Centroid Optimization of K-Means Using Ant Colony Optimization for Culinary MSME Clustering

Micro, Small, and Medium Enterprises (MSMEs) are economic activities conducted by individuals or groups, particularly in the culinary sector. The rapid expansion of culinary MSMEs, especially in tourism-oriented regions such as the Special Region of Yogyakarta, necessitates effective data clustering to systematically analyze their characteristics. High-quality clustering plays a crucial role in supporting informed decision-making, including business development planning, MSME assistance programs, and the formulation of well-targeted policies. This study applies the K-Means algorithm to cluster culinary MSME data; however, its performance is sensitive to centroid initialization, which may result in suboptimal clustering outcomes. To address this limitation, Ant Colony Optimization (ACO) is employed as a centroid optimization approach. ACO is a metaheuristic algorithm inspired by the foraging behavior of ant colonies, where pheromone trails guide the search toward optimal solutions. The results indicate that the integration of ACO enhances clustering performance compared to K-Means. The silhouette scores obtained are 0.88 and 0.89 for two clusters, 0.80 and 0.86 for three clusters, and 0.80 and 0.92 for four clusters for K-Means and ACO-optimized K-Means, respectively. These findings demonstrate that ACO effectively improves centroid initialization, with four clusters identified as the optimal configuration.