Learning Dynamic Pricing Rules for Flight Tickets

摘要

It is possible and necessary to adjust the flight ticket prices for each airlines dynamically in order to increase online travel agencies' revenues. Unfortunately, the demands and the availability of flight tickets change following very complex patterns so that it is very hard, if not impossible, to adopt mathematical models to describe them and to derive analytical solutions. We apply reinforcement learning approach to learn dynamic pricing rules from a passenger simulator which can generate passengers' responses according to flight tickets' prices. In order to make passenger simulator more realistic, it adjusts it's inherent models based on historical data and up-to-date data continuously. The experimental results on a real-world data set show that our approach can learn dynamic pricing rules efficiently.