Social Interaction Study Of Online Communities And Its Applications

The increasing popularity of online community is witnessed by a huge number of users these years. An online community can take the form of an information system where anyone can post content, such as a bulletin board system, social network and micro-blog. The essential feature of the online community is the interactive relations between participants. The analysis of online community opens up new research questions. For example, 1) data regarding relations are defined and embedded in pre-defined structures provided by the online community structure itself. Therefore, much harder to extract; 2) what are the topological characteristics of the relationship network of online community? And what about their evolution and community structure? 3) Social interactions on online communities involve both positive and negative relationships, how can we predict one’s current attitude to the other under a given topic; 4) In the online community, some individuals use multiple usernames or disguise themselves as other users(usually called “sock puppet”) to communicate with others. How can we identify these sock puppets automatically? To address these issues, new techniques and paradigms are required to acquire and analyze massive data from online communities.In this dissertation, we analyzed the social interactions of online users, and constructed some interaction networks. The topological characteristics, evolution model, community structures and users’ current attitudes of these networks have been further studied. Finally, we proposed a novel sock puppet detection algorithm. The main contributions are as follows:1) A statistics-based approach that integrates the concept of fuzzy association rules(FAR) with that of sliding window(SW) has been propsed to efficiently extract the main text content from web pages.2) We construct a Similar-View Network model to model the interactions of online users. Then a “last updating time” network evolution model is proposed, which can maintain the robustness of scale free networks and the “small-world effect”. Finally, a fast parallel clustering algorithm(FPCA) has been proposed to detect the communities in online social networks.3) A game theory based method to analyze the interactive patterns in online communities, which is the first in its kind has been proposed.4) We propose an automatic sock puppet detection algorithm which combines authorship-identification techniques and link analysis. Compared to traditional methods, our approach conforms to the practical meanings of sock puppet communities; it can be applied at run time in a dynamic environment like sock puppets; and it increases the efficiency of link analysis.