Hyper-heuristics applied to class and exam timetabling problems

Abstract

Combinatorial optimisation algorithms can be both slow and fragile. That is, the quality of results produced can vary considerably with the problem and with the parameters chosen and the user must hope or the best or search for problem-specific good parameters. The idea of hyper-heuristics is to search for a good, fast, deterministic algorithm built from easily-understood heuristics that shows good performance across a range of problems. In this paper we show how the idea can be applied to class and exam timetabling problems and report results on nontrivial problems. Unlike many optimisation algorithms, the generated algorithm does not involve and solution-improving search step, it is purely constructive.