Trust-based Collaborative Filtering Recommendation in E-commerce

Collaborative filtering is the dominant techniques used by today's E-commerce recommender systems, but the sparsity problem of the user-item matrix is one of the main limitations of collaborative filtering. To deal with the sparsity problem of collaborative filtering, this paper proposes a trust-based collaborative filtering algorithm. This method uses users' ratings for items to calculate the direct trust between users, then based on trust inferences produces a trust matrix to find the nearest neighbors and make recommendations for a given user. Compared with traditional collaborative filtering, the proposed method can provide additional information to help alleviate sparsity. The experimental results show that the trust-based collaborative filtering algorithm can significantly improve recommendation performance.