Early Stage Detection of Chronic Kidney Disease using Machine Learning

Chronic Kidney Disease (CKD) is an illness that is long term and progresses over time causing people to lose their kidney function. This study aims to detect CKD at an early stage through the use of machine learning (ML) techniques, XGBoost, Gradient Boosting, AdaBoost, Random Forest, and Support Vector Machine (SVM); they were trained on clinically relevant biomarkers selected specifically for the early detection of the disease. A dataset of 1,659 patient records was refined so that the emphasis was on early stage risk factors. The results demonstrate consistently high predictive performance across all models, with accuracy exceeding 91% and F1-scores above 95%. AdaBoost got the best performance overall with an F1-score of 96.3% and recall of 99.7%, indicating its strong ability to correctly identify CKD cases while reducing false negatives which is an essential requirement in clinical diagnostics. The predictive performance of the model was high even when the Receiver Operating Characteristic Area Under the Curve (ROC-AUC) scores were relatively low. The novelty of this research lies in its concentration on the selection of biomarkers during the initial stages, guided by clinical thresholds, along with the comparison between various ensemble algorithms using a clinically meaningful criterion framework. This work highlights how machine learning algorithms can be used as a potential diagnostic tool for early CKD screening and can support timely intervention and better healthcare outcomes in the real world scenario.