Social Recommendation with Interpersonal Influence

Social recommendation, that an individual recommends an item to another, has gained popularity and success in web applications such as online sharing and shopping services. It is largely different from a traditional recommendation where an automatic system recommends an item to a user. In a social recommendation, the interpersonal influence plays a critical role but is usually ignored in traditional recommendation systems, which recommend items based on user-item utility. In this paper, we propose an approach to model the utility of a social recommendation through combining three factors, i.e. receiver interests, item qualities and interpersonal influences. In our approach, values of all factors can be learned from user behaviors. Experiments are conducted to compare our approach with three conventional methods in social recommendation prediction.Empirical results show the effectiveness of our approach, where an increase by 26% in prediction accuracy can be observed.