摘要：AbstractTimely and accurate forecasting of short-time traffic volume is essential for Intelligent Transportation Systems (ITS). With previous traffic volume forecasting model, traffic agencies could make relevant decisions on traffic managements and controls. In the past decades, many advanced models have been proposed and achieved a significant improvement in forecasting. However, multiple-step ahead traffic forecasting has not attracted its deserved attention. In this paper, we proposed a Gaussian Mixed Model embedded Back Propagation network (GMM-BP) to implement multiple-step ahead traffic forecasting. To test the effectiveness of the proposed method, we have designed a numerical experiment using PeMS dataset and compared its performance to single Gaussian Mixed Model (GMM) and Back Propagation (BP) network with the Mean Relative Deviation (MRD) criterion and “acceptance region”. The results show that our proposed method in this paper outperforms single GMM and BP network, and especially in multiple-step ahead traffic volume forecasting.
