文献类型：期刊文献

中文题名：基于用户的协同过滤(UserCF)新闻推荐算法研究

英文题名：Study on News Recommendation Arithmetic Base on UserCF

作者：潘丽芳[1];张大龙[2];李慧[3]

第一作者：潘丽芳

机构：[1]长治医学院;[2]北京数字探索科技有限公司;[3]北京联合大学管理学院

第一机构：长治医学院,山西长治046000

年份：2018

卷号：32

期号：4

起止页码：26-30

中文期刊名：山西师范大学学报：自然科学版

基金：北京市自然科学基金资助项目(9164028)

语种：中文

中文关键词：基于用户的协同过滤算法;UserCF;用户相似度计算;新闻推荐;推荐算法

外文关键词：UserCF;news recommendation;similarity computation scheme

摘要：新闻更新快、易受流行和热门item的影响,用户兴趣也在不断的变化,因此就新闻推荐来说更应该关注其推荐的时效性.目前存在的用户相似度计算公式没有考虑时间因素,文中改进了原有用户相似度计算方法,增加了时间衰减因子,共同喜欢新闻i的用户u和v,产生行为的时间越远,这两个用户在新闻i上的相似度就会越小.文中提出的算法应用于http://www. show-ease. com网站的新闻推荐.通过收集7月4日到7月21日对推荐新闻的点击量和未推荐新闻的点击量,得出推荐新闻的点击量比未推荐新闻的点击量提高了31%到52%.文中提出的用户相似度计算方法在实践中取得了较好的效果.
With the development of information science,the news transmitted as an explosive way with the character of convenience on the internet,and it also update quickly with the change of user＇s interests and social focus. In other words,the timelines play a more and more important role in news recommendation. So the current similarity computation method,which haven＇t considered this yet,was urgent need to improve with the timelines. Here we further improve news recommendation algorithm with time decay factor,as implemented in the http：//www. show-ease. com（a website for news recommendation）,which means that when user u and user v reading the same news,the reading behavior was farther,the value of similarity between user u and user v will be smaller. By comparing the clicks with and without recommended news during 4 July to 21 July with total 17 days,we get the conclusion that the number of clicks was improved from 31 % to 52 % when with our recommendation strategy,which is further indicate that the news recommendation strategy,proposed in this paper,has achieved good effect in news recommendation.