Advancing Health Diagnostics: AI-Powered CVD-REF Framework for Precise and Early Risk Assessment

Deprivation of Critical Care systems are a major cause of fatality worldwide, highlighting it’s need for saving human lives. This study proposes a novel hybrid ensemble model, which integrates Random Forests, Gradient Boosting Machines (GBM), and Neural Networks to enhance the predictive accuracy diagnostics. The methodology combines data pre-processing, feature selection, and ensemble learning, ensuring robust and reliable predictions. Comprehensive data pre-processing includes K-Nearest Neighbours (KNN) imputation for missing values, Z-Score normalization for scaling, and Polynomial Feature Generation for non-linear feature interactions. Feature selection performed using Recursive Feature Elimination (RFE) and Mutual Information relevant variable retention. The proposed model produces 98.55% accuracy, very surpassing nine baseline models, that includes XGBoost, Random Forests, and Neural Networks. Additional metrics such as precision (97.80%), recall (98.12%), F1-Score (98.00%), and ROC-AUC (99.12%) further validate the model's robustness. This framework not only demonstrates superior accuracy but also ensures computational efficiency, making it viable for deployment in real-world healthcare settings.