摘要
针对用户兴趣迁移与用户对推荐行为无自主控制两个典型问题,提出了一种新颖的个性化协同过滤推荐策略。通过一种信任管理机制,用户自己管理信任关系。各用户拥有一个列表用来存放信任用户以及信任关系程度,通过用户自主行为可以控制系统的推荐。在推荐过程中,将信任度与相似度合并,得到复合权值,并以此进行推荐。同时,通过比较实际评价值与期望评价值,实时调整信任关系程度。实验结果表明,该策略提高了系统推荐的准确率,一定程度上增强了用户的信心。
A novel personalized CF recommendation strategy is proposed,which is used to deal with the issues of transfer of user profiles and without control of recommendation.On one part,a mechanism is introduced for user to manage his own trust relationship,which could increase user confidence for the system.A trust table is adopted for a single user to keep his own trust neighbors,trust degree is changed or viewed.On another part trust value is used as a complementary factor to user similarity,which makes the recommendation more accurate.At the same time,according to the analysis of diversity between the rare rating and the expectation,the degree of trust relationship is changed adaptively.The experiment shows that the recommendation method has a better performance than traditional CF method,and consumer confidence is strengthened.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第2期566-569,共4页
Computer Engineering and Design
基金
中央高校基本科研业务费专项基金项目(XDJK2010C037
SWU1009021)
关键词
协同过滤
推荐系统
信任度
信任管理
兴趣迁移
collaborative filtering
recommender system
trust value
trust management
transfer of profile
