Abstract

Automatic data clustering plays a vital role in unsupervised learning, where the number of clusters and their structure are often unknown beforehand. This paper presents a novel and computationally efficient clustering algorithm based on a Chaotic Butterfly Optimization Algorithm (CBOA). The proposed approach enhances the original Butterfly Optimization Algorithm by incorporating a chaotic perturbation mechanism and adaptive movement strategies to improve convergence speed and solution diversity. Unlike traditional clustering methods, CBOA does not require prior knowledge of the number of clusters. It automatically determines the optimal cluster count using a combined fitness function based on internal cluster validity indices, including the Silhouette index, Davies-Bouldin index, and CS index. Experimental results on UCI datasets and Shape datasets demonstrate the superiority of the proposed method in terms of clustering accuracy, compactness, and separation. Comparative analysis further confirms that CBOA outperforms classical clustering algorithms and recent metaheuristic-based approaches, while significantly reducing computational time. This makes it a promising tool for large-scale and complex clustering tasks.

Authors

R. Jensi
V V College of Engineering, India

Keywords

Automatic Clustering, Swarm Intelligence, Cluster Validity Indices, Chaotic Map, Cluster Structure Detection