Improved RBF Neural Network for Short-Time Passenger Flow Forecast of Rail Transit

In recent years, urban rail transit has attracted attention of travelers for its advantages of convenience, speed, and environmental protection, development of urban rail has become trend of urban development in China. So urban rail transit passenger flow forecasting is indispensable in rail transit construction and operation. Accurate passenger flow forecasting is of great significance in meeting passenger travel needs, improving the service level of operating departments, and ensuring the economic benefits of operating enterprises. However, the prediction accuracy of the traditional prediction model is not high in the field of short-term prediction, so an improved RBF neural network model is proposed to solve problem of urban rail transit passenger flow prediction. RBF neural network has advantages of simple structure and fast learning speed, and it takes work to fall into local minimization. However, in practical application, the selection of initial values of some parameters by the RBF neural network will affect the output results. The RBF neural network is improved by adopting the beetle antennae search(BAS) algorithm. To avoid the BAS algorithm from falling into the local optimum, the variable step size factor and Metropolis criterion of the simulated annealing algorithm are introduced into the BAS algorithm. The min-max normalization algorithm normalizes the data and predicts the passenger flow. Taking actual passenger flow of Dongmen Bridge Station of Chengdu Metro as an example, the results show that prediction error of improved model is lower.