Understanding individual modal choice decisions is fundamental when developing and planning transportation systems. Developing precise travel behavior models and testing their forecasting capabilities also have an important role. This study provides empirical analysis of the temporal transferability of Multinomial Logit and Neuro-Fuzzy Multinomial Logit models. In the estimation context, the results of the models are compared with the actual modal choices to investigate their accuracy. In the application context, the evaluation of the models is based on predictive performances, where models re-calibrated from a small data sample and models transferred from the estimation context are evaluated. Accordingly, a sensitivity analysis that aims to examine travelers' behavior under the transportation system changes is presented. Overall, the Neuro-Fuzzy Multinomial Logit model performs better than the Multinomial Logit model. However, both models do not show satisfactory forecasting behavior when directly transferred to the application context. The classical Logit model is not able to correctly represent the influence of Parking Cost on modal choices, although this attribute is identified as statistically important for modeling modal choices. However, the neuro-fuzz model has shown that travelers are very sensitive to variations on parking fees, which was implemented as a strategic policy action so as to motivate changes in travelers' behavior towards sustainable transport modes. The results of this study suggest that travelers' behavior could be better explained by incorporating the neuro-fuzzy utility functions into the Multinomial Logit model rather than using the classical Multinomial Logit structure. JEL Classification: R4 Transportation Studies R41 Transportation: Demand, Supply and Congestion Safety and Accidents, Transportation Noise