Case-Based Initialisation of Metaheuristics for Examination Timetabling

Abstract

Examination timetabling problems are traditionally solved by choosing a solu-tion procedure from a plethora of heuristic algorithms based either on a direct construction principle or on some incremental improvement procedure. A num-ber of hybrid approaches have also been examined in which a sequential heuris-tic and a metaheuristic are employed successively. As a rule, best results for a problem instance are obtained by implementing heuristics with domain-specific knowledge. However, solutions of this kind are not easily adoptable across dif-ferent problem classes. In order to lessen the need for a problem-specific knowl-edge we developed a novel solution approach to examination timetabling by in-corporating the case-based reasoning methodology. A solution to a given prob-lem is constructed by implementing case-based reasoning to select a sequential heuristic, which produces a good initial solution for the Great Deluge meta-heuristic. A series of computational experiments on benchmark problems were conducted which subsequently demonstrate that this approach gives compara-ble or better results than solutions generated not only by a single Great Deluge algorithm, but also the state-of-the-art approaches.