Hybrid Ensemble Stacking Techniques for Coronary Artery Disease Prediction using Machine Learning Algorithms

Abstract

Throughout history, humanity has been plagued by several outbreaks that have claimed numerous lives. Since coronary artery disease is among the most fatal illnesses that humanity has faced in the modern era, it has been recognized in our time. It links several Coronary Artery Disease (CAD) risk factors to the critical requirement for precise, reliable, and workable methods for early identification and management. In light of this, we suggest a technique called Hybrid Ensemble Stacking that combines Naive Bayes (NB), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), and Ada Boosting for the prediction of CAD illnesses. To combine the forecasts of the basis models, a meta-logistic regression model is utilized. According to a quantitative study, the ensemble model and brute force feature selection method together produce a classification accuracy for heart disease of up to 92.66%. The suggested stacking model has demonstrated its effectiveness and outperforms current methods in the categorization of cardiac disorders. Several classification issues have been solved successfully using ensemble techniques. The suggested method was constructed using the Sani dataset, which contains 303 nearly completed records. Using Min-Max Normalization, the data are pre-processed to making it suitable for a Machine Learning (ML) model. SMOTE and SelectKBest technique were applied to   increases the accuracy and efficiency of a model. Using the metrics such as accuracy, precision, recall, F1, ROC and log-loss, the outcomes produced by the suggested model had the greatest performance.