Iterated local search in nurse rostering problem

Abstract

This paper presents how to solve a nurse rostering problem over the real datasets of Centre hospitalier régional de Trois-Rivières hospital in Canada. Due to the complexity of this problem with plenty of hard constraints, we propose an advanced Iterated Local Search, combining Tabu Search with 2 moves: Single Shift Move and Worst-Scheduled Nurse Swap. Greedy Shuffling with Steepest Descent is also used to improve the solution. Experimental results of our proposed algorithm on 5 real datasets improve the current schedules provided by the hospital. Our experimental results satisfy all of the hard constraints and objectives.