Research On Flight Inventory Control Optimization Considering Passenger Transfer Behavior

Revenue management is an important technology of the airlines to increase income.With the rise of low-cost carriers,and the development of e-commerce and online travel agents,the booking differences between different fare classes are shrinking.These changes have brought challenges to traditional revenue management theory.The traditional theory assumes that passenger demand in different classes is independent among each other.But in fact,when the passenger's reservation request is rejected,he or she can choose to purchase a higher-priced class.And passengers will also consider whether there is a discount ticket even they are willing to buy a full-price ticket.Therefore,it is of great practical significance to study the revenue management with passenger transfer behavior.This paper mainly studies the demand forecasting and inventory control problems of passenger transfer behavior.Firstly,the existing seat inventory control as well as passenger choice models are introduced in detail.It is pointed out that under the increasingly popular simplified fare structure,passenger transfer behavior will occur more frequently.Then based on the multinominal logit model,the possibility of passengers selecting fare products are quantified.New forecasting methods of buy-up and buy-down probabilities are proposed,and the maximum likelihood and expectation maximization algorithm are used for parameter estimation.Then,a multi-fare class control model that considers buy-up and buy-down behavior simultaneously on a single leg is constructed,which improves the classic expected marginal seat revenue method.Under O&D network conditions,several optimization models incorporating buy-up behavior are put forward.Two approximate models of dynamic programming are combined with partitioned booking limit and bid price strategies to generate specific control policies.In the simulation experiment,the passenger arrival process is simulated as a non-homogeneous Poisson distribution.By comparing the models and strategies proposed in this paper with those without considering passenger transfer behavior,it is found that the revenue performance can be effectively improved.And that the larger the buy-up probability is,the more total revenues the airline will obtain,which shows the practicability of the algorithms.