A Novel Collaborative Filtering Algorithm Based on Trust

This paper proposes a novel collaborative filtering method based on trust, which combines direct and indirect trust information to further improve the precision of recommendations. First, user behaviors and user trust propagation are utilized to generate direct and indirect trust. Second, overall trust is calculated based on the direct and indirect trust obtained. Third, by combining the overall trust and similarity with certain weights, trust similarity can be calculated. To verify the performance of the proposed algorithm, the dataset in Epinions is used, and the Mean Absolute Error (MAE) of the proposed algorithm and the traditional collaborative filtering method is compared. The results show that the proposed algorithm outperforms the traditional collaborative filtering algorithm in terms of prediction accuracy.