Passenger Flow Prediction for Urban Rail Transit Stations Considering Weather Conditions

Abstract

Precise prediction of urban rail transit passenger flow is essential for the development of organizing plans by the rail transit management and operation department, and also is the fundament to achieving passenger transport guarantees. This study collected Ningbo rail transit Route 2 passenger flow data and candidates of key driving factors including station type, population and employment position density, transfer facilities, main land area within an 800 m radius, particularly considering weather conditions, and then Random Forest was applied for passenger flow prediction. The prediction results show that the models considering the weather factors is superior to the models without consideration, mean absolute deviation (MAD) and mean absolute percentage deviation (MAPD) are reduced by 14.40 and 57.55%, respectively. The model involved weather factors is more accurate under hot and heavy rain weather conditions. Employment position, population density and commercial service facilities land area within an 800 m radius of the station, are the most important factors influencing the passenger flow, while average temperature is more likely to affect the passenger flow than precipitation. These results suggest that the passenger flow forecasting model based on random forest can achieve rapid calculation under different weather conditions, and provide important data basis for urban rail transit passenger flow density warning, passenger flow guidance and operation scheduling.