11.522: UIS Research Seminar (Fall 2009) - Discussion notes

Tuesday, October 20, 2009, 5:15 - 7:00 PM

Estimating Activity Destinations at Disaggregated Level: Towards an Activity-Based Approach

Discussion Leader: Shan Jiang

Traditional Four-Step Modeling (FSM) of travel demand and network allocation has been widely employed in the past half century, especially in the performance analysis of transportation systems.  However, there are many limitations of the FSM approach, including ignoring the spatio-temporal characteristics of household travel behavior, assuming a fixed pattern of underlying activities, and lacking integration with land use forecasting models.  In other words, the derived nature of the demand for transportation is not reflected well in the FSM methodology. (McNally & Rindt, 2008)

In contrast, activity-based approach has generated a much richer framework of estimating activity patterns at the individual and household level (Waddell, et al, 2008, 2002). However, the requirements for disaggregated data have greatly increased, ranging from population data, employment data, land use data, and travel survey data.  Employment data (on the travel destination side) is usually obtained from proprietary sources, which adds another layer of barriers to widely apply such modeling approach in addition to the expensive travel survey data acquisition.  This is particular true in cases outside of the North America. 

Recent developments in spatially-detailed, GIS-based data sources are making it practical to consider new methods for modeling urban activity in ways that can facilitate travel demand estimation.  Massive amounts of data on land use, point of interests (POI), high resolution orthophotos, public events, urban sensing, etc. are becoming available online.  These data, together with modern techniques for geoprocessing and data fusion, offer new possibilities for deriving activity destinations.  In urban settings, such analyses can also link travel patterns with different activity patterns in ways that can be usefully incorporated into models of land use and transportation interactions.

This research proposes to develop and analyze data fusion and data mining methods that use such data to calibrate models of urban activity that can substitute for traditional Trip Generation and Trip Distribution steps in the FSM at more disaggregated level.  The methods will be developed and illustrated by using Boston Metropolitan Area, MA, as an example, and applied to its counterpart of a Portuguese metropolitan area, Lisbon, after testing the reliability and robustness of the methods.  Data sources include land use, derived 'point of interest' information, road networks, high resolution orthophotos, and proprietary databases of destinations (for model validation).  This new approach for estimating activities and incorporating them into travel demand may also be beneficial for cities that lack detailed survey data for building Activity-Based Models but wish to test the sensitivity of travel behavior to policy options and ITS implementations that are likely to alter activity patterns.

Readings:

(1)      McNally, M. G. and Rindt, C. R. (2008).  The Activity-Based Approach. Handbook of Transportation Modeling (Edited by Hensher, D.A., and Button, K. J.), Chapter 4. Amsterdam; London: Elsevier.

(2)      Waddell, P., Wang, L., and Charlton, B. (2008). Integration of Parcel-Level Land Use Model and Activity-Based Travel Model. TRB 87th Annual Meeting Compendium of Papers DVD, TRB. Washington D.C. (Note: restricted use only).

Or Use the Following Video Presentation as a Substitution:

Waddell, P. (2008). "UrbanSim: Simulation, Planning and Visualization." Talk at the GeoDa Center for Geospatial Analysis and Computation, ASU, AZ.

(3)      Waddell, P. (2002). UrbanSim: Modeling Urban Development for Land Use, Transportation and Environmental Planning. Preprint of an article in the Journal of the American Planning Association, Vol. 68 No. 3, Summer 2002, pages 297-314.

Other Background Readings:

(4)      McNally, M. G. (2008). The Four Step Model. Handbook of Transportation Modeling (Edited by Hensher, D.A., and Button, K. J.), Chapter 3. Amsterdam; London: Elsevier.

(5)      Axhausen, K. (2008). Definition of Movement and Activity for Transport Modeling. Handbook of Transportation Modeling (Edited by Hensher, D.A., and Button, K. J.), Chapter 16. Amsterdam; London: Elsevier.

(6)      Sebastini, F. (2002). Machine Learning in Automated Text Categorization. ACM Computing Surveys, Vol. 34, No. 1, pp. 1–47.

Discussion Questions:

(1)      What are the fundamental differences between the traditional Four Step Model and the Activity Based Model of travel demand modeling? What are the data input structures for these two demand modeling approaches?

(2)      What are the advantageous applications of the Activity Based Model in urban management policy and decision making, compared with the Four Step Model?

(3)      Compare the Waddell 2002 and 2008 paper, and discuss:

a.  What is the evolution of spatial analysis unit used in UrbanSim?

b.  What are the arguments underling the changes of the geographic units?

c.  What are the advantages and disadvantages of the different types of geographic analysis units, both analytically and realistically?

Hint: Note the Grid Cell structure in the earlier version of UrbanSim model and the Parcel and Building structure in the most recent version of UrbanSim. Compare the applications of these two spatial structures in urban modeling and urban policy making.

(4)      In order to develop data mining and data fusion algorithms to estimate activity destinations,  incorporating with both aggregate data (e.g. census data at low resolution grid cell level, or in large geographic analysis units. i.e. TAZs) and disaggregate data (such as derived POI data using machine learning method), discuss the following:

a.  What are the potential measures to validate the reliability and robustness of the (machine learning) algorithms? Potential measures may include accessibility to different opportunities. What else? What will be a good structure of the accessibility measures?

b.  This research also proposes to apply spatial interpolation (or other non-parametric estimation) methods in estimating relatively high resolution activity destinations by using POI samples and aggregated data sources. What would be the potential shortcomings or disadvantages of this approach?