The role of Occam's razor in activity based modeling

Abstract

This paper reports the results of a study conducted to assess the impact of the number of attibutes on the accuracy of tree induction algorithm and to find a good criterion to select the final model. To that effect, activity-travel diary data collected in the Netherlands are used for analysis. The results indicate that significantly smaller trees can be derived with a maximum loss of 2% in accuracy. Transport variables appear to be the most discriminating attributes in order to determine the transport mode chosen for work-related trips. For non-work tours, household characteristics are most important for this purpose.