Usefulness of sequence alignment methods in validating rule-based activity-based models

Abstract

The aim of this paper is to achieve a better understanding of rule-based activity-based models by proposing a new level of validation on the process model level already existing in the ALBATROSS model. To that effect the work activity process model, which includes six different decision steps, is investigated. Each decision step or facet is evaluated during the prediction of individuals’ schedules. The comportment of execution in the process model contains activation dependency between modeled decisions, which branches the execution and evaluation of each agent under examination. This yields a sequence of decisions for each agent where the Sequence Alignment Method (SAM) is employed to evaluate predicted with observed decision sequences, as it utterly fits for assessing the analysis of decision sequences on this level. The original CHAID decision trees at each decision step utilized in ALBATROSS are compared with other well known induction methods chosen to appraise the purpose of the analyses on the process model level. Additionally, the performance of the models is compared at three existing validation levels: the classifier or decision step level using confusion matrix statistics, the work activity trip Origin-Destination (OD) matrix level and time of day work activity start time level, using the correlation coefficient to determine the degree of correspondence between the observed and predicted values. The results of validation on the proposed process model level show conformance to those already existing to validate rue-based activity-based models, with additional information to help in better understanding the process model's behaviour.