Long-Term Prediction of Coronary Artery Disease via Ensemble Machine Learning Algorithms

Coronary artery disease (CAD) is the leading cause of death worldwide, necessitating early detection methods that are non-invasive, cost-effective, and reliable. In this study, the effectiveness of various machine learning (ML) models in predicting CAD was evaluated, with a focus on addressing class imbalance using the Synthetic Minority Oversampling Technique (SMOTE). The Framingham CAD dataset was utilized, and SMOTE was applied with different k-values to balance the data, examining the impact on prediction performance. Eight significant features—age, diaBP, glucose, heart rate, sysBP, totChol, cigsPerDay, and BMI—were determined during preprocessing and used for further analysis. Among the models tested, the StackingC classifier demonstrated superior performance, achieving an accuracy of 95.81%, sensitivity of 95.9%, specificity of 95.7%, and an AUROC of 99.2% for k=1. These findings highlight the potential of the StackingC model as a robust tool for CAD prediction, offering a promising non-invasive method for early diagnosis.