Machine learning modelling and explainability of coronary heart disease based on Mediterranean diet

Background: Coronary heart disease (CHD) occurs due to the narrowing or blockage of coronary arteries caused by atherosclerosis. It is one of the leading factors of widespread mortality and morbidity. The latest research highlighted the importance of the Mediterranean diet (MD) as an excellent cardioprotective nutritional regimen because of its abundant content of monounsaturated fats, antioxidant-rich compounds, and anti-inflammatory nutrients. Conventional CHD risk models frequently overlook food habits, highlighting the need for sophisticated predictive modeling that includes lifestyle aspects. Objectives: We aim to use machine learning (ML) for the prediction of CHD by combining adherence to the MD with clinical characteristics. Method: For the present study, we employed Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Decision Tree (DT), Adaptive Boosting (AdaBoost), Multilayer Perceptron (MLP) Classifier, Gaussian Naive Bayes (GNB) on the MD dataset and its overall diversity on cumulative preventive effects against CHD. The dataset was published on 26 April 2021 by Mendeley. Result: The results, as shown in this study, indicate that RF performed excellently with 0.90, 0.95, 0.95, and 0.90 as accuracy, precision, recall, and F-1 score values, respectively. Shapley additive explanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) showed that Glucose, high-density lipoprotein cholesterol (HDL-C), bread, and chocolate have a high impact on CHD prediction. Conclusion: ML models have shown the great potential that MD has as a cardioprotective nutritional regimen for the prediction of CHD.