Using decision tree induction systems for modeling space-time behavior

Discrete choice models are commonly used to predict individuals' activity a nd travel choices either separately or simultaneously in activity schedulin g models. This paper investigates the possibilities of decision tree induct ion systems as an alternative approach. The ability of decision frees to re present heuristic decision rules is evaluated and a method of capturing int eractions across decisions In a sequential decision model is outlined. Deci sion tree induction algorithms, such as C4.5, CART, and CHAID, are suited t o device the decision rules from empirical data. A case study to illustrate the approach considers decisions of individuals when they are faced with t he choice to combine different out-of-home activities into a multipurpose, multistop trip or make a trip fbr each activity separately. Data fron a lar ge-scale activity diary survey are used to induce the decision rules. Possi ble directions of future research are identified.