An adaptive coevolutionary memetic algorithm for examination timetabling problems

In this paper, we present an adaptive coevolutionary memetic algorithm (ACMA) for examination timetabling problems. In our proposed algorithm, the evolutionary search is conducted in two spaces: the heuristic space and the solution space. In the heuristic space, a hyper-heuristic approach is utilised to generate the initial population, and then basic evolutionary operators are used to find the potential heuristic lists. The evolutionary strategy in the heuristic space is regarded as a global search procedure. In the solution space, according to the solution structure, some specific evolutionary operators are designed for expanding the scope of search in solution space. This scheme can be viewed as the local search procedure. By combining two different strategies, the cooperation between them will eventually increase the diversities in the population. In order to determine which space should be selected at each generation, an adaptive parameter is designed based on the proportion of feasible solutions in the current population. Experimental results demonstrated that ACMA obtains competitive results and outperforms the compared approaches on some benchmark instances.