Modeling and estimation of travel behaviors using bayesian network

Computer simulation method has been used to measure the effect of new Intelligent Transportation Systems (ITS). Prior works which make use of the method simulated simplified travel demand, thus it is necessary to represent the demand correctly. Most of existing researches for forecasting travel behaviors need survey data of travel activity in target city. The data is called Person Trip (PT) data. Therefore, they are not able to be applied to cities where the survey was not conducted. In this paper, we propose a method for modeling and estimating travel behaviors, using Bayesian network (BN). BN is constructed based on dependency zone and trip characteristics. The zones are characterized by the important facilities for travelers. Our method is able to apply to the cities since the zone characteristics are available without PT data. In addition, the dependency is represented as graph structure obtained by using K2 algorithm. Our experimental results show the effectiveness of our method for estimating the behaviors.