Abstract: Reliable short-term traffic flow forecasting is an important foundation for the intelligent transportation system.In order to improve the accuracy of the short-term traffic flow forecasting and increase its adaptabilityin different traffic states,a combination forecasting model based on the analysis of traffic flow characteristics and space-time two-dimensional impact factors is presented to reflect the characteristics and influencing factors.The model has three sub-models:time-series model,space-related model and combinationforecasting model.The single forecast models includes single adaptive exponential smoothing model and RBF neural network model.The combination coefficientis obtained adaptively based on the smoothing percentage relative error of thetwo single forecast sub-modules as input by using the neural network as a learning algorithm.Finally,the traffic flow data are measured respectively in flat peak and peak hours to verify the validity and reliability of the model.The results show that the combination model can produce more precise forecasting than that of two individual models and adapt to different traffic states better.

Key words: intelligent transport system, traffic flow prediction, exponent smoothness method, RBF neural network