PEMS08: Computation Time Metrics for a Traffic Forecasting Model

5.5 KB of data in XLS format, uploaded by Pei Shi on figshare in June 2026. The dataset contains computation time metrics for the PEMS08 traffic dataset, used in experiments evaluating the AMGST traffic forecasting model. The model integrates adaptive multi-graph convolution and spatiotemporal attention to predict traffic speed and flow.

Use Cases

• Benchmarking model training efficiency based on reported computation time.
• Comparing the computational cost of different graph convolution modules for traffic forecasting.
• Analyzing the relationship between model architecture complexity and runtime performance.

Strengths

• Dataset is small (5.5 KB), ensuring quick download and easy inspection.
• Shared under a permissive CC-BY-4.0 license, allowing for broad reuse.

Limitations

• The specific columns and row count are unknown, limiting assessment of data structure.
• The description focuses on the model methodology, with minimal details on the computation time data itself.
• Last updated 2026-06-04 17:36:39; freshness should be verified.

Provenance

Source

figshare

Collection Method

Likely generated as experimental results from evaluating the AMGST model.

Freshness

Last updated 2026-06-04 17:36:39.