Machine learning model for accurate prediction of coronary artery disease by incorporating error reduction methodologies

Abstract

Coronary artery disease (CAD) remains a leading cause of mortality worldwide, with an especially high burden in developing countries such as India. In light of increasing patient loads and limited medical resources, there is an urgent need for accurate and reliable diagnostic support systems. This study introduces a machine learning (ML) framework that aims to enhance CAD prediction accuracy by specifically addressing the reduction of false negatives (FN), which are critical in medical diagnostics. Utilizing a stacked ensemble model comprising five base classifiers and a meta-classifier, the framework integrates cost-sensitive learning, classification threshold tuning, engineered features, and manual weighting strategies. The model was developed using a clinically acquired dataset from the Jawaharlal Institute of postgraduate medical education and research (JIPMER), consisting of 428 patient records with 36 original features. Evaluation metrics show that the proposed model achieved an accuracy of 92.19%, sensitivity of 98%, and an F1-score of 95.15%. These improvements are significant in a clinical context, potentially reducing missed diagnoses and improving patient outcomes. The model is intended for deployment in cardiology outpatient settings and demonstrates a scalable, adaptable approach to medical diagnostics.