Skip to Main content Skip to Navigation
Conference papers

A maximum entropy approach to update airlines demand distributions

Ioana-Codruta Bilegan 1 Carlos Albert Nunes Cosenza 2 Sergio Gonzalez-Rojo 3 Felix Mora-Camino 3
MAIAA - ENAC - Laboratoire de Mathématiques Appliquées, Informatique et Automatique pour l'Aérien
3 LAAS-DISCO - Équipe DIagnostic, Supervision et COnduite
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : In order to design an efficient market-reactive revenue management system for airlines, it is necessary to have at hand a representative probability distribution of demand by period, by fare class and by order of arrival. For this, a dual geometric programming problem can be formulated according to the principle of maximum entropy. Making use of the corresponding primal form of the geometric program, the large-scale convex optimization problem is transformed into a non-constrained non-convex minimization problem. The solution strategy proposed consists of two main steps: first, solve the primal form of the geometric program and second, compute by geometric inversion the updated probability distributions. A numerical solution of the non-convex primal geometric program is obtained using an ad hoc designed genetic algorithm. Its performances had been evaluated under different simulation scenarios involving various fare classes, several forecasting periods and different demand profiles, showing satisfactory results.
Complete list of metadatas
Contributor : Laurence Porte <>
Submitted on : Thursday, December 5, 2013 - 4:58:42 PM
Last modification on : Monday, May 18, 2020 - 4:50:33 PM
Document(s) archivé(s) le : Thursday, March 6, 2014 - 4:36:58 AM


Files produced by the author(s)



Ioana-Codruta Bilegan, Carlos Albert Nunes Cosenza, Sergio Gonzalez-Rojo, Felix Mora-Camino. A maximum entropy approach to update airlines demand distributions. IPDO 2004, Inverse Problems, Design and Optimization Symposium, Mar 2004, Rio de Janeiro, Brazil. pp 19-29, ⟨10.1080/17415970600573825⟩. ⟨hal-00914276⟩



Record views


Files downloads