摘要
协同过滤算法根据用户项目评分数据进行推荐,但评分数据通常很稀疏,使得用户无法获得满意的推荐,尤其是新用户。而信任网络以及社交网络能提供用户之间的关系数据,可用于推荐算法中。基于二值信任网络,提出GenTrust算法预测新的信任关系,扩展信任网络;并提出IndegreeTrust算法,区分被同一用户信任的所有用户。采用Epinions.com数据集,实验结果表明改进算法相比基于原始信任网络的算法准确率有所提升。
Collaborative filtering technique predicts items for users according to user-item marking data. However, the marking data is usually too sparse to make users, especially the new users, get satisfied recommendations. The trust-aware network or social network could be used to provide relationship data between the users, and is able to be used for recommendation algorithm. This paper, based on binary trustaware network, proposes GenTrust algorithm to predict new trust relationship in order to extend the trust-aware network, and IndegreeTrust algorithm to differentiate the users trusted by the same user. An evaluation on Epinions. com dataset shows that the improved algorithm has enhancement in its accuracy compared with the algorithm based on primitive trust-aware network.
出处
《计算机应用与软件》
CSCD
北大核心
2012年第12期157-160,共4页
Computer Applications and Software
关键词
推荐系统
协同过滤
二值信任网络
Recommender system Collaborative filtering Binary trust-aware network