A Multi-attribute Probabilistic Matrix Factorization Model for Personalized Recommendation

Abstract

Recommender systems can interpret personalized preferences and recommend the most relevant choices to the benefit of countless users in the era of information explosion. Attempts to improve the performance of recommendation systems have hence been the focus of much research effort. Few attempts, however, recommend on the basis of both social relationship and the content of items users have tagged. This paper proposes a new recommending model incorporating social relationship and items’ description information with probabilistic matrix factorization called SCT-PMF. Meanwhile, we take full advantage of the scalability of probabilistic matrix factorization, which helps to overcome data sparsity as well. Experiments demonstrate that SCT-PMF is scalable and outperforms several baselines (PMF, LDA, CTR) for recommending.