Studying patterns of use of transport modes through data mining - Application to U.S. national household travel survey data set

Data collection activities related to travel require large amounts of financial and human resources to be conducted successfully. When available resources are scarce, the information hidden in these data sets needs to be exploited, both to increase their added value and to gain support among decision makers not to discontinue such efforts. This study assessed the use of a data mining technique, association analysis, to understand better the patterns of mode use from the 2009 U.S. National Household Travel Survey. Only variables related to self-reported levels of use of the different transportation means are considered, along with those useful to the socioeconomic characterization of the respondents. Association rules potentially showed a substitution effect between cars and public transportation, in economic terms but such an effect was not observed between public transportation and nonmotorized modes (e.g., bicycling and walking). This effect was a policy-relevant finding, because transit marketing should be targeted to car drivers rather than to bikers or walkers for real improvement in the environmental performance of any transportation system. Given the competitive advantage of private modes extensively discussed in the literature, modal diversion from car to transit is seldom observed in practice. However, after such a factor was controlled, the results suggest that modal diversion should mainly occur from cars to transit rather than from nonmotorized modes to transit