Research On Personalized Recommendation Method Based On Improved Collaborative Filtering

The rapid development of the Internet has made the acquisition of information become more and more convenient,but it also has made it difficult to filter out useful information for users.Personalized recommendation system came into being under the background of multiple and complex information.Since the collaborative filtering method has high recommendation accuracy and is easy to implement in engineering,it is widely used in the recommended field.However,the existing collaborative filtering methods face data sparsity problem and ignore the influence of user's interest change trends,users' interest preference scope and time on the recommendation result.As a result,a collection of items recommended fail to meet users' needs.This paper focuses on the relationship between the trends of users' interest,the scope of users' interest preferences and the impact of time on the collaborative filtering methods.In order to improve the recommendation accuracy and diversity,we proposed two improved collaborative filtering methods.The main work of this paper is as follows:(1)In order to finely describe the changing trends of users' interests,this paper firstly constructs a user-user covariance matrix to replace the user-user similarity matrix used in the traditional collaborative filtering methods.And then we propose a collaborative filtering method based on user-user covariance matrix,which improves the diversity of recommendation under the premise of ensuring recommendation accuracy.(2)In order to consider the users' interest preference more comprehensively,this paper integrates the interest preference scope and time into traditional collaborative filtering methods,which improves the traditional nearest neighbor recommendation method.And then we propose an improved collaborative filtering method based on the users' preference scope and time.The proposed method integrates time into the quantifying the scope users' preference and defines the inclusion relationship between users based on the scope of interest preference,which improves the accuracy and diversity of recommendations.(3)In addition,this paper conduct extensive experiments on the public available datasets to evaluate the improved collaborative filtering methods based on user-user covariance matrix and users' preference scope.We can see from experimental results that the two improved collaborative filtering methods have obvious improvements on recommendation performances.