摘要
协同过滤面临着用户评分数据极为稀疏的问题,为改善不同稀疏程度数据上的推荐效果,提出基于直接评分与间接评分的协同过滤算法。针对直接评分,定义加权用户相似性和加权项目相似性度量标准,构造直接推荐用户集合与直接推荐项目集合,给出直接评分权重的计算方法;针对间接评分,构造相似评分集合,定义评分相似性度量标准。定义综合评分权重,在直接评分与间接评分的基础上得到最终推荐结果。大量实验结果表明,该算法在不同稀疏程度的数据上均具有较高的推荐质量。
Existing collaborative filtering algorithms face challenge of extremely sparse users ’rating data.To improve the recommendation effects on datasets with different sparse levels,a collaborative filtering algorithm based on both direct rating and indirect rating was proposed.For direct rating,weighted user similarity and weighted item similarity were defined, direct recommendation user set and direct recommendation item set were constructed, and computation method of direct rating weight was presented.For indirect rating,similar rating set was constructed,and rating similarity was defined.Further-more,comprehensive rating weight was defined,which was used for producing the final recommendation results.Extensive ex-perimental results show that the proposed algorithm can achieve high recommendation quality on datasets with different sparse levels.
出处
《计算机工程与设计》
北大核心
2015年第5期1228-1232,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61105045)
北方工业大学科研人才提升计划基金项目(CCXZ201303)
关键词
推荐系统
协同过滤
直接评分
间接评分
综合评分
recommender system
collaborative filtering
direct rating
indirect rating
comprehensive rating
