An Ensemble Clustering Framework Based on Majority Voting: Comparative Analysis with Enhanced Clustering Algorithms

¹ Research Scholar, Department of Computer Engineering, MET’s Institute of Engineering, Adgaon, Nashik - 422003, Maharashtra, India
² Professor, Department of AI & DS, MET’s Institute of Engineering, Adgaon, Nashik - 422003. (Affiliated to Savitribai Phule Pune University, Pune), Maharashtra, India

Published online: 20 November 2025

Abstract

The growing complexity of data in modern applications demands clustering methods that are both accurate and stable. Traditional algorithms often struggle with initialization sensitivity, noise, or inconsistent results across datasets. In this paper, we present an ensemble clustering framework based on majority voting, designed to integrate the strengths of multiple enhanced algorithms—K-Means, K-Medoids, Fuzzy C-Means, Expectation-Maximization, and DBSCAN— while reducing their individual limitations. Each base algorithm is optimized for improved convergence and consistency. The ensemble mechanism assigns final cluster labels through a majority voting rule, ensuring balanced outcomes across diverse data distributions. Experiments conducted on a synthetic dataset of 16,000 instances demonstrate that the proposed model achieves a Silhouette score of 0.7043, matching the best-performing individual methods while maintaining higher robustness and stability. The results confirm that this straightforward voting-based ensemble approach delivers consistent and interpretable clustering performance, making it a practical choice for high-dimensional and noisy datasets encountered in real-world applications.