Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Local Perspective-based Model for Overlapping Community Detection

Gaofeng Zhou, Rui-Feng Wang, Kangning Cui

TL;DR

The paper tackles overlapping community detection in large-scale networks by introducing LQ-GCN, a local-perspective model that fuses a Bernoulli-Poisson community-affiliation mechanism with local modularity within a two-layer GCN backbone. The end-to-end framework learns a node–community affiliation matrix $F$, balancing $L_{BP}$ and $L_{LQ}$ losses to refine community boundaries and intra-community coherence. Empirical results on six real datasets show substantial gains over state-of-the-art baselines, with improvements in ONMI up to 33% and Recall up to 26.3%, particularly on large graphs and when node features are informative. Ablation studies validate the contributions of local modularity and the modified convolution, and parameter analysis identifies a robust threshold setting for overlaps. Overall, LQ-GCN offers a scalable, accurate approach for detecting overlapping communities by exploiting local structure in conjunction with graph features.

Abstract

Community detection, which identifies densely connected node clusters with sparse between-group links, is vital for analyzing network structure and function in real-world systems. Most existing community detection methods based on GCNs primarily focus on node-level information while overlooking community-level features, leading to performance limitations on large-scale networks. To address this issue, we propose LQ-GCN, an overlapping community detection model from a local community perspective. LQ-GCN employs a Bernoulli-Poisson model to construct a community affiliation matrix and form an end-to-end detection framework. By adopting local modularity as the objective function, the model incorporates local community information to enhance the quality and accuracy of clustering results. Additionally, the conventional GCNs architecture is optimized to improve the model capability in identifying overlapping communities in large-scale networks. Experimental results demonstrate that LQ-GCN achieves up to a 33% improvement in Normalized Mutual Information (NMI) and a 26.3% improvement in Recall compared to baseline models across multiple real-world benchmark datasets.

A Local Perspective-based Model for Overlapping Community Detection

TL;DR

The paper tackles overlapping community detection in large-scale networks by introducing LQ-GCN, a local-perspective model that fuses a Bernoulli-Poisson community-affiliation mechanism with local modularity within a two-layer GCN backbone. The end-to-end framework learns a node–community affiliation matrix , balancing and losses to refine community boundaries and intra-community coherence. Empirical results on six real datasets show substantial gains over state-of-the-art baselines, with improvements in ONMI up to 33% and Recall up to 26.3%, particularly on large graphs and when node features are informative. Ablation studies validate the contributions of local modularity and the modified convolution, and parameter analysis identifies a robust threshold setting for overlaps. Overall, LQ-GCN offers a scalable, accurate approach for detecting overlapping communities by exploiting local structure in conjunction with graph features.

Abstract

Community detection, which identifies densely connected node clusters with sparse between-group links, is vital for analyzing network structure and function in real-world systems. Most existing community detection methods based on GCNs primarily focus on node-level information while overlooking community-level features, leading to performance limitations on large-scale networks. To address this issue, we propose LQ-GCN, an overlapping community detection model from a local community perspective. LQ-GCN employs a Bernoulli-Poisson model to construct a community affiliation matrix and form an end-to-end detection framework. By adopting local modularity as the objective function, the model incorporates local community information to enhance the quality and accuracy of clustering results. Additionally, the conventional GCNs architecture is optimized to improve the model capability in identifying overlapping communities in large-scale networks. Experimental results demonstrate that LQ-GCN achieves up to a 33% improvement in Normalized Mutual Information (NMI) and a 26.3% improvement in Recall compared to baseline models across multiple real-world benchmark datasets.

Paper Structure

This paper contains 15 sections, 10 equations, 3 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Methodology
  3. GCN Architecture
  4. Loss Function
  5. Bernoulli-Poisson Loss $L_{BP}$
  6. Local Modularity Loss $L_{LQ}$
  7. LQ-GCN Algorithm
  8. Experiments and Evaluation
  9. Datasets and Evaluation Metrics
  10. Comparative Experiments
  11. Ablation Study
  12. Effect of Local Modularity
  13. Effect of Modified Convolution
  14. Parameter Sensitivity Analysis
  15. Conclusion

Figures (3)

  • Figure 1: The proposed LQ-GCN Model Framework
  • Figure 2: The runtime of community detection algorithms.
  • Figure 3: The influence of the threshold.