Merge branch 'master' of git.wanmeizhensuo.com:ML/ffm-baseline

change main

eda/recommended_indexs/code/getRate.py

@@ -26,9 +26,9 @@ def result2file(fpath):
 	1.5 无点击用户占比(=无点击用户数/有曝光用户数)
 2.Top特征
 	2.1 用户点击次数分布(第一列：用户点击次数；第二列：独立用户数量)
-	2.2 Top 100 diary(sorted by ctr)
-	2.3 Top 100 Answer(sorted by ctr)
-	2.4 Top 100 Question(sorted by click times)
+	2.2 Top 100 diary (sorted by ctr)
+	2.3 Top 100 Answer (sorted by ctr)
+	2.4 Top 100 Question (sorted by click times)
 
 
 

eda/test/config.py

-DIRECTORY_PATH="/data2/models/eda/recommended_indexs/"
\ No newline at end of file
+DIRECTORY_PATH="/data2/models/eda/test/"
\ No newline at end of file

eda/test/getTopFeatures.py

@@ -24,8 +24,8 @@ class TopFeatures(object):
 
 	def get_click_times(self):
 		# rtype : dict
-		if self.cid_type[-2] == 'e':
-			self.cid_type = self.cid_type.replace(' ','')
+		if self.platform[-2] == 'e':
+			self.platform = self.platform.replace(' ','')
 		sql = "select cid,count(cid) from data_feed_click \
 		where from_unixtime(time,'%Y-%m-%d')=date_add(curdate(), interval -{0} day) \
 		and device_type{1} and cid_type='{2}' \
@@ -36,8 +36,8 @@ class TopFeatures(object):
 
 	def get_impression_times(self):
 		# rtype : dict
-		if self.cid_type[-2] == 'e':
-			self.cid_type = self.cid_type[:-6] + ' ' + self.cid_type[:-6:]
+		if self.platform[-2] == 'e':
+			self.platform = self.platform[:-6] + ' ' + self.platform[-6:]
 		sql = "select cid,count(cid) from data_feed_exposure \
 		where from_unixtime(time,'%Y-%m-%d')=date_add(curdate(), interval -{0} day) \
 		and device_type{1} and cid_type='{2}' \
@@ -97,7 +97,7 @@ class TopFeatures(object):
 			tplt = "{0:\u3000<4}\t{1:\u3000<12}\t{2:\u3000^6}\t{3:\u3000^6}\t{4:\u3000<8}\t{5:\u3000^15}\n"
 			f.write("Top {0} {1}\n".format(self.top_n,self.cid_type))
 			sep = "=================================================================\n"
-			header = tplt.format("平台","{}_id".format(self.cid_type),"点击数","曝光数","点击率","{}链接".format(self.cid_type))
+			header = tplt.format("平台","{0}_id".format(self.cid_type),"点击数","曝光数","点击率","{1}链接".format(self.cid_type,self.cid_type))
 			f.write(sep)
 			f.write(header)
 			for i in result_lst:
@@ -112,26 +112,68 @@ class TopFeatures(object):
 
 
 def main():
+
+	#1. Top diary
+	top_diary_all = TopFeatures(1, "all", "diary", 100)
+	clk_diary_times_all = top_diary_all.get_click_times()
+	imp_diary_times_all = top_diary_all.get_impression_times()
+	clk_diary_ctr_all = top_diary_all.get_result(clk_diary_times_all, imp_diary_times_all, 2, "ctr")
+
+	top_diary_ios = TopFeatures(1, "ios", "diary", 100)
+	clk_diary_times_ios = top_diary_ios.get_click_times()
+	imp_diary_times_ios = top_diary_ios.get_impression_times()
+	clk_diary_ctr_ios = top_diary_ios.get_result(clk_diary_times_ios, imp_diary_times_ios, 2, "ctr")
+
+	top_diary_android = TopFeatures(1, "android", "diary", 100)
+	clk_diary_times_android = top_diary_android.get_click_times()
+	imp_diary_times_android = top_diary_android.get_impression_times()
+	clk_diary_ctr_android = top_diary_android.get_result(clk_diary_times_android, imp_diary_times_android, 2, "ctr")
+
+	result_lst = [clk_diary_ctr_all, clk_diary_ctr_ios, clk_diary_ctr_android]
+	output_path = DIRECTORY_PATH + "top100_ctr_diary_{}.txt".format(get_yesterday_date())
+	top_diary_all.result2file(result_lst, output_path)
+
+	#2. Top answer
 	top_answer_all = TopFeatures(1, "all", "answer", 100)
 	clk_answer_times_all = top_answer_all.get_click_times()
 	imp_answer_times_all = top_answer_all.get_impression_times()
-	clk_answer_ctr_all = top_answer_all.get_result(clk_answer_times_all, imp_answer_times_all, 4, "ctr")
+	clk_answer_ctr_all = top_answer_all.get_result(clk_answer_times_all, imp_answer_times_all, 2, "ctr")
 
 	top_answer_ios = TopFeatures(1, "ios", "answer", 100)
 	clk_answer_times_ios = top_answer_ios.get_click_times()
 	imp_answer_times_ios = top_answer_ios.get_impression_times()
-	clk_answer_ctr_ios = top_answer_ios.get_result(clk_answer_times_ios, imp_answer_times_ios, 4, "ctr")
+	clk_answer_ctr_ios = top_answer_ios.get_result(clk_answer_times_ios, imp_answer_times_ios, 2, "ctr")
 
 	top_answer_android = TopFeatures(1, "android", "answer", 100)
 	clk_answer_times_android = top_answer_android.get_click_times()
 	imp_answer_times_android = top_answer_android.get_impression_times()
-	clk_answer_ctr_android = top_answer_android.get_result(clk_answer_times_android, imp_answer_times_android, 4, "ctr")
+	clk_answer_ctr_android = top_answer_android.get_result(clk_answer_times_android, imp_answer_times_android, 2, "ctr")
 
 	result_lst = [clk_answer_ctr_all, clk_answer_ctr_ios, clk_answer_ctr_android]
 	output_path = DIRECTORY_PATH + "top100_ctr_answer_{}.txt".format(get_yesterday_date())
 	top_answer_all.result2file(result_lst, output_path)
 
 
+	#3. Top question
+	top_question_all = TopFeatures(1, "all", "question", 100)
+	clk_question_times_all = top_question_all.get_click_times()
+	imp_question_times_all = top_question_all.get_impression_times()
+	clk_question_ctr_all = top_question_all.get_result(clk_question_times_all, imp_question_times_all, 2, "ctr")
+
+	top_question_ios = TopFeatures(1, "ios", "question", 100)
+	clk_question_times_ios = top_question_ios.get_click_times()
+	imp_question_times_ios = top_question_ios.get_impression_times()
+	clk_question_ctr_ios = top_question_ios.get_result(clk_question_times_ios, imp_question_times_ios, 2, "ctr")
+
+	top_question_android = TopFeatures(1, "android", "question", 100)
+	clk_question_times_android = top_question_android.get_click_times()
+	imp_question_times_android = top_question_android.get_impression_times()
+	clk_question_ctr_android = top_question_android.get_result(clk_question_times_android, imp_question_times_android, 2, "ctr")
+
+	result_lst = [clk_question_ctr_all, clk_question_ctr_ios, clk_question_ctr_android]
+	output_path = DIRECTORY_PATH + "top100_ctr_question_{}.txt".format(get_yesterday_date())
+	top_question_all.result2file(result_lst, output_path)
+
 
 if __name__ == '__main__':
 	main()