Optimized Ensemble Learning for Chronic Kidney Disease Prognostication: A Stratified Cross-Validation Approach

Chronic Kidney Disease (CKD) is a critical global health issue necessitating early and accurate diagnosis to mitigate its progression. This study presents a comprehensive machine learning framework for CKD prediction, employing advanced ensemble methods optimized through rigorous hyperparameter tuning and validated using 10 -fold Stratified Cross-Validation. The models implemented include Support Vector Machines (SVM), LightGBM, CatBoost, Random Forest, and a Stacking ensemble integrating LightGBM, SVM and Random Forest. The framework addresses challenges in handling categorical data with Label Encoding, imputation of missing values, and feature scaling for models sensitive to data distributions. Experimental results demonstrate that the Stacking ensemble achieves superior performance with an accuracy of $\mathbf{9 9. 5 \%}$, outperforming individual models, with CatBoost and Random Forest also exhibiting strong predictive power. Feature importance analysis further enhances interpretability, aligning with clinical markers relevant to CKD. This work underscores the potential of optimized ensemble learning models to improve diagnostic accuracy and interpretability in clinical decision support systems for CKD.