Comparison of Supervised Learning Models for the Prediction of Coronary Artery Disease

Abstract

Cardiovascular diseases and Coronary Artery Disease (CAD) are the leading causes of mortality among people of different ages and conditions. The use of different and not so invasive biomarkers to detect these types of diseases joined with Machine Learning techniques seems promising for early detection of these illnesses. In the present work, we have used the Sani Z-Alizadeh dataset, which comprises a set of different medical features extracted with not invasive methods and used with different machine learning models. The comparisons performed showed that the best results were using a complete set and a subset of features as input for the Random Forest and XGBoost algorithms. Considering the results obtained, we believe that using a complete set of features gives insights that the features should also be analyzed by considering the medical advances and findings of how these markers influence a CAD disease's presence.