An Order Based Evolutionary Approach to Dual Objective Examination Timetabling

This paper explores a simple bi-objective evolutionary approach to the examination timetabling problem. The new algorithm handles two hard constraints: 1) avoiding examination clashes and 2) respecting the given maximum seating capacity; while simultaneously minimizing two objective functions: 1) the overall length of the examination period and 2) the total proximity cost An order based representation with a greedy decoder ensures that neither of the hard constraints is violated, and produces only feasible timetables. At the same time the dual objectives are attacked and the multi-objective evolutionary algorithm (MOEA) attempts to pack all the examinations into as short a period as possible while, at the same time, favoring a good spread of examinations for individual students. Most other published timetabling algorithms require the number time slots to be fixed in advance of any optimization for soft constraints, such as proximity costs. Smart genetic and heuristic operators used in the present study ensure that a good set of non-dominated results is produced by the new MOEA, covering a range of timetable lengths