Abstract: At present, the large number of users has led to speed bottleneck problem, and the addition of new users has led to the cold start problem in collaborative filtering algorithm. A collaborative filtering algorithm is proposed based on clustering of the users trust and interests. The algorithm clusters the users based on their trust degree and scoring similarity. The experimental results show that, this method not only accelerates the results recommended, but also improves the accuracy of recommendation.

Key words: recommender system, collaborative filtering, clustering, users trust