摘要： 传统协同过滤算法最后的预测值是用户最近邻评价的加权平均值，过于强调相似度的作用。除相似度以外，信任也是影响推荐结果的因素之一。该文提出以用户的评价个数和为他人提供推荐的次数为要素的可计算的信任模型与算法以及基于信任因子的协同过滤算法。该算法改变传统推荐过程中，用户之间的相似度唯一决定预测结果的现状，提高了推荐的精度。并通过一系列实验证明了该设想和算法的优越性。

关键词: 协同过滤, 信任, 相似度, 最近邻

Abstract: Traditional collaborative filtering algorithm is a weighted average prediction algorithm based on nearest neighbors’ ratings. This kind of predictive methodology only considers similarity between users while trust is also an important effective parameter in real life. This paper suggests that the traditional emphasis on user similarity may be overstated and there are additional factors having an important role to play in guiding recommendations, and trustworthiness of users must be an important consideration. It proposes computational model of trust and then a predictive algorithm based on factor of it. The experimental results prove the validity and superiority of the proposed algorithm.

Key words: collaborative filtering, trust, similarity, nearest neighbor

中图分类号: 

• TP301.6