Genetic programming: on the programming of computers by means of natural selection
Genetic programming: on the programming of computers by means of natural selection
Proceedings of the 2008 annual research conference of the South African Institute of Computer Scientists and Information Technologists on IT research in developing countries: riding the wave of technology
Computers and Operations Research
HM '09 Proceedings of the 6th International Workshop on Hybrid Metaheuristics
Hi-index | 0.00 |
A hyper-heuristic for the examination timetabling problem searches a space of constructive heuristic combinations instead of a space of examination timetables. The most optimal heuristic combination found by the search is used to construct the examination timetable. The focus of a hyper-heuristic is to generalize well rather than producing the best result for one or more problem sets in the domain. A metaheuristic such as evolutionary algorithms is usually employed to explore the heuristic space. This study reports on an empirical investigation conducted to test how the structure of the heuristic combination affects the success of the search of an evolutionary algorithm (EA) hyper-heuristic for the uncapacitated examination timetabling problem. Two structures, namely, one that combines low-level construction heuristics linearly, and applies them sequentially and a second which combines heuristics hierarchically and applies them simultaneously are investigated. The performance of the EA-based hyper-heuristic using both structures is tested on a set of eight uncapacitated examination timetabling problems. The study has revealed that the representation used does have an impact on the success of the evolutionary algorithm. In this domain the linear combination and sequential application of heuristics produced better results. The EAs with both representations were also found to perform better than other hyper-heuristic methods applied to the same problem.