A particle swarm optimization approach with refinement procedure for nurse rostering problem

This study attempts to develop a model satisfying the rules of a typical hospital environment based both on published research data and on requirements of a local hospital under study. A mathematical formulation for the studied nurse rostering problem (NRP) is presented first. Due to the combinatorial nature of the NRP model, a particle swarm optimization (PSO) approach is proposed to solve this highly complicated NRP. The structure of the problem constraints is analyzed and used as base for generating workstretch patterns. These patterns serve as the base for generating fast initial solutions, and will later be improved upon by the proposed PSO algorithm. This study also proposes a simple yet effective procedure for attempting possible refinements on the solutions obtained by the PSO before reporting the final solutions. When fair shift assignment is considered as the decision objective, computational results show that the proposed PSO algorithm with refinement procedure is able to produce optimal solutions in all real test problems in a very efficient manner.