基于隐含信任关系的概率张量分解推荐算法

Recommendation Based on Probabilistic Tensor Decomposition with Implicit Trust Relation

上下文信息以及用户之间的信任关系在推荐系统中起着重要作用。现有的推荐算法存在以下两个主要问题:只是利用上下文信息对数据进行预处理,只考虑了用户之间的直接信任关系对推荐系统的影响。针对上述问题,提出了一种基于隐含信任关系的概率张量分解推荐算法,将上下文融入到模型中的同时还考虑了用户之间的隐含信任关系对推荐系统的影响。实验结果表明,提出的算法提高了推荐精度。

Contextual information and the trust relation play important roles in the recommender system.However,existing recommendation algorithms have two obvious problems：these methods only exploit context information for data preprocessing and only take the explicit trust relation.Aiming at handling the problems above,an enhanced method was proposed.This method not only integrates the context into the model,but also considers the implicit trust relation on the recommender system.The experimental results show that the proposed method improves the accuracy of recommendation.