Study of Traffic Flow Prediction Based on BP Neural Network

Abstract

In this paper the back propagation (BP) neural network algorithm is applied to predict the traffic flow of urban road. The neuron structure needs 48 input nodes and 48 output nodes, so the frame of 48-20-48 is selected. First train an ideal input network with lower error square sum, then take the trained weight vector as initial value of the next input vector. The network training is realized by functions of adaptive learning rate and additional momentum method. The design can forecast 5-minute vehicle flow in future by the current related traffic flow and provide effective information for traffic department. The simulation by Matlab shows that the method with power learning ability and adaptability has high application value.