Ensemble-based community detection for dynamic networks

Community detection is crucial for understanding complex systems in network science. However, traditional methods often face practical issues due to the variability of the results influenced by resolution parameters. Ensemble-based community detection techniques have been proposed to address this problem by aggregating results from multiple analyses to enhance reliability, and suggested global and local metrics for robust community detection. In this study, we explore the applicability of these ensemble-based techniques to dynamic networks by applying them to simulated networks with evolving community structures. Using the partition inconsistency measure, a global metric assessing overall structural stability, we identified time points where stable community configurations changed. Furthermore, by analyzing the trajectories of membership inconsistency, a local metric quantifying node-level assignment community consistency, we detected nodes that were initially affected by dynamic changes in community structure. These findings demonstrate that ensemble-based community detection methods are effective tools for analyzing dynamic networks. This method has the potential to enhance our understanding of temporal dynamics in complex networks and aid in predicting future states across various domains.