Context & Summary
A Road Traffic Accident (RTA) is an unexpected event occurring on the road involving vehicles and other road users. With increasing vehicle numbers year on year, reducing RTAs is a significant challenge.
This project involves developing and training statistical models (XGBoost, Random Forest) to identify causes of RTAs based on location, surface, and climate. We leverage open-source UK government data to deploy these models on a web-based dashboard for real-time prediction and exploration.
Key Stakeholders
Deploying Computer Vision solutions to accident-prone locations. Statistical modelling identifies where solutions are most useful.
Road Users
Access to user-friendly modelling dashboards helps plan journeys and view potential risk areas, promoting safer usage.
Dept of Transport
Insight into feature importance and systematic causes of RTAs assists with intervention and infrastructure planning.
Technologies Used
Frontend & Infrastructure
- • Next.js (App Router)
- • Tailwind CSS & Lucide Icons
- • AWS CDK & S3
- • Grafana Monitoring
ML & Data
- • Python 3.10+
- • XGBoost & Random Forest
- • Pandas & NumPy
- • Jupyter Notebooks
Future Goals
We aim to implement grid-based mapping (splitting the map by lat/long) for granular severity predictions (scored 0-1). This will improve upon basic heatmaps and allow for specific "danger zone" identification.