The competition is divided into three tracks. There will be three data sets for each track: the early set, the late set, and the hidden set. Competitors can enter one, two or all of the tracks if they wish to. There will be two tracks on course timetabling and one on examination timetabling. 'Early' Datasets will be released at the beginning of the competition with a further 'late' set, being released two weeks before the deadline. Hidden datasets will also be used for each track to judge who the finalists and eventual winners should be. For information related to formulation of the problem, dataset formats and method of evaluation, please refer to the dedicated web pages.
Datasets will be provided by the eventMAP research Group. Four early datasets will be provided initially with a further four late datasets released prior to the competition end date. Four hidden datasets will be held back for validation purposes.
Track1: Examination Timetabling
This is divided into two tracks. Both of these tracks are distinct and represent methods of course timetable construction which are used to varying degrees within institutions. One aim of the competition is to combine both of the general models associated with these tracks into one ‘course timetabling’ model.
After student enrolment, the timetable is constructed in such a way that all students can attend the events on which they are enrolled. This approach, which formed the basis of the 2003 competition to course timetabling, has been given various names in the literature including the Class Timetabling Problem and Event Timetabling.
This track is an extension of the problem model used in the first competition with extra constraints to move further in the direction of real-world timetabling. These extra constraints make finding feasibility more difficult, this means there will be a shift of emphasis from soft constraints to hard constraints.
Track 3: Curriculum based Course Timetabling
The Curriculum-based timetabling problem consists of the weekly scheduling of the lectures for several university courses within a given number of rooms and time periods, where conflicts between courses are set according to the curricula published by the University and not on the basis of enrolment data.
Real datasets will be used from the university of Udine, Italy.
Last Updated: Wednesday, October 1, 2008 10:16 AM