Using NARX Neural Network for Prediction of Urban Rail Transit Passenger Flow

Abstract

Prediction of urban rail transit passenger flow is very important. If there is a large traffic volume in the future, we can reduce the departure interval time in advance to avoid traffic congestion. The passenger flow of rail transit is complex and nonlinear. In order to make more precise real-time passenger flow prediction, the paper takes the passenger flow of Chongqing rail transit as the research object, introduces the NARX neural network with external input and establishes the prediction model. NARX neural network not only has excellent simulation performance but also has outstanding performance in prediction, so it can provide more accurate judgment on the future trends. Compared with other neural networks, its convergence is faster and its normality is also better. Therefore, it is an outstanding prediction tool and suitable for the prediction of passenger flow data in the complex network of rail transit. In addition, compared with the prediction effect of existing methods all over the world, such as Support Vector Machine, we can know that nonlinear recurrent neural network algorithm provides more precise prediction on the passenger flow of rail transits.