摘要

Aimed at the inadequacy of the existing prediction algorithm of traffic flow, an improved BP (Back Propagation) wavelet neural network algorithm is presented. The learning rate of the weight, and the learning rate of dilation and translation factor in the algorithm are both selected by an array which includes the most used learning rate, so it avoids the blindness in determining the learning rate. Simulations results show that the algorithm has faster training time, and is superior to the moving average algorithm for the prediction of traffic flow.

全文