Representing and predicting tourist choice behavior : a rule-based vs. a utility-based approach

As an alternative to the more traditional utility-based models. this article proposes and tests a rulebased approach to modeling tourist choice behavior. The CHAID tree-induction algorithm is used to formulate and test tourists' decision rules for the choice of season. The article shows that a set of 100 exclusive and exhaustive decision rules equals the predictive performance of a MNL model involving 255 parameters. Moreover, it is shown that the number of parameters of a MNL model can be reduced without decreasing the predictive performance by using CHAID preprocessed condition states. Overall, however, none of the models clearly outperforms the other models.