Heart Disease Prediction: Leveraging CNNs and XAI Modeling for Interpretability

This study examines heart disease prediction using machine learning models, focusing on interpretability, leveraging a dataset of 1,319 entries containing nine health-related attributes. We applied One-dimensional Convolutional Neural Networks (1D CNN) and Logistic Regression for classification, achieving an overall accuracy of 80%. Our analysis includes Local Interpretable Model-Agnostic Explanations (LIME) to provide transparency into the predictions by identifying influential features, including age, troponin, CKMB, glucose, and systolic blood pressure. The model’s binary classification results, with a precision of approximately 75% and F1-scores favoring class 1 (patients with heart disease), reflect high predictive reliability. Macro and weighted averages, around 0.79 and 0.80, respectively, further indicate balanced performance despite class imbalances. Local interpretations also illustrate predictive support from critical health markers, with individual predictions yielding notable intercepts, confirming both positive and negative diagnoses. These results demonstrate the value of interpretable machine learning for advancing clinical decision-making and offer insights into ke