Probabilistic neural networks for the identification of traffic state

The paper proposes a method of real-time traffic state estimation for highways based on probabilistic neural network (PNN). In China, the traffic composition of highways is different from urban road, motorway. The large vehicles have a higher proportion. Poor operating performance and overloading of large vehicles are serious and constitute a serious threat to road traffic safety. So the proportion of large vehicles is chosen as one of classification indicators. The number of classification indicators is determined by calculating the correlation coefficient between each other. Reduction of parameters reduces the complexity of model. The new method is verified by the empirical data comes from G101 national highways, Shunyi District of Beijing. To evaluate the performance of the method, fuzzy C-mean clustering (FCM) algorithm is also applied to the classification problem. The results prove that the new method improves the stability and accuracy of identification.