The Research On Collaborative Filtering Recommendation Algorithm Based On Score Contribution

The development of information technology and Internet data of explosive growth,People have to spend a lot of time and effort to search and browse the information they need with so much data on the Internet.Although the traditional search engine,information retrieval technology has been able to meet people's demand,but in the face of the query needs of different users in different backgrounds,different period,different purposes,the traditional information retrieval technology seemed to be stretched.Personalized recommendation technology is to solve this problem and is proposed.It is different users to provide different services,to meet the needs of each user different and personalized recommendation technology by collecting and analyzing users' information to learn user's interests and behavior,in order to achieve the initiative for every user recommended.Personalized service technology can improve the quality of service and access efficiency of the site,so as to attract more visitors.In recent years,more and more research on personalized recommendation technology,the recommendation algorithm based on collaborative filtering both in academia and industry have been vigorous development.Based on the comparative analysis of the similarity algorithm commonly used recommendation algorithm,points out the algorithm application scenarios and their limitations.Based on the traditional similarity algorithm,a novel neighborhood generation method is proposed,and the collaborative filtering recommendation algorithm based on the score contribution is proposed.Based on score with the recommendation algorithm innovation are proposed to score contribution coefficient concept,the establishment of sparse processing model in order to establish the neighborhood generation method.At the same time,the weighted average method predict the score,using MAE and RMSE method of accuracy of measurement.In order to verify the score with recommendation algorithm effectiveness of rating data prediction problem,this paper adopts site movielens data set of algorithm in the experiment,respectively,the neighborhood generation method based on user,based on project and mixed user item based collaborative filtering recommendation algorithm and with the traditional neighborhood generation method were compared and analyzed.The experimental results show that the evaluation method is applied to the evaluation of the project based on the project,the user and the user project.Through the experimental analysis,the collaborative filtering recommendation algorithm based on the score contribution in this paper is a reference to the similarity computation of the recommendation algorithm and the sparse processing of the score data.