CARDIAC DISEASE PREDICTION USING MACHINE LEARNING AND DEEP LEARNING: A COMPREHENSIVE MULTI-MODEL CLINICAL EVALUATION WITH RECALL OPTIMIZATION

Abstract

Cardiovascular diseases (CVDs) remain the leading cause of mortality worldwide, claiming approximately 17.9 million deaths annually. Early detection of high-risk individuals is critical to reducing mortality rates, yet this remains challenging due to the multifactorial and nonlinear nature of cardiac risk. This paper presents a comprehensive review and experimental evaluation of machine learning (ML) and deep learning (DL) approaches for cardiac disease prediction, examining 32 model configurations systematically assessed across a five-phase study on 10,585 patient records characterised by 26 clinical variables. Starting from baseline classifiers and building toward ensemble methods, neural networks, automated hyperparameter tuning, probability calibration, and a clinically motivated recall-maximisation framework, the work traces the incremental contribution of each methodological step. Four primary contributions are synthesised: (i) three engineered composite features—Clinical Risk Score, Cardiac Load, and Metabolic Index—constructed from cardiovascular domain knowledge and validated through mutual information analysis; (ii) a comparative evaluation of four class-imbalance correction strategies, from standard SMOTE to an aggressive 3× oversampling scheme with asymmetric cost weighting; (iii) isotonic probability calibration enabling a validated three-tier patient risk stratification in which the high-risk tier demonstrates a 94.2% observed cardiac event rate; and (iv) a threshold-adjusted Random Forest achieving 99.21% sensitivity, with only 6 of 755 confirmed cardiac events undetected. This review demonstrates that data preparation—specifically clinical feature engineering and class-imbalance correction—contributes more to predictive performance than model architecture choice, and traces a reproducible path from general-purpose ML classifiers toward a clinically oriented, patient-safety-governed screening tool.