Enhancing Coronary Disease Prediction Using Stacking and Voting-Based Ensemble Machine Learning Models

Machine learning (ML) has emerged as a very important tool in healthcare, especially in the early detection of critical diseases. Heart disease is one of the major global health hazards for which accurate predictive models are needed to assist in timely intervention. This paper reports an elaborative assessment of the advanced ML techniques, such as Stacking-Classifier, and VotingClassifier on the Framingham Heart Study dataset to enhance heart disease prediction. It is seen that, among the reported and studied hybrid and traditional ML models, the StackingClassifier achieved the highest accuracy of 92.56%, while the VotingClassifier scored an accuracy of 89.99%. Both the hybrid models outplayed traditional boosts such as AdaBoost, Gradient Boosting, and CatBoost by an average increase in accuracy of around 3-6%. Notably, the above-mentioned hybrid models have been able to provide accuracy close to or, in some cases, higher than Deep Belief Networks (DBNs) with lower computational overhead. These results further show that hybrid ML models can be a reliable, comparatively accurate, and computationally efficient alternative for the prediction of heart diseases and may enable rapid clinical decisions that improve patient outcomes.