update feature

train/train_service.py

@@ -72,11 +72,29 @@ def loadData(data_path):
     print("读取数据耗时ms:{}".format(timestmp2 - timestmp1))
     return df
 
+def getWeight(x):
+    res = 1
+    try:
+        p = int(x)
+        if p > 0 and p <= 5:
+            res = 2
+        elif p > 5 and p <= 10:
+            res = 3
+        elif p > 10:
+            res = 4
+
+    except Exception as e:
+        print(e)
+
+    return res
+
 
 def getDataSet(df,shuffleSize = 10000,batchSize=128):
     # print(df.dtypes)
     labels = df.pop('label')
-    dataSet = tf.data.Dataset.from_tensor_slices((dict(df), labels)).shuffle(shuffleSize).batch(batchSize)
+    df["page_stay"] = labels["label"].map(lambda x:1)
+    weights = df.pop('page_stay')
+    dataSet = tf.data.Dataset.from_tensor_slices((dict(df), labels,weights)).shuffle(shuffleSize).batch(batchSize)
     return dataSet
 
 def getTrainColumns(train_columns,data_vocab):
@@ -182,8 +200,9 @@ if __name__ == '__main__':
     splitTimestamp = int(time.mktime(time.strptime(addDays(-1), "%Y%m%d")))
 
     # redis中加载数据字典
-    print("redis 中加载模型字典...")
+    print("加载模型字典...")
     data_vocab = json.load(open(configUtils.VOCAB_PATH,mode='r',encoding='utf-8'))
+    print("字典keys:",str(data_vocab.keys()))
     # data_vocab = getDataVocabFromRedis(VERSION)
     assert data_vocab