修改特征工程文件

eda/esmm/Model_pipline/feature_engineering.py

@@ -104,15 +104,15 @@ def feature_engineer():
 
     unique_values = []
     db = pymysql.connect(host='172.16.40.158', port=4000, user='root', passwd='3SYz54LS9#^9sBvC', db='jerry_test')
-    sql = "select distinct stat_date from esmm_train_data_dwell"
+    sql = "select distinct stat_date from esmm_train_data_share"
     unique_values.extend(get_unique(db,sql))
 
     db = pymysql.connect(host='172.16.40.158', port=4000, user='root', passwd='3SYz54LS9#^9sBvC', db='jerry_test')
-    sql = "select distinct ucity_id from esmm_train_data_dwell"
+    sql = "select distinct ucity_id from esmm_train_data_share"
     unique_values.extend(get_unique(db, sql))
 
     db = pymysql.connect(host='172.16.40.158', port=4000, user='root', passwd='3SYz54LS9#^9sBvC', db='jerry_test')
-    sql = "select distinct ccity_name from esmm_train_data_dwell"
+    sql = "select distinct ccity_name from esmm_train_data_share"
     unique_values.extend(get_unique(db, sql))
 
     db = pymysql.connect(host='172.16.40.158', port=4000, user='root', passwd='3SYz54LS9#^9sBvC', db='jerry_test')
@@ -162,15 +162,16 @@ def feature_engineer():
     # unique_values.append("video")
 
     db = pymysql.connect(host='172.16.40.158', port=4000, user='root', passwd='3SYz54LS9#^9sBvC', db='jerry_test')
-    sql = "select max(stat_date) from esmm_train_data_dwell"
+    sql = "select max(stat_date) from esmm_train_data_share"
     validate_date = con_sql(db, sql)[0].values.tolist()[0]
     print("validate_date:" + validate_date)
     temp = datetime.datetime.strptime(validate_date, "%Y-%m-%d")
     start = (temp - datetime.timedelta(days=180)).strftime("%Y-%m-%d")
     print(start)
+    print("这是分享数据")
 
     db = pymysql.connect(host='172.16.40.158', port=4000, user='root', passwd='3SYz54LS9#^9sBvC')
-    sql = "select distinct doctor.hospital_id from jerry_test.esmm_train_data_dwell e " \
+    sql = "select distinct doctor.hospital_id from jerry_test.esmm_train_data_share e " \
           "left join eagle.src_zhengxing_api_service service on e.diary_service_id = service.id " \
           "left join eagle.src_zhengxing_api_doctor doctor on service.doctor_id = doctor.id " \
           "where e.stat_date >= '{}'".format(start)
@@ -190,13 +191,13 @@ def feature_engineer():
                       29 + apps_number + level2_number + level3_number + len(unique_values)))
     value_map = dict(zip(unique_values, temp))
 
-    sql = "select e.y,e.z,e.stat_date,e.ucity_id,feat.level2_ids,e.ccity_name,u.device_type,u.manufacturer," \
+    sql = "select e.y,e.z,e.s,e.stat_date,e.ucity_id,feat.level2_ids,e.ccity_name,u.device_type,u.manufacturer," \
           "u.channel,c.top,cut.time,dl.app_list,feat.level3_ids,doctor.hospital_id," \
           "wiki.tag as tag1,question.tag as tag2,search.tag as tag3,budan.tag as tag4," \
           "ot.tag as tag5,sixin.tag as tag6,cart.tag as tag7,doris.search_tag2,doris.search_tag3," \
           "k.treatment_method,k.price_min,k.price_max,k.treatment_time,k.maintain_time,k.recover_time," \
           "e.device_id,e.cid_id " \
-          "from jerry_test.esmm_train_data_dwell e left join jerry_test.user_feature u on e.device_id = u.device_id " \
+          "from jerry_test.esmm_train_data_share e left join jerry_test.user_feature u on e.device_id = u.device_id " \
           "left join jerry_test.cid_type_top c on e.device_id = c.device_id " \
           "left join jerry_test.cid_time_cut cut on e.cid_id = cut.cid " \
           "left join jerry_test.device_app_list dl on e.device_id = dl.device_id " \
@@ -226,7 +227,7 @@ def feature_engineer():
                     "tag1", "tag2", "tag3", "tag4", "tag5", "tag6", "tag7",
                     "ucity_id", "ccity_name", "device_type", "manufacturer", "channel", "top", "time",
                     "hospital_id", "treatment_method", "price_min", "price_max", "treatment_time",
-                    "maintain_time", "recover_time", "search_tag2", "search_tag3","cid_id","device_id")\
+                    "maintain_time", "recover_time", "search_tag2", "search_tag3","cid_id","device_id","s")\
         .rdd.repartition(200).map(
         lambda x: (x[0], float(x[1]), float(x[2]), app_list_func(x[3], app_list_map), app_list_func(x[4], leve2_map),
                    app_list_func(x[5], leve3_map), app_list_func(x[6], leve2_map), app_list_func(x[7], leve2_map),
@@ -237,40 +238,37 @@ def feature_engineer():
                     value_map.get(x[20], 9), value_map.get(x[21], 10),
                     value_map.get(x[22], 11), value_map.get(x[23], 12), value_map.get(x[24], 13),
                     value_map.get(x[25], 14), value_map.get(x[26], 15)],
-                   app_list_func(x[27], leve2_map), app_list_func(x[28], leve3_map),x[13],x[29],x[30]
+                   app_list_func(x[27], leve2_map), app_list_func(x[28], leve3_map),x[13],x[29],x[30],float(x[31])
                    ))
 
 
     rdd.persist(storageLevel= StorageLevel.MEMORY_ONLY_SER)
 
     # TODO 上线后把下面train fliter 删除，因为最近一天的数据也要作为训练集
-    print("最近一天的数据没有加到训练集里")
+
     train = rdd.filter(lambda x: x[0] != validate_date).map(
         lambda x: (x[1], x[2], x[3], x[4], x[5], x[6], x[7], x[8], x[9],
-                   x[10], x[11], x[12], x[13], x[14], x[15],x[16],x[17],x[18]))
+                   x[10], x[11], x[12], x[13], x[14], x[15],x[16],x[17],x[18],x[19]))
     f = time.time()
     spark.createDataFrame(train).toDF("y", "z", "app_list", "level2_list", "level3_list",
                                       "tag1_list", "tag2_list", "tag3_list", "tag4_list",
                                       "tag5_list", "tag6_list", "tag7_list", "ids",
-                                      "search_tag2_list","search_tag3_list","city","cid_id","uid") \
+                                      "search_tag2_list","search_tag3_list","city","cid_id","uid","s") \
         .repartition(1).write.format("tfrecords").save(path=path + "tr/", mode="overwrite")
     h = time.time()
     print("train tfrecord done")
     print((h - f) / 60)
 
-    print("训练集样本总量：")
-    print(rdd.count())
-
     get_pre_number()
 
     test = rdd.filter(lambda x: x[0] == validate_date).map(
         lambda x: (x[1], x[2], x[3], x[4], x[5], x[6], x[7], x[8], x[9],
-                   x[10], x[11], x[12], x[13], x[14], x[15],x[16],x[17],x[18]))
+                   x[10], x[11], x[12], x[13], x[14], x[15],x[16],x[17],x[18],x[19]))
 
     spark.createDataFrame(test).toDF("y", "z", "app_list", "level2_list", "level3_list",
                                      "tag1_list", "tag2_list", "tag3_list", "tag4_list",
                                      "tag5_list", "tag6_list", "tag7_list", "ids",
-                                     "search_tag2_list","search_tag3_list","city","cid_id","uid") \
+                                     "search_tag2_list","search_tag3_list","city","cid_id","uid","s") \
         .repartition(1).write.format("tfrecords").save(path=path + "va/", mode="overwrite")
 
     print("va tfrecord done")
@@ -389,3 +387,9 @@ if __name__ == '__main__':
     # get_predict(validate_date, value_map, app_list_map, leve2_map, leve3_map)
 
     spark.stop()
+
+
+
+
+
+