Short-time Passenger Flow Forecast of Urban Rail Transit Based on the CEEMDAN-BLSTM Model

Abstract

Subway short-term passenger flow forecast can help the subway operation department to optimize the train driving interval, improve the operation level and save the station air-conditioning electric energy. According to the different characteristics of passenger flow change on workdays and holidays, in order to avoid the influence of different characteristics between data, the prediction model of workdays and holidays passenger flow data are built respectively, and the Pearson correlation coefficient was used to analyze the correlation degree between the historical passenger flow data and the predicted value. Due to the large fluctuation of passenger flow data, there will be large error in directly predicting the original data. Adopt complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) algorithm to decompose passenger flow data, then based on the timing change of the passenger flow, bi-directional long short-term memory network (BLSTM) improved by LSTM is employed to predict each decomposed series. Finally, all the predicted outputs are merged as the overall output. To verify the superiority of the model, the LSTM, BLSTM, EMD-BLSTM, STL-BLSTM model is compared with the CEEMDAN-BLSTM model proposed in this paper, and the results show that the proposed model has a significant accuracy improvement compared to the traditional prediction model.