| With the rapid development of data acquisition technology and storage technology, andthe popularity of Internet technology in the worldwide, various types of massive data areendless. Faced with these vast amounts of data, how to find useful information hidden in thesedata has become an urgent problem in the field of information. Recommending system usesthe technology of knowledge discovery to provide personalized recommendations, it is one ofthe most effective methods of filtering out overload information. So far, it has beensuccessfully applied to e-commerce systems, recommendation systems of books, movies, andvideo, and so on.Recently, despite recommendation system has been developed rapidly in theory andpractical, there are also some problems. With the surge in the number of users, expanding thescope of the system, there are insurmountable problems at cold start, recommendation qualityand real-time. To solve these problems, we have a useful exploration and research in one ofthe core issues in the recommendation system recommendation algorithm. The main contentsare as follows:(1) In this paper directly using two tripartite graphs weighted, the network has asecondary proliferation, we propose a new strategy based on tripartite graphs recommendationalgorithm, for sloving the problem of poor recommendation quality, the traditional tripartitegraphs lead to the loss of information.(2) Proposed recommendation algorithm based on N-partite graphs, and gives thetheoretical derivation and demonstration.(3) We improve the K-means clustering algorithm and apply it in customer segmentation,in order to overcome the problem of real time in traditional recommendation systems.(4) To solve the cold start problem, the improved K-means algorithm is applied toclassify the new users, combined with the recommendation algorithm based on tripartitegraphs; we propose a method of combinated cluster offline and online recommendation.Experiments demonstrate the effectiveness of the proposed method. |
PDF Full Text Request