Short-Time Passenger Flow Prediction for Urban Rail Transit Considering Data Noise Reduction

In order to reduce the influence of data fluctuations on the prediction accuracy of short-term passenger flow, a wavelet transform-support vector machine (WT-SVR) combined model is proposed. According to the correlation of passenger flow data, the appropriate wavelet basis function is selected to reduce the impact of external environment changes on the original passenger flow data. We construct a denoising passenger flow signal, and input support vector machine for machine learning. Using the passenger flow data of two categories in Xi’an, the results show that compared with the single SVR model, the determinable coefficients of passenger flow prediction are increased by 2.72% and 5.69%, respectively, and the computing time is reduced by 24.04% and 30.59%, respectively. This combined model can accurately and quickly predict short-term passenger flow, providing a reference for the operation and management of urban rail stations.