Multilayered Railway Passenger Demand Estimation Considering Nested Choices: A Computational Graph-based Learning Framework

Abstract

This paper attempts to develop a systematic and theoretically consistent machine learning model to quantify and estimate multi-level railway passenger demand, capturing overall demand patterns. This includes estimating boarding and alighting passengers at stations along a corridor, origin–destination trips between stations, and passenger flows loaded onto train lines and corresponding train line segments. Observations are transformed into loss functions and mapped onto a hierarchical flow network to simultaneously estimate these multilayered demand variables. By incorporating a Nested Logit model into the hierarchical flow network, the learning model can further calibrate a series of interpretable parameters associated with key attributes in rail line planning (e.g., line frequency, fare levels, and travel time), enabling policy-sensitive analyses. Unlike pure discrete choice models, the proposed estimation model is constrained by line-based capacity constraints to prevent passenger flow overload on specific train lines and address inconsistencies between different observations. The nonlinear estimation model is reformulated as a computational graph and solved using backpropagation algorithms with off-the-shelf machine learning solvers (e.g., TensorFlow). To validate the applicability of our approach, we conduct a real-world case study on the Beijing–Shanghai high-speed rail corridor. The evolution of multiple loss functions in the case study demonstrates the accuracy and convergence of the method. Fourteen days of ticket sales data (10 days for training and 4 days for validation) are utilized to demonstrate the applicability of the proposed method.

Keywords

Railway passenger demand / Hierarchical flow network / Nested Logit model / Computational graph / Backpropagation

References

[1] SmallK, VerhoefETThe economics of urban transportation, 2007, Abingdon. Routledge.

[2] SunJ, GuoJ, WuX, ZhuQ, WuD, XianK, ZhouX. Analyzing the impact of traffic congestion mitigation: From an explainable neural network learning framework to marginal effect analyses. Sensors, 2019, 19102254.

[3] WuX, GuoJ, XianK, ZhouX. Hierarchical travel demand estimation using multiple data sources: a forward and backward propagation algorithmic framework on a layered computational graph. Transp Res Part C: Emerg Technol, 2018, 96: 321-346.

[4] GoossensJW, van HoeselS, KroonL. On solving multi-type railway line planning problems. Eur J Oper Res, 2006, 168(2): 403-424.

[5] NuzzoloA, CrisalliU, GangemiF. A behavioural choice model for the evaluation of railway supply and pricing policies. Transp Res Part A: Policy Pract, 2000, 34(5): 395-404.

[6] BlaineyS. Trip end models of local rail demand in England and Wales. J Transp Geogr, 2010, 18(1): 153-165.

[7] Blainey S , Mulley C (2013) Using geographically weighted regression to forecast rail demand in the Sydney region. Australas Transp Res Forum. 1-16

[8] BlaineyS, PrestonJ. Modelling local rail demand in South Wales. Transp Plan Technol, 2010, 33(1): 55-73.

[9] CorderaR, SañudoR, dell’OlioL, IbeasÁ. Trip distribution model for regional railway services considering spatial effects between stations. Transp Policy, 2018, 67: 77-84.

[10] FitzRoyF, SmithI. The demand for rail transport in European countries. Transp Policy, 1995, 2(3): 153-158.

[11] FitzRoy F, Smith I (1998) Passenger rail demand in 14 western European countries: a comparative time series study. Int J Transp Econ. 299-312.

[12] KoppelmanFS, BhatCA self instructing course in mode choice modeling: multinomial and nested logit models, 2006, Department of Transportation, Federal Transit Administration, Washington, DC. U.S.

[13] TrainKEDiscrete choice methods with simulation, 2009, Cambridge. Cambridge University Press.

[14] CederAPublic transit planning and operation: Modeling, practice and behavior, 2016, Boca Raton. CRC Press.

[15] Sheffi Y (1985) Urban transportation networks. (Vol. 6). Prentice-Hall, Englewood Cliffs.

[16] TangJ, ZhenF, CaoJ, MokhtarianPL. How do passengers use travel time? A case study of Shanghai-Nanjing high speed rail. Transportation, 2018, 45: 451-477.

[17] ZhangJ, ChenF, YangL, MaW, JinG, GaoZ. Network-wide link travel time and station waiting time estimation using automatic fare collection data: a computational graph approach. IEEE Trans Intell Transp Syst, 2022, 23(11): 21034-21049.

[18] RezikaWY, PriyantoS, IrawanMZ. Estimation urban railway demand in Yogyakarta using bivariate ordered probit model. MATEC Web of Conf, 2018, 18102009.

[19] KatoH, KanekoY, InoueM. Comparative analysis of transit assignment: evidence from urban railway system in the Tokyo Metropolitan Area. Transportation, 2010, 37: 775-799.

[20] Espinosa-ArandaJL, García-RódenasR, del CarmenR-F, López-GarcíaML, AnguloE. High-speed railway scheduling based on user preferences. Eur J Oper Res, 2015, 246(3): 772-786.

[21] HalyalS, MulangiRH, HarshaMM. Forecasting public transit passenger demand: With neural networks using APC data. Case Stud Transp Policy, 2022, 10(2): 965-975

[22] WangS, WangQ, ZhaoJ. Deep neural networks for choice analysis: Extracting complete economic information for interpretation. Transp Res Part C: Emerg Technol, 2020, 118102701.

[23] LiC, BaiL, LiuW, YaoL, WallerST. Graph neural network for robust public transit demand prediction. IEEE Trans Intell Transp Syst, 2020, 23(5): 4086-4098

[24] LeeJ, ChoiY, KimJ. A multi-task deep learning framework for forecasting sparse demand of demand responsive transit. Expert Syst Appl, 2024, 250123833.

[25] NoursalehiP, KoutsopoulosHN, ZhaoJ. Dynamic origin-destination prediction in urban rail systems: a multi-resolution spatio-temporal deep learning approach. IEEE Trans Intell Transp Syst, 2021, 23(6): 5106-5115

[26] Toqué F, Khouadjia M, Come E, Trepanier M, Oukhellou L (2017) Short & long term forecasting of multimodal transport passenger flows with machine learning methods. In: 2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC): 560-566. https://doi.org/10.1109/ITSC.2017.8317939

[27] García-GarcíaJC, García-RódenasR, López-GómezJA, Martín-BaosJÁ. A comparative study of machine learning, deep neural networks and random utility maximization models for travel mode choice modelling. Transp Res Procedia, 2022, 62: 374-382.

[28] KimT, ZhouX, PendyalaRM. Computational graph-based framework for integrating econometric models and machine learning algorithms in emerging data-driven analytical environments. Transp A: Transp Sci, 2022, 18(3): 1346-1375.

[29] WangS, WangQ, BaileyN, ZhaoJ. Deep neural networks for choice analysis: a statistical learning theory perspective. Transp Res Part B: Methodol, 2021, 148: 60-81.

[30] WangS, MoB, ZhengY, HessS, ZhaoJ. Comparing hundreds of machine learning and discrete choice models for travel demand modeling: An empirical benchmark. Transp Res Part B Methodol, 2024.

[31] Zhao X, Yan X, Yu A, Van Hentenryck P (2018) Modeling stated preference for mobility-on-demand transit: a comparison of machine learning and logit models. arXiv preprint. https://doi.org/10.48550/arXiv.1811.01315

[32] Došilović FK, Brčić M, Hlupić N (2018) Explainable artificial intelligence: A survey, In: 2018 41st International convention on information and communication technology, electronics and microelectronics (MIPRO): 0210-0215. https://doi.org/10.23919/MIPRO.2018.8400040

[33] BengioY, GoodfellowI, CourvilleADeep learning, 2017, Cambridge. MIT Press. 1

[34] MaW, PiX, QianS. Estimating multi-class dynamic origin-destination demand through a forward-backward algorithm on computational graphs. Transp Res Part C: Emerg Technol, 2020, 119102747.

[35] Vasilev I, Slater D, Spacagna G, Roelants P, Zocca V (2019) Python Deep Learning: Exploring deep learning techniques and neural network architectures with Pytorch, Keras, and TensorFlow. Packt, Birmingham

[36] Bartholomew-BiggsM, BrownS, ChristiansonB, DixonL. Automatic differentiation of algorithms. J Comput Appl Math, 2000, 124(1–2): 171-190.

[37] KarniadakisGE, KevrekidisIG, LuL, PerdikarisP, WangS, YangL. Physics-informed machine learning. Nat Rev Phys, 2021, 3(6): 422-440.

[38] LuJ, LiC, WuXB, ZhouXS. Physics-informed neural networks for integrated traffic state and queue profile estimation: a differentiable programming approach on layered computational graphs. Transp Res Part C: Emerg Technol, 2023, 153104224.

[39] ZhangJ, MaoS, YangL, MaW, LiS, GaoZ. Physics-informed deep learning for traffic state estimation based on the traffic flow model and computational graph method. Inform Fusion, 2024, 101101971.

[40] ZhangS, ZhangJ, YangL, ChenF, LiS, GaoZ. Physics guided deep learning-based model for short-term origin-destination demand prediction in urban rail transit systems under pandemic. Engineering, 2024, 41: 276-296.

[41] KimT, LuJ, PendyalaRM, ZhouXS. Computational graph-based mathematical programming reformulation for integrated demand and supply models. Transp Res Part C: Emerg Technol, 2024, 164104671.

[42] LiuZ, YinY, BaiF, GrimmDK. End-to-end learning of user equilibrium with implicit neural networks. Transp Res Part C: Emerg Technol, 2023, 150104085.

[43] LiuJ, ZhouX. Capacitated transit service network design with boundedly rational agents. Transp Res Part B: Methodol, 2016, 93: 225-250

[44] BoyceDE, ZhangYF, LupaMR. Introducing" feedback" into four-step travel forecasting procedure versus equilibrium solution of combined model. Transp Res Rec, 1994, 144365

[45] PeetersLCyclic railway timetable optimization, 2003, Rotterdam. Erasmus University Rotterdam.

[46] FuH, NieL, MengL, SperryBR, HeZ. A hierarchical line planning approach for a large-scale high speed rail network: the China case. Transp Res Part A: Policy Practi, 2015, 75: 61-83.

[47] NieB, NieL, FuH, WuX. Weekly line planning to improve the match between rail supply and passenger demand. Comput-Aided Civ Infrastr Eng, 2023, 38(13): 1810-1833.

[48] Schöbel A, Scholl S (2006) Line planning with minimal traveling time. 5th Workshop on Algorithmic Methods and Models for Optimization of Railways (ATMOS'05). Schloss Dagstuhl-Leibniz-Zentrum für Informatik. https://doi.org/10.1109/CarpathianCC.2018.8399626

[49] PatrikssonMThe traffic assignment problem: models and methods, 2015, Mineola. Dover Publications.

[50] WuX, NieL, XuM, YanF. A perishable food supply chain problem considering demand uncertainty and time deadline constraints: modeling and application to a high-speed railway catering service. Transp Res Part E: Logist Transp Rev, 2018, 111: 186-209.

[51] GlowinskiRFitzgibbonW, ItoY, KuznetsovY, PironneauO. On alternating direction methods of multipliers: a historical perspective. Modeling, simulation and optimization for science and technology, 2014, Dordrecht. Springer. 5982

[52] YaoY, ZhuX, DongH, WuS, WuH, TongLC, ZhouX. ADMM-based problem decomposition scheme for vehicle routing problem with time windows. Transp Res Part B: Methodol, 2019, 129: 156-174.

[53] Martín-BaosJÁ, García-RódenasR, López-GarcíaML, Rodriguez-BenitezL. Discrete choice modeling using kernel logistic regression. Transp Res Procedia, 2020, 47: 457-464.

[54] Martín-BaosJÁ, García-RódenasR, Rodriguez-BenitezL. Revisiting kernel logistic regression under the random utility models perspective. An interpretable machine-learning approach. Transp Lett, 2021, 13(3): 151-162.

[55] Martín-BaosJÁ, García-RódenasR, Rodriguez-BenitezL. A Python package for performing penalized maximum likelihood estimation of conditional logit models using Kernel Logistic Regression. Transp Res Proced, 2021, 58: 61-68.

RIGHTS & PERMISSIONS

This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply