A Comprehensive Evaluation for Coronary Heart Disease Prediction Using Deep Learning Models

Coronary heart disease (CHD) remains a leading cause of global death, underscoring the need for accurate and reliable predictive models that support early detection and prevention. In this study, we present a comprehensive evaluation of deep learning architectures aimed at predicting coronary heart disease. We focus on three custom-built deep learning model Convolutional Neural Networks (CNN), Multilayer Perceptions (MLP), and a hybrid model combining various neural architectures-alongside a pre-trained Tabular Data-Specific model, TabNET. Using the UCI dataset, we train and assess each model's performance based on key metrics, including accuracy, precision, recall, and F1-score. Hybrid model, achieved a remarkable level accuracy of 98.54%, with a weighted average precision of 100%, recall of 97.09%, and F1-score of 98.52%, reflecting its strong predictive capability on the dataset. Additionally, we explore the interpretability of these models, a crucial aspect for medical applications, by analyzing the contribution of individual features within each architecture. Our results highlight the advantages and limitations of these models in the circumstance of coronary heart disease prediction, with Hybrid architecture emerging as a particularly effective model. This analysis not only deepens the understanding of deep learning models for medical prediction but also offers valuable guidance for future research and the development of predictive tools in clinical settings.