A tensor-based Bayesian probabilistic model for citywide personalized travel time estimation

Kun Tang, Shuyan Chen, Zhiyuan Liu, Aemal J. Khattak

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

Urban travel time information is of great importance for many levels of traffic management and operation. This paper develops a tensor-based Bayesian probabilistic model for citywide and personalized travel time estimation, using the large-scale and sparse GPS trajectories generated by taxicabs. Combined with the knowledge learned from historical trajectories, travel times of different drivers on all road segments in some time slots are modeled with a 3-order tensor. This tensor-based modeling approach incorporates both the spatial correlation between different road segments and the person-specific variation between different drivers, as well as the coarse-grain temporal correlation between recent and historical traffic conditions and the fine-grain temporal correlation between different time slots. To account for the variability caused by the intrinsic uncertainties in urban road network, each travel time entry in the built tensor is treated as a variable following a log-normal distribution. With the help of the fully Bayesian treatment, the model achieves automatic hyper-parameter tuning and model complexity controlling, and therefore the problem of over-fitting is prevented even when the used data is large-scale and sparse. The proposed model is applied to a real case study on the citywide road network of Beijing, China, using the large-scale and sparse GPS trajectories collected from over 32,670 taxicabs for a period of two months. Empirical results of extensive experiments demonstrate that the proposed model provides an effective and robust approach for urban travel time estimation and outperforms the considered competing methods.

Original languageEnglish (US)
Pages (from-to)260-280
Number of pages21
JournalTransportation Research Part C: Emerging Technologies
Volume90
DOIs
StatePublished - May 2018

Fingerprint

Travel time
Tensors
travel
Taxicabs
Trajectories
Global positioning system
road network
agricultural product
driver
Normal distribution
road
traffic
Tuning
Statistical Models
time
uncertainty
China
human being
experiment
Experiments

Keywords

  • Bayesian treatment
  • CANDECOMP/PARAFAC (CP) factorization
  • Log-normal distribution
  • Over-fitting
  • Tensor

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Automotive Engineering
  • Transportation
  • Computer Science Applications

Cite this

A tensor-based Bayesian probabilistic model for citywide personalized travel time estimation. / Tang, Kun; Chen, Shuyan; Liu, Zhiyuan; Khattak, Aemal J.

In: Transportation Research Part C: Emerging Technologies, Vol. 90, 05.2018, p. 260-280.

Research output: Contribution to journalArticle

@article{bd6f59c50a43486f8ea3a3add5f6f409,
title = "A tensor-based Bayesian probabilistic model for citywide personalized travel time estimation",
abstract = "Urban travel time information is of great importance for many levels of traffic management and operation. This paper develops a tensor-based Bayesian probabilistic model for citywide and personalized travel time estimation, using the large-scale and sparse GPS trajectories generated by taxicabs. Combined with the knowledge learned from historical trajectories, travel times of different drivers on all road segments in some time slots are modeled with a 3-order tensor. This tensor-based modeling approach incorporates both the spatial correlation between different road segments and the person-specific variation between different drivers, as well as the coarse-grain temporal correlation between recent and historical traffic conditions and the fine-grain temporal correlation between different time slots. To account for the variability caused by the intrinsic uncertainties in urban road network, each travel time entry in the built tensor is treated as a variable following a log-normal distribution. With the help of the fully Bayesian treatment, the model achieves automatic hyper-parameter tuning and model complexity controlling, and therefore the problem of over-fitting is prevented even when the used data is large-scale and sparse. The proposed model is applied to a real case study on the citywide road network of Beijing, China, using the large-scale and sparse GPS trajectories collected from over 32,670 taxicabs for a period of two months. Empirical results of extensive experiments demonstrate that the proposed model provides an effective and robust approach for urban travel time estimation and outperforms the considered competing methods.",
keywords = "Bayesian treatment, CANDECOMP/PARAFAC (CP) factorization, Log-normal distribution, Over-fitting, Tensor",
author = "Kun Tang and Shuyan Chen and Zhiyuan Liu and Khattak, {Aemal J.}",
year = "2018",
month = "5",
doi = "10.1016/j.trc.2018.03.004",
language = "English (US)",
volume = "90",
pages = "260--280",
journal = "Transportation Research Part C: Emerging Technologies",
issn = "0968-090X",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - A tensor-based Bayesian probabilistic model for citywide personalized travel time estimation

AU - Tang, Kun

AU - Chen, Shuyan

AU - Liu, Zhiyuan

AU - Khattak, Aemal J.

PY - 2018/5

Y1 - 2018/5

N2 - Urban travel time information is of great importance for many levels of traffic management and operation. This paper develops a tensor-based Bayesian probabilistic model for citywide and personalized travel time estimation, using the large-scale and sparse GPS trajectories generated by taxicabs. Combined with the knowledge learned from historical trajectories, travel times of different drivers on all road segments in some time slots are modeled with a 3-order tensor. This tensor-based modeling approach incorporates both the spatial correlation between different road segments and the person-specific variation between different drivers, as well as the coarse-grain temporal correlation between recent and historical traffic conditions and the fine-grain temporal correlation between different time slots. To account for the variability caused by the intrinsic uncertainties in urban road network, each travel time entry in the built tensor is treated as a variable following a log-normal distribution. With the help of the fully Bayesian treatment, the model achieves automatic hyper-parameter tuning and model complexity controlling, and therefore the problem of over-fitting is prevented even when the used data is large-scale and sparse. The proposed model is applied to a real case study on the citywide road network of Beijing, China, using the large-scale and sparse GPS trajectories collected from over 32,670 taxicabs for a period of two months. Empirical results of extensive experiments demonstrate that the proposed model provides an effective and robust approach for urban travel time estimation and outperforms the considered competing methods.

AB - Urban travel time information is of great importance for many levels of traffic management and operation. This paper develops a tensor-based Bayesian probabilistic model for citywide and personalized travel time estimation, using the large-scale and sparse GPS trajectories generated by taxicabs. Combined with the knowledge learned from historical trajectories, travel times of different drivers on all road segments in some time slots are modeled with a 3-order tensor. This tensor-based modeling approach incorporates both the spatial correlation between different road segments and the person-specific variation between different drivers, as well as the coarse-grain temporal correlation between recent and historical traffic conditions and the fine-grain temporal correlation between different time slots. To account for the variability caused by the intrinsic uncertainties in urban road network, each travel time entry in the built tensor is treated as a variable following a log-normal distribution. With the help of the fully Bayesian treatment, the model achieves automatic hyper-parameter tuning and model complexity controlling, and therefore the problem of over-fitting is prevented even when the used data is large-scale and sparse. The proposed model is applied to a real case study on the citywide road network of Beijing, China, using the large-scale and sparse GPS trajectories collected from over 32,670 taxicabs for a period of two months. Empirical results of extensive experiments demonstrate that the proposed model provides an effective and robust approach for urban travel time estimation and outperforms the considered competing methods.

KW - Bayesian treatment

KW - CANDECOMP/PARAFAC (CP) factorization

KW - Log-normal distribution

KW - Over-fitting

KW - Tensor

UR - http://www.scopus.com/inward/record.url?scp=85044730494&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044730494&partnerID=8YFLogxK

U2 - 10.1016/j.trc.2018.03.004

DO - 10.1016/j.trc.2018.03.004

M3 - Article

AN - SCOPUS:85044730494

VL - 90

SP - 260

EP - 280

JO - Transportation Research Part C: Emerging Technologies

JF - Transportation Research Part C: Emerging Technologies

SN - 0968-090X

ER -