 1), datetime.datetime(2017, 2, 1))\n",
    "jdata_train_df_201702 = getUserIsLastTwoMonthFirstBuy(jdata_train_df_201702, jdata_user_order_data, datetime.datetime(2016, 12, 1), datetime.datetime(2017, 1, 1))\n",
    "jdata_train_df_201701 = getUserIsLastTwoMonthFirstBuy(jdata_train_df_201701, jdata_user_order_data, datetime.datetime(2016, 11, 1), datetime.datetime(2016, 12, 1))\n",
    "jdata_train_df_201612 = getUserIsLastTwoMonthFirstBuy(jdata_train_df_201612, jdata_user_order_data, datetime.datetime(2016, 10, 1), datetime.datetime(2016, 11, 1))\n",
    "jdata_train_df_201611 = getUserIsLastTwoMonthFirstBuy(jdata_train_df_201611, jdata_user_order_data, datetime.datetime(2016, 9, 1), datetime.datetime(2016, 10, 1))\n",
    "jdata_test_df = getUserIsLastTwoMonthFirstBuy(jdata_test_df, jdata_user_order_data, datetime.datetime(2017, 3, 1), datetime.datetime(2017, 4, 1))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lab-zhao.yinhu/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:25: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n"
     ]
    }
   ],
   "source": [
    "#　添加一个冷启动特征，用户是否是上3个月第一次购买特征\n",
    "def getUserIsLastThreeMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate):\n",
    "    jdata_df_copy = jdata_df.copy()\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "            jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "            jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_one = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "        jdata_user_order_data_pre = jdata_user_order_data[(jdata_user_order_data.o_date < start_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_one_pivot_table = pd.pivot_table(jdata_user_order_data_one, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_one_pivot_table.rename(columns={'o_id':'one_count'}, inplace=True)\n",
    "    jdata_user_order_data_one_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_pre_pivot_table = pd.pivot_table(jdata_user_order_data_pre, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_pre_pivot_table.rename(columns={'o_id':'pre_count'}, inplace=True)\n",
    "    jdata_user_order_data_pre_pivot_table.reset_index(inplace=True)\n",
    "    jdata_df_copy = pd.merge(jdata_df_copy, jdata_user_order_data_one_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_df_copy = pd.merge(jdata_df_copy, jdata_user_order_data_pre_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_df_copy['one_count'] = jdata_df_copy['one_count'].fillna(-1)\n",
    "    jdata_df_copy['pre_count'] = jdata_df_copy['pre_count'].fillna(-1)\n",
    "    jdata_df_copy['user_cate' + str(cate) + '_isLastThreeMonthFirstBuy'] = 0\n",
    "    jdata_df_copy['user_cate' + str(cate) + '_isLastThreeMonthFirstBuy'][(jdata_df_copy.one_count != -1) & (jdata_df_copy.pre_count == -1)] = 1\n",
    "    jdata_df = pd.merge(jdata_df, jdata_df_copy[['user_id', 'user_cate' + str(cate) + '_isLastThreeMonthFirstBuy']], on=['user_id'], how='left')\n",
    "    return jdata_df\n",
    "\n",
    "def getUserIsLastThreeMonthFirstBuy(jdata_df, jdata_user_order_data, start_time, end_time):\n",
    "    jdata_df = getUserIsLastThreeMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, -1)\n",
    "    jdata_df = getUserIsLastThreeMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, -2)\n",
    "    jdata_df = getUserIsLastThreeMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, 101)\n",
    "    jdata_df = getUserIsLastThreeMonthFirstBuyMonthly(jdata_df, jdata_user_order_data, start_time, end_time, 30)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserIsLastThreeMonthFirstBuy(jdata_train_df_201704, jdata_user_order_data, datetime.datetime(2017, 1, 1), datetime.datetime(2017, 2, 1))\n",
    "jdata_train_df_201703 = getUserIsLastThreeMonthFirstBuy(jdata_train_df_201703, jdata_user_order_data, datetime.datetime(2016, 12, 1), datetime.datetime(2017, 1, 1))\n",
    "jdata_train_df_201702 = getUserIsLastThreeMonthFirstBuy(jdata_train_df_201702, jdata_user_order_data, datetime.datetime(2016, 11, 1), datetime.datetime(2016, 12, 1))\n",
    "jdata_train_df_201701 = getUserIsLastThreeMonthFirstBuy(jdata_train_df_201701, jdata_user_order_data, datetime.datetime(2016, 10, 1), datetime.datetime(2016, 11, 1))\n",
    "jdata_train_df_201612 = getUserIsLastThreeMonthFirstBuy(jdata_train_df_201612, jdata_user_order_data, datetime.datetime(2016, 9, 1), datetime.datetime(2016, 10, 1))\n",
    "jdata_train_df_201611 = getUserIsLastThreeMonthFirstBuy(jdata_train_df_201611, jdata_user_order_data, datetime.datetime(2016, 8, 1), datetime.datetime(2016, 9, 1))\n",
    "jdata_test_df = getUserIsLastThreeMonthFirstBuy(jdata_test_df, jdata_user_order_data, datetime.datetime(2017, 2, 1), datetime.datetime(2017, 3, 1))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lab-zhao.yinhu/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:11: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  # This is added back by InteractiveShellApp.init_path()\n"
     ]
    }
   ],
   "source": [
    "# 统计某用户在滑窗区间内下单某个类目商品的次数，逐个月进行处理\n",
    "def getUserOrderNumberMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_order_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_order_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_order_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_order_future.drop_duplicates(['user_id', 'o_id'], inplace=True)\n",
    "    jdata_user_order_data_order_future_pivot_table = pd.pivot_table(jdata_user_order_data_order_future, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_order_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_order_future_pivot_table.rename(columns={'o_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_order_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserOrderNumber(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserOrderNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_orderNumber')\n",
    "        jdata_df = getUserOrderNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_orderNumber')\n",
    "        jdata_df = getUserOrderNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_orderNumber')\n",
    "        jdata_df = getUserOrderNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_orderNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserOrderNumber(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserOrderNumber(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserOrderNumber(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserOrderNumber(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserOrderNumber(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserOrderNumber(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserOrderNumber(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n",
    "                                                                                                           "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内购买某个类目商品的次数，逐个月进行处理\n",
    "def getUserBuyNumberMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['sku_id'], aggfunc=len)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'sku_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBuyNumber(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBuyNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_buyNumber')\n",
    "        jdata_df = getUserBuyNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_buyNumber')\n",
    "        jdata_df = getUserBuyNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_buyNumber')\n",
    "        jdata_df = getUserBuyNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_buyNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyNumber(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBuyNumber(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBuyNumber(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBuyNumber(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBuyNumber(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBuyNumber(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBuyNumber(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lab-zhao.yinhu/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:11: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  # This is added back by InteractiveShellApp.init_path()\n"
     ]
    }
   ],
   "source": [
    "# 统计某用户在滑窗区间内购买某个类目商品的天数，逐个月进行处理\n",
    "def getUserBuyDayNumberMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_buy_future.drop_duplicates(['user_id', 'o_date'], inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'o_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBuyDayNumber(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBuyDayNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_buyDayNumber')\n",
    "        jdata_df = getUserBuyDayNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_buyDayNumber')\n",
    "        jdata_df = getUserBuyDayNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_buyDayNumber')\n",
    "        jdata_df = getUserBuyDayNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_buyDayNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyDayNumber(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBuyDayNumber(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBuyDayNumber(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBuyDayNumber(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBuyDayNumber(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBuyDayNumber(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBuyDayNumber(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内购买某个类目商品的件数，逐个月进行处理\n",
    "def getUserBuyCountMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['o_sku_num'], aggfunc=sum)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'o_sku_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBuyCount(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBuyCountMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_buyCount')\n",
    "        jdata_df = getUserBuyCountMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_buyCount')\n",
    "        jdata_df = getUserBuyCountMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_buyCount')\n",
    "        jdata_df = getUserBuyCountMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_buyCount')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyCount(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBuyCount(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBuyCount(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBuyCount(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBuyCount(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBuyCount(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBuyCount(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内购买某个类目商品在当月第几天的一些统计特征(min,max,mean)，逐个月进行处理\n",
    "def getUserBuyDayMinMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['day'], aggfunc=min)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'day':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBuyDayMin(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBuyDayMinMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_buyDayMin')\n",
    "        jdata_df = getUserBuyDayMinMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_buyDayMin')\n",
    "        jdata_df = getUserBuyDayMinMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_buyDayMin')\n",
    "        jdata_df = getUserBuyDayMinMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_buyDayMin')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyDayMin(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBuyDayMin(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBuyDayMin(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBuyDayMin(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBuyDayMin(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBuyDayMin(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBuyDayMin(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def getUserBuyDayMaxMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['day'], aggfunc=max)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'day':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBuyDayMax(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBuyDayMaxMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_buyDayMax')\n",
    "        jdata_df = getUserBuyDayMaxMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_buyDayMax')\n",
    "        jdata_df = getUserBuyDayMaxMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_buyDayMax')\n",
    "        jdata_df = getUserBuyDayMaxMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_buyDayMax')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyDayMax(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBuyDayMax(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBuyDayMax(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBuyDayMax(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBuyDayMax(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBuyDayMax(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBuyDayMax(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def getUserBuyDayMeanMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['day'], aggfunc=mean)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'day':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBuyDayMean(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBuyDayMeanMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_buyDayMean')\n",
    "        jdata_df = getUserBuyDayMeanMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_buyDayMean')\n",
    "        jdata_df = getUserBuyDayMeanMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_buyDayMean')\n",
    "        jdata_df = getUserBuyDayMeanMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_buyDayMean')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyDayMean(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBuyDayMean(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBuyDayMean(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBuyDayMean(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBuyDayMean(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBuyDayMean(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBuyDayMean(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内购买某个类目商品的月份数，逐个月进行处理\n",
    "def getUserBuyMonthNumberMonthly(jdata_df, jdata_user_order_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.o_date >= start_time) & (jdata_user_order_data.o_date < end_time) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_buy_future.drop_duplicates(['user_id', 'month'], inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['month'], aggfunc=len)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'month':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBuyMonthNumber(jdata_df, jdata_user_order_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBuyMonthNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_buyMonthNumber')\n",
    "        jdata_df = getUserBuyMonthNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_buyMonthNumber')\n",
    "        jdata_df = getUserBuyMonthNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_buyMonthNumber')\n",
    "        jdata_df = getUserBuyMonthNumberMonthly(jdata_df, jdata_user_order_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_buyMonthNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyMonthNumber(jdata_train_df_201704, jdata_user_order_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBuyMonthNumber(jdata_train_df_201703, jdata_user_order_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBuyMonthNumber(jdata_train_df_201702, jdata_user_order_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBuyMonthNumber(jdata_train_df_201701, jdata_user_order_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBuyMonthNumber(jdata_train_df_201612, jdata_user_order_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBuyMonthNumber(jdata_train_df_201611, jdata_user_order_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBuyMonthNumber(jdata_test_df, jdata_user_order_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的操作个数，逐个月进行处理\n",
    "def getUserSkuActionNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate)]\n",
    "    jdata_user_action_data_future.drop_duplicates(['user_id', 'sku_id'], inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['sku_id'], aggfunc=len)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'sku_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserSkuActionNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserSkuActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_skuActionNumber')\n",
    "        jdata_df = getUserSkuActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_skuActionNumber')\n",
    "        jdata_df = getUserSkuActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_skuActionNumber')\n",
    "        jdata_df = getUserSkuActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_skuActionNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserSkuActionNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserSkuActionNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserSkuActionNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserSkuActionNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserSkuActionNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserSkuActionNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserSkuActionNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的操作天数，逐个月进行处理\n",
    "def getUserSkuActionDayNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate)]\n",
    "    jdata_user_action_data_future.drop_duplicates(['user_id', 'a_date'], inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_date'], aggfunc=len)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_date':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserSkuActionDayNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserSkuActionDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_skuActionDayNumber')\n",
    "        jdata_df = getUserSkuActionDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_skuActionDayNumber')\n",
    "        jdata_df = getUserSkuActionDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_skuActionDayNumber')\n",
    "        jdata_df = getUserSkuActionDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_skuActionDayNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserSkuActionDayNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserSkuActionDayNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserSkuActionDayNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserSkuActionDayNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserSkuActionDayNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserSkuActionDayNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserSkuActionDayNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的操作总数，逐个月进行处理\n",
    "def getUserActionNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate)]\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_num'], aggfunc=sum)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserActionNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_actionNumber')\n",
    "        jdata_df = getUserActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_actionNumber')\n",
    "        jdata_df = getUserActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_actionNumber')\n",
    "        jdata_df = getUserActionNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_actionNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserActionNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserActionNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserActionNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserActionNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserActionNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserActionNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserActionNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的浏览个数，逐个月进行处理\n",
    "def getUserSkuBrowseNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate) & (jdata_user_action_data.a_type == 1)]\n",
    "    jdata_user_action_data_future.drop_duplicates(['user_id', 'sku_id'], inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['sku_id'], aggfunc=len)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'sku_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserSkuBrowseNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserSkuBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_skuBrowseNumber')\n",
    "        jdata_df = getUserSkuBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_skuBrowseNumber')\n",
    "        jdata_df = getUserSkuBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_skuBrowseNumber')\n",
    "        jdata_df = getUserSkuBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_skuBrowseNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserSkuBrowseNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserSkuBrowseNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserSkuBrowseNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserSkuBrowseNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserSkuBrowseNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserSkuBrowseNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserSkuBrowseNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的浏览天数，逐个月进行处理\n",
    "def getUserSkuBrowseDayNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate) & (jdata_user_action_data.a_type == 1)]\n",
    "    jdata_user_action_data_future.drop_duplicates(['user_id', 'a_date'], inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_date'], aggfunc=len)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_date':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserSkuBrowseDayNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserSkuBrowseDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_skuBrowseDayNumber')\n",
    "        jdata_df = getUserSkuBrowseDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_skuBrowseDayNumber')\n",
    "        jdata_df = getUserSkuBrowseDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_skuBrowseDayNumber')\n",
    "        jdata_df = getUserSkuBrowseDayNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_skuBrowseDayNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserSkuBrowseDayNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserSkuBrowseDayNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserSkuBrowseDayNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserSkuBrowseDayNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserSkuBrowseDayNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserSkuBrowseDayNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserSkuBrowseDayNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的浏览总数，逐个月进行处理\n",
    "def getUserBrowseNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate) & (jdata_user_action_data.a_type == 1)]\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_num'], aggfunc=sum)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserBrowseNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_browseNumber')\n",
    "        jdata_df = getUserBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_browseNumber')\n",
    "        jdata_df = getUserBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_browseNumber')\n",
    "        jdata_df = getUserBrowseNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_browseNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBrowseNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserBrowseNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserBrowseNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserBrowseNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserBrowseNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserBrowseNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserBrowseNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的关注个数，逐个月进行处理\n",
    "def getUserSkuFocusNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30)) & (jdata_user_action_data.a_type == 2)]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30)) & (jdata_user_action_data.a_type == 2)]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate) & (jdata_user_action_data.a_type == 2)]\n",
    "    jdata_user_action_data_future.drop_duplicates(['user_id', 'sku_id'], inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['sku_id'], aggfunc=len)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'sku_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserSkuFocusNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserSkuFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_skuFocusNumber')\n",
    "        jdata_df = getUserSkuFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_skuFocusNumber')\n",
    "        jdata_df = getUserSkuFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_skuFocusNumber')\n",
    "        jdata_df = getUserSkuFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_skuFocusNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserSkuFocusNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserSkuFocusNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserSkuFocusNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserSkuFocusNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserSkuFocusNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserSkuFocusNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserSkuFocusNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内对某个类目商品的关注总数，逐个月进行处理\n",
    "def getUserFocusNumberMonthly(jdata_df, jdata_user_action_data, start_time, end_time, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30)) & (jdata_user_action_data.a_type == 2)]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30)) & (jdata_user_action_data.a_type == 2)]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date >= start_time) & (jdata_user_action_data.a_date < end_time) & (jdata_user_action_data.cate == cate) & (jdata_user_action_data.a_type == 2)]\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_num'], aggfunc=sum)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserFocusNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 101, 'user_cate101_last' + str(month) + 'Month_focusNumber')\n",
    "        jdata_df = getUserFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, 30, 'user_cate30_last' + str(month) + 'Month_focusNumber')\n",
    "        jdata_df = getUserFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -1, 'user_targetCate_last' + str(month) + 'Month_focusNumber')\n",
    "        jdata_df = getUserFocusNumberMonthly(jdata_df, jdata_user_action_data, featureMonthBegin, featureMonthEnd, -2, 'user_relatedCate_last' + str(month) + 'Month_focusNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserFocusNumber(jdata_train_df_201704, jdata_user_action_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserFocusNumber(jdata_train_df_201703, jdata_user_action_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserFocusNumber(jdata_train_df_201702, jdata_user_action_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserFocusNumber(jdata_train_df_201701, jdata_user_action_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserFocusNumber(jdata_train_df_201612, jdata_user_action_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserFocusNumber(jdata_train_df_201611, jdata_user_action_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserFocusNumber(jdata_test_df, jdata_user_action_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户在滑窗区间内的评价次数，逐个月进行处理\n",
    "def getUserCommentNumberMonthly(jdata_df, jdata_user_comment_score_data, start_time, end_time, score_level, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if score_level == 0:\n",
    "        jdata_user_comment_score_data_future = jdata_user_comment_score_data[(jdata_user_comment_score_data.comment_create_tm >= start_time) & (jdata_user_comment_score_data.comment_create_tm < end_time)]\n",
    "    else:\n",
    "        jdata_user_comment_score_data_future = jdata_user_comment_score_data[(jdata_user_comment_score_data.comment_create_tm >= start_time) & (jdata_user_comment_score_data.comment_create_tm < end_time) & (jdata_user_comment_score_data.score_level == score_level)]\n",
    "    jdata_user_comment_score_data_future_pivot_table = pd.pivot_table(jdata_user_comment_score_data_future, index=['user_id'], values=['score_level'], aggfunc=len)\n",
    "    jdata_user_comment_score_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_comment_score_data_future_pivot_table.rename(columns={'score_level':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_comment_score_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserCommentNumber(jdata_df, jdata_user_action_data, FeatureMonthList):\n",
    "    for featureMonthBegin, featureMonthEnd, month in FeatureMonthList:\n",
    "        jdata_df = getUserCommentNumberMonthly(jdata_df, jdata_user_comment_score_data, featureMonthBegin, featureMonthEnd, 1, 'user_score1_last' + str(month) + 'Month_commentNumber')\n",
    "        jdata_df = getUserCommentNumberMonthly(jdata_df, jdata_user_comment_score_data, featureMonthBegin, featureMonthEnd, 2, 'user_score2_last' + str(month) + 'Month_commentNumber')\n",
    "        jdata_df = getUserCommentNumberMonthly(jdata_df, jdata_user_comment_score_data, featureMonthBegin, featureMonthEnd, 3, 'user_score3_last' + str(month) + 'Month_commentNumber')\n",
    "        jdata_df = getUserCommentNumberMonthly(jdata_df, jdata_user_comment_score_data, featureMonthBegin, featureMonthEnd, 0, 'user_score0_last' + str(month) + 'Month_commentNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserCommentNumber(jdata_train_df_201704, jdata_user_comment_score_data, FeatureMonthList_201704)\n",
    "jdata_train_df_201703 = getUserCommentNumber(jdata_train_df_201703, jdata_user_comment_score_data, FeatureMonthList_201703)\n",
    "jdata_train_df_201702 = getUserCommentNumber(jdata_train_df_201702, jdata_user_comment_score_data, FeatureMonthList_201702)\n",
    "jdata_train_df_201701 = getUserCommentNumber(jdata_train_df_201701, jdata_user_comment_score_data, FeatureMonthList_201701)\n",
    "jdata_train_df_201612 = getUserCommentNumber(jdata_train_df_201612, jdata_user_comment_score_data, FeatureMonthList_201612)\n",
    "jdata_train_df_201611 = getUserCommentNumber(jdata_train_df_201611, jdata_user_comment_score_data, FeatureMonthList_201611)\n",
    "jdata_test_df = getUserCommentNumber(jdata_test_df, jdata_user_comment_score_data, FeatureMonthList_201705)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户距离上次购买商品的时间，逐个月进行处理\n",
    "def getUserLastBuyDeltaTimeMonthly(jdata_df, jdata_user_order_data, year, month, cate, newColName):\n",
    "    time_boundary = datetime.datetime(year, month, 1)\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_future = jdata_user_order_data[['user_id', 'o_date']][(jdata_user_order_data.o_date < time_boundary) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_future = jdata_user_order_data[['user_id', 'o_date']][(jdata_user_order_data.o_date < time_boundary) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_future = jdata_user_order_data[['user_id', 'o_date']][(jdata_user_order_data.o_date < time_boundary) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_future = jdata_user_order_data_future.sort_index(by=['user_id', 'o_date'], ascending=False)\n",
    "    jdata_user_order_data_future = jdata_user_order_data_future.drop_duplicates(['user_id'])\n",
    "    jdata_user_order_data_future.rename(columns={'o_date':newColName}, inplace=True)\n",
    "    jdata_df = pd.merge(jdata_df, jdata_user_order_data_future, on=['user_id'], how='left')\n",
    "    jdata_df[newColName] = jdata_df[newColName].map(lambda x: 99999999 if pd.isnull(x) else (time_boundary - x).days)\n",
    "    return jdata_df\n",
    "\n",
    "def getUserLastBuyDeltaTime(jdata_df, jdata_user_order_data, year, month):\n",
    "    jdata_df = getUserLastBuyDeltaTimeMonthly(jdata_df, jdata_user_order_data, year, month, -1, 'user_targetCate_lastBuyDeltaTime')\n",
    "    jdata_df = getUserLastBuyDeltaTimeMonthly(jdata_df, jdata_user_order_data, year, month, -2, 'user_relatedCate_lastBuyDeltaTime')\n",
    "    jdata_df = getUserLastBuyDeltaTimeMonthly(jdata_df, jdata_user_order_data, year, month, 101, 'user_cate101_lastBuyDeltaTime')\n",
    "    jdata_df = getUserLastBuyDeltaTimeMonthly(jdata_df, jdata_user_order_data, year, month, 30, 'user_cate30_lastBuyDeltaTime')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201611 = getUserLastBuyDeltaTime(jdata_train_df_201611, jdata_user_order_data, 2016, 11)\n",
    "jdata_train_df_201612 = getUserLastBuyDeltaTime(jdata_train_df_201612, jdata_user_order_data, 2016, 12)\n",
    "jdata_train_df_201701 = getUserLastBuyDeltaTime(jdata_train_df_201701, jdata_user_order_data, 2017, 1)\n",
    "jdata_train_df_201702 = getUserLastBuyDeltaTime(jdata_train_df_201702, jdata_user_order_data, 2017, 2)\n",
    "jdata_train_df_201703 = getUserLastBuyDeltaTime(jdata_train_df_201703, jdata_user_order_data, 2017, 3)\n",
    "jdata_train_df_201704 = getUserLastBuyDeltaTime(jdata_train_df_201704, jdata_user_order_data, 2017, 4)\n",
    "jdata_test_df = getUserLastBuyDeltaTime(jdata_test_df, jdata_user_order_data, 2017, 5)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户上次购买商品的数量，逐个月进行处理\n",
    "def getUserLastBuyNumberMonthly(jdata_df, jdata_user_order_data, year, month, cate, newColName):\n",
    "    time_boundary = datetime.datetime(year, month, 1)\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_future = jdata_user_order_data[['user_id', 'o_date', 'o_sku_num']][(jdata_user_order_data.o_date < time_boundary) & ((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_order_data_future = jdata_user_order_data[['user_id', 'o_date', 'o_sku_num']][(jdata_user_order_data.o_date < time_boundary) & ((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_order_data_future = jdata_user_order_data[['user_id', 'o_date', 'o_sku_num']][(jdata_user_order_data.o_date < time_boundary) & (jdata_user_order_data.cate == cate)]\n",
    "    jdata_user_order_data_future = jdata_user_order_data_future.sort_index(by=['user_id', 'o_date'], ascending=False)\n",
    "    jdata_user_order_data_future = jdata_user_order_data_future.drop_duplicates(['user_id'])\n",
    "    jdata_user_order_data_future.rename(columns={'o_sku_num':newColName}, inplace=True)\n",
    "    jdata_df = pd.merge(jdata_df, jdata_user_order_data_future[['user_id', newColName]], on=['user_id'], how='left')\n",
    "    jdata_df[newColName] = jdata_df[newColName].fillna(0)\n",
    "    return jdata_df\n",
    "\n",
    "def getUserLastBuyNumber(jdata_df, jdata_user_order_data, year, month):\n",
    "    jdata_df = getUserLastBuyNumberMonthly(jdata_df, jdata_user_order_data, year, month, -1, 'user_targetCate_lastBuyNumber')\n",
    "    jdata_df = getUserLastBuyNumberMonthly(jdata_df, jdata_user_order_data, year, month, -2, 'user_relatedCate_lastBuyNumber')\n",
    "    jdata_df = getUserLastBuyNumberMonthly(jdata_df, jdata_user_order_data, year, month, 101, 'user_cate101_lastBuyNumber')\n",
    "    jdata_df = getUserLastBuyNumberMonthly(jdata_df, jdata_user_order_data, year, month, 30, 'user_cate30_lastBuyNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201611 = getUserLastBuyNumber(jdata_train_df_201611, jdata_user_order_data, 2016, 11)\n",
    "jdata_train_df_201612 = getUserLastBuyNumber(jdata_train_df_201612, jdata_user_order_data, 2016, 12)\n",
    "jdata_train_df_201701 = getUserLastBuyNumber(jdata_train_df_201701, jdata_user_order_data, 2017, 1)\n",
    "jdata_train_df_201702 = getUserLastBuyNumber(jdata_train_df_201702, jdata_user_order_data, 2017, 2)\n",
    "jdata_train_df_201703 = getUserLastBuyNumber(jdata_train_df_201703, jdata_user_order_data, 2017, 3)\n",
    "jdata_train_df_201704 = getUserLastBuyNumber(jdata_train_df_201704, jdata_user_order_data, 2017, 4)\n",
    "jdata_test_df = getUserLastBuyNumber(jdata_test_df, jdata_user_order_data, 2017, 5)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计用户上次购买时间间隔/数量的特征\n",
    "def getUserLastDeltatimeNumber(jdata_df):\n",
    "    jdata_df_copy = jdata_df.copy()\n",
    "    jdata_df_copy1 = jdata_df_copy['user_targetCate_lastBuyNumber'].map(lambda x: 1 if x==0 else x)\n",
    "    jdata_df['user_targetCate_lastDeltatimeNumber'] = jdata_df['user_targetCate_lastBuyDeltaTime'] / jdata_df_copy1\n",
    "    jdata_df_copy2 = jdata_df_copy['user_relatedCate_lastBuyNumber'].map(lambda x: 1 if x==0 else x)\n",
    "    jdata_df['user_relatedCate_lastDeltatimeNumber'] = jdata_df['user_relatedCate_lastBuyDeltaTime'] / jdata_df_copy2\n",
    "    jdata_df_copy3 = jdata_df_copy['user_cate101_lastBuyNumber'].map(lambda x: 1 if x==0 else x)\n",
    "    jdata_df['user_cate101_lastDeltatimeNumber'] = jdata_df['user_cate101_lastBuyDeltaTime'] / jdata_df_copy3\n",
    "    jdata_df_copy4 = jdata_df_copy['user_cate30_lastBuyNumber'].map(lambda x: 1 if x==0 else x)\n",
    "    jdata_df['user_cate30_lastDeltatimeNumber'] = jdata_df['user_cate30_lastBuyDeltaTime'] / jdata_df_copy4\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201611 = getUserLastDeltatimeNumber(jdata_train_df_201611)\n",
    "jdata_train_df_201612 = getUserLastDeltatimeNumber(jdata_train_df_201612)\n",
    "jdata_train_df_201701 = getUserLastDeltatimeNumber(jdata_train_df_201701)\n",
    "jdata_train_df_201702 = getUserLastDeltatimeNumber(jdata_train_df_201702)\n",
    "jdata_train_df_201703 = getUserLastDeltatimeNumber(jdata_train_df_201703)\n",
    "jdata_train_df_201704 = getUserLastDeltatimeNumber(jdata_train_df_201704)\n",
    "jdata_test_df = getUserLastDeltatimeNumber(jdata_test_df)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月购买某个类目商品的次数，逐个月进行处理\n",
    "def getUserMonthlyBuyNumberMonthly(jdata_df, jdata_user_order_data, cate, newColName, time_boundary, length):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.cate == cate) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['sku_id'], aggfunc=len)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'sku_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyBuyNumber(jdata_df, jdata_user_order_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyBuyNumberMonthly(jdata_df, jdata_user_order_data, 101, 'user_101Cate_monthlyBuyNumber', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyNumberMonthly(jdata_df, jdata_user_order_data, 30, 'user_30Cate_monthlyBuyNumber', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyNumberMonthly(jdata_df, jdata_user_order_data, -1, 'user_targetCate_monthlyBuyNumber', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyNumberMonthly(jdata_df, jdata_user_order_data, -2, 'user_relatedCate_monthlyBuyNumber', time_boundary, length)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyBuyNumber(jdata_train_df_201704, jdata_user_order_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyBuyNumber(jdata_train_df_201703, jdata_user_order_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyBuyNumber(jdata_train_df_201702, jdata_user_order_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyBuyNumber(jdata_train_df_201701, jdata_user_order_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyBuyNumber(jdata_train_df_201612, jdata_user_order_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyBuyNumber(jdata_train_df_201611, jdata_user_order_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyBuyNumber(jdata_test_df, jdata_user_order_data, datetime.datetime(2017, 5, 1), 12)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月购买某个类目商品的下单次数，逐个月进行处理\n",
    "def getUserMonthlyOrderNumberMonthly(jdata_df, jdata_user_order_data, cate, newColName, time_boundary, length):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.cate == cate) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    jdata_user_order_data_buy_future.drop_duplicates(['user_id', 'o_id'], inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['o_id'], aggfunc=len)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'o_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyOrderNumber(jdata_df, jdata_user_order_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyOrderNumberMonthly(jdata_df, jdata_user_order_data, 101, 'user_101Cate_monthlyOrderNumber', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyOrderNumberMonthly(jdata_df, jdata_user_order_data, 30, 'user_30Cate_monthlyOrderNumber', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyOrderNumberMonthly(jdata_df, jdata_user_order_data, -1, 'user_targetCate_monthlyOrderNumber', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyOrderNumberMonthly(jdata_df, jdata_user_order_data, -2, 'user_relatedCate_monthlyOrderNumber', time_boundary, length)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyOrderNumber(jdata_train_df_201704, jdata_user_order_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyOrderNumber(jdata_train_df_201703, jdata_user_order_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyOrderNumber(jdata_train_df_201702, jdata_user_order_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyOrderNumber(jdata_train_df_201701, jdata_user_order_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyOrderNumber(jdata_train_df_201612, jdata_user_order_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyOrderNumber(jdata_train_df_201611, jdata_user_order_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyOrderNumber(jdata_test_df, jdata_user_order_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月购买某个类目商品的件数，逐个月进行处理\n",
    "def getUserMonthlyBuyCountMonthly(jdata_df, jdata_user_order_data, cate, newColName, time_boundary, length):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.cate == cate) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['o_sku_num'], aggfunc=sum)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'o_sku_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyBuyCount(jdata_df, jdata_user_order_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyBuyCountMonthly(jdata_df, jdata_user_order_data, 101, 'user_101Cate_monthlyBuyCount', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyCountMonthly(jdata_df, jdata_user_order_data, 30, 'user_30Cate_monthlyBuyCount', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyCountMonthly(jdata_df, jdata_user_order_data, -1, 'user_targetCate_monthlyBuyCount', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyCountMonthly(jdata_df, jdata_user_order_data, -2, 'user_relatedCate_monthlyBuyCount', time_boundary, length)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyBuyCount(jdata_train_df_201704, jdata_user_order_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyBuyCount(jdata_train_df_201703, jdata_user_order_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyBuyCount(jdata_train_df_201702, jdata_user_order_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyBuyCount(jdata_train_df_201701, jdata_user_order_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyBuyCount(jdata_train_df_201612, jdata_user_order_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyBuyCount(jdata_train_df_201611, jdata_user_order_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyBuyCount(jdata_test_df, jdata_user_order_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月购买某个类目商品的天数，逐个月进行处理\n",
    "def getUserMonthlyBuyDayMonthly(jdata_df, jdata_user_order_data, cate, newColName, time_boundary, length):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "        else:\n",
    "            jdata_user_order_data_buy_future = jdata_user_order_data[((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30)) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    else:\n",
    "        jdata_user_order_data_buy_future = jdata_user_order_data[(jdata_user_order_data.cate == cate) & (jdata_user_order_data.o_date < time_boundary)]\n",
    "    jdata_user_order_data_buy_future.drop_duplicates(['user_id', 'o_date'], inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table = pd.pivot_table(jdata_user_order_data_buy_future, index=['user_id'], values=['o_date'], aggfunc=len)\n",
    "    jdata_user_order_data_buy_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_buy_future_pivot_table.rename(columns={'o_date':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_order_data_buy_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyBuyDay(jdata_df, jdata_user_order_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyBuyDayMonthly(jdata_df, jdata_user_order_data, 101, 'user_101Cate_monthlyBuyDay', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyDayMonthly(jdata_df, jdata_user_order_data, 30, 'user_30Cate_monthlyBuyDay', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyDayMonthly(jdata_df, jdata_user_order_data, -1, 'user_targetCate_monthlyBuyDay', time_boundary, length)\n",
    "    jdata_df = getUserMonthlyBuyDayMonthly(jdata_df, jdata_user_order_data, -2, 'user_relatedCate_monthlyBuyDay', time_boundary, length)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyBuyDay(jdata_train_df_201704, jdata_user_order_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyBuyDay(jdata_train_df_201703, jdata_user_order_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyBuyDay(jdata_train_df_201702, jdata_user_order_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyBuyDay(jdata_train_df_201701, jdata_user_order_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyBuyDay(jdata_train_df_201612, jdata_user_order_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyBuyDay(jdata_train_df_201611, jdata_user_order_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyBuyDay(jdata_test_df, jdata_user_order_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月对某个类目商品的关注总数，逐个月进行处理\n",
    "def getUserMonthlyFocusNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30)) & (jdata_user_action_data.a_type == 2)]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30)) & (jdata_user_action_data.a_type == 2)]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & (jdata_user_action_data.cate == cate) & (jdata_user_action_data.a_type == 2)]\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_num'], aggfunc=sum)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyFocusNumber(jdata_df, jdata_user_action_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyFocusNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 101, 'user_cate101_monthlyFocusNumber')\n",
    "    jdata_df = getUserMonthlyFocusNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 30, 'user_cate30_monthlyFocusNumber')\n",
    "    jdata_df = getUserMonthlyFocusNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -1, 'user_targetCate_monthlyFocusNumber')\n",
    "    jdata_df = getUserMonthlyFocusNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -2, 'user_relatedCate_monthlyFocusNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyFocusNumber(jdata_train_df_201704, jdata_user_action_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyFocusNumber(jdata_train_df_201703, jdata_user_action_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyFocusNumber(jdata_train_df_201702, jdata_user_action_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyFocusNumber(jdata_train_df_201701, jdata_user_action_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyFocusNumber(jdata_train_df_201612, jdata_user_action_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyFocusNumber(jdata_train_df_201611, jdata_user_action_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyFocusNumber(jdata_test_df, jdata_user_action_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月对某个类目商品的浏览总数，逐个月进行处理\n",
    "def getUserMonthlyBrowseNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30)) & (jdata_user_action_data.a_type == 1)]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & (jdata_user_action_data.cate == cate) & (jdata_user_action_data.a_type == 1)]\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_num'], aggfunc=sum)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyBrowseNumber(jdata_df, jdata_user_action_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyBrowseNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 101, 'user_cate101_monthlyBrowseNumber')\n",
    "    jdata_df = getUserMonthlyBrowseNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 30, 'user_cate30_monthlyBrowseNumber')\n",
    "    jdata_df = getUserMonthlyBrowseNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -1, 'user_targetCate_monthlyBrowseNumber')\n",
    "    jdata_df = getUserMonthlyBrowseNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -2, 'user_relatedCate_monthlyBrowseNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyBrowseNumber(jdata_train_df_201704, jdata_user_action_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyBrowseNumber(jdata_train_df_201703, jdata_user_action_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyBrowseNumber(jdata_train_df_201702, jdata_user_action_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyBrowseNumber(jdata_train_df_201701, jdata_user_action_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyBrowseNumber(jdata_train_df_201612, jdata_user_action_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyBrowseNumber(jdata_train_df_201611, jdata_user_action_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyBrowseNumber(jdata_test_df, jdata_user_action_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月对某个类目商品的操作总数，逐个月进行处理\n",
    "def getUserMonthlyActionNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & (jdata_user_action_data.cate == cate)]\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_num'], aggfunc=sum)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_num':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyActionNumber(jdata_df, jdata_user_action_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyActionNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 101, 'user_cate101_monthlyActionNumber')\n",
    "    jdata_df = getUserMonthlyActionNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 30, 'user_cate30_monthlyActionNumber')\n",
    "    jdata_df = getUserMonthlyActionNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -1, 'user_targetCate_monthlyActionNumber')\n",
    "    jdata_df = getUserMonthlyActionNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -2, 'user_relatedCate_monthlyActionNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyActionNumber(jdata_train_df_201704, jdata_user_action_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyActionNumber(jdata_train_df_201703, jdata_user_action_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyActionNumber(jdata_train_df_201702, jdata_user_action_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyActionNumber(jdata_train_df_201701, jdata_user_action_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyActionNumber(jdata_train_df_201612, jdata_user_action_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyActionNumber(jdata_train_df_201611, jdata_user_action_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyActionNumber(jdata_test_df, jdata_user_action_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月对某个类目商品的操作天数，逐个月进行处理\n",
    "def getUserMonthlyActionDayNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & (jdata_user_action_data.cate == cate)]\n",
    "    jdata_user_action_data_future.drop_duplicates(['user_id', 'a_date'], inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['a_date'], aggfunc=len)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'a_date':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyActionDayNumber(jdata_df, jdata_user_action_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyActionDayNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 101, 'user_cate101_monthlyActionDayNumber')\n",
    "    jdata_df = getUserMonthlyActionDayNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 30, 'user_cate30_monthlyActionDayNumber')\n",
    "    jdata_df = getUserMonthlyActionDayNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -1, 'user_targetCate_monthlyActionDayNumber')\n",
    "    jdata_df = getUserMonthlyActionDayNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -2, 'user_relatedCate_monthlyActionDayNumber')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyActionDayNumber(jdata_train_df_201704, jdata_user_action_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyActionDayNumber(jdata_train_df_201703, jdata_user_action_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyActionDayNumber(jdata_train_df_201702, jdata_user_action_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyActionDayNumber(jdata_train_df_201701, jdata_user_action_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyActionDayNumber(jdata_train_df_201612, jdata_user_action_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyActionDayNumber(jdata_train_df_201611, jdata_user_action_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyActionDayNumber(jdata_test_df, jdata_user_action_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计某用户平均每个月对某个类目商品的操作商品数，逐个月进行处理\n",
    "def getUserMonthlyActionSkuNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, cate, newColName):\n",
    "    jdata_train_df_dealMonth = jdata_df\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate == 101) | (jdata_user_action_data.cate == 30))]\n",
    "        else:\n",
    "            jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & ((jdata_user_action_data.cate != 101) & (jdata_user_action_data.cate != 30))]\n",
    "    else:\n",
    "        jdata_user_action_data_future = jdata_user_action_data[(jdata_user_action_data.a_date < time_boundary) & (jdata_user_action_data.cate == cate)]\n",
    "    jdata_user_action_data_future.drop_duplicates(['user_id', 'sku_id'], inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table = pd.pivot_table(jdata_user_action_data_future, index=['user_id'], values=['sku_id'], aggfunc=len)\n",
    "    jdata_user_action_data_future_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_action_data_future_pivot_table.rename(columns={'sku_id':newColName}, inplace=True)\n",
    "    jdata_train_df_dealMonth = pd.merge(jdata_train_df_dealMonth, jdata_user_action_data_future_pivot_table, on=['user_id'], how='left')\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName].fillna(0)\n",
    "    jdata_train_df_dealMonth[newColName] = jdata_train_df_dealMonth[newColName] / length\n",
    "    return jdata_train_df_dealMonth\n",
    "\n",
    "def getUserMonthlyActionSkuNumber(jdata_df, jdata_user_action_data, time_boundary, length):\n",
    "    jdata_df = getUserMonthlyActionSkuNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 101, 'user_cate101_monthlyActionSkuCount')\n",
    "    jdata_df = getUserMonthlyActionSkuNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, 30, 'user_cate30_monthlyActionSkuCount')\n",
    "    jdata_df = getUserMonthlyActionSkuNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -1, 'user_targetCate_monthlyActionSkuCount')\n",
    "    jdata_df = getUserMonthlyActionSkuNumberMonthly(jdata_df, jdata_user_action_data, time_boundary, length, -2, 'user_relatedCate_monthlyActionSkuCount')\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserMonthlyActionSkuNumber(jdata_train_df_201704, jdata_user_action_data, datetime.datetime(2017, 4, 1), 11)\n",
    "jdata_train_df_201703 = getUserMonthlyActionSkuNumber(jdata_train_df_201703, jdata_user_action_data, datetime.datetime(2017, 3, 1), 10)\n",
    "jdata_train_df_201702 = getUserMonthlyActionSkuNumber(jdata_train_df_201702, jdata_user_action_data, datetime.datetime(2017, 2, 1), 9)\n",
    "jdata_train_df_201701 = getUserMonthlyActionSkuNumber(jdata_train_df_201701, jdata_user_action_data, datetime.datetime(2017, 1, 1), 8)\n",
    "jdata_train_df_201612 = getUserMonthlyActionSkuNumber(jdata_train_df_201612, jdata_user_action_data, datetime.datetime(2016, 12, 1), 7)\n",
    "jdata_train_df_201611 = getUserMonthlyActionSkuNumber(jdata_train_df_201611, jdata_user_action_data, datetime.datetime(2016, 11, 1), 6)\n",
    "jdata_test_df = getUserMonthlyActionSkuNumber(jdata_test_df, jdata_user_action_data, datetime.datetime(2017, 5, 1), 12)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 统计用户平均购买时间间隔特征\n",
    "def getUserBuyDeltaTimeFeatureMonthly(jdata_df, jdata_user_order_data, cate):\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            jdata_user_order_data_copy = jdata_user_order_data[((jdata_user_order_data.cate == 101) | (jdata_user_order_data.cate == 30))].copy()\n",
    "        else:\n",
    "            jdata_user_order_data_copy = jdata_user_order_data[((jdata_user_order_data.cate != 101) & (jdata_user_order_data.cate != 30))].copy()\n",
    "    else:\n",
    "        jdata_user_order_data_copy = jdata_user_order_data[(jdata_user_order_data.cate == cate)].copy()\n",
    "    jdata_user_order_data_copy.drop_duplicates(['user_id', 'o_date'], inplace=True)\n",
    "    jdata_user_order_data_copy = jdata_user_order_data_copy.sort_index(by=['user_id', 'o_date'], ascending=True)\n",
    "    jdata_user_order_data_copy['last_user_id'] = jdata_user_order_data_copy['user_id'].shift(1)\n",
    "    jdata_user_order_data_copy['last_user_id'] = jdata_user_order_data_copy['user_id'] == jdata_user_order_data_copy['last_user_id']\n",
    "    jdata_user_order_data_copy['last_o_date'] = jdata_user_order_data_copy['o_date'].shift(1)\n",
    "    jdata_user_order_data_copy_ave = jdata_user_order_data_copy[jdata_user_order_data_copy.last_user_id == True]\n",
    "    jdata_user_order_data_copy_ave['user_deltaTimeAve'] = jdata_user_order_data_copy_ave['o_date'] - jdata_user_order_data_copy_ave['last_o_date']\n",
    "    jdata_user_order_data_copy_ave['user_deltaTimeAve'] = jdata_user_order_data_copy_ave['user_deltaTimeAve'].map(lambda x: x.days)\n",
    "    jdata_user_order_data_copy_ave_pivot_table = pd.pivot_table(jdata_user_order_data_copy_ave, index=['user_id'], values=['user_deltaTimeAve'], aggfunc={min, max, mean})\n",
    "    jdata_user_order_data_copy_ave_pivot_table.reset_index(inplace=True)\n",
    "    jdata_user_order_data_copy_ave_pivot_table.columns = ['user_id', 'user_cate'+ str(cate) + '_deltaTime_max', 'user_cate' + str(cate) + '_deltaTime_mean', 'user_cate' + str(cate) + '_deltaTime_min']\n",
    "    jdata_df = pd.merge(jdata_df, jdata_user_order_data_copy_ave_pivot_table, on=['user_id'], how='left')\n",
    "    return jdata_df\n",
    "\n",
    "def getUserBuyDeltaTimeFeature(jdata_df, jdata_user_order_data):\n",
    "    jdata_df = getUserBuyDeltaTimeFeatureMonthly(jdata_df, jdata_user_order_data, -1)\n",
    "    jdata_df = getUserBuyDeltaTimeFeatureMonthly(jdata_df, jdata_user_order_data, -2)\n",
    "    jdata_df = getUserBuyDeltaTimeFeatureMonthly(jdata_df, jdata_user_order_data, 101)\n",
    "    jdata_df = getUserBuyDeltaTimeFeatureMonthly(jdata_df, jdata_user_order_data, 30)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyDeltaTimeFeature(jdata_train_df_201704, jdata_user_order_data)\n",
    "jdata_train_df_201703 = getUserBuyDeltaTimeFeature(jdata_train_df_201703, jdata_user_order_data)\n",
    "jdata_train_df_201702 = getUserBuyDeltaTimeFeature(jdata_train_df_201702, jdata_user_order_data)\n",
    "jdata_train_df_201701 = getUserBuyDeltaTimeFeature(jdata_train_df_201701, jdata_user_order_data)\n",
    "jdata_train_df_201612 = getUserBuyDeltaTimeFeature(jdata_train_df_201612, jdata_user_order_data)\n",
    "jdata_train_df_201611 = getUserBuyDeltaTimeFeature(jdata_train_df_201611, jdata_user_order_data)\n",
    "jdata_test_df = getUserBuyDeltaTimeFeature(jdata_test_df, jdata_user_order_data)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 统计用户上次购买商品时间间隔－平均购买时间间隔特征\n",
    "def getUserBuyDeltaTimeLastAve(jdata_df):\n",
    "    jdata_df['user_cate30_last-ave'] = jdata_df['user_cate30_lastBuyDeltaTime'] - jdata_df['user_cate30_deltaTime_mean'] \n",
    "    jdata_df['user_cate101_last-ave'] = jdata_df['user_cate101_lastBuyDeltaTime'] - jdata_df['user_cate101_deltaTime_mean'] \n",
    "    jdata_df['user_cate-1_last-ave'] = jdata_df['user_targetCate_lastBuyDeltaTime'] - jdata_df['user_cate-1_deltaTime_mean'] \n",
    "    jdata_df['user_cate-2_last-ave'] = jdata_df['user_relatedCate_lastBuyDeltaTime'] - jdata_df['user_cate-2_deltaTime_mean'] \n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyDeltaTimeLastAve(jdata_train_df_201704)\n",
    "jdata_train_df_201703 = getUserBuyDeltaTimeLastAve(jdata_train_df_201703)\n",
    "jdata_train_df_201702 = getUserBuyDeltaTimeLastAve(jdata_train_df_201702)\n",
    "jdata_train_df_201701 = getUserBuyDeltaTimeLastAve(jdata_train_df_201701)\n",
    "jdata_train_df_201612 = getUserBuyDeltaTimeLastAve(jdata_train_df_201612)\n",
    "jdata_train_df_201611 = getUserBuyDeltaTimeLastAve(jdata_train_df_201611)\n",
    "jdata_test_df = getUserBuyDeltaTimeLastAve(jdata_test_df)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "def getBayesSmoothParam(origion_rate):\n",
    "    origion_rate_mean = origion_rate.mean()\n",
    "    origion_rate_var = origion_rate.var()\n",
    "    alpha = origion_rate_mean / origion_rate_var * (origion_rate_mean * (1 - origion_rate_mean) - origion_rate_var)\n",
    "    beta = (1 - origion_rate_mean) / origion_rate_var * (origion_rate_mean * (1 - origion_rate_mean) - origion_rate_var)\n",
    "    print('origion_rate_mean : ', origion_rate_mean)\n",
    "    print('origion_rate_var : ', origion_rate_var)\n",
    "    print('alpha : ', alpha)\n",
    "    print('beta : ', beta)\n",
    "    return alpha, beta\n",
    "\n",
    "# 添加用户一个月前购买数量占三个月内购买数量的比例特征,进行贝叶斯平滑\n",
    "def getUserBuyNumberLastOneInThreeRatioMonthly(jdata_df, cate):\n",
    "    if ((cate == -1) | (cate == -2)):\n",
    "        if cate == -1:\n",
    "            ratioName = 'user_targetCate_buyNumber_lastOneInThreeRatio'\n",
    "            threeMonthNumberName = 'user_targetCate_last3Month_buyNumber'\n",
    "            oneMonthNumberName = 'user_targetCate_last1Month_buyNumber'\n",
    "        else:\n",
    "            ratioName = 'user_relatedCate_buyNumber_lastOneInThreeRatio'\n",
    "            threeMonthNumberName = 'user_relatedCate_last3Month_buyNumber'\n",
    "            oneMonthNumberName = 'user_relatedCate_last1Month_buyNumber'\n",
    "    else:\n",
    "        ratioName = 'user_cate' + str(cate) + '_buyNumber_lastOneInThreeRatio'\n",
    "        threeMonthNumberName = 'user_cate' + str(cate) + '_last3Month_buyNumber'\n",
    "        oneMonthNumberName = 'user_cate' + str(cate) + '_last1Month_buyNumber'\n",
    "    jdata_df_copy = jdata_df.copy()\n",
    "    jdata_df_copy[ratioName] = -1\n",
    "    jdata_df_copy[ratioName][jdata_df_copy[threeMonthNumberName] != 0] = jdata_df_copy[oneMonthNumberName][jdata_df_copy[threeMonthNumberName] != 0] / jdata_df_copy[threeMonthNumberName][jdata_df_copy[threeMonthNumberName] != 0] \n",
    "    alpha, beta = getBayesSmoothParam(jdata_df_copy[ratioName][jdata_df_copy[ratioName] != -1])\n",
    "    jdata_df_copy[ratioName] = (jdata_df_copy[oneMonthNumberName] + alpha) / (jdata_df_copy[threeMonthNumberName] + alpha + beta)\n",
    "    jdata_df = pd.merge(jdata_df, jdata_df_copy[['user_id', ratioName]], on=['user_id'], how='left')\n",
    "    return jdata_df\n",
    "\n",
    "def getUserBuyNumberLastOneInThreeRatio(jdata_df):\n",
    "    jdata_df = getUserBuyNumberLastOneInThreeRatioMonthly(jdata_df, 101)\n",
    "    jdata_df = getUserBuyNumberLastOneInThreeRatioMonthly(jdata_df, 30)\n",
    "    jdata_df = getUserBuyNumberLastOneInThreeRatioMonthly(jdata_df, -1)\n",
    "    jdata_df = getUserBuyNumberLastOneInThreeRatioMonthly(jdata_df, -2)\n",
    "    return jdata_df\n",
    "\n",
    "jdata_train_df_201704 = getUserBuyNumberLastOneInThreeRatio(jdata_train_df_201704)\n",
    "jdata_train_df_201703 = getUserBuyNumberLastOneInThreeRatio(jdata_train_df_201703)\n",
    "jdata_train_df_201702 = getUserBuyNumberLastOneInThreeRatio(jdata_train_df_201702)\n",
    "jdata_train_df_201701 = getUserBuyNumberLastOneInThreeRatio(jdata_train_df_201701)\n",
    "jdata_train_df_201612 = getUserBuyNumberLastOneInThreeRatio(jdata_train_df_201612)\n",
    "jdata_train_df_201611 = getUserBuyNumberLastOneInThreeRatio(jdata_train_df_201611)\n",
    "jdata_test_df = getUserBuyNumberLastOneInThreeRatio(jdata_test_df)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(len(jdata_train_df_201704))\n",
    "print(len(jdata_train_df_201704.columns.values))\n",
    "print(jdata_train_df_201704.columns.values)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "jdata_train_df_baseline = pd.concat([jdata_train_df_201611, jdata_train_df_201612, jdata_train_df_201701, jdata_train_df_201702, jdata_train_df_201703, jdata_train_df_201704])\n",
    "print(len(jdata_train_df_baseline))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 导出训练集和测试集预处理结果\n",
    "def exportResult(df, fileName):\n",
    "    df.to_csv('./%s.csv' % fileName, header=True, index=False)\n",
    "    \n",
    "exportResult(jdata_train_df_baseline, 'chusai_jdata_train_df_baseline')\n",
    "exportResult(jdata_test_df, 'chusai_jdata_test_df_baseline')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}