融合用户相似度和信任传播重组信任矩阵算法

摘要

Aiming at the problems of Collaborative Filtering( CF) , such as data sparsity and cold start, an algorithm of reconstructing trust matrix is proposed in this paper, which integrates user similarity and weighted trust propagation.Specifically, the trust relation-ship of those users whose similarity falls below a certain threshold is removed firstly. Then the users of rating matrix is added into trust matrix when the similarity between the users exceeds a certain threshold. Finally, weighted trust propagation is considered, in order to incorporate more trusted neighbors as well as distinguish trusted neighbors in a shorter distance with those in a longer distance.Experimental results on FilmTrust and Epinions data sets show that the proposed method can achieve superior prediction accuracy and solve cold user problem better.%针对协同过滤面临的一些本质问题,如数据稀疏和冷启动,本文提出了融合用户相似度和加权的信任传播来重组信任矩阵的方法. 首先,将原始信任矩阵中用户相似度低于某一阈值的信任关系去掉;其次,将评分矩阵中用户相似度高于某一阈值的用户对添加到信任矩阵中;最后,考虑加权的信任传播,以此找到更多的信任邻居并对不同距离的信任邻居进行区分. 在Epinions和FilmTrust数据集上进行的对比实验结果表明,重组信任矩阵的方法能够有效地提高推荐精度,并在一定程度上解决了冷启动问题.