A Novel System of Disaggregate Models for Travel Demand Modelling, Using Decision Tree and Random Forest Concepts

This research attempts to address the gap between research and practical transport demand models. Evolution of travel demand models started from the early paper and pencil versions of conventional four-step models of the late 50s and proceeded to the activity-based models. During this transition the emphasis shifted from aggregate to disaggregate models, whereby researchers increasingly paid attention to individual decision making regarding daily activities. This alteration engendered a conspicuous disparity between aggregate and disaggregate models regarding their practicality, precision and policy sensitivity level. This disparity is the main attention of this paper. Unlike the four-step models, a highly disaggregate travel demand modelling structure is proposed in this paper in which trip purpose, mode of transport, time of day, commute distance and attributes of trip destination are modelled. Unlike activity-based models, the proposed method is computational efficient and cost effective making it attractive for small or medium size cities. Destination choice problem is decomposed from the framework by using the attributes of the destination rather than its actual location. All the decisions are modelled using Decision Tree (DT), a Modified version of Decision Tree (MDT), and Random Forest (RF). The model estimation practice is done using the Victorian Integrated Survey of Travel and Activity (VISTA) of 2007 as the training set, and its transferability is examined by utilizing the 2009 year of the same dataset. As the results show RF and DT system of models have a high disaggregate accuracy, whereas MDT and RF system of models predict the aggregate trip patterns in the test dataset more similar to the observations.

  • Supplemental Notes:
    • This paper was sponsored by TRB committee ADB40 Transportation Demand Forecasting. Alternate title: Novel System of Disaggregate Models for Travel Demand Modeling Using Decision Tree and Random Forest Concepts
  • Corporate Authors:

    Transportation Research Board

    500 Fifth Street, NW
    Washington, DC  United States  20001
  • Authors:
    • Ghasri, Milad
    • Rashidi, Taha Hossein
    • Waller, S Travis
  • Conference:
  • Date: 2015

Language

  • English

Media Info

  • Media Type: Digital/other
  • Features: Figures; References; Tables;
  • Pagination: 17p
  • Monograph Title: TRB 94th Annual Meeting Compendium of Papers

Subject/Index Terms

Filing Info

  • Accession Number: 01557350
  • Record Type: Publication
  • Report/Paper Numbers: 15-3781
  • Files: TRIS, TRB, ATRI
  • Created Date: Dec 30 2014 1:14PM