Abstract:Accurate short-period traffic flow forecast plays an important role in reasonable traffic induction and controlling, reducing traffic congestion. This paper takes gray model as foundation, making use of the advantage of simple algorithm, using a small amount of data modeling of gray model, and BP neural network's good performance on forecast of nonlinear systems, correcting gray forecast model via BP neural network. We finally build a synthesis model. In view of such shortcomings of BP neural network which is easy to fall into local minimum while calculating, and improve the algorithm combining momentum method and adoptive learning rate method. The results show that synthesis model can exert the advantage and weaken the shortcoming of each single model.

Prediction of Traffic Flow Based on Gray Theory and BP Neural Network