Psychological Preference-based Optimization Framework with Interactive Genetic Algorithms on the Nurse Scheduling Problem

One of the challenges of the nurse scheduling problems (NSPs) is the great number of constraints, and another challenge is that psychological preference of chief nurse is usually difficult to clearly define. Typically, solutions to conquer the first challenge rely on operational research (OR) methods, and the objective functions are constructed manually to conquer the second challenge. However, OR methods are usually time consuming and it is too strong an assumption that the constructed objective function is close to the psychological preference in chief nurses’ minds. This paper presents a psychological preference-based optimization framework (PPOF) with a guidable hill-climbing decoder (HCD) and the active interactive genetic algorithm (AIGA). The guidable HCD yields satisfactory performance in speed, and it reduces the highly-constrained search space to a constraint-less one which is easy for genetic algorithms to optimize. To meet the psychological preference, PPOF adopts AIGA to tune HCD to approximate the chief nurses’ psychological preference. The experiment shows that PPOF is able to arrange a feasible monthly schedule of a realistic NSP of the National Taiwan University Hospital within several seconds.