Nicola Hewett
Improving the prediction of hotspots using Bayesian models.
Email: [email protected]
Supervisors
- School of Engineering
- School of Mathematics, Statistics and Physics
- PTV Group
- Dr Karsten Kremer
Project description
We are developing sophisticated Bayesian hierarchical models for real-time prediction of road safety hotspots.
Current practice for treating such hotspots is almost always reactive. So a threshold level of collisions is overtopped during a pre-determined observation period. We then apply a solution (such as road safety cameras). But methodology can now predict collision counts at potential hotspots in future time periods. This provides a more proactive treatment of road safety hotspots.
It is preferable to identify road safety hotspots based on predicted rather than observed counts. It avoids the need to wait until collisions (and casualties/fatalities) occur before applying road safety schemes. Of course, we must have faith in the underpinning methodology and the predictions produced by the statistical models used. Thus, we will start by carrying out a thorough literature review and trial of existing methods for site-based prediction modelling applied to accidents observed at discrete time periods. We will then consider improved variations on these models. Finally, we will extend these models to support real-time modelling and prediction over a road network.
Our work builds on recent advances made through Newcastle's research partnership with transport planning and logistics software company PTV Group.
Road safety practitioners across the world use PTV Visum. It is the world’s leading software for traffic analyses, forecasts and data management. PTV will make significant improvements to existing methodology available through Visum.