Original scientific paper

https://doi.org/10.17559/TV-20230227000384

Multi-Step Subway Passenger Flow Prediction under Large Events Using Website Data

Qun Tu ; School of Economics and Management, Beijing University of Chemical Technology, Beijing 100029, China, No.15, Beisanhuandong Road, Chaoyang District, Beijing, China
Guining Geng ; 360 Digital Security Technology Group Co., Ltd., Beijing 100015, China, No. 6, Jiuxianqiao Road, Chaoyang District, Beijing, China
Qianqian Zhang ; School of Information, Beijing Wuzi University, Beijing 101149, China, No. 1 Fuhe Street, Tongzhou District, Beijing, China

Abstract

An accurate and reliable forecasting method of the subway passenger flow provides the operators with more valuable reference to make decisions, especially in reducing energy consumption and controlling potential risks. However, due to the non-recurrence and inconsistency of large events (such as sports games, concerts or urban marathons), predicting passenger flow under large events has become a very challenging task. This paper proposes a method for extracting event-related information from websites and constructing a multi-step station-level passenger flow prediction model called DeepSPE (Deep Learning for Subway Passenger Flow Forecasting under Events). Experiments on the actual data set of the Beijing subway prove the superiority of the model and the effectiveness of website data in subway passenger flow forecasting under events.

Keywords

deep learning; intelligent transportation systems; multi-task learning; subway passenger flow forecasting

Hrčak ID:

307745

URI

https://hrcak.srce.hr/307745

Publication date:

31.8.2023.

Visits: 877 *