美购精排模型

mlp/train.py → models/train_service.py

@@ -21,7 +21,7 @@ one_hot_columns = ["item_"+c for c in ["service_type","doctor_type","doctor_famo
 version = "v1"
 data_path_train = "/data/files/service_feature_{}_train.csv".format(version)
 data_path_test = "/data/files/service_feature_{}_test.csv".format(version)
-model_file = "service_mlp_"+version
+model_file = "service/1"
 
 def is_float(s):
     try:
@@ -192,6 +192,10 @@ def predict(model_path,df):
 
 
 if __name__ == '__main__':
+
+    curTime = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
+    print("train_service执行时间:{}".format(curTime))
+
     # redis中加载数据字典
     print("redis 中加载模型字典...")
     data_vocab = getDataVocabFromRedis(version)

shell/train_service.sh

+source /srv/envs/serviceRec/bin/activate
+python /srv/apps/serviceRec/models/train_service.py > /srv/apps/serviceRec/models/train_service_log.log
\ No newline at end of file

spark/featureEng.py

@@ -593,7 +593,7 @@ if __name__ == '__main__':
     endDay = addDays(-1)
     startDay = addDays(-(1 + int(trainDays)))
 
-    print(startDay,endDay)
+    print("train_data start:{} end:{}".format(startDay,endDay))
 
     itemDF = get_service_feature_df(spark)
     # 行为数据
@@ -675,40 +675,4 @@ if __name__ == '__main__':
     splitAndSaveTrainingTestSamplesByTimeStamp(samplesWithUserFeatures, splitTimestamp, TRAIN_FILE_PATH)
     print("write to hdfs success...")
 
-
-
-
-
-    # # user画像数据
-    # # userDF = get_user_portrait(spark)
-    #
-    # # 数据处理
-    #
-    # # 数据写入
-    # item_csv_dir = "/service_item"
-    # user_csv_dir = "/service_user"
-    # action_csv_dir = "/service_action"
-    # os.system("hdfs dfs -rmr {}".format(item_csv_dir))
-    # os.system("hdfs dfs -rmr {}".format(user_csv_dir))
-    # os.system("hdfs dfs -rmr {}".format(action_csv_dir))
-    #
-    # itemDF.write.option("header", "true").option("delimiter", "|").csv(item_csv_dir)
-    # print("service_item write successful", flush=True)
-    #
-    # user_endDay = addDays(-1,format="%Y-%m-%d")
-    # user_startDay = addDays(-(1 + int(trainDays)),format="%Y-%m-%d")
-    # userTmpTable = "user_tag3_portrait"
-    # userDF.createOrReplaceTempView(userTmpTable)
-    # user_sql = "select * from {} where date between '{}' and '{}' ".format(userTmpTable,user_startDay,user_endDay)
-    # userDF = spark.sql(user_sql)
-    # userDF.write.option("header", "true").option("delimiter", "|").csv(user_csv_dir)
-    # print("service_user write successful", flush=True)
-    #
-    # actionTmpTable = "action"
-    # actionDF.createOrReplaceTempView(actionTmpTable)
-    # action_sql = "select * from {}".format(actionTmpTable)
-    # actionDF = spark.sql(action_sql)
-    # actionDF.write.option("header", "true").option("delimiter", "|").csv(action_csv_dir)
-    # print("service_action write successful", flush=True)
-
     spark.stop()
\ No newline at end of file