A Diabetes Prediction System Based on Federated Learning
Junzhe Liu, Xi Lü, Haolin Yang, Lianzhi Zhuang · 2022 International Conference on Big Data, Information and Computer Network (BDICN) · 2022
In recent decades, with the rapid development of society, people’s quality of life has improved, but the incidence of diseases such as diabetes has become more and more serious. According to authoritative literature, the occurrence of diabetes may be related to a person’s age, living environment, living habits, genes, past medical history and other factors. We can get more accurate models through machine learning training on a large amount of data, but in order to protect patients' private information between different hospitals, it is impossible to conduct training directly through data transmission. In view of this situation, we propose a solution combining homomorphic encryption and federated learning, and design and implement a privacy-protected diabetes prediction system based on federated learning. The experimental results demonstrated that it broke the phenomenon of information isolation among hospitals and successfully collected patient information from different hospitals, which could not only improve the accuracy of the trained model, but also effectively protect the privacy of patients. Therefore, this work is very innovative and practical, and has strong practical and application significance under the current social background. It can provide solutions for diabetes treatment in the medical field in the future, and it is expected to provide new ideas for multi-party data combination in different fields in the future.