Study On Collaborative Filtering Recommender Algorithm Based On Friend Relationship And Popularity

With the rapid development of Internet technology and the fast expansion of information resources, the problem of information overload is becoming more and more serious. Recommendation system is one of the most effective ways to solve the problem of information overload, and recommendation algorithm is the core of which. The recommendation method based on collaborative filtering is widely used, which forecasts the user’s preference on the basis of user-item rating data, and then makes recommendation for the user.This is paper, we introduce friend relationship of a user and social popularity of an item on the basis of collaborative filtering recommendation algorithm of SVD++ to improve recommendation accuracy of which, and the contributions of the paper is mainly in the following aspects:(1) we put a collaborative filtering recommendation algorithm based on friend relationship on the basis of SVD++. Firstly, the method measures familiarity between user and friend with Jaccard coefficient. Secondly, the impact on amount of resources rated in common by user and his friend which is handled with Shrink, is introduced in interest similarity calculation of them. Finally, the friend relationship strength is added to implicit feedback information of SVD++. We have carried out multi-group experiments on Epinions dataset, and the results show that the method has high performance.(2) According to Zipf’s law, most users prefer commodity with a higher popularity. Therefore, the social popularity information of an item is introduced to establish FPSVD++that based on the (1) in this paper, as an another implicit feedback information. The experimental results show that the proposed model improves the recommendation performance in further.