Reliability of travel time affects travel behavior such as departure time decision, transportation mode choice and also route choice. Therefore, the influence of travel time uncertainty of urban railway on railway commuters was focused on in this study. Internet survey was conducted to collect data from railway commuters living in Tokyo metropolitan area. Stated preference for railway service was executed in the survey. Four alternatives were presented to the respondent and each respondent was requested to choose the most preferable service. Average travel time, shortest travel time, longest travel time uncertainty, variability of travel time, congestion level in a vehicle and fare were considered as the compared factors. Meanwhile, it is thought that extent of interest in travel time reliability depends on trip purpose. Therefore, four kinds of trip purpose were considered in the survey. Four kinds of trip purpose were to commute, to attend business meeting, to go shopping and to go to airport. The data obtained by the stated preference choice experiments was used for parameter estimation of railway service choice model. At first, multinomial logit model were estimated by trip purpose and the weight for the travel time uncertainty was verified. Subsequently, latent class logit model was estimated and the validity of considering multiclass to estimate choice behavior model was examined. According to the Bayesian information criterion, it was demonstrated that latent class logit model was more useful to explain the choice behavior in travel for shopping and going to airport.

travel time reliability; stated preference survey; choice model; latent class model