A Hybrid XGBoost-Deep Neural Network Framework with SHAP Explainability for Early Chronic Kidney Disease Detection

View at publisher Summarise with VIRAT
Authors: Dhanabalan R, Hariprakash G, G. Indra Navaroj
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Chronic Kidney Disease or CKD is a disease that is usually silent and progressive and often diagnosed in its later stages hence detecting it at the earliest stage is paramount. However, it is extremely difficult to be detected early. Traditional diagnosis involves clinical tests and their analysis by professionals that involves high labor and costly, not easy to scale up to large population to conduct mass screening. The current project suggests an automated CKD prediction system, which combines conventional machine learning features like SVM, Random Forest, and XGBoost with a Deep Neural Network through effective preprocessing, dealing with the imbalance in classes, and model interpretability with SHAP. It is desired to achieve high accuracy and sensitivity to detect potential cases of early detection and interpretability to give clinical trust. The newness of the system will lie in the integration of XGBoost with DNN to enhance the outputs of the systems in prediction and having a deployable web application based on Flask to detect CKD in real-time. It will reduce the clinical load in diagnostics, assist clinical workers in making more valid decisions and offer an AI-based tool that can be adapted to largescale solutions in early CKD intervention.

Fields of Study & Subjects

Artificial Intelligence in Healthcare Chronic Kidney Disease and Diabetes Machine Learning in Healthcare

Publication details

Journal / source
International Conference on Intelligent Sustainable Systems (ICISS)
Pages
1198–1203
Publication date
4 March 2026
Language
EN