DATA-DRIVEN EARLY DIAGNOSIS OF CHRONIC KIDNEY DISEASE: DEVELOPMENT AND EVALUATION OF AN EXPLAINABLE AI MODEL

Chronic Kidney Disease (CKD) is becoming an increasingly common health concern and, if not identified early, canresult in severe complications and premature death. The integration of Artificial Intelligence (AI) and MachineLearning (ML) can assist in detecting CKD at earlier stages, potentially minimizing kidney deterioration. However,the lack of interpretability in AI models often deters clinical adoption. To address this, explainable AI (XAI)methodologies are employed to enhance the transparency of CKD prediction models. This project focuses oncreating a data-centric and interpretable model that quantifies the influence of specific clinical indicators inforecasting CKD. The proposed approach utilizes an Extreme Gradient Boosting classifier with three key features—hemoglobin level, specific gravity, and hypertension—validated using 5-fold cross-validation. The modeldemonstrates strong performance on new, unseen data, offering a cost-effective solution for early CKD diagnosis.