摘要
内容消费性平台优化页面内容的排序具有十分重要的意义。分析捕捉用户在页面浏览时的点击和反馈行为,以用户的兴趣标签与点击内容的主题匹配定义用户的正向点击和反向点击,以用户对内容的偏好评价定义反馈,以逻辑回归模型为基准抽取用户历史的数据进行回归分析,用户实时的点击和反馈行为对内容推荐流中数据进行重排序。以Epinion数据集作为测试数据集,实验结果表明:新的算法比单一采用逻辑回归模型能更为明显地提升数据AUC。
To optimize the ranking of the page content for content consumption platform is very important.The users＇ clicks and feedback behavior when they are reviewing pages were analyzed and captured and the user＇ s forward and reverse click were defined and matched on their interest label and the theme of their click content.We defined the feedback on the user＇ s preference for content and had regression analysis of extracting the user＇ s history data based on logistic regression model,and we reordered the in the content recommendation flow by user clicks and feedback of real-time behavior.Taking Epinion data set as the testing data set,the experimental results show that the new algorithm can improve the data AUC more significantly than the single use of logistic regression model.
出处
《机床与液压》
北大核心
2016年第12期129-135,共7页
Machine Tool & Hydraulics
基金
supported by National Science and Technology Support Program (71102065)
Chongqing Basic Science and Research Project in Cutting-Edge Technologies (cstc2015jcyjA40049)
Chongqing Municipal Education Committee Science and Technology Project Support Program (KJ1403209)
关键词
点击和反馈
内容推荐
逻辑回归
重排序
AUC
Click and feedback
Content recommendation
Logistic regression
Reorder
AUC
