Topology Structure Research And Opinion Leader Identifying In BBS

With the widespread use of the Internet, BBS (Bulletin Board Systems) has become an indispensable platform for information exchange, due to its property of large quantity, huge scale and numerous users. Corresponding research has been made on the network structure and user behaviors in BBS by scholars ranging from domestic to overseas. In BBS opinion leaders, who play an important role in leading public opinions, tend to affect most of the people and become an important node in the process of message spreading, so identifying opinion leaders in the online community has practical significance, and become one of the research focus in user behavior study.There are two major branches in BBS network research: topology study of BBS network and user behavior study. Topology measuring is the base for opinion leaders mining. In our research different network models are selected for different research purposes, and we first introduce a new algorithm for mining opinion leaders. Interest is a group of topics about similar things. We use the method of text clustering to find groups of topics with the same interest from all topics in BBS. And through these groups we can find users sharing the same interest in BBS. In the following research, we use sentiment analysis to judge the authenticity of users'emotion. According to the improvement above, we first propose a LeaderRank algorithm to identify opinion leaders. Carefully designed experiment proves that LeaderRank can effectively improve the accuracy of opinion leaders identifying.This research utilizes a university BBS as the research object. We make some complex network analysis and study the problems about the number of connections and network between boards. We first propose an interest-based three-layer model for BBS. Based on this model we propose a LeaderRank algorithm and improve the accuracy of identify opinion leaders algorithm.