Hybrid Graph Embeddings and Louvain Algorithm for Unsupervised Community Detection

2025-09-27

AI Summary

This work addresses the challenge of requiring a pre-specified number of communities in unsupervised community detection. Methodologically, it proposes a novel framework integrating Graph Neural Networks (GNNs) with the Louvain algorithm: first, GNNs learn expressive node embeddings to enhance structural representation; second, embedding-based distances are incorporated into the Louvain modularity optimization, coupled with an adaptive community merging strategy to dynamically determine the number of communities and improve partition quality. Its key contribution lies in being the first to introduce a GNN-driven embedding mechanism into the Louvain algorithm, thereby eliminating reliance on prior knowledge of community count. Extensive experiments on multiple real-world graph datasets demonstrate that the method significantly outperforms conventional Louvain and state-of-the-art unsupervised baselines, achieving average improvements of 5.2%–9.8% in NMI and F1-score while effectively suppressing redundant community generation.

Technology Category

• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Machine Learning: Graph-based Machine Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

This paper proposes a novel community detection method that integrates the Louvain algorithm with Graph Neural Networks (GNNs), enabling the discovery of communities without prior knowledge. Compared to most existing solutions, the proposed method does not require prior knowledge of the number of communities. It enhances the Louvain algorithm using node embeddings generated by a GNN to capture richer structural and feature information. Furthermore, it introduces a merging algorithm to refine the results of the enhanced Louvain algorithm, reducing the number of detected communities. To the best of our knowledge, this work is the first one that improves the Louvain algorithm using GNNs for community detection. The improvement of the proposed method was empirically confirmed through an evaluation on real-world datasets. The results demonstrate its ability to dynamically adjust the number of detected communities and increase the detection accuracy in comparison with the benchmark solutions.

Problem

• Proposes unsupervised community detection without prior knowledge
• Enhances Louvain algorithm using GNN-generated node embeddings
• Introduces merging algorithm to refine detected community structure

Innovation

• Integrates Louvain algorithm with Graph Neural Networks
• Uses GNN embeddings to capture structural information
• Introduces merging algorithm to refine community results