Dynamic air ticket pricing using reinforcement learning method

This paper studies a dynamic air ticket pricing problem in strategic and myopic passengers co-existence market. The or can be further divided into high-valuation low-valuation groups according to how they evaluate their purchases. have different levels. When the airline sets price, every passenger makes his her purchase decision type level, might select “wait” “leave (the market)”. firstly proposes algorithm which utilities of both are considered. reinforcement learning (RL) is employed deal with progressive decision-making framework, formulated as discrete finite Markov process (MDP) Q-learning adopted solve problem. By using this method, adaptively decide price based on behaviors time-varying demand. effects proportion level analyzed. computational results show higher is, smaller increase adopt, larger put use under same level. If higher, should gentle step by when strategy adopted. uses price-cut policy, adjustment small. In addition, mainly affects high-price periods low-price periods. fixed, lower slope is. These findings provide some references for make more precise flexible decisions.