Salvador Urban Network Transportation (SUNT): A Landmark Spatiotemporal Dataset for Public Transportation

Marcos Vinícius dos Santos Ferreira, Matheus Carvalho de Souza, Tatiane Nogueira Rios, Islame Felipe da Costa Fernandes, Danilo Oliveira Andrade, Joao Gama, Albert Bifet, RICARDO RIOS

Keywords: Graph Neural Network, Time Series, Spatial Dataset, Public Transportation, Urban Mobility

TL;DR: This paper introduces a novel dataset comprising temporal and geospatial details of public transportation infrastructure and passenger trips in Salvador, Brazil. Machine learning models trained on this dataset have also been published as benchmarks.

Abstract: Efficient public transportation management is essential for the development of large urban centers, providing several benefits such as comprehensive coverage of population mobility, improvement of the local economy with the offer of new jobs and the decrease of transport costs, better control of traffic congestion, and significant reduction of environmental impact limiting gas emissions and pollution. Realizing these benefits requires carefully pursuing two essential pathways: (i) deeply understanding the population and transit patterns and (ii) using intelligent approaches to model multiple relations and characteristics efficiently. This work addresses these challenges by providing a novel dataset that includes various public transportation components alongside machine learning models trained to understand and predict different real-world behaviors. Our dataset comprises daily information from about 710,000 passengers in Salvador, one of Brazil's largest cities, and local public transportation data with approximately 2,000 vehicles operating across nearly 400 lines, connecting almost 3,000 stops and stations. As benchmarks, we have fine-tuned diverse Graph Neural Networks to perform inference on vertices and edges, undertaking both regression and classification tasks. These models leverage temporal and spatial features concerning passengers and transportation data. We emphasize the greatest advantage of using our dataset lies in different possibilities of modeling a real-world urban mobility dataset, reproducing our results, overcoming our models, and investigating several other open-problem situations listed in this manuscript as future work, which include the designing of new methods, optimization strategies, and environmental approaches. Our dataset, codes, and models are available at https://github.com/suntdataset/sunt.git.

Primary Area: datasets and benchmarks

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.

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: 9985