Research On The Urban Road Traffic Flow Prediction

With the improvement of the intelligent traffic technique, vehicle guidance system which is an important part of the intelligent traffic system has become an effective way for traffic management department to guide the urban traffic. The key technique of the vehicle guidance system is the prediction of the urban traffic state, i.e. use historical traffic data and real-time traffic data to predict the traffic flow of the urban cross and road section, in order to provide reliable proof for providing the most appropriate route for travelers, balancing the traffic flow, optimizing the traffic management measure, improving traffic control and so on. Hence, the prediction of urban road traffic flow is very important.In this paper, the meaning of the prediction of the urban road flow prediction is presented and many kinds of prediction methods both home and abroad are analyzed. Then, the spatio-temporal characteristics of urban traffic flow is analyzed based on the actual road section flow between Huan Cheng Bei Road and Mo Gan Shan Road in the Hangzhou city. Aiming at the nonlinearity, time-varying and uncertainty of traffic flow, artificial neural network which is a mathematical model simulating the biological structure of human brain shows the advantage of distributed processing, self-organizing, self-adaptive and self-learning. This paper used artificial neural network to carry out the research of short-term traffic flow prediction of the Hangzhou city.Then, the historical correlation of traffic flow and the influence of upstream road section are considered together. Single point prediction model and multiple points prediction model are built in this paper. Three kinds of neural network for traffic flow prediction model, i.e. BP, RBF and GRNN are built with m language, based on Matab2007R, to predict the actual road section flow between Huan Cheng Bei Road and Mo Gan Shan Road in the Hangzhou city. With many experiments, the prediction accuracy of the three kinds of neural network is compared. BP neural network takes the longest simulation time and leads to the second best result. GRNN uses least time but also leads to the worst result. RBF takes the second shortest time and leads to the best prediction accuracy. Considering the simulation time and prediction accuracy, RBF neural network is the best model for short-term traffic flow prediction in Hangzhou city. Finally, some existing questions are raised in short-term prediction of urban road section for future research.