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.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve The paper aims to address the issue of supply and demand imbalance in supply chain networks. Specifically, while accurate demand forecasting is crucial for supply planning, demand forecasting alone is insufficient to create effective supply plans. Successful supply planning requires maximizing the predictability of demand and supply throughout the execution period to ensure that the supply plan can effectively match demand, avoiding excess or insufficient inventory. However, in complex supply chain networks, accurate supply forecasting becomes highly challenging due to factors such as dynamic interactions between nodes, the cascading effects of supply delays, resource availability, production capacity, and logistics capabilities. As a result, actual supply execution often deviates from the initial plan. To address this issue, the authors propose a Graph-based Supply Prediction (GSP) probabilistic model. This model utilizes a Graph Neural Network (GNN) with an attention mechanism to predict supply, inventory, and imbalances using historical data, demand forecasts, and initial supply plans as inputs. ### Main Contributions 1. **Proposed a new GNN-based general method**: This method is used to predict the quantity and timing of events in the context of graph-structured problems, especially when there is no one-to-one mapping to actual events. The method relies on edge-level and/or node-level aggregated quantities at specific time granularities (e.g., daily or weekly) as labeled data. 2. **Introduced a new prediction error metric sMACE**: This metric is used to evaluate the inaccuracy of predictions caused by uncertainties in supply quantity and timing. It is used in the loss function to train the GNN-based event increment prediction model. 3. **Developed the GSP model**: Specifically designed for supply chain network scenarios, it predicts the quantity and timing of shipment events. Experimental results show that the GSP model significantly improves the accuracy of supply and inventory predictions across the network, aiding in the optimization of supply plan execution. ### Experimental Results Experiments were conducted using historical data from the complex supply chain network of a global consumer goods company. The results show that the GSP model significantly outperforms planned shipments and the Croston method in daily supply prediction (sMACE) and weekly inventory prediction (wMAPE). The specific values are as follows: - **GSP (α=0.0)**: sMACE 102.90 ± 0.1%, wMAPE 31.03 ± 0.3%, constraint error κ 3.99 ± 0.03% - **GSP (α=0.5)**: sMACE 99.94 ± 0.1%, wMAPE 30.43 ± 0.3%, constraint error κ 3.69 ± 0.02% - **GSP (α=1.0)**: sMACE 100.01 ± 0.2%, wMAPE 30.44 ± 0.4%, constraint error κ 3.69 ± 0.04% - **Planned Shipments**: sMACE 279.72%, wMAPE 34.65 ± 0.5%, constraint error κ 3.67 ± 0.03% - **Croston Method**: sMACE 1541.79%, wMAPE 55.06 ± 0.4%, constraint error κ 3.47 ± 0.02% These results indicate that the GSP model has a significant advantage in improving the accuracy of supply and inventory predictions, aiding in the optimization of supply chain execution.

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

Generative Probabilistic Planning for Optimizing Supply Chain Networks

Applying graph neural network to SupplyGraph for supply chain network

BikeNet: Accurate Bike Demand Prediction Using Graph Neural Networks for Station Rebalancing

Learning production functions for supply chains with graph neural networks

Graph Neural Networks in Supply Chain Analytics and Optimization: Concepts, Perspectives, Dataset and Benchmarks

SupplyGraph: A Benchmark Dataset for Supply Chain Planning using Graph Neural Networks

AGNP: Network-wide Short-Term Probabilistic Traffic Speed Prediction and Imputation

Probabilistic Demand Forecasting with Graph Neural Networks

Spatial Graph Neural Networks for Accurate Product Demand Prediction

Intricate Supply Chain Demand Forecasting Based on Graph Convolution Network

HKTGNN: Hierarchical Knowledge Transferable Graph Neural Network-based Supply Chain Risk Assessment

Discovering Supply Chain Links with Augmented Intelligence

Scalable Spatiotemporal Graph Neural Networks

Predicting future production system bottlenecks with a graph neural network approach

A dynamical spatial-temporal graph neural network for traffic demand prediction

Leveraging Graph Neural Networks and Multi-Agent Reinforcement Learning for Inventory Control in Supply Chains

Using graph neural network to conduct supplier recommendation based on large-scale supply chain

Application Of Recurrent Neural Networks Prediction In Supply Chain

G-networks and the optimization of supply chains

Predicting Green Supply Chain Impact With SNN-Stacking Model in Digital Transformation Context