The Use of Attitudinal Data in Building a Modal Split Model for the Urban Transportation Planning

Abstract

It is essential to clarify the modal choice structure of each transportation system when one plans for the comprehensive system of urban transportation which supports the city activities. If it might be so, a model for predicting modal choice, i.e., the modal split model, is to be constructed for the urban transportation planning.The purpose of this paper is to build a modal split model with attitudinal data. This is because, it would be essential, when the choice of transportation systems is analyzed, to focus upon the people's subjectivity, for urban activities seem to have become more complex due presumably to diversification of the people's subjectivity. In other words, we are controlled and adiministered by our own subjectivity, so information in the real world can only be obtained through the attitudinal data.The major results of this study are summarized as follows.1) Through the attitudinal survey, the following problem arises. How can the numbers of items in question be minimized? This problem is related to the most proper choice of explaining factors and the classification level of them. Therefore, in this study, the tables of orthogonal arrays are adopted to reduce the numbers of items in question.2) This study discusses the optimal modal split model estimated by the attitudinal survey in which the tables of orthogonal arrays were used. In the results, a logit model, whose theoretical basis is well-established, is the most suitable model.3) The following problem arises when we use our modal split model built with the attitudinal data. How precisely can we predict our behavior from the data obtained from the attitudinal survey? This problem is related to the issue of “the discrepancy between attitude and behavior” which is the great concern of the researchers who are engaged in the attitudinal survey. Therefore, we compare forecasts of attitudinal model with observed data of origin-destination survey. In the results, the prediction values correspond with the observation values almost perfectly.