Leveraging Federated Learning With XGBoost for Secure and Privacy-Preserving Hemochromatosis Diagnosis in Distributed Healthcare Systems

Abstract

This paper illustrates a platform of feature-rich, privacy-conscious hemochromatosis diagnosis with the help of federated learning and XGBoost. A genetic dataset of features, which included species, gene identifiers, genetic entities, disease associations and evidence codes, was used to apply the performed model. Gene ID, Gene Symbol and Genetic Entity Type were the principal biological attributes that were utilized to provide important genetic correlations with hemochromatosis across the distributed healthcare systems. Using a federated architecture, patient data will be stored at the location, but one can collaboratively train the model and grant access to patient data to adequate standards under the Health Insurance Portability and Accountability Act. The XGBoost model was more precise as compared to the random Forest as it had an exact diagnosis of 95.6 % when compared to the random forest at 82.7 %. These findings, on the whole, suggest that XGBoost in a federated architecture has the capacity to make initial diagnosis of hemochromatosis using available structured biological information in a secure, scalable and most accurate fashion. Keywords- Federated Learning, XG-Boost, Random Forest, Hemochromatosis, Distributed Healthcare, Machine Learning.

Ähnliche Arbeiten