(Passively-Collected Data)
Line 2: Line 2:
 
The flexibility of destination choice models comes at a cost. While it is possible to represent the selection of trip destinations more rigorously, destination choice models tend to require more data and data with higher fidelity than traditional [[Category:Spatial interaction models#Gravity Models|gravity models]] (link to Category:Spatial interaction models#Gravity Models). There are two types of data that are relevant for destination choice models. [[Destination Choice: Data Sources#Observed Choice Data|Observed choice data]] describe origin-destination flows that have been observed in a survey, by counting or by passive data collection. [[Destination Choice: Data Sources#Explanatory Data|Explanatory data]], on the other hand, refer to input data that describe either destinations or characteristics of the decision maker who chooses the destination.
 
The flexibility of destination choice models comes at a cost. While it is possible to represent the selection of trip destinations more rigorously, destination choice models tend to require more data and data with higher fidelity than traditional [[Category:Spatial interaction models#Gravity Models|gravity models]] (link to Category:Spatial interaction models#Gravity Models). There are two types of data that are relevant for destination choice models. [[Destination Choice: Data Sources#Observed Choice Data|Observed choice data]] describe origin-destination flows that have been observed in a survey, by counting or by passive data collection. [[Destination Choice: Data Sources#Explanatory Data|Explanatory data]], on the other hand, refer to input data that describe either destinations or characteristics of the decision maker who chooses the destination.
  
=Observed Choice Data=
+
==Observed Choice Data==
 
[[File:DestinationChoiceArrows.png|thumb|left]]Observed Choice Data describe actually chosen origin-destination pairs. Usually provided at a zone-to-zone resolution, these data provide at a minimum the origin and destination of individual trips. For [[Tour-based models|tour-based]] or [[Activity-Based Models|activity-based models]], an entire trip chain is provided, such as going from home to work, from work to a restaurant and the restaurant back home. Often, such data are stratified by trip purpose, mode, time of day and various socio-demographic characteristics of the traveler.  
 
[[File:DestinationChoiceArrows.png|thumb|left]]Observed Choice Data describe actually chosen origin-destination pairs. Usually provided at a zone-to-zone resolution, these data provide at a minimum the origin and destination of individual trips. For [[Tour-based models|tour-based]] or [[Activity-Based Models|activity-based models]], an entire trip chain is provided, such as going from home to work, from work to a restaurant and the restaurant back home. Often, such data are stratified by trip purpose, mode, time of day and various socio-demographic characteristics of the traveler.  
 
   
 
   
==Household Travel Surveys==
+
===Household Travel Surveys===
 
The most common source for Observed Choice Data are [[Household travel surveys|household travel surveys]]. Origins and destinations usually are collected at the address level and translated into traffic analysis zones (TAZ) for data analysis. Long-distance data commonly are provided at a coarser geography, such as counties or metropolitan areas. Surveys have the benefit that they tend to provide rich information on the socio-demographic characteristics of the traveler as well as the purpose of the trip. In addition to individual trips, surveys also commonly allow the analyst to identify entire tours.
 
The most common source for Observed Choice Data are [[Household travel surveys|household travel surveys]]. Origins and destinations usually are collected at the address level and translated into traffic analysis zones (TAZ) for data analysis. Long-distance data commonly are provided at a coarser geography, such as counties or metropolitan areas. Surveys have the benefit that they tend to provide rich information on the socio-demographic characteristics of the traveler as well as the purpose of the trip. In addition to individual trips, surveys also commonly allow the analyst to identify entire tours.
  

Revision as of 20:11, 24 May 2017

This page is open for editing because it is part of the Incubator. Have something to add? Please register so you can contribute. Have an option you would like to share? Please click on the 'Talk' button to enter the dialogue. The TF Resource Volunteers appreciate your feedback and interest.

The flexibility of destination choice models comes at a cost. While it is possible to represent the selection of trip destinations more rigorously, destination choice models tend to require more data and data with higher fidelity than traditional (link to Category:Spatial interaction models#Gravity Models). There are two types of data that are relevant for destination choice models. Observed choice data describe origin-destination flows that have been observed in a survey, by counting or by passive data collection. Explanatory data, on the other hand, refer to input data that describe either destinations or characteristics of the decision maker who chooses the destination.

Observed Choice Data

DestinationChoiceArrows.png
Observed Choice Data describe actually chosen origin-destination pairs. Usually provided at a zone-to-zone resolution, these data provide at a minimum the origin and destination of individual trips. For tour-based or activity-based models, an entire trip chain is provided, such as going from home to work, from work to a restaurant and the restaurant back home. Often, such data are stratified by trip purpose, mode, time of day and various socio-demographic characteristics of the traveler.

Household Travel Surveys

The most common source for Observed Choice Data are household travel surveys. Origins and destinations usually are collected at the address level and translated into traffic analysis zones (TAZ) for data analysis. Long-distance data commonly are provided at a coarser geography, such as counties or metropolitan areas. Surveys have the benefit that they tend to provide rich information on the socio-demographic characteristics of the traveler as well as the purpose of the trip. In addition to individual trips, surveys also commonly allow the analyst to identify entire tours.

Census Journey-to-Work

The Census Transportation Planning Products Program (CTPP), hosted by the American Association of State Highway and Transportation Officials (AASHTO) in cooperation with the states, collects data on work trips. The rich dataset includes where people live and work, their journey to work commuting patterns and the modes they use for getting to work. Commute flow data are provided at the county-to-county level by mode. No trip purposes other than journey to work are represented in this dataset.

Passively Collected Data

In contrast to survey data, passively collected data do not ask people about their travel behavior but rather collect data passively through cellular phones, GPS devices or other location-revealing technologies. While these data generally do not provide socio-demographic details of the traveler nor information on the motivation for a trip, these data have proven to be powerful because of their magnitude of coverage.

While surveys often cover 1 or 2 percent of the population, it is not uncommon for passively collected data to cover a quarter or a third of actual travel movements. An important benefit is the coverage of many origin-destination pairs. As an example, a household travel survey for the state of Tennessee covered with over 10,000 households roughly 40,000 origin-destination pairs, which is only 0.3 percent of all possible origin-destination pairs in Tennessee. Cell phone data, on the other hand, was able to capture 26 percent of all origin-destination pairs. There are many origin-destination pairs that are traveled very rarely (particularly from one rural area to another rural area in a different part of the state). Hence, it is expected that cell phone data was able to capture almost all origin-destination pairs that are actually traveled. The almost complete coverage has important benefits for the estimation of destination choice models, as the choice set is covered representatively.

A disadvantage of passively collected data is the lack of socio-economic characteristics of the traveler as well as missing information about the trip purpose. Sometimes, trip purposes are imputed by analyzing origin and destination. A trip from a residential area to a central business district could be classified as a home-to-work trip. The error margin of this imputation, however, is unknown. For this reason, passively collected data commonly are used either in applications where trip purposes are not distinguished or for model calibration or validation purposes.

Traffic Counts

Traffic counts may be used to impute origin-destination flows. In the so-called origin-destination matrix estimation (ODME), sometimes also called synthetic matrix estimation (SME), a trip matrix is synthesized that matches traffic count data.

ODME is a method to create a synthetic trip tables that resembles count data after assignment (Willumsen 1981[1]). Such models often suffered from unexpectedly large differences in outcomes due to small changes in inputs (Aerde, Rakha and Paramahamsan 2003[2]) as well as their inability to reconcile inconsistent or erroneous traffic counts (Hazelton 2003[3]). As traffic counts do not distinguish trip purposes or user classes, ODME cannot provide trip tables by purpose or trip tables that distinguish travelers by socio-economic characteristics. Given these shortcomings, trip matrices generated with ODME flows are only used if no other origin-destination data sources are available.

Explanatory Data

In addition to observed choice data, destination choice models also need information on possible destinations, such as retail facilities, parks or hotels. Similarly, information about the travelers, such as age, sex or income, are relevant when estimating destination choice models. These data often are called Explanatory Data, or size term data.

Land Use Data

The most common attractor used in destination choice modeling are socio-economic data, i.e. population and employment. Employment often is distinguished by industry category, such as manufacturing, retail, office and other. Obviously, shopping trips are mostly attracted by retail employment, while trips for visiting most frequently are attracted by population. An important limitation of these data is that categories tend to be rather broad. Retail employment, for example, includes destinations as diverse as bakeries and car dealers, two very different retail facilities that in reality would attract very different trips. Further, is has been shown that larger facilities tend to attract more trips per employees than smaller facilities. Nevertheless, zonal land use data are the most common data source for modeling trip destinations.

Passively-Collected Data

While land use data usually are based on census and business registration data, passively-collected data are gathered from online data sources, such as Facebook, Foursquare, Google or Twitter. Often, these data are also called Location-Based Social Network (LBSN) data. These websites provides Application Programming Interfaces (API) that allow downloading the location, type and size of various trip attractors. Trip attractors include, for example:

  • Restaurants and Bars
  • Hotels
  • Parks
  • Ski Resorts
  • Grocery Stores
  • and many more destinations (depending on the LBSN site)

Most LBSN websites allow downloading a small sample for free, while larger samples require paying a fee.

Choice-maker Data

Network Data

References

  1. Willumsen, L. G. (1981) Simplified transport models based on traffic counts. In: Transportation 10 (3):257-278. doi: 10.1007/BF00148462.
  2. Aerde, Michel, Hesham Rakha, and Harinarayan Paramahamsan (2003) Estimation of Origin-Destination Matrices: Relationship Between Practical and Theoretical Considerations. In: Transportation Research Record: Journal of the Transportation Research Board 1831 (-1):122-130. doi: 10.3141/1831-14.
  3. Hazelton, M. L. (2003) Some comments on origin-destination matrix estimation. In: Transportation Research Part A - Policy and Practice 37 (10):811-822. doi: Doi 10.1016/S0965-8564(03)00044-2.