Data-Driven Airline Fleet Assignment

The airline operational management involves the cooperation among different departments and the allocation of different resources.In the planning stage,the fleet assignment problem is the key to matching the market demand and resource supply,and thus it is crucial to the efficient use of resources and profit maximization.As the new resource of airlines,operational data is the cornerstone of intelligent transformation.To improve the airline’s operational management,this paper makes use of the operational data to optimize the globality,robustness,and accuracy of the fleet assignment decision.First,to deal with the suboptimality resulting from determining the flight slots,flight schedule,and fleet assignment sequentially,an integrated slot decision and biannual fleet assignment model is proposed to improve the result’s globality.Specifically,multiple monthly networks are combined to construct a biannual network,and an integrated biannual model is proposed.The computational results show that a profit improvement can be achieved through integrating the slot decision into the fleet assignment problem.Second,considering the airlines cannot accurately predict passenger demands when assigning a fleet to each flight,this paper proposes a distributionally robust fleet assignment model to improve the result’s robustness.Specifically,based on the statistical information from the available historical data,a first-order ambiguity set of passenger demand is constructed.Then,based on the demand ambiguity set,a distributionally robust model is proposed.According to the dual theory,the proposed model is reformulated and a tractable mathematical model is obtained.The computational results show that a profit improvement can be achieved through considering the demand uncertainty in the fleet assignment problem.Third,since most traditional fleet assignment models use constant price,which cannot capture the correlation between the assigned capacity and ticket price,this paper proposes a fleet assignment model considering this correlation to improve the result’s accuracy.Specifically,the revenue is approximated by piecewise linear functions,and thus a mixed integer programming model is obtained.Identifying different revenue functions into three typical groups,the model is modified to reduce the number of variables and constraints.Moreover,a segment-generation-based heuristic is developed to solve the models efficiently.The computational results show that a profit improvement can be achieved through considering the correlation between the assigned capacity and ticket price in the fleet assignment problem.