Hybrid explainable machine learning for population level cardiovascular disease risk prediction using satellite environmental data

Abstract

Cardiovascular disease (CVD) remains one of the main causes of death worldwide, and better risk prediction is needed to support early prevention and public health planning. Most existing prediction studies focus mainly on clinical variables, while environmental and regional influences are often underused. This study proposes a Hybrid Explainable Artificial Intelligence and Machine Learning (HXAI-ML) framework for area-level CVD risk prediction using health-related and satellite-derived environmental data at the Core-Based Statistical Area (CBSA) level. The dataset comprised 329 observations and 80 variables. Data preprocessing included median imputation, standardization, correlation-based feature filtering, and binary class labeling. To address class imbalance, several resampling strategies were evaluated, including Random Oversampling, SMOTE, Tomek Links, Instance Hardness Threshold, and hybrid combinations of these methods. Five ensemble classifiers were then assessed: Decision Tree, Random Forest, Extra Trees, Gradient Boosting, and Extreme Gradient Boosting. Model performance was evaluated using accuracy, precision, recall, F1-score, ROC-AUC, Kappa, Matthews Correlation Coefficient, MAE, MSE, and RMSE. Among all tested configurations, SMOTE combined with Extreme Gradient Boosting achieved the strongest performance, with 95.45% accuracy, 97.67% precision, 95.45% recall, and 96.55% F1-score. To enhance transparency, SHAP, LIME, and Permutation Importance Analysis were applied to identify and explain the most influential predictors. Overall, the proposed framework offers a robust and interpretable approach for population-level cardiovascular risk assessment and environmental health decision-making.