Hyper-heuristics for grouping problems

Hyper-heuristics emerge as domain independent methodologies to solve hard computational search problems by performing search over the heuristics rather than directly solutions. One of the main goals of hyper-heuristic research is to support and investigate into the development of more general approaches applicable across different problem domains. Grouping problems requires partitioning of a set of items into mutually disjoint subsets subject to constraints. In this study, high level selection hyper-heuristics are investigated embedding a set of low level heuristics for grouping problems based on an efficient representation, referred to as linear linkage encoding. The empirical results over multi-objective and single objective grouping problems, such as graph coloring, examination timetabling, data clustering and bin packing show that the proposed grouping hyper-heuristic framework is sufficiently general providing high quality solutions at each domain.