Trust prediction in online social rating networks

Abstract Previous research on trust prediction in online social rating networks focused on users’ history of interactions in evaluating the reputation of other users, service providers or products. Nevertheless, these approaches still suffer from malicious or inconsistent recommendations where the prediction problem turns out to be rather difficult. In this paper, we take up the challenge of coming up with a better solution by introducing a new global trust computation model that makes use of the recommendations made by what we call trusted parties in weighing users ratings. These entities gain higher reputation levels compared to others. Further, a sampling approach is proposed to capture new developments or changes during runtime based on a trust propagation technique. The proposed techniques are applied on the Epinion.com dataset. Empirical work shows the effectiveness of the proposed techniques in trust prediction compared to related work.