Article: Energy supply chain financial risk prediction based on GNN and multi-source time series data Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.1 pp.60 - 76 Abstract: The energy industry is characterised by strong cyclicality, network infection risk, and multi-source data heterogeneity, making it difficult for traditional static assessment models of supply chain finance to meet dynamic early warning needs. To address these challenges, this paper proposes a dynamic heterogeneous graph neural network (DHGNN) model, which integrates a dynamic graph structure with cross-modal fusion of multi-source time series data, including finance, logistics, and public opinion. The core innovations of the model include a spatiotemporal attention mechanism, a dynamic graph construction module, and an industry-specific indicator system. Verification on the energy industry dataset demonstrates that the early warning accuracy reaches 96.2% (F1 = 0.974, AUC (area under the ROC curve) = 0.962), with an average early warning time of 6.8 months ahead of schedule, which is 40% higher than existing models, and the risk transmission path identification accuracy increased by 32%. Inderscience Publishers - linking academia, business and industry through research

Title: Energy supply chain financial risk prediction based on GNN and multi-source time series data

Authors: Pei Wan; Juanjuan Jiang

Addresses: School of Management, Guangzhou City University of Technology, Guangdong, 510800, China ' School of Management, Guangzhou City University of Technology, Guangdong, 510800, China

Abstract: The energy industry is characterised by strong cyclicality, network infection risk, and multi-source data heterogeneity, making it difficult for traditional static assessment models of supply chain finance to meet dynamic early warning needs. To address these challenges, this paper proposes a dynamic heterogeneous graph neural network (DHGNN) model, which integrates a dynamic graph structure with cross-modal fusion of multi-source time series data, including finance, logistics, and public opinion. The core innovations of the model include a spatiotemporal attention mechanism, a dynamic graph construction module, and an industry-specific indicator system. Verification on the energy industry dataset demonstrates that the early warning accuracy reaches 96.2% (F1 = 0.974, AUC (area under the ROC curve) = 0.962), with an average early warning time of 6.8 months ahead of schedule, which is 40% higher than existing models, and the risk transmission path identification accuracy increased by 32%.

Keywords: graph neural network; GNN; multimodal time series fusion; risk dynamic prediction; energy supply chain finance.

DOI: 10.1504/IJICT.2026.151313

International Journal of Information and Communication Technology, 2026 Vol.27 No.1, pp.60 - 76

Received: 12 Sep 2025
Accepted: 19 Oct 2025
Published online: 22 Jan 2026 *

Title: Energy supply chain financial risk prediction based on GNN and multi-source time series data

Keep up-to-date