Transferability of National Household Travel Survey Data to Local Areas: Disaggregate Approach

Traditionally, Metropolitan Planning Organizations are required to have their models calibrated on a continuing basis using new data. However, new survey data required to support these models do not exist in most urban areas. This makes it very difficult to calibrate existing models or develop new travel demand models using emerging modeling techniques. As a result, the need to assess potential approaches and develop knowledge on how to transfer and use the information collected in one context in another context is becoming critical. This paper presents the process of developing a model that can facilitate household travel data transferability. This can reduce or eliminate the need for a large data collection in the application context. Household records from the 2001 National Household Travel Survey are clustered into several homogeneous groups representing various household lifestyles. The clustering model is based on several variables that are easily obtainable from available sources. In addition to household characteristics, variables representing land-use, urban form, and transportation system characteristics are included in the analysis. Various statistical measures of travel behavior are estimated for each cluster. Using an artificial neural network model, cluster membership rules and travel statistics are then transferred to other geographic areas for the purpose of comparison and validation of the transferability method.