Over the past decade overlapping community detection has gained more focus in network science. Most of the existing algorithms for overlapping community detection are computationally expensive. These algorithms may not be scalable for dynamic and large networks. Improving the computational efficiency of overlapping community detection algorithms while maintaining high accuracy is an ongoing challenge. The key step in fuzzy clustering-based overlapping community detection is identifying the most optimal community centers. With FCM (Fuzzy C-means), any data point can be a member of multiple clusters to varying degrees. This research proposed a novel, accurate, and efficient approach for identifying overlapping communities using Particle Swarm Optimization (PSO) and Graph Neural Network (GNN). PSO maximizes the model's accuracy on FCM-derived community labels by fine-tuning GNN parameters to improve the model's performance on community detection tasks. On the Facebook dataset, a GNN optimized with PSO significantly outperforms GNN optimization using conventional techniques like Markov Stability and Modularity Optimization, achieving a high accuracy of 99.62%.