Research And Applications On Recommendation System Based On Cross-Domain

With the fast development of Internet, people are in a time of explosive growth of information. It’s more and more difficult to find out the right information from the information sea. Recommendation system is the most popular and effective technique to solve this problem. Lots of the Internet commpanys are using recommendation system, however, it’s hard to say satisfy to the result. The current recommendation system mainly faces challenges of the data sparsity problem, cold start problem and mass data processing, etc. To solve these problems, the paper finishes the work as following:(1) The paper studies the current mainstream recommendation algorithms deeply and analyzes the advantages and disadvantages of each recommendation algorithm, and thus lays the foundation for the study of cross-domain information recommendation system.(2) To solve the sparsity and cold start problem, the paper proposes two cross-domain recommendation algorithms. With the help of social network, the algroithm introduces the social intimate degree into reommendation system. The paper designs4experiments using the Baidu movie recommendation algorithm contest’s dataset. Experiments show that social and rating cross-domain recommendation system can improve the prediction acccuracy of system.(3) The paper uses Hadoop to design a parallel algorithm of the cross-domain model’s key algorithm in this paper. Experiments show that the parallel algorithm can improve the algorithm’s ability of handling large data.The paper gives a better way to solve the problem of the current recommendation algorithms, and have a good reference value in the cross-domain recommendation system.