Hyperheuristics form a new concept that provides a more general procedure for optimization. Their goal is to manage existing low-level heuristics to solve a large number of problems without specific parameter tuning. A classification was proposed by Burke et al. divides the hyperheuristics into two classes: Heuristic selection which is the strategy for choosing or selecting heuristics and heuristic generation which is an approach for generating new heuristics from components of existing heuristics. The two classes operate with both on construction and perturbation heuristics (Burke et al. 2010).
In our work, we are interested to the generation hyperheuristic. As we mentioned before, generation hyperheuristic generates new heuristics based on components of existing heuristics. Therefore, we use the genetic programming (GP) as a technic to generate new heuristics. As the genetic programming uses the tree representation, we represent our heuristic as a tree, its nodes are simple operators and the leaves represent functions specific to the problem.
For our first contribution, we chose as a problem domain to test our hyperheuristic a scheduling problem which is the permutation flow shop problem. We integrated the simulated annealing approach in order to improve the model.
For our second contribution, we will treat another problem domain using the same components of our first model. In other word, we will adapt our algorithm to treat the Patient Admission Scheduling problem.
Our final challenge is to adapt our model in order to deal with multi-objective problems.