Unified Probabilistic Matrix Factorization Recommendation Algorithm Fusing Social Tagging

Traditional recommendation systems only use the users' rating information for the calculation and the recommendation. We can obtain the latent feature of the users or the resources to some extent but cannot get enough semantic interpretation which affects recommendation results. In order to solve this problem, we propose a neighborhood-aware unified probabilistic matrix factorization recommendation algorithm which combines social tags. First, we calculate the similarity between the users and the resources through the similarity of the tags to make neighborhood selection. Second, we construct a user-resource rating matrix, a user-tag tagging matrix and a resources-tag correlation matrix, and use the unified probability matrix factorization to get the latent feature vectors of three matrices to recommend by optimizing training model parameter. The experimental results show that the proposed algorithm can effectively use the semantics of the tags and improve the recommendation quality.