Properties of Yeditepe Examination Timetabling Benchmark Instances

Examination timetabling is a type of educational timetabling which is a highly challenging field for the researchers and practitioners. Examination timetabling problems require a search for the best assignment of examinations into a fixed number of time-slots possibly along with other resources, such as, a set of rooms with certain capacities, subject to a set of constraints. There are two common types of constraints: hard and soft. The hard constraints must not be violated, while the soft constraints represent preferences that can be infringed. Examination timetabling problems are proven to be NP-complete (Even et al 1976). A recent survey on exam examination timetabling can be found in Qu et al (2009). There are many variants of examination timetabling problems due to the fact that each educational institution have their own rules, regulations and expectations resulting with various constraints. This situation also makes it extremely difficult to compare different solution methods. Not only comparability but also reproducibility of the results is vital for the research community, as pointed out in Schaerf and Gaspero (2006). McCollum (2006) discusses real world issues in examination and course timetabling. Although practitioners and researchers have to deal with different aspects of examination timetabling, it has been always of interest for both communities to design robust and flexible approaches that can solve new problem instances. ITC2007 (http://www.cs.qub.ac.uk/itc2007/) competition is organised considering the real world examination timetabling complexities and capturing them within the problem instances. The state of the art method for examination timetabling turned out to be a hybrid multistage approach combining Iterated Forward Search (IFS) for feasible initial solution construction and great deluge for improvement as described in Müller (2009). The source code of the solver is available from http://www.unitime.org/itc2007. Yeditepe University (Faculty of Engineering) data set contains real problem instances from a total of eight semesters in three consecutive years. Bilgin et al (2007) modified the initial data set provided in Özcan and Ersoy (2005) with new properties and also generated a variant of Toronto benchmarks (Carter et al 1996) that fits into the problem formulation which will be referred to as modified Toronto benchmark. This problem is a capacitated variant of examination timetabling. There is a maximum capacity of seating available during