Understanding opinion leaders in bulletin board systems: Structures and algorithms

In this paper we conduct a measurement study on a campus bulletin board system (BBS). BBS users post/reply articles and form an interpersonal social network. Applying a social network analysis, we analyze the common characteristics of opinion leaders on BBS. Our results reveal the structural characteristics of this interpersonal network to be scale-free and to be of a small-world property. The existence of opinion leaders is clearly demonstrated. We also proposed a LeaderRank algorithm to identify opinion leaders based on community discovery and emotion mining methods. The performance of this algorithm is evaluated using real-world datasets and our experiments show that the identification of interest groups and the emotion property shown in post/reply articles helps to find opinion leaders on BBS. Finally, we investigated the relationship between opinion leaders and BBS boards, and we found that most of the opinion leaders are only active on few BBS boards.