Recommendation Algorithm Based On Nonnegative Matrix Factorization And Fuzzy Clustering

Thanks to the vigorous development of computer networks,we have lived in an era of information explosion.The advent of the information age has deepened the expansion of existing data.Some of the current recommended technologies and outdated information retrieval methods are not perfect To solve the problems related to big data.At the same time,users can not obtain their own target information data efficiently and accurately in the huge amount of data information in the Internet.Therefore,the recommender system is born.Nowadays,the collaborative filtering recommendation algorithm(CF)is at the core of the commonly used recommendation system.However,the Slope One algorithm performs well in many CF algorithms.It has the advantages of simplicity and high efficiency,and the recommended effect when the data is dense However,in practice,it is found that Slope One collaborative filtering recommendation algorithm also has some problems: sparseness and scalability issues,and can not be recommended when the data in the recommended system is sparse.In this paper,some improvements have been made to the above two aspects.By using fuzzy clustering technology,the potential information of users is captured,the accuracy of recommendation is promoted and the clustering mechanism is used to improve the scalability.At the same time,by using non-negative matrix factorization(NMF)Solve the problem of data sparse,the main work of this paper are as follows:On the one hand,according to the current Slope One recommendation algorithm,when the data is sparse,the recommendation effect is not good(that is,the sparseness problem).By using the method of nonnegative matrix factorization,the original user item scoring matrix is completed Dimension reduction and normalization to improve the sparsity of scoring matrix and improve the efficiency of the original Slope One algorithm.This paper proves this experimentally.On the other hand,the traditional Slope One algorithm combined with NMFtechnology,the further introduction of fuzzy clustering method,according to different users on the degree of preference of the project,the users are grouped into different user groups,while using membership degree of the weight coefficient To adjust deviations from different clusters,predict and complete the recommendation using the Slope One algorithm.This method to some extent to solve the problem of scalability.In the specific experiment,this paper compares the two algorithms with some traditional algorithms after the improvement,and the final experimental data also proves that the improved algorithm does have better recommendation accuracy and also has good MAE and MRSE values.