Aiming for Parsimony in the Sequential Analysis

This paper aims at a better understanding in the impact of simplification in a sequential analysis of activitydiary data using a feature selection method. To this effect, the predictive performance of the Albatross model, which incorporates nine different facets of activity-travel behaviour, based on the original full decision trees is compared with the performance of the model based on trimmed decision trees. The more parsimonious models are derived by first using a feature selection method to determine the irrelevant variables which are then left out of the further model building process. The results indicate that significantly smaller decision trees can be used for modelling the different choice facets of the sequential system without loosing much too much in predictive power. The performance of the models is compared at two levels: the choice facet level, at which we compare the performance of each facet separately and the trip level, comparing the correlation coefficients that determine the strength of the associations between the observed and the predicted origindestination matrices. The results indicate that the model based on the trimmed decision trees predicts activity diary schedules with a minimum loss of accuracy at the choice facet level. Moreover, the results show a slightly better performance at the trip matrix level.