Abstract

Recommender systems have become de facto tools for suggesting items that are of potential interest to users. Predicting a user’s rat-ing on an item is the fundamental recommendation task. Tradi-tional methods that generate predictions by analyzing the user-item rating matrix perform poorly when the matrix is sparse. Recent ap-proaches use data from social networks to improve accuracy. How-ever, most of the social-network based recommender systems only consider direct friendships and they are less effective when the tar-geted user has few social connections. In this paper, we propose two alternative models that incorporate the overlapping commu-nity regularization into the matrix factorization framework. Our empirical study on four real datasets shows that our approaches outperform the state-of-the-art algorithms in both traditional and social-network based recommender systems regarding both cold-start users and normal users. 1.