PeMSD4 & PeMSD8: Traffic Flow Data for Spatiotemporal Prediction

This dataset supports the reproducibility of a research paper on integrated spatiotemporal feature alignment for traffic flow prediction. It contains two processed sub-datasets, PeMSD4 and PeMSD8, each including traffic data and a distance matrix representing sensor relationships. The original data is sourced from the California Department of Transportation's Performance Measurement System (PeMS).

Use Cases

• Traffic flow prediction based on integrated spatiotemporal feature alignment.
• Benchmarking graph neural network models based on the provided sensor graph (distance matrix).
• Reproducing results for the paper 'Integrated Spatiotemporal Feature Alignment with Graph Convolutional and Gated Recurrent Networks for Traffic Flow Prediction'.
• Analyzing sensor network relationships for traffic management based on the distance.csv files.

Strengths

• Explicitly supports reproducibility for a specific, cited research paper.
• Includes two established benchmark datasets (PeMSD4 and PeMSD8) for traffic prediction.
• Provides pre-processed data in .npz format alongside a distance matrix, facilitating graph-based modeling.
• Released under a permissive CC-BY-4.0 license.

Limitations

• Column names and row counts are not specified in the provided metadata, limiting immediate understanding of data structure.
• The dataset is a processed derivative; users must refer to the original PeMS source for raw data and full context.
• The 'last updated' date of 2026-03-22 appears to be a future date, indicating a potential metadata error.

Provenance

Source

California Department of Transportation (PeMS) Performance Measurement System (https://pems.dot.ca.gov/).

Collection Method

Processed and compiled by the author to support the reproducibility of a specific research paper.

Freshness

Last updated date is listed as 2026-03-22, which is a future date and may be inaccurate.

Geography

California, USA (based on the source, Caltrans PeMS).