A Bayesian dynamic approach to modelling flow through a traffic network.

Wright, Benjamin John (2006). A Bayesian dynamic approach to modelling flow through a traffic network. PhD thesis The Open University.

DOI: https://doi.org/10.21954/ou.ro.0000fe35

Abstract

Hourly traffic flows through complicated motorway junctions form time series with a hierarchical element. Multiregression Dynamic Models provide a Bayesian framework for forecasting the time series while incorporating this hierarchical structure. Calculation of forecasts is computationally inexpensive and the model is designed in order to have readily interpretable parameters wherever possible. Expert intervention is straightforward in this system for periods of unusual activity, whether they were anticipated or not. Fundamental change to the road layout and behaviour can also be incorporated into the model without untenable complication. In this thesis work is centred on applying the methodology to one particular junction, including finding techniques to aid in using the methodology in this application. The theory of Multiregression Dynamic Models is furthered in order to do this- specifically through modelling certain parameter constraints imposed by the nature of the problem and the production of prior covariances for quantities modelled. The model was found to perform competitively in comparison to common other approaches, and the hierarchical structure offered significant advantages when expert intervention was applied.