Abstract
Analyzing topical user influence in online social networks is conducive to better advertisement injection, information dissemination, and user behavior analysis. In this paper, we propose a new approach to measure topical user influence in online social networks. Specifically, by comprehensively considering users' social relationships, posting and forwarding behaviors, and posts content, we define two metrics of user intimacy and social circle difference to measure how influential users rank on different topics. The dataset obtained from Sina Weibo is utilized to evaluate the effects of our proposed approach and several comparison methods. Specifically, the advantages of our approach are evaluated from several aspects, including the correlations between influence rankings on different topics, the spread ability of high influential users, and the spread balance of high influential users. The extensive experimental results show that our approach is superior in mining influential users on different topics. Specifically, in our approach, the user influence rankings on different topics are less correlated, the users with high influence rankings achieve higher spread scope, and the higher a user's influence ranking, the more evenly the user's posts being spread among different communities.
Highlights
Along with the expeditious development of Internet, online social networks are growing rapidly and becoming main platforms for people making friends and sharing information
In this paper we propose a new approach iSCD which considers the user intimacy and social circle difference, to measure topical user influence and discover high influential users on different topics in online social networks
We take into account the topology structures of social networks, the user behaviors of posting and forwarding information, and the post content to evaluate topic-based user influence
Summary
Along with the expeditious development of Internet, online social networks are growing rapidly and becoming main platforms for people making friends and sharing information. The number of followers and other information in online social networks have been utilized to measure user influence by some researchers. Liu et al [8] and Bi et al [9] proposed an improved LDA model [10] by considering users’ posts and relationships in social networks, the topical user influence is obtained through model training. We propose a new approach to measure topical user influence, which comprehensively takes into account users’ post contents, forwarding behaviors and the community structures of social networks.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
