The Importment And Application Of Bi Partial Graph Personalized Recommendation

With the advent of the Internet era, the concept of "big data" has been more and more familiar with people. Being faced of massive data, provide users with fast and accurate services become more and more important. Recommendation technology is used to provide users with efficient information technology.With the rapid increase of the size of the recommendation systems’users, the system comes to the problem of users cold start, real-time and accuracy of shortage and other issues. In this paper, aiming at these problems, this paper is about the study of the Bi partial graph personalized recommendation algorithm,the specific studies are as follows:(1)For the Bi partial graph recommendation algorithm, user item connected edge weights are 1. The behavior of time and user behavior history data in frequency considered two aspects which determine to the weights. This paper proposed a specific method of edge weighted. Experimental results verify the effectiveness of the weighted strategy.(2)For the Bi partial graph recommendation algorithm, the number of users increases, resulting in reduced efficiency of recommendation algorithm and the problem of poor real-time. This paper Proposed the use of K-Medoids clustering algorithm to solve these problem. A cluster of user, thus the user segment, and then the subdivision user sub classes were recommended. The experimental results show that thanks to the clustering,the recommend efficiency is improved.(3) For the user cold start problem, using the method of SimRank to join the user in a subclass. In order to give new users what are recommended.Finally, the recommendation algorithm add to a UI application. This application is a WEB application for recommendation algorithm, experimental results show that it can improve the user experience effect. This system vertified the feasibility of the recommendation algorithm’s improvement.