Comparative Study of Four Methods in Hierarchical Cluster Analysis

https://doi.org/10.64943/ljmas.v3i4.183

Authors

  • Hanadi A. Amhimmid Mathematics Department, Faculty of Science, Omar Almukhtar University, Albeida, Libya
  • Fatma Alzahra A. Aljehany Statistics Department, Faculty of Science, Benghazi University, Benghazi, Libya
  • Mohamed A. Mohamed Department of Statistics, Faculty of Science, Sebha University, Sebha, Libya
  • Kasem A. Farag Mathematics Department, Faculty of Science, Omar Almukhtar University, Albeida, Libya

Keywords:

Hierarchical Clustering, Linkage Methods, Ward's Method, Cluster Analysis, Performance Evaluation.

Abstract

Hierarchical cluster analysis represents a fundamental technique in unsupervised machine learning and exploratory data analysis, with applications spanning numerous scientific disciplines. This study presents a comprehensive comparative analysis of four principal hierarchical clustering methods: single linkage, complete linkage, average linkage, and Ward's method. The primary objective is to evaluate the performance characteristics, strengths, and limitations of each approach across diverse data structures and clustering scenarios. Through systematic simulation studies implemented in R software, we generated synthetic datasets with varying cluster properties, including different shapes, densities, and noise levels. Performance evaluation utilized multiple metrics including silhouette coefficients, cophenetic correlation, and cluster validity indices. Results demonstrate that Ward's method consistently produces the most compact and well-separated clusters for spherical cluster structures, achieving superior silhouette scores (mean = 0.78) compared to other methods. Complete linkage showed robust performance against outliers but exhibited sensitivity to cluster size variations. Single linkage effectively identified elongated clusters but suffered from chaining effects in noisy datasets. Average linkage provided balanced performance across different scenarios, serving as a reliable middle-ground approach. The findings reveal significant performance dependencies on data characteristics, suggesting that method selection should be guided by prior knowledge of underlying cluster structures. This research contributes to the understanding of hierarchical clustering method selection and provides practical guidelines for practitioners in choosing appropriate algorithms for specific data analysis contexts.

Published

2025-10-02

How to Cite

Hanadi A. Amhimmid, Fatma Alzahra A. Aljehany, Mohamed A. Mohamed, & Kasem A. Farag. (2025). Comparative Study of Four Methods in Hierarchical Cluster Analysis. Libyan Journal of Medical and Applied Sciences, 3(4), 11–16. https://doi.org/10.64943/ljmas.v3i4.183

Issue

Section

Applied Science