结合信任和用户关系的微博关注推荐算法  

Microblog followee recommendation algorithm combining with trust and user relationship

作　　者：王梦佳 贺智明[1] 郑剑[1] Wang Mengjia;He Zhiming;Zheng Jian(School of Information Engineering,Jiangxi University of Science ＆ Technology,Ganzhou Jiangxi 341000,China)

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《计算机应用研究》2018年第12期3560-3563,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61462034;61563019);江西省教育厅科学技术研究资助项目(GJJ13415);江西理工大学科研基金重点资助项目(NSFJ2014-K11);江西省研究生创新专项基金资助项目(YC2016-S316)

摘　　要：用户关注推荐在微博信息分享平台中扮演着非常重要的角色。现有的用户关注推荐模式大多采取内容及背景相似性度量用户距离,且多数使用单一因素进行排序推荐,没有针对用户群体进行分类,推荐效果一般。提出了一种针对非名人用户的新型关注推荐算法RTLR,该算法同时使用用户关系和互动行为信息进行推荐,并结合多因素建立逻辑回归模型。从国内用户最多的新浪微博中采集真实数据进行实验,结果显示新型模型有效地提高了关注推荐的准确性。Users followee recommendation plays an important role in information sharing over microblogging platforms. The majority existing followee recommendation schemes adopt either content or background information relevance to measure the user distance,and most use a single factor for followee ranking,there is no classification for user groups,suffering poor performance. Thus this paper proposed a novel followee recommendation algorithm RTLR for non-celebrity that took into user the information sources of both followee relationship and interaction behavior,and established logistic regression model combined with multi-factors. It conducted experiments on real-world datasets collected from Sina Weibo,the largest microblogging system in China. The results demonstrate that the novel scheme effectively improves the accuracy of followee recommendation.

关 键 词：信任度 用户行为 用户关系 关注推荐 逻辑回归 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]