Development of nurse scheduling support system using interactive EA

The nurse scheduling problem (NSP) is the task of allocating nurses in three shifts under constraints. A nurse-in-chief needs a lot of time and effort to make a schedule. Studies on automatic scheduling have been done. A genetic algorithm (GA) has been applied to this nurse scheduling problem. However, this type of conventional scheduling method could not satisfy the user. It has been difficult to incorporate the evaluation criteria of the nurse-in-chief. An interactive GA has been studied. In the process of this GA, humans directly evaluate the offspring. The interactive GA has been applied to the fields of arts and Kansei engineering. The paper presents a nurse scheduling support system utilizing an interactive evolutionary algorithm (EA). From the past schedule, and users operation, the system learns the user's evaluation criteria of the system, and generates schedules under the obtained criteria, and displays some of the best schedules to the user. Then she can decide which parts of the schedules to adopt, or to be improved. The system reschedules the parts that need improvement. Repeating this process, the system and the user help each other to find a satisfactory schedule. The paper also presents a hardware implementation of EA for this interactive nurse scheduling system. For human-machine interaction, quick response of the system is indispensable.