摘要
电子商务个性化推荐成为客户关系管理的重要内容,协同过滤算法是应用最为广泛的个性化推荐技术,但传统的协同过滤推荐算法并不适合用户多兴趣情况下的个性化推荐。在分析原因的基础上,通过组合基于用户的协同过滤和基于项目的协同过滤算法,先求解目标项目的相似项目集,在目标项目的相似项目集上再采用基于用户的协同过滤算法。这种基于相似项目的邻居用户协同推荐方法,能很好地处理用户多兴趣下的个性化推荐问题,尤其当候选推荐项目的内容属性相差较大时,该方法性能更优。最后,用EachMovie数据库对算法进行了仿真实验,实验表明该算法准确率更高。
Personalized recommendation for E-commerce has become increasingly important for customer relationship management, and the collaborative filtering algorithm is the most widely used method in personalized recommendation. But typical collaborative filtering algorithm is not suitable for user's multiple interest recommendation. A new algorithm which applied user-based collaborative filtering method based on similar item-set of target item by combing user-based collaborative filtering and project-based collaborative filtering was presented. The experiment conducted by applying EachMovie dataset indicated that new algorithm was more accurate, especially in the situation when users had many completely different interests.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2004年第12期1610-1615,共6页
Computer Integrated Manufacturing Systems
基金
教育部博士点基金资助项目(2000000601)
国家自然科学基金资助项目(70371004)。~~
关键词
协同过滤
个性化推荐
推荐系统
客户关系管理
collaborative filtering
personalized recommendation
recommendation systems
customer relationship management
