Machine Learning and Daily Activity Patterns

Understanding the time-use activity patterns of population cohorts in the region will contribute greatly to modeling spatio-temporal urban transportation demand models. In this paper we present a comprehensive modeling methodology to forecast and replicate individual’s travel behavior within the Scheduler for Activities, Locations, and Travel (SALT) framework. The SALT modeling framework comprises a series of micro-behavioral modules that employ behaviorally-realistic econometric, advanced-machine learning, and data-mining techniques to construct the 24-hour activity schedule and the corresponding travel linked with activities accomplished by individuals. A state-of-art three-dimensional, four-stage pattern recognition model is developed to identify population clusters with homogeneous time-use daily activity patterns, and to derive a representative set of activity patterns in each cluster. A new agent-based inference model is developed to predict various facets of the daily activity agenda, such as stop number, activity type, and activity sequential arrangement. In the next phase, temporal attributes of each activity in the agenda are predicted and the 24-hour activity schedule of all individuals is formed through a heuristic decision rule-based algorithm. The data used for the analysis is from the large Halifax Space-Time-Activity-Research (STAR) household survey, which provides GPS-validated time-diary data for 2,778 person-days. Results show that the SALT scheduling model is able to assemble the traveler’s schedule with an average 82% accuracy in the 24-hour period. The proposed simulation modeling framework is useful for urban and transport modelers to advance transportation demand management for different segments of the urban population, as well as to analyze environmental mitigation and transport policy scenarios.

• Authors:
  • Hafezi, Mohammad Hesam
  • Daisy, Naznin Sultana
  • Millward, Hugh
  • Liu, Lei
• Conference:
  • Transportation Research Board 98th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2019-1-13 to 2019-1-17
• Date: 2019

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; References; Tables;
• Pagination: 21p

Subject/Index Terms

• TRT Terms: Activity choices; Machine learning; Travel behavior; Travel demand; Travel patterns; Urban transportation
• Subject Areas: Operations and Traffic Management; Planning and Forecasting; Transportation (General);

Filing Info

• Accession Number: 01715850
• Record Type: Publication
• Report/Paper Numbers: 19-03221
• Files: TRIS, TRB, ATRI
• Created Date: Sep 3 2019 9:44AM