Ensemble classifiers with hybrid feature selection approach for diagnosis of coronary artery disease

Abstract

Coronary artery disease (CAD) is a common type of cardiovascular disease with a high mortality rate worldwide. As symptoms may not be recognized until, after the cardiac attack, early diagnosis and treatment are critical to lowering mortality. The proposed study focuses on the creation of an intelligent ensemble system for the accurate detection of CAD. This paper presents the hybrid feature selection method based on Lasso, random forest-based boruta, and recursive feature elimination methods. The significance of a feature is determined by the score each approach provides. Machine learning techniques such as random forest, support vector machine, K-nearest neighbor, logistic regression, decision tree, and Naive Bayes are developed as base classifiers. Then, ensemble techniques like bagging and boosting models are created using base classifiers. The Z-Alizadeh Sani dataset was used to build and test the model. The bagged random forest model achieved 97.6% accuracy and 100% recall. The CatBoost model achieved 97.7% accuracy and 99.0% recall. Compared to traditional classifiers, the ensemble models achieved higher accuracy and can be used to assist clinicians in diagnosing coronary artery disease