A choice-based mixed integer programming formulation for network revenue management

Airlines routinely use revenue management techniques to improve their revenue performance by optimizing the available fares for the various flights that they operate within their network. These approaches often assume independent passenger demand, which essentially ignores the interactions between various fare classes and routing options. In this article, we build on previous studies that explicitly incorporate passenger choice and fare class/routing interactions using a MultiNomial Logit discrete choice model by presenting an alternative Mixed Integer Programming formulation for the problem. The formulation yields substantial reduction in the number of variables over the Choice-based Deterministic Linear Program while maintaining revenue performance. Using small and larger network examples from the literature, we demonstrate performance improvements that one can obtain in comparison to popular leg-and network-based revenue management methods currently used in practice, and explain how the solution can be implemented in real-life systems.