Collaborative ?ltering is a widely adopted approach to recommendation, but sparse and high dimensional data are often barriers to providing high quality recommendations. Meanwhile, the traditional methods only utilize the information of the user-item rating matrix but ignore the trust relations between users, so their recommendation precision is often unsatisfactory. To address such issues, this paper constructs an user-preference matrix to reduce the data dimension and clusters the users by k-means clustering algorithm. Incorporating trust relationship, an improved similarity method is proposed to compute the similarity value. Then we find the nearest neighbor in the target user’s category according to the similarity; and predict the user’s prediction score by the nearest neighbor. At last we recommend the items with high prediction score to the user. This improved method has been tested via MovieLens 100K in order to make a comparison with the traditional techniques. The results have indicated that the proposed method can enhance performance of recommender systems.
展开▼