Regular Article

Finding modular structure in multiplex networks by sequential intra-layer edge elimination

¹ Instituto de Física, Universidade Federal da Bahia, 40170-115 Salvador, Bahia, Brazil
² Centre for Data and Knowledge Integration for Health (CIDACS), Instituto Gonçalo Moniz, Fundação Oswaldo Cruz (FIOCRUZ), 41745-715 Salvador, Brazil

^a e-mail: randrade@ufba.br

Received: 11 February 2019
Received in final form: 1 August 2019
Published online: 18 May 2020

Abstract

Identifying modular properties of multiplex networks presents additional subtleties to the already corresponding difficult problem in single layer networks. One most evident issue is the presence of conflicting module partitions, when two or more layers have very clear community structure that differ from one another. Based on the well known Newman-Girvan method, a framework (MultiNG) to reach this goal is developed and tested. Taking into account that the community structure of any multiplex layer can be evaluated in advance, and that multiplex nodes have an intrinsic identity, MultiNG is targeted at finding one sole global structure, meaning that to any node a single module is assigned in all layers. As a consequence, inter-layer connections are preserved throughout the process, and only intra-layer edges are eligible to be eliminated. The reliability of the procedure is tested by investigating different cases, as synthetic multiplex networks and multiplex networks obtained from real data. Results are compared with those obtained by other methods.