Learning Cheap and Novel Flight Itineraries
As a leading travel meta search engine, Skyscanner is dedicated to provide the best flight deals available on the Internet. Towards this goal, we consider the problem of efficiently constructing cheap and novel flight itineraries resulting from combining legs from different ticket providers. We analyze the factors that contribute towards the competitiveness of such itineraries and formulate the problem of predicting competitive itinerary combinations. We consider a variety of supervised learning approaches to model the proposed prediction problem and put forward a number of practical considerations for implementing them in production.