GRAPHINE : Enhancing Spatiotemporal Supply Chain Forecasting Using Virtual Node-Augmented Graph Diffusion for Improved Fuel Efficiency

Authors: Abdulelah S. Alshehri, Azmine Toushik Wasi, Mahfuz Ahmed Anik, MD Shafikul Islam, and Mohamed Kamel Hadj-Kali

Link to the paper on the International Journal of Production Research : https://doi.org/10.1080/00207543.2026.2685179

Abstract

Spatiotemporal forecasting in supply chain networks demands modelling complex spatial dependencies and nonlinear temporal dynamics. Traditional models often fail to capture the heterogeneity, structural sparsity, and long-term dependencies of real-world supply chains. To address this, we propose GRAPHINE, a Virtual Node Diffusion-Convolutional Recurrent Neural Network tailored for supply chain forecasting. GRAPHINE leverages diffusion-based learning, using bidirectional random walks to model spatial relations and a recurrent encoder-decoder with scheduled sampling to capture temporal patterns. It introduces virtual nodes to aggregate global context and applies a learnable gating mechanism that allows each node to regulate global influence based on local features, preventing oversmoothing and preserving specificity. Evaluated on the SCG dataset for demand and inventory forecasting, GRAPHINE achieves reductions of 38.71% in MSE and of 21.71% in RMSE over state-of-the-art baselines, which may imply potential reductions in fuel use on the order of 5.3% in scenario-specific settings under commonly used elasticity assumptions in logistics planning. GRAPHINE thus establishes a new decision-aid framework for applying advanced spatiotemporal graph neural architectures to complex production and supply chain problems, with tangible implications for production planning, operational cost efficiency, and environmental sustainability.

Installation

To set up the project, follow these steps:

1. Clone the repository:

   git clone https://github.com/AI4CHEMIA/GRAPHINE.git
   cd GRAPHINE

   (Note: Replace your-username with the actual GitHub username or organization once the repository is hosted.)

2. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

   Alternatively, you can install the package in editable mode:

   pip install -e .

Usage

Data Preparation

Place your processed supply chain data files into the data/processed/ directory. Ensure the CSV files are correctly formatted as expected by the data_preprocessing.py module.

Training the Model

To train the GNN model, run the train.py script from the root directory of the project. You can specify data paths and training parameters via command-line arguments or modify the config.yaml file.

Example command-line usage:

python train.py \
  --edges_path data/processed/Edges.csv \
  --node_features_path data/processed/NodeFeatures.csv \
  --output_path data/processed/Output.csv \
  --epochs 200 \
  --learning_rate 0.01 \
  --hidden_dim 32 \
  --v_node_weight 0.1 \
  --train_ratio 0.8 \
  --sampling_percentage 0.6 \
  --model_save_path models/trained_model.pth \
  --tag GRAPHINE_Trial

Project Structure

GRAPHINE/
├── data/
│   ├── Edges(Plant)-Uniques.csv
│   ├── NOdeFeatures.csv
│   └── NOdeFeaturesSO.csv
├── src/
│   ├── __init__.py
│   ├── models/
│   │   ├── __init__.py
│   │   ├── diffconv.py
│   │   ├── dgc_rnn.py
│   │   └── temporal_gnn.py
│   └── utils/
│       ├── __init__.py
│       └── data_preprocessing.py
├── requirements.txt
├── setup.py
├── train.py
└── README.md

• data/: Directory to store raw and processed data files.
• src/: Contains the core Python modules of the project.
  • src/models/: Defines the GNN model architectures (DiffConv, DGC_RNN, TemporalGNN).
  • src/utils/: Provides utility functions, such as data preprocessing.
• config.yaml: Configuration file for project parameters.
• requirements.txt: Lists all Python dependencies.
• setup.py: Setup script for packaging the project.
• train.py: Main script for training the GNN model.
• README.md: This documentation file.

Citation

@article{Alshehri09062026,
author = {Abdulelah S. Alshehri and Azmine Toushik Wasi and Mahfuz Ahmed Anik and MD Shafikul Islam and Mohamed Kamel Hadj-Kali},
title = {GRAPHINE: enhancing spatiotemporal supply chain forecasting using virtual node-augmented graph diffusion for improved fuel efficiency},
journal = {International Journal of Production Research},
volume = {0},
number = {0},
pages = {1--26},
year = {2026},
publisher = {Taylor \& Francis},
doi = {10.1080/00207543.2026.2685179},


URL = { 
    
        https://doi.org/10.1080/00207543.2026.2685179
    
    

},
eprint = { 
    
        https://doi.org/10.1080/00207543.2026.2685179
    
    

}
,
    abstract = { Spatiotemporal forecasting in supply chain networks demands modelling complex spatial dependencies and nonlinear temporal dynamics. Traditional models often fail to capture the heterogeneity, structural sparsity, and long-term dependencies of real-world supply chains. To address this, we propose GRAPHINE, a Virtual Node Diffusion-Convolutional Recurrent Neural Network tailored for supply chain forecasting. GRAPHINE leverages diffusion-based learning, using bidirectional random walks to model spatial relations and a recurrent encoder-decoder with scheduled sampling to capture temporal patterns. It introduces virtual nodes to aggregate global context and applies a learnable gating mechanism that allows each node to regulate global influence based on local features, preventing oversmoothing and preserving specificity. Evaluated on the SCG dataset for demand and inventory forecasting, GRAPHINE achieves reductions of 38.71\% in MSE and of 21.71\% in RMSE over state-of-the-art baselines, which may imply potential reductions in fuel use on the order of 5.3\% in scenario-specific settings under commonly used elasticity assumptions in logistics planning. GRAPHINE thus establishes a new decision-aid framework for applying advanced spatiotemporal graph neural architectures to complex production and supply chain problems, with tangible implications for production planning, operational cost efficiency, and environmental sustainability. }
}