The Research Of Recommendation Technology Based On Collaborative Filter

As the most widely used recommendation technology which has achieved greater success currently, the collaborative filtering recommendation technology select users (or items) which have higher similarity with the target user(or item) as its nearest neighbors according to the target user’s(or item’s) accessed data and evaluated information. Then the target user’s(or item’s)predicted score can be obtained according to the scores of the neighbors and then items can be recommended to the target user. In practical applications,however, collaborative filtering recommendation is facing many problems such as score data sparse, cold start and poor algorithm scalability and so on.This paper focuses on collaborative filtering recommendation algorithm and then proposed two improved algorithms from a different perspective for the existing problems of the algorithm.On the one hand,cluster users using the users’background information with fuzzy clustering technology to improve the item similarity calculation method;on the other hand, make full use of user rating information, and highlight the special role of common user ratings.Specific research content of this paper includes the following three aspects:(1)Since the item similarity calculation method is not only inaccurate but also time complexity is high in terms of the high-dimensional sparse rating matrix,and therefore,the paper conduct fuzzy clustering on users using their back ground information,which considers the item similarity on each user group from the perspective of user groups,and then proposes a fuzzy clustering recommendation algorithm based on weighted item similarity.With the proposed algrithm, the accuracy of searching the item nearest neighbors and the quality of collaborative filtering recommendation algorithm can be improved on the condition of extremely sparse data and high user dimension.(2)User rating information reflects the user’s preference, but due to the extremely sparse rating matrix, how to make full use of the user ratings is particularly important for discovering use.interest and making recommendation.This paper divides user ratings into two types,and in the scoring matrix, the number of common user rating reflects the similarity between users to some extent.So the article will introduce user common rating factor into similarity calculation,which similarity calculation method is selected dynamically according to the common rating factors and a collaborative filtering recommendation algorithm integrated into co-ratings impact factor is proposed.The poposed algorithm makes full use of ratings data, and the score filled method based on singular value decomposition is used to improve the data sparsity and the quality of recommendation is improved as well.(3) For the proposed algorithms, the paper carries on some contrast experiment along with some traditional collaborative filtering algorithms in selected data sets. Experimental results show that the proposed algorithms can can effectively improve the quality of recommendation in the case of high dimension and sparse user ratings matrix.