Research On Mining Social Circles Based On Microblog User Similarity

Microblog is the most popular social media and online communication platform nowadays, more and more users choose to build personal social circles in there. Meanwhile, the behavior of users' own posting or following may lead to the microblog contains numerous and complex content and relational data. If using these data to identify the social circles accurately from microblog users' huge virtual network of relationships, it not only facilitates the management of user's personal relationship network, but also has practical significance in the personalized recommendation, influence spreading and public opinions pre-warning.With the analysis of pages and content features from microblog, we summarized the requirements of the needed data firstly. Then, the data collection system was designed and implemented based on web crawler, which crawled 430 user nodes' data in the Sina Weibo successfully with the seed of author's account in whose followers and fans list.The characteristics of the microblog user relationship result from the one-way follow mechanism were analyzed in detail based on the above crawled data. A user similarity measure method was proposed with a fusion of relationship and content information to describe the connection tightness between users better, in which used the method to extraction the topic distribution of tweet content that based on LDA model. Then, a microblog social circle mining method called CMUS based on the user fusion similarity and node clustering was designed to solve the problem of center user's circles mining in microblog.Finally, the comparative experiments on the acquired user data set verified the validity of the fusion user similarity. When compared to the other mining methods, the CMUS obtained more accurate circles that closer to the user's real situation from the last experimental result.