Abstract: To resolve the problems of data sparsity and cold-start of the collaborative filtering algorithms in recommender systems, the trust information has being introduced to recommender systems, it can effectively alleviate these problems above by adding the users’ explicit trust information. But the explicit trust information is more difficult to obtain or sparse, in order to better improve the efficiency of recommendation, in the basis of TrustSVD algorithm based on explicit trust, this paper introduces the implicit trust information, and proposes a singular value decomposition (SVD) algorithm based on double trust mechanism, named EITrustSVD. At the same time of getting a reliable recommendation by using the explicit trust, the EITrustSVD algorithm gets a recommendation related to user performance through implicit trust. The experimental results show that the proposed algorithm is a better solution to the problem of cold-start, and has a better recommendation accuracy.

Key words: trust recommendation, collaborative filtering, singular value decomposition (SVD), recommender systems, implicit trust

摘要： 为解决传统协同过滤算法中存在的数据稀疏与冷启动问题，社会化信任推荐机制被引入推荐系统，通过加入用户的显式信任信息，可有效地缓解上述问题。但是显式信任较难获取，并且数据较为稀疏，为了更好地提高推荐效率，在基于显式信任的TrustSVD算法的基础上，加入用户的隐式信任信息，提出了一种基于双信任机制的奇异值分解（singular value decomposition，SVD）算法EITrustSVD。在利用显式信任获得可靠推荐的同时，通过隐式信任的影响获得与用户喜好相关的推荐。通过实验证明，该方法可以较好地解决冷启动问题，且能提高推荐的准确率。

关键词: 信任推荐, 协同过滤, 奇异值分解（SVD）, 推荐系统, 隐式信任