Consideration of different travel strategies and choice set sizes in transit path choice modelling

Path choice modelling is typically conducted by considering a subset of paths, not the universal set of all feasible paths as this is computationally challenging. This study proposes a two-stage modelling approach. In the first stage, it develops a new probabilistic importance sampling protocol by using fuzzy logic. In the second stage, it tested different structures of the discrete choice models, where different strategy attributes are considered along with the traditional variables. The results prove that the new sampling protocol performs better than traditional sampling protocol. Again, the inclusion of the strategy attributes proves to yield better prediction. The results of the study recommend considering different strategy attributes in the path generation process as well as in the transit assignment models. The study also discusses the effect of the choice set size on the model performance. Household travel survey data of south-east Queensland, Australia is used to develop the models.