Short-term Passenger Flow Prediction and Operation Optimization for Urban Rail Transit Based on Multi-model Comparison

College of Big Data and Artificial Intelligence, Guangdong University of Finance, Qingyuan, 511510, China

^* Corresponding author’s email: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

With the continuous growth of public transportation demand in large cities, metro operations are facing increasingly severe passenger flow pressure. Accurate short-term passenger flow forecasting is very important. It helps transportation systems run smoothly. Passenger experience improves because of it. Operational management also gets better. Based on Shenzhen’s subway passenger flow and meteorological data, this paper compares multiple models. The goal is to forecast short-term subway passenger flow. Weather factors affect passenger flow. The paper analyzed this effect separately, it is found that rainfall and strong winds significantly reduce passenger numbers, while high temperatures lead to an increase. The study compares the predictive performance of three models—CNN-LSTM, XGBoost, and STGCN — revealing that STGCN outperforms in MAE, RMSE, MAPE, and R² metrics, effectively capturing spatiotemporal dependencies; XGBoost offers high training efficiency, making it suitable for real-time scenarios; and CNN-LSTM demonstrates strong trend fitting. This research helps with model selection. Different cities can use these criteria for passenger flow forecasting. The findings also support operations. They provide data for optimizing subway management.