Wavelet-Bayesian hierarchical stochastic model for short-term traffic flow at noncritical junctions

Abstract

In ITS (Intelligent Transportation System) equipped urban transportation systems noncritical junctions are often ignored in short-term traffic condition prediction algorithms as the traffic data collection systems in these junctions are not adequate. The paper proposes a shortterm traffic volume model based on a combination of discrete wavelet transform (DWT) and Bayesian hierarchical methodology (BHM) applicable to non-critical junctions lacking continuous data collection systems. Unlike typical short-term traffic condition forecasting algorithms, large traffic flow datasets including information on current traffic scenarios are not required for the proposed model. In this model, a non-functional representation of the daily ‘trend’ of urban traffic flow observations is achieved using DWT while the fluctuations in the traffic flow in addition to the variations represented by the ‘trend’ are modeled as a stochastic process using BHM. The time-varying variance (within day) of these fluctuations over the ‘trend’ in urban traffic flow observations at a signalized intersection has been estimated in the model. The effectiveness and the accuracy of the model have been compared with a conventional short-term traffic flow forecasting time-series model based on Holt-Winters Exponential Smoothing (HWES) technique. Both the models are applied at two signalized intersections at the city-centre of Dublin and their performances have been discussed.