Classification of Heart Disease Using the Ensemble SVM Method

Abstract

Cardiovascular disease (CVD), particularly coronary heart disease, remains the leading cause of global mortality. It makes early detection essential for effective prevention of the diseases. Machine learning offers a promising alternative for rapid and accurate prediction. This study investigates the performance of Support Vector Machine (SVM) classifiers enhanced through an ensemble stacking approach. In this study, we employed three SVM kernels, including linear, RBF, and polynomial, using GridSearchCV to obtain accuracies of 97.1%, 97.2%, and 96.3%, respectively. Experimental results show that the optimized stacking ensemble achieved the highest accuracy of 97.5%, with TP=91, FN=4, FP=1, and TN=104. This model outperformed individual SVM kernels and surpassed several existing methods, including ANN and hybrid SVM–NN approaches. The findings confirm that integrating multiple optimized SVM kernels enhances classification accuracy, stability, and robustness for heart disease prediction. The proposed ensemble-based SVM model provides a valuable contribution to medical diagnostics by improving early detection reliability and supporting preventive strategies for cardiovascular diseases