A Microblog User Recommendation System Based On PSO Algorithm

With the rapid development of microblog application, microblog platform goes into the stage of infomation overload, and it becomes more and more difficult for user to obtain useful information and follow highly correlated users, because they are increasing everyday. Therefore an effective microblog user recommendation system has a key role to improve user experience. This dissertation is mainly focused on microblog crawler, evaluation of user influence and user clustering. The aim of this dissertation is to efficiently eliminate celebrity effect and zombie fans effect, and recommend highly correlated users.This dissertation presents a Microblog Crawler System based on Simulated Logining (MCSSL) to solve the problem caused by microblog login security mechanism and information types. Data standard and storage model are established according to application requirements. The useful information is extracted by MCSSL, which has higher accuracy rate and completeness than gathering by API.The user’s behaviors of microblog match the five principles of swarm intelligence. The user influence evaluation algorithm based on Particle Swarm Optimization (PSO) is proposed in this paper. Firstly, microblog data is statisticed including the numbers of comment, forward and so on. Secondly, the variable is defined as the velocity hange in our method. Thirdly, the infulence value will be iteratively calculated based on user’s behavior.This paper presents a K-means optimization algorithm based on property weight, and the microblog user recommendation system based on PSO is designed. Activity degree, intimacy degree and influence degree are taken into account to cluster relevant users by the proposed algorithm, which has ability to recommend results for center user. Experiment results show that the designed recommendation system has superior precision and recall than the recommendation system based on PageRank.