Community Detection in Social Networks Using Graph Attention Network with Enhanced Embeddings

The widespread use of social networks, besides the diverse nature of user interactions, has led to increasingly complex and multi-layered structures within these networks. Unveiling these relationships and semantic structures necessitates deep and precise analyses. Community detection in social networks is one of the key topics in this area and includes wide-ranging applications such as recommender systems, fraud detection, and social interaction analysis. In recent years, the use of deep learning methods in this field of study has increased significantly due to their ability to model complex relationships. This research introduces a novel approach that uses a combination of metapath-based embeddings and graph neural networks (GNN) to improve the performance of community detection. Within this framework, the Metapath2Vec embedding model is used to create node embeddings and the GAT model, an attention-based graph neural network, is used to extract structural and semantic information from heterogeneous networks. The proposed model is evaluated on the heterogeneous DBLP dataset and accuracy (ACC) and F1score metrics are used to measure its performance. Experimental results underscore that the proposed method outperforms several existing state-of-the-art models and achieves a high ranking.