Leveraging Graph Neural Networks for Intelligent Supply Chain Risk Management in the Era of Industry 4.0

The complexity of global supply chains introduces significant risks, ranging from supplier failure to transportation disruptions, which can critically impact industry performance. This study presents a novel approach to supply chain risk detection using Graph Neural Networks (GNNs), leveraging the structural dependencies between suppliers, products, and logistics. Data collected from the UCI Machine Learning Repository was preprocessed and modeled to capture both node-level and relational features. Comparative experiments with baseline models including Logistic Regression, Random Forest, and Gradient Boosting revealed that GNN-based models, particularly Graph Attention Networks (GATs), outperformed traditional methods across all evaluation metrics. The GAT achieved an accuracy of 92.1% and an AUC of 0.97, significantly surpassing classical models in identifying high-risk suppliers. Bar chart and ROC curve analyses further validated the superiority of graph-based learning in extracting meaningful patterns from relational data. These results demonstrate the transformative potential of integrating GNNs into U.S. industry supply chains, where real-time risk detection can be embedded into AI-driven decision-support systems, enabling more resilient and adaptive supply chain management.