A New Initialisation Method for Examination Timetabling Heuristics

Abstract

Timetabling problems are widespread, but are particularly prevalent in the educational domain. When sufficiently large, these are often only effectively tackled by timetabling meta-heuristics. The effectiveness of these in turn are often largely dependant on their initialisation protocols. There are a number of different initialisation approaches used in the literature for starting examination timetabling heuristics. We present a new iterative initialisation algorithm here --- which attempts to generate high-quality and legal solutions, to feed into a heuristic optimiser. The proposed approach is empirically verified on the ITC 2007 and Yeditepe benchmark sets. It is compared to popular initialisation approaches commonly employed in exam timetabling heuristics: the largest degree, largest weighted degree, largest enrollment, and saturation degree graph-colouring approaches, and random schedule allocation. The effectiveness of these approaches are also compared via incorporation in an exemplar evolutionary algorithm. The results show that the proposed method is capable of producing feasible solutions for all instances, with better quality and diversity compared to the alternative methods. It also leads to improved optimiser performance.