Machine Learning Algorithms for Predicting CKD Progression: A Real-World Hospital Dataset Analysis

Abstract

Background. Chronic Kidney Disease (CKD) is a progressive condition associated with substantial global morbidity and mortality. Early detection remains critical for reducing complications and slowing progression to end-stage kidney disease. Traditional diagnostic approaches depend on laboratory markers that may not fully capture nonlinear interactions among clinical parameters. Machine learning offers promising capabilities for improving early identification and supporting clinical decision-making. Methods. This study developed an end-to-end machine learning framework for CKD prediction using the Early Stage CKD dataset. The workflow included rigorous data preprocessing, exploratory data analysis, and feature engineering prior to model development. A Random Forest classifier was trained using an 80/20 stratified split, and performance was assessed using accuracy, precision, recall, F1-score, confusion matrix, and ROC–AUC. To enhance transparency, SHAP (SHapley Additive exPlanations) analysis was applied to interpret feature contributions and validate clinical relevance. Results. The Random Forest model demonstrated excellent predictive performance, achieving an accuracy of 96.25% and a ROC–AUC of 1.00. The confusion matrix indicated zero false positives and only three false negatives, reflecting strong diagnostic reliability. SHAP analysis identified hemoglobin, serum creatinine, packed cell volume, and specific gravity as the most influential predictors, aligning with established CKD biomarkers. Conclusion. The proposed machine learning framework offers a robust, interpretable approach for early CKD prediction. Its strong performance and explainability make it suitable for integration into real-world clinical decision-support systems, particularly in resource-limited healthcare settings.