Graph Neural Networks for Complex Relationship Modeling in Supply Chain Analytics

Abstract

Research on the use of graph neural networks (GNNs) in supply chain management is still scarce, despite the fact that they have lately acquired popularity in the fields of language, image processing, bioinformatics, and transportation. Because supply chains are graph-like by nature, they are perfect for GNN approaches, which may optimise and resolve challenging issues. This research uses the Supply Graph dataset, a standard for graph-based supply chain analysis, to examine how GNNs might be used to demand forecasting in supply chain networks. Utilising cutting-edge GNN techniques, we improve forecasting model accuracy, reveal hidden relationships, and handle the temporal complexity that comes with supply chain processes. Because of their intrinsic graph-like structure, supply chain networks are excellent candidates for the use of GNN techniques. Therefore, it possible to anticipate, optimise, and resolve even the most challenging supply chain issues. Since graphs allow researchers to examine linkages and improve networks in addition to identifying patterns, their use makes it possible to conduct thorough data analysis. Graphs' fundamental ideas, applications, and analytical techniques for complex system analysis are all examined in this paper. The research offers key analytical methodologies, including graph clustering methods, shortest route algorithms, and network centrality measurements.  These findings demonstrate the usefulness and need of graph-based models for solving real-world problems via their in-structure analysis