Research On Community Evolution Based On Network Topological Structure

The structure formed by both nodes and the links between the nodes in complex network is called network topological structure. Social network is a kind of typical complex network in which nodes represent of people and the connection between nodes represent of the relationship of people. Community a kind of sub-network of social network which has key characters. As time goes by, the communities structure changed, which we call community evolution, and we call the decompose, occur, shrink, expand, split or merge key events. During the evolution, the communities’ structure parameters will also change. In social network, events represent group behavior tendency and may mean some interest or social factors occurrence, and the prediction contributes to find these factors earlier and use them to guide network behavior. So it’s significant to predict the key events both in aspects of research and application.Community mergence covers more than one community, but previous research can only predict single community behavior tendency. This paper improve the predict function with a prediction model which can predict whether any pair of communities will merge or not. By analyzing structure parameters’change during the evolution, we improve the direct factors and we design link prediction algorithm at community level to extract indirect factors, which are used to weaken the uncertainty brought by direct factors and improve the prediction accuracy. After that, we make both SVM classifier model and community mergence tendency model designed in this paper, and then put true social network data into the two models, analyze prediction results and get the conclusion that tendency model gets better performance in normal social network, and the director factor and indirector factor proposed in this paper is efficient in the prediction.The basic steps to predict kinds of events are almost the same in past researches. For doing the research with convenience and achieving the prediction’s application, we design a prediction system in this paper. The system is divided into three main modules (community extraction module, key factor extraction module and community evolution prediction module) as the prediction’s steps, the input and output modules, and data storage module. We design and implement the modules in details for community mergence prediction as it’s the focus of this paper.Finally, we make conclusion of this paper’s research work, and propose the future research direction of algorithm improvement and system designation.