proposal of a sequential method for spatial interpolation of mode choice

The main objective of this study is to propose a sequential method for spatial interpolation of mode choice for household locations where choices are unobserved based on Decision Tree analysis and Geostatistics. Initially, Decision Tree analysis was applied in order to estimate the probability of mode choice in surveyed households, thus determining the numeric variable to be estimated by Ordinary Kriging. The data used is from the Origin-Destination Survey and Urban Transportation Evaluation Survey, carried out in 2007/2008 in the city of São Carlos (São Paulo/Brazil). The study area selected for geoestatistical modeling is a small region of the city with 110 sampling points. The mode choice was estimated for the study area revealing a tendency of increasing the probability of car usage from the center to the periphery of region. The proposed method can be an alternative to traditional approaches in both non-spatial modeling, especially for the case of lack of data from stated preference survey, as in spatial modeling, allowing estimation in various geographic coordinates.