Chronic Kidney Disease Stage Prediction using Multi-Dimensional Healthcare Data

Chronic Kidney Disease (CKD) prediction is essential in healthcare and can influence the fate of patients due to early detection of such diseases. Most of the currently adopted techniques are majorly dependent on clinical data, while this study analyses predicting CKD stages from NHANES medical datasets through advanced machine learning models. Different models are employed, where XGBoost serves as a baseline classification model, and Random Forest, K-Nearest Neighbors (KNN), and Gradient Boosting are used for comparison purposes. In addition to this, ensemble learning methodologies, soft-voting, and stacking classifiers are also used to combine the predictions of individual models for great performance. The results indicate that there was very high classification accuracy, with the stacking classifier at 96% accuracy and a soft-voting classifier at 95%. Along with that, the Transformer Tab Net classifier attains a classification accuracy of 96%. It demonstrated its effectiveness in predicting CKD stages.