Abstract
This paper presents a dynamic pricing framework for airline revenue management, leveraging simulation-based Approximate Policy Iteration (API) technique to optimize pricing and capacity control decisions. The framework incorporates a price-sensitive multinomial logit (MNL) function for customer choice modeling to learn the bid prices representing seat opportunity cost to guide pricing and capacity control decisions. We compare two policies: (1) typical capacity control that optimizes product offer sets under fixed prices, and (2) a dynamic pricing policy that jointly optimizes both product offer sets and associated prices. The generation of the dynamic pricing policy uses epsilon-greedy exploration and neighborhood local search for offer set optimization, and constrained linear regression to estimate the value function in the dynamic programming (DP) model. Simulation results show that the dynamic pricing policy significantly increases revenue.
Recommended Citation
Huang, Kuancheng and Chung, Pei-Chen, "Solving a Dynamic Revenue Management Problem of Airlines by Bid Price Learning and Approximate Policy Iteration" (2025). ICEB 2025 Proceedings (Hanoi, Vietnam). 43.
https://aisel.aisnet.org/iceb2025/43