Improving alighting stop inference accuracy in the trip chaining method using neural networks

Behrang Assemi (), Azalden Alsger, Mahboobeh Moghaddam, Mark Hickman and Mahmoud Mesbah
Additional contact information
Behrang Assemi: Queensland University of Technology (QUT)
Azalden Alsger: University of Queensland
Mahboobeh Moghaddam: University of Queensland
Mark Hickman: University of Queensland
Mahmoud Mesbah: Amirkabir University of Technology

Public Transport, 2020, vol. 12, issue 1, No 3, 89-121

Abstract: Abstract Public transport origin–destination (OD) estimation based on smartcard data has increasingly been used for transit network planning, passengers’ behaviour analyses and network demand forecasting. While various OD estimation methods using the trip-chaining approach have been developed in recent years, the validity of these estimation methods has not extensively been investigated. This study examines the errors in OD estimation caused by inaccurate inference of alighting stops, to improve the accuracy of the existing trip-chaining algorithms. The distribution of errors is evaluated both at the stop level and the public transport zonal level, given the geographical attributes of each zone and the attributes of the smartcard transactions. While the results show significant associations between zone attributes as well as transaction attributes and the alighting stop inference errors, they undermine the existing algorithm’s assumption that ‘travellers alight a public transport service at a stop which is the closest to their next boarding stop’. Accordingly, this study proposes and evaluates the application of a probabilistic approach using neural networks to infer alighting stops based on a combination of transactional and public transit network attributes. The proposed method is validated using 138,122 smartcard transactions obtained during a normal day in Southeast Queensland, Australia. The results show that our method can improve the accuracy of the existing algorithm by inferring the exact location of 79.5% of the alighting stops and reducing the mean alighting estimation error from 1689 to 503 m for incorrectly estimated alighting stops. At the zonal level, the proposed method also improves the accuracy of the existing algorithm by more than 5%. Finally, the study provides both researchers and practitioners with a method to improve the accuracy of the trip-chaining algorithm and OD estimation, and presents a list of practical guidelines for more effective planning and operation of public transit services.

Keywords: Origin–destination (OD) estimation; Alighting stop inference; Trip-chaining method; Error distribution; Neural network; Deep learning; Public transport; Smartcard data (search for similar items in EconPapers)
JEL-codes: R41 (search for similar items in EconPapers)
Date: 2020
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (3)

Downloads: (external link)
http://link.springer.com/10.1007/s12469-019-00218-9 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:pubtra:v:12:y:2020:i:1:d:10.1007_s12469-019-00218-9

Ordering information: This journal article can be ordered from
https://www.springer ... search/journal/12469

DOI: 10.1007/s12469-019-00218-9

Access Statistics for this article

Public Transport is currently edited by Stefan Voß

More articles in Public Transport from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().