Hidden community detection based on microblog by opinion-consistent analysis

The content or topic of post on the social network such as microblog, forum are usually reflected user's interests. Traditional community detection methods only consider explicit information of users. So that data analysis is limited in user predefined attributes. In order to solve this problem, a hidden community detection framework is proposed in this paper called opinion-consistent hidden community (OCHC) framework. Firstly, we collect and process post comments on facebook. Then, the post topic that the target user participated in can be defined through topic identification by the selected ontology, Wikipedia. Moreover, opinion-consistency between users and the target user is discovered by sentiment analysis. In brief, opinion mining and sentiment analysis are used to track the users who have the similar opinion on the specific topics. Besides, users focus on different features with different scopes on facebook can be found by multi-level OCHC framework that we proposed in this paper. Communities of opinion-consistent users are clustered Multi-level OCHC model. There are two major improvements of OCHC framework, one is that post topic is decided by topic identification instead of user-self, and the other is that user opinions are also considered during analysis phrase on OCHC framework. In experiment results, accuracy of topic identification promoted 5.5% than other methods and the time complexity reached 26 times faster than other one. On quantitative measurements of Polarity and Multi-Dimension sentiment analysis methods are performed well.