Cardiovascular Disease Detection Using Machine Learning Models

Cardiovascular diseases (CVDs) are the leading cause of death worldwide. Therefore, early detection of CVD is crucial for managing the health outcome as well as reducing the burden on healthcare. Artificial intelligence-based solutions have been studied in this domain to improve CVD detection and management. In this article, our objective is to detect the onset of CVD with high accuracy using biomarkers that are easily accessible in the community. We used three publicly available datasets from Kaggle and the UCI Repository to build machine learning models for the detection of CVD. Three different feature subset selection methods - genetic algorithm, recursive feature elimination, and Chi-square-based techniques - were applied to the datasets that supported the Ensemble-based Extreme Tree Classifier (ETC) model, achieving the best results with 98.35%, 99.61%, and 94.02% accuracy on these datasets. After feature subset selection, the RF-based model outperformed the existing models in the literature for all three datasets. We also identified the features (e.g., age, gender, fasting blood sugar, type of chest pain, cholesterol, and exercise-induced angina) that contributed to the improvement in the AI model's performance. We believe the proposed model will support the early detection of CVD with high accuracy in a clinical setup and reduce the healthcare burden.