Customer Choice Based Airline Schedule Model

In this paper, we consider the incremental airline schedule design integratedwith customer choice model. The incremental airline schedule design integrates flightschedule design problem and fleet assignment. Given the basic list of flight legs fromthe airline company, the incremental airline schedule design selects the set of flightlegs that will be included in the final schedule while assigning aircraft type for eachflight legs in the final schedule.Our model also captures customer’s preference by integrating conjoint-analysis in the airline scheduling model. Conjoint-analysis is widely used method tomeasure customer preference for each level of attribute. Our model measures utility atthe itinerary level. Among the different models, we selected Multinomial Logit(MNL) to measure more accurate utility.Previously, there are studies that attempted to integrate product line designand airline schedule design. Previous works are itinerary based fleet assignmentmodels which include spill cost and recapture revenue. The major con about itinerarybased schedule models is that accurate number of passengers cannot be captured dueto the estimated demand for each itinerary. As result, spilled demand and recaptureddemand are counted more than once in a certain Origin-Destination (OD) pair.To have more accurate and realistic airline schedule, we approach the airlinescheduling problem based on OD pairs. Instead of estimating demand for eachitinerary, we estimate demand for each OD pair in the network. Given the demand foreach OD pairs, our model will calculate customer preference for each itinerary andoptimize more realistic profit-maximizing airline schedule capturing interactionbetween supply and demand.Our model is Mixed Integer Nonlinear programming (MINLP) problem dueto integer decision variables, nonlinear objective, and nonlinear constraints. Theclassic way to solve MINLP problem is separating integer part and nonlinear part andsolve independently. We choose to implement our model with Branch-and-Bound algorithm, the most commonly used algorithm solving integer programming, andSequential Quadratic Programming (SQP) method, one of the widely used method tosolve Nonlinear programming (NLP) problems.The implementation is done in C++and compiled with Microsoft VisualStudio2008. For NLP solver, E04UCC function from Numerical Algorithms GroupC library is used. E04UCC function solves minimization of nonlinear programmingproblem using SQP method. We use the flight leg and aircraft information from majorU.S. airline to conduct case studies.There are two case studies presented in this paper. In the first case the fleetassignment is fixed, and in the second case fleet assignment is not fixed. The result ofour model is then compared with Passenger Mix Model (PMM), which finds mostprofitable number of passengers to capture over the schedule. Based on the casestudies, we observe that our model finds more realistic airline schedule that takescustomer preference into account.The result of various test cases shows that our model optimizes the price ofitinerary while assigning aircraft to each flight legs based on customer’s preference.Since the demand of customer is estimated based on the OD pair, not based on theitinerary, the total number of passengers traveling from one airport to other airportcontains spilled passengers and re-captured passengers without double counting re-captured passengers. Therefore, the result of our model can estimate more realisticand accurate demand for the future.