Ieee federated Machine Learning White Paper
View original pdf
Navigate this page:
Categorization of Federated Learning
Authorized licensed use limited to University of Malta. Downloaded on December 24,2022 at 11:03:39 UTC from IEEE Xplore. Restrictions apply.
Copyright © 2021 IEEE. All rights reserved.
OF FEDERATED MACHINE
The application and commercialization of federated machine learning in the industry include different use cases, which have different features of the federated machine learning function. Federated machine learning application areas are broadly categorized as three types, according to requirements from
different marketing sectors
, i.e., business-to-consumer (BC, business-to-business (BB, and business-to-government (B2G). Each category has respective requirements on FML methods concerning privacy protection, model performances, and efficiency etc.
Finance or financial services is an important area that can be greatly improved with the use of AI and big data. Traditionally financial companies or banks make business decisions based on their data such as
information from bank accounts
, credit card use, and loan history, which might be insufficient to evaluate customers financial risks because these data only present a small part of user behavior needed for risk modeling.
, customers yearly income, real estate ownership, and shopping history may provide
more valuable information
, but these are the private information of users that need to be protected. In financial application scenarios, regulatory requirements and privacy concerns prevent banks and financial companies from sharing their data. The main risks faced by financial institutions are overdue loans and fraudulent loans caused by user credit risk and even fraud. Traditional financial institutions may only know users borrowing history and behavior locally, but they know
little about users interests
, consumption tendencies, behavior, and other private information. To conduct modeling without involving privacy leakage and improve the
assessment of risks of loans
, the traditional practice is to provide each institution with a separate model and integrate all model’s results to get the evaluation result. However, this modeling method often has low performance and the obtained result may not be accurate enough. Federated machine learning can solve this problem by jointly modeling the users overall behavior across many sectors and financial institutions, without compromising model performances. By adopting FML methods, each data holder can exchange encryption parameters to jointly train a model and obtain more reliable evaluation results when the data is retained locally. It can help financial institutions avoid risks more effectively.
Share with your friends:
The database is protected by copyright ©ininet.org 2023
prior written permission