|PALADINO STEFANO||Cycle: XXX |
Section: Computer Science and Engineering
Tutor: AMIGONI FRANCESCO Abstract:
Dynamic pricing of flight tickets with online learning techniques
The revenues of airline companies comes mainly from the margin they get selling flight tickets. Therefore, airline companies try to set the optimal price to obtain the maximum revenue from each sold flight. If they set too low prices, they could lose the revenue from customer who would be willing to pay for higher prices. On the other hand, if they set too high prices, customers could buy tickets at lower prices from other competitor companies. In these scenarios, we study the problem of finding the best pricing policies for flight tickets using game theory and machine learning techniques. This task is very challenging, due to the high competition of the different companies and the huge amount of data that needs to be processed and evaluated. We focus on a setting affected by lack of information, so it is necessary to use automatic learning techniques in order to explore pricing configurations whose economic return is unknown. We design new algorithms and we empirically evaluate them to study their computational complexity and to derive their asymptotic properties in terms of convergence to solutions. First we implement our algorithms as prototypes and we evaluate them using the real data of one of the major online travel agency (OTA) in Europe. Then we integrate our prototypes within the real pricing system of the OTA and we monitor the performance of the system.
Advisor: Nicola Gatti