Paper Details

TRB 91st Annual Meeting (January 22-26, 2012)

Event Number:	600
Event Title:	Taking Urban Data to New Heights: New Sources, New Techniques,and New Applications
Event Date:	Jan 24 2012 2:00PM- 3:45PM
Event Location:	Hilton, International Center
Event Description:
Event Agenda:

An Efficient Automatic Approach for Variable Selection to Visualize 2009 National Household Travel Survey Data (12-3219)

To maximize the utility and relevance of the 2009 National Household Travel Survey data program to the user community, a geospatial data visualization tool is being developed at Federal Highway Administration to provide support in data dissemination reporting for easy access to indicators of travel behavior, or measure variables. 20 key measure variables are manually selected with domain knowledge. The performance of such a measure variable is affected by a set of relevant variables representing the characteristics of traveling persons, their households, and their vehicles in the survey. Due to the large number of variables and significant amount of data, it is extremely difficult to allow users to conveniently understand and capture the key relevant variables to these measure variables. Therefore, it is necessary to adopt an automatic and efficient approach to effectively select key variables based on their importance or impacts to the corresponding measure variable, and present these key relevant variables in the visualization tool for users to explore the impacts of these variables to the corresponding measure variable. This paper introduces the Information Gain method from machine learning field into transportation field to automatically and efficiently select key relevant variables for a given measure variable. Major findings through the application of Information Gain are intuitive and consistent with domain knowledge and were validated by domain experts, and other findings that are not intuitive to domain experts but have strong relations to the corresponding measure variables are also identified, and they are invaluable findings to travel behavior analysts and modelers. All findings are beneficial to policy makers, planners, and travel behavior modelers to explore the relationship between these key variables and the corresponding measure variable of interest for decision making.

Authors
     Lu, Qifeng , MacroSys, LLC
     Fang, Bingsong , MacroSys, LLC
     Han, Xiaoli , MacroSys, LLC