Application of AI Models Integrating Multi-Source Data in Risk Early Warning and Dynamic Optimization of Smart Supply Chain in Cross-Border E-Commerce
Year: 2026
Type: conference-paper
Abstract: Cross-border e-commerce supply chains face multiple risks such as demand fluctuations, logistics disruptions, and public opinion shocks. This article uses a multi-view heterogeneous graph neural network to map entities such as suppliers, ports, inventory units (SKUs), and public opinion events to a unified semantic space, and utilizes a cross modal attention mechanism to capture the nonlinear dependencies between public opinion text and logistics temporal data. This article further constructs a dual time convolutional network, which uses the large core global branch to perceive systemic risks such as port congestion, and the small core local branch to identify anomalies such as sudden changes in public opinion. Combined with adaptive thresholds, it achieves minute level risk warning. Finally, the near end strategy optimization algorithm is used to train the intelligent agent, taking risk probability, inventory level, etc., as state inputs, and outputting a mixed action of procurement adjustment and path switching. A reward function is constructed based on inventory turnover rate, performance cost, and risk exposure to drive dynamic optimization. Experiments show that the model achieves an F1 score of 91.1% in risk warning, with a warning response time of only 4.7 minutes under sudden events. After dynamic optimization, the number of supply chain disruptions and overall operating costs in extreme scenarios are reduced, providing a data-driven intelligent solution for improving supply chain resilience in complex environments.
Source: Procedia Computer Science
Author: Xu Zhao
Institution: Liaoning Vocational University of Technology
Language: English
Subfield: Business and International Management
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7163133643