Innovative modeling for assessing driver behavior towards informaiton provision: A comparison between neutral networks and discrete modeling in route choice

Abstract

This paper proposes a different type of travel behavior model, under the effort to formulate route choice behavior in the presence of information. Until nowadays discrete choice models and especially the multinomial logit model are the most common used. In order to encounter problems associated with these models, which are based on the theory of utility maximization, we propose the implementation of models using Artificial Neural Networks, and especially supervised feedforward neural network models, exploiting their advantages and their functional similarity to the neural structure of the brain. A case study in the outer ring road of Thessaloniki (Greece) is examined, with the assistance of traffic simulation software. The results underscore the advantages of NN models, not only by means of prediction potential but also in demonstrating limited demand of time and effort, when someone tries to simulate a complex phenomenon like human behavior in a complex environment like city road network.