Research On Improved Collaborative Filtering Algorithm With Combination Of Social Network Analysis

With the rapid development of Internet technology, there exist a variety of WEBapplications. These applications provide users with more and more information and services,however a lot of information is flooded on the Internet—the so-called "information overload"problem——is becoming more and more serious. Many solutions to this problem come intobeing. As one of tools, recommendation system is more mature, better technology as so far.Recommender system is an important content of e-commerce technology and has beenwidely used in this field.Among the many personalized recommendation technology, themost widely used and most successful one is the collaborative filtering recommendationtechnologies. But collaborative filtering algorithm is also facing data sparsity, cold start andaccuracy issues. Based on user relationships, we construct a social network, whose nodes arefor users, edges for the relationship between them, then we propose a collaborative filteringalgorithm based on social network analysis. First, based on attributes of items and users,weuse cluster technology on them respectively, to solve the new item and new users problem.Then using social network analysis techniques to conduct subgroup analysis, mining two–steps-up user subgroups, based on the propagation theory, compute the direct and indirecttrust between users, finally we merge traditional similarity with user trust and the context ofuser——an improved similarity algorithm——to obtain a more accurate set of nearestneighbors to give reasonable and accurate recommendations.In the end, based on the improved algorithm, we conduct the related empirical analysis,and the results show that the proposed algorithm is better than the traditional collaborativefiltering algorithms in the accuracy of the recommendation...