A HYBRID MACHINE LEARNING APPROACH FOR HEART DISEASE PREDICTION WITH SMOTE AND ENSEMBLE MODELS

Abstract

Heart disease remains one of the leading causes of mortality worldwide, empha-sizing the need for accurate and early diagnostic systems. This study proposes a machine learning–based framework for heart disease prediction using clinical data, with a particular focus on addressing class imbalance and improving predictive re-liability. To mitigate the effects of data imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was applied prior to model training. Three machine learning models Logistic Regression, Random Forest, and XGBoost—were developed and evaluated using multiple performance metrics, including accuracy, precision, recall, F1-score, confusion matrix, and Receiver Operating Characteristic (ROC) Area Un-der the Curve (AUC) analysis. Experimental results demonstrate that ensemble and boosting-based models significantly outperform traditional linear classifiers. Among the evaluated models, XGBoost achieved the best performance, attaining an accuracy of (87.87) with superior ROC–AUC characteristics and reduced false negative rates. Comparative analysis with recent state-of-the-art studies further validates the effec-tiveness and competitiveness of the proposed framework. The findings indicate that the proposed approach offers a robust and clinically reliable solution for early heart disease prediction and can support decision-making in healthcare systems.