Article: Real-time early warning of business and financial risks via temporal knowledge graphs and dynamic graph neural networks Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.85 pp.20 - 46 Abstract: Supply and guarantee chains among enterprises create complex networks with significant risk transmission effects. Traditional early warning models relying on isolated financial indicators cannot capture this dynamic evolution, causing delayed and inaccurate warnings. This paper proposes temporal knowledge graph - dynamic graph neural network, a real-time framework integrating temporal knowledge graphs and dynamic graph neural networks. It constructs a multi-relation temporal knowledge graph containing enterprises, transactions, and guarantees to depict risk transmission paths, and designs a time-aware graph attention mechanism. Experiments on public datasets show the proposed method achieves 0.962 area under the curve, 3.2% higher than graphshield, and 0.781 normalised discounted cumulative gain at position 10, 8.7% higher than dynamic graph neural network with static relationships. Results demonstrate this technology effectively identifies early risk signals in network environments, providing timely and interpretable decision support for business and finance risk control. Inderscience Publishers - linking academia, business and industry through research

Title: Real-time early warning of business and financial risks via temporal knowledge graphs and dynamic graph neural networks

Authors: Hui Liu

Addresses: School of Accounting, Anhui Wenda University of Information Engineering, Anhui, 231201, China

Abstract: Supply and guarantee chains among enterprises create complex networks with significant risk transmission effects. Traditional early warning models relying on isolated financial indicators cannot capture this dynamic evolution, causing delayed and inaccurate warnings. This paper proposes temporal knowledge graph - dynamic graph neural network, a real-time framework integrating temporal knowledge graphs and dynamic graph neural networks. It constructs a multi-relation temporal knowledge graph containing enterprises, transactions, and guarantees to depict risk transmission paths, and designs a time-aware graph attention mechanism. Experiments on public datasets show the proposed method achieves 0.962 area under the curve, 3.2% higher than graphshield, and 0.781 normalised discounted cumulative gain at position 10, 8.7% higher than dynamic graph neural network with static relationships. Results demonstrate this technology effectively identifies early risk signals in network environments, providing timely and interpretable decision support for business and finance risk control.

Keywords: temporal knowledge graph; dynamic graph neural network; business and finance risk; early warning.

DOI: 10.1504/IJICT.2026.155485

International Journal of Information and Communication Technology, 2026 Vol.27 No.85, pp.20 - 46

Received: 03 Mar 2026
Accepted: 06 May 2026
Published online: 03 Aug 2026 *

Title: Real-time early warning of business and financial risks via temporal knowledge graphs and dynamic graph neural networks

Keep up-to-date