A New Hybrid Clustering Algorithm for Improving Results of Recommender Systems

Recommender systems are used to recommending interest items to users. A widely used recommendation technique in recommender system is collaborative filtering. This technique, assumes that users, who share the preferences on some items, share these preferences on the other items. Clustering methods can be used for collaborative filtering technique. In this paper, a new hybrid clustering method is presented to improve the recommender system results. The proposed method utilizes both user profiles and user-item rating matrix as its information sources. Moreover, a new heuristic method is presented to ensemble clusters. K-means method is used as the clustering method. Then, the set of items will be recommended to the new user based on its detected ensemble cluster. The results of experiments on MovieLens dataset show that the proposed method enhances the efficiency of recommender systems