Multiple Objectives in Travel Demand Modeling

Abstract

Traditional techniques for estimating travel demand models cannot always identify a model if the quality of the input data is poor. These techniques do not allow modelers to easily predefine types of travel behaviors that they or their clients believe cannot be true. Models estimated with the best academic practice also may occasionally fail important validation tests. These factors often lead practitioners to determine model parameters through an inefficient trial-and-error process. A multiobjective model estimation procedure is presented that overrules solutions that cannot meet either statistical or political criteria. This procedure is not intended to criticize the traditional modeling approach, but it illustrates that a more pragmatic approach is available and works efficiently. This conclusion is illustrated in the estimation of a demand model for Dublin, Ireland.