GNN-based Probabilistic Supply and Inventory Predictions in Supply Chain Networks

TL;DR

This paper tackles the challenge of aligning supply and demand across complex supply chain networks by introducing GSP, a probabilistic, graph neural network framework that predicts edge-level shipment quantities/timings and node-level inventories. It uses Graph Attention Networks to produce network-wide embeddings, and models planned events with a probabilistic delta distribution via GumbelSoftmax, while enforcing capacity constraints through iterative inference. A loss combining cumulative edge-level and node-level errors (controlled by $\alpha$) enables coherent optimization of both shipment timing/quantities and inventory levels. Experiments on a large-scale, real-world supply chain dataset show GSP achieving superior daily and weekly predictive accuracy over baselines, with notable improvements when balancing edge and node predictions, and the approach offers actionable corrections to supply plans. The work advances AI-enabled supply chain planning by providing a theoretically principled, scalable framework for integrated demand-supply-inventory prediction under cascading network dynamics.

Abstract

Successful supply chain optimization must mitigate imbalances between supply and demand over time. While accurate demand prediction is essential for supply planning, it alone does not suffice. The key to successful supply planning for optimal and viable execution lies in maximizing predictability for both demand and supply throughout an execution horizon. Therefore, enhancing the accuracy of supply predictions is imperative to create an attainable supply plan that matches demand without overstocking or understocking. However, in complex supply chain networks with numerous nodes and edges, accurate supply predictions are challenging due to dynamic node interactions, cascading supply delays, resource availability, production and logistic capabilities. Consequently, supply executions often deviate from their initial plans. To address this, we present the Graph-based Supply Prediction (GSP) probabilistic model. Our attention-based graph neural network (GNN) model predicts supplies, inventory, and imbalances using graph-structured historical data, demand forecasting, and original supply plan inputs. The experiments, conducted using historical data from a global consumer goods company's large-scale supply chain, demonstrate that GSP significantly improves supply and inventory prediction accuracy, potentially offering supply plan corrections to optimize executions.

Figures (8)

• Figure 1: Supply Chain Network (Example)
• Figure 2: Graph Node Embedding
• Figure 3: Cumulative Supply Prediction Error
• Figure 4: Edge-level Shipments (Example)
• Figure 5: Node-level Relationship (Example)