The Fuzzy Self-organizing Map (FSOM) is a supervised neuro-fuzzy model designed for prediction tasks. In this study, we adapt FSOM for unsupervised clustering by introducing an optimized framework that automatically tunes key parameters: such as rule centers, spreads, and learning rates, using metaheuristic algorithms. A new objective function based on clustering validity indices guides this optimization. Experiments on benchmark datasets show that the proposed model significantly enhances clustering performance and consistently outperforms traditional methods in terms of Silhouette and Dunn indices.