Community Detection Using Graph Attention Networks Clustering Algorithm

Abstract

Community detection plays a crucial role in various domains, including social network analysis, recommendation systems, and biological networks. However, current community detection methods face challenges in capturing the complex structural patterns and dynamic nature of real-world networks. In this study, we propose an innovative approach to enhance community detection by integrating graph attention networks (GATs) with two distinct clustering algorithms: k-means and agglomerative. Our proposed model leverages GATs to capture and extract new node features, providing a more comprehensive representation of the network. Subsequently, a clustering algorithm is applied to identify communities based on the learned embeddings. Our findings demonstrate that GATs combined with k-means outperform existing community detection methods. Moreover, this approach serves as a proof-of-concept, highlighting the broader potential of applying graph neural networks to unsupervised network clustering.