Hot And Personalized Followee Recommendation On Microblog

Microblog as a new hot killer SNS (Social network service) application, recently,itattracted lots of scholar to study and microblog recommendation is one of mainstreamresearch. Due to the special of microblogging network structure, Traditionalrecommendation algorithm is only used for recommendation can lead to recommendationimprecise problem, while the lost of considerations about user abstrutes can effectrecommendation quality. This paper analyzes the characteristics of microblog, focusing onthe research of microblog followee recommendation algorithm, mainly analysis andresearch the hot followee recommendation and personalized followee recommendation.First of all, hot followee recommendation. This paper analyzes the micro-bloginformation dissemination, select five factors to describe user’s behavior patterns in orderto identify active users, than according to the user’s reputation and microblog influence inthe design of hot user identification method,at last hot users will eventually be highlyrecommended as hot followees.Secondly, personalized followee recommendation. Aim at solving the problemofinaccurate prediction score, proposed a recommendation algorithm based on userattributes and bayesian classification. Firstly, according to the traditional user-basedcollaborative filtering algorithm to calculate the initial score for target users,secondlycombine weibo user attributes and bayesian classification to get the probability that targetthe project, last combine this two factor mprove prediction score.and compare thisalgorithm with traditional recommendation algorithm.Finally, the experimental process is designed on Eclipse to verify and analyze the twopoint contents in this paper. Mainly detail analyze the experimental results, verified theeffectiveness of the identification of hot user and the rationality of personalized followeerecommendation based on Bayesian classification.