Tree-Based Federated Learning Models to Support the Diagnosis of Chronic Kidney Disease

Chronic Kidney Disease (CKD) is a widespread condition with significant clinical impact. Therefore, early detection is essential to reduce complications and healthcare costs. Researchers increasingly explore Machine Learning (ML) techniques to support clinical decision-making. However, centralizing clinical data raises concerns about privacy and data heterogeneity across healthcare institutions. In this work, we evaluate the performance of tree-based Federated Learning (FL) models for CKD diagnosis on a dataset from the UCI Machine Learning Repository. We compare FL models with a traditional centralized approach. We assess performance using the area under the curve, F1-score, recall, and balanced accuracy. The results show that tree-based FL models achieve performance comparable to centralized models for CKD diagnosis.