Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Development of a graph neural network surrogate for travel demand modelling

Nikita Makarov, Santhanakrishnan Narayanan, Constantinos Antoniou

TL;DR

This paper advances travel demand modelling by introducing a GATv3-based surrogate framework, synthetic data augmentation, and a fine-grained classification approach that balances stability with numerical precision. It demonstrates that GATv3 can outperform other GNNs in classification tasks, while a deeply configured GCNII can excel in fine-grained predictions when supplemented with synthetic data. The work highlights the relative advantages of classification over regression for travel demand surrogates and underscores challenges in extending GAT-based architectures to large, complex transport networks. Practically, the findings offer actionable guidance for applying GATv3 and fine-grained classification to classification-heavy transportation tasks like section control and congestion warning systems, while pointing to data-augmentation and error-propagation issues as key areas for future work.

Abstract

As urban environments grow, the modelling of transportation systems becomes increasingly complex. This paper advances the field of travel demand modelling by introducing advanced Graph Neural Network (GNN) architectures as surrogate models, addressing key limitations of previous approaches. Building on prior work with Graph Convolutional Networks (GCNs), we introduce GATv3, a new Graph Attention Network (GAT) variant that mitigates over-smoothing through residual connections, enabling deeper and more expressive architectures. Additionally, we propose a fine-grained classification framework that improves predictive stability while achieving numerical precision comparable to regression, offering a more interpretable and efficient alternative. To enhance model performance, we develop a synthetic data generation strategy, which expands the augmented training dataset without overfitting. Our experiments demonstrate that GATv3 significantly improves classification performance, while the GCN model shows unexpected dominance in fine-grained classification when supplemented with additional training data. The results highlight the advantages of fine-grained classification over regression for travel demand modelling tasks and reveal new challenges in extending GAT-based architectures to complex transport scenarios. Notably, GATv3 appears well-suited for classification-based transportation applications, such as section control and congestion warning systems, which require a higher degree of differentiation among neighboring links. These findings contribute to refining GNN-based surrogates, offering new possibilities for applying GATv3 and fine-grained classification in broader transportation challenges.

Development of a graph neural network surrogate for travel demand modelling

TL;DR

This paper advances travel demand modelling by introducing a GATv3-based surrogate framework, synthetic data augmentation, and a fine-grained classification approach that balances stability with numerical precision. It demonstrates that GATv3 can outperform other GNNs in classification tasks, while a deeply configured GCNII can excel in fine-grained predictions when supplemented with synthetic data. The work highlights the relative advantages of classification over regression for travel demand surrogates and underscores challenges in extending GAT-based architectures to large, complex transport networks. Practically, the findings offer actionable guidance for applying GATv3 and fine-grained classification to classification-heavy transportation tasks like section control and congestion warning systems, while pointing to data-augmentation and error-propagation issues as key areas for future work.

Abstract

As urban environments grow, the modelling of transportation systems becomes increasingly complex. This paper advances the field of travel demand modelling by introducing advanced Graph Neural Network (GNN) architectures as surrogate models, addressing key limitations of previous approaches. Building on prior work with Graph Convolutional Networks (GCNs), we introduce GATv3, a new Graph Attention Network (GAT) variant that mitigates over-smoothing through residual connections, enabling deeper and more expressive architectures. Additionally, we propose a fine-grained classification framework that improves predictive stability while achieving numerical precision comparable to regression, offering a more interpretable and efficient alternative. To enhance model performance, we develop a synthetic data generation strategy, which expands the augmented training dataset without overfitting. Our experiments demonstrate that GATv3 significantly improves classification performance, while the GCN model shows unexpected dominance in fine-grained classification when supplemented with additional training data. The results highlight the advantages of fine-grained classification over regression for travel demand modelling tasks and reveal new challenges in extending GAT-based architectures to complex transport scenarios. Notably, GATv3 appears well-suited for classification-based transportation applications, such as section control and congestion warning systems, which require a higher degree of differentiation among neighboring links. These findings contribute to refining GNN-based surrogates, offering new possibilities for applying GATv3 and fine-grained classification in broader transportation challenges.
Paper Structure (18 sections, 2 equations, 4 figures, 2 tables)

This paper contains 18 sections, 2 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Literature review
  3. GNNs for transport planning
  4. Surrogate modelling in the field of transportation
  5. Summary and research gaps
  6. Methodological framework
  7. Graph attention networks
  8. Synthetic data generation
  9. Experimental setup
  10. classification task (comparison between GCNII, GATv2 and GATv3)
  11. Fine grained classification
  12. Overview of the setup
  13. Optimal Bucket Strategy
  14. Use of additional synthetic data
  15. Experimental results
  16. ...and 3 more sections

Figures (4)

  • Figure 1: Methodological framework
  • Figure 2: Predicted output vs true output for the GCNII - Buckets-e90 - extra data model on the test dataset. Each point is a prediction for a single link, with both values being in veh/h. An optimal prediction would be along the light blue line. Notice that for true values above 2500, the model constantly predicts around 2600.
  • Figure 3: Predicted output vs true output corresponding to the test dataset for the GCNII - Buckets-nl54 - extra data model. Each point is a prediction for a single link, with both values being in veh/h. An optimal predicted would be along the light blue line. Observe that for true values over 2500, the model still predicts proportionally higher output values.
  • Figure 4: $MAE_{\, \ge10}$ with respect to graph size, i.e. number of nodes for the trained GNN model on the test dataset. Each point is the average absolute error of a single graph, with the errors coming from incorrect link transport usage prediction. Highly similar plots are observed for all GNNs.