把最近一天的数据集放进训练集

tensnsorflow/multi.py

@@ -91,7 +91,7 @@ def feature_engineer():
                   "channel", "top", "time", "hospital_id","treatment_method", "price_min",
                   "price_max", "treatment_time","maintain_time", "recover_time","y","z",).rdd
     rdd.persist()
-
+    # 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
     train = rdd.filter(lambda x: x[3]!= validate_date).map(lambda x: (app_list_func(x[0], app_list_map), app_list_func(x[1], leve2_map),
                             app_list_func(x[2], leve3_map),value_map[x[3]],value_map[x[4]],
                             value_map[x[5]],value_map[x[6]],value_map[x[7]],value_map[x[8]],
@@ -114,71 +114,71 @@ def feature_engineer():
     return validate_date,value_map,app_list_map,leve2_map,leve3_map
 
 
-# def get_predict(date,value_map,app_list_map,level2_map,level3_map):
-#     sql = "select e.y,e.z,e.label,e.ucity_id,feat.level2_ids,e.ccity_name," \
-#           "u.device_type,u.manufacturer,u.channel,c.top,e.device_id,e.cid_id,cut.time," \
-#           "dl.app_list,e.hospital_id,feat.level3_ids," \
-#           "k.treatment_method,k.price_min,k.price_max,k.treatment_time,k.maintain_time,k.recover_time " \
-#           "from esmm_pre_data e left join user_feature u on e.device_id = u.device_id " \
-#           "left join cid_type_top c on e.device_id = c.device_id " \
-#           "left join cid_time_cut cut on e.cid_id = cut.cid " \
-#           "left join device_app_list dl on e.device_id = dl.device_id " \
-#           "left join diary_feat feat on e.cid_id = feat.diary_id " \
-#           "left join train_Knowledge_network_data k on feat.level2 = k.level2_id"
-#
-#     features = ["app_list", "level2_ids", "level3_ids","ucity_id", "ccity_name", "device_type", "manufacturer",
-#                 "channel", "top", "time", "hospital_id",
-#                 "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time", "recover_time"]
-#     df = spark.sql(sql)
-#
-#     df = df.na.fill(dict(zip(features, features)))
-#     df = df.drop_duplicates(["ucity_id", "level2_ids", "ccity_name", "device_type", "manufacturer",
-#                              "device_id","cid,id","label",
-#                              "channel", "top", "time", "app_list", "hospital_id", "level3_ids"])
-#
-#     rdd = df.select("app_list", "level2_ids", "level3_ids","ucity_id","device_id","cid_id","label", "y", "z",
-#                     "ccity_name", "device_type","manufacturer", "channel", "top", "time", "hospital_id",
-#               "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time",
-#               "recover_time") \
-#         .rdd.map(lambda x: (app_list_func(x[0], app_list_map), app_list_func(x[1], level2_map),
-#                         app_list_func(x[2], level3_map), x[3],x[4],x[5],x[6],x[7],x[8],
-#                             value_map[x[3]], value_map[x[9]],
-#                         value_map[x[10]], value_map[x[11]], value_map[x[12]], value_map[x[13]],
-#                         value_map[x[14]], value_map[x[15]], value_map[x[16]], value_map[x[17]],
-#                         value_map[x[18]], value_map[x[19]], value_map[x[20]], value_map[x[21]],
-#                         value_map[date]))
-#
-#     rdd.persist()
-#
-#     native_pre = spark.createDataFrame(rdd.filter(lambda x:x[6] == 0).map(lambda x:(x[3],x[4],x[5])))\
-#         .toDF("city","uid","cid_id")
-#     print("native")
-#     print(native_pre.count())
-#     native_pre.write.csv('/recommend', mode='overwrite', header=True)
-#
-#     spark.createDataFrame(rdd.filter(lambda x: x[6] == 0)
-#                                    .map(lambda x: (x[0], x[1], x[2],x[9],x[10],x[11],x[12],x[13],x[14],x[15],
-#                                                    x[16,x[17],x[18],x[19],x[20],x[21],x[22],x[23]]))) \
-#         .toDF("app_list", "level2_ids", "level3_ids","ucity_id",
-#              "ccity_name", "device_type","manufacturer", "channel", "top", "time", "hospital_id",
-#               "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time",
-#               "recover_time", "stat_date").write.csv('/recommend/native', mode='overwrite', header=True)
-#
-#     nearby_pre = spark.createDataFrame(rdd.filter(lambda x: x[6] == 1).map(lambda x: (x[3], x[4], x[5]))) \
-#         .toDF("city", "uid", "cid_id")
-#     print("nearby")
-#     print(nearby_pre.count())
-#     nearby_pre.write.csv('/recommend', mode='overwrite', header=True)
-#
-#     spark.createDataFrame(rdd.filter(lambda x: x[6] == 1)
-#                           .map(lambda x: (x[0], x[1], x[2], x[9], x[10], x[11], x[12], x[13], x[14], x[15],
-#                                           x[16, x[17], x[18], x[19], x[20], x[21], x[22], x[23]]))) \
-#         .toDF("app_list", "level2_ids", "level3_ids", "ucity_id",
-#                "ccity_name", "device_type", "manufacturer", "channel", "top", "time", "hospital_id",
-#               "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time",
-#               "recover_time","stat_date").write.csv('/recommend/nearby', mode='overwrite', header=True)
-#
-#     rdd.unpersist()
+def get_predict(date,value_map,app_list_map,level2_map,level3_map):
+    sql = "select e.y,e.z,e.label,e.ucity_id,feat.level2_ids,e.ccity_name," \
+          "u.device_type,u.manufacturer,u.channel,c.top,e.device_id,e.cid_id,cut.time," \
+          "dl.app_list,e.hospital_id,feat.level3_ids," \
+          "k.treatment_method,k.price_min,k.price_max,k.treatment_time,k.maintain_time,k.recover_time " \
+          "from esmm_pre_data e left join user_feature u on e.device_id = u.device_id " \
+          "left join cid_type_top c on e.device_id = c.device_id " \
+          "left join cid_time_cut cut on e.cid_id = cut.cid " \
+          "left join device_app_list dl on e.device_id = dl.device_id " \
+          "left join diary_feat feat on e.cid_id = feat.diary_id " \
+          "left join train_Knowledge_network_data k on feat.level2 = k.level2_id"
+
+    features = ["app_list", "level2_ids", "level3_ids","ucity_id", "ccity_name", "device_type", "manufacturer",
+                "channel", "top", "time", "hospital_id",
+                "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time", "recover_time"]
+    df = spark.sql(sql)
+
+    df = df.na.fill(dict(zip(features, features)))
+    df = df.drop_duplicates(["ucity_id", "level2_ids", "ccity_name", "device_type", "manufacturer",
+                             "device_id","cid,id","label",
+                             "channel", "top", "time", "app_list", "hospital_id", "level3_ids"])
+
+    rdd = df.select("app_list", "level2_ids", "level3_ids","ucity_id","device_id","cid_id","label", "y", "z",
+                    "ccity_name", "device_type","manufacturer", "channel", "top", "time", "hospital_id",
+              "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time",
+              "recover_time") \
+        .rdd.map(lambda x: (app_list_func(x[0], app_list_map), app_list_func(x[1], level2_map),
+                        app_list_func(x[2], level3_map), x[3],x[4],x[5],x[6],x[7],x[8],
+                            value_map[x[3]], value_map[x[9]],
+                        value_map[x[10]], value_map[x[11]], value_map[x[12]], value_map[x[13]],
+                        value_map[x[14]], value_map[x[15]], value_map[x[16]], value_map[x[17]],
+                        value_map[x[18]], value_map[x[19]], value_map[x[20]], value_map[x[21]],
+                        value_map[date]))
+
+    rdd.persist()
+
+    native_pre = spark.createDataFrame(rdd.filter(lambda x:x[6] == 0).map(lambda x:(x[3],x[4],x[5])))\
+        .toDF("city","uid","cid_id")
+    print("native")
+    print(native_pre.count())
+    native_pre.write.csv('/recommend', mode='overwrite', header=True)
+
+    spark.createDataFrame(rdd.filter(lambda x: x[6] == 0)
+                                   .map(lambda x: (x[0], x[1], x[2],x[9],x[10],x[11],x[12],x[13],x[14],x[15],
+                                                   x[16,x[17],x[18],x[19],x[20],x[21],x[22],x[23]]))) \
+        .toDF("app_list", "level2_ids", "level3_ids","ucity_id",
+             "ccity_name", "device_type","manufacturer", "channel", "top", "time", "hospital_id",
+              "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time",
+              "recover_time", "stat_date").write.csv('/recommend/native', mode='overwrite', header=True)
+
+    nearby_pre = spark.createDataFrame(rdd.filter(lambda x: x[6] == 1).map(lambda x: (x[3], x[4], x[5]))) \
+        .toDF("city", "uid", "cid_id")
+    print("nearby")
+    print(nearby_pre.count())
+    nearby_pre.write.csv('/recommend', mode='overwrite', header=True)
+
+    spark.createDataFrame(rdd.filter(lambda x: x[6] == 1)
+                          .map(lambda x: (x[0], x[1], x[2], x[9], x[10], x[11], x[12], x[13], x[14], x[15],
+                                          x[16, x[17], x[18], x[19], x[20], x[21], x[22], x[23]]))) \
+        .toDF("app_list", "level2_ids", "level3_ids", "ucity_id",
+               "ccity_name", "device_type", "manufacturer", "channel", "top", "time", "hospital_id",
+              "treatment_method", "price_min", "price_max", "treatment_time", "maintain_time",
+              "recover_time","stat_date").write.csv('/recommend/nearby', mode='overwrite', header=True)
+
+    rdd.unpersist()
 
 
 def con_sql(db,sql):
@@ -235,8 +235,8 @@ if __name__ == '__main__':
     ti = pti.TiContext(spark)
     ti.tidbMapDatabase("jerry_test")
     spark.sparkContext.setLogLevel("WARN")
-    test()
-    # validate_date, value_map, app_list_map, leve2_map, leve3_map = feature_engineer()
-    # get_predict(validate_date, value_map, app_list_map, leve2_map, leve3_map)
+
+    validate_date, value_map, app_list_map, leve2_map, leve3_map = feature_engineer()
+    get_predict(validate_date, value_map, app_list_map, leve2_map, leve3_map)