Ensemble Learning Approaches for Cardiovascular Disease Prediction

Cardiovascular Disease (CVD) is a primary cause of mortality worldwide, which requires precise and timely predictions to implement effective prevention and intervention strategies. Existing prediction methods, dependent on a single classifier, frequently prove inadequate due to data complexity and class imbalance, leading to incorrect classification. This study aimed to compare the performance of ensemble learning approaches for predicting CVD using clinical datasets. Six ensemble classifiers were evaluated: XGBoost and Gradient Boosting (Boosting), Random Forest and Decision Tree (Bagging), and K-Nearest-Neighbor and Logistic Regression (Stacking). Boosting classifiers achieved the highest accuracy (73.0%) and G-mean (72.0%) for CVD prediction, while the bagging classifier for Random Forest achieved similar precision in identifying high-risk patients with CVD. Boosting classifiers demonstrate strong discriminative power in differentiating between risk, as confirmed by the ROC AUC (0.80), indicating their effectiveness for reliable CVD prediction.