Causal structure learning for travel mode choice using structural restrictions and model averaging algorithm

Tai Yu Ma, Joseph Ying Jun Chow, Jia Xu

Research output: Contribution to journalArticle

Abstract

This work contributes to develop a new methodology to identify empirical-data-driven causal structure of a domain knowledge. We propose an algorithm as a two-stage procedure by first drawing relevant prior partial relationships between variables and using them as structure constraints in a structure learning task of Bayesian networks (BNs). The latter is then based on a model averaging approach to obtain a statistically sound BN. The empirical study focuses on modeling commuters' travel mode choice. We present experimental results on testing the design of prior restrictions, the effect of resampling size and learning algorithms, and the effect of random draw on fitted BN structure. The results show that the proposed method can capture more sophisticated relationships between the variables that are missing in both decision tree models and random utility models.

Original languageEnglish (US)
Pages (from-to)1-27
Number of pages27
JournalTransportmetrica A: Transport Science
DOIs
StateAccepted/In press - Dec 20 2016

Fingerprint

Bayesian networks
travel
learning
Decision trees
Learning algorithms
commuter
Acoustic waves
Testing
methodology
knowledge

Keywords

  • Bayesian networks
  • causal structure
  • structure learning algorithm
  • travel mode choice

ASJC Scopus subject areas

  • Transportation
  • Engineering(all)

Cite this

@article{ecf6cdd3a4ab423189a7d25a6a6cb7c6,
title = "Causal structure learning for travel mode choice using structural restrictions and model averaging algorithm",
abstract = "This work contributes to develop a new methodology to identify empirical-data-driven causal structure of a domain knowledge. We propose an algorithm as a two-stage procedure by first drawing relevant prior partial relationships between variables and using them as structure constraints in a structure learning task of Bayesian networks (BNs). The latter is then based on a model averaging approach to obtain a statistically sound BN. The empirical study focuses on modeling commuters' travel mode choice. We present experimental results on testing the design of prior restrictions, the effect of resampling size and learning algorithms, and the effect of random draw on fitted BN structure. The results show that the proposed method can capture more sophisticated relationships between the variables that are missing in both decision tree models and random utility models.",
keywords = "Bayesian networks, causal structure, structure learning algorithm, travel mode choice",
author = "Ma, {Tai Yu} and Chow, {Joseph Ying Jun} and Jia Xu",
year = "2016",
month = "12",
day = "20",
doi = "10.1080/23249935.2016.1265019",
language = "English (US)",
pages = "1--27",
journal = "Transportmetrica A: Transport Science",
issn = "2324-9935",
publisher = "Taylor and Francis",

}

TY - JOUR

T1 - Causal structure learning for travel mode choice using structural restrictions and model averaging algorithm

AU - Ma, Tai Yu

AU - Chow, Joseph Ying Jun

AU - Xu, Jia

PY - 2016/12/20

Y1 - 2016/12/20

N2 - This work contributes to develop a new methodology to identify empirical-data-driven causal structure of a domain knowledge. We propose an algorithm as a two-stage procedure by first drawing relevant prior partial relationships between variables and using them as structure constraints in a structure learning task of Bayesian networks (BNs). The latter is then based on a model averaging approach to obtain a statistically sound BN. The empirical study focuses on modeling commuters' travel mode choice. We present experimental results on testing the design of prior restrictions, the effect of resampling size and learning algorithms, and the effect of random draw on fitted BN structure. The results show that the proposed method can capture more sophisticated relationships between the variables that are missing in both decision tree models and random utility models.

AB - This work contributes to develop a new methodology to identify empirical-data-driven causal structure of a domain knowledge. We propose an algorithm as a two-stage procedure by first drawing relevant prior partial relationships between variables and using them as structure constraints in a structure learning task of Bayesian networks (BNs). The latter is then based on a model averaging approach to obtain a statistically sound BN. The empirical study focuses on modeling commuters' travel mode choice. We present experimental results on testing the design of prior restrictions, the effect of resampling size and learning algorithms, and the effect of random draw on fitted BN structure. The results show that the proposed method can capture more sophisticated relationships between the variables that are missing in both decision tree models and random utility models.

KW - Bayesian networks

KW - causal structure

KW - structure learning algorithm

KW - travel mode choice

UR - http://www.scopus.com/inward/record.url?scp=85006918724&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85006918724&partnerID=8YFLogxK

U2 - 10.1080/23249935.2016.1265019

DO - 10.1080/23249935.2016.1265019

M3 - Article

SP - 1

EP - 27

JO - Transportmetrica A: Transport Science

JF - Transportmetrica A: Transport Science

SN - 2324-9935

ER -