A PRELIMINARY STUDY ON A MULTI-BANK JOINT CREDITRISK CONTROL MODEL BASED ON FEDERATEDLEARNING

Abstract

The increasing complexity of credit risk management in the banking sector, coupled with stringent data privacy regulations, has limited the ability of financial institutions to collaboratively develop robust risk control models. Traditional approaches often rely on centralized data sharing, which raises significant privacy and security concerns. This study proposes a multi-bank joint credit risk control model leveraging federated learning, a decentralized machine learning technique that enables collaborative model training without exposing raw data. The primary objective is to enhance the accuracy and generalizability of credit risk prediction while preserving data privacy. Using a simulated dataset representing heterogeneous credit data from multiple banks, we implemented a federated learning framework with a logistic regression baseline and a deep neural network variant. Experimental results demonstrate that the proposed model achieves predictive performance comparable to centralized training methods, with an average F1-score improvement of 7.3% over isolated bank-specific models. Additionally, the framework effectively addresses non-IID (non-independent and identically distributed) data challenges across institutions. The findings highlight the potential of federated learning as a scalable and privacy-preserving solution for multi-institutional credit risk management, offering practical implications for regulatory technology and collaborative financial ecosystems.