As an important mode of transportation connecting urban areas and airports, airport shuttle buses are affected by multiple coupled factors, including flight dynamics, metro operations, employee commuting, and weather conditions, resulting in significant nonlinearity, time variability, and fluctuations in short-term passenger flow.Specifically, flight arrivals generate passenger demand, but passengers must go through processes such as disembarkation, baggage claim, and walking to the bus stop before boarding, leading to a lagged driving effect of flight information on bus ridership.Metro operation conditions may change passengers’ mode choices, causing noticeable fluctuations in airport bus demand when metro services are unavailable or headways vary.In addition, concentrated employee commuting creates stable but prominent demand peaks during specific periods.The temporal mismatch and superposition of these factors make airport bus passenger flow patterns more complex, posing challenges to operational scheduling, capacity allocation, and accurate short-term passenger flow prediction.To enhance the accuracy of short-term passenger flow forecasting for airport shuttles, the GTC (Gamma Temporal Convolution)-TCN (Temporal Convolutional Network)-LSTM (Long Short-Term Memory)-Attention model is proposed, which integrates multi-source data and a time-delay modeling mechanism.The model incorporates heterogeneous data sources, including historical airport shuttle ridership, flight dynamics, metro operations, employee commuting data, and weather conditions, to build a comprehensive multidimensional feature input framework, thereby improving the representation of complex spatiotemporal flow patterns.A Gamma-based temporal convolution kernel is introduced to simulate the time-delay distribution between flight arrivals and passenger boarding, effectively capturing the lagged effect of flight information on shuttle ridership.The model architecture combines TCN, LSTM, and an attention mechanism to respectively model local dynamics, long-term dependencies, and responses to key driving factors, enabling accurate short-term forecasting.To show the performance of the proposed model, a simulation analysis is implemented based on the data from Xianyang International Airport.The empirical results indicate that the proposed model achieves an R2 of 0.912 5, an RMSE of 28.55, and an MAE of 19.82 on the test set, significantly outperforming multiple baseline models.Compared with a single-feature model using only historical ridership data and a multi-feature model without the Gamma mechanism, the proposed model reduces the average prediction error by 13.6% and 8.0%, respectively.These results demonstrate that the proposed model can effectively handle complex variations in shuttle demand driven by multi-source data, offering high predictive accuracy and strong practical value.
HeJ, LiY, ChaoY.Research on prediction and optimization of airport express passenger flow based on fusion intelligence network model [J].Appl Sci, 2024, 14(24): 11886.
[5]
ZhaoJ D, ShenJ, LiuL W.Bus passenger flow classification prediction driven by CNN-GRU model and multi-source data [J].J Traffic Transp Eng, 2021, 21(5): 265-273.
FuX, ZuoY, WuJ, et al.Short-term prediction of metro passenger flow with multi-source data: A neural network model fusing spatial and temporal features [J].Tunn Undergr Space Technol, 2022, 124: 104486.
[8]
HuZ A, DengJ C, YangJ H, et al.Rail transit station clustering and its impact on passenger flow forecasting [J].J Transp Syst Eng Inf Technol, 2023, 23(6): 227-238.
CuiH, SiB, WangJ, et al.Short-term origin-destination flow prediction for urban rail network: A deep learning method based on multi-source big data [J].Complex Intell Syst, 2024, 10(4): 4675-4696.
[11]
ChengY, LiH, SunS, et al.Short-term subway passenger flow forecasting approach based on multi-source data fusion [J].Inf Sci, 2024, 679: 121109.
[12]
ZhangY, SunK, WenD, et al.Deep learning for metro short-term origin-destination passenger flow forecasting considering section capacity utilization ratio [J].IEEE Trans Intell Transp Syst, 2023, 24(8): 7943-7960.
[13]
ZhaoL Q, LiR S, TangS X, et al.Short-term urban rail passenger flow prediction using temporal convolutional network-long short-term memory (TCN-LSTM) based on multidimensional predictable features [J].J Beijing Univ Chem Technol (Natural Science), 2024, 51(5): 86-96.
ZhengY, ZhengW, YinZ, et al.Predicting regional-level bus stop passenger flow with a multigraph fusion spatio-temporal graph attention network [J].J Transp Eng Part A: Syst, 2025, 151(2): 04024103.
[16]
WangT Z, XuJ H, ChenJ H, et al.Short-term passenger flow prediction method for urban rail transit considering station classification [J].Comput Eng Appl, 2024, 60(19): 343-353.
HuangW, LinY, LinB, et al.Modeling and predicting the occupancy in a China hub airport terminal using Wi-Fi data [J].Energy Build, 2019, 203: 109439.
[19]
XingZ W, KanB, ZhuS J, et al.A dynamic model of arriving passengers based on dual value driving [J].J Beijing Univ Aeronaut Astronaut, 2024, 50(12): 3645-3653.