Mothe Josiane ; Haroutunian Mariam
Communities in real world complex networks correspond to hidden structures that are composed of nodes tightly connected among themselves and weakly connected with other nodes in the network. There are various applications of automatic community detection in computer science, medicine, machine learning, sociology, etc. In this paper, we first present the existing community detection algorithms and evaluation measures used in order to consider the algorithms effectiveness. We then report a deep comparison of the algorithms using both large scale real world complex networks and artificial networks generated from stochastic block model. We found that Louvain algorithm is consistently the best across both the measures and the networks (8 real world and many varied synthetic networks) we tested. Fast Greedy and Leading Eigenvector algorithms are also good alternatives. Moreover, compared to related work, our paper considers both more algorithms and more networks.
oai:noad.sci.am:135911
Information Theories and Applications
karenmkhitaryan@gmail.com ; josiane.Mothe@irit.fr ; armar@ipia.sci.am
Institute for Informatics and Automation Problems ; Institutde Recherche enInformatique de Toulouse
Mar 3, 2021
Jul 27, 2020
38
https://noad.sci.am/publication/149504
Հրատարակության անուն | Ամսաթիվ |
---|---|
Karen Mkhitaryan, DETECTING COMMUNITIES FROM NETWORKS: COMPARISONOF ALGORITHMS ON REAL AND SYNTHETIC NETWORKS | Mar 3, 2021 |