Layer Clustering-Enhanced Stochastic Block Model for Community Detection in Multiplex Networks

Nowadays, multiplex data are often collected, and the study of multiplex-network (MN)s’ community detection is a cutting-edge topic. multiplex-network (MN) layers can be grouped by clustering, and there are correlations between network layers that are assigned to the same cluster. Although the differences between network layers entail that the node community membership can differ across the layers, Stochastic-Block-Models (SBM)-based MN-community-detection methods current available are theoretically constrained to assume the same node community membership across the layers. Here, we propose a new SBM-based MN-community-detection algorithm, which surpasses this theoretical constraint by exploiting a two-stage procedure. Numerical experiments show that the proposed algorithm can be more accurate and robust than multilayer-Louvain algorithm, and may help to contain some inference issues of classical monolayer SBM. Finally, results on two real-world datasets suggest that our algorithm can mine meaningful relationships between network layers.