Urban Road Congestion Probability Prediction Based On Bayesian Network

With the rapid development of urban economy and the constant improvement of city modernization, contradiction of urban road supply and demand is more serious than ever. Urban road congestion causes the reducing of traffic capacity which seriouly influences people’s daily travelling. The effective prediction of traffic congestion has become one of the important subjects of social theory and practice research.A reasonable traffic congestion probability prediction is able to help traffic managers to make efficient decisions to reduce the negative effect of traffic congestion. The existing researchs, like time series method, grey prediction method, neural network, etc have made well application achievements in some congestion prediction fields. However, they rarely consider the dependency and uncertainty of traffic congestion. Bayesian Network (BN) is applicable to analyzing uncertainty and probabilistic things, and able to deal with the dependencies between variables. It is base on directed graph to express the probabilistic relationship between variables. In consideration of the characteristics of traffic congestion prediction and Bayesian Network from different perspectives, this research proposes a Bayesian Network to predict the probability of urban traffic congestion. The input variables of BN are determined based on public transportation, road lengths, GDP (Gross Domestic Product), vehicle numbers, etc to comprehensive reflect the characteristics of traffic congestion and to ensure the reasonable of results.This research focuses on built-up areas of megalopolis. Based on different urban transport development policies, the road congestion probabilities of the research area are analyzed. So the most appropriate policy will be decided to ease traffic congestion in built-up areas. The results show that the established BN is fully reflected the reasonable prediction of road network traffic congestion probability under the influence of different transport policies in research area. Furthermore, if some measures of curb traffic congestion are not adopted, the road network traffic congestion probability of research area will soon be over50.00%. This will cause troubles in the normal operation of traffic system. In addition, because of the induced traffic volume, only increasing the supply of newly built road on the degree of ease the traffic congestion is very limited. Simply encourage rail transportation is also difficult to reverse the reality which traffic congestion probability is high. In comparison, both road construction and bus system development at the same time can better mitigate traffic congestion than other policies in a short period of time.