Discovering Massive High-Value Users from Sina Weibo Based on Quality and Activity

Guangzhi Zhang,Rongfang Bie

DOI: https://doi.org/10.1109/CyberC.2013.42

2013-01-01

Abstract:This paper proposes a method to discover the high-value users who are both high-quality and high-activity. First, a Trust Transfer Model is introduced to capture users with high-quality in Sina Weibo that is a social web in China. Then, analysis and discussion are shown to identify the users are high-quality. Next, fresh users who are busy reposting are captured into the dataset filled with the high-quality users. Considering that reposting stands for the high-activity, hence the four degrees including reposting are proposed to judge one user's quality and activity. We discuss the effects of the four degrees. And then, the method called "WeiboRank" based on the four degrees is proposed to mine the high-value users. Finally, testing for the degree of coverage based on the "Top10 Hot Microblog" in Weibo presents a relatively high credibility of our dataset. The evaluation indicates that the users mined in this paper are quite high-quality and high-value. In conclusion, we will continue efforts to discover the high-value users in microblog and believe that discovering and maintaining a dataset filled with the high-value users which is not too big is quite significant for both academic research and business applications.