Hybrid Harmony Search Algorithm for Nurse Rostering Problem

This paper addresses the nurse rostering problem (NRP), whose objective is to minimize a total penalty caused by the roster. A large number of constraints required to be considered could cause a great difficulty of handling the NRP. A hybrid harmony search algorithm (HHSA) with a greedy local search is proposed to solve the NRP. A personal schedule is divided into several blocks, in which a subset of constraints is considered in advance. Based on these blocks, the pitch adjustment and randomization are carried out. Every time a roster is improvised, a coverage repairing procedure is applied to make the shift constraints satisfied, and the greedy local search is used to improve the roster's quality. The proposed HHAS was tested on many well known real-world problem instances and competitive solutions were obtained.