24th Conference of the International Federation of Operational Research Societies
Abstract Submission

140. XGNN-RCI: Explainable Graph Neural Networks with Regional Contagion Intelligence for Systemic Risk Prediction in South Asian Emerging Markets

Invited abstract in session Sustainable and Generative AI in SCM for Emerging Economies , stream OR for Development.

Authors (first author is the speaker)

1. BUSHRA SAEED
School of economics and management, southeast University
2. WEI YIN
School of economics and management, SOUTHEAST UNIVERSITY

Abstract

Systemic crises propagate through interconnected regional networks. We introduce XGNN-RCI, an explainable graph neural network for systemic risk prediction in South Asian markets. Our Regional Contagion Index captures cross-border spillovers with 33% predictive importance. RCI-Attention dynamically adapts message passing, yielding 27% F1 improvements. Analyzing India, Pakistan, Bangladesh, Sri Lanka data achieves F1=0.816, AUC=0.946. Validated on 2022 Sri Lankan crisis with one-quarter advance detection. Deployment-ready for central banks.

Keywords

Status: accepted


Back to the list of papers