Top-N Trustee Recommendation with Binary User Trust Feedback

Ke Xu, Yi Cai, Huaqing Min, Jieyu Chen

Abstract: Trust is one of the most important types of social information since we are more likely to accept viewpoints from whom we trust. Trustee recommendation aims to provide a target individual with a list of candidate users she might be trust. However, most existing work on this topic focuses on the use of trusters’ interest but ignores the influence of trustees for recommendation. In this article, we propose a simple but effective method with the incorporation of both interest and influence of users for trustee recommendation based on binary user-user trust feedback. Specifically, we first introduce LDA twice on truster-documents corpus and trustee-documents corpus respectively to discover interest communities of users and influence communities of users. We then perform matrix factorization method on each community and finally design a merge method to rank the top-N trustees for a target user. Experimental results on Epinions dataset demonstrate that our proposed method outperforms other counterparts by large margins.