Collaborative Filtering Recommendation Methods And Applied Research

The appearance of Internet offer usrs a lot of information, but at the same time, it costs so much for users to get the effective information. Therefore, Internet company all want to use better recommendation methods to enhance their user experience of web applications, improve the quality of their service. Collaborative filtering recommendation methods are successfully and widely applied in Web 1.0 era. But with the emergence of Web2.0 technology, existing collaborative filtering recommendation technology is no longer applicable to the new network environment, to use the new data types. Therefore, this thesis base on correlation theory, starts with social network, reaserachs on the integration of community discovery and collaborative filtering recommendation algorithm method, using large amounts of data about users in social network, and provides theoretical and practical support for the recommendation service by the website.Firstly, related research background and significance are elaborated. Then, related researches are reviewed in twoaspects:collaborative filtering recommendation methods from three different angles, and the network community discovery methods, especially for the non-overlapping community discovery methods. Lastly main contents of this paper are presented.Then, the related concepts and theoretical basis are introduced. Mainly about social networks, social commerce, Girvan-Newman algorithm, K-cores algorithm, module maximization algorithm, users based collaborative filtering algorithm, item based collaborative filtering algorithms and other related concepts and methods.Based on the assumption that users in different communities show different preferences in social e-commence, a kind of collaborative filtering recommendation algorithm based on social-computing and users’contents is proposed. Community discovery methods are used in the social network, different kinds of datas provided by the web are applied in the progress of similarity calculation and recommended value calculation. According to the all above,using the datas which are obtained from the social network, we prove the algorithm proposed in the third chapter is effective from three aspects:community classification results, the differences of recommended results and the precision of recommended results. Result shows that, the proposed algorithm can effectively use the social network data and applied in the social e-commerce sites. At last, Research achievements are summarized and future researchdirections are also presented.