Mukara: Highway Traffic Volume Predictions for England and Wales

Model variants from an ablation study for a deep learning framework predicting daily traffic volumes on highway trunk roads. The Mukara framework was trained on eight years of data from England and Wales, incorporating socioeconomic and network features. The dataset is authored by Yue Li and was last updated in April 2026.

Use Cases

• Ablation testing of deep learning architectures based on the described Mukara framework
• Evaluating feature importance for traffic prediction based on population, employment, land use, and road network inputs
• Benchmarking spatial transferability of traffic models using region-based cross-validation results
• Comparing data-driven models against traditional four-step travel demand models for long-term forecasting

Strengths

• Model performance metrics are provided, including a mean GEH of 50.74, MAE of 8,989 vehicles/day, and R² of 0.583
• Training data spans eight years, providing a substantial temporal basis for model development
• Performance was validated under a stringent region-based spatial cross-validation scheme, demonstrating spatial transferability

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• The dataset is very small (5.5 KB), indicating it likely contains summary or model variant data rather than raw observations

Provenance

Source

figshare

Collection Method

Likely derived from an ablation study of the proposed Mukara deep learning framework.

Time Range

Training data covers eight years; specific years are not stated.

Freshness

Last updated 2026-04-16 17:41:59; freshness should be verified

Geography

England and Wales