Research On The Collaborative Filtering Recommendation Algorithm Based On Social Network Analysis

With the rapid development of internet technology, e-commerce has been widelyused. E-commerce system is facing serious information overload problem whilebring the information transmission and information services. People can’t effectivefind the information they need with large amouts of information, so waste a lot oftime and experience. E-commerce system in order to provide satisfactory services,personalized recommendation system emerges as the personalized services.Personalized recommendation system’s purpose is to increase sales ability of theenterprise and attract customers, improve the quality of sevice.Recommendation system is the important research content of e-commercetechnology, has been widely used in the field of e-commerce. In numerouspersonalized recommendation technologies, the most widely used and mostsuccessful collaborative filtering is recommendation technology. But thecollaborative filtering recommendation algorithm is faced with many problems, suchas: data sparsity, cold start and low recommendation precision.In fully understand the principle of collaborative filtering recommendationtechnology and its problems, in view of the data sparse problem, and cold startproblems in the collaborative filtering recommendation technology, this paper putforward the clustering algorithm based on social network analysis. First of all,clustering of the item based on the item attributes, at the same time on therelationship between the users to condensing subgroup analysis based on socialnetwork analysis technology. Then combined with the user’ operation informationfor the item to clustering. In view of the recommendation accuracy problem, mainlyis to consider the user’s neighbors, combined with the analysis of the trustrelationship between users with social network analysis, the paper put forward thealgorithm to improve the calculation of similarity between users. First of all, throughthe analysis to determine trust value of trust relationship between users, consider theimpact on user behavior based on trust relationship between the users. With the trust value applied to the calculation of similarity, get more accurate neighbor to givemore precise recommendation results.Finally, this paper used the MovieLens datasets to test and verification. Verifythe nearest neighbor query efficiency after clustering, and the recommenderprecision. The experimental results show that the clustering algorithm based onsocial network analysis in nearest neighbor query efficiency is superior to thetraditional clustering algorithm, and recommender precision is superior to thetraditional collaborative filtering recommendation algorithm.