The Microblogging Information Filtering Based On User Analysis

With the development of micro-blog, a new way of communication appears. The users publish and share information with others through micro-blog, then plenty of information grows up rapidly. In the case of information overload, the valuable information may be hidden. For this reason, information filtering problem has gradually become a hot topic. The users expect gain valuable information in a quick way, and then read or share.In this paper, we solve information filtering problem though two approaches named user classification and user influence. For user classification, we consider user behavior and users’ personal information. Users’personal information includes user profile, social tagging, authentication information, and micro-blog content. User behavior focuses on "@someone" and "#topic". By analyzing the different attributes for the user classification problem, we eventually propose a method based on multi-attribute fusion for user classification.For the study of the influence of users, we dig influential users based on different standards of measurement. First, we analyze the traditional measures including influence of fans, influence of comments, influence of retweets, and the influence on the number of micro-blog contents. Second, we consider the value of microblogging, the proliferation influence of message propagation, and activity level of the user in the micro-blog platform. We propose three new methods to show the influence of user, including tweet influence, behavior influence, and active influence. Third, we present a user influence model which fuses three new methods mentioned above. Finally, for different influence indicators, we analyze their internal relations, and describe the reasons among them.In this paper, we treat Sina Weibo as a carrier and achieve two approaches on the data set. Though experimental result, we analysis the contribution of user classification and user influence on information filtering problem.