摘要
针对协同过滤算法在计算相似度时未考虑项目之间内在联系的问题,提出一种结合关联规则的协同过滤改进算法.改进算法首先应用Apriori算法,挖掘项目之间的强关联规则,尤其是两个以上项目之间的强关联规则;接着,将强关联规则过滤和拆分;最后将拆分后的强关联规则集成到相似度矩阵中.为适应时效变化和防止用户作弊,在计算相似度时,改进算法还考虑到项目的时效和用户可信度等因素.Movie Lens上的实验结果表明,改进算法提高了推荐系统准确性、有效性和适用性.
Collaborative filtering algorithm does not consider the inner relevance between projects when calculating item（user） similar- ity. In order to solve the problem, an improving collaborative filtering algorithm incorporating association rules is presented. Improving algorithm uses Apriori algorithm in the first place to mine strong association rules between projects especially strong association rules between three or more projects. Then, these strong association rules are filtered and split. Lastly, strong association rules after the split is integrated into the item similarity matrix. In order to adapt to change over time and prevent users from cheating, improving algorithm is also considering the factors such as time and user credibility level when calculating item similarity. The experiment result on the MovieLens shows that the algorithm can improve the accuracy, effectiveness and applicability of the recommendation results.
出处
《小型微型计算机系统》
CSCD
北大核心
2016年第2期287-292,共6页
Journal of Chinese Computer Systems
基金
广东省教育部产学研结合项目(2012B091100003
2012B091000058)资助
广东省专业镇中小微企业服务平台建设项目(2012B040500034)资助
关键词
关联规则
协同过滤
用户可信度
APRIORI
推荐算法
association rules
collaborative filtering
user credibility
Apriori
recommendation algorithm
