Using μ Genetic Algorithms for Hyper-heuristic Development: A Preliminary Study on Bin Packing Problems

Hyper-heuristics have gained attention from the community in the last decade, given their capability to deal with complex combinatorial problems. Although they do not guarantee optimality, they provide a mechanism to benefit from the strengths of different heuristics and obtain good-quality solutions. So, by selectively applying such heuristics, the search is optimized. Although evolutionary algorithms have successfully been used to generate hyper-heuristics for various problems, the number of objective function evaluations is always an issue. We propose using a
$\mu$
Genetic Algorithm to address this severe criticism when producing hyper-heuristics. We tested our approach on the one-dimensional Bin Packing Problem, and the results suggest that this approach is competitive concerning the heuristics applied in isolation.