Passenger Flow Prediction of Urban Rail Transit based on Dual- Channel and Multi-Factor

Abstract

At present, in the process of subway passenger flow forecasting, there are often problems such as single forecasting model and incomplete construction features, which lead to the phenomenon of low forecasting accuracy. To solve these problems, this paper proposes a subway passenger flow prediction model based on dual-channel and multi-factor (DCMF-CNN-LSTM). Based on the comprehensive consideration of the influence of time characteristics, space characteristics and external characteristics on subway passenger flow, the model adopts dual-channel experimental route to predict subway passenger flow. First of all, based on the analysis of the passenger flow between adjacent stations, the passenger flow of the same station in the same time period, peak hours, weather and other factors on the passenger flow, the spatial characteristics, time characteristics and external characteristics of subway passenger flow models are established. Secondly, a dual-channel experimental route was established. Channel 1 was composed of CNN+LSTM+Attention model to extract the spatial and external features affecting passenger flow and complete the passenger flow prediction influenced by spatial and external factors; channel 2 was composed of LSTM model to extract the time features affecting passenger flow and complete the passenger flow prediction influenced by time factors. Finally, the random forest algorithm is used to fusion the prediction results. The experimental results show that the proposed prediction method based on dual-channel  and multi-factorhas the lowest prediction error and higher practicability compared with other prediction methods.