Comparative Evaluation of XGBoost and Support Vector Machine Algorithms for Heart Disease Prediction Using Healthcare Datasets

Abstract

This study examines the efficacy of advanced machine learning algorithms—Support Vector Machines (SVM) and Extreme Gradient Boosting (XGBoost)—in predicting heart disease using the IEEE DataPort Heart Disease Dataset. The analytical approach included rigorous data preprocessing, such as handling missing values and normalising features, followed by correlation-based feature selection to retain variables strongly associated with heart disease. Both models underwent extensive hyperparameter optimisation using GridSearchCV to enhance predictive performance. Evaluation metrics encompassed accuracy, precision, recall, F1 score, and AUC-ROC. SVM achieved notable results, including an accuracy of 0.8613 and an AUC-ROC of 0.9092, while XGBoost outperformed it across all metrics, with an accuracy of 0.8950 and an AUC-ROC of 0.9526. These findings underscore XGBoost’s superior robustness and classification ability. The study highlights the potential of these models to advance diagnostic precision in cardiology, enabling earlier intervention and personalised treatment strategies for improved patient outcomes.