Prediction of Urban Road Congestion Using a Bayesian Network Approach

A reasonable prediction of the possibility of traffic congestion is able to help traffic managers to make efficient decisions to reduce the negative effect of traffic congestion. Previous research has made valuable analyses in this field. However, they rarely consider the dependency and uncertainty of traffic jams. In consideration of the characteristics of traffic congestion from different perspectives, this research proposes a Bayesian Network (BN) analysis approach to predict the possibility of urban traffic congestion. The built-up area of Beijing is taken as the study area of this research. A comprehensive set of variables are utilized to reflect the characteristics of traffic congestion from various viewpoints. The BN method is used to analyze the uncertainty and probability of traffic congestion, and is proved to be fully capable of representing the stochastic nature of traffic congestion. Furthermore, the difference of the congestion probabilities because of applying different urban transport development policies is analyzed in comparison. The study results show that the both road construction and bus system development at the same time can obviously mitigate traffic congestion for the built-up area of Beijing. In the future research, the impact of the comprehensive development of various travel modes on urban traffic congestion needs further studies.