Identifying Cohesive Local Community Structures in Networks

Identifying community structure in networks is an important topic in data mining research. One of the challenges is to find local communities without requiring the global knowledge of the entire network. Exiting techniques have several limitations. First, there is no widely accepted definition for community. Second, these algorithms either lack good stopping criteria or depend on predefined threshold parameters. In this research I propose a local cohesion based algorithm to identify local communities in networks. This algorithm is grounded on the widely accepted group cohesion definition in social network analysis research. The algorithm is self-contained and does not depend on predefined threshold parameter to terminate the identification process. The evaluation results show that the proposed algorithm is more effective than the benchmark algorithm and can identify meaningful local communities in very large networks such as the Amazon co-purchasing network.