A Trajectory Probability Network for City-Scale Road Volume Prediction

ICLR 2025 Conference Submission10916 Authors

Keywords: Data Mining, Traffic Volume Prediction, Learning on Graph

TL;DR: A deep-learning model for city-scale road volume estimation with ultral incomplete data, fast and accurate.

Abstract: City-scale road volume prediction is a fundamental task in traffic management. However, the observation data are often incomplete and biased, posting a challenge for accurate prediction. Existing methods address this issue through interpolation techniques or manual priors, but they typically provide only a deterministic restoration, overlooking the influence of other potential scenarios. To overcome these limitations, we propose a novel neural network-based probabilistic model, the Trajectory Probability Network (TraPNet), which predicts traffic volume through the aggregation of the joint distribution of potential trajectories. TraPNet makes full use of current observations, historical data, and road network information to offer a comprehensive inference of road volumes. Unlike autoregressive methods, TraPNet makes predictions in a single step, substantially reducing computational time while maintaining high predictive accuracy. Experiments on real-world road networks demonstrate that TraPNet outperforms state-of-the-art methods, and can keep the advantage with only 20\% observation ratio. The code will be made publicly available.

Primary Area: learning on graphs and other geometries & topologies

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 10916