Unveiling cabdrivers’ dining behavior patterns for site selection of ‘taxi canteen’ using taxi trajectory data

Pengxiang Zhao, Xintao Liu, Mei Po Kwan, Wenzhong Shi

Research output: Journal article publicationJournal articleAcademic researchpeer-review

15 Citations (Scopus)


In recent years, some big cities in China have established a number of ‘taxi canteens,’ which are special cafeterias intended only for cabdrivers to dine and rest. To dine and rest at the appropriate time is one of the most concerned problems for cabdrivers, since long dining delay may affect their health and driving safety, and arbitrary parking for dining will be fined and may even cause dangerous traffic accidents. The establishment of ‘taxi canteens’ is expected to mitigate these problems. But little has been done to examine and optimize the site selection of taxi canteens. This paper presents a data-driven approach to allocate ‘taxi canteens’ throughout a city with the main objective of minimizing the total distance between all dining demand locations identified from taxi GPS trajectories and the corresponding closest ‘taxi canteen’ locations. We propose a dining event detection method that considers four features using support vector machine and further identifies the spatiotemporal patterns of cabdrivers’ dining. A constrained optimization model is proposed to select locations for ‘taxi canteens.’ A case study is conducted in Wuhan, China, to evaluate how the identification of cabdrivers’ dining behavior patterns can support the site selection of ‘taxi canteens.’ The results indicate that the proposed method has superior performance.

Original languageEnglish
JournalTransportmetrica A: Transport Science
Publication statusAccepted/In press - 1 Jan 2018


  • constrained optimization model
  • dining patterns
  • site selection
  • support vector machine
  • Taxi canteen

ASJC Scopus subject areas

  • Transportation
  • General Engineering


Dive into the research topics of 'Unveiling cabdrivers’ dining behavior patterns for site selection of ‘taxi canteen’ using taxi trajectory data'. Together they form a unique fingerprint.

Cite this