What question did this study set out to answer?

The study aims to improve the early detection and prediction of CKD severity using machine learning techniques.

May 15, 2026PeerJ Computer Science0 citationsOpen Access

Enhanced interpretable detection and stage prediction of chronic kidney disease through an ensemble learning approach

SRShabana Ramzan SKSidra Khalid MIMuhammad Munwar Iqbal

Key Points

The study aims to improve the early detection and prediction of CKD severity using machine learning techniques.
Utilized machine learning models like Random Forest, Logistic Regression, K-Nearest Neighbors, and Gradient Boosting.
Employed ensemble stacking and techniques like scaling and Recursive Feature Elimination for optimization.
Applied Local Interpretable Model-Agnostic Explanations for model explainability.
Achieved a peak accuracy of 99.16% for CKD detection using the ensemble stacking model.
Achieved 92% accuracy for CKD stage classification with the Gradient Boosting model.
Identified key features influencing predictions effectively.

Abstract

Chronic kidney disease (CKD) presents a significant global health challenge due to its life-threatening consequences and the necessity for earlier detection. This study aims to develop a dual framework for the early detection and severity prediction of CKD using advanced machine learning (ML) techniques. We utilized various ML models, including Random Forest (RF), Logistic Regression (LR), K-Nearest Neighbors (KNN), Gradient Boosting (GB), ensemble stacking, and Artificial Neural Network (ANN), enhancing their performance through techniques such as scaling, Recursive Feature Elimination (RFE), and the application of Local Interpretable Model-Agnostic Explanations (LIME) for explainability. Our findings show that the ensemble stacking model outperformed individual models in both accuracy and reliability, while also identifying key features that influence predictions. Furthermore, the ANN model excelled in recognizing complex patterns for CKD stage prediction, supported by KNN and GB models for optimal results. This study significantly advances CKD detection and stage prediction through ML and ensemble models, achieving a peak accuracy of 99.16% with a stacking ensemble model in detection and 92% with a GB model in stage classification, particularly excelling in differentiating intermediate CKD stages.

Demander à l'IA

Bookmark

View Full Paper