Current research approaches have not adequately considered the potential uncertainties of different modes. This paper develops an inference diagram (ID) approach to model travel mode choice with the consideration of travel time/cost uncertainty. The proposed ID model captures probabilistic relations between a trip's uncertain attributes and certain attributes. Given a trip's certain attributes, the uncertain attributes (e.g., travel time) can be inferred based on the Bayesian theorem; and the

more »

... pected utility (EU) theorem is utilized for choice decision making. Simulation-based optimization (SBO) is applied to estimating the parameters of the ID model. There are two major contributions in this paper: 1) the ID model considers uncertain attributes inference in the decisionmaking process of travel mode choice; 2) the ID model is capable of considering travelers' heterogeneous tastes of risk. A real-world travel survey data conducted in the Washington D.C. area is used for a case study. We construct ID models with different tastes of risk; a traditional Logit model is fitted for comparison. The results indicate the ID model is superior in classification accuracy compared with MNL models. With travelers' heterogeneous tastes of risk, the ID model is capable of providing interesting findings in travel behavior research.