使用贝叶斯网络的社交网络信任驱动推荐方法  

Trust-driven recommendation method using Bayesian network in social networks

作　　者：王东[1,2] 陈志[1] 岳文静[3] 刘亚威[1] WANG Dong;CHEN Zhi;YUE Wenjing;LIU Yawei(College of Computer, Nanjing University of Posts and Telecommunications, Nanjing 210023, China;Department of Information Engineering, Nantong College of Science and Technology, Nantong, Jiangsu 226007, China;College of Telecommunication and Information Engineering, Nanjing University of Posts and Telecommunications,Nanjing 210003, China)

机构地区：[1]南京邮电大学计算机学院,南京210023 [2]南通科技职业学院信息工程系,江苏南通226007 [3]南京邮电大学通信与信息工程学院,南京210003

出　　处：《计算机工程与应用》2017年第9期146-151,共6页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61501253;No.60905040);中国博士后科学基金项目(No.2013M531393);江苏省"六大人才高峰"第十一批高层次人才选拔培养资助项目(No.XXRJ-009);江苏省基础研究计划(自然科学基金)项目(No.BK20131382;No.BK20151506);江苏省博士后科研资助计划项目(No.1102102C);江苏省2014年度普通高校研究生实践创新计划项目(No.SJZZ_0107)

摘　　要：为处理推荐行为来源复杂、路径多样、不信任陌生推荐等问题,提出一种在社交网络中信任驱动推荐方法。该方法利用贝叶斯网络,计算用户评分的先验概率分布以及朋友之间的联合条件概率,预测用户在该环境下的评分并将推荐给用户。在信任驱动推荐过程中,预测评分既考虑到用户的偏好,也考虑到用户的社会关系;此外,用户的信息交换只限于朋友之间,能够有效保护用户的隐私。实验结果表明,所提出的推荐方法在预测准确率和推荐覆盖率上具有良好的性能。To deal with the problems such as the source intricacy of recommendation behaviors,path diversity,distrusting strangers’recommendation,a trust-driven recommendation method in social networks is proposed.The method uses Bayesian network,calculates the prior probability distribution of users’scores and the joint conditional probability between the friends,predicts users’ratings in the environment and provides the appropriate recommendations for the users.In the process of trust-driven recommending,rating prediction takes into account not only users’preference,but also users’social relationships;in addition,the information exchanges of the users are limited between friends,which can effectively protect the user privacy.Experimental results show that the proposed recommendation method has good performance on prediction accuracy and recommend coverage.

关 键 词：推荐系统 社交网络 贝叶斯网络 信任驱动 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]