Skip to content

F
ffm-baseline
Commit ecafcb00 authored by 张彦钊's avatar 张彦钊
Browse files
Options

增加日记是否是视频日记特征

parent 2594abf6
Hide whitespace changes
Inline Side-by-side
Showing with 19 additions and 15 deletions
eda/esmm/Model_pipline/feature_engineering.py
View file @ ecafcb00
......@@ -159,12 +159,14 @@ def feature_engineer():
sql = "select distinct recover_time from knowledge"
unique_values.extend(get_unique(db, sql))
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"
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=100)).strftime("%Y-%m-%d")
start = (temp - datetime.timedelta(days=1)).strftime("%Y-%m-%d")
print(start)
db = pymysql.connect(host='172.16.40.158', port=4000, user='root', passwd='3SYz54LS9#^9sBvC')
......@@ -177,15 +179,15 @@ def feature_engineer():
"channel", "top", "time", "stat_date", "hospital_id",
"treatment_method", "price_min", "price_max", "treatment_time", "maintain_time", "recover_time",
"app_list", "level3_ids", "level2_ids", "tag1", "tag2", "tag3", "tag4", "tag5", "tag6", "tag7",
"search_tag2", "search_tag3"]
"search_tag2", "search_tag3","is_video"]
unique_values.extend(features)
print("unique_values length")
print(len(unique_values))
print("特征维度：")
print(apps_number + level2_number + level3_number + len(unique_values))
temp = list(range(28 + apps_number + level2_number + level3_number,
28 + apps_number + level2_number + level3_number + len(unique_values)))
temp = list(range(29 + apps_number + level2_number + level3_number,
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," \
......@@ -193,7 +195,7 @@ def feature_engineer():
"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 " \
"e.device_id,e.cid_id,video.is_video " \
"from jerry_test.esmm_train_data_dwell 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 " \
......@@ -210,13 +212,14 @@ def feature_engineer():
"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 " \
"left join jerry_test.search_doris doris on e.device_id = doris.device_id and e.stat_date = doris.get_date " \
"left join jerry_test.diary_video video on e.cid_id = video.cid and e.stat_date = video.stat_date " \
"where e.stat_date >= '{}'".format(start)
df = spark.sql(sql)
df = df.drop_duplicates(["ucity_id", "level2_ids", "ccity_name", "device_type", "manufacturer",
"channel", "top", "time", "stat_date", "app_list", "hospital_id", "level3_ids",
"tag1", "tag2", "tag3", "tag4", "tag5", "tag6", "tag7"])
"tag1", "tag2", "tag3", "tag4", "tag5", "tag6", "tag7","is_video"])
df = df.na.fill(dict(zip(features, features)))
......@@ -224,7 +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","is_video","cid_id","device_id")\
.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),
......@@ -234,8 +237,8 @@ def feature_engineer():
value_map.get(x[16], 5), value_map.get(x[17], 6), value_map.get(x[18], 7), value_map.get(x[19], 8),
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]
value_map.get(x[25], 14), value_map.get(x[26], 15),value_map.get(x[29], 16)],
app_list_func(x[27], leve2_map), app_list_func(x[28], leve3_map),x[13],x[30],x[31]
))
......@@ -283,7 +286,7 @@ def get_predict(date,value_map,app_list_map,leve2_map,leve3_map):
"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," \
"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," \
"ot.tag as tag5,sixin.tag as tag6,cart.tag as tag7,doris.search_tag2,doris.search_tag3,video.is_video," \
"k.treatment_method,k.price_min,k.price_max,k.treatment_time,k.maintain_time,k.recover_time " \
"from jerry_test.esmm_pre_data e " \
"left join jerry_test.user_feature u on e.device_id = u.device_id " \
......@@ -299,13 +302,14 @@ def get_predict(date,value_map,app_list_map,leve2_map,leve3_map):
"left join jerry_test.sixin_tag sixin on e.device_id = sixin.device_id " \
"left join jerry_test.cart_tag cart on e.device_id = cart.device_id " \
"left join jerry_test.knowledge k on feat.level2 = k.level2_id " \
"left join jerry_test.search_doris doris on e.device_id = doris.device_id and e.stat_date = doris.get_date"
"left join jerry_test.search_doris doris on e.device_id = doris.device_id and e.stat_date = doris.get_date " \
"left join jerry_test.diary_video video on e.cid_id = video.cid and e.stat_date = video.stat_date"
features = ["ucity_id", "ccity_name", "device_type", "manufacturer",
"channel", "top", "time", "hospital_id",
"treatment_method", "price_min", "price_max", "treatment_time", "maintain_time", "recover_time",
"app_list", "level3_ids", "level2_ids", "tag1", "tag2", "tag3", "tag4", "tag5", "tag6", "tag7",
"search_tag2", "search_tag3"]
"search_tag2", "search_tag3","is_video"]
df = spark.sql(sql)
df = df.drop_duplicates(["ucity_id", "device_id", "cid_id"])
......@@ -316,7 +320,7 @@ def get_predict(date,value_map,app_list_map,leve2_map,leve3_map):
"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") \
"maintain_time", "recover_time", "search_tag2", "search_tag3","is_video") \
.rdd.repartition(200).map(lambda x: (x[0], float(x[1]), float(x[2]), x[3], x[4], x[5],
app_list_func(x[6], app_list_map), app_list_func(x[7], leve2_map),
app_list_func(x[8], leve3_map), app_list_func(x[9], leve2_map),
......@@ -330,8 +334,8 @@ def get_predict(date,value_map,app_list_map,leve2_map,leve3_map):
value_map.get(x[23], 9), value_map.get(x[24], 10),
value_map.get(x[25], 11), value_map.get(x[26], 12),
value_map.get(x[27], 13), value_map.get(x[28], 14),
value_map.get(x[29], 15)], app_list_func(x[30], leve2_map),
app_list_func(x[31], leve3_map)))
value_map.get(x[29], 15),value_map.get(x[32], 16)],
app_list_func(x[30], leve2_map),app_list_func(x[31], leve3_map)))
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment