Analyzing Clustering Algorithms for Non-Linear Data to Evaluate Robustness and Scalability

Jahnu Tanai Kumar Hindupur, Navaneeth A D, Hida Fathima P H, Swati Sharma

2025

Abstract

Clustering algorithms are fundamental in unsupervised machine learning, but they face significant challenges when applied to non-linear and complex data geometries. This study evaluates the performance of three clustering methods—K-Means, DBSCAN, and Hierarchical Clustering—on a Synthetic Circle Dataset and a Random Non-Synthetic Dataset. The Synthetic Circle Dataset, designed with concentric circular clusters, exposes the limitations of K-Means, which assumes convex cluster boundaries. In contrast, DBSCAN effectively detects non-linear clusters but is sensitive to parameter selection. Hierarchical Clustering demonstrates flexibility and interpretability through dendrogram visualizations, though it becomes computationally expensive for larger datasets. Quantitative metrics, including the Silhouette Score, Adjusted Rand Index, and Calinski-Harabasz Index, are employed to assess cluster quality. Visual comparisons reinforce that K-Means performs well on uniform, random data, while DBSCAN and Hierarchical Clustering excel at identifying complex structures. However, challenges such as parameter tuning and scalability persist. This study highlights the importance of selecting clustering techniques suited to data geometry and complexity. Future advancements, including adaptive parameter tuning, hybrid clustering approaches, and kernel-based methods, are proposed to address existing limitations. These findings provide a foundation for improving clustering algorithms to handle real-world datasets with irregular patterns, noise, and diverse densities.

Paper Citation

in Harvard Style

Hindupur J., A D N., P H H. and Sharma S. (2025). Analyzing Clustering Algorithms for Non-Linear Data to Evaluate Robustness and Scalability. In Proceedings of the 3rd International Conference on Futuristic Technology - Volume 3: INCOFT; ISBN 978-989-758-763-4, SciTePress, pages 770-780. DOI: 10.5220/0013659600004664

in Bibtex Style

@conference{incoft25,
author={Jahnu Tanai Kumar Hindupur and Navaneeth A D and Hida Fathima P H and Swati Sharma},
title={Analyzing Clustering Algorithms for Non-Linear Data to Evaluate Robustness and Scalability},
booktitle={Proceedings of the 3rd International Conference on Futuristic Technology - Volume 3: INCOFT},
year={2025},
pages={770-780},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013659600004664},
isbn={978-989-758-763-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 3rd International Conference on Futuristic Technology - Volume 3: INCOFT
TI - Analyzing Clustering Algorithms for Non-Linear Data to Evaluate Robustness and Scalability
SN - 978-989-758-763-4
AU - Hindupur J.
AU - A D N.
AU - P H H.
AU - Sharma S.
PY - 2025
SP - 770
EP - 780
DO - 10.5220/0013659600004664
PB - SciTePress