Optimizing Gaussian Mixture Models with the Black-Winged Kite Algorithm for Enhanced Clustering Performance

The Gaussian Mixture Model (GMM) is a powerful tool for unsupervised learning, but its performance is highly sensitive to the selection of hyperparameters, such as the number of clusters and the regularization parameter. Traditional optimization methods, including the Expectation-Maximization algorithm, are often plagued by convergence to local optima and high computational costs. To address these challenges, this study introduces a novel metaheuristic-driven framework that employs the Black-winged Kite Algorithm (BKA) for the automated optimization of GMM parameters. Inspired by the foraging and migration behaviors of the black-winged kite, BKA excels in global exploration and demonstrates robust convergence capabilities. Our proposed BKA-GMM method automates the tuning process, effectively navigating the complex parameter space to identify optimal configurations. Extensive experiments on five benchmark datasets — including high-dimensional and large-scale scenarios — demonstrate that BKA-GMM significantly outperforms state-of-the-art optimizers, such as PSO, GWO, DE, and Bayesian Optimization, in terms of clustering accuracy, stability, and computational efficiency. The results confirm BKA-GMM as a superior and practical solution for intelligent clustering tasks in complex data environments.