Trust Prediction Based on Interactive Relations Strength

Abstract

In online social network, trust is the basis of reliable interaction among users, and interaction relations also affect trust establishment. Although many researchers have studied approaches of trust model and prediction, most trust prediction methods are based on the existing trust network, and lack the in-depth study of user interaction and contents; therefore, it is not conducive to implement those trust prediction models, at the same time it also limits the scope of their applications. To deal with these issues, this paper presents a novel trust prediction framework based on both a trust network and the interactive contexts between users, and a kind of measurement mechanism is put forward to evaluate the strength of interaction relations. Combined with the existing trust network, a trust prediction threshold value is learned and used to predict unknown trust relations. Empirical experiments conducted on Epinions dataset show that the unknown trust relations can be effectively predicted combining with the user’s interaction behaviors, and the proposed method can improve the performance of the trust prediction model.