Adaptive Clustering with Weighted Centroids: A Hybrid Approach for Scalable and Accurate Data Partitioning

Vishal Kaushik, Abdul Aleem

2025

Abstract

Clustering is a well-known task in machine learning, which is normally exposed to noise, non-standard cluster sizes, and uneven data sets. This paper provides a hybrid adaptive algorithm on the basis of weighted k-means and DBSCAN which combines the strengths to resolve the limitations. The method proposed utilizes weighted areas for dynamic adjustment of priorities and data density, robustness against imbalances, and increase the noise. DBSCAN, if integrated into optimization the algorithm handles nonlinear and irregular boundaries well. Three different data types, namely, blobs, months, and Gaussian mixtures, were tested on this algorithm. The experimental results indicate high clustering accuracy, with an adjusted rand index (ARI) of 0.92 on the blobs dataset, outperforming traditional k-means (0.85) and weighted k-means (0.88). Scalability analysis reveals efficient runtime memory usage, with some compensation for improved efficiency and accuracy. Sensitivity analysis confirms the flexibility of the algorithm for changes in hyperparameters, including the number of clusters, weighting, and DBSCAN parameters. Visual proof, such as ARI and runtime comparison charts, confirms the superiority of the hybrid approach with regard to accuracy and efficiency. Utility was demonstrated through the data sets; it is indeed capable of solving real challenges in the world. Further work, combining deep learning for automatic feature extraction with the extended method for streaming or online clustering applications, opens up the way to even more flexible and dynamic solutions to clustering problems.

Paper Citation

in Harvard Style

Kaushik V. and Aleem A. (2025). Adaptive Clustering with Weighted Centroids: A Hybrid Approach for Scalable and Accurate Data Partitioning. In Proceedings of the 3rd International Conference on Futuristic Technology - Volume 1: INCOFT; ISBN 978-989-758-763-4, SciTePress, pages 735-743. DOI: 10.5220/0013584800004664

in Bibtex Style

@conference{incoft25,
author={Vishal Kaushik and Abdul Aleem},
title={Adaptive Clustering with Weighted Centroids: A Hybrid Approach for Scalable and Accurate Data Partitioning},
booktitle={Proceedings of the 3rd International Conference on Futuristic Technology - Volume 1: INCOFT},
year={2025},
pages={735-743},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013584800004664},
isbn={978-989-758-763-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 3rd International Conference on Futuristic Technology - Volume 1: INCOFT
TI - Adaptive Clustering with Weighted Centroids: A Hybrid Approach for Scalable and Accurate Data Partitioning
SN - 978-989-758-763-4
AU - Kaushik V.
AU - Aleem A.
PY - 2025
SP - 735
EP - 743
DO - 10.5220/0013584800004664
PB - SciTePress