Machine Learning for Coronary Heart Disease Risk Prediction: A Conventional and Ensemble Classification Model Approach

Abstract

Coronary Heart Disease (CHD) is the leading cause of illness and mortality worldwide. This is driven by risk factors such as high cholesterol, hypertension, smoking, diabetes, obesity and genetic predisposition. Early detection and management of CHD are critical and involve lifestyle changes, pharmacological interventions, angioplasty, or bypass grafting.  Traditional diagnostic methods require effective and timely medical interventions or the early detection of CHD. The proposed work suggests a stacking classifier for early prediction of CHD. Conventional classification models are typically simple, interpretable and efficient, making them suitable for classification problems. The performance of these models is limited when dealing with complex or high-dimensional data. Ensemble models address this limitation by combining the predictions of several models to improve accuracy. Bagging, Voting, and stacking classifiers are more sophisticated ensemble methods that use this concept further by using a metamodel to aggregate the predictions of multiple base models. In this study, several conventional models, such as the SGD Classifier, Logistic Regression, K-Nearest Neighbor, Naive Bayes, Decision Tree, and Random Forest Classifiers are applied to the dataset to predict whether the patient has 10 year risk of Coronary Heart Disease (CHD). Boosting models such as Gradient Boosting, AdaBoost, and XGBoost are also applied. The proposed voting classifier resulted in better accuracy, precision, recall, and other metrics compared with the conventional, boosting, and bagging models.