Data Mining of Travel Surveys Using Bayesian Network Learning

Data mining techniques are potentially useful to discover relationships in data that may be overlooked in more theory driven or parametric approaches. In this paper we consider the problem of discovering dependency relationships in travel diary data by using Bayesian Network learning algorithms. We propose and compare two strategies which consider context, situational and socio-demographic variables either simultaneously with travel-related variables or as independent variables for explaining/predicting the behavioral variables. An application of these strategies on MON data – a national travel data set from The Netherlands – illustrates the potentials and limitations of the two approaches depending on the purpose of the analysis.