jdata_feature_maker.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import lightgbm as lgb\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import math\n",
    "import gc\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import log_loss\n",
    "from sklearn import preprocessing\n",
    "lbl = preprocessing.LabelEncoder()\n",
    "import warnings\n",
    "from SBBTree_ONLINE import SBBTree\n",
    "from datetime import datetime, timedelta\n",
    "warnings.filterwarnings('ignore')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%run com_util.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(108333, 6) (99446, 4) (8033923, 5) (881692, 6) (233366, 4)\n"
     ]
    }
   ],
   "source": [
    "path = 'D:/learningdata/Mlearing/JData/B'\n",
    "sku_basic_info=pd.read_csv(path+'/jdata_sku_basic_info.csv')\n",
    "user_basic_info=pd.read_csv(path+'/jdata_user_basic_info.csv')\n",
    "user_action=pd.read_csv(path+'/jdata_user_action.csv')\n",
    "user_order=pd.read_csv(path+'/jdata_user_order.csv')\n",
    "user_comment_score=pd.read_csv(path+'/jdata_user_comment_score.csv')\n",
    "#sku_basic_info = sku_basic_info[(sku_basic_info.cate==30)|(sku_basic_info.cate==101)]\n",
    "print (sku_basic_info.shape,user_basic_info.shape,user_action.shape,user_order.shape,user_comment_score.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "####处理数据\n",
    "user_action['a_date'] = pd.to_datetime(user_action['a_date'])\n",
    "user_order['o_date'] = pd.to_datetime(user_order['o_date'])\n",
    "user_comment_score['comment_create_tm'] = pd.to_datetime(user_comment_score['comment_create_tm'])\n",
    "\n",
    "# sort by datetime\n",
    "user_action = user_action.sort_values(['user_id','a_date'])\n",
    "user_order = user_order.sort_values(['user_id','o_date'])\n",
    "user_comment_score = user_comment_score.sort_values(['user_id','comment_create_tm'])\n",
    "\n",
    "# year month day\n",
    "user_order['o_year'] = user_order['o_date'].dt.year\n",
    "user_order['o_month'] = user_order['o_date'].dt.month\n",
    "user_order['o_day'] = user_order['o_date'].dt.day\n",
    "\n",
    "user_action['a_year'] = user_action['a_date'].dt.year\n",
    "user_action['a_month'] = user_action['a_date'].dt.month\n",
    "user_action['a_day'] = user_action['a_date'].dt.day\n",
    "\n",
    "user_comment_score['comment_year'] = user_comment_score['comment_create_tm'].dt.year\n",
    "user_comment_score['comment_month'] = user_comment_score['comment_create_tm'].dt.month\n",
    "user_comment_score['comment_day'] = user_comment_score['comment_create_tm'].dt.day\n",
    "\n",
    "#Order Comment User Sku\n",
    "Order_Comment_User_Sku = user_order.\\\n",
    "                            merge(user_comment_score,on=['user_id','o_id'],how='left').\\\n",
    "                            merge(user_basic_info,on='user_id',how='left').\\\n",
    "                            merge(sku_basic_info,on='sku_id',how='left')\n",
    "# Action User Sku\n",
    "Action_User_Sku = user_action.\\\n",
    "                        merge(user_basic_info,on='user_id',how='left').\\\n",
    "                        merge(sku_basic_info,on='sku_id',how='left')\n",
    "\n",
    "Action_User_Sku = Action_User_Sku.drop(['age','sex','user_lv_cd','price','cate','para_1','para_2','para_3'],axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###dataset split :          其中dataset为提取label，，，feature为提取特征\n",
    "###dataset3: 20170901-20170930         feature3:20170601-20170831  \n",
    "###dataset2: 20170801-20170831         feature2:20170501-20170731\n",
    "###dataset1: 20170701-20170731         featrue1:20170401-20170630\n",
    "\n",
    "######此处有可能出现的问题就是当commentdate在时间窗口内的时候，o_date不在\n",
    "\n",
    "dataset3_time = [datetime(2017,9,1),datetime(2017,9,30)]\n",
    "dataset2_time = [datetime(2017,8,1),datetime(2017,8,31)]\n",
    "dataset1_time = [datetime(2017,7,1),datetime(2017,7,31)]\n",
    "feature3_time = [datetime(2017,6,1),datetime(2017,8,31)] \n",
    "feature2_time = [datetime(2017,5,1),datetime(2017,7,31)]\n",
    "feature1_time = [datetime(2017,4,1),datetime(2017,6,30)]\n",
    "\n",
    "feature1_OC = Order_Comment_User_Sku[((Order_Comment_User_Sku.o_date>=feature1_time[0])&(Order_Comment_User_Sku.o_date<=feature1_time[1]))|((Order_Comment_User_Sku.comment_create_tm>=feature1_time[0])&(Order_Comment_User_Sku.comment_create_tm<=feature1_time[1]))]\n",
    "feature1_A = Action_User_Sku[(Action_User_Sku.a_date>=feature1_time[0])&(Action_User_Sku.a_date<=feature1_time[1])]\n",
    "feature1 = feature1_OC.merge(feature1_A,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "feature2_OC = Order_Comment_User_Sku[((Order_Comment_User_Sku.o_date>=feature2_time[0])&(Order_Comment_User_Sku.o_date<=feature2_time[1]))|((Order_Comment_User_Sku.comment_create_tm>=feature2_time[0])&(Order_Comment_User_Sku.comment_create_tm<=feature2_time[1]))]\n",
    "feature2_A = Action_User_Sku[(Action_User_Sku.a_date>=feature2_time[0])&(Action_User_Sku.a_date<=feature2_time[1])]\n",
    "feature2 = feature2_OC.merge(feature2_A,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "feature3_OC = Order_Comment_User_Sku[((Order_Comment_User_Sku.o_date>=feature3_time[0])&(Order_Comment_User_Sku.o_date<=feature3_time[1]))|((Order_Comment_User_Sku.comment_create_tm>=feature3_time[0])&(Order_Comment_User_Sku.comment_create_tm<=feature3_time[1]))]\n",
    "feature3_A = Action_User_Sku[(Action_User_Sku.a_date>=feature3_time[0])&(Action_User_Sku.a_date<=feature3_time[1])]\n",
    "feature3 = feature3_OC.merge(feature3_A,on=['user_id','sku_id'],how='left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 4) (0, 4)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "218"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "###构造label\n",
    "feature1 = feature1.fillna(0)\n",
    "dataset1 = feature1[(featrue1.o_date!=0)|(feature.a_date!=0)|(feature.comment_create_tm!=0)]\n",
    "dataset1 = dataset1[['sku_id','user_id','o_day','o_date']]\n",
    "dataset1['label_1'] = 1\n",
    "dataset1['label_2'] = dataset1['o_day']\n",
    "dataset1 = dataset1[['sku_id','user_id','label_1','label_2']]\n",
    "dataset1 = dataset1.drop_duplicates()\n",
    "\n",
    "\n",
    "dataset2 = feature2[((feature2.o_date>=dataset2_time[0])&(feature2.o_date<=dataset2_time[1]))&((feature2.cate==30) | (feature2.cate==101))]\n",
    "dataset2 = dataset2[['sku_id','user_id','o_day','o_date']]\n",
    "dataset2['label_1'] = 1 \n",
    "dataset2['label_2'] = dataset2['o_day']\n",
    "dataset2 = dataset2[['sku_id','user_id','label_1','label_2']]\n",
    "dataset2 = dataset2.drop_duplicates()\n",
    "print(dataset2.shape,dataset1.shape)\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###商品特征:(sku_feature)\n",
    "\n",
    "'''\n",
    "预测目标:cate = {30,101}\n",
    "\n",
    "自带特征：价格price，商品参数para_1，品类cate\n",
    "\n",
    "在规定的时间窗口内：\n",
    "    每个商品在时间窗口内被购买的次数m1\n",
    "    M1/12此商品在京东7天的平均购买次数m2\n",
    "    M1/6此商品在京东15天的平均购买次数m3\n",
    "    M1/3此商品在京东31天的平均购买次数m4\n",
    "\n",
    "    每个商品在时间窗口内被购买的总数m5\n",
    "    M5/12此商品在京东7天的平均购买数目m6\n",
    "    M5/6此商品在京东15天的平均购买数目m7\n",
    "    M5/3此商品在京东31天的平均购买数目m8\n",
    "\n",
    "    每个商品在时间窗口内被浏览的总次数m9\n",
    "    M9/12此商品在京东7天的平均浏览数m10\n",
    "    M9/6此商品在京东15天的平均浏览数m11\n",
    "    M9/3此商品在京东31天的平均浏览数m12\n",
    "\n",
    "    每个商品在时间窗口内被关注的总次数m13\n",
    "    M13/12此商品在京东7天的平均浏览数m14\n",
    "    M13/6此商品在京东15天的平均浏览数m15\n",
    "    M13/3此商品在京东31天的平均浏览数m16\n",
    "\n",
    "    每个商品在时间窗口内被评论的总次数m17\n",
    "    M13/12此商品在京东7天的平均评论数m18\n",
    "    M13/6此商品在京东15天的平均评论数m19\n",
    "    M13/3此商品在京东31天的平均评论数m20\n",
    "\n",
    "\n",
    "    \n",
    "    此商品在时间窗口内被购买的平均时间间隔m21\n",
    "    此商品在时间窗口内被浏览的平均时间间隔m22\n",
    "    此商品在时间窗口内被关注的平均时间间隔m23\n",
    "    此商品在时间窗口内被评论的平均时间间隔m24\n",
    "    \n",
    "    此商品在时间窗口内最早被购买的距离月初的平均时间间隔m25\n",
    "    此商品在时间窗口内最早被浏览的距离月初的平均时间间隔m26\n",
    "    此商品在时间窗口内最早被关注的距离月初的平均时间间隔m27\n",
    "    此商品在时间窗口内最早被评论的距离月初的平均时间间隔m28\n",
    "    \n",
    "    此商品在时间窗口内最晚被购买的距离月末的平均时间间隔m29\n",
    "    此商品在时间窗口内最晚被浏览的距离月末的平均时间间隔m30\n",
    "    此商品在时间窗口内最晚被关注的距离月末的平均时间间隔m31\n",
    "    此商品在时间窗口内最晚被评论的距离月末的平均时间间隔m32\n",
    "    \n",
    "    每个商品在时间窗口内被购买的总金额m33\n",
    "    M33/12此商品在京东7天的平均消费金额m34\n",
    "    M33/6此商品在京东15天的平均消费金额m35\n",
    "    M33/3此商品在京东31天的平均消费金额m36\n",
    "    \n",
    "    此商品在时间窗口内最早购买的时间m37\n",
    "    此商品在时间窗口内最早浏览的时间m38\n",
    "    此商品在时间窗口内最早关注的时间m39\n",
    "    此商品在时间窗口内最早评论的时间m40\n",
    "    \n",
    "    此商品在时间窗口内最晚购买的时间m41\n",
    "    此商品在时间窗口内最晚浏览的时间m42\n",
    "    此商品在时间窗口内最晚关注的时间m43\n",
    "    此商品在时间窗口内最晚评论的时间m44\n",
    "\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\n自带特征：年龄age,用户等级user_lv_cd\\n\\n在规定的时间窗口内:\\n\\n    每个用户在时间窗口内购买的次数u1\\n    u1/12此用户在京东7天的平均购买次数u2\\n    u1/6此用户在京东15天的平均购买次数u3\\n    u1/3此用户在京东31天的平均购买次数u4\\n\\n    每个用户在时间窗口内购买的总数u5\\n    u5/12此用户在京东7天的平均购买数目u6\\n    u5/6此用户在京东15天的平均购买数目u7\\n    u5/3此用户在京东31天的平均购买数目u8\\n\\n    每个用户在时间窗口内浏览的总次数u9\\n    u9/12此用户在京东7天的平均浏览数u10\\n    u9/6此用户在京东15天的平均浏览数u11\\n    u9/3此用户在京东31天的平均浏览数u12\\n\\n    每个用户在时间窗口内关注的总次数u13\\n    u13/12此用户在京东7天的平均浏览数u14\\n    u13/6此用户在京东15天的平均浏览数u15\\n    u13/3此用户在京东31天的平均浏览数u16\\n\\n    每个用户在时间窗口内评论的总次数u13\\n    u13/12此用户在京东7天的平均评论数u14\\n    u13/6此用户在京东15天的平均评论数u15\\n    u13/3此用户在京东31天的平均评论数u16\\n\\n    每个用户在时间窗口内购买的总金额u17\\n    U17/12此用户在京东7天的平均消费金额u18\\n    U13/6此用户在京东15天的平均消费金额u19\\n    U13/3此用户在京东31天的平均消费金额u20\\n    \\n    此用户在时间窗口内购买的平均时间间隔u21\\n    此用户在时间窗口内浏览的平均时间间隔u21\\n    此用户在时间窗口内关注的平均时间间隔u22\\n    此用户在时间窗口内评论的平均时间间隔u23\\n\\n    此用户在时间窗口内最早购买的距离月初的平均时间间隔u24\\n    此用户在时间窗口内最早浏览的距离月初的平均时间间隔u25\\n    此用户在时间窗口内最早关注的距离月初的平均时间间隔u26\\n    此用户在时间窗口内最早评论的距离月初的平均时间间隔u27\\n    \\n    此用户在时间窗口内最晚购买的距离月末的平均时间间隔u28\\n    此用户在时间窗口内最晚浏览的距离月末的平均时间间隔u29\\n    此用户在时间窗口内最晚关注的距离月末的平均时间间隔u30\\n    此用户在时间窗口内最晚评论的距离月末的平均时间间隔u31\\n    \\n    每个用户在时间窗口内购买的总金额u33\\n    u33/12此用户在京东7天的平均消费金额u34\\n    u33/6此用户在京东15天的平均消费金额u35\\n    u33/3此用户在京东31天的平均消费金额u36\\n\\n    此用户在时间窗口内最早购买的时间u37\\n    此用户在时间窗口内最早浏览的时间u38\\n    此用户在时间窗口内最早关注的时间u39\\n    此用户在时间窗口内最早评论的时间u40\\n    \\n    此用户在时间窗口内最晚购买的时间u41\\n    此用户在时间窗口内最晚浏览的时间u42\\n    此用户在时间窗口内最晚关注的时间u43\\n    此用户在时间窗口内最晚评论的时间u44\\n\\n'"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "###用户特征：(user_feature)\n",
    "\n",
    "'''\n",
    "自带特征：年龄age,用户等级user_lv_cd\n",
    "\n",
    "在规定的时间窗口内:\n",
    "\n",
    "    每个用户在时间窗口内购买的次数u1\n",
    "    u1/12此用户在京东7天的平均购买次数u2\n",
    "    u1/6此用户在京东15天的平均购买次数u3\n",
    "    u1/3此用户在京东31天的平均购买次数u4\n",
    "\n",
    "    每个用户在时间窗口内购买的总数u5\n",
    "    u5/12此用户在京东7天的平均购买数目u6\n",
    "    u5/6此用户在京东15天的平均购买数目u7\n",
    "    u5/3此用户在京东31天的平均购买数目u8\n",
    "\n",
    "    每个用户在时间窗口内浏览的总次数u9\n",
    "    u9/12此用户在京东7天的平均浏览数u10\n",
    "    u9/6此用户在京东15天的平均浏览数u11\n",
    "    u9/3此用户在京东31天的平均浏览数u12\n",
    "\n",
    "    每个用户在时间窗口内关注的总次数u13\n",
    "    u13/12此用户在京东7天的平均浏览数u14\n",
    "    u13/6此用户在京东15天的平均浏览数u15\n",
    "    u13/3此用户在京东31天的平均浏览数u16\n",
    "\n",
    "    每个用户在时间窗口内评论的总次数u13\n",
    "    u13/12此用户在京东7天的平均评论数u14\n",
    "    u13/6此用户在京东15天的平均评论数u15\n",
    "    u13/3此用户在京东31天的平均评论数u16\n",
    "\n",
    "    每个用户在时间窗口内购买的总金额u17\n",
    "    U17/12此用户在京东7天的平均消费金额u18\n",
    "    U13/6此用户在京东15天的平均消费金额u19\n",
    "    U13/3此用户在京东31天的平均消费金额u20\n",
    "    \n",
    "    此用户在时间窗口内购买的平均时间间隔u21\n",
    "    此用户在时间窗口内浏览的平均时间间隔u21\n",
    "    此用户在时间窗口内关注的平均时间间隔u22\n",
    "    此用户在时间窗口内评论的平均时间间隔u23\n",
    "\n",
    "    此用户在时间窗口内最早购买的距离月初的平均时间间隔u24\n",
    "    此用户在时间窗口内最早浏览的距离月初的平均时间间隔u25\n",
    "    此用户在时间窗口内最早关注的距离月初的平均时间间隔u26\n",
    "    此用户在时间窗口内最早评论的距离月初的平均时间间隔u27\n",
    "    \n",
    "    此用户在时间窗口内最晚购买的距离月末的平均时间间隔u28\n",
    "    此用户在时间窗口内最晚浏览的距离月末的平均时间间隔u29\n",
    "    此用户在时间窗口内最晚关注的距离月末的平均时间间隔u30\n",
    "    此用户在时间窗口内最晚评论的距离月末的平均时间间隔u31\n",
    "    \n",
    "    每个用户在时间窗口内购买的总金额u33\n",
    "    u33/12此用户在京东7天的平均消费金额u34\n",
    "    u33/6此用户在京东15天的平均消费金额u35\n",
    "    u33/3此用户在京东31天的平均消费金额u36\n",
    "\n",
    "    此用户在时间窗口内最早购买的时间u37\n",
    "    此用户在时间窗口内最早浏览的时间u38\n",
    "    此用户在时间窗口内最早关注的时间u39\n",
    "    此用户在时间窗口内最早评论的时间u40\n",
    "    \n",
    "    此用户在时间窗口内最晚购买的时间u41\n",
    "    此用户在时间窗口内最晚浏览的时间u42\n",
    "    此用户在时间窗口内最晚关注的时间u43\n",
    "    此用户在时间窗口内最晚评论的时间u44\n",
    "\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\n在规定的时间窗口内:\\n\\n    用户对此商品在时间窗口内购买的次数um1\\n    UM1/12此用户对此商品在京东7天的平均购买次数um2\\n    UM1/6此用户对此商品在京东15天的平均购买次数um3\\n    UM1/3此用户对此商品在京东31天的平均购买次数um4\\n\\n    用户对此商品在时间窗口内购买的总数um5\\n    UM5/12此用户对此商品在京东7天的平均购买数目um6\\n    UM5/6此用户对此商品在京东15天的平均购买数目um7\\n    UM5/3此用户对此商品在京东31天的平均购买数目um8\\n\\n    用户在对此商品时间窗口内浏览的总次数um9\\n    UM9/12此用户对此商品在京东7天的平均浏览数um10\\n    UM9/6此用户对此商品在京东15天的平均浏览数um11\\n    UM9/3此用户对此商品在京东31天的平均浏览数um12\\n\\n    用户对此商品在时间窗口内关注的总次数um13\\n    UM13/12用户对此商品在京东7天的平均浏览数um14\\n    UM13/6用户对此商品在京东15天的平均浏览数um15\\n    UM13/3用户对此商品在京东31天的平均浏览数um16\\n\\n    用户对此商品在时间窗口内评论的总次数um13\\n    UM13/12此用户对此商品在京东7天的平均评论数um14\\n    UM13/6此用户对此商品在京东15天的平均评论数um15\\n    UM13/3此用户对此商品在京东31天的平均评论数um16\\n\\n    用户对此商品在时间窗口内购买的总金额um17\\n    UM17/12此用户对此商品在京东7天的平均消费金额um18\\n    UM13/6此用户对此商品在京东15天的平均消费金额um19\\n    UM13/3此用户对此商品在京东31天的平均消费金额um20\\n    \\n    此用户对此商品在时间窗口内购买的平均时间间隔um21\\n    此用户对此商品在时间窗口内浏览的平均时间间隔um21\\n    此用户对此商品在时间窗口内关注的平均时间间隔um22\\n    此用户对此商品在时间窗口内评论的平均时间间隔um23\\n\\n    此用户在时间窗口内对此商品最早购买的距离月初的平均时间间隔um24\\n    此用户在时间窗口内对此商品最早浏览的距离月初的平均时间间隔um25\\n    此用户在时间窗口内对此商品最早关注的距离月初的平均时间间隔um26\\n    此用户在时间窗口内对此商品最早评论的距离月初的平均时间间隔um27\\n    \\n    此用户在时间窗口内对此商品最晚购买的距离月末的平均时间间隔um28\\n    此用户在时间窗口内对此商品最晚浏览的距离月末的平均时间间隔um29\\n    此用户在时间窗口内对此商品最晚关注的距离月末的平均时间间隔um30\\n    此用户在时间窗口内对此商品最晚评论的距离月末的平均时间间隔um31\\n\\n    用户在时间窗口内对该商品购买的总金额um33\\n    M17/12此用户在京东7天该商品的平均消费金额um34\\n    M13/6此用户在商品在京东15天该商品的平均消费金额um35\\n    M13/3此用户在商品在京东31天该商品的平均消费金额um36\\n    \\n    此用户对此商品在时间窗口内最早购买的时间u37\\n    此用户对此商品在时间窗口内最早浏览的时间u38\\n    此用户对此商品在时间窗口内最早关注的时间u39\\n    此用户对此商品在时间窗口内最早评论的时间u40\\n    \\n    此用户对此商品在时间窗口内最晚购买的时间u41\\n    此用户对此商品在时间窗口内最晚浏览的时间u42\\n    此用户对此商品在时间窗口内最晚关注的时间u43\\n    此用户对此商品在时间窗口内最晚评论的时间u44\\n'"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#####商品用户特征(user_sku_feature)\n",
    "\n",
    "'''\n",
    "在规定的时间窗口内:\n",
    "\n",
    "    用户对此商品在时间窗口内购买的次数um1\n",
    "    UM1/12此用户对此商品在京东7天的平均购买次数um2\n",
    "    UM1/6此用户对此商品在京东15天的平均购买次数um3\n",
    "    UM1/3此用户对此商品在京东31天的平均购买次数um4\n",
    "\n",
    "    用户对此商品在时间窗口内购买的总数um5\n",
    "    UM5/12此用户对此商品在京东7天的平均购买数目um6\n",
    "    UM5/6此用户对此商品在京东15天的平均购买数目um7\n",
    "    UM5/3此用户对此商品在京东31天的平均购买数目um8\n",
    "\n",
    "    用户在对此商品时间窗口内浏览的总次数um9\n",
    "    UM9/12此用户对此商品在京东7天的平均浏览数um10\n",
    "    UM9/6此用户对此商品在京东15天的平均浏览数um11\n",
    "    UM9/3此用户对此商品在京东31天的平均浏览数um12\n",
    "\n",
    "    用户对此商品在时间窗口内关注的总次数um13\n",
    "    UM13/12用户对此商品在京东7天的平均浏览数um14\n",
    "    UM13/6用户对此商品在京东15天的平均浏览数um15\n",
    "    UM13/3用户对此商品在京东31天的平均浏览数um16\n",
    "\n",
    "    用户对此商品在时间窗口内评论的总次数um13\n",
    "    UM13/12此用户对此商品在京东7天的平均评论数um14\n",
    "    UM13/6此用户对此商品在京东15天的平均评论数um15\n",
    "    UM13/3此用户对此商品在京东31天的平均评论数um16\n",
    "\n",
    "    用户对此商品在时间窗口内购买的总金额um17\n",
    "    UM17/12此用户对此商品在京东7天的平均消费金额um18\n",
    "    UM13/6此用户对此商品在京东15天的平均消费金额um19\n",
    "    UM13/3此用户对此商品在京东31天的平均消费金额um20\n",
    "    \n",
    "    此用户对此商品在时间窗口内购买的平均时间间隔um21\n",
    "    此用户对此商品在时间窗口内浏览的平均时间间隔um21\n",
    "    此用户对此商品在时间窗口内关注的平均时间间隔um22\n",
    "    此用户对此商品在时间窗口内评论的平均时间间隔um23\n",
    "\n",
    "    此用户在时间窗口内对此商品最早购买的距离月初的平均时间间隔um24\n",
    "    此用户在时间窗口内对此商品最早浏览的距离月初的平均时间间隔um25\n",
    "    此用户在时间窗口内对此商品最早关注的距离月初的平均时间间隔um26\n",
    "    此用户在时间窗口内对此商品最早评论的距离月初的平均时间间隔um27\n",
    "    \n",
    "    此用户在时间窗口内对此商品最晚购买的距离月末的平均时间间隔um28\n",
    "    此用户在时间窗口内对此商品最晚浏览的距离月末的平均时间间隔um29\n",
    "    此用户在时间窗口内对此商品最晚关注的距离月末的平均时间间隔um30\n",
    "    此用户在时间窗口内对此商品最晚评论的距离月末的平均时间间隔um31\n",
    "\n",
    "    用户在时间窗口内对该商品购买的总金额um33\n",
    "    M17/12此用户在京东7天该商品的平均消费金额um34\n",
    "    M13/6此用户在商品在京东15天该商品的平均消费金额um35\n",
    "    M13/3此用户在商品在京东31天该商品的平均消费金额um36\n",
    "    \n",
    "    此用户对此商品在时间窗口内最早购买的时间u37\n",
    "    此用户对此商品在时间窗口内最早浏览的时间u38\n",
    "    此用户对此商品在时间窗口内最早关注的时间u39\n",
    "    此用户对此商品在时间窗口内最早评论的时间u40\n",
    "    \n",
    "    此用户对此商品在时间窗口内最晚购买的时间u41\n",
    "    此用户对此商品在时间窗口内最晚浏览的时间u42\n",
    "    此用户对此商品在时间窗口内最晚关注的时间u43\n",
    "    此用户对此商品在时间窗口内最晚评论的时间u44\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "'f3_buytimes_30101_o_id_sku_id_nunique'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\indexes\\base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   2524\u001b[0m             \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2525\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2526\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 'f3_buytimes_30101_o_id_sku_id_nunique'",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-12-0c8c3d2d9f1b>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     13\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     14\u001b[0m \u001b[1;31m###m2-m4\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 15\u001b[1;33m \u001b[0msample_sku3\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f3_buytimes_30101_in_7days_'\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msample_sku3\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f3_buytimes_30101_o_id_sku_id_nunique'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;36m12\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     16\u001b[0m \u001b[0msample_sku3\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f3_buytimes_30_in_7days_'\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msample_sku3\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f3_buytimes_30_o_id_sku_id_nunique'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;36m12\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     17\u001b[0m \u001b[0msample_sku3\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f3_buytimes_101_in_7days_'\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msample_sku3\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f3_buytimes_101_o_id_sku_id_nunique'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;36m12\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   2137\u001b[0m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_multilevel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2138\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2139\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_column\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2140\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2141\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_getitem_column\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m_getitem_column\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   2144\u001b[0m         \u001b[1;31m# get column\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2145\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mis_unique\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2146\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_item_cache\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2147\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2148\u001b[0m         \u001b[1;31m# duplicate columns & possible reduce dimensionality\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\generic.py\u001b[0m in \u001b[0;36m_get_item_cache\u001b[1;34m(self, item)\u001b[0m\n\u001b[0;32m   1840\u001b[0m         \u001b[0mres\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcache\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1841\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mres\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1842\u001b[1;33m             \u001b[0mvalues\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_data\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1843\u001b[0m             \u001b[0mres\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_box_item_values\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvalues\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1844\u001b[0m             \u001b[0mcache\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mres\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\internals.py\u001b[0m in \u001b[0;36mget\u001b[1;34m(self, item, fastpath)\u001b[0m\n\u001b[0;32m   3841\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   3842\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0misna\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 3843\u001b[1;33m                 \u001b[0mloc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   3844\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   3845\u001b[0m                 \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0marange\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0misna\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\indexes\\base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   2525\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2526\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2527\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_maybe_cast_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2528\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2529\u001b[0m         \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mmethod\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtolerance\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mtolerance\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 'f3_buytimes_30101_o_id_sku_id_nunique'"
     ]
    }
   ],
   "source": [
    "####商品特征\n",
    "#######dataset3_sku\n",
    "#######\n",
    "####for dataset3\n",
    "sku3 = feature3[['sku_id','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','price','cate','para_1','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = sku3[['sku_id','price','cate']]\n",
    "t = t[(t.cate==30)|(t.cate==101)]\n",
    "\n",
    "t.drop_duplicates(inplace=True)\n",
    "###m1\n",
    "sample_sku3 = feat_nunique(t,sku3[(sku3.cate==101)|(sku3.cate==30)],['sku_id'],'o_id','f3_buytimes_30101_')\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3[(sku3.cate==30)],['sku_id'],'o_id','f3_buytimes_30_')\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3[(sku3.cate==101)],['sku_id'],'o_id','f3_buytimes_101_')\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3[(sku3.cate!=101)&(sku3.cate!=30)],['sku_id'],'o_id','f3_buytimes_no_30101_')\n",
    "\n",
    "###m2-m4\n",
    "sample_sku3['f3_buytimes_30101_in_7days_'] = sample_sku3['f3_buytimes_30101_o_id_sku_id_nunique']/12\n",
    "sample_sku3['f3_buytimes_30_in_7days_'] = sample_sku3['f3_buytimes_30_o_id_sku_id_nunique']/12\n",
    "sample_sku3['f3_buytimes_101_in_7days_'] = sample_sku3['f3_buytimes_101_o_id_sku_id_nunique']/12\n",
    "sample_sku3['f3_buytimes_no_30101_in_7days_'] = sample_sku3['f3_buytimes_no_30101_o_id_sku_id_nunique']/12\n",
    "\n",
    "sample_sku3['f3_buytimes_30101_in_15days_'] = sample_sku3['f3_buytimes_30101_o_id_sku_id_nunique']/6\n",
    "sample_sku3['f3_buytimes_30_in_15days_'] = sample_sku3['f3_buytimes_30_o_id_sku_id_nunique']/6\n",
    "sample_sku3['f3_buytimes_101_in_15days_'] = sample_sku3['f3_buytimes_101_o_id_sku_id_nunique']/6\n",
    "sample_sku3['f3_buytimes_no_30101_in_15days_'] = sample_sku3['f3_buytimes_no_30101_o_id_sku_id_nunique']/6\n",
    "\n",
    "sample_sku3['f3_buytimes_30101_in_31days_'] = sample_sku3['f3_buytimes_30101_o_id_sku_id_nunique']/3\n",
    "sample_sku3['f3_buytimes_30_in_31days_'] = sample_sku3['f3_buytimes_30_o_id_sku_id_nunique']/3\n",
    "sample_sku3['f3_buytimes_101_in_31days_'] = sample_sku3['f3_buytimes_101_o_id_sku_id_nunique']/3\n",
    "sample_sku3['f3_buytimes_no_30101_in_31days_'] = sample_sku3['f3_buytimes_no_30101_o_id_sku_id_nunique']/3\n",
    "\n",
    "###m5\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[(sku3.cate==101)|(sku3.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30101_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[(sku3.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[(sku3.cate==101)],['sku_id'],'o_sku_num','f3_buynums_101_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[(sku3.cate!=101)&(sku3.cate!=30)],['sku_id'],'o_sku_num','f3_buynums_no_30101_')\n",
    "\n",
    "###m6-m8\n",
    "sample_sku3['f3_buynums_30101_in_7days_'] = sample_sku3['f3_buynums_30101_o_sku_num_sku_id_sum']/12\n",
    "sample_sku3['f3_buynums_30_in_7days_'] = sample_sku3['f3_buynums_30_o_sku_num_sku_id_sum']/12\n",
    "sample_sku3['f3_buynums_101_in_7days_'] = sample_sku3['f3_buynums_101_o_sku_num_sku_id_sum']/12\n",
    "sample_sku3['f3_buynums_no_30101_in_7days_'] = sample_sku3['f3_buynums_no_30101_o_sku_num_sku_id_sum']/12\n",
    "\n",
    "sample_sku3['f3_buynums_30101_in_15days_'] = sample_sku3['f3_buynums_30101_o_sku_num_sku_id_sum']/6\n",
    "sample_sku3['f3_buynums_30_in_15days_'] = sample_sku3['f3_buynums_30_o_sku_num_sku_id_sum']/6\n",
    "sample_sku3['f3_buynums_101_in_15days_'] = sample_sku3['f3_buynums_101_o_sku_num_sku_id_sum']/6\n",
    "sample_sku3['f3_buynums_no_30101_in_15days_'] = sample_sku3['f3_buynums_no_30101_o_sku_num_sku_id_sum']/6\n",
    "\n",
    "sample_sku3['f3_buynums_30101_in_31days_'] = sample_sku3['f3_buynums_30101_o_sku_num_sku_id_sum']/3\n",
    "sample_sku3['f3_buynums_30_in_31days_'] = sample_sku3['f3_buynums_30_o_sku_num_sku_id_sum']/3\n",
    "sample_sku3['f3_buynums_101_in_31days_'] = sample_sku3['f3_buynums_101_o_sku_num_sku_id_sum']/3\n",
    "sample_sku3['f3_buynums_no_30101_in_31days_'] = sample_sku3['f3_buynums_no_30101_o_sku_num_sku_id_sum']/3\n",
    "\n",
    "###m9\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate==101)|(sku3.cate==30))&(sku3.a_type==1)],['sku_id'],'a_num','f3_scannums_30101_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate==30))&(sku3.a_type==1)],['sku_id'],'a_num','f3_scannums_30_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate==101))&(sku3.a_type==1)],['sku_id'],'a_num','f3_scannums_101_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate!=101)&(sku3.cate!=30))&(sku3.a_type==1)],['sku_id'],'a_num','f3_scannums_no_30101_')\n",
    "###m10-m12\n",
    "sample_sku3['f3_scannums_30101_in_7days_'] = sample_sku3['f3_scannums_30101_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_scannums_30_in_7days_'] = sample_sku3['f3_scannums_30_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_scannums_101_in_7days_'] = sample_sku3['f3_scannums_101_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_scannums_no_30101_in_7days_'] = sample_sku3['f3_scannums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_sku3['f3_scannums_30101_in_15days_'] = sample_sku3['f3_scannums_30101_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_scannums_30_in_15days_'] = sample_sku3['f3_scannums_30_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_scannums_101_in_15days_'] = sample_sku3['f3_scannums_101_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_scannums_no_30101_in_15days_'] = sample_sku3['f3_scannums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_sku3['f3_scannums_30101_in_31days_'] = sample_sku3['f3_scannums_30101_a_num_sku_id_sum']/3\n",
    "sample_sku3['f3_scannums_30_in_31days_'] = sample_sku3['f3_scannums_30_a_num_sku_id_sum']/3\n",
    "sample_sku3['f3_scannums_101_in_31days_'] = sample_sku3['f3_scannums_101_a_num_sku_id_sum']/3\n",
    "sample_sku3['f3_scannums_no_30101_in_31days_'] = sample_sku3['f3_scannums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "###m13\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate==101)|(sku3.cate==30))&(sku3.a_type==2)],['sku_id'],'a_num','f3_attnums_30101_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate==30))&(sku3.a_type==2)],['sku_id'],'a_num','f3_attnums_30_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate==101))&(sku3.a_type==2)],['sku_id'],'a_num','f3_attnums_101_')\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[((sku3.cate!=101)&(sku3.cate!=30))&(sku3.a_type==2)],['sku_id'],'a_num','f3_attnums_no_30101_')\n",
    "###m14-m16\n",
    "sample_sku3['f3_attnums_30101_in_7days_'] = sample_sku3['f3_attnums_30101_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_attnums_30_in_7days_'] = sample_sku3['f3_attnums_30_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_attnums_101_in_7days_'] = sample_sku3['f3_attnums_101_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_attnums_no_30101_in_7days_'] = sample_sku3['f3_attnums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_sku3['f3_attnums_30101_in_15days_'] = sample_sku3['f3_attnums_30101_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_attnums_30_in_15days_'] = sample_sku3['f3_attnums_30_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_attnums_101_in_15days_'] = sample_sku3['f3_attnums_101_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_attnums_no_30101_in_15days_'] = sample_sku3['f3_attnums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_sku3['f3_attnums_30101_in_31days_'] = sample_sku3['f3_attnums_30101_a_num_sku_id_sum']/3\n",
    "sample_sku3['f3_attnums_30_in_31days_'] = sample_sku3['f3_attnums_30_a_num_sku_id_sum']/3\n",
    "sample_sku3['f3_attnums_101_in_31days_'] = sample_sku3['f3_attnums_101_a_num_sku_id_sum']/3\n",
    "sample_sku3['f3_attnums_no_30101_in_31days_'] = sample_sku3['f3_attnums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "\n",
    "###m17\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3[(sku3.cate==101)|(sku3.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30101_')\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3[(sku3.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30_')\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3[(sku3.cate==101)],['sku_id'],'comment_create_tm','f3_commentnums_101_')\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3[(sku3.cate!=101)&(sku3.cate!=30)],['sku_id'],'comment_create_tm','f3_commentnums_no_30101_')\n",
    "\n",
    "###m18-20\n",
    "sample_sku3['f3_commentnums_30101_in_7days_'] = sample_sku3['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_sku3['f3_commentnums_30_in_7days_'] = sample_sku3['f3_commentnums_30_comment_create_tm_sku_id_nunique']/12\n",
    "sample_sku3['f3_commentnums_101_in_7days_'] = sample_sku3['f3_commentnums_101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_sku3['f3_commentnums_no_30101_in_7days_'] = sample_sku3['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/12\n",
    "\n",
    "sample_sku3['f3_commentnums_30101_in_15days_'] = sample_sku3['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_sku3['f3_commentnums_30_in_15days_'] = sample_sku3['f3_commentnums_30_comment_create_tm_sku_id_nunique']/6\n",
    "sample_sku3['f3_commentnums_101_in_15days_'] = sample_sku3['f3_commentnums_101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_sku3['f3_commentnums_no_30101_in_15days_'] = sample_sku3['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/6\n",
    "\n",
    "sample_sku3['f3_commentnums_30101_in_31days_'] = sample_sku3['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_sku3['f3_commentnums_30_in_31days_'] = sample_sku3['f3_commentnums_30_comment_create_tm_sku_id_nunique']/3\n",
    "sample_sku3['f3_commentnums_101_in_31days_'] = sample_sku3['f3_commentnums_101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_sku3['f3_commentnums_no_30101_in_31days_'] = sample_sku3['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/3\n",
    "\n",
    "###m21\n",
    "t1 = sample_sku3[['sku_id','f3_buytimes_30101_o_id_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate==101)|(sku3.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate==101)|(sku3.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_buy_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_buytimes_30_o_id_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_buy_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_buytimes_101_o_id_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate==101)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate==101)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_buy_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_buytimes_no_30101_o_id_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate!=101)&(sku3.cate!=30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate!=101)&(sku3.cate!=30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_no_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_buy_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m22\n",
    "t1 = sample_sku3[['sku_id','f3_scannums_30101_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==1)&((sku3.cate==101)|(sku3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==1)&((sku3.cate==101)|(sku3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_scan_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_scannums_30_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==1)&((sku3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==1)&((sku3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_scan_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_scannums_101_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==1)&((sku3.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==1)&((sku3.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_scan_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_scannums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==1)&((sku3.cate!=30)&(sku3.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==1)&((sku3.cate!=30)&(sku3.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_scan_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m23\n",
    "t1 = sample_sku3[['sku_id','f3_attnums_30101_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==2)&((sku3.cate==101)|(sku3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==2)&((sku3.cate==101)|(sku3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_att_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_attnums_30_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==2)&((sku3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==2)&((sku3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_att_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_attnums_101_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==2)&((sku3.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==2)&((sku3.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_att_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_sku3[['sku_id','f3_attnums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = sku3[(sku3.a_type==2)&((sku3.cate!=30)&(sku3.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.a_type==2)&((sku3.cate!=30)&(sku3.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_att_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m24\n",
    "t1 = sample_sku3[['sku_id','f3_commentnums_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate==101)|(sku3.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate==101)|(sku3.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_comment_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "t1 = sample_sku3[['sku_id','f3_commentnums_30_comment_create_tm_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_comment_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_sku3[['sku_id','f3_commentnums_101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate==101)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate==101)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_comment_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_sku3[['sku_id','f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = sku3[(sku3.cate!=101)&(sku3.cate!=30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[(sku3.cate!=101)&(sku3.cate!=30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_comment_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m25-m32还没写，正在考虑\n",
    "sku3_a_o_date = sku3[((pd.to_datetime(sku3.a_date)<=pd.to_datetime(sku3.o_date.fillna('2018-01-01'))))]\n",
    "sku3_a_o_date = sku3_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "sku3_a_o_date = sku3_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "sku3_a_o_date['o_a_gap']=(pd.to_datetime(sku3_a_o_date.o_date)-pd.to_datetime(sku3_a_o_date.a_date)).dt.days \n",
    "FLAG='sku3_a_o_'\n",
    "sample_sku3 = feat_count(sample_sku3,sku3_a_o_date[sku3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_sku3 = feat_mean(sample_sku3,sku3_a_o_date[(sku3_a_o_date.o_a_gap>=-1)&((sku3_a_o_date.cate==101)|(sku3_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30101_')\n",
    "sample_sku3 = feat_mean(sample_sku3,sku3_a_o_date[(sku3_a_o_date.o_a_gap>=-1)&((sku3_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30_')\n",
    "sample_sku3 = feat_mean(sample_sku3,sku3_a_o_date[(sku3_a_o_date.o_a_gap>=-1)&((sku3_a_o_date.cate==101))],['sku_id'],'o_a_gap',FLAG+'101_')\n",
    "sample_sku3 = feat_mean(sample_sku3,sku3_a_o_date[(sku3_a_o_date.o_a_gap>=-1)&((sku3_a_o_date.cate!=101)&(sku3_a_o_date.cate!=30))],['sku_id'],'o_a_gap',FLAG+'no_30101_')\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "###m33\n",
    "t1 = sample_sku3[['sku_id','f3_buynums_o_sku_num_sku_id_sum']] \n",
    "t2 = sku3[['sku_id','price']]\n",
    "t3 = t1.merge(t2,on=['sku_id'],how='left')\n",
    "t3['f3_total_price'] = t3['f3_buynums_o_sku_num_sku_id_sum']*t3['price']\n",
    "t3 = t3[['sku_id','f3_total_price']]\n",
    "sample_sku3 = sample_sku3.merge(t3,on=['sku_id'],how='left')\n",
    "\n",
    "###m34-m36\n",
    "sample_sku3['f3_total_price_in_7days_'] = sample_sku3['f3_total_price']/12\n",
    "sample_sku3['f3_total_price_in_15days_'] = sample_sku3['f3_total_price']/6\n",
    "sample_sku3['f3_total_price_in_31days_'] = sample_sku3['f3_total_price']/3\n",
    "\n",
    "###m37-m40\n",
    "sample_sku3 = feat_min(sample_sku3,sku3,['sku_id'],'o_day','f3_buytime_early_')\n",
    "sample_sku3 = feat_min(sample_sku3,sku3[sku3.a_type==1],['sku_id'],'a_day','f3_scantime_early_')\n",
    "sample_sku3 = feat_min(sample_sku3,sku3[sku3.a_type==2],['sku_id'],'a_day','f3_atttime_early_')\n",
    "sample_sku3 = feat_min(sample_sku3,sku3,['sku_id'],'comment_day','f3_commenttime_early_')\n",
    "\n",
    "###m41-m44\n",
    "sample_sku3 = feat_max(sample_sku3,sku3,['sku_id'],'o_day','f3_buytime_late_')\n",
    "sample_sku3 = feat_max(sample_sku3,sku3[sku3.a_type==1],['sku_id'],'a_day','f3_scantime_late_')\n",
    "sample_sku3 = feat_max(sample_sku3,sku3[sku3.a_type==2],['sku_id'],'a_day','f3_atttime_late_')\n",
    "sample_sku3 = feat_max(sample_sku3,sku3,['sku_id'],'comment_day','f3_commenttime_late_')\n",
    "\n",
    "###\n",
    "sample_sku3 = sample_sku3.drop_duplicates()\n",
    "gc.collect()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sku3 = feature3[['sku_id','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','price','cate','para_1','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = sku3[['sku_id','price','cate']]\n",
    "t = t[(t.cate==30)|(t.cate==101)]\n",
    "\n",
    "t.drop_duplicates(inplace=True)\n",
    "###m1\n",
    "sample_sku3 = feat_nunique(t,sku3,['sku_id'],'o_id','f3_buytimes_')\n",
    "###m2-m4\n",
    "sample_sku3['f3_buytimes_in_7days_'] = sample_sku3['f3_buytimes_o_id_sku_id_nunique']/12\n",
    "sample_sku3['f3_buytimes_in_15days_'] = sample_sku3['f3_buytimes_o_id_sku_id_nunique']/6\n",
    "sample_sku3['f3_buytimes_in_31days_'] = sample_sku3['f3_buytimes_o_id_sku_id_nunique']/3\n",
    "\n",
    "###m5\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3,['sku_id'],'o_sku_num','f3_buynums_')\n",
    "###m6-m8\n",
    "sample_sku3['f3_buynums_in_7days_'] = sample_sku3['f3_buynums_o_sku_num_sku_id_sum']/12\n",
    "sample_sku3['f3_buynums_in_15days_'] = sample_sku3['f3_buynums_o_sku_num_sku_id_sum']/6\n",
    "sample_sku3['f3_buynums_in_31days_'] = sample_sku3['f3_buynums_o_sku_num_sku_id_sum']/3\n",
    "\n",
    "###m9\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[sku3.a_type==1],['sku_id'],'a_num','f3_scannums_')\n",
    "###m10-m12\n",
    "sample_sku3['f3_scannums_in_7days_'] = sample_sku3['f3_scannums_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_scannums_in_15days_'] = sample_sku3['f3_scannums_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_scannums_in_31days_'] = sample_sku3['f3_scannums_a_num_sku_id_sum']/3\n",
    "\n",
    "###m13\n",
    "sample_sku3 = feat_sum(sample_sku3,sku3[sku3.a_type==2],['sku_id'],'a_num','f3_attnums_')\n",
    "###m14-m16\n",
    "sample_sku3['f3_attnums_in_7days_'] = sample_sku3['f3_attnums_a_num_sku_id_sum']/12\n",
    "sample_sku3['f3_attnums_in_15days_'] = sample_sku3['f3_attnums_a_num_sku_id_sum']/6\n",
    "sample_sku3['f3_attnums_in_31days_'] = sample_sku3['f3_attnums_a_num_sku_id_sum']/3\n",
    "\n",
    "###m17\n",
    "sample_sku3 = feat_nunique(sample_sku3,sku3,['sku_id'],'comment_create_tm','f3_commentnums_')\n",
    "###m18-20\n",
    "sample_sku3['f3_commentnums_in_7days_'] = sample_sku3['f3_commentnums_comment_create_tm_sku_id_nunique']/12\n",
    "sample_sku3['f3_commentnums_in_15days_'] = sample_sku3['f3_commentnums_comment_create_tm_sku_id_nunique']/6\n",
    "sample_sku3['f3_commentnums_in_31days_'] = sample_sku3['f3_commentnums_comment_create_tm_sku_id_nunique']/3\n",
    "\n",
    "###m21\n",
    "t1 = sample_sku3[['sku_id','f3_buytimes_o_id_sku_id_nunique']]\n",
    "t2 = sku3.groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3.groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_buy_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m22\n",
    "t1 = sample_sku3[['sku_id','f3_scannums_a_num_sku_id_sum']]\n",
    "t2 = sku3[sku3.a_type==1].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[sku3.a_type==1].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_scan_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "sample_sku3\n",
    "\n",
    "###m23\n",
    "t1 = sample_sku3[['sku_id','f3_attnums_a_num_sku_id_sum']]\n",
    "t2 = sku3[sku3.a_type==2].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3[sku3.a_type==2].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_att_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m24\n",
    "t1 = sample_sku3[['sku_id','f3_commentnums_comment_create_tm_sku_id_nunique']]\n",
    "t2 = sku3.groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku3.groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_comment_day_gap']]\n",
    "sample_sku3 = sample_sku3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m25-m32还没写，正在考虑\n",
    "sku3_a_o_date = sku3[((pd.to_datetime(sku3.a_date)<=pd.to_datetime(sku3.o_date.fillna('2018-01-01'))))]\n",
    "sku3_a_o_date = sku3_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "sku3_a_o_date = sku3_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "sku3_a_o_date['o_a_gap']=(pd.to_datetime(sku3_a_o_date.o_date)-pd.to_datetime(sku3_a_o_date.a_date)).dt.days \n",
    "FLAG='sku3_a_o_'\n",
    "sample_sku3 = feat_count(sample_sku3,sku3_a_o_date[sku3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_sku3 = feat_mean(sample_sku3,sku3_a_o_date[sku3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###m33\n",
    "t1 = sample_sku3[['sku_id','f3_buynums_o_sku_num_sku_id_sum']] \n",
    "t2 = sku3[['sku_id','price']]\n",
    "t3 = t1.merge(t2,on=['sku_id'],how='left')\n",
    "t3['f3_total_price'] = t3['f3_buynums_o_sku_num_sku_id_sum']*t3['price']\n",
    "t3 = t3[['sku_id','f3_total_price']]\n",
    "sample_sku3 = sample_sku3.merge(t3,on=['sku_id'],how='left')\n",
    "\n",
    "###m34-m36\n",
    "sample_sku3['f3_total_price_in_7days_'] = sample_sku3['f3_total_price']/12\n",
    "sample_sku3['f3_total_price_in_15days_'] = sample_sku3['f3_total_price']/6\n",
    "sample_sku3['f3_total_price_in_31days_'] = sample_sku3['f3_total_price']/3\n",
    "\n",
    "###m37-m40\n",
    "sample_sku3 = feat_min(sample_sku3,sku3,['sku_id'],'o_day','f3_buytime_early_')\n",
    "sample_sku3 = feat_min(sample_sku3,sku3[sku3.a_type==1],['sku_id'],'a_day','f3_scantime_early_')\n",
    "sample_sku3 = feat_min(sample_sku3,sku3[sku3.a_type==2],['sku_id'],'a_day','f3_atttime_early_')\n",
    "sample_sku3 = feat_min(sample_sku3,sku3,['sku_id'],'comment_day','f3_commenttime_early_')\n",
    "\n",
    "###m41-m44\n",
    "sample_sku3 = feat_max(sample_sku3,sku3,['sku_id'],'o_day','f3_buytime_late_')\n",
    "sample_sku3 = feat_max(sample_sku3,sku3[sku3.a_type==1],['sku_id'],'a_day','f3_scantime_late_')\n",
    "sample_sku3 = feat_max(sample_sku3,sku3[sku3.a_type==2],['sku_id'],'a_day','f3_atttime_late_')\n",
    "sample_sku3 = feat_max(sample_sku3,sku3,['sku_id'],'comment_day','f3_commenttime_late_')\n",
    "\n",
    "###\n",
    "sample_sku3 = sample_sku3.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "ename": "MemoryError",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mMemoryError\u001b[0m                               Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-17-d68b34805b40>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     50\u001b[0m \u001b[0mt\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f2_average_buy_day_gap'\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mt\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'o_date_max'\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m-\u001b[0m \u001b[0mt\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'o_date_min'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdays\u001b[0m\u001b[1;33m/\u001b[0m\u001b[0mt\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'f2_buytimes_o_id_sku_id_nunique'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     51\u001b[0m \u001b[0mt\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mt\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'sku_id'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;34m'f2_average_buy_day_gap'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 52\u001b[1;33m \u001b[0msample_sku2\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msample_sku2\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmerge\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mt\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mon\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'sku_id'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mhow\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'left'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     53\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     54\u001b[0m \u001b[1;31m###m22\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36mmerge\u001b[1;34m(self, right, how, on, left_on, right_on, left_index, right_index, sort, suffixes, copy, indicator, validate)\u001b[0m\n\u001b[0;32m   5368\u001b[0m                      \u001b[0mright_on\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mright_on\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mleft_index\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mleft_index\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   5369\u001b[0m                      \u001b[0mright_index\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mright_index\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msort\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0msort\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msuffixes\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0msuffixes\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 5370\u001b[1;33m                      copy=copy, indicator=indicator, validate=validate)\n\u001b[0m\u001b[0;32m   5371\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   5372\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mround\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdecimals\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\reshape\\merge.py\u001b[0m in \u001b[0;36mmerge\u001b[1;34m(left, right, how, on, left_on, right_on, left_index, right_index, sort, suffixes, copy, indicator, validate)\u001b[0m\n\u001b[0;32m     56\u001b[0m                          \u001b[0mcopy\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mcopy\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mindicator\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mindicator\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     57\u001b[0m                          validate=validate)\n\u001b[1;32m---> 58\u001b[1;33m     \u001b[1;32mreturn\u001b[0m \u001b[0mop\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_result\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     59\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     60\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\reshape\\merge.py\u001b[0m in \u001b[0;36mget_result\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    580\u001b[0m                 self.left, self.right)\n\u001b[0;32m    581\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 582\u001b[1;33m         \u001b[0mjoin_index\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mleft_indexer\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mright_indexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_join_info\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    583\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    584\u001b[0m         \u001b[0mldata\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mrdata\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mleft\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_data\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mright\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_data\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\reshape\\merge.py\u001b[0m in \u001b[0;36m_get_join_info\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    746\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    747\u001b[0m             (left_indexer,\n\u001b[1;32m--> 748\u001b[1;33m              right_indexer) = self._get_join_indexers()\n\u001b[0m\u001b[0;32m    749\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    750\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mright_index\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\reshape\\merge.py\u001b[0m in \u001b[0;36m_get_join_indexers\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    725\u001b[0m                                   \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mright_join_keys\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    726\u001b[0m                                   \u001b[0msort\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msort\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 727\u001b[1;33m                                   how=self.how)\n\u001b[0m\u001b[0;32m    728\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    729\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_get_join_info\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\Software\\Anaconda3\\envs\\tensorflow\\lib\\site-packages\\pandas\\core\\reshape\\merge.py\u001b[0m in \u001b[0;36m_get_join_indexers\u001b[1;34m(left_keys, right_keys, sort, how, **kwargs)\u001b[0m\n\u001b[0;32m   1064\u001b[0m     \u001b[0mjoin_func\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m_join_functions\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mhow\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1065\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1066\u001b[1;33m     \u001b[1;32mreturn\u001b[0m \u001b[0mjoin_func\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlkey\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mrkey\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcount\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1067\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1068\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas/_libs/join.pyx\u001b[0m in \u001b[0;36mpandas._libs.join.left_outer_join\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mMemoryError\u001b[0m: "
     ]
    }
   ],
   "source": [
    "####商品特征\n",
    "#######dataset2_sku\n",
    "#######\n",
    "####for dataset2\n",
    "sku2 = feature2[['sku_id','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','price','cate','para_1','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = sku2[['sku_id','price','cate','a_date','o_date']]\n",
    "t = sku2[(sku2.cate==30)|(sku2.cate==101)]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###m1\n",
    "sample_sku2 = feat_nunique(t,sku2,['sku_id'],'o_id','f2_buytimes_')\n",
    "###m2-m4\n",
    "sample_sku2['f2_buytimes_in_7days_'] = sample_sku2['f2_buytimes_o_id_sku_id_nunique']/12\n",
    "sample_sku2['f2_buytimes_in_15days_'] = sample_sku2['f2_buytimes_o_id_sku_id_nunique']/6\n",
    "sample_sku2['f2_buytimes_in_31days_'] = sample_sku2['f2_buytimes_o_id_sku_id_nunique']/3\n",
    "\n",
    "###m5\n",
    "sample_sku2 = feat_sum(sample_sku2,sku2,['sku_id'],'o_sku_num','f2_buynums_')\n",
    "###m6-m8\n",
    "sample_sku2['f2_buynums_in_7days_'] = sample_sku2['f2_buynums_o_sku_num_sku_id_sum']/12\n",
    "sample_sku2['f2_buynums_in_15days_'] = sample_sku2['f2_buynums_o_sku_num_sku_id_sum']/6\n",
    "sample_sku2['f2_buynums_in_31days_'] = sample_sku2['f2_buynums_o_sku_num_sku_id_sum']/3\n",
    "\n",
    "###m9\n",
    "sample_sku2 = feat_sum(sample_sku2,sku2[sku2.a_type==1],['sku_id'],'a_num','f2_scannums_')\n",
    "###m10-m12\n",
    "sample_sku2['f2_scannums_in_7days_'] = sample_sku2['f2_scannums_a_num_sku_id_sum']/12\n",
    "sample_sku2['f2_scannums_in_15days_'] = sample_sku2['f2_scannums_a_num_sku_id_sum']/6\n",
    "sample_sku2['f2_scannums_in_31days_'] = sample_sku2['f2_scannums_a_num_sku_id_sum']/3\n",
    "\n",
    "###m13\n",
    "sample_sku2 = feat_sum(sample_sku2,sku2[sku2.a_type==2],['sku_id'],'a_num','f2_attnums_')\n",
    "###m14-m16\n",
    "sample_sku2['f2_attnums_in_7days_'] = sample_sku2['f2_attnums_a_num_sku_id_sum']/12\n",
    "sample_sku2['f2_attnums_in_15days_'] = sample_sku2['f2_attnums_a_num_sku_id_sum']/6\n",
    "sample_sku2['f2_attnums_in_31days_'] = sample_sku2['f2_attnums_a_num_sku_id_sum']/3\n",
    "\n",
    "###m17\n",
    "sample_sku2 = feat_nunique(sample_sku2,sku2,['sku_id'],'comment_create_tm','f2_commentnums_')\n",
    "###m18-20\n",
    "sample_sku2['f2_commentnums_in_7days_'] = sample_sku2['f2_commentnums_comment_create_tm_sku_id_nunique']/12\n",
    "sample_sku2['f2_commentnums_in_15days_'] = sample_sku2['f2_commentnums_comment_create_tm_sku_id_nunique']/6\n",
    "sample_sku2['f2_commentnums_in_31days_'] = sample_sku2['f2_commentnums_comment_create_tm_sku_id_nunique']/3\n",
    "\n",
    "###m21\n",
    "t1 = sample_sku2[['sku_id','f2_buytimes_o_id_sku_id_nunique']]\n",
    "t2 = sku2.groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku2.groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f2_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f2_buytimes_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f2_average_buy_day_gap']]\n",
    "sample_sku2 = sample_sku2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m22\n",
    "t1 = sample_sku2[['sku_id','f2_scannums_a_num_sku_id_sum']]\n",
    "t2 = sku2[sku2.a_type==1].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku2[sku2.a_type==1].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f2_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f2_scannums_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f2_average_scan_day_gap']]\n",
    "sample_sku2 = sample_sku2.merge(t,on=['sku_id'],how='left')\n",
    "sample_sku2\n",
    "\n",
    "###m23\n",
    "t1 = sample_sku2[['sku_id','f2_attnums_a_num_sku_id_sum']]\n",
    "t2 = sku2[sku2.a_type==2].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku2[sku2.a_type==2].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f2_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f2_attnums_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f2_average_att_day_gap']]\n",
    "sample_sku2 = sample_sku2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m24\n",
    "t1 = sample_sku2[['sku_id','f2_commentnums_comment_create_tm_sku_id_nunique']]\n",
    "t2 = sku2.groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku2.groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f2_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f2_commentnums_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f2_average_comment_day_gap']]\n",
    "sample_sku2 = sample_sku2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m25-m32还没写，正在考虑\n",
    "\n",
    "sku2_a_o_date = sku2[((pd.to_datetime(sku2.a_date)<=pd.to_datetime(sku2.o_date.fillna('2018-01-01'))))]\n",
    "sku2_a_o_date = sku2_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "sku2_a_o_date = sku2_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "sku2_a_o_date['o_a_gap']=(pd.to_datetime(sku2_a_o_date.o_date)-pd.to_datetime(sku2_a_o_date.a_date)).dt.days \n",
    "FLAG='sku2_a_o_'\n",
    "sample_sku2 = feat_count(sample_sku2,sku2_a_o_date[sku2_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_sku2 = feat_mean(sample_sku2,sku2_a_o_date[sku2_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###m33\n",
    "t1 = sample_sku2[['sku_id','f2_buynums_o_sku_num_sku_id_sum']] \n",
    "t2 = sku2[['sku_id','price']]\n",
    "t3 = t1.merge(t2,on=['sku_id'],how='left')\n",
    "t3['f2_total_price'] = t3['f2_buynums_o_sku_num_sku_id_sum']*t3['price']\n",
    "t3 = t3[['sku_id','f2_total_price']]\n",
    "sample_sku2 = sample_sku2.merge(t3,on=['sku_id'],how='left')\n",
    "\n",
    "###m34-m36\n",
    "sample_sku2['f2_total_price_in_7days_'] = sample_sku2['f2_total_price']/12\n",
    "sample_sku2['f2_total_price_in_15days_'] = sample_sku2['f2_total_price']/6\n",
    "sample_sku2['f2_total_price_in_31days_'] = sample_sku2['f2_total_price']/3\n",
    "\n",
    "###m27-m40\n",
    "sample_sku2 = feat_min(sample_sku2,sku2,['sku_id'],'o_day','f2_buytime_early_')\n",
    "sample_sku2 = feat_min(sample_sku2,sku2[sku2.a_type==1],['sku_id'],'a_day','f2_scantime_early_')\n",
    "sample_sku2 = feat_min(sample_sku2,sku2[sku2.a_type==2],['sku_id'],'a_day','f2_atttime_early_')\n",
    "sample_sku2 = feat_min(sample_sku2,sku2,['sku_id'],'comment_day','f2_commenttime_early_')\n",
    "\n",
    "###m41-m44\n",
    "sample_sku2 = feat_max(sample_sku2,sku2,['sku_id'],'o_day','f2_buytime_late_')\n",
    "sample_sku2 = feat_max(sample_sku2,sku2[sku2.a_type==1],['sku_id'],'a_day','f2_scantime_late_')\n",
    "sample_sku2 = feat_max(sample_sku2,sku2[sku2.a_type==2],['sku_id'],'a_day','f2_atttime_late_')\n",
    "sample_sku2 = feat_max(sample_sku2,sku2,['sku_id'],'comment_day','f2_commenttime_late_')\n",
    "\n",
    "\n",
    "sample_sku2 = sample_sku2.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "142"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "####商品特征\n",
    "#######dataset1_sku\n",
    "#######\n",
    "####for dataset1\n",
    "sku1 = feature1[['sku_id','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','price','cate','para_1','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = sku1[['sku_id','price','cate']]\n",
    "t = sku1[(sku1.cate==30)|(sku1.cate==101)]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###m1\n",
    "sample_sku1 = feat_nunique(t,sku1,['sku_id'],'o_id','f1_buytimes_')\n",
    "###m2-m4\n",
    "sample_sku1['f1_buytimes_in_7days_'] = sample_sku1['f1_buytimes_o_id_sku_id_nunique']/12\n",
    "sample_sku1['f1_buytimes_in_15days_'] = sample_sku1['f1_buytimes_o_id_sku_id_nunique']/6\n",
    "sample_sku1['f1_buytimes_in_31days_'] = sample_sku1['f1_buytimes_o_id_sku_id_nunique']/3\n",
    "\n",
    "###m5\n",
    "sample_sku1 = feat_sum(sample_sku1,sku1,['sku_id'],'o_sku_num','f1_buynums_')\n",
    "###m6-m8\n",
    "sample_sku1['f1_buynums_in_7days_'] = sample_sku1['f1_buynums_o_sku_num_sku_id_sum']/12\n",
    "sample_sku1['f1_buynums_in_15days_'] = sample_sku1['f1_buynums_o_sku_num_sku_id_sum']/6\n",
    "sample_sku1['f1_buynums_in_31days_'] = sample_sku1['f1_buynums_o_sku_num_sku_id_sum']/3\n",
    "\n",
    "###m9\n",
    "sample_sku1 = feat_sum(sample_sku1,sku1[sku1.a_type==1],['sku_id'],'a_num','f1_scannums_')\n",
    "###m10-m12\n",
    "sample_sku1['f1_scannums_in_7days_'] = sample_sku1['f1_scannums_a_num_sku_id_sum']/12\n",
    "sample_sku1['f1_scannums_in_15days_'] = sample_sku1['f1_scannums_a_num_sku_id_sum']/6\n",
    "sample_sku1['f1_scannums_in_31days_'] = sample_sku1['f1_scannums_a_num_sku_id_sum']/3\n",
    "\n",
    "###m13\n",
    "sample_sku1 = feat_sum(sample_sku1,sku1[sku1.a_type==2],['sku_id'],'a_num','f1_attnums_')\n",
    "###m14-m16\n",
    "sample_sku1['f1_attnums_in_7days_'] = sample_sku1['f1_attnums_a_num_sku_id_sum']/12\n",
    "sample_sku1['f1_attnums_in_15days_'] = sample_sku1['f1_attnums_a_num_sku_id_sum']/6\n",
    "sample_sku1['f1_attnums_in_31days_'] = sample_sku1['f1_attnums_a_num_sku_id_sum']/3\n",
    "\n",
    "###m17\n",
    "sample_sku1 = feat_nunique(sample_sku1,sku1,['sku_id'],'comment_create_tm','f1_commentnums_')\n",
    "###m18-20\n",
    "sample_sku1['f1_commentnums_in_7days_'] = sample_sku1['f1_commentnums_comment_create_tm_sku_id_nunique']/12\n",
    "sample_sku1['f1_commentnums_in_15days_'] = sample_sku1['f1_commentnums_comment_create_tm_sku_id_nunique']/6\n",
    "sample_sku1['f1_commentnums_in_31days_'] = sample_sku1['f1_commentnums_comment_create_tm_sku_id_nunique']/3\n",
    "\n",
    "###m21\n",
    "t1 = sample_sku1[['sku_id','f1_buytimes_o_id_sku_id_nunique']]\n",
    "t2 = sku1.groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku1.groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f1_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f1_buytimes_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f1_average_buy_day_gap']]\n",
    "sample_sku1 = sample_sku1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m22\n",
    "t1 = sample_sku1[['sku_id','f1_scannums_a_num_sku_id_sum']]\n",
    "t2 = sku1[sku1.a_type==1].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku1[sku1.a_type==1].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f1_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f1_scannums_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f1_average_scan_day_gap']]\n",
    "sample_sku1 = sample_sku1.merge(t,on=['sku_id'],how='left')\n",
    "sample_sku1\n",
    "\n",
    "###m23\n",
    "t1 = sample_sku1[['sku_id','f1_attnums_a_num_sku_id_sum']]\n",
    "t2 = sku1[sku1.a_type==2].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku1[sku1.a_type==2].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f1_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f1_attnums_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f1_average_att_day_gap']]\n",
    "sample_sku1 = sample_sku1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m24\n",
    "t1 = sample_sku1[['sku_id','f1_commentnums_comment_create_tm_sku_id_nunique']]\n",
    "t2 = sku1.groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = sku1.groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f1_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f1_commentnums_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f1_average_comment_day_gap']]\n",
    "sample_sku1 = sample_sku1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m25-m32还没写，正在考虑\n",
    "sku1_a_o_date = sku1[((pd.to_datetime(sku1.a_date)<=pd.to_datetime(sku1.o_date.fillna('2018-01-01'))))]\n",
    "sku1_a_o_date = sku1_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "sku1_a_o_date = sku1_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "sku1_a_o_date['o_a_gap']=(pd.to_datetime(sku1_a_o_date.o_date)-pd.to_datetime(sku1_a_o_date.a_date)).dt.days \n",
    "FLAG='sku1_a_o_'\n",
    "sample_sku1 = feat_count(sample_sku1,sku1_a_o_date[sku1_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_sku1 = feat_mean(sample_sku1,sku1_a_o_date[sku1_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###m33\n",
    "t1 = sample_sku1[['sku_id','f1_buynums_o_sku_num_sku_id_sum']] \n",
    "t2 = sku1[['sku_id','price']]\n",
    "t3 = t1.merge(t2,on=['sku_id'],how='left')\n",
    "t3['f1_total_price'] = t3['f1_buynums_o_sku_num_sku_id_sum']*t3['price']\n",
    "t3 = t3[['sku_id','f1_total_price']]\n",
    "sample_sku1 = sample_sku1.merge(t3,on=['sku_id'],how='left')\n",
    "\n",
    "###m34-m36\n",
    "sample_sku1['f1_total_price_in_7days_'] = sample_sku1['f1_total_price']/12\n",
    "sample_sku1['f1_total_price_in_15days_'] = sample_sku1['f1_total_price']/6\n",
    "sample_sku1['f1_total_price_in_31days_'] = sample_sku1['f1_total_price']/3\n",
    "\n",
    "###m17-m40\n",
    "sample_sku1 = feat_min(sample_sku1,sku1,['sku_id'],'o_day','f1_buytime_early_')\n",
    "sample_sku1 = feat_min(sample_sku1,sku1[sku1.a_type==1],['sku_id'],'a_day','f1_scantime_early_')\n",
    "sample_sku1 = feat_min(sample_sku1,sku1[sku1.a_type==2],['sku_id'],'a_day','f1_atttime_early_')\n",
    "sample_sku1 = feat_min(sample_sku1,sku1,['sku_id'],'comment_day','f1_commenttime_early_')\n",
    "\n",
    "###m41-m44\n",
    "sample_sku1 = feat_max(sample_sku1,sku1,['sku_id'],'o_day','f1_buytime_late_')\n",
    "sample_sku1 = feat_max(sample_sku1,sku1[sku1.a_type==1],['sku_id'],'a_day','f1_scantime_late_')\n",
    "sample_sku1 = feat_max(sample_sku1,sku1[sku1.a_type==2],['sku_id'],'a_day','f1_atttime_late_')\n",
    "sample_sku1 = feat_max(sample_sku1,sku1,['sku_id'],'comment_day','f1_commenttime_late_')\n",
    "\n",
    "\n",
    "sample_sku1 = sample_sku1.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "133"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "###用户特征\n",
    "#######dataset3_user\n",
    "#######\n",
    "####for dataset3\n",
    "user3 = feature3[['user_id','sku_id','price','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','age','user_lv_cd','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = user3[['user_id','age','user_lv_cd','cate']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###u1\n",
    "sample_user3 = feat_nunique(t,user3,['user_id'],'o_id','f3_buytimes_')\n",
    "###u2-u4\n",
    "sample_user3['f3_buytimes_in_7days_'] = sample_user3['f3_buytimes_o_id_user_id_nunique']/12\n",
    "sample_user3['f3_buytimes_in_15days_'] = sample_user3['f3_buytimes_o_id_user_id_nunique']/6\n",
    "sample_user3['f3_buytimes_in_31days_'] = sample_user3['f3_buytimes_o_id_user_id_nunique']/3\n",
    "\n",
    "###u5\n",
    "sample_user3 = feat_sum(sample_user3,user3,['user_id'],'o_sku_num','f3_buynums_')\n",
    "###u6-u8\n",
    "sample_user3['f3_buynums_in_7days_'] = sample_user3['f3_buynums_o_sku_num_user_id_sum']/12\n",
    "sample_user3['f3_buynums_in_15days_'] = sample_user3['f3_buynums_o_sku_num_user_id_sum']/6\n",
    "sample_user3['f3_buynums_in_31days_'] = sample_user3['f3_buynums_o_sku_num_user_id_sum']/3\n",
    "\n",
    "###u9\n",
    "sample_user3 = feat_sum(sample_user3,user3[user3.a_type==1],['user_id'],'a_num','f3_scannums_')\n",
    "###u10-u12\n",
    "sample_user3['f3_scannums_in_7days_'] = sample_user3['f3_scannums_a_num_user_id_sum']/12\n",
    "sample_user3['f3_scannums_in_15days_'] = sample_user3['f3_scannums_a_num_user_id_sum']/6\n",
    "sample_user3['f3_scannums_in_31days_'] = sample_user3['f3_scannums_a_num_user_id_sum']/3\n",
    "\n",
    "###u13\n",
    "sample_user3 = feat_sum(sample_user3,user3[user3.a_type==2],['user_id'],'a_num','f3_attnums_')\n",
    "###u14-u16\n",
    "sample_user3['f3_attnums_in_7days_'] = sample_user3['f3_attnums_a_num_user_id_sum']/12\n",
    "sample_user3['f3_attnums_in_15days_'] = sample_user3['f3_attnums_a_num_user_id_sum']/6\n",
    "sample_user3['f3_attnums_in_31days_'] = sample_user3['f3_attnums_a_num_user_id_sum']/3\n",
    "\n",
    "###u17\n",
    "sample_user3 = feat_nunique(sample_user3,user3,['user_id'],'comment_create_tm','f3_commentnums_')\n",
    "###u18-20\n",
    "sample_user3['f3_commentnums_in_7days_'] = sample_user3['f3_commentnums_comment_create_tm_user_id_nunique']/12\n",
    "sample_user3['f3_commentnums_in_15days_'] = sample_user3['f3_commentnums_comment_create_tm_user_id_nunique']/6\n",
    "sample_user3['f3_commentnums_in_31days_'] = sample_user3['f3_commentnums_comment_create_tm_user_id_nunique']/3\n",
    "\n",
    "###u21\n",
    "t1 = sample_user3[['user_id','f3_buytimes_o_id_user_id_nunique']]\n",
    "t2 = user3.groupby(['user_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user3.groupby(['user_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_o_id_user_id_nunique']\n",
    "t = t[['user_id','f3_average_buy_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "###u22\n",
    "t1 = sample_user3[['user_id','f3_scannums_a_num_user_id_sum']]\n",
    "t2 = user3[user3.a_type==1].groupby(['user_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user3[user3.a_type==1].groupby(['user_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_a_num_user_id_sum']\n",
    "t = t[['user_id','f3_average_scan_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u23\n",
    "t1 = sample_user3[['user_id','f3_attnums_a_num_user_id_sum']]\n",
    "t2 = user3[user3.a_type==2].groupby(['user_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user3[user3.a_type==2].groupby(['user_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_a_num_user_id_sum']\n",
    "t = t[['user_id','f3_average_att_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u24\n",
    "t1 = sample_user3[['user_id','f3_commentnums_comment_create_tm_user_id_nunique']]\n",
    "t2 = user3.groupby(['user_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user3.groupby(['user_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_comment_create_tm_user_id_nunique']\n",
    "t = t[['user_id','f3_average_comment_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u25-u32还没写，正在考虑\n",
    "user3_a_o_date = user3[((pd.to_datetime(user3.a_date)<=pd.to_datetime(user3.o_date.fillna('2018-01-01'))))]\n",
    "user3_a_o_date = user3_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user3_a_o_date = user3_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user3_a_o_date['o_a_gap']=(pd.to_datetime(user3_a_o_date.o_date)-pd.to_datetime(user3_a_o_date.a_date)).dt.days \n",
    "FLAG='user3_a_o_'\n",
    "sample_user3 = feat_count(sample_user3,user3_a_o_date[user3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user3 = feat_mean(sample_user3,user3_a_o_date[user3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "\n",
    "\n",
    "###u33\n",
    "t1 = user3[['user_id','o_sku_num','price']] \n",
    "t1['sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user3 = feat_sum(sample_user3,t1,['user_id'],'sku_total_price','f3_user_total_consumption_')\n",
    "\n",
    "###u34-u36\n",
    "sample_user3['f3_total_consumption_in_7days_'] = sample_user3['f3_user_total_consumption_sku_total_price_user_id_sum']/12\n",
    "sample_user3['f3_total_consumption_in_15days_'] = sample_user3['f3_user_total_consumption_sku_total_price_user_id_sum']/6\n",
    "sample_user3['f3_total_consumption_in_31days_'] = sample_user3['f3_user_total_consumption_sku_total_price_user_id_sum']/3\n",
    "\n",
    "\n",
    "###u37-u40\n",
    "sample_user3 = feat_min(sample_user3,user3,['user_id'],'o_day','f3_buytime_early_')\n",
    "sample_user3 = feat_min(sample_user3,user3[user3.a_type==1],['user_id'],'a_day','f3_scantime_early_')\n",
    "sample_user3 = feat_min(sample_user3,user3[user3.a_type==2],['user_id'],'a_day','f3_atttime_early_')\n",
    "sample_user3 = feat_min(sample_user3,user3,['user_id'],'comment_day','f3_commenttime_early_')\n",
    "\n",
    "###u41-u44\n",
    "sample_user3 = feat_max(sample_user3,user3,['user_id'],'o_day','f3_buytime_late_')\n",
    "sample_user3 = feat_max(sample_user3,user3[user3.a_type==1],['user_id'],'a_day','f3_scantime_late_')\n",
    "sample_user3 = feat_max(sample_user3,user3[user3.a_type==2],['user_id'],'a_day','f3_atttime_late_')\n",
    "sample_user3 = feat_max(sample_user3,user3,['user_id'],'comment_day','f3_commenttime_late_')\n",
    "\n",
    "\n",
    "sample_user3 = sample_user3.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "user3 = feature3[['user_id','sku_id','price','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','age','user_lv_cd','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = user3[['user_id','age','user_lv_cd','cate']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###m1\n",
    "sample_user3 = feat_nunique(t,user3[(user3.cate==101)|(user3.cate==30)],['sku_id'],'o_id','f3_buytimes_30101_')\n",
    "sample_user3 = feat_nunique(sample_user3,user3[(user3.cate==30)],['sku_id'],'o_id','f3_buytimes_30_')\n",
    "sample_user3 = feat_nunique(sample_user3,user3[(user3.cate==101)],['sku_id'],'o_id','f3_buytimes_101_')\n",
    "sample_user3 = feat_nunique(sample_user3,user3[(user3.cate!=101)&(user3.cate!=30)],['sku_id'],'o_id','f3_buytimes_no_30101_')\n",
    "\n",
    "###m2-m4\n",
    "sample_user3['f3_buytimes_30101_in_7days_'] = sample_user3['f3_buytimes_30101_o_id_sku_id_nunique']/12\n",
    "sample_user3['f3_buytimes_30_in_7days_'] = sample_user3['f3_buytimes_30_o_id_sku_id_nunique']/12\n",
    "sample_user3['f3_buytimes_101_in_7days_'] = sample_user3['f3_buytimes_101_o_id_sku_id_nunique']/12\n",
    "sample_user3['f3_buytimes_no_30101_in_7days_'] = sample_user3['f3_buytimes_no_30101_o_id_sku_id_nunique']/12\n",
    "\n",
    "sample_user3['f3_buytimes_30101_in_15days_'] = sample_user3['f3_buytimes_30101_o_id_sku_id_nunique']/6\n",
    "sample_user3['f3_buytimes_30_in_15days_'] = sample_user3['f3_buytimes_30_o_id_sku_id_nunique']/6\n",
    "sample_user3['f3_buytimes_101_in_15days_'] = sample_user3['f3_buytimes_101_o_id_sku_id_nunique']/6\n",
    "sample_user3['f3_buytimes_no_30101_in_15days_'] = sample_user3['f3_buytimes_no_30101_o_id_sku_id_nunique']/6\n",
    "\n",
    "sample_user3['f3_buytimes_30101_in_31days_'] = sample_user3['f3_buytimes_30101_o_id_sku_id_nunique']/3\n",
    "sample_user3['f3_buytimes_30_in_31days_'] = sample_user3['f3_buytimes_30_o_id_sku_id_nunique']/3\n",
    "sample_user3['f3_buytimes_101_in_31days_'] = sample_user3['f3_buytimes_101_o_id_sku_id_nunique']/3\n",
    "sample_user3['f3_buytimes_no_30101_in_31days_'] = sample_user3['f3_buytimes_no_30101_o_id_sku_id_nunique']/3\n",
    "\n",
    "###m5\n",
    "sample_user3 = feat_sum(sample_user3,user3[(user3.cate==101)|(user3.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30101_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[(user3.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[(user3.cate==101)],['sku_id'],'o_sku_num','f3_buynums_101_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[(user3.cate!=101)&(user3.cate!=30)],['sku_id'],'o_sku_num','f3_buynums_no_30101_')\n",
    "\n",
    "###m6-m8\n",
    "sample_user3['f3_buynums_30101_in_7days_'] = sample_user3['f3_buynums_30101_o_sku_num_sku_id_sum']/12\n",
    "sample_user3['f3_buynums_30_in_7days_'] = sample_user3['f3_buynums_30_o_sku_num_sku_id_sum']/12\n",
    "sample_user3['f3_buynums_101_in_7days_'] = sample_user3['f3_buynums_101_o_sku_num_sku_id_sum']/12\n",
    "sample_user3['f3_buynums_no_30101_in_7days_'] = sample_user3['f3_buynums_no_30101_o_sku_num_sku_id_sum']/12\n",
    "\n",
    "sample_user3['f3_buynums_30101_in_15days_'] = sample_user3['f3_buynums_30101_o_sku_num_sku_id_sum']/6\n",
    "sample_user3['f3_buynums_30_in_15days_'] = sample_user3['f3_buynums_30_o_sku_num_sku_id_sum']/6\n",
    "sample_user3['f3_buynums_101_in_15days_'] = sample_user3['f3_buynums_101_o_sku_num_sku_id_sum']/6\n",
    "sample_user3['f3_buynums_no_30101_in_15days_'] = sample_user3['f3_buynums_no_30101_o_sku_num_sku_id_sum']/6\n",
    "\n",
    "sample_user3['f3_buynums_30101_in_31days_'] = sample_user3['f3_buynums_30101_o_sku_num_sku_id_sum']/3\n",
    "sample_user3['f3_buynums_30_in_31days_'] = sample_user3['f3_buynums_30_o_sku_num_sku_id_sum']/3\n",
    "sample_user3['f3_buynums_101_in_31days_'] = sample_user3['f3_buynums_101_o_sku_num_sku_id_sum']/3\n",
    "sample_user3['f3_buynums_no_30101_in_31days_'] = sample_user3['f3_buynums_no_30101_o_sku_num_sku_id_sum']/3\n",
    "\n",
    "###m9\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate==101)|(user3.cate==30))&(user3.a_type==1)],['sku_id'],'a_num','f3_scannums_30101_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate==30))&(user3.a_type==1)],['sku_id'],'a_num','f3_scannums_30_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate==101))&(user3.a_type==1)],['sku_id'],'a_num','f3_scannums_101_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate!=101)&(user3.cate!=30))&(user3.a_type==1)],['sku_id'],'a_num','f3_scannums_no_30101_')\n",
    "###m10-m12\n",
    "sample_user3['f3_scannums_30101_in_7days_'] = sample_user3['f3_scannums_30101_a_num_sku_id_sum']/12\n",
    "sample_user3['f3_scannums_30_in_7days_'] = sample_user3['f3_scannums_30_a_num_sku_id_sum']/12\n",
    "sample_user3['f3_scannums_101_in_7days_'] = sample_user3['f3_scannums_101_a_num_sku_id_sum']/12\n",
    "sample_user3['f3_scannums_no_30101_in_7days_'] = sample_user3['f3_scannums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_user3['f3_scannums_30101_in_15days_'] = sample_user3['f3_scannums_30101_a_num_sku_id_sum']/6\n",
    "sample_user3['f3_scannums_30_in_15days_'] = sample_user3['f3_scannums_30_a_num_sku_id_sum']/6\n",
    "sample_user3['f3_scannums_101_in_15days_'] = sample_user3['f3_scannums_101_a_num_sku_id_sum']/6\n",
    "sample_user3['f3_scannums_no_30101_in_15days_'] = sample_user3['f3_scannums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_user3['f3_scannums_30101_in_31days_'] = sample_user3['f3_scannums_30101_a_num_sku_id_sum']/3\n",
    "sample_user3['f3_scannums_30_in_31days_'] = sample_user3['f3_scannums_30_a_num_sku_id_sum']/3\n",
    "sample_user3['f3_scannums_101_in_31days_'] = sample_user3['f3_scannums_101_a_num_sku_id_sum']/3\n",
    "sample_user3['f3_scannums_no_30101_in_31days_'] = sample_user3['f3_scannums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "###m13\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate==101)|(user3.cate==30))&(user3.a_type==2)],['sku_id'],'a_num','f3_attnums_30101_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate==30))&(user3.a_type==2)],['sku_id'],'a_num','f3_attnums_30_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate==101))&(user3.a_type==2)],['sku_id'],'a_num','f3_attnums_101_')\n",
    "sample_user3 = feat_sum(sample_user3,user3[((user3.cate!=101)&(user3.cate!=30))&(user3.a_type==2)],['sku_id'],'a_num','f3_attnums_no_30101_')\n",
    "###m14-m16\n",
    "sample_user3['f3_attnums_30101_in_7days_'] = sample_user3['f3_attnums_30101_a_num_sku_id_sum']/12\n",
    "sample_user3['f3_attnums_30_in_7days_'] = sample_user3['f3_attnums_30_a_num_sku_id_sum']/12\n",
    "sample_user3['f3_attnums_101_in_7days_'] = sample_user3['f3_attnums_101_a_num_sku_id_sum']/12\n",
    "sample_user3['f3_attnums_no_30101_in_7days_'] = sample_user3['f3_attnums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_user3['f3_attnums_30101_in_15days_'] = sample_user3['f3_attnums_30101_a_num_sku_id_sum']/6\n",
    "sample_user3['f3_attnums_30_in_15days_'] = sample_user3['f3_attnums_30_a_num_sku_id_sum']/6\n",
    "sample_user3['f3_attnums_101_in_15days_'] = sample_user3['f3_attnums_101_a_num_sku_id_sum']/6\n",
    "sample_user3['f3_attnums_no_30101_in_15days_'] = sample_user3['f3_attnums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_user3['f3_attnums_30101_in_31days_'] = sample_user3['f3_attnums_30101_a_num_sku_id_sum']/3\n",
    "sample_user3['f3_attnums_30_in_31days_'] = sample_user3['f3_attnums_30_a_num_sku_id_sum']/3\n",
    "sample_user3['f3_attnums_101_in_31days_'] = sample_user3['f3_attnums_101_a_num_sku_id_sum']/3\n",
    "sample_user3['f3_attnums_no_30101_in_31days_'] = sample_user3['f3_attnums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "\n",
    "###m17\n",
    "sample_user3 = feat_nunique(sample_user3,user3[(user3.cate==101)|(user3.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30101_')\n",
    "sample_user3 = feat_nunique(sample_user3,user3[(user3.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30_')\n",
    "sample_user3 = feat_nunique(sample_user3,user3[(user3.cate==101)],['sku_id'],'comment_create_tm','f3_commentnums_101_')\n",
    "sample_user3 = feat_nunique(sample_user3,user3[(user3.cate!=101)&(user3.cate!=30)],['sku_id'],'comment_create_tm','f3_commentnums_no_30101_')\n",
    "\n",
    "###m18-20\n",
    "sample_user3['f3_commentnums_30101_in_7days_'] = sample_user3['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user3['f3_commentnums_30_in_7days_'] = sample_user3['f3_commentnums_30_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user3['f3_commentnums_101_in_7days_'] = sample_user3['f3_commentnums_101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user3['f3_commentnums_no_30101_in_7days_'] = sample_user3['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/12\n",
    "\n",
    "sample_user3['f3_commentnums_30101_in_15days_'] = sample_user3['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user3['f3_commentnums_30_in_15days_'] = sample_user3['f3_commentnums_30_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user3['f3_commentnums_101_in_15days_'] = sample_user3['f3_commentnums_101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user3['f3_commentnums_no_30101_in_15days_'] = sample_user3['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/6\n",
    "\n",
    "sample_user3['f3_commentnums_30101_in_31days_'] = sample_user3['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user3['f3_commentnums_30_in_31days_'] = sample_user3['f3_commentnums_30_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user3['f3_commentnums_101_in_31days_'] = sample_user3['f3_commentnums_101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user3['f3_commentnums_no_30101_in_31days_'] = sample_user3['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/3\n",
    "\n",
    "###m21\n",
    "t1 = sample_user3[['sku_id','f3_buytimes_30101_o_id_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate==101)|(user3.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate==101)|(user3.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_buy_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_buytimes_30_o_id_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_buy_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_buytimes_101_o_id_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate==101)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate==101)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_buy_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_buytimes_no_30101_o_id_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate!=101)&(user3.cate!=30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate!=101)&(user3.cate!=30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_no_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_buy_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m22\n",
    "t1 = sample_user3[['sku_id','f3_scannums_30101_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==1)&((user3.cate==101)|(user3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==1)&((user3.cate==101)|(user3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_scan_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_scannums_30_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==1)&((user3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==1)&((user3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_scan_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_scannums_101_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==1)&((user3.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==1)&((user3.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_scan_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_scannums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==1)&((user3.cate!=30)&(user3.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==1)&((user3.cate!=30)&(user3.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_scan_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m23\n",
    "t1 = sample_user3[['sku_id','f3_attnums_30101_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==2)&((user3.cate==101)|(user3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==2)&((user3.cate==101)|(user3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_att_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_attnums_30_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==2)&((user3.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==2)&((user3.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_att_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_attnums_101_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==2)&((user3.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==2)&((user3.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_att_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user3[['sku_id','f3_attnums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = user3[(user3.a_type==2)&((user3.cate!=30)&(user3.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.a_type==2)&((user3.cate!=30)&(user3.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_att_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m24\n",
    "t1 = sample_user3[['sku_id','f3_commentnums_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate==101)|(user3.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate==101)|(user3.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_comment_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "t1 = sample_user3[['sku_id','f3_commentnums_30_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_comment_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_user3[['sku_id','f3_commentnums_101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate==101)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate==101)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_comment_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_user3[['sku_id','f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user3[(user3.cate!=101)&(user3.cate!=30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user3[(user3.cate!=101)&(user3.cate!=30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_comment_day_gap']]\n",
    "sample_user3 = sample_user3.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m25-m32还没写，正在考虑\n",
    "user3_a_o_date = user3[((pd.to_datetime(user3.a_date)<=pd.to_datetime(user3.o_date.fillna('2018-01-01'))))]\n",
    "user3_a_o_date = user3_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user3_a_o_date = user3_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user3_a_o_date['o_a_gap']=(pd.to_datetime(user3_a_o_date.o_date)-pd.to_datetime(user3_a_o_date.a_date)).dt.days \n",
    "FLAG='user3_a_o_'\n",
    "sample_user3 = feat_count(sample_user3,user3_a_o_date[user3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user3 = feat_mean(sample_user3,user3_a_o_date[(user3_a_o_date.o_a_gap>=-1)&((user3_a_o_date.cate==101)|(user3_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30101_')\n",
    "sample_user3 = feat_mean(sample_user3,user3_a_o_date[(user3_a_o_date.o_a_gap>=-1)&((user3_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30_')\n",
    "sample_user3 = feat_mean(sample_user3,user3_a_o_date[(user3_a_o_date.o_a_gap>=-1)&((user3_a_o_date.cate==101))],['sku_id'],'o_a_gap',FLAG+'101_')\n",
    "sample_user3 = feat_mean(sample_user3,user3_a_o_date[(user3_a_o_date.o_a_gap>=-1)&((user3_a_o_date.cate!=101)&(user3_a_o_date.cate!=30))],['sku_id'],'o_a_gap',FLAG+'no_30101_')\n",
    "\n",
    "\n",
    "###u33\n",
    "t1 = user3[['user_id','o_sku_num','price','cate']] \n",
    "t1['sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user3 = feat_sum(sample_user3,t1[(t1.cate==30)|(t1.cate==101)],['user_id'],'sku_total_price','f3_user_total_consumption_30101_')\n",
    "sample_user3 = feat_sum(sample_user3,t1[(t1.cate==30)],['user_id'],'sku_total_price','f3_user_total_consumption_30_')\n",
    "sample_user3 = feat_sum(sample_user3,t1[(t1.cate==101)],['user_id'],'sku_total_price','f3_user_total_consumption_101_')\n",
    "sample_user3 = feat_sum(sample_user3,t1[(t1.cate!=30)&(t1.cate!=101)],['user_id'],'sku_total_price','f3_user_total_consumption_no_30101_')\n",
    "\n",
    "###u34-u36\n",
    "sample_user3['f3_total_consumption_30101_in_7days_'] = sample_user3['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/12\n",
    "sample_user3['f3_total_consumption_30_in_7days_'] = sample_user3['f3_user_total_consumption_30_sku_total_price_user_id_sum']/12\n",
    "sample_user3['f3_total_consumption_101_in_7days_'] = sample_user3['f3_user_total_consumption_101_sku_total_price_user_id_sum']/12\n",
    "sample_user3['f3_total_consumption_no_30101_in_7days_'] = sample_user3['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/12\n",
    "\n",
    "sample_user3['f3_total_consumption_30101_in_15days_'] = sample_user3['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/6\n",
    "sample_user3['f3_total_consumption_30_in_15days_'] = sample_user3['f3_user_total_consumption_30_sku_total_price_user_id_sum']/6\n",
    "sample_user3['f3_total_consumption_101_in_15days_'] = sample_user3['f3_user_total_consumption_101_sku_total_price_user_id_sum']/6\n",
    "sample_user3['f3_total_consumption_no_30101_in_15days_'] = sample_user3['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/6\n",
    "\n",
    "sample_user3['f3_total_consumption_30101_in_31days_'] = sample_user3['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/3\n",
    "sample_user3['f3_total_consumption_30_in_31days_'] = sample_user3['f3_user_total_consumption_30_sku_total_price_user_id_sum']/3\n",
    "sample_user3['f3_total_consumption_101_in_31days_'] = sample_user3['f3_user_total_consumption_101_sku_total_price_user_id_sum']/3\n",
    "sample_user3['f3_total_consumption_no_30101_in_31days_'] = sample_user3['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/3\n",
    "\n",
    "\n",
    "###u37-u40\n",
    "sample_user3 = feat_min(sample_user3,user3[()],['user_id'],'o_day','f3_buytime_early_')\n",
    "sample_user3 = feat_min(sample_user3,user3[user3.a_type==1],['user_id'],'a_day','f3_scantime_early_')\n",
    "sample_user3 = feat_min(sample_user3,user3[user3.a_type==2],['user_id'],'a_day','f3_atttime_early_')\n",
    "sample_user3 = feat_min(sample_user3,user3,['user_id'],'comment_day','f3_commenttime_early_')\n",
    "\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==101)|(user3.cate==30)],['user_id'],'o_day','f3_buytime_early_30101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==30)],['user_id'],'o_day','f3_buytime_early_30_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==101)],['user_id'],'o_day','f3_buytime_early_101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate!=101)&(user3.cate!=30)],['user_id'],'o_day','f3_buytime_early_no_30101_')\n",
    "\n",
    "sample_user3 = feat_min(sample_user3,user3[((user3.cate==101)|(user3.cate==30))&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_early_30101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==30)&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_early_30_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==101)&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_early_101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[((user3.cate!=101)&(user3.cate!=30))&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_early_30101_')\n",
    "\n",
    "sample_user3 = feat_min(sample_user3,user3[((user3.cate==101)|(user3.cate==30))&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_early_30101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==30)&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_early_30_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==101)&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_early_101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[((user3.cate!=101)&(user3.cate!=30))&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_early_no_30101_')\n",
    "\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==101)|(user3.cate==30)],['user_id'],'comment_day','f3_commenttime_early_30101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==30)],['user_id'],'comment_day','f3_commenttime_early_30_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate==101)],['user_id'],'comment_day','f3_commenttime_early_101_')\n",
    "sample_user3 = feat_min(sample_user3,user3[(user3.cate!=101)&(user3.cate!=30)],['user_id'],'comment_day','f3_commenttime_early_no_30101_')\n",
    "\n",
    "\n",
    "###u41-u44\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==101)|(user3.cate==30)],['user_id'],'o_day','f3_buytime_late_30101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==30)],['user_id'],'o_day','f3_buytime_late_30_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==101)],['user_id'],'o_day','f3_buytime_late_101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate!=101)&(user3.cate!=30)],['user_id'],'o_day','f3_buytime_late_no_30101_')\n",
    "\n",
    "sample_user3 = feat_max(sample_user3,user3[((user3.cate==101)|(user3.cate==30))&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_late_30101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==30)&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_late_30_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==101)&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_late_101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[((user3.cate!=101)&(user3.cate!=30))&(user3.a_type==1)],['user_id'],'a_day','f3_scantime_late_30101_')\n",
    "\n",
    "sample_user3 = feat_max(sample_user3,user3[((user3.cate==101)|(user3.cate==30))&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_late_30101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==30)&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_late_30_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==101)&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_late_101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[((user3.cate!=101)&(user3.cate!=30))&(user3.a_type==2)],['user_id'],'a_day','f3_atttime_late_no_30101_')\n",
    "\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==101)|(user3.cate==30)],['user_id'],'comment_day','f3_commenttime_late_30101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==30)],['user_id'],'comment_day','f3_commenttime_late_30_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate==101)],['user_id'],'comment_day','f3_commenttime_late_101_')\n",
    "sample_user3 = feat_max(sample_user3,user3[(user3.cate!=101)&(user3.cate!=30)],['user_id'],'comment_day','f3_commenttime_late_no_30101_')\n",
    "\n",
    "\n",
    "sample_user3 = sample_user3.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "133"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "###用户特征\n",
    "#######dataset2_user\n",
    "#######\n",
    "####for dataset2\n",
    "user2 = feature2[['user_id','sku_id','price','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','age','user_lv_cd','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = user2[['user_id','age','user_lv_cd']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###u1\n",
    "sample_user2 = feat_nunique(t,user2,['user_id'],'o_id','f2_buytimes_')\n",
    "###u2-u4\n",
    "sample_user2['f2_buytimes_in_7days_'] = sample_user2['f2_buytimes_o_id_user_id_nunique']/12\n",
    "sample_user2['f2_buytimes_in_15days_'] = sample_user2['f2_buytimes_o_id_user_id_nunique']/6\n",
    "sample_user2['f2_buytimes_in_31days_'] = sample_user2['f2_buytimes_o_id_user_id_nunique']/3\n",
    "\n",
    "###u5\n",
    "sample_user2 = feat_sum(sample_user2,user2,['user_id'],'o_sku_num','f2_buynums_')\n",
    "###u6-u8\n",
    "sample_user2['f2_buynums_in_7days_'] = sample_user2['f2_buynums_o_sku_num_user_id_sum']/12\n",
    "sample_user2['f2_buynums_in_15days_'] = sample_user2['f2_buynums_o_sku_num_user_id_sum']/6\n",
    "sample_user2['f2_buynums_in_31days_'] = sample_user2['f2_buynums_o_sku_num_user_id_sum']/3\n",
    "\n",
    "###u9\n",
    "sample_user2 = feat_sum(sample_user2,user2[user2.a_type==1],['user_id'],'a_num','f2_scannums_')\n",
    "###u10-u12\n",
    "sample_user2['f2_scannums_in_7days_'] = sample_user2['f2_scannums_a_num_user_id_sum']/12\n",
    "sample_user2['f2_scannums_in_15days_'] = sample_user2['f2_scannums_a_num_user_id_sum']/6\n",
    "sample_user2['f2_scannums_in_31days_'] = sample_user2['f2_scannums_a_num_user_id_sum']/3\n",
    "\n",
    "###u13\n",
    "sample_user2 = feat_sum(sample_user2,user2[user2.a_type==2],['user_id'],'a_num','f2_attnums_')\n",
    "###u14-u16\n",
    "sample_user2['f2_attnums_in_7days_'] = sample_user2['f2_attnums_a_num_user_id_sum']/12\n",
    "sample_user2['f2_attnums_in_15days_'] = sample_user2['f2_attnums_a_num_user_id_sum']/6\n",
    "sample_user2['f2_attnums_in_31days_'] = sample_user2['f2_attnums_a_num_user_id_sum']/3\n",
    "\n",
    "###u17\n",
    "sample_user2 = feat_nunique(sample_user2,user2,['user_id'],'comment_create_tm','f2_commentnums_')\n",
    "###u18-20\n",
    "sample_user2['f2_commentnums_in_7days_'] = sample_user2['f2_commentnums_comment_create_tm_user_id_nunique']/12\n",
    "sample_user2['f2_commentnums_in_15days_'] = sample_user2['f2_commentnums_comment_create_tm_user_id_nunique']/6\n",
    "sample_user2['f2_commentnums_in_31days_'] = sample_user2['f2_commentnums_comment_create_tm_user_id_nunique']/3\n",
    "\n",
    "###u21\n",
    "t1 = sample_user2[['user_id','f2_buytimes_o_id_user_id_nunique']]\n",
    "t2 = user2.groupby(['user_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user2.groupby(['user_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f2_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f2_buytimes_o_id_user_id_nunique']\n",
    "t = t[['user_id','f2_average_buy_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "###u22\n",
    "t1 = sample_user2[['user_id','f2_scannums_a_num_user_id_sum']]\n",
    "t2 = user2[user2.a_type==1].groupby(['user_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user2[user2.a_type==1].groupby(['user_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f2_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f2_scannums_a_num_user_id_sum']\n",
    "t = t[['user_id','f2_average_scan_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u23\n",
    "t1 = sample_user2[['user_id','f2_attnums_a_num_user_id_sum']]\n",
    "t2 = user2[user2.a_type==2].groupby(['user_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user2[user2.a_type==2].groupby(['user_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f2_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f2_attnums_a_num_user_id_sum']\n",
    "t = t[['user_id','f2_average_att_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u24\n",
    "t1 = sample_user2[['user_id','f2_commentnums_comment_create_tm_user_id_nunique']]\n",
    "t2 = user2.groupby(['user_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user2.groupby(['user_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f2_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f2_commentnums_comment_create_tm_user_id_nunique']\n",
    "t = t[['user_id','f2_average_comment_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u25-u32还没写，正在考虑\n",
    "user2_a_o_date = user2[((pd.to_datetime(user2.a_date)<=pd.to_datetime(user2.o_date.fillna('2018-01-01'))))]\n",
    "user2_a_o_date = user2_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user2_a_o_date = user2_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user2_a_o_date['o_a_gap']=(pd.to_datetime(user2_a_o_date.o_date)-pd.to_datetime(user2_a_o_date.a_date)).dt.days \n",
    "FLAG='user2_a_o_'\n",
    "sample_user2 = feat_count(sample_user2,user2_a_o_date[user2_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user2 = feat_mean(sample_user2,user2_a_o_date[user2_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###u33\n",
    "t1 = user2[['user_id','o_sku_num','price']] \n",
    "t1['sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user2 = feat_sum(sample_user2,t1,['user_id'],'sku_total_price','f2_user_total_consumption_')\n",
    "\n",
    "###u34-u36\n",
    "sample_user2['f2_total_consumption_in_7days_'] = sample_user2['f2_user_total_consumption_sku_total_price_user_id_sum']/12\n",
    "sample_user2['f2_total_consumption_in_15days_'] = sample_user2['f2_user_total_consumption_sku_total_price_user_id_sum']/6\n",
    "sample_user2['f2_total_consumption_in_31days_'] = sample_user2['f2_user_total_consumption_sku_total_price_user_id_sum']/3\n",
    "\n",
    "\n",
    "###u37-u40\n",
    "sample_user2 = feat_min(sample_user2,user2,['user_id'],'o_day','f2_buytime_early_')\n",
    "sample_user2 = feat_min(sample_user2,user2[user2.a_type==1],['user_id'],'a_day','f2_scantime_early_')\n",
    "sample_user2 = feat_min(sample_user2,user2[user2.a_type==2],['user_id'],'a_day','f2_atttime_early_')\n",
    "sample_user2 = feat_min(sample_user2,user2,['user_id'],'comment_day','f2_commenttime_early_')\n",
    "\n",
    "###u41-u44\n",
    "sample_user2 = feat_max(sample_user2,user2,['user_id'],'o_day','f2_buytime_late_')\n",
    "sample_user2 = feat_max(sample_user2,user2[user2.a_type==1],['user_id'],'a_day','f2_scantime_late_')\n",
    "sample_user2 = feat_max(sample_user2,user2[user2.a_type==2],['user_id'],'a_day','f2_atttime_late_')\n",
    "sample_user2 = feat_max(sample_user2,user2,['user_id'],'comment_day','f2_commenttime_late_')\n",
    "\n",
    "sample_user2 = sample_user2.drop_duplicates()\n",
    "gc.collect()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "user2 = feature3[['user_id','sku_id','price','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','age','user_lv_cd','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = user2[['user_id','age','user_lv_cd','cate']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###m1\n",
    "sample_user2 = feat_nunique(t,user2[(user2.cate==101)|(user2.cate==30)],['sku_id'],'o_id','f3_buytimes_30101_')\n",
    "sample_user2 = feat_nunique(sample_user2,user2[(user2.cate==30)],['sku_id'],'o_id','f3_buytimes_30_')\n",
    "sample_user2 = feat_nunique(sample_user2,user2[(user2.cate==101)],['sku_id'],'o_id','f3_buytimes_101_')\n",
    "sample_user2 = feat_nunique(sample_user2,user2[(user2.cate!=101)&(user2.cate!=30)],['sku_id'],'o_id','f3_buytimes_no_30101_')\n",
    "\n",
    "###m2-m4\n",
    "sample_user2['f3_buytimes_30101_in_7days_'] = sample_user2['f3_buytimes_30101_o_id_sku_id_nunique']/12\n",
    "sample_user2['f3_buytimes_30_in_7days_'] = sample_user2['f3_buytimes_30_o_id_sku_id_nunique']/12\n",
    "sample_user2['f3_buytimes_101_in_7days_'] = sample_user2['f3_buytimes_101_o_id_sku_id_nunique']/12\n",
    "sample_user2['f3_buytimes_no_30101_in_7days_'] = sample_user2['f3_buytimes_no_30101_o_id_sku_id_nunique']/12\n",
    "\n",
    "sample_user2['f3_buytimes_30101_in_15days_'] = sample_user2['f3_buytimes_30101_o_id_sku_id_nunique']/6\n",
    "sample_user2['f3_buytimes_30_in_15days_'] = sample_user2['f3_buytimes_30_o_id_sku_id_nunique']/6\n",
    "sample_user2['f3_buytimes_101_in_15days_'] = sample_user2['f3_buytimes_101_o_id_sku_id_nunique']/6\n",
    "sample_user2['f3_buytimes_no_30101_in_15days_'] = sample_user2['f3_buytimes_no_30101_o_id_sku_id_nunique']/6\n",
    "\n",
    "sample_user2['f3_buytimes_30101_in_31days_'] = sample_user2['f3_buytimes_30101_o_id_sku_id_nunique']/3\n",
    "sample_user2['f3_buytimes_30_in_31days_'] = sample_user2['f3_buytimes_30_o_id_sku_id_nunique']/3\n",
    "sample_user2['f3_buytimes_101_in_31days_'] = sample_user2['f3_buytimes_101_o_id_sku_id_nunique']/3\n",
    "sample_user2['f3_buytimes_no_30101_in_31days_'] = sample_user2['f3_buytimes_no_30101_o_id_sku_id_nunique']/3\n",
    "\n",
    "###m5\n",
    "sample_user2 = feat_sum(sample_user2,user2[(user2.cate==101)|(user2.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30101_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[(user2.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[(user2.cate==101)],['sku_id'],'o_sku_num','f3_buynums_101_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[(user2.cate!=101)&(user2.cate!=30)],['sku_id'],'o_sku_num','f3_buynums_no_30101_')\n",
    "\n",
    "###m6-m8\n",
    "sample_user2['f3_buynums_30101_in_7days_'] = sample_user2['f3_buynums_30101_o_sku_num_sku_id_sum']/12\n",
    "sample_user2['f3_buynums_30_in_7days_'] = sample_user2['f3_buynums_30_o_sku_num_sku_id_sum']/12\n",
    "sample_user2['f3_buynums_101_in_7days_'] = sample_user2['f3_buynums_101_o_sku_num_sku_id_sum']/12\n",
    "sample_user2['f3_buynums_no_30101_in_7days_'] = sample_user2['f3_buynums_no_30101_o_sku_num_sku_id_sum']/12\n",
    "\n",
    "sample_user2['f3_buynums_30101_in_15days_'] = sample_user2['f3_buynums_30101_o_sku_num_sku_id_sum']/6\n",
    "sample_user2['f3_buynums_30_in_15days_'] = sample_user2['f3_buynums_30_o_sku_num_sku_id_sum']/6\n",
    "sample_user2['f3_buynums_101_in_15days_'] = sample_user2['f3_buynums_101_o_sku_num_sku_id_sum']/6\n",
    "sample_user2['f3_buynums_no_30101_in_15days_'] = sample_user2['f3_buynums_no_30101_o_sku_num_sku_id_sum']/6\n",
    "\n",
    "sample_user2['f3_buynums_30101_in_31days_'] = sample_user2['f3_buynums_30101_o_sku_num_sku_id_sum']/3\n",
    "sample_user2['f3_buynums_30_in_31days_'] = sample_user2['f3_buynums_30_o_sku_num_sku_id_sum']/3\n",
    "sample_user2['f3_buynums_101_in_31days_'] = sample_user2['f3_buynums_101_o_sku_num_sku_id_sum']/3\n",
    "sample_user2['f3_buynums_no_30101_in_31days_'] = sample_user2['f3_buynums_no_30101_o_sku_num_sku_id_sum']/3\n",
    "\n",
    "###m9\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate==101)|(user2.cate==30))&(user2.a_type==1)],['sku_id'],'a_num','f3_scannums_30101_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate==30))&(user2.a_type==1)],['sku_id'],'a_num','f3_scannums_30_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate==101))&(user2.a_type==1)],['sku_id'],'a_num','f3_scannums_101_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate!=101)&(user2.cate!=30))&(user2.a_type==1)],['sku_id'],'a_num','f3_scannums_no_30101_')\n",
    "###m10-m12\n",
    "sample_user2['f3_scannums_30101_in_7days_'] = sample_user2['f3_scannums_30101_a_num_sku_id_sum']/12\n",
    "sample_user2['f3_scannums_30_in_7days_'] = sample_user2['f3_scannums_30_a_num_sku_id_sum']/12\n",
    "sample_user2['f3_scannums_101_in_7days_'] = sample_user2['f3_scannums_101_a_num_sku_id_sum']/12\n",
    "sample_user2['f3_scannums_no_30101_in_7days_'] = sample_user2['f3_scannums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_user2['f3_scannums_30101_in_15days_'] = sample_user2['f3_scannums_30101_a_num_sku_id_sum']/6\n",
    "sample_user2['f3_scannums_30_in_15days_'] = sample_user2['f3_scannums_30_a_num_sku_id_sum']/6\n",
    "sample_user2['f3_scannums_101_in_15days_'] = sample_user2['f3_scannums_101_a_num_sku_id_sum']/6\n",
    "sample_user2['f3_scannums_no_30101_in_15days_'] = sample_user2['f3_scannums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_user2['f3_scannums_30101_in_31days_'] = sample_user2['f3_scannums_30101_a_num_sku_id_sum']/3\n",
    "sample_user2['f3_scannums_30_in_31days_'] = sample_user2['f3_scannums_30_a_num_sku_id_sum']/3\n",
    "sample_user2['f3_scannums_101_in_31days_'] = sample_user2['f3_scannums_101_a_num_sku_id_sum']/3\n",
    "sample_user2['f3_scannums_no_30101_in_31days_'] = sample_user2['f3_scannums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "###m13\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate==101)|(user2.cate==30))&(user2.a_type==2)],['sku_id'],'a_num','f3_attnums_30101_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate==30))&(user2.a_type==2)],['sku_id'],'a_num','f3_attnums_30_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate==101))&(user2.a_type==2)],['sku_id'],'a_num','f3_attnums_101_')\n",
    "sample_user2 = feat_sum(sample_user2,user2[((user2.cate!=101)&(user2.cate!=30))&(user2.a_type==2)],['sku_id'],'a_num','f3_attnums_no_30101_')\n",
    "###m14-m16\n",
    "sample_user2['f3_attnums_30101_in_7days_'] = sample_user2['f3_attnums_30101_a_num_sku_id_sum']/12\n",
    "sample_user2['f3_attnums_30_in_7days_'] = sample_user2['f3_attnums_30_a_num_sku_id_sum']/12\n",
    "sample_user2['f3_attnums_101_in_7days_'] = sample_user2['f3_attnums_101_a_num_sku_id_sum']/12\n",
    "sample_user2['f3_attnums_no_30101_in_7days_'] = sample_user2['f3_attnums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_user2['f3_attnums_30101_in_15days_'] = sample_user2['f3_attnums_30101_a_num_sku_id_sum']/6\n",
    "sample_user2['f3_attnums_30_in_15days_'] = sample_user2['f3_attnums_30_a_num_sku_id_sum']/6\n",
    "sample_user2['f3_attnums_101_in_15days_'] = sample_user2['f3_attnums_101_a_num_sku_id_sum']/6\n",
    "sample_user2['f3_attnums_no_30101_in_15days_'] = sample_user2['f3_attnums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_user2['f3_attnums_30101_in_31days_'] = sample_user2['f3_attnums_30101_a_num_sku_id_sum']/3\n",
    "sample_user2['f3_attnums_30_in_31days_'] = sample_user2['f3_attnums_30_a_num_sku_id_sum']/3\n",
    "sample_user2['f3_attnums_101_in_31days_'] = sample_user2['f3_attnums_101_a_num_sku_id_sum']/3\n",
    "sample_user2['f3_attnums_no_30101_in_31days_'] = sample_user2['f3_attnums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "\n",
    "###m17\n",
    "sample_user2 = feat_nunique(sample_user2,user2[(user2.cate==101)|(user2.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30101_')\n",
    "sample_user2 = feat_nunique(sample_user2,user2[(user2.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30_')\n",
    "sample_user2 = feat_nunique(sample_user2,user2[(user2.cate==101)],['sku_id'],'comment_create_tm','f3_commentnums_101_')\n",
    "sample_user2 = feat_nunique(sample_user2,user2[(user2.cate!=101)&(user2.cate!=30)],['sku_id'],'comment_create_tm','f3_commentnums_no_30101_')\n",
    "\n",
    "###m18-20\n",
    "sample_user2['f3_commentnums_30101_in_7days_'] = sample_user2['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user2['f3_commentnums_30_in_7days_'] = sample_user2['f3_commentnums_30_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user2['f3_commentnums_101_in_7days_'] = sample_user2['f3_commentnums_101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user2['f3_commentnums_no_30101_in_7days_'] = sample_user2['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/12\n",
    "\n",
    "sample_user2['f3_commentnums_30101_in_15days_'] = sample_user2['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user2['f3_commentnums_30_in_15days_'] = sample_user2['f3_commentnums_30_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user2['f3_commentnums_101_in_15days_'] = sample_user2['f3_commentnums_101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user2['f3_commentnums_no_30101_in_15days_'] = sample_user2['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/6\n",
    "\n",
    "sample_user2['f3_commentnums_30101_in_31days_'] = sample_user2['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user2['f3_commentnums_30_in_31days_'] = sample_user2['f3_commentnums_30_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user2['f3_commentnums_101_in_31days_'] = sample_user2['f3_commentnums_101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user2['f3_commentnums_no_30101_in_31days_'] = sample_user2['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/3\n",
    "\n",
    "###m21\n",
    "t1 = sample_user2[['sku_id','f3_buytimes_30101_o_id_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate==101)|(user2.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate==101)|(user2.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_buy_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_buytimes_30_o_id_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_buy_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_buytimes_101_o_id_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate==101)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate==101)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_buy_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_buytimes_no_30101_o_id_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate!=101)&(user2.cate!=30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate!=101)&(user2.cate!=30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_no_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_buy_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m22\n",
    "t1 = sample_user2[['sku_id','f3_scannums_30101_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==1)&((user2.cate==101)|(user2.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==1)&((user2.cate==101)|(user2.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_scan_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_scannums_30_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==1)&((user2.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==1)&((user2.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_scan_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_scannums_101_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==1)&((user2.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==1)&((user2.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_scan_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_scannums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==1)&((user2.cate!=30)&(user2.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==1)&((user2.cate!=30)&(user2.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_scan_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m23\n",
    "t1 = sample_user2[['sku_id','f3_attnums_30101_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==2)&((user2.cate==101)|(user2.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==2)&((user2.cate==101)|(user2.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_att_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_attnums_30_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==2)&((user2.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==2)&((user2.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_att_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_attnums_101_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==2)&((user2.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==2)&((user2.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_att_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user2[['sku_id','f3_attnums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = user2[(user2.a_type==2)&((user2.cate!=30)&(user2.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.a_type==2)&((user2.cate!=30)&(user2.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_att_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m24\n",
    "t1 = sample_user2[['sku_id','f3_commentnums_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate==101)|(user2.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate==101)|(user2.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_comment_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "t1 = sample_user2[['sku_id','f3_commentnums_30_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_comment_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_user2[['sku_id','f3_commentnums_101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate==101)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate==101)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_comment_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_user2[['sku_id','f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user2[(user2.cate!=101)&(user2.cate!=30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user2[(user2.cate!=101)&(user2.cate!=30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_comment_day_gap']]\n",
    "sample_user2 = sample_user2.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m25-m32还没写，正在考虑\n",
    "user2_a_o_date = user2[((pd.to_datetime(user2.a_date)<=pd.to_datetime(user2.o_date.fillna('2018-01-01'))))]\n",
    "user2_a_o_date = user2_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user2_a_o_date = user2_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user2_a_o_date['o_a_gap']=(pd.to_datetime(user2_a_o_date.o_date)-pd.to_datetime(user2_a_o_date.a_date)).dt.days \n",
    "FLAG='user2_a_o_'\n",
    "sample_user2 = feat_count(sample_user2,user2_a_o_date[user2_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user2 = feat_mean(sample_user2,user2_a_o_date[(user2_a_o_date.o_a_gap>=-1)&((user2_a_o_date.cate==101)|(user2_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30101_')\n",
    "sample_user2 = feat_mean(sample_user2,user2_a_o_date[(user2_a_o_date.o_a_gap>=-1)&((user2_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30_')\n",
    "sample_user2 = feat_mean(sample_user2,user2_a_o_date[(user2_a_o_date.o_a_gap>=-1)&((user2_a_o_date.cate==101))],['sku_id'],'o_a_gap',FLAG+'101_')\n",
    "sample_user2 = feat_mean(sample_user2,user2_a_o_date[(user2_a_o_date.o_a_gap>=-1)&((user2_a_o_date.cate!=101)&(user2_a_o_date.cate!=30))],['sku_id'],'o_a_gap',FLAG+'no_30101_')\n",
    "\n",
    "\n",
    "###u33\n",
    "t1 = user2[['user_id','o_sku_num','price','cate']] \n",
    "t1['sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user2 = feat_sum(sample_user2,t1[(t1.cate==30)|(t1.cate==101)],['user_id'],'sku_total_price','f3_user_total_consumption_30101_')\n",
    "sample_user2 = feat_sum(sample_user2,t1[(t1.cate==30)],['user_id'],'sku_total_price','f3_user_total_consumption_30_')\n",
    "sample_user2 = feat_sum(sample_user2,t1[(t1.cate==101)],['user_id'],'sku_total_price','f3_user_total_consumption_101_')\n",
    "sample_user2 = feat_sum(sample_user2,t1[(t1.cate!=30)&(t1.cate!=101)],['user_id'],'sku_total_price','f3_user_total_consumption_no_30101_')\n",
    "\n",
    "###u34-u36\n",
    "sample_user2['f3_total_consumption_30101_in_7days_'] = sample_user2['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/12\n",
    "sample_user2['f3_total_consumption_30_in_7days_'] = sample_user2['f3_user_total_consumption_30_sku_total_price_user_id_sum']/12\n",
    "sample_user2['f3_total_consumption_101_in_7days_'] = sample_user2['f3_user_total_consumption_101_sku_total_price_user_id_sum']/12\n",
    "sample_user2['f3_total_consumption_no_30101_in_7days_'] = sample_user2['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/12\n",
    "\n",
    "sample_user2['f3_total_consumption_30101_in_15days_'] = sample_user2['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/6\n",
    "sample_user2['f3_total_consumption_30_in_15days_'] = sample_user2['f3_user_total_consumption_30_sku_total_price_user_id_sum']/6\n",
    "sample_user2['f3_total_consumption_101_in_15days_'] = sample_user2['f3_user_total_consumption_101_sku_total_price_user_id_sum']/6\n",
    "sample_user2['f3_total_consumption_no_30101_in_15days_'] = sample_user2['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/6\n",
    "\n",
    "sample_user2['f3_total_consumption_30101_in_31days_'] = sample_user2['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/3\n",
    "sample_user2['f3_total_consumption_30_in_31days_'] = sample_user2['f3_user_total_consumption_30_sku_total_price_user_id_sum']/3\n",
    "sample_user2['f3_total_consumption_101_in_31days_'] = sample_user2['f3_user_total_consumption_101_sku_total_price_user_id_sum']/3\n",
    "sample_user2['f3_total_consumption_no_30101_in_31days_'] = sample_user2['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/3\n",
    "\n",
    "\n",
    "###u37-u40\n",
    "sample_user2 = feat_min(sample_user2,user2[()],['user_id'],'o_day','f3_buytime_early_')\n",
    "sample_user2 = feat_min(sample_user2,user2[user2.a_type==1],['user_id'],'a_day','f3_scantime_early_')\n",
    "sample_user2 = feat_min(sample_user2,user2[user2.a_type==2],['user_id'],'a_day','f3_atttime_early_')\n",
    "sample_user2 = feat_min(sample_user2,user2,['user_id'],'comment_day','f3_commenttime_early_')\n",
    "\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==101)|(user2.cate==30)],['user_id'],'o_day','f3_buytime_early_30101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==30)],['user_id'],'o_day','f3_buytime_early_30_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==101)],['user_id'],'o_day','f3_buytime_early_101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate!=101)&(user2.cate!=30)],['user_id'],'o_day','f3_buytime_early_no_30101_')\n",
    "\n",
    "sample_user2 = feat_min(sample_user2,user2[((user2.cate==101)|(user2.cate==30))&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_early_30101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==30)&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_early_30_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==101)&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_early_101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[((user2.cate!=101)&(user2.cate!=30))&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_early_30101_')\n",
    "\n",
    "sample_user2 = feat_min(sample_user2,user2[((user2.cate==101)|(user2.cate==30))&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_early_30101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==30)&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_early_30_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==101)&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_early_101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[((user2.cate!=101)&(user2.cate!=30))&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_early_no_30101_')\n",
    "\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==101)|(user2.cate==30)],['user_id'],'comment_day','f3_commenttime_early_30101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==30)],['user_id'],'comment_day','f3_commenttime_early_30_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate==101)],['user_id'],'comment_day','f3_commenttime_early_101_')\n",
    "sample_user2 = feat_min(sample_user2,user2[(user2.cate!=101)&(user2.cate!=30)],['user_id'],'comment_day','f3_commenttime_early_no_30101_')\n",
    "\n",
    "\n",
    "###u41-u44\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==101)|(user2.cate==30)],['user_id'],'o_day','f3_buytime_late_30101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==30)],['user_id'],'o_day','f3_buytime_late_30_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==101)],['user_id'],'o_day','f3_buytime_late_101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate!=101)&(user2.cate!=30)],['user_id'],'o_day','f3_buytime_late_no_30101_')\n",
    "\n",
    "sample_user2 = feat_max(sample_user2,user2[((user2.cate==101)|(user2.cate==30))&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_late_30101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==30)&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_late_30_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==101)&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_late_101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[((user2.cate!=101)&(user2.cate!=30))&(user2.a_type==1)],['user_id'],'a_day','f3_scantime_late_30101_')\n",
    "\n",
    "sample_user2 = feat_max(sample_user2,user2[((user2.cate==101)|(user2.cate==30))&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_late_30101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==30)&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_late_30_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==101)&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_late_101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[((user2.cate!=101)&(user2.cate!=30))&(user2.a_type==2)],['user_id'],'a_day','f3_atttime_late_no_30101_')\n",
    "\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==101)|(user2.cate==30)],['user_id'],'comment_day','f3_commenttime_late_30101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==30)],['user_id'],'comment_day','f3_commenttime_late_30_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate==101)],['user_id'],'comment_day','f3_commenttime_late_101_')\n",
    "sample_user2 = feat_max(sample_user2,user2[(user2.cate!=101)&(user2.cate!=30)],['user_id'],'comment_day','f3_commenttime_late_no_30101_')\n",
    "\n",
    "\n",
    "sample_user2 = sample_user2.drop_duplicates()\n",
    "gc.collect()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "133"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "###用户特征\n",
    "#######dataset1_user\n",
    "#######\n",
    "####for dataset1\n",
    "user1 = feature1[['user_id','sku_id','price','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','age','user_lv_cd','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = user1[['user_id','age','user_lv_cd']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###u1\n",
    "sample_user1 = feat_nunique(t,user1,['user_id'],'o_id','f1_buytimes_')\n",
    "###u2-u4\n",
    "sample_user1['f1_buytimes_in_7days_'] = sample_user1['f1_buytimes_o_id_user_id_nunique']/12\n",
    "sample_user1['f1_buytimes_in_15days_'] = sample_user1['f1_buytimes_o_id_user_id_nunique']/6\n",
    "sample_user1['f1_buytimes_in_31days_'] = sample_user1['f1_buytimes_o_id_user_id_nunique']/3\n",
    "\n",
    "###u5\n",
    "sample_user1 = feat_sum(sample_user1,user1,['user_id'],'o_sku_num','f1_buynums_')\n",
    "###u6-u8\n",
    "sample_user1['f1_buynums_in_7days_'] = sample_user1['f1_buynums_o_sku_num_user_id_sum']/12\n",
    "sample_user1['f1_buynums_in_15days_'] = sample_user1['f1_buynums_o_sku_num_user_id_sum']/6\n",
    "sample_user1['f1_buynums_in_31days_'] = sample_user1['f1_buynums_o_sku_num_user_id_sum']/3\n",
    "\n",
    "###u9\n",
    "sample_user1 = feat_sum(sample_user1,user1[user1.a_type==1],['user_id'],'a_num','f1_scannums_')\n",
    "###u10-u12\n",
    "sample_user1['f1_scannums_in_7days_'] = sample_user1['f1_scannums_a_num_user_id_sum']/12\n",
    "sample_user1['f1_scannums_in_15days_'] = sample_user1['f1_scannums_a_num_user_id_sum']/6\n",
    "sample_user1['f1_scannums_in_31days_'] = sample_user1['f1_scannums_a_num_user_id_sum']/3\n",
    "\n",
    "###u13\n",
    "sample_user1 = feat_sum(sample_user1,user1[user1.a_type==2],['user_id'],'a_num','f1_attnums_')\n",
    "###u14-u16\n",
    "sample_user1['f1_attnums_in_7days_'] = sample_user1['f1_attnums_a_num_user_id_sum']/12\n",
    "sample_user1['f1_attnums_in_15days_'] = sample_user1['f1_attnums_a_num_user_id_sum']/6\n",
    "sample_user1['f1_attnums_in_31days_'] = sample_user1['f1_attnums_a_num_user_id_sum']/3\n",
    "\n",
    "###u17\n",
    "sample_user1 = feat_nunique(sample_user1,user1,['user_id'],'comment_create_tm','f1_commentnums_')\n",
    "###u18-20\n",
    "sample_user1['f1_commentnums_in_7days_'] = sample_user1['f1_commentnums_comment_create_tm_user_id_nunique']/12\n",
    "sample_user1['f1_commentnums_in_15days_'] = sample_user1['f1_commentnums_comment_create_tm_user_id_nunique']/6\n",
    "sample_user1['f1_commentnums_in_31days_'] = sample_user1['f1_commentnums_comment_create_tm_user_id_nunique']/3\n",
    "\n",
    "###u21\n",
    "t1 = sample_user1[['user_id','f1_buytimes_o_id_user_id_nunique']]\n",
    "t2 = user1.groupby(['user_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user1.groupby(['user_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f1_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f1_buytimes_o_id_user_id_nunique']\n",
    "t = t[['user_id','f1_average_buy_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "###u22\n",
    "t1 = sample_user1[['user_id','f1_scannums_a_num_user_id_sum']]\n",
    "t2 = user1[user1.a_type==1].groupby(['user_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user1[user1.a_type==1].groupby(['user_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f1_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f1_scannums_a_num_user_id_sum']\n",
    "t = t[['user_id','f1_average_scan_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u23\n",
    "t1 = sample_user1[['user_id','f1_attnums_a_num_user_id_sum']]\n",
    "t2 = user1[user1.a_type==2].groupby(['user_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user1[user1.a_type==2].groupby(['user_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f1_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f1_attnums_a_num_user_id_sum']\n",
    "t = t[['user_id','f1_average_att_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u24\n",
    "t1 = sample_user1[['user_id','f1_commentnums_comment_create_tm_user_id_nunique']]\n",
    "t2 = user1.groupby(['user_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['user_id'],how='left')\n",
    "t3 = user1.groupby(['user_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['user_id'],how='left')\n",
    "t['f1_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f1_commentnums_comment_create_tm_user_id_nunique']\n",
    "t = t[['user_id','f1_average_comment_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['user_id'],how='left')\n",
    "\n",
    "\n",
    "###u25-u32还没写，正在考虑\n",
    "user1_a_o_date = user1[((pd.to_datetime(user1.a_date)<=pd.to_datetime(user1.o_date.fillna('2018-01-01'))))]\n",
    "user1_a_o_date = user1_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user1_a_o_date = user1_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user1_a_o_date['o_a_gap']=(pd.to_datetime(user1_a_o_date.o_date)-pd.to_datetime(user1_a_o_date.a_date)).dt.days \n",
    "FLAG='user1_a_o_'\n",
    "sample_user1 = feat_count(sample_user1,user1_a_o_date[user1_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user1 = feat_mean(sample_user1,user1_a_o_date[user1_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###u33\n",
    "t1 = user1[['user_id','o_sku_num','price']] \n",
    "t1['sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user1 = feat_sum(sample_user1,t1,['user_id'],'sku_total_price','f1_user_total_consumption_')\n",
    "\n",
    "###u34-u36\n",
    "sample_user1['f1_total_consumption_in_7days_'] = sample_user1['f1_user_total_consumption_sku_total_price_user_id_sum']/12\n",
    "sample_user1['f1_total_consumption_in_15days_'] = sample_user1['f1_user_total_consumption_sku_total_price_user_id_sum']/6\n",
    "sample_user1['f1_total_consumption_in_31days_'] = sample_user1['f1_user_total_consumption_sku_total_price_user_id_sum']/3\n",
    "\n",
    "###u37-u40\n",
    "sample_user1 = feat_min(sample_user1,user1,['user_id'],'o_day','f1_buytime_early_')\n",
    "sample_user1 = feat_min(sample_user1,user1[user1.a_type==1],['user_id'],'a_day','f1_scantime_early_')\n",
    "sample_user1 = feat_min(sample_user1,user1[user1.a_type==2],['user_id'],'a_day','f1_atttime_early_')\n",
    "sample_user1 = feat_min(sample_user1,user1,['user_id'],'comment_day','f1_commenttime_early_')\n",
    "\n",
    "###u41-u44\n",
    "sample_user1 = feat_max(sample_user1,user1,['user_id'],'o_day','f1_buytime_late_')\n",
    "sample_user1 = feat_max(sample_user1,user1[user1.a_type==1],['user_id'],'a_day','f1_scantime_late_')\n",
    "sample_user1 = feat_max(sample_user1,user1[user1.a_type==2],['user_id'],'a_day','f1_atttime_late_')\n",
    "sample_user1 = feat_max(sample_user1,user1,['user_id'],'comment_day','f1_commenttime_late_')\n",
    "\n",
    "sample_user1 = sample_user1.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "user1 = feature3[['user_id','sku_id','price','o_id','o_date','o_sku_num','o_month','o_day','comment_create_tm','score_level','comment_year','comment_month','comment_day','age','user_lv_cd','a_date','a_num','a_type','a_year','a_month','a_day']]\n",
    "t = user1[['user_id','age','user_lv_cd','cate']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###m1\n",
    "sample_user1 = feat_nunique(t,user1[(user1.cate==101)|(user1.cate==30)],['sku_id'],'o_id','f3_buytimes_30101_')\n",
    "sample_user1 = feat_nunique(sample_user1,user1[(user1.cate==30)],['sku_id'],'o_id','f3_buytimes_30_')\n",
    "sample_user1 = feat_nunique(sample_user1,user1[(user1.cate==101)],['sku_id'],'o_id','f3_buytimes_101_')\n",
    "sample_user1 = feat_nunique(sample_user1,user1[(user1.cate!=101)&(user1.cate!=30)],['sku_id'],'o_id','f3_buytimes_no_30101_')\n",
    "\n",
    "###m2-m4\n",
    "sample_user1['f3_buytimes_30101_in_7days_'] = sample_user1['f3_buytimes_30101_o_id_sku_id_nunique']/12\n",
    "sample_user1['f3_buytimes_30_in_7days_'] = sample_user1['f3_buytimes_30_o_id_sku_id_nunique']/12\n",
    "sample_user1['f3_buytimes_101_in_7days_'] = sample_user1['f3_buytimes_101_o_id_sku_id_nunique']/12\n",
    "sample_user1['f3_buytimes_no_30101_in_7days_'] = sample_user1['f3_buytimes_no_30101_o_id_sku_id_nunique']/12\n",
    "\n",
    "sample_user1['f3_buytimes_30101_in_15days_'] = sample_user1['f3_buytimes_30101_o_id_sku_id_nunique']/6\n",
    "sample_user1['f3_buytimes_30_in_15days_'] = sample_user1['f3_buytimes_30_o_id_sku_id_nunique']/6\n",
    "sample_user1['f3_buytimes_101_in_15days_'] = sample_user1['f3_buytimes_101_o_id_sku_id_nunique']/6\n",
    "sample_user1['f3_buytimes_no_30101_in_15days_'] = sample_user1['f3_buytimes_no_30101_o_id_sku_id_nunique']/6\n",
    "\n",
    "sample_user1['f3_buytimes_30101_in_31days_'] = sample_user1['f3_buytimes_30101_o_id_sku_id_nunique']/3\n",
    "sample_user1['f3_buytimes_30_in_31days_'] = sample_user1['f3_buytimes_30_o_id_sku_id_nunique']/3\n",
    "sample_user1['f3_buytimes_101_in_31days_'] = sample_user1['f3_buytimes_101_o_id_sku_id_nunique']/3\n",
    "sample_user1['f3_buytimes_no_30101_in_31days_'] = sample_user1['f3_buytimes_no_30101_o_id_sku_id_nunique']/3\n",
    "\n",
    "###m5\n",
    "sample_user1 = feat_sum(sample_user1,user1[(user1.cate==101)|(user1.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30101_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[(user1.cate==30)],['sku_id'],'o_sku_num','f3_buynums_30_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[(user1.cate==101)],['sku_id'],'o_sku_num','f3_buynums_101_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[(user1.cate!=101)&(user1.cate!=30)],['sku_id'],'o_sku_num','f3_buynums_no_30101_')\n",
    "\n",
    "###m6-m8\n",
    "sample_user1['f3_buynums_30101_in_7days_'] = sample_user1['f3_buynums_30101_o_sku_num_sku_id_sum']/12\n",
    "sample_user1['f3_buynums_30_in_7days_'] = sample_user1['f3_buynums_30_o_sku_num_sku_id_sum']/12\n",
    "sample_user1['f3_buynums_101_in_7days_'] = sample_user1['f3_buynums_101_o_sku_num_sku_id_sum']/12\n",
    "sample_user1['f3_buynums_no_30101_in_7days_'] = sample_user1['f3_buynums_no_30101_o_sku_num_sku_id_sum']/12\n",
    "\n",
    "sample_user1['f3_buynums_30101_in_15days_'] = sample_user1['f3_buynums_30101_o_sku_num_sku_id_sum']/6\n",
    "sample_user1['f3_buynums_30_in_15days_'] = sample_user1['f3_buynums_30_o_sku_num_sku_id_sum']/6\n",
    "sample_user1['f3_buynums_101_in_15days_'] = sample_user1['f3_buynums_101_o_sku_num_sku_id_sum']/6\n",
    "sample_user1['f3_buynums_no_30101_in_15days_'] = sample_user1['f3_buynums_no_30101_o_sku_num_sku_id_sum']/6\n",
    "\n",
    "sample_user1['f3_buynums_30101_in_31days_'] = sample_user1['f3_buynums_30101_o_sku_num_sku_id_sum']/3\n",
    "sample_user1['f3_buynums_30_in_31days_'] = sample_user1['f3_buynums_30_o_sku_num_sku_id_sum']/3\n",
    "sample_user1['f3_buynums_101_in_31days_'] = sample_user1['f3_buynums_101_o_sku_num_sku_id_sum']/3\n",
    "sample_user1['f3_buynums_no_30101_in_31days_'] = sample_user1['f3_buynums_no_30101_o_sku_num_sku_id_sum']/3\n",
    "\n",
    "###m9\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate==101)|(user1.cate==30))&(user1.a_type==1)],['sku_id'],'a_num','f3_scannums_30101_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate==30))&(user1.a_type==1)],['sku_id'],'a_num','f3_scannums_30_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate==101))&(user1.a_type==1)],['sku_id'],'a_num','f3_scannums_101_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate!=101)&(user1.cate!=30))&(user1.a_type==1)],['sku_id'],'a_num','f3_scannums_no_30101_')\n",
    "###m10-m12\n",
    "sample_user1['f3_scannums_30101_in_7days_'] = sample_user1['f3_scannums_30101_a_num_sku_id_sum']/12\n",
    "sample_user1['f3_scannums_30_in_7days_'] = sample_user1['f3_scannums_30_a_num_sku_id_sum']/12\n",
    "sample_user1['f3_scannums_101_in_7days_'] = sample_user1['f3_scannums_101_a_num_sku_id_sum']/12\n",
    "sample_user1['f3_scannums_no_30101_in_7days_'] = sample_user1['f3_scannums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_user1['f3_scannums_30101_in_15days_'] = sample_user1['f3_scannums_30101_a_num_sku_id_sum']/6\n",
    "sample_user1['f3_scannums_30_in_15days_'] = sample_user1['f3_scannums_30_a_num_sku_id_sum']/6\n",
    "sample_user1['f3_scannums_101_in_15days_'] = sample_user1['f3_scannums_101_a_num_sku_id_sum']/6\n",
    "sample_user1['f3_scannums_no_30101_in_15days_'] = sample_user1['f3_scannums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_user1['f3_scannums_30101_in_31days_'] = sample_user1['f3_scannums_30101_a_num_sku_id_sum']/3\n",
    "sample_user1['f3_scannums_30_in_31days_'] = sample_user1['f3_scannums_30_a_num_sku_id_sum']/3\n",
    "sample_user1['f3_scannums_101_in_31days_'] = sample_user1['f3_scannums_101_a_num_sku_id_sum']/3\n",
    "sample_user1['f3_scannums_no_30101_in_31days_'] = sample_user1['f3_scannums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "###m13\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate==101)|(user1.cate==30))&(user1.a_type==2)],['sku_id'],'a_num','f3_attnums_30101_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate==30))&(user1.a_type==2)],['sku_id'],'a_num','f3_attnums_30_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate==101))&(user1.a_type==2)],['sku_id'],'a_num','f3_attnums_101_')\n",
    "sample_user1 = feat_sum(sample_user1,user1[((user1.cate!=101)&(user1.cate!=30))&(user1.a_type==2)],['sku_id'],'a_num','f3_attnums_no_30101_')\n",
    "###m14-m16\n",
    "sample_user1['f3_attnums_30101_in_7days_'] = sample_user1['f3_attnums_30101_a_num_sku_id_sum']/12\n",
    "sample_user1['f3_attnums_30_in_7days_'] = sample_user1['f3_attnums_30_a_num_sku_id_sum']/12\n",
    "sample_user1['f3_attnums_101_in_7days_'] = sample_user1['f3_attnums_101_a_num_sku_id_sum']/12\n",
    "sample_user1['f3_attnums_no_30101_in_7days_'] = sample_user1['f3_attnums_no_30101_a_num_sku_id_sum']/12\n",
    "\n",
    "sample_user1['f3_attnums_30101_in_15days_'] = sample_user1['f3_attnums_30101_a_num_sku_id_sum']/6\n",
    "sample_user1['f3_attnums_30_in_15days_'] = sample_user1['f3_attnums_30_a_num_sku_id_sum']/6\n",
    "sample_user1['f3_attnums_101_in_15days_'] = sample_user1['f3_attnums_101_a_num_sku_id_sum']/6\n",
    "sample_user1['f3_attnums_no_30101_in_15days_'] = sample_user1['f3_attnums_no_30101_a_num_sku_id_sum']/6\n",
    "\n",
    "sample_user1['f3_attnums_30101_in_31days_'] = sample_user1['f3_attnums_30101_a_num_sku_id_sum']/3\n",
    "sample_user1['f3_attnums_30_in_31days_'] = sample_user1['f3_attnums_30_a_num_sku_id_sum']/3\n",
    "sample_user1['f3_attnums_101_in_31days_'] = sample_user1['f3_attnums_101_a_num_sku_id_sum']/3\n",
    "sample_user1['f3_attnums_no_30101_in_31days_'] = sample_user1['f3_attnums_no_30101_a_num_sku_id_sum']/3\n",
    "\n",
    "\n",
    "\n",
    "###m17\n",
    "sample_user1 = feat_nunique(sample_user1,user1[(user1.cate==101)|(user1.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30101_')\n",
    "sample_user1 = feat_nunique(sample_user1,user1[(user1.cate==30)],['sku_id'],'comment_create_tm','f3_commentnums_30_')\n",
    "sample_user1 = feat_nunique(sample_user1,user1[(user1.cate==101)],['sku_id'],'comment_create_tm','f3_commentnums_101_')\n",
    "sample_user1 = feat_nunique(sample_user1,user1[(user1.cate!=101)&(user1.cate!=30)],['sku_id'],'comment_create_tm','f3_commentnums_no_30101_')\n",
    "\n",
    "###m18-20\n",
    "sample_user1['f3_commentnums_30101_in_7days_'] = sample_user1['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user1['f3_commentnums_30_in_7days_'] = sample_user1['f3_commentnums_30_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user1['f3_commentnums_101_in_7days_'] = sample_user1['f3_commentnums_101_comment_create_tm_sku_id_nunique']/12\n",
    "sample_user1['f3_commentnums_no_30101_in_7days_'] = sample_user1['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/12\n",
    "\n",
    "sample_user1['f3_commentnums_30101_in_15days_'] = sample_user1['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user1['f3_commentnums_30_in_15days_'] = sample_user1['f3_commentnums_30_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user1['f3_commentnums_101_in_15days_'] = sample_user1['f3_commentnums_101_comment_create_tm_sku_id_nunique']/6\n",
    "sample_user1['f3_commentnums_no_30101_in_15days_'] = sample_user1['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/6\n",
    "\n",
    "sample_user1['f3_commentnums_30101_in_31days_'] = sample_user1['f3_commentnums_30101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user1['f3_commentnums_30_in_31days_'] = sample_user1['f3_commentnums_30_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user1['f3_commentnums_101_in_31days_'] = sample_user1['f3_commentnums_101_comment_create_tm_sku_id_nunique']/3\n",
    "sample_user1['f3_commentnums_no_30101_in_31days_'] = sample_user1['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']/3\n",
    "\n",
    "###m21\n",
    "t1 = sample_user1[['sku_id','f3_buytimes_30101_o_id_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate==101)|(user1.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate==101)|(user1.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_buy_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_buytimes_30_o_id_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate==30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate==30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_30_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_buy_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_buytimes_101_o_id_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate==101)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate==101)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_buy_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_buytimes_no_30101_o_id_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate!=101)&(user1.cate!=30)].groupby(['sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate!=101)&(user1.cate!=30)].groupby(['sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_no_30101_o_id_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_buy_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m22\n",
    "t1 = sample_user1[['sku_id','f3_scannums_30101_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==1)&((user1.cate==101)|(user1.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==1)&((user1.cate==101)|(user1.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_scan_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_scannums_30_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==1)&((user1.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==1)&((user1.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_scan_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_scannums_101_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==1)&((user1.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==1)&((user1.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_scan_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_scannums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==1)&((user1.cate!=30)&(user1.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==1)&((user1.cate!=30)&(user1.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_scan_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m23\n",
    "t1 = sample_user1[['sku_id','f3_attnums_30101_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==2)&((user1.cate==101)|(user1.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==2)&((user1.cate==101)|(user1.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30101_att_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_attnums_30_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==2)&((user1.cate==30))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==2)&((user1.cate==30))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_30_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_30_att_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_attnums_101_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==2)&((user1.cate==101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==2)&((user1.cate==101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_101_att_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###\n",
    "t1 = sample_user1[['sku_id','f3_attnums_no_30101_a_num_sku_id_sum']]\n",
    "t2 = user1[(user1.a_type==2)&((user1.cate!=30)&(user1.cate!=101))].groupby(['sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.a_type==2)&((user1.cate!=30)&(user1.cate!=101))].groupby(['sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_no_30101_a_num_sku_id_sum']\n",
    "t = t[['sku_id','f3_average_no_30101_att_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "###m24\n",
    "t1 = sample_user1[['sku_id','f3_commentnums_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate==101)|(user1.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate==101)|(user1.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30101_comment_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "t1 = sample_user1[['sku_id','f3_commentnums_30_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate==30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate==30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_30_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_30_comment_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_user1[['sku_id','f3_commentnums_101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate==101)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate==101)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_101_comment_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "\n",
    "t1 = sample_user1[['sku_id','f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']]\n",
    "t2 = user1[(user1.cate!=101)&(user1.cate!=30)].groupby(['sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['sku_id'],how='left')\n",
    "t3 = user1[(user1.cate!=101)&(user1.cate!=30)].groupby(['sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_no_30101_comment_create_tm_sku_id_nunique']\n",
    "t = t[['sku_id','f3_average_no_30101_comment_day_gap']]\n",
    "sample_user1 = sample_user1.merge(t,on=['sku_id'],how='left')\n",
    "\n",
    "###m25-m32还没写，正在考虑\n",
    "user1_a_o_date = user1[((pd.to_datetime(user1.a_date)<=pd.to_datetime(user1.o_date.fillna('2018-01-01'))))]\n",
    "user1_a_o_date = user1_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user1_a_o_date = user1_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user1_a_o_date['o_a_gap']=(pd.to_datetime(user1_a_o_date.o_date)-pd.to_datetime(user1_a_o_date.a_date)).dt.days \n",
    "FLAG='user1_a_o_'\n",
    "sample_user1 = feat_count(sample_user1,user1_a_o_date[user1_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user1 = feat_mean(sample_user1,user1_a_o_date[(user1_a_o_date.o_a_gap>=-1)&((user1_a_o_date.cate==101)|(user1_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30101_')\n",
    "sample_user1 = feat_mean(sample_user1,user1_a_o_date[(user1_a_o_date.o_a_gap>=-1)&((user1_a_o_date.cate==30))],['sku_id'],'o_a_gap',FLAG+'30_')\n",
    "sample_user1 = feat_mean(sample_user1,user1_a_o_date[(user1_a_o_date.o_a_gap>=-1)&((user1_a_o_date.cate==101))],['sku_id'],'o_a_gap',FLAG+'101_')\n",
    "sample_user1 = feat_mean(sample_user1,user1_a_o_date[(user1_a_o_date.o_a_gap>=-1)&((user1_a_o_date.cate!=101)&(user1_a_o_date.cate!=30))],['sku_id'],'o_a_gap',FLAG+'no_30101_')\n",
    "\n",
    "\n",
    "###u33\n",
    "t1 = user1[['user_id','o_sku_num','price','cate']] \n",
    "t1['sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user1 = feat_sum(sample_user1,t1[(t1.cate==30)|(t1.cate==101)],['user_id'],'sku_total_price','f3_user_total_consumption_30101_')\n",
    "sample_user1 = feat_sum(sample_user1,t1[(t1.cate==30)],['user_id'],'sku_total_price','f3_user_total_consumption_30_')\n",
    "sample_user1 = feat_sum(sample_user1,t1[(t1.cate==101)],['user_id'],'sku_total_price','f3_user_total_consumption_101_')\n",
    "sample_user1 = feat_sum(sample_user1,t1[(t1.cate!=30)&(t1.cate!=101)],['user_id'],'sku_total_price','f3_user_total_consumption_no_30101_')\n",
    "\n",
    "###u34-u36\n",
    "sample_user1['f3_total_consumption_30101_in_7days_'] = sample_user1['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/12\n",
    "sample_user1['f3_total_consumption_30_in_7days_'] = sample_user1['f3_user_total_consumption_30_sku_total_price_user_id_sum']/12\n",
    "sample_user1['f3_total_consumption_101_in_7days_'] = sample_user1['f3_user_total_consumption_101_sku_total_price_user_id_sum']/12\n",
    "sample_user1['f3_total_consumption_no_30101_in_7days_'] = sample_user1['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/12\n",
    "\n",
    "sample_user1['f3_total_consumption_30101_in_15days_'] = sample_user1['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/6\n",
    "sample_user1['f3_total_consumption_30_in_15days_'] = sample_user1['f3_user_total_consumption_30_sku_total_price_user_id_sum']/6\n",
    "sample_user1['f3_total_consumption_101_in_15days_'] = sample_user1['f3_user_total_consumption_101_sku_total_price_user_id_sum']/6\n",
    "sample_user1['f3_total_consumption_no_30101_in_15days_'] = sample_user1['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/6\n",
    "\n",
    "sample_user1['f3_total_consumption_30101_in_31days_'] = sample_user1['f3_user_total_consumption_30101_sku_total_price_user_id_sum']/3\n",
    "sample_user1['f3_total_consumption_30_in_31days_'] = sample_user1['f3_user_total_consumption_30_sku_total_price_user_id_sum']/3\n",
    "sample_user1['f3_total_consumption_101_in_31days_'] = sample_user1['f3_user_total_consumption_101_sku_total_price_user_id_sum']/3\n",
    "sample_user1['f3_total_consumption_no_30101_in_31days_'] = sample_user1['f3_user_total_consumption_no_30101_sku_total_price_user_id_sum']/3\n",
    "\n",
    "\n",
    "###u37-u40\n",
    "sample_user1 = feat_min(sample_user1,user1[()],['user_id'],'o_day','f3_buytime_early_')\n",
    "sample_user1 = feat_min(sample_user1,user1[user1.a_type==1],['user_id'],'a_day','f3_scantime_early_')\n",
    "sample_user1 = feat_min(sample_user1,user1[user1.a_type==2],['user_id'],'a_day','f3_atttime_early_')\n",
    "sample_user1 = feat_min(sample_user1,user1,['user_id'],'comment_day','f3_commenttime_early_')\n",
    "\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==101)|(user1.cate==30)],['user_id'],'o_day','f3_buytime_early_30101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==30)],['user_id'],'o_day','f3_buytime_early_30_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==101)],['user_id'],'o_day','f3_buytime_early_101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate!=101)&(user1.cate!=30)],['user_id'],'o_day','f3_buytime_early_no_30101_')\n",
    "\n",
    "sample_user1 = feat_min(sample_user1,user1[((user1.cate==101)|(user1.cate==30))&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_early_30101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==30)&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_early_30_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==101)&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_early_101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[((user1.cate!=101)&(user1.cate!=30))&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_early_30101_')\n",
    "\n",
    "sample_user1 = feat_min(sample_user1,user1[((user1.cate==101)|(user1.cate==30))&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_early_30101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==30)&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_early_30_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==101)&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_early_101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[((user1.cate!=101)&(user1.cate!=30))&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_early_no_30101_')\n",
    "\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==101)|(user1.cate==30)],['user_id'],'comment_day','f3_commenttime_early_30101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==30)],['user_id'],'comment_day','f3_commenttime_early_30_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate==101)],['user_id'],'comment_day','f3_commenttime_early_101_')\n",
    "sample_user1 = feat_min(sample_user1,user1[(user1.cate!=101)&(user1.cate!=30)],['user_id'],'comment_day','f3_commenttime_early_no_30101_')\n",
    "\n",
    "\n",
    "###u41-u44\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==101)|(user1.cate==30)],['user_id'],'o_day','f3_buytime_late_30101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==30)],['user_id'],'o_day','f3_buytime_late_30_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==101)],['user_id'],'o_day','f3_buytime_late_101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate!=101)&(user1.cate!=30)],['user_id'],'o_day','f3_buytime_late_no_30101_')\n",
    "\n",
    "sample_user1 = feat_max(sample_user1,user1[((user1.cate==101)|(user1.cate==30))&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_late_30101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==30)&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_late_30_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==101)&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_late_101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[((user1.cate!=101)&(user1.cate!=30))&(user1.a_type==1)],['user_id'],'a_day','f3_scantime_late_30101_')\n",
    "\n",
    "sample_user1 = feat_max(sample_user1,user1[((user1.cate==101)|(user1.cate==30))&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_late_30101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==30)&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_late_30_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==101)&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_late_101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[((user1.cate!=101)&(user1.cate!=30))&(user1.a_type==2)],['user_id'],'a_day','f3_atttime_late_no_30101_')\n",
    "\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==101)|(user1.cate==30)],['user_id'],'comment_day','f3_commenttime_late_30101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==30)],['user_id'],'comment_day','f3_commenttime_late_30_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate==101)],['user_id'],'comment_day','f3_commenttime_late_101_')\n",
    "sample_user1 = feat_max(sample_user1,user1[(user1.cate!=101)&(user1.cate!=30)],['user_id'],'comment_day','f3_commenttime_late_no_30101_')\n",
    "\n",
    "\n",
    "sample_user1 = sample_user1.drop_duplicates()\n",
    "gc.collect()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###用户-商品组合特征\n",
    "#####dataset3_user_sku\n",
    "#####\n",
    "###for dataset3\n",
    "user_sku3 = feature3[(feature3.cate==30)|(feature3.cate==101)]\n",
    "t = user_sku3[['user_id','sku_id']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###um1\n",
    "sample_user_sku3 = feat_nunique(t,user_sku3,['user_id','sku_id'],'o_id','f3_buytimes_')\n",
    "###um2-um4\n",
    "sample_user_sku3['f3_buytimes_in_7days_'] = sample_user_sku3['f3_buytimes_o_id_user_id_sku_id_nunique']/12\n",
    "sample_user_sku3['f3_buytimes_in_15days_'] = sample_user_sku3['f3_buytimes_o_id_user_id_sku_id_nunique']/6\n",
    "sample_user_sku3['f3_buytimes_in_31days_'] = sample_user_sku3['f3_buytimes_o_id_user_id_sku_id_nunique']/3\n",
    "\n",
    "###um5\n",
    "sample_user_sku3 = feat_sum(sample_user_sku3,user_sku3,['user_id','sku_id'],'o_sku_num','f3_buynums_')\n",
    "###um6-um8\n",
    "sample_user_sku3['f3_buynums_in_7days_'] = sample_user_sku3['f3_buynums_o_sku_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku3['f3_buynums_in_15days_'] = sample_user_sku3['f3_buynums_o_sku_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku3['f3_buynums_in_31days_'] = sample_user_sku3['f3_buynums_o_sku_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um9\n",
    "sample_user_sku3 = feat_sum(sample_user_sku3,user_sku3[user_sku3.a_type==1],['user_id','sku_id'],'a_num','f3_scannums_')\n",
    "###um10-um12\n",
    "sample_user_sku3['f3_scannums_in_7days_'] = sample_user_sku3['f3_scannums_a_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku3['f3_scannums_in_15days_'] = sample_user_sku3['f3_scannums_a_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku3['f3_scannums_in_31days_'] = sample_user_sku3['f3_scannums_a_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um13\n",
    "sample_user_sku3 = feat_sum(sample_user_sku3,user_sku3[user_sku3.a_type==2],['user_id','sku_id'],'a_num','f3_attnums_')\n",
    "###um14-um16\n",
    "sample_user_sku3['f3_attnums_in_7days_'] = sample_user_sku3['f3_attnums_a_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku3['f3_attnums_in_15days_'] = sample_user_sku3['f3_attnums_a_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku3['f3_attnums_in_31days_'] = sample_user_sku3['f3_attnums_a_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um17\n",
    "sample_user_sku3 = feat_nunique(sample_user_sku3,user_sku3,['user_id','sku_id'],'comment_create_tm','f3_commentnums_')\n",
    "###um18-20\n",
    "sample_user_sku3['f3_commentnums_in_7days_'] = sample_user_sku3['f3_commentnums_comment_create_tm_user_id_sku_id_nunique']/12\n",
    "sample_user_sku3['f3_commentnums_in_15days_'] = sample_user_sku3['f3_commentnums_comment_create_tm_user_id_sku_id_nunique']/6\n",
    "sample_user_sku3['f3_commentnums_in_31days_'] = sample_user_sku3['f3_commentnums_comment_create_tm_user_id_sku_id_nunique']/3\n",
    "\n",
    "###um21\n",
    "t1 = sample_user_sku3[['user_id','f3_buytimes_o_id_user_id_sku_id_nunique','sku_id']]\n",
    "t2 = user_sku3.groupby(['user_id','sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku3.groupby(['user_id','sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f3_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f3_buytimes_o_id_user_id_sku_id_nunique']\n",
    "t = t[['user_id','sku_id','f3_average_buy_day_gap']]\n",
    "sample_user_sku3 = sample_user_sku3.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "###um22\n",
    "t1 = sample_user_sku3[['user_id','f3_scannums_a_num_user_id_sku_id_sum','sku_id']]\n",
    "t2 = user_sku3[user_sku3.a_type==1].groupby(['user_id','sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku3[user_sku3.a_type==1].groupby(['user_id','sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f3_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_scannums_a_num_user_id_sku_id_sum']\n",
    "t = t[['user_id','sku_id','f3_average_scan_day_gap']]\n",
    "sample_user_sku3 = sample_user_sku3.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um23\n",
    "t1 = sample_user_sku3[['user_id','sku_id','f3_attnums_a_num_user_id_sku_id_sum']]\n",
    "t2 = user_sku3[user_sku3.a_type==2].groupby(['user_id','sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku3[user_sku3.a_type==2].groupby(['user_id','sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f3_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f3_attnums_a_num_user_id_sku_id_sum']\n",
    "t = t[['user_id','sku_id','f3_average_att_day_gap']]\n",
    "sample_user_sku3 = sample_user_sku3.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um24\n",
    "t1 = sample_user_sku3[['user_id','sku_id','f3_commentnums_comment_create_tm_user_id_sku_id_nunique']]\n",
    "t2 = user_sku3.groupby(['user_id','sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku3.groupby(['user_id','sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f3_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f3_commentnums_comment_create_tm_user_id_sku_id_nunique']\n",
    "t = t[['user_id','sku_id','f3_average_comment_day_gap']]\n",
    "sample_user_sku3 = sample_user_sku3.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um25-um32还没写，正在考虑\n",
    "user_sku3_a_o_date = user_sku3[((pd.to_datetime(user_sku3.a_date)<=pd.to_datetime(user_sku3.o_date.fillna('2018-01-01'))))]\n",
    "user_sku3_a_o_date = user_sku3_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user_sku3_a_o_date = user_sku3_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user_sku3_a_o_date['o_a_gap']=(pd.to_datetime(user_sku3_a_o_date.o_date)-pd.to_datetime(user_sku3_a_o_date.a_date)).dt.days \n",
    "FLAG='user_sku3_a_o_'\n",
    "sample_user_sku3 = feat_count(sample_user_sku3,user_sku3_a_o_date[user_sku3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user_sku3 = feat_mean(sample_user_sku3,user_sku3_a_o_date[user_sku3_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###um33\n",
    "t1 = user_sku3[['user_id','sku_id','o_sku_num','price']] \n",
    "t1['user_sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user_sku3 = feat_sum(sample_user_sku3,t1,['user_id','sku_id'],'user_sku_total_price','f3_user_total_consumption_')\n",
    "\n",
    "###um34-um36\n",
    "sample_user_sku3['f3_total_consumption_in_7days_'] = sample_user_sku3['f3_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/12\n",
    "sample_user_sku3['f3_total_consumption_in_15days_'] = sample_user_sku3['f3_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/6\n",
    "sample_user_sku3['f3_total_consumption_in_31days_'] = sample_user_sku3['f3_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/3\n",
    "\n",
    "###um37-um40\n",
    "sample_user_sku3 = feat_min(sample_user_sku3,user_sku3,['user_id','sku_id'],'o_day','f3_buytime_early_')\n",
    "sample_user_sku3 = feat_min(sample_user_sku3,user_sku3[user_sku3.a_type==1],['user_id','sku_id'],'a_day','f3_scantime_early_')\n",
    "sample_user_sku3 = feat_min(sample_user_sku3,user_sku3[user_sku3.a_type==2],['user_id','sku_id'],'a_day','f3_atttime_early_')\n",
    "sample_user_sku3 = feat_min(sample_user_sku3,user_sku3,['user_id','sku_id'],'comment_day','f3_commenttime_early_')\n",
    "\n",
    "###um41-um44\n",
    "sample_user_sku3 = feat_max(sample_user_sku3,user_sku3,['user_id','sku_id'],'o_day','f3_buytime_late_')\n",
    "sample_user_sku3 = feat_max(sample_user_sku3,user_sku3[user_sku3.a_type==1],['user_id','sku_id'],'a_day','f3_scantime_late_')\n",
    "sample_user_sku3 = feat_max(sample_user_sku3,user_sku3[user_sku3.a_type==2],['user_id','sku_id'],'a_day','f3_atttime_late_')\n",
    "sample_user_sku3 = feat_max(sample_user_sku3,user_sku3,['user_id','sku_id'],'comment_day','f3_commenttime_late_')\n",
    "\n",
    "\n",
    "sample_user_sku3 = sample_user_sku3.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###用户-商品组合特征\n",
    "#####dataset2_user_sku\n",
    "#####\n",
    "###for dataset2\n",
    "user_sku2 = feature2[(feature2.cate==30)|(feature2.cate==101)]\n",
    "t = user_sku2[['user_id','sku_id']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###um1\n",
    "sample_user_sku2 = feat_nunique(t,user_sku2,['user_id','sku_id'],'o_id','f2_buytimes_')\n",
    "###um2-um4\n",
    "sample_user_sku2['f2_buytimes_in_7days_'] = sample_user_sku2['f2_buytimes_o_id_user_id_sku_id_nunique']/12\n",
    "sample_user_sku2['f2_buytimes_in_15days_'] = sample_user_sku2['f2_buytimes_o_id_user_id_sku_id_nunique']/6\n",
    "sample_user_sku2['f2_buytimes_in_31days_'] = sample_user_sku2['f2_buytimes_o_id_user_id_sku_id_nunique']/3\n",
    "\n",
    "###um5\n",
    "sample_user_sku2 = feat_sum(sample_user_sku2,user_sku2,['user_id','sku_id'],'o_sku_num','f2_buynums_')\n",
    "###um6-um8\n",
    "sample_user_sku2['f2_buynums_in_7days_'] = sample_user_sku2['f2_buynums_o_sku_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku2['f2_buynums_in_15days_'] = sample_user_sku2['f2_buynums_o_sku_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku2['f2_buynums_in_31days_'] = sample_user_sku2['f2_buynums_o_sku_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um9\n",
    "sample_user_sku2 = feat_sum(sample_user_sku2,user_sku2[user_sku2.a_type==1],['user_id','sku_id'],'a_num','f2_scannums_')\n",
    "###um10-um12\n",
    "sample_user_sku2['f2_scannums_in_7days_'] = sample_user_sku2['f2_scannums_a_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku2['f2_scannums_in_15days_'] = sample_user_sku2['f2_scannums_a_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku2['f2_scannums_in_31days_'] = sample_user_sku2['f2_scannums_a_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um13\n",
    "sample_user_sku2 = feat_sum(sample_user_sku2,user_sku2[user_sku2.a_type==2],['user_id','sku_id'],'a_num','f2_attnums_')\n",
    "###um14-um16\n",
    "sample_user_sku2['f2_attnums_in_7days_'] = sample_user_sku2['f2_attnums_a_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku2['f2_attnums_in_15days_'] = sample_user_sku2['f2_attnums_a_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku2['f2_attnums_in_31days_'] = sample_user_sku2['f2_attnums_a_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um17\n",
    "sample_user_sku2 = feat_nunique(sample_user_sku2,user_sku2,['user_id','sku_id'],'comment_create_tm','f2_commentnums_')\n",
    "###um18-20\n",
    "sample_user_sku2['f2_commentnums_in_7days_'] = sample_user_sku2['f2_commentnums_comment_create_tm_user_id_sku_id_nunique']/12\n",
    "sample_user_sku2['f2_commentnums_in_15days_'] = sample_user_sku2['f2_commentnums_comment_create_tm_user_id_sku_id_nunique']/6\n",
    "sample_user_sku2['f2_commentnums_in_31days_'] = sample_user_sku2['f2_commentnums_comment_create_tm_user_id_sku_id_nunique']/3\n",
    "\n",
    "###um21\n",
    "t1 = sample_user_sku2[['user_id','f2_buytimes_o_id_user_id_sku_id_nunique','sku_id']]\n",
    "t2 = user_sku2.groupby(['user_id','sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku2.groupby(['user_id','sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f2_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f2_buytimes_o_id_user_id_sku_id_nunique']\n",
    "t = t[['user_id','sku_id','f2_average_buy_day_gap']]\n",
    "sample_user_sku2 = sample_user_sku2.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "###um22\n",
    "t1 = sample_user_sku2[['user_id','f2_scannums_a_num_user_id_sku_id_sum','sku_id']]\n",
    "t2 = user_sku2[user_sku2.a_type==1].groupby(['user_id','sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku2[user_sku2.a_type==1].groupby(['user_id','sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f2_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f2_scannums_a_num_user_id_sku_id_sum']\n",
    "t = t[['user_id','sku_id','f2_average_scan_day_gap']]\n",
    "sample_user_sku2 = sample_user_sku2.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um23\n",
    "t1 = sample_user_sku2[['user_id','sku_id','f2_attnums_a_num_user_id_sku_id_sum']]\n",
    "t2 = user_sku2[user_sku2.a_type==2].groupby(['user_id','sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku2[user_sku2.a_type==2].groupby(['user_id','sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f2_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f2_attnums_a_num_user_id_sku_id_sum']\n",
    "t = t[['user_id','sku_id','f2_average_att_day_gap']]\n",
    "sample_user_sku2 = sample_user_sku2.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um24\n",
    "t1 = sample_user_sku2[['user_id','sku_id','f2_commentnums_comment_create_tm_user_id_sku_id_nunique']]\n",
    "t2 = user_sku2.groupby(['user_id','sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku2.groupby(['user_id','sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f2_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f2_commentnums_comment_create_tm_user_id_sku_id_nunique']\n",
    "t = t[['user_id','sku_id','f2_average_comment_day_gap']]\n",
    "sample_user_sku2 = sample_user_sku2.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###u25-u32还没写，正在考虑\n",
    "user_sku2_a_o_date = user_sku2[((pd.to_datetime(user_sku2.a_date)<=pd.to_datetime(user_sku2.o_date.fillna('2018-01-01'))))]\n",
    "user_sku2_a_o_date = user_sku2_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user_sku2_a_o_date = user_sku2_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user_sku2_a_o_date['o_a_gap']=(pd.to_datetime(user_sku2_a_o_date.o_date)-pd.to_datetime(user_sku2_a_o_date.a_date)).dt.days \n",
    "FLAG='user_sku2_a_o_'\n",
    "sample_user_sku2 = feat_count(sample_user_sku2,user_sku2_a_o_date[user_sku2_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user_sku2 = feat_mean(sample_user_sku2,user_sku2_a_o_date[user_sku2_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###um33\n",
    "t1 = user_sku2[['user_id','sku_id','o_sku_num','price']] \n",
    "t1['user_sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user_sku2 = feat_sum(sample_user_sku2,t1,['user_id','sku_id'],'user_sku_total_price','f2_user_total_consumption_')\n",
    "\n",
    "###um34-um36\n",
    "sample_user_sku2['f2_total_consumption_in_7days_'] = sample_user_sku2['f2_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/12\n",
    "sample_user_sku2['f2_total_consumption_in_15days_'] = sample_user_sku2['f2_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/6\n",
    "sample_user_sku2['f2_total_consumption_in_31days_'] = sample_user_sku2['f2_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/3\n",
    "\n",
    "\n",
    "###um37-um40\n",
    "sample_user_sku2 = feat_min(sample_user_sku2,user_sku2,['user_id','sku_id'],'o_day','f2_buytime_early_')\n",
    "sample_user_sku2 = feat_min(sample_user_sku2,user_sku2[user_sku2.a_type==1],['user_id','sku_id'],'a_day','f2_scantime_early_')\n",
    "sample_user_sku2 = feat_min(sample_user_sku2,user_sku2[user_sku2.a_type==2],['user_id','sku_id'],'a_day','f2_atttime_early_')\n",
    "sample_user_sku2 = feat_min(sample_user_sku2,user_sku2,['user_id','sku_id'],'comment_day','f2_commenttime_early_')\n",
    "\n",
    "###um41-um44\n",
    "sample_user_sku2 = feat_max(sample_user_sku2,user_sku2,['user_id','sku_id'],'o_day','f2_buytime_late_')\n",
    "sample_user_sku2 = feat_max(sample_user_sku2,user_sku2[user_sku2.a_type==1],['user_id','sku_id'],'a_day','f2_scantime_late_')\n",
    "sample_user_sku2 = feat_max(sample_user_sku2,user_sku2[user_sku2.a_type==2],['user_id','sku_id'],'a_day','f2_atttime_late_')\n",
    "sample_user_sku2 = feat_max(sample_user_sku2,user_sku2,['user_id','sku_id'],'comment_day','f2_commenttime_late_')\n",
    "\n",
    "\n",
    "sample_user_sku2 = sample_user_sku2.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###用户-商品组合特征\n",
    "#####dataset1_user_sku\n",
    "#####\n",
    "###for dataset1\n",
    "user_sku1 = feature1[(feature1.cate==30)|(feature1.cate==101)]\n",
    "t = user_sku1[['user_id','sku_id']]\n",
    "t.drop_duplicates(inplace=True)\n",
    "###um1\n",
    "sample_user_sku1 = feat_nunique(t,user_sku1,['user_id','sku_id'],'o_id','f1_buytimes_')\n",
    "###um2-um4\n",
    "sample_user_sku1['f1_buytimes_in_7days_'] = sample_user_sku1['f1_buytimes_o_id_user_id_sku_id_nunique']/12\n",
    "sample_user_sku1['f1_buytimes_in_15days_'] = sample_user_sku1['f1_buytimes_o_id_user_id_sku_id_nunique']/6\n",
    "sample_user_sku1['f1_buytimes_in_31days_'] = sample_user_sku1['f1_buytimes_o_id_user_id_sku_id_nunique']/3\n",
    "\n",
    "###um5\n",
    "sample_user_sku1 = feat_sum(sample_user_sku1,user_sku1,['user_id','sku_id'],'o_sku_num','f1_buynums_')\n",
    "###um6-um8\n",
    "sample_user_sku1['f1_buynums_in_7days_'] = sample_user_sku1['f1_buynums_o_sku_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku1['f1_buynums_in_15days_'] = sample_user_sku1['f1_buynums_o_sku_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku1['f1_buynums_in_31days_'] = sample_user_sku1['f1_buynums_o_sku_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um9\n",
    "sample_user_sku1 = feat_sum(sample_user_sku1,user_sku1[user_sku1.a_type==1],['user_id','sku_id'],'a_num','f1_scannums_')\n",
    "###um10-um12\n",
    "sample_user_sku1['f1_scannums_in_7days_'] = sample_user_sku1['f1_scannums_a_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku1['f1_scannums_in_15days_'] = sample_user_sku1['f1_scannums_a_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku1['f1_scannums_in_31days_'] = sample_user_sku1['f1_scannums_a_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um13\n",
    "sample_user_sku1 = feat_sum(sample_user_sku1,user_sku1[user_sku1.a_type==2],['user_id','sku_id'],'a_num','f1_attnums_')\n",
    "###um14-um16\n",
    "sample_user_sku1['f1_attnums_in_7days_'] = sample_user_sku1['f1_attnums_a_num_user_id_sku_id_sum']/12\n",
    "sample_user_sku1['f1_attnums_in_15days_'] = sample_user_sku1['f1_attnums_a_num_user_id_sku_id_sum']/6\n",
    "sample_user_sku1['f1_attnums_in_31days_'] = sample_user_sku1['f1_attnums_a_num_user_id_sku_id_sum']/3\n",
    "\n",
    "###um17\n",
    "sample_user_sku1 = feat_nunique(sample_user_sku1,user_sku1,['user_id','sku_id'],'comment_create_tm','f1_commentnums_')\n",
    "###um18-20\n",
    "sample_user_sku1['f1_commentnums_in_7days_'] = sample_user_sku1['f1_commentnums_comment_create_tm_user_id_sku_id_nunique']/12\n",
    "sample_user_sku1['f1_commentnums_in_15days_'] = sample_user_sku1['f1_commentnums_comment_create_tm_user_id_sku_id_nunique']/6\n",
    "sample_user_sku1['f1_commentnums_in_31days_'] = sample_user_sku1['f1_commentnums_comment_create_tm_user_id_sku_id_nunique']/3\n",
    "\n",
    "###um21\n",
    "t1 = sample_user_sku1[['user_id','f1_buytimes_o_id_user_id_sku_id_nunique','sku_id']]\n",
    "t2 = user_sku1.groupby(['user_id','sku_id'])['o_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','o_date':'o_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku1.groupby(['user_id','sku_id'])['o_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','o_date':'o_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f1_average_buy_day_gap'] = (t['o_date_max'] - t['o_date_min']).dt.days/t['f1_buytimes_o_id_user_id_sku_id_nunique']\n",
    "t = t[['user_id','sku_id','f1_average_buy_day_gap']]\n",
    "sample_user_sku1 = sample_user_sku1.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "###um22\n",
    "t1 = sample_user_sku1[['user_id','f1_scannums_a_num_user_id_sku_id_sum','sku_id']]\n",
    "t2 = user_sku1[user_sku1.a_type==1].groupby(['user_id','sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku1[user_sku1.a_type==1].groupby(['user_id','sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f1_average_scan_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f1_scannums_a_num_user_id_sku_id_sum']\n",
    "t = t[['user_id','sku_id','f1_average_scan_day_gap']]\n",
    "sample_user_sku1 = sample_user_sku1.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um23\n",
    "t1 = sample_user_sku1[['user_id','sku_id','f1_attnums_a_num_user_id_sku_id_sum']]\n",
    "t2 = user_sku1[user_sku1.a_type==2].groupby(['user_id','sku_id'])['a_date'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku1[user_sku1.a_type==2].groupby(['user_id','sku_id'])['a_date'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','a_date':'a_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f1_average_att_day_gap'] = (t['a_date_max'] - t['a_date_min']).dt.days/t['f1_attnums_a_num_user_id_sku_id_sum']\n",
    "t = t[['user_id','sku_id','f1_average_att_day_gap']]\n",
    "sample_user_sku1 = sample_user_sku1.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um24\n",
    "t1 = sample_user_sku1[['user_id','sku_id','f1_commentnums_comment_create_tm_user_id_sku_id_nunique']]\n",
    "t2 = user_sku1.groupby(['user_id','sku_id'])['comment_create_tm'].max().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','comment_create_tm':'comment_date_max'})\n",
    "t = t2.merge(t1,on=['user_id','sku_id'],how='left')\n",
    "t3 = user_sku1.groupby(['user_id','sku_id'])['comment_create_tm'].min().reset_index().rename(columns={'user_id':'user_id','sku_id':'sku_id','comment_create_tm':'comment_date_min'})\n",
    "t = t.merge(t3,on=['user_id','sku_id'],how='left')\n",
    "t['f1_average_comment_day_gap'] = (t['comment_date_max'] - t['comment_date_min']).dt.days/t['f1_commentnums_comment_create_tm_user_id_sku_id_nunique']\n",
    "t = t[['user_id','sku_id','f1_average_comment_day_gap']]\n",
    "sample_user_sku1 = sample_user_sku1.merge(t,on=['user_id','sku_id'],how='left')\n",
    "\n",
    "\n",
    "###um25-um32还没写，正在考虑\n",
    "user_sku1_a_o_date = user_sku1[((pd.to_datetime(user_sku1.a_date)<=pd.to_datetime(user_sku1.o_date.fillna('2018-01-01'))))]\n",
    "user_sku1_a_o_date = user_sku1_a_o_date.sort_values(by=['sku_id','a_date','o_date']).drop_duplicates(['sku_id','a_date'],keep=\"first\")\n",
    "user_sku1_a_o_date = user_sku1_a_o_date[['sku_id','a_date','o_date','cate']]\n",
    "user_sku1_a_o_date['o_a_gap']=(pd.to_datetime(user_sku1_a_o_date.o_date)-pd.to_datetime(user_sku1_a_o_date.a_date)).dt.days \n",
    "FLAG='user_sku1_a_o_'\n",
    "sample_user_sku1 = feat_count(sample_user_sku1,user_sku1_a_o_date[user_sku1_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG+'order_')\n",
    "sample_user_sku1 = feat_mean(sample_user_sku1,user_sku1_a_o_date[user_sku1_a_o_date.o_a_gap>=-1],['sku_id'],'o_a_gap',FLAG)\n",
    "\n",
    "###um33\n",
    "t1 = user_sku1[['user_id','sku_id','o_sku_num','price']] \n",
    "t1['user_sku_total_price'] = t1['o_sku_num'] * t1['price']\n",
    "sample_user_sku1 = feat_sum(sample_user_sku1,t1,['user_id','sku_id'],'user_sku_total_price','f1_user_total_consumption_')\n",
    "\n",
    "###um34-um36\n",
    "sample_user_sku1['f1_total_consumption_in_7days_'] = sample_user_sku1['f1_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/12\n",
    "sample_user_sku1['f1_total_consumption_in_15days_'] = sample_user_sku1['f1_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/6\n",
    "sample_user_sku1['f1_total_consumption_in_31days_'] = sample_user_sku1['f1_user_total_consumption_user_sku_total_price_user_id_sku_id_sum']/3\n",
    "\n",
    "\n",
    "###um37-um40\n",
    "sample_user_sku1 = feat_min(sample_user_sku1,user_sku1,['user_id','sku_id'],'o_day','f1_buytime_early_')\n",
    "sample_user_sku1 = feat_min(sample_user_sku1,user_sku1[user_sku1.a_type==1],['user_id','sku_id'],'a_day','f1_scantime_early_')\n",
    "sample_user_sku1 = feat_min(sample_user_sku1,user_sku1[user_sku1.a_type==2],['user_id','sku_id'],'a_day','f1_atttime_early_')\n",
    "sample_user_sku1 = feat_min(sample_user_sku1,user_sku1,['user_id','sku_id'],'comment_day','f1_commenttime_early_')\n",
    "\n",
    "###um41-um44\n",
    "sample_user_sku1 = feat_max(sample_user_sku1,user_sku1,['user_id','sku_id'],'o_day','f1_buytime_late_')\n",
    "sample_user_sku1 = feat_max(sample_user_sku1,user_sku1[user_sku1.a_type==1],['user_id','sku_id'],'a_day','f1_scantime_late_')\n",
    "sample_user_sku1 = feat_max(sample_user_sku1,user_sku1[user_sku1.a_type==2],['user_id','sku_id'],'a_day','f1_atttime_late_')\n",
    "sample_user_sku1 = feat_max(sample_user_sku1,user_sku1,['user_id','sku_id'],'comment_day','f1_commenttime_late_')\n",
    "\n",
    "\n",
    "sample_user_sku1 = sample_user_sku1.drop_duplicates()\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "####最后把三个day的特征加上，之前加会有memoryerror\n",
    "####最后把三个day的特征加上，之前加会有memoryerror\n",
    "####最后把三个day的特征加上，之前加会有memoryerror\n",
    "\n",
    "sample3 = sample_user_sku3.merge(sample_sku3,on=['sku_id'],how='left')\n",
    "sample3 = sample3.merge(sample_user3,on=['user_id'],how='left')\n",
    "test_cv = sample3[['user_id']]\n",
    "\n",
    "sample2 = sample_user_sku2.merge(sample_sku2,on=['sku_id'],how='left')\n",
    "sample2 = sample2.merge(sample_user2,on=['user_id'],how='left')\n",
    "\n",
    "sample1 = sample_user_sku1.merge(sample_sku1,on=['sku_id'],how='left')\n",
    "sample1 = sample1.merge(sample_user1,on=['user_id'],how='left')\n",
    "\n",
    "print(sample3.shape,sample2.shape,sample1.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1389415, 117) (1347333, 117)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "14"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gc.collect()\n",
    "train_dataset2 = sample2.merge(dataset2,on=['user_id','sku_id'],how='left')\n",
    "train_dataset1 = sample1.merge(dataset1,on=['user_id','sku_id'],how='left')\n",
    "train_dataset2 = train_dataset2.drop(['user_id','sku_id'],axis=1)\n",
    "train_dataset1 = train_dataset1.drop(['user_id','sku_id'],axis=1)\n",
    "print(train_dataset2.shape,train_dataset1.shape)\n",
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_dataset2 = train_dataset2.fillna(0)\n",
    "train_dataset1 = train_dataset1.fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "115\n"
     ]
    }
   ],
   "source": [
    "feature_combine = sample1.drop(['user_id','sku_id'],axis=1)\n",
    "feature_list = feature_combine.columns.tolist()\n",
    "# train_feature_label, val_feature_label\n",
    "# test_feature\n",
    "print (len(feature_list))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 1282992 entries, 0 to 1282991\n",
      "Columns: 117 entries, user_id to comment_day\n",
      "dtypes: float64(96), int64(21)\n",
      "memory usage: 1.1 GB\n"
     ]
    }
   ],
   "source": [
    "sample3.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "ename": "MemoryError",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mMemoryError\u001b[0m                               Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-24-ddd10a0882ee>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     24\u001b[0m \u001b[0mX_pred\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msample3\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdrop\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'user_id'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;34m'sku_id'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0maxis\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     25\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 26\u001b[1;33m \u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0my\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     27\u001b[0m \u001b[0mtest_cv\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'probility'\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX_pred\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\learningdata\\Mlearing\\jupyter\\JData\\SBBTree_ONLINE.py\u001b[0m in \u001b[0;36mfit\u001b[1;34m(self, X, y)\u001b[0m\n\u001b[0;32m     41\u001b[0m                                 \u001b[0mX_train\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mX\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mtrain_index\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     42\u001b[0m                                 \u001b[0my_train\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0my\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mtrain_index\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 43\u001b[1;33m                                 \u001b[0mX_test\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mX\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mtest_index\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     44\u001b[0m                                 \u001b[0my_test\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0my\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mtest_index\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     45\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mMemoryError\u001b[0m: "
     ]
    }
   ],
   "source": [
    "params = {\n",
    "    'task': 'train',\n",
    "    'boosting_type': 'gbdt',\n",
    "    'objective': 'regression',\n",
    "    'metric': {'l2', 'auc'},\n",
    "    'num_leaves': 31,\n",
    "    'learning_rate': 0.05,\n",
    "    'feature_fraction': 0.9,\n",
    "    'bagging_fraction': 0.8,\n",
    "    'bagging_freq': 5,\n",
    "    'verbose': 0\n",
    "}\n",
    "\n",
    "model = SBBTree(params=params, stacking_num=5, bagging_num=3,  bagging_test_size=0.33, num_boost_round=10000, early_stopping_rounds=200)\n",
    "\n",
    "# train 下个月购买次数预测 回归模型\n",
    "\n",
    "train_data = train_dataset2.drop(['label_1','label_2'],axis=1)\n",
    "train_label = train_dataset2['label_1']\n",
    "\n",
    "X = train_data.values \n",
    "y = train_label.values\n",
    "\n",
    "X_pred = sample3.drop(['user_id','sku_id'],axis=1).values\n",
    "\n",
    "model.fit(X,y)\n",
    "test_cv['probility'] = model.predict(X_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "params = {\n",
    "    'task': 'train',\n",
    "    'boosting_type': 'gbdt',\n",
    "    'objective': 'regression',\n",
    "    'metric': {'l2'},\n",
    "    'num_leaves': 31,\n",
    "    'learning_rate': 0.05,\n",
    "    'feature_fraction': 0.9,\n",
    "    'bagging_fraction': 0.8,\n",
    "    'bagging_freq': 5,\n",
    "    'verbose': 0\n",
    "}\n",
    "model = SBBTree(params=params, stacking_num=5, bagging_num=3,  bagging_test_size=0.33, num_boost_round=10000, early_stopping_rounds=200)\n",
    "\n",
    "# train 当月首次购买时间预测 回归模型\n",
    "train_data = train_dataset2.drop(['label_1','label_2'],axis=1)\n",
    "train_label = train_dataset2['label_2']\n",
    "\n",
    "X = train_data.values \n",
    "y = train_label.values\n",
    "\n",
    "X_pred = sample3.drop(['user_id','sku_id'],axis=1).values\n",
    "\n",
    "model.fit(X,y)\n",
    "test_cv['buy_gap'] = model.predict(X_pred)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}