Penalty weight adjustment in cooperative GA for nurse scheduling

This paper describes a technique of penalty weight adjustment executed in the Cooperative Genetic Algorithm applied to the nurse scheduling problem. The nurse scheduling is very complex task, because many requirements must be considered. These requirements, or the constraints, are implemented by penalty functions in this research. In real hospital, several changes of the schedule often happen. Such changes of the shift schedule yields various inconveniences, for example, imbalance of the number of the holidays and the number of the attendance. Such inconvenience causes the fall of the nursing level of the nurse organization. Reoptimization of the schedule including the changes is very hard task and requires very long computing time. We consider that this problem is caused by the solution space having many local minima. We propose a technique to adjust penalty weight through the optimization to escape from the local minima. By means of the penalty adjustment, the optimization finishes in one-tenth computation time by the conventional technique.