International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
Modeling two-destination choice with GPS data in the context of non-work trip chains
Arthur Huang, David Levinson

Last modified: 11 May 2015


This research focuses on non-work vehicle trips. According to the 2009 National Household Travel Survey, non-work trips make up approximately 90% of total trips. Non-work trips include a spectrum of trip purposes, such as social/recreational trips, shopping trips, family/personal errands, and school/church activities. Because of its significant share in daily travel, it is of interest to investigate non-work travel behavior and the influencing factors using empirical data. In addition, the advancements of GPS and GIS technologies provide new opportunities for investigating non-work destination choice at the microscopic level. This research models how land use and road network structure influence two-destination choice in the context of non-work trip chains, using the in-vehicle GPS travel data in the Minneapolis-St. Paul Metropolitan Area.

The in-vehicle GPS data collection process lasted from September to December of 2008, during which 141 surveyed subjects made over 20,000 trips. The data collection consisted of three stages: (1) recruiting subjects and installing GPS devices in participants' vehicles, (2) collecting GPS trips from GPS devices, and (3) creating and analyzing trip trajectories. The trip trajectories were drawn based on the GPS points in the underlying the Twin Cities Metropolitan Planning Network.

This research proposes a new framework that explicitly considers non-work, two-destination choice in a home-based trip chain. In modeling non-work, multiple-destination choice, we propose a new choice set formation approach which combines survival analysis and random selection for two-destination choice in a trip chain. After testing three prototypes, we recommend the major/minor-destination approach for modeling two-destination choice. The modeling procedure can be expanded to trip chains of more than two destinations.

The key contributions of this research include:

  1. Proposing a model that explicitly considers home-based two-destination choice (i.e., home $\rightarrow$ destination 1 $\rightarrow$ destination 2 $\rightarrow$ home).
  2. Proposing and empirically testing three methods to form choice sets for modeling home-based two-destination choice.
  3. Empirically applying the reference-dependent theory to the two-destination choice problem.
  4. Investigating the impacts of land use and characteristics of travel routes on two-destination choice based on the in-vehicle GPS data.

The key findings are summarized as follows:


  1. Accessibility and road network structure influence two-destination choice
  2. Total travel time and travel time between the major destination and work influence the attractiveness of a two-destination trip chain the most.
  3. Greater accessibility and diversity of services at the two destinations make them more attractive. The land uses at the two destinations do not exert the same level of influence.
  4. More dissimilar destination in a trip chain are more likely to be selected, which may be due to greater complementarity of services one may engage in by chaining the trips.
  5. Route-specific network measures impact destination choice. The travel route with fewer turns per unit time makes a trip chain more attractive.
  6. A trip chain producing a higher travel time saving ratio is more attractive to travelers. %Individuals' perception of travel time saving is shown to influence non-work destination choice.
  7. Individuals' familiarity factors reflected by distance to home, work, and downtown also plays a role in destination choice in non-work trip chains.

The findings shed light on transportation and land use planning in several aspects. First, shaping an attractive retail zone needs careful planning. It not only concerns increasing the number of services and diversity of services in the zone itself, but also may be related with the types of services in other destinations to provide complementary services. Second, travel routes' network structure influences where people drive to shop. Third, a major destination's land use characteristics influence destination choice more than a minor destination in a trip chain. It may be speculated that if one's major destinations of interest are all far away from home, one has to drive a long distance for non-work trip purposes. It provides food for thought when it comes to the policy question of how to reduce the amount of vehicle travel.


Conference registration is required in order to view papers.