Implementation Of Parallelized Method For Local Community Detection In Weighted Complex Networks

Community structure has already been found in many real complex networks. Nodes insuch structure link densely to nodes inside the structure and meanwhile they link loosely tonodes outside of the structure. Studying such structure is meaningful to know more aboutcomplex network, because such structure represents some clustering features of complexnetwork. Nowadays, how to use the information of networks more effectively, and how tohandle the large networks are the important problems in the research of community detection.This paper proposes a weighted method to detect community structure based on localcentral vertex. Firstly, the maximal weighted node is proposed; the incident weight ofmaximal weighted node is no less than any of its neighbors. And then the weighted method,WLCV, starts to find local community by initializing every local maximal weighted node asinitial community and detects community by adding nodes into initial community step by step.By utilizing the Map-Reduce framework on Hadoop, a parallelized scheme of the method isproposed. It is shown that our method performs better than the other weighted methods withhigher accuracy and stability while detecting communities on both real and syntheticnetworks. It is also proved that the parallelized scheme works well on large weighted complexnetworks.