Abstract:

     Passenger flow forecast is the basic reference for the design and operational management of urban rail transit and has become an important procedure in the construction of urban rail transit. We combine a linearity based ARIMA model and a nonlinearity based RBF neural network and establish an ARIMARBF prediction model by the analysis of such time sequence characteristics as the weekly change and nonstability of passenger flow and the mechanism of ARIMA and RBF models. We then apply the prediction model to the forecast for passenger flow of Beijing daily urban rail transit and receive a better prediction result, fully considering the linear and nonlinear characteristics of urban rail transit passenger flow.

Key words: urban rail transit, passenger flow forecast, combination forecast, RBF neural network, ARIMA model