A Hyperheuristic Approach to Belgian Nurse Rostering Problems

The nurse rostering problem involves the assignment of shifts to nurses with respect to several constraints like workload, legal restrictions, and contractual agreements [9]. The complexity of the problem attracts many researchers around the world. Since the regulations, requirements, and agreements differ from country to country, the models and solution methods differ as well [9]. The solution methods to tackle the nurse rostering problems are as diverse as the problem definitions. Estimation of distribution algorithm [1], constraint programming [6], case-based reasoning [2], and hybrid meta-heuristic methods [8] are examples of solution methods that cope with the nurse rostering problem. The success of metaheuristic methods depends on the weights of the objective function. Parr and Thompson propose a method for setting weights in an objective function for the nurse rostering problem in [15]. The nurse rostering problem is not the only scheduling problem encountered in hospitals. Beliën and Demeulemeester integrate nurse rostering and surgery scheduling in a single problem definition and address the resulting problem with a branch-andprice approach [3]. Gendreau et al. tackle the physician scheduling in emergency rooms in Canada using mathematical programming, tabu search, constraint programming and column generation [11]. This study is focused on the nurse rostering problems in Belgium. The instances of nurse rostering problems in Belgium are very diverse. Problem properties like workload, number of nurses, schedule periods, and contractual agreements vary among different institutions as well as among the wards within the same institution. Bilgin et al. propose