Multi-Source Data-Driven Estimation Model for Passenger Flow Management in Urban Rail Transit.

The accurate estimation of real-time passenger flow in urban rail transit (URT) networks plays a crucial role in optimizing the operation and management of urban rail transit systems, with profound implications for daily operation scheduling, passenger flow control, and safety management. Most existing studies have relied solely on a single data source such as Automatic Fare Collection (AFC) data for real-time passenger flow estimation. However, the inherent data uploading delay in the urban rail transit AFC system often leads to the delayed acquisition of passenger flow information. This severely limits the timeliness and accuracy of passenger flow estimation, thereby affecting the efficiency of URT operation and management. To address this critical challenge, this study proposes a multi-source data-driven estimation model to achieve the fusion of multi-source heterogeneous data covering the uploaded AFC data, historical passenger flow data and mobile phone signaling data collected from mainstream sensors such as through-beam photoelectric sensors and RFID/NFC sensors. In the proposed model, various types of information are taken into account by extracting features of the different data sources. The advantages of the proposed model are validated by utilizing multi-source data from Chengdu, China. The experimental results demonstrate that the proposed model achieves higher accuracy compared to existing benchmark models. Compared with the second-best-performing model, it reduces the MAE by 8.1%, RMSE by 10.7%, and MAPE by 17.2% at the 15 min time granularity, which shows that the proposed model has effective performance in terms of accuracy and stability.