Hyper-heuristics with a dynamic heuristic set for the home care scheduling problem

A hyper-heuristic performs search over a set of other search mechanisms. During the search, it does not require any problem-dependent data. This structure makes hyper-heuristics problem-independent indirect search mechanisms. In this study, we propose a learning strategy to explore elite heuristic subsets for different phases of a search. For that purpose, we apply a number of hyper-heuristics with the proposed approach to a set of home care scheduling problem instances. The results show that the learning strategy increases the performance of the different hyper-heuristics by excluding some heuristics from the heuristic set over the tested problem instances.