An Enhanced Local Modularity Measure

Recently, detection of community structure in networks has drawn a lot of attention. In this case, most of the developed methods need global knowledge of the network which is not applicable to real world graphs, since, they are are too large or evolve too quickly .Besides, we may be interested in the community structures of some given nodes, not all nodes .So, detecting the community of a specific node is more appropriate .several local modularity measures have been developed. Amongst, local modularity R works well in case of performance and simple agglomeration mechanism. But it has low recall information retrieval measure due to its predetermined number of agglomerated nodes. In this paper, we have changed its stopping criteria to multiple regression analysis. Hence, its recall parameter is improved leading to more accurate measure. Moreover, its performance is optimized, because, this new stopping criteria and agglomeration process works simultaneously leading to lower execution time. We validate our method on two real-world networks whose community structures are known .The result shows that our method can achieve higher recall as well as better performance.