A Trust Enhanced Recommender System Based On Distrust Model

The accumulated users' social relationships in the social media, is an important available data source for relieving the Data Sparsity and Cold-Start problems in recommender system.Compared to the recommendation from strangers, people tend to receive the recommendation from friends or whom they trust. That is, people are much similar to their trustee. But the distrust relations also exists in large user community with the trust, and distrust is not the negative of trust. Distrust and trust are two different dimensions of social relationship, and they have different affect for users. Now, there already has some trust-aware recommender system incorporate the trust relationship into the recommendation process successfully, but there rarely has research about how to incorporate the distrust relationship.In this paper, we first based on the intuition that users' trust relationships is correlate with the interaction data between users, proposed a distrust predict model (disTPM) with users'interaction data which is review rating. Then, we proposed matrix factorization based distrust enhanced recommender system (MF-disTRS) that properly incorporates both distrust and trust relationships at the same time. The MF-disTRS method aims to improve the quality of recommendations and mitigate the data sparsity and cold-start issues. Besides, we also introduced the optimization way for the MF-disTRS method with the mini-batch SGD algorithm.At last, through the evaluation of MF-disTRS on the Extend Epinions dataset, the experimental result shows that distrust relationship can have positive effect in recommendation with respect to the accuracy compared to the standard trust-aware recommender systems. The result also proves that the distrust prediction model can predict the distrust between users accurately.