A passage time prediction method of gates for passengers in subway stations

Predicting the passenger passage time of gates can help management personnel adjust subway station facilities in a timely manner to better adapt to the changing passenger demands, improve the overall performance of the subway network, and help better coordinate evacuation and rescue work in emergency situations. An improved time prediction model for the passenger traveling through gates is established based on the random forest optimized by the energy valley optimizer, providing a theoretical basis for scheduling, personnel management, and resource allocation. The passage time is taken as the output variable of the prediction model, and the factors related to the subway station gates and passengers that affect the output results are taken as input variables. The importance of different input variables on the impact of gate passage time is ranked based on the SHAP method, and the number of passengers is the most important influence factor. A gate simulation system based on the MassMotion software is built, and simulation experiments in different scenarios are conducted by changing the number and width of gates, as well as the number and attributes of passengers, to obtain traffic data under rich scene parameters, providing training and validation sets for the time prediction model. Compared with the random forest and the BP neural network methods, the accuracy of the prediction model for the passenger passage time of gates is concentrated between 85%- 100%, and the RMSE and MAE can be as low as 2.93 and 2.66.