A robust ant colony optimization-based algorithm for community mining in large scale oriented social graphs

摘要

Community detection plays a key role in such important fields as biology, sociology and computer science. For example, detecting the communities in protein-protein interactions networks helps in understanding their functionalities. Most existing approaches were devoted to community mining in undirected social networks (either weighted or not). In fact, despite their ubiquity, few proposals were interested in community detection in oriented social networks. For example, in a friendship network, the influence between individuals could be asymmetric; in a networked environment, the flow of information could be unidirectional. In this paper, we propose an algorithm, called AC0D1G, for community detection in oriented social networks. ACODIC uses an objective function based on measures of density and purity and incorporates the information about edge orientations in the social graph. ACODIC uses ant colony for its optimization. Simulation results on real-world as well as power law artificial benchmark networks reveal a good robustness of ACODIC and an efficiency in computing the real structure of the network.