Modularity Maximization Adjusted by Neural Networks

• Published in: Neural Information Processing pp. 287 - 294
• Main Authors: Carvalho, Desiree Maldonado, Resende, Hugo, Nascimento, Mariá C. V.
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2014
• Series: Lecture Notes in Computer Science
• Subjects: community detection in networks; graph clustering; modularity maximization; neural networks
• ISBN: 3319126369; 9783319126364
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-12637-1_36

Graphs are combinatorial structures suitable for modelling various real systems. The high clustering tendency observed in many of these graphs has led a large number of researches, among them, we point out the modularity maximization-based community detection algorithms. Although very effective a few studies suggest that, for some networks, this approach does not find the expected communities due to a resolution limit of the measure. In this paper, we propose a way to automatically choose the value of the resolution parameter considered in the modularity by using neural networks. In the computational experiments, we observed that the proposed strategy outperformed another strategies from the literature for hundreds of artificial graphs considering the expected communities.