Research On User Based Collaborative Filtering Recommendation Algorithm

With the rise of e-commerce and network communication,the Internet has become an important tool for people to get information and shopping.Resulting in explosive growth of data,that is,information overload.It is very difficult for users to find the information they need from the vast ocean of information on the Internet.Therefore,the recommendation system emerges as the times require.With the development of recommender system,various technologies have been applied to recommender systems.Collaborative filtering algorithm is the most widely applied and efficient algorithm.The "nearest neighbor" idea is a fundamental of user based collaborative filtering algorithm.The algorithm is based on the premise that a user and similar users have the same type of preference.The target users love project which the similar neighbors love,and the target user does not score or comment on them.User based collaborative filtering recommendation algorithm has some problems such as unstable user interestwhich lead to low recommendation accuracy,and inaccurate user relationship.In view of the above problems,this paper improves the user based collaborative filtering recommendation algorithm.The main work is as follows:In view of the data sparseness,and the traditional algorithm neglects the problem of the user's interest from the key words.A clustering algorithm combining the user interest is proposed.Using user data and project attribute data,we can calculate user preferences for keywords based on RF-IIF(Rating Frequency-Inverse Item Frequency)method,then K-means clustering.Then we use logistic function to get users' interest in projects,clear user preferences,find similar neighbors of target users in clusters,and recommend users' top-N projects to users.Finally,the Movielens dataset is used to test,and the experimental simulation results show that the algorithm improves the accuracy and efficiency of the recommendation.In view of the inaccurate measurement of user relations in traditional algorithms,an asymmetric similarity calculation method is proposed.With the SVD(Singular Value Decomposition)of matrix,using the number of potential features of the user,to caculate the proportion of the number of common scoring items that account for the total of all the scoring items of the user,resulting in asymmetric similarity between users.So as to clear the relative relationship between users.The Movielens data set is used for testing.The simulation results show that the average absolute error of the algorithm is always better than the traditional algorithm with the increase of the number of neighbors,and the accuracy of user relationship is judged accurately.The prediction score is closer to the actual score than the traditional algorithm.Finally,combined with the combination of user interest clustering method and asymmetric user similarity algorithm,an improved recommendation algorithm based on user collaborative filtering is proposed.Finally,the algorithm is compared with the Movielens data set.The experimental results show that the improved algorithm can improve the accuracy of the recommendation while alleviating the problem of data sparsity.