Extraction Of Social Network Relations And Its Collaborative Recommendation Algorithm

With the development of information technology, the Internet is gradually replacingthe traditional books to become one of the most important information platform. Peoplecan get all kinds of information via the Internet. With the development of the Internet,social networking site has become one of the main ways for users to obtain information.User-generated content and content-sharing is the most important feature of the socialnetworking site. Users will spend a lot of time and effort to extract useful informationfrom social networking sites, and how to help users obtain valid information from themass of information is an urgent problem. In addition, the mobile Internet is booming, therapid growth of Internet applications make difficult for users to find valuable applicationsfrom various homogenous applications. Faced with these problems, personalizedrecommendation system is widely used to solve the information overload problem, andcollaborative filtering algorithm is the most widely used algorithm in the recommendationsystem.With the growth of users and recommend items, collaborative filtering algorithmrevealed some problems. First, the sparsity of the matrix. The similarity calculated basedon user ratings information influence by the sparsity of the matrix. Second, the scalability.With the growth of users and the number of recommended items, the system performancedecline quickly. Third, cold start problem. New Recommended items and new users areunable to effectively provide recommendations. In some case, the missing of user ratingsinformation let the collaborative filtering algorithm fail completely.In order to solve the above problems, the introduction of social network context, weproposed a new collaborative recommendation algorithm based on social networkrelations. First, extract social network relations and its strength. Get social networkassociations and mining associations intensity, to predict the strength of the relationshipbetween the members of the association. Then, the use of social network relationshipbetween the user to redefine the similarity between users, proposes a collaborative recommendation algorithm based on social network relations.The experimental results show the new collaborative recommendation algorithm caneffectively improve the similarity calculation and system scalability, and to a certaindegree against the sparsity of the matrix. In sparse matrix case, the new collaborativerecommendation algorithm is better than traditional collaborative filtering algorithms.