Research On Collaborative Filtering Recommendation System Based On Clustering Algorithm For Cold Start Problem

With the rapid development of information technology and the Internet,human society has entered a new information era.Although the continuous enrichment of the network world has brought many conveniences to us,it also leads to the problem of information overload.At present,the amount of Internet information has far exceeded the amount that we can deal with so that we have to spend a lot of time and energy to find the information we need.Under such a background,personalized recommendation system came into being.Collaborative filtering recommendation algorithm has become an important research field since 1990 s.Although this technology has achieved a lot of successful applications,there are also many problems need to be considerd.Such as the cold start problem,the data sparsity problem,the real-time response problem and the user privacy problem.In order to solve these problems,researchers proposed the collaborative filtering recommendation algorithm based on clustering.This algorithm clusters similar users or items into the same set which simplifies the process of finding nearest neighbors and reduced time consumption.Since the clustering process can be done offline,the real-time responsiveness of the recommendation system can be improved.This paper makes theoretical research on the recommendation system and clustering technology and proposes a collaborative filtering recommendation algorithm based on clustering.Experiments validate the effectiveness of this new algorithm.Here are the main contents of this paper:(1)This paper makes comparison of traditional similarity calculation methods and points out their shortcomings.A new similarity calculation method is proposed which involves user attribute,rating tendency and confidence degree.This new method improves the rationality and accuracy of similarity calculation and relieves the cold start problem to some extent.(2)We propose a new recommendation algorithm named FKCR which combines the new similarity caluculation with the K-means clustering algorithm.FOA algorithm is used to optimize the K-means algorithm.With the help of K-means and FOA algorithm,FKCR acquire higher efficiency and accuracy in finding similar neighbors.Finally,we verified in expriments the effect of the proposed algorithm in efficiency and accuracy under normal and cold start condition.