[Back]


Diploma and Master Theses (authored and supervised):

Y. Darmaputra:
"An Application of Heuristic Route Search Techniques for a Scalable Flight Search System";
Supervisor: R. Pichler, R. Baumgartner; Institut für Informationssysteme, Arbeitsbereich Datenbanken & Artificial Intelligence, 2008; final examination: 06-2008.



English abstract:
Most of the fiight search engines that are currently available in the Internet can be considered as meta-search engine. They forward the route query to other websites such as online travel agents, major airlines, and low-cost carriers. The result from each source is collected and aggrcgated before presented to user. There arc several weaknesses for this approach.
First, not all available routes can be found. Take low-cost airlines (which empha.size on direct flights) as example. If the airline doesn`t have a direct flight from X to Y, then trying to seareh for route from X to Y, would give no result, although it has route from X to Z and Z to Y. Second, they can not mix fiights between airlines in different alliances. An airline would never promote fiights from other airline, except if the other airline is in the Same alliance or if there exists some codeshare agreements.
In this thesis we propose a mashup solution for this problem. A mashup application does not own the data. lt uses data from other resources (called content provider) to create a new application with new feature and functionality that is not offered by any of the content provider. The web data extraction technology from Lixto Visual Developer is used to wrap data.
The fiight search problem is treated as graph search problem with airports as the nodes and the pair of airports where exists direct flight between them as the edges. We consider the scalability of the current fiight search engines by introducing hub identification heuristic. Instead of analyzing and evaluating all possible routes to reach the destination, this heuristic gives hint on which hub airports that are possibly containing the best routes (in tenns of shortest flight duration). Hence, the system can limit the search by only considering a fraction from all possible routes. Performing the search this way also ensures system scalability and increases the system`s responsiveness by shortening the query procesing time.
The term interesting route is defined as a list of routes which match with the user`s preference. An interesting route for one may not be interesting for the other. Therefore, three sorting criteria arc used for evaluating and ranking the search results. The search is also very customizable, for example user can choose airline preferences (e.g. only searching from low-cost airlines), transit time preference (e.g. transits between 1-3 hours), which airports/routes to be chosen for transits/stops (e.g. avoid non-Schengen airports), and searching for fiights from dose airports (e.g. include other airports within particular distance from the original departure airport).

Created from the Publication Database of the Vienna University of Technology.