AMGST: Traffic Forecasting Benchmark Datasets for Speed and Flow

AMGST is a collection of four public traffic datasets used for benchmarking a novel traffic forecasting model. The datasets contain speed and flow measurements, as described in the research paper by Pei Shi. The data was last updated on 2026-06-04.

Use Cases

• Benchmarking traffic forecasting models based on speed and flow measurements.
• Training graph-based neural networks to capture spatial correlations in traffic patterns.
• Evaluating the performance of attention mechanisms for modeling long-term temporal dependencies in traffic data.

Strengths

• Dataset is small at 5.5 KB, facilitating quick download and initial exploration.
• Data is sourced from four public traffic datasets, providing a basis for comparative benchmarking.
• The associated research paper demonstrates the dataset's use in validating a novel forecasting model.

Limitations

• Row count is unknown, which may limit suitability assessment for large-scale training.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Pei Shi via figshare.

Collection Method

Aggregated from four public traffic datasets, as described in the associated research paper.

Freshness

Last updated 2026-06-04 17:36:33; freshness should be verified.