Enhancing value-growth stock classification using K-Means clustering with Mahalanobis distance

This study investigates the integration of K-Means clustering and Mahalanobis Distance (MD) for classifying S&P 500 stocks into value and growth categories. Unlike traditional approaches relying on single financial ratios, this method accounts for multivariate dependencies among financial metrics. Using Refinitiv data from 2008 to 2023, the study compares standard K-Means MD, K-Means++ MD, and a novel K-Means++ MD Hybrid approach. The hybrid algorithm, which combines robust initialization with Mahalanobis-based distance computation, demonstrates superior cluster stability and alignment with traditional valuegrowth classifications. Performance evaluations show that the resulting portfolios exhibit economic interpretability and consistency over time. These findings highlight the methodological advantages of incorporating covariance structures in unsupervised financial classification.