An Effective Cardiovascular Risk Prediction Using PCA and mRMR Feature Selection with Multi-Model Explainable AI Framework

Cardiovascular disease (CVD) continues to be the leading cause of death worldwide, claiming around 18 million lives each year. Predicting risk early and intervening promptly is especially important in resource-limited healthcare settings. This study introduces a machine learning framework for cardiovascular risk prediction using the Framingham Heart Study dataset, which includes 4,240 records. The proposed framework combines dimensional reduction technique PCA (Principal Component Analysis) and feature selection method mRMR (minimum Redundancy Maximum Relevance) with seven machine learning classifiers Extra Trees Classifier, Random Forest, Gradient Boosting, CatBoost, XGBoost, Decision Tree, and Multilayer Perceptron. To make the predictions interpretable for clinicians, SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations) were applied as a dual explainable AI (XAI) framework. Overall, the framework performed best when PCA was combined with the Extra Trees Classifier, reaching 0.9763 accuracy, 0.9797 recall, 0.9780 precision, 0.9762 F1-score, and 0.9763 ROC-AUC using 10 -fold stratified cross-validation with standard deviation. XAI analysis highlighted age, systolic blood pressure, and glucose as the most important predictors. By balancing high predictive performance with interpretability, this framework provides a transparent and scalable tool for cardiovascular risk assessment, which could be particularly useful in developing regions.