Pricing Game of Flight Ticket Using Reinforcement Learning

Flight ticket pricing not only significantly impacts airlines’ operations but also plays a crucial role in consumer welfare and the sustainable development of the air transportation industry. The pricing of tickets has consistently been a focus of research for scholars and experts in the aviation industry. However, existing works often lack the research on integrating multiple factors into decision-making across continuous decision points. To address this gap, we propose a new game approach involving multiple intelligent agents across consecutive states to simulate the sequential game process of parallel flights using reinforcement learning theory. The action space represents the price space, action selection corresponds to price decisions, and the state space represents the set of decision points. Ultimately, the optimal action sequence output by the intelligent agents represents equilibrium solution of sequential game. Particularly, maximizing enterprise profits is the core objective at each decision point, while considering multiple factors such as market demand, enterprise supply and passenger buying preferences. Simulation experiments illustrate action sequences converge rapidly to stable points. By analyzing the convexity of the utility function, we demonstrate the existence of Nash equilibrium points. Our findings show that total revenue increases with higher passenger preferences, decreases with greater demand sensitivity, and increases with higher supply sensitivity. Therefore, this study offers novel insights into dynamic ticket pricing across the entire sales range and presents the solution for sequential game.