Variable neighborhood search algorithm for nurse rostering problem

Abstract

Variable neighborhood search algorithm is effective for the nurse rostering,and the perturbation method used in it has significant effect on its performance.In order to improve the satisfaction of nurses in the nurse rostering problem,an Improved Variable Neighborhood Search(IVNS) algorithm was proposed.The algorithm used three neighborhood structures,when using any neighborhood could not improve the current solution further,a method for perturbing the current optimal solution was designed:firstly,two days in the rostering period were randomly selected,then a group of nurses were selected and their shifts in these two days were exchanged under the restriction of hard constraints.Comparison experiments with a Hybrid Variable Neighborhood Search(HVNS) algorithm were carried out on the benchmarks provided by the first international nurse rostering competition in 2010,and the results in the Sprint-early,Medium-early and Long-early instance groups show that,the optimal value of the IVNS algorithm is not inferior to HVNS at least,and its average value is superior to HVNS;the maximum variance of IVNS algorithm is 0.72,which means the fluctuation range is small,and the solution performance is stable.The IVNS disturbance program makes small disturbance to the existing project,and the current local optimal value can effectively jump out,enhancing the optimization ability of variable neighborhood search algorithm.Compared with HVNS algorithm,its performance is better.