Machine Learning Model Comparison for Traffic Accident Severity in Kent, UK (2022-2024)

A study proposing an explainable, spatially segmented machine learning framework to examine urban-rural heterogeneity in crash outcomes. It uses disaggregated accident data from Kent, UK, covering the period 2022 to 2024, and compares five models. The dataset was authored by Huashan Ye and last updated on 2026-04-24.

Use Cases

• Comparing machine learning model performance for predicting accident severity based on the described five-model test.
• Analyzing spatial disparities in risk factors between urban and rural areas based on the described segmentation framework.
• Applying SHAP explanations to understand feature importance differences across spatial contexts as described in the study.
• Informing context-sensitive transport safety interventions based on the identified behavioral and infrastructure-driven risk profiles.

Strengths

• Data covers a defined geographic area (Kent, UK) and a recent time range (2022–2024).
• The study framework explicitly treats urban and rural systems as separate analytical units.
• Model performance is benchmarked across five tested algorithms, with Random Forest identified as best.

Limitations

• The dataset is very small at 5.5 KB, indicating limited scope or a summary of results rather than raw data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale modeling.

Provenance

Source

figshare, author Huashan Ye.

Collection Method

Likely contains results from a machine learning study comparing models on accident data.

Time Range

2022–2024

Freshness

Last updated 2026-04-24 17:52:25; freshness should be verified.

Geography

Kent, United Kingdom