Research On Collaborative Filtering Recommendation Algorithm Adapting To User Interest Changes

With the rapid development of computer technology,the Internet,and mobile communications,the amount of information in people's lives has been increasing exponentially.It has brought convenience to people and also caused “information overload”.How to find out what someone need from these massive information has become a research hotspot under the current big data situation.Because the recommendation system can effectively extract the content required by users from the massive information and is popular with users,it has been applied to various fields.With the rise of the recommendation system,the corresponding recommendation algorithm is also highly valued by the people from the business community and the academia.At present,the recommendation system that sees collaborative filtering recommendation algorithm as the core is the most widely used.The main research content of this paper is to compose the Propagation Algorithm and Fuzzy C-Means Clustering Algorithm(AP-FCM)and the time weight function to reduce the influence of the data sparsity and user interest change over time on the traditional collaborative filtering advancing algorithm.First,we use AP-FCM clustering algorithm to reduce the data sparse proportion.Because the Fuzzy C-Means Clustering Algorithm(FCM)has main two defects that the cluster center is unstable and the number of clusters needs to be set artificially,and the affinity propagation cluster algorithm(AP)can automatically generate the centers and the number of clusters,we combine the advantages of two algorithms and proposed the AP-FCM clustering algorithm.The experimental results show that the AP-FCM algorithm is not only better than the traditional FCM clustering algorithm in the selection of cluster centers,but also can effectively reduce the data sparse proportion.Second,introduce the time weight function.The scores for different time periods are given different weight values for attenuation to reflect that the user's interest changes with time.The experimental results show that the time weight function can effectively reflect the changes of user's interest over time.Finally,we combine the AP-FCM clustering algorithm and the time weight function,and propose a collaborative filtering recommendation algorithm based on AP-FCM and time weight function.The algorithm is compared with the traditional collaborative filtering recommendation algorithm on the data set Movie Lens.The results show that the collaborative filtering recommendation algorithm based on AP-FCM and time weight function is superior to the traditional collaborative filtering recommendation in accuracy.