An approach of solving itinerary construction problems using real life data

This paper investigates a computational intelligence approach to solving the itinerary construction problem in airline scheduling using real booking data. In our case, the result of itinerary construction is defined as an optimal set of direct and connecting flight connections of different airlines between given origin and destination airports. A set of rules restricts the number of possible flight connections treating restrictions in time, location and passenger commodity. In order to increase the variety of possible sets of flights using various itineraries we propose a way of analyzing numerous data of real passenger bookings to improve the itinerary construction generation. Therefore, heuristics are implemented on a Java open-source platform combining a particular connection building routine, an adapted hill-climbing algorithm for meta-optimization of the parameters of the heuristics, and a connection to a database containing the Marketing Information Data Tape (MIDT) of real booking data of airlines. The feasibility of this approach is shown; the quality of the results is proven.