Benchmarking Evolutionary Community Detection Algorithms in Dynamic Networks

TL;DR

This work tackles the challenge of evaluating evolutionary modularity-based community detection in dynamic networks. It introduces a benchmarking framework that generates synthetic, ground-truth graphs via the LFR model and applies nine evolving transformations across Noise, Morphing, and Disruptive scenarios, paired with three evaluation metrics rooted in AMI: Stability, Correctness, and Delay. The authors also propose NeGMA, a neighbourhood-based generalization of Louvain, alongside two existing evolutionary variants, αGMA and sGMA. Experimental results show that NeGMA achieves a well-balanced performance in correctness and stability and superior responsiveness to abrupt changes, while αGMA and sGMA excel in particular aspects (stability or speed) but fail to consistently detect new or instantaneous communities. Overall, the framework and NeGMA enable robust, ground-truth-driven analysis of dynamic modularity-based CD methods with potential applicability to real-world evolving networks.

Abstract

In dynamic complex networks, entities interact and form network communities that evolve over time. Among the many static Community Detection (CD) solutions, the modularity-based Louvain, or Greedy Modularity Algorithm (GMA), is widely employed in real-world applications due to its intuitiveness and scalability. Nevertheless, addressing CD in dynamic graphs remains an open problem, since the evolution of the network connections may poison the identification of communities, which may be evolving at a slower pace. Hence, naively applying GMA to successive network snapshots may lead to temporal inconsistencies in the communities. Two evolutionary adaptations of GMA, sGMA and $α$GMA, have been proposed to tackle this problem. Yet, evaluating the performance of these methods and understanding to which scenarios each one is better suited is challenging because of the lack of a comprehensive set of metrics and a consistent ground truth. To address these challenges, we propose (i) a benchmarking framework for evolutionary CD algorithms in dynamic networks and (ii) a generalised modularity-based approach (NeGMA). Our framework allows us to generate synthetic community-structured graphs and design evolving scenarios with nine basic graph transformations occurring at different rates. We evaluate performance through three metrics we define, i.e. Correctness, Delay, and Stability. Our findings reveal that $α$GMA is well-suited for detecting intermittent transformations, but struggles with abrupt changes; sGMA achieves superior stability, but fails to detect emerging communities; and NeGMA appears a well-balanced solution, excelling in responsiveness and instantaneous transformations detection.

Figures (6)

• Figure 1: Evaluation of static and evolutionary CD algorithms on the real-world SocioPatterns dataset.
• Figure 2: Proposed NeGMA initialisation of nodes at snapshot $t$ starting from snapshot $t-1$. Different colours represent different communities; squares represent new nodes.
• Figure 3: Noise. Median gain and bootstrapped 99% confidence interval in Correctness and Stability.
• Figure 4: Dynamic evolution of the graph perturbed by the intermittence transformation, showcasing community assignment. Zoom in a sample of snapshots from 15 to 90 with step 15 where changes occur.
• Figure 5: Morphing. Median gain and bootstrapped 99% confidence interval in Correctness and Stability.