The Collaborative Filtering Algorithm Based On User-item Double Clustering

”Information Overload” makes us difficult to obtain effective information,.The personalized recommendation algorithm is a solution to this problem.In the recommendation algorithm,the collaborative filtering algorithm is the best development and the most widely used algorithm.First,the collaborative filtering algorithm finds similar neighbors by analyzing user attributes.Then the algorithm provide recommendations for the target user by using these similar users(or project)rating.But there has some problem in he collaborative filtering algorithm,such as sparseness problem,because the project dimension is too high and the user's score data is too small;User interest changes,User interest is always changing;The cold start problem,because of new users and new project information is too small,the algorithm can not provide services for them.These problems lead to a lower accuracy of the recommended results for collaborative filtering algorithms.Aiming at these problems,this paper proposes a new algorithm,the collaborative filtering algorithm based on user-item double clustering.These algorithm presents some solutions to these problems based on the previous study.To solve the problem of data sparsity,the algorithm use the result of user-item double clustering to fing neighbors of the target.Then fill the blank rating item by using the data of these neighbors.The data obtained in this way is credible.To solve the problem of user interest changes,The algorithm uses the Ebbinghaus forgetting curve to simulate the process of user interest with time.When we calculate the similarity between users,we weighted the score by time.In this way,the similarity we get is more in line with the user's current situation.To solve the problem of the cold start problem,two new similarity calculation methods are used in the algorithm,fusion of user attributes for user similarity and fusion of item attributes for item similarity.New users and new projects can also find neighbors by using the user attributes and the project attributes.At last,we give the experimental evaluations and analysis of the collaborative filtering algorithm based on user-item double clustering.In this experimental we using dataset Movie Lens.In the experiment process,the improvement of the algorithm is carried out.The result of the experiment shows that the new algorithm has improved recommended accuracy compared to the traditional collaborative filtering algorithm.