Traffic Flow Prediction Based on GM-RBF

In order to provide more reliable data for traffic control and guidance system, traffic flow prediction is very important. This paper proposes a traffic flow prediction method based on GM-RBF combined model. Firstly, build a GM(1, 1) prediction model to predict multiple sequences. Secondly, the GM model and RBF neural network are combined in application, and the residual feedback is established by the RBF model, so as to solve the problem of low accuracy of the grey model in the prediction. Taking the data every 15 min of the A12 highway in Suffolk, UK from January 1, 2019 to January 20, 2019 as a sample, taking into account the influence of weather factors on road traffic flow, the traffic flow per 15 min in a day on January 21, 2019 is predicted. The experimental results show that MAPE of GM-RBF model is 2.302911%, MAE is 3.625, RMSE is 5.6807. Compared with GM model and RBF prediction model, the error evaluation index of GM-RBF combination model has been significantly improved in accuracy. Therefore, the combined model has good applicability in traffic flow prediction.