Research On Personalized Recommendation In Social Networks

With the expansion of the Internet scale and the widespread popularity of the mobile terminals in recent years, social networks such as Facebook, Twitter and Tumblr have quickly developed into every aspect of people’s lives, where people can enjoy posting messages, sharing releases, and keeping up with their friends’ updates anywhere and anytime. However, continuously expanding of users and rapid updating of information in social networks make the data into exponential growth. A large number of users find the information they received everyday contains not only their friends’ sharing, but also something that they are not interested in, which will become spam and submerge the interested part in some serious cases.How to provide personalized services to users in social networks have become the problems that the recommendation system faced with. Traditional recommendation methods have made some achievements, but they ignore the characteristics of the social networks and the users’ relationship in the process of recommendation when applied in social networks, making the results far from satisfactory.To solve the problems above, this paper studies the community structure of social networks, and focuses on the influence of user’s preference and relationship for personalized recommendation. According to social networks’ properties, a novel friends recommendation method based on the user’s preference and an improved collaborative filtering are proposed to make up for the deficiencies of existing recommendation methods in this paper. The simulation results show that the proposed methods are more suitable for social networks and have a higher accuracy.The works of this paper are mainly in the following areas:Based on the feature of the social networks, a new community discovery algorithm called CUPC(Community of User Preference Clustering) is presented.In order to make up for the shortcomings of the existing friends recommendation such as Friend of Friend and Mutual Friends, a new friends recommendation method based on CUPC is proposed, which takes both the users’ relationship and the users’ preference into consideration.Combined CUPC community discovery algorithm with collaborative filtering, an improved personalized recommendation method is put forward, which can reduce the scope of the nearest neighbors in the first stage of CF by clustering the users according to their preference and relations, improving the recommendation accuracy at the same time.Finally, the proposed personalized recommendation methods are applied to real social networks of MovieLens and Last.fm. The results of simulations show that the proposed methods in this paper are more suitable for personalized recommendation in social networks.