Open AccessCommunity Structure Analysis using Fast Louvain Method in Real World Networks
Author(s) -
Laxmi Chaudhary,
Buddha Singh,
Neeru Meena
Publication year - 2019
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.d8113.118419
Subject(s) - jaccard index , computer science , community structure , complex network , modularity (biology) , data mining , context (archaeology) , metric (unit) , clique percolation method , visualization , theoretical computer science , quality (philosophy) , social network analysis , artificial intelligence , world wide web , pattern recognition (psychology) , mathematics , social media , paleontology , philosophy , operations management , epistemology , combinatorics , biology , economics , genetics
Recently, in complex networks detection of Community structure has gained so much attention. It adds a lot of value to social, biological and communication networks. The community structure is a convoluted framework thus analyzing it helps in deep visualization and a better understanding of complex networks. Moreover, it also helps in finding hidden patterns, predicting link in various types of networks, recommending a product to name a few. In this context, this paper proposes an agglomerative greedy method, referred to as Fast Louvain Method (FLM), based on Jaccard cosine shared metric (JCSM) to deal with the issues of community structure detection. Specifically, Jaccard cosine shared metric (JCSM) is developed to find the similarity between the nodes in a network. We have utilized modularity quality function for assessing community quality considering the local changes in this network. We test the method performance in different real-world network datasets i.e. collaboration networks, communication networks, online social networks, as well as another miscellaneous networks. The results also determined the computation time for unveiling the communities. This proposed method gave an improved output of modularity, community goodness, along with computation time for detecting communities’ number as well as community structure. Extensive experimental analysis showed that the method outperforms the existing methods.