An Explainable Ensemble Learning Framework Optimized by MSSO for Kidney Disease Prediction

Chronic kidney disease (CKD) is a long-term clinical situation associated with gradual deterioration of kidney function, increased risk of morbidity if not identified at an early stage. Reliable classification of CKD using clinical data can assist healthcare professionals in timely diagnosis and treatment planning. The application of, machine learning (ML) techniques approaches a significant potential in medical decision-support systems due to their ability to analyze complex and heterogeneous datasets. In this study, a systematic ML-based framework is presented for CKD classification using supervised learning models, including Logistic Regression (LR), Decision Tree (DT), k-Nearest Neighbors (KNN), Gradient Boosting (GB), Extreme Gradient Boosting (XGB), and Random Forest (RF), integrated with the proposed Feature Optimization (FO) Multi-Stage Stacking Optimizer (MSSO). Comprehensive data preprocessing is performed to address missing values, categorical feature representation, and feature scaling. All models undergo 10-fold cross-validation (CV) to provide a thorough and unbiased measure of their predictive performance. Model performance is evaluated through a comparative analysis using standard classification metrics. Experimental findings indicate that the RF classifier optimized through MSSO achieves the highest predictive performance with an accuracy of 99.50%, demonstrating superior stability and generalization compared to the other models. The results suggest that ensemble learning methods, when combined with appropriate data preprocessing and the MSSO optimization strategy, can provide effective and reliable support for automated kidney disease classification, offering practical relevance for clinical decision-making and digital healthcare systems.