A Hybrid K-Means Decision Tree Model for Enhanced Classification Performance

Shilpa Naskar

Abstract: This paper presents a hybrid machine learning architecture that integrates K-Means clustering with Decision Tree classification to enhance predictive performance on heterogeneous datasets. Traditional classifiers often assume that data originates from a uniform distribution, and this assumption weakens performance when the data contains internal subgroups or latent structures. The proposed model addresses this limitation by initially partitioning the dataset using unsupervised K-Means clustering, followed by training a separate Decision Tree classifier on each cluster-specific subset. During prediction, each test instance is assigned to the nearest cluster centroid and subsequently classified by the Decision Tree associated with that cluster. This approach combines the structural strengths of both unsupervised and supervised learning techniques to achieve greater accuracy, improved interpretability, and more efficient learning on diverse datasets. Experimental analysis shows that the hybrid model adapts well to non-uniform data distributions and offers performance improvements over traditional single- model classifiers.

Keywords: K-Means Clustering, Decision Tree (DT), Hybrid Model, Machine Learning, Classification, Unsupervised Learning, Supervised Learning, Elbow Method

How to Cite?: Shilpa Naskar, "A Hybrid K-Means Decision Tree Model for Enhanced Classification Performance", Volume 14 Issue 12, December 2025, International Journal of Science and Research (IJSR), Pages: 26-29, https://www.ijsr.net/getabstract.php?paperid=MR251129161258, DOI: https://dx.doi.org/10.21275/MR251129161258