Optimizing Myocardial Infarction Prediction Performance Using Advanced Machine Learning Techniques

The rising prevalence of Myocardial Infarction (MI) and limited clinical resources highlight the need for accurate, automated diagnostic tools. This study presents a Machine Learning (ML) framework for early MI prediction using both structured health records and Electrocardiogram (ECG) data. Multiple ML algorithms—including ridge classifier, radius neighbor classifier, linear SVC, and extra trees classifier—are evaluated on two publicly available datasets and two clinical datasets collected from hospitals. The additional trees classifier achieves the highest training accuracy of 1.00, with consistent performance across datasets. For ECG-based diagnosis, a deep learning model combining Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) is developed using the ECG Heartbeat Categorization Dataset. It classifies five heartbeat types: normal, Fusion of Paced and Normal (FPAN), Fusion of Ventricular and Normal (FVAN), Atrial Premature Contractions (APC), and Premature Ventricular Contractions (PVC). The model achieves a testing accuracy of 0.98, supported by strong precision and recall across classes. The novelty of this study lies in its integration of public and real-world datasets, noise-augmented training to improve ECG robustness, and a multi-class CNN–RNN framework that enhances generalizability beyond conventional binary classifiers. The proposed approach contributes to more reliable and interpretable cardiovascular diagnostics, with strong potential for clinical deployment and improved patient outcomes.