A Recommendation Technique Based on the Social Networks and Sequential Behaviors

Abstract

Collaborative filtering is one of the most successful method for recommendation. But it also has disadvantages such as cold-start problems and data sparseness which influence the result of recommendation. The probability Matrix Factorization Technique is one of the method of model based CF and it always ignore the relationship of users. Sequential behaviors is often one of the most important factor which is easy to be ignored. In this paper, in order to solving the problems of collaborative filtering, we not only consider the effect on the user rating matrix but also take the social networks and time factor into account. Putting these important factors in the PMF model is an effective way for making recommendation.