Examination Timetabling: A New Formulation

Since the mid 1990’s, with the implementation of increasingly flexible modular course structures in many UK Universities, the central production and coordination of the associated examination timetable has become increasingly difficult with more examination offerings having to be timetabled in such a manner as to provide students with a maximum distribution of their exams throughout the examination session. Of course, we must also ensure that time and resources usage maximised. Universities, struggling with rising student numbers, more flexibility in choice and less time to examine, have increasingly relied upon automation of this task to produce efficient timetables which satisfy these constraints e.g. [1,2]. Although strong in some respects, unfortunately, many of the search methodologies currently described in the literature have some limitations in terms of potential application in a wide number of differing institutions. The examination timetabling problem has long represented an area where new and exciting techniques have been trialed at an early stage of their development. This, in large part, is related to the inherently straightforward nature of the timetabling problem which can be expressed by the following. Students are examined over a designated time period, within a finite area of space, in such a way as to ensure they do not have two exams at the same time. This ‘hard’ constraint must be satisfied for a solution to be viable. The quality of the overall solution is measured by factors such as how well an individual’s exams are distributed throughout the designated time period e.g. soft constraints. Both type of constraints, hard and soft, were documented in some detail for UK Universities in 1996 [1]. All subsequent research has been trialed on datasets which have been in existence from the middle part of that decade. The initiative described in this presentation sets out to update the situation with regard to the examination timetabling problem by investigating the changes in the problem along with providing new updated datasets for techniques to be subsequently trialed. It is well reported that there is a gap between theory and practice in scheduling research (eg [3]). A major contributor to the work presented in this abstract is our spin out company, EventMAP Limited. From a practitioner’s point of view, the company has reported the steady increase in complexity of the examination problem over the last five years.