Spatial-temporal dynamics for enhanced temporal link prediction in stock market crisis forecasting

Zhang, Ruizhi, Wei, Wei, Wu, Fan ORCID: https://orcid.org/0009-0009-0002-7434, Yang, Qiming, Chen, Jing ORCID: https://orcid.org/0000-0001-7135-2116, Liu, Anqi ORCID: https://orcid.org/0000-0002-9224-084X and Li, Yuhua ORCID: https://orcid.org/0000-0003-2913-4478 2026. Spatial-temporal dynamics for enhanced temporal link prediction in stock market crisis forecasting. Presented at: 2026 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Bellevue, WA, USA, 4-7 October 2026. SMC 2026 Conference Proceedings. IEEE, Item availability restricted.

	PDF (Provisional file) - Accepted Post-Print Version Download (17kB)
	PDF - Accepted Post-Print Version Restricted to Repository staff only Download (1MB)

Abstract

With the rapid development of production and consumption, stock market crises occur periodically, often driven by economic instability, market volatility, and shifts in investor sentiment. Predicting such crises remains a significant challenge for both scholars and investors. This paper addresses the crisis forecasting problem through temporal link prediction and proposes the CF2TLP model. It analyzes structured financial networks and explores potential contagion effects as a mechanism of information propagation within the network. Specifically, the model employs Graph Convolutional Networks (GCN) to capture complex spatial dependencies between nodes and integrates spatial features with temporal dynamics to construct comprehensive feature representations. It further leverages the transformer model to learn nonlinear evolutionary patterns and latent temporal structures, effectively characterizing the spatiotemporal complexity of the system. Experiments on stock data from U.S. technology companies demonstrate that the model effectively captures a notable increase in contagion effects prior to the collapse of Silicon Valley Bank (SVB), with both AUC and precision values significantly surpassing historical peaks, providing a reliable signal for crisis early warning. This result offers policymakers and investors valuable decision-making support and actionable insights.

Item Type:	Conference or Workshop Item - published (Paper)
Status:	In Press
Schools:	Schools > Computational & Mathematical Sciences
Publisher:	IEEE
Date of First Compliant Deposit:	11 August 2026
Last Modified:	11 Aug 2026 14:06
URI:	https://orca.cardiff.ac.uk/id/eprint/188912

Actions (repository staff only)

Edit Item