Prediction of Short-term Passenger Flow of Urban Rail Transit based on Data Decomposition

Abstract

With the continuous advancement of urbanization, traffic congestion has become more and more serious, and rail transit has the characteristics of fast and safe, and is more and more favored by residents. However, urban rail transit has serious passenger flow congestion problems in the morning and evening peaks or holidays, so it is particularly important for the subway department to formulate an operation plan in advance. Short-term passenger flow prediction can not only provide data support for the operation and scheduling of subway departments, but also provide better services for passenger travel, so short-term passenger flow prediction has higher accuracy requirements. In this paper, the original passenger flow data is decomposed by VMD to reduce the fluctuation of the original passenger flow, and then combined with the neural network with better nonlinear signal fitting results to predict. Considering that the fluctuation of some components after decomposition is small, MLR is selected to predict the low frequency. components, so as to build a prediction model based on VMD-MLR-BiGRU, which can improve the prediction accuracy and shorten the prediction time. In order to verify the superiority of the proposed model, single models such as MLR, GRU, Bi-GRU, etc., and combined models such as EMD-Bi-GRU and VMD-Bi-GRU are used for comparison.