Passenger Flow Prediction in Urban Rail Transit Based on Transformer

Accurately predicting passenger flow in urban rail transit is of critical importance for ensuring operational safety, enhancing efficiency, and optimizing costs. To enhance the accuracy of metro passenger flow prediction, this study proposes a passenger flow prediction model based on the Transformer deep learning framework. It is conducted using Automatic Fare Collection (AFC) data from Shanghai Metro Line 5. In addition, clustering algorithms are employed to perform cluster analysis on the stations. Finally, the accuracy and practicality of the Transformer-based model for metro passenger flow prediction are validated through comparative experiments. This model is capable of predicting future passenger flow in rail transit with minute-level precision, thereby assisting subway operators in enhancing train scheduling. It helps in the prevention of resource wastage and facilitates the rational planning of departure frequencies and shifts to accommodate variations in passenger flow during peak and off-peak periods. This ultimately reduces passenger waiting times and improves the overall operational efficiency.