How Robust are Communities in Temporal Networks? A Comparative Analysis using Community Detection Algorithms
Moyi Tian
, Brown University
Communities often represent structural and functional clusters in networked systems. Communities have been found to be essential building blocks for understanding the robustness of critical infrastructures and the diffusion of information in online social networks. Previous work in community robustness analysis has focused on studying changes in the community structure of networks as a response of edge rewiring and node/edge removal. However, many real networked systems are constantly evolving and increasing their connectivity. Thus, there is a growing need to understand the limits of the robustness of communities with respect to expanding density. We hypothesize that the choice of algorithm used for detecting communities has an effect on the robustness of the associated community structure. Here we use state-of-the-art community detection algorithms (i.e., Infomap, Label propagation, Leiden, and Louvain) to understand their effect when studying the robustness of community structures in temporal networks. We test this hypothesis in both synthetic Lancichinetti–Fortunato–Radicch benchmark networks under random addition of edges and on email networks with increasing edges over time. Our preliminary results indicate that the robustness of communities is heavily dependent on the chosen community detection algorithm. We suggest that for understanding the robustness of communities in temporal networks, a careful selection of the community detection algorithm is imperative.