Katia Andrade

Development of Transport Mode Choice Model by Using Adaptive Neuro-Fuzzy Inference System

Developing precise travel behavior models is important for estimating traffic demand and, consequently, for planning transportation systems. A study is presented that suggests a hybrid model that combines a stochastic model with a neuro-fuzzy inference system. The model is applied for estimating traveler behavior in the context of the problem of transport mode choice. Particularly, the multinomial logit model with neuro-fuzzy utility functions is developed to investigate shopping traveler preferences regarding the modes of bus, subway, and automobile. The model is evaluated by comparing its results with the results of a multinomial logit model. Moreover, the probabilities of selecting a transport mode obtained by applying the two models are compared with the actual transport mode choices, which show better performance of the proposed model. In addition, the model demonstrates good performance by estimating a large number of right choices during the validation process. A sensitivity analysis demonstrates the influence of time variations of mode subway on the probabilities of selecting a transport mode. The analysis highlights different behaviors of the models caused by the different utility functions. The results confirm that the proposed model can describe uncertainties regarding traveler decisions on the time of transport mode choice.

Investigating behavior of active cyclists

The behavior of active cyclists in Japan and their motivations for cycling were investigated. In Japan, bicycles were often used either as the main transportation mode or as part of a multimodal journey. However, few Japanese transportation policies and strategies address cycling. The current investigation thus addressed the reasons why people cycled although the cycling system had yet to achieve high levels of service. Commuters’ views on cycling were investigated in a behavioral analysis. In a statistical analysis, discrete choice models were estimated to assess the factors that had a strong influence on cycling. Results of the behavioral and statistical analyses are discussed with a focus on unimodal commuting trips. The findings of the models suggest that the single-person household type influences the choice of nonmotorized modes and that owning a car is not significant when people choose to cycle. Those who regularly cycle to perform other activities and those who regularly exercise are more likely to commute by bicycle. The possibility of sharing sidewalk space with pedestrians appears to have a strong positive impact on cycling. In addition, longer distances to public transport stations motivate the use of bicycle in multimodal journeys. The insight gained can be used by transport planners and government bodies in Japan and elsewhere in planning for more cycling-friendly cities.