A probabilistic model for estimating travel choice

A travel choice model is proposed. It is derived from the hypothesis that travel alternatives can be represented by certain system attributes which affect travel behaviour and that travellers make their choices according to the weights (values) they place on these attributes. The weights are treated as random variables which vary among individuals. The model computes the probability that an alternative is chosen by an individual. Total share for this alternative is then estimated by finding the distributions of these weights and aggregating over the population. New alternatives can be studied once their attribute values are specified. Equations for direct and cross elasticities of choice as well as the implicit dollar value of attributes are also derived. To demonstrate model application, the route selection for an air trip in a corridor that could be served by both short-take-off-and-landing and conventional air transport systems is studied. Although the proposed model was calibrated with deficient data, the results indicate the model is workable. Comparison with results of a logit model calibrated on the same data showed the proposed model gives a better fit, at least for the data used. Further testing with more suitable data is recommended.