增加日期特征，并把预测集的日期特征减少一天

tensnsorflow/feature_engineering.py

+
 import pandas as pd
 import pymysql
 import datetime
@@ -24,6 +25,7 @@ def get_data():
     print("validate_date:" + validate_date)
     temp = datetime.datetime.strptime(validate_date, "%Y-%m-%d")
     start = (temp - datetime.timedelta(days=60)).strftime("%Y-%m-%d")
+    pre_date = (temp - datetime.timedelta(days=1)).strftime("%Y-%m-%d")
     print(start)
     db = pymysql.connect(host='10.66.157.22', port=4000, user='root', passwd='3SYz54LS9#^9sBvC', db='jerry_test')
     sql = "select e.y,e.z,e.stat_date,e.ucity_id,e.clevel1_id,e.ccity_name," \
@@ -50,7 +52,7 @@ def get_data():
 
     unique_values = []
     features = ["ucity_id", "clevel1_id", "ccity_name", "device_type", "manufacturer",
-                "channel", "top", "time"]
+                "channel", "top", "time", "stat_date"]
     for i in features:
         df[i] = df[i].astype("str")
         df[i] = df[i].fillna("lost")
@@ -72,12 +74,10 @@ def get_data():
     value_map = dict(zip(unique_values,temp))
 
     df = df.drop("device_id", axis=1)
-    train = df[df["stat_date"] != validate_date]
-    train = train.drop("stat_date",axis=1)
-    test = df[df["stat_date"] == validate_date]
-    test = test.drop("stat_date", axis=1)
+    train = df[df["stat_date"] != validate_date+"stat_date"]
+    test = df[df["stat_date"] == validate_date+"stat_date"]
     for i in ["ucity_id", "clevel1_id", "ccity_name", "device_type", "manufacturer",
-                "channel", "top", "l1", "time", "l2"]:
+                "channel", "top", "l1", "time", "stat_date","l2"]:
         train[i] = train[i].map(value_map)
         test[i] = test[i].map(value_map)
 
@@ -89,7 +89,7 @@ def get_data():
     write_csv(train, "tr",100000)
     write_csv(test, "va",80000)
 
-    return validate_date,value_map
+    return pre_date,value_map
 
 
 def write_csv(df,name,n):
@@ -116,13 +116,14 @@ def get_predict(date,value_map):
                             6: "device_type", 7: "manufacturer", 8: "channel", 9: "top", 10: "l1",11:"l2",
                             12: "device_id", 13: "cid_id", 14: "time"})
 
+    df["stat_date"] = date
 
     print("predict shape")
     print(df.shape)
     df["uid"] = df["device_id"]
     df["city"] = df["ucity_id"]
     features = ["ucity_id", "clevel1_id", "ccity_name", "device_type", "manufacturer",
-                "channel", "top", "time"]
+                "channel", "top", "time", "stat_date"]
     for i in features:
         df[i] = df[i].astype("str")
         df[i] = df[i].fillna("lost")
@@ -139,7 +140,7 @@ def get_predict(date,value_map):
     nearby_pre = nearby_pre.drop("label", axis=1)
 
     for i in ["ucity_id", "clevel1_id", "ccity_name", "device_type", "manufacturer",
-                "channel", "top", "l1", "time","l2"]:
+                "channel", "top", "l1", "time", "stat_date","l2"]:
         native_pre[i] = native_pre[i].map(value_map)
         # TODO 没有覆盖到的类别会处理成na，暂时用0填充，后续完善一下
         native_pre[i] = native_pre[i].fillna(0)

tensnsorflow/pipe.sh

@@ -32,15 +32,15 @@ rm ${DATA_PATH}/nearby/nearby_*
 
 
 echo "train..."
-${PYTHON_PATH} ${MODEL_PATH}/train.py --ctr_task_wgt=0.9 --learning_rate=0.0001 --deep_layers=256,128,64,32 --dropout=0.3,0.3,0.3,0.3 --optimizer=Adam --num_epochs=2 --embedding_size=16 --batch_size=1024 --field_size=10 --feature_size=1460 --l2_reg=0.005 --log_steps=100 --num_threads=36 --model_dir=${DATA_PATH}/model_ckpt/DeepCvrMTL/ --data_dir=${DATA_PATH} --task_type=train
+${PYTHON_PATH} ${MODEL_PATH}/train.py --ctr_task_wgt=0.9 --learning_rate=0.0001 --deep_layers=256,128,64,32 --dropout=0.3,0.3,0.3,0.3 --optimizer=Adam --num_epochs=2 --embedding_size=16 --batch_size=1024 --field_size=11 --feature_size=1460 --l2_reg=0.005 --log_steps=100 --num_threads=36 --model_dir=${DATA_PATH}/model_ckpt/DeepCvrMTL/ --data_dir=${DATA_PATH} --task_type=train
 
 
 echo "infer native..."
-${PYTHON_PATH} ${MODEL_PATH}/train.py --ctr_task_wgt=0.9 --learning_rate=0.0001 --deep_layers=256,128,64,32 --dropout=0.3,0.3,0.3,0.3 --optimizer=Adam --num_epochs=1 --embedding_size=16 --batch_size=1024 --field_size=10 --feature_size=1460 --l2_reg=0.005 --log_steps=100 --num_threads=36 --model_dir=${DATA_PATH}/model_ckpt/DeepCvrMTL/ --data_dir=${DATA_PATH}/native --task_type=infer > ${DATA_PATH}/infer.log
+${PYTHON_PATH} ${MODEL_PATH}/train.py --ctr_task_wgt=0.9 --learning_rate=0.0001 --deep_layers=256,128,64,32 --dropout=0.3,0.3,0.3,0.3 --optimizer=Adam --num_epochs=1 --embedding_size=16 --batch_size=1024 --field_size=11 --feature_size=1460 --l2_reg=0.005 --log_steps=100 --num_threads=36 --model_dir=${DATA_PATH}/model_ckpt/DeepCvrMTL/ --data_dir=${DATA_PATH}/native --task_type=infer > ${DATA_PATH}/infer.log
 
 
 echo "infer nearby..."
-${PYTHON_PATH} ${MODEL_PATH}/train.py --ctr_task_wgt=0.9 --learning_rate=0.0001 --deep_layers=256,128,64,32 --dropout=0.3,0.3,0.3,0.3 --optimizer=Adam --num_epochs=1 --embedding_size=16 --batch_size=1024 --field_size=10 --feature_size=1460 --l2_reg=0.005 --log_steps=100 --num_threads=36 --model_dir=${DATA_PATH}/model_ckpt/DeepCvrMTL/ --data_dir=${DATA_PATH}/nearby --task_type=infer > ${DATA_PATH}/infer.log
+${PYTHON_PATH} ${MODEL_PATH}/train.py --ctr_task_wgt=0.9 --learning_rate=0.0001 --deep_layers=256,128,64,32 --dropout=0.3,0.3,0.3,0.3 --optimizer=Adam --num_epochs=1 --embedding_size=16 --batch_size=1024 --field_size=11 --feature_size=1460 --l2_reg=0.005 --log_steps=100 --num_threads=36 --model_dir=${DATA_PATH}/model_ckpt/DeepCvrMTL/ --data_dir=${DATA_PATH}/nearby --task_type=infer > ${DATA_PATH}/infer.log
 
 echo "sort and 2sql"
 ${PYTHON_PATH} ${MODEL_PATH}/sort_to_sql.py

tensnsorflow/test.py

@@ -29,7 +29,7 @@ def gen_tfrecords(in_file):
 
     for i in range(df.shape[0]):
         feats = ["ucity_id", "clevel1_id", "ccity_name", "device_type", "manufacturer",
-                "channel", "top", "l1", "time","l2"]
+                "channel", "top", "l1", "time", "stat_date","l2"]
         id = np.array([])
         for j in feats:
             id = np.append(id,df[j][i])

tensnsorflow/train.py

@@ -53,7 +53,7 @@ def input_fn(filenames, batch_size=32, num_epochs=1, perform_shuffle=False):
         features = {
             "y": tf.FixedLenFeature([], tf.float32),
             "z": tf.FixedLenFeature([], tf.float32),
-            "ids": tf.FixedLenFeature([10], tf.int64)
+            "ids": tf.FixedLenFeature([11], tf.int64)
         }
 
         parsed = tf.parse_single_example(record, features)