Recommendation With Social Roles

Social recommender is an active research area. Most previous social recommenders adopt existing social networks to augment recommendations which are based on user preferences. In this contribution, we propose to simultaneously infer the social influence network and the user preferences in a matrix factorization framework. Furthermore, we assume that the influence strength is dependent on the social roles of users. We present an incremental clustering algorithm to detect dynamic social roles. Comprehensive experiments on real data sets demonstrate the efficiency and effectiveness of our model to generate precise recommendations.