Finding Closely Communicating Community Based on Ant Colony Clustering Model

Abstract

The investigation of community structures in networks is an important issue in many domains and disciplines. However, Most of the present algorithms consider only structure of the network, ignoring some additional conditions such as direction, weight, semantic, etc. In this paper the behaviors of each vertex are focus. Based on the previous work, two limitations of swarm similarity in closely community detection is outlined and the closely communicating community is defined clearly. The method for measuring relationship propinquity is proposed which considers multi-views to calculate the propinquity. Ant colony clustering model is applied into the closely communicating community detection. The improvement of community detection model is mainly in global problem space and local communication propinquity. Based on the method proposed, the Email Digger is implemented. Our method is successfully tested and evaluated on the Enron email dataset, and shows that the method is effective at identifying closely communicating communities.