SRSP-PMF: A Novel Probabilistic Matrix Factorization Recommendation Algorithm Using Social Reliable Similarity Propagation

Recommendation systems have received great attention for their commercial value in today’s online business world. Although matrix factorization is one of the most popular and most effective recommendation methods in recent years, it also encounters the data sparsity problem and the cold-start problem, which leads it is very difficult problem to further improve recommendation accuracy. In this paper, we propose a novel factor analysis approach to solve this hard problem by incorporating additional sources of information about the users and items into recommendation systems. Firstly, it introduces some unreasonable prior hypothesises to the features while using probabilistic matrix factorization algorithm (PMF). Then, it points out that it is neccesary to give two new hypothesises about conditional probability distribution of user and item feature and buliding some concepts such as social relation, social reliable similarity propagation metrics, and social reliable similarity propagation algorithm (SRSP). Finally, a kind of a novel recommendation algorithm is proposed based on SRSP and probabilistic matrix factorization (SRSP-PMF). The experimental results show that our method performs much better than the state-of-the-art approaches to long tail recommendation.