Conformal Load Forecasting on Graphs

This repository contains the official source code for the paper:
"Conformal Load Prediction with Transductive Graph Autoencoders"
Published in Machine Learning, Volume 114, Article 54, 2025.
DOI: 10.1007/s10994-024-06713-w.

Overview

This repository implements the methods described in the paper for conformal load prediction on graph-structured data using transductive graph autoencoders. The approach leverages conformal prediction techniques to provide uncertainty estimates for edge weight predictions in graphs, with applications in transportation systems and other domains.

How to Run

To reproduce the results from Table 1 in the paper, run the following command:

sh Table-1.sh

To reproduce the results from Table 2 in the paper, run the following command:

sh Table-2.sh

Acknowledgement

We acknowledge several GitHub resources that we used in our research.

• For providing the datasets: https://github.com/bstabler/TransportationNetworks/tree/master;
• For providing the data preprocessing code: https://github.com/000Justin000/ssl_edge;
• For providing the conformal quantile regression code: https://github.com/snap-stanford/conformalized-gnn.

We sincerely appreciate their efforts.

Citation

If you find this repository or our work helpful, please consider citing our paper:

@article{luo2025conformal,
  title={Conformal load prediction with transductive graph autoencoders},
  author={Luo, Rui and Colombo, Nicolo},
  journal={Machine Learning},
  volume={114},
  number={3},
  pages={1--22},
  year={2025},
  publisher={Springer}
}