Article

Graph Neural Network-Based Risk Propagation Model for FinTech Interbank Networks in the Digital Economy

---

Abstract

The increasing interdependence between FinTech platforms and traditional banks has created a complex and dynamic financial ecosystem that is highly vulnerable to contagion. Existing systemic risk models are predominantly linear and static, lacking the ability to capture the temporal and relational characteristics of digital financial interactions. This study proposes a Graph Neural Network (GNN)-based risk propagation model designed to identify, simulate, and explain systemic contagion within FinTech interbank networks. The model represents each financial institution as a node and its exposures and transaction dependencies as directed, weighted edges, allowing it to learn multi-hop propagation dynamics in real time. Using integrated datasets of interbank exposures, payment transactions, and digital liquidity indicators, the model was trained and validated under temporal cross-validation schemes. Experimental results show that the proposed GNN outperforms baseline models Logistic Regression, Random Forest, and static Graph Convolutional Network (GCN) achieving an AUROC of 0.938 and an AUPRC of 0.812. The simulation of shock scenarios reveals that distress originating from major FinTech hubs can propagate system-wide within three propagation steps. Robustness and explainability analyses confirm the model’s reliability and transparency, identifying short-term funding ratio and transaction volatility as the most influential drivers of systemic risk. The findings highlight the significance of graph-based deep learning in enhancing early-warning systems and strengthening macroprudential surveillance in the digital economy.

Keywords: Graph Neural Network (GNN), FinTech, Systemic Risk, Interbank Network, Risk Propagation, Explainable AI (XAI), Digital Economy

How to Cite:

Yulianto, K. E. & Arifianto, M., (2025) “Graph Neural Network-Based Risk Propagation Model for FinTech Interbank Networks in the Digital Economy”, FinTech Innovation Journal 1(4), 309-324. doi: https://doi.org/10.63913/ftij.v1i4.84