Akademik Veri Yönetim Sistemi
2023 International Symposium on Networks, Computers and Communications, ISNCC 2023, Doha, Qatar, 23 - 26 Ekim 2023, (Tam Metin Bildiri)
Machine Learning (ML) techniques for decision-making have become a topic of popular interest for many organizations. Such models, however, depend on data availability which is not always possible. Banks, for example, often have limited data about their customers. While this data could be supplemented with information from other banks, data sharing leads to privacy concerns due to regulations such as the General Data Protection Regulation (GDPR) and the California Customer Privacy Act (CCPA). An alternative method proposed to solve such a problem is Federated Learning (FL), where the data is not shared, but the models created by each entity are shared to keep privacy. FL has been applied across various sectors and industries, but not all. In this paper, we implement an FL algorithm in a new sector, namely, banking and loans. We used an architecture where multiple banks within the same country, each possessing different customer information, want to evaluate eligibility for loan provision. Our objective is to design an FL model where the banks share the data without exposing any customer hidden information. We implemented our FL model using Flower on the publicly available Loan dataset on Kaggle. Our model achieved an accuracy of 81%.