A Novel Fine-Grained User Trust Relation Prediction for Improving Recommendation Accuracy

Recommender Systems (RSs) are playing an important role in improving user satisfaction as they can recommend items which might be highly interested to users. In recent years, it has been observed that social relations including factors such as trust and distrust among users are very useful in improving recommendation accuracy. However, traditional recommendation methods like Collaborative Filtering (CF) usually neglect social relations and even for those methods which considered social relations often fail to uncover different types of positive and negative social relations, which hinders the improvement of recommendation accuracy. To solve such a problem, in this paper, we first divide user trust relations into four fine-grained types, including strong trust, weak trust, strong distrust and weak distrust, which help to thoroughly exploit the trust and distrust relations among users. Afterwards, we propose a trust prediction framework based on a SVD algorithm to obtain weighted social relations. Finally, we employ two examples on rating prediction to demonstrate how to use fine-grained user trust relations. Experimental results based on Extended Epinions dataset show that our proposed approach based on fine-grained user trust relations can achieve better accuracy than other conventional approaches.