Explainable GNN Framework Enhances Banking Surveillance
Key Takeaways
- 1New ST-GAT model detects early bank distress signals.
- 2Introduces advanced methods for macro-prudential supervision.
- 3Enhances national financial stability through data-driven insights.
The study presents the Spatial-Temporal Graph Attention Network (ST-GAT), designed to detect early warning signs of distress in U.S. banks. This framework models over 8,100 FDIC-insured entities using historical data from 2010 to 2024. It showcases a high AUPRC score, indicating effectiveness in identifying financial vulnerabilities, with publicly available data supporting its development and validation.
Source
arXiv cs.LG (Machine Learning)https://arxiv.org/abs/2604.14232
Read original