Research on Optimization and Management of Supply Chain Networks Based on Graph Neural Networks

In recent years, there has been a notable increase in the complexity and scale of supply chain networks, accompanied by a growing necessity for effective optimisation and management strategies. Conventional methodologies frequently prove inadequate for addressing the intricate dynamics inherent to the supply chain, particularly in the context of market volatility. In this study, we propose a supply chain network optimisation model based on a graph neural network (GNN), which is capable of effectively representing and learning complex relational data. The proposed model constructs a graph structure, wherein each node represents an entity within the supply chain, such as suppliers, manufacturers, and distributors. Edges represent the interactions and dependencies between these entities. The graph neural network enables the model to capture multi-hop dependencies and the flow of information through the network, thereby facilitating adaptive decision-making in response to fluctuations in demand, disruptions and cost changes. Furthermore, the optimisation process incorporates reinforcement learning, which enables the model to learn the optimal strategy in simulated supply chain scenarios, thereby enhancing overall efficiency and resilience. The experimental analysis demonstrates that the GNN-based model outperforms the traditional method on numerous performance indicators, exhibiting markedly enhanced accuracy and efficiency.