摘要
针对传统协同过滤推荐算法的不足,依据现实生活经验,认为在协同过滤推荐过程中考虑用户的专家信任因素十分必要。详细阐述专家信任的概念以及利用用户评分数据计算专家信任度的方法,提出一种基于专家优先信任的协同过滤推荐算法。在公开数据集GroupLens上的实验结果表明,该算法预测用户评分的精度和成功率都明显优于传统的最近邻法。
In view of the limitations of traditional collaborative filtering recommendation algorithm and according to real life experience, this paper argues that integrating expert factor of user with collaborative filtering is necessary, elaborates on the concept of expert trust and its computing method by using rating data, and proposes a collaborative filtering algorithm based on expert prior trust. The experimental results on an open dataset named GroupLens show that, the algorithm in prediction accuracy and success rates are superior to the traditional nearest neighbor method.
出处
《图书情报工作》
CSSCI
北大核心
2012年第11期105-108,共4页
Library and Information Service
基金
北京高校图书馆科研基金项目"基于社会网络拓展图书馆服务领域的研究探索"(项目编号:BJT2011313)研究成果之一
关键词
协同过滤
信任度
最近邻
专家信任优先
collaborative filtering trustworthiness nearest neighbor expert prior trust
