基于混合信任模型的协同过滤推荐算法
2016-08-19彭玉



摘要:协同过滤技术在面临评分数据稀疏性问题时,推荐效果较差。为了提高推荐系统的推荐精度,可以将用户在社会网络中的交互信息加入到推荐系统中来,弥补评分数据不足的问题。在本文中,提出一种基于混合信任模型的协同过滤推荐算法,该算法将用户在社会网络中的历史评级信息和交互频率合并计算用户之间的直接信任度,并通过一定的信任传递规则将没有直接关联的用户间的间接信任度也计算出来,组成稠密的用户信任矩阵。最后基于用户信任矩阵和用户评分矩阵共同来计算用户之间的相似度并预测目标用户对未评分项目的评分,得到推荐集。实验结果表明本文提出算法可以提高数据密度,改善协同过滤推荐技术的稀疏性问题,有效提高推荐精度。
关键词:协同过滤;信任模型;基于信任的推荐算法;相似度;社会网路
中图分类号:TP31 文献标识码:A 文章编号:1009-3044(2016)20-0257-04
Abstract: In the face of rating data's sparsity problem, the recommendation effect of collaborative filtering technology is not so good. In order to improve the accuracy of recommendation system, we can add the interaction with the user information in social networks into the recommended system, in order to make up for the shortage of data rates. In this paper, we set up a hybrid trust model based on collaborative filtering algorithm, which combine the user rating information and interaction frequency in a social network, and calculate direct-trust-relationship between users. Following, the indirect-trust-relationship is also calculated by using some certain trust rules, and at last it forms a dense matrix of user trust. Finally, we calculate the similarity between users and predict the target ratings based on user-trust-matrix and user-scoring-matrix and get recommended set. Experimental results show that the proposed algorithm can increase data density, improve the sparsity problem of collaborative filtering technology and effectively improve the recommendation accuracy.
Key words:Collaborative filtering; Trust model; Trust-based recommender algorithm; Similarity;Social networks
1 引言
一个社交网站,比如脸谱网、微博,已经成为互联网用户不可或缺的一部分。信息技术和互联网技术的普及和迅猛的发展,加剧了信息过载问题,增加了用户获得有用信息的成本。推荐系统作为一种过滤技术,可以有效缓解信息过载问题,为用户提供优质的个性化服务[1]。在传统的个性化推荐算法中,协同过滤算法无疑是最成功的一种算法。协同过滤推荐算法是基于某些特定项目的用户之间的相似性偏好来预测推荐结果的。更一般而言,如果他们在某些项目上有类似的利益,他们最有可能对一些其他项目感兴趣。……
