5. Conclusions
This study contributes to the literature by addressing the issue of how to obtain a demand-oriented fare table in the context of revenue management and partially fills the research gap of price framing indicated by Guillet and Mohammed (2015). By giving an identical seat service, this study aims to reveal how passengers make their choices of booking classes in terms of RM-centric attributes and shows the trade-off effect among fares and fences. First of all, the modelling results of MNL do not show satisfactory outcomes and such unexpected results may come from the violation of IIA assumption in MNL. With the use of ML model, all the applied fences including departure time, booking time, ticket validity, changing fee, refund, and fare are shown to have significant in- fluences. Secondly, this study also considers the phenomenon of passenger heterogeneity toward the same seat service. The utilization of ML model can provide standard deviation information of the attributes. This study also reveals that heterogeneity does exist in the fences of booking time, ticket validity, changing fee, and fare. Passengers do possess different attitudes on these fences. Overall speaking, the mixed logit model may fit the data well and obtain more plausible estimation than the multinomial logit model. Thirdly, by combining the five studied fences, this study demonstrates how to generate one hundred and sixty two booking classes with corresponding fences/fares and provides a fare table for practice use.
