Hybrid Harmony Search for Nurse Rostering Problems

Abstract

In this paper, a Hybrid Harmony Search Algorithm (HHSA) is presented for Nurse Rostering Problem (NRP) using the dataset proposed by the First International Nurse Rostering Competition (INRC2010). NRP is tackled by assigning daily shifts to nurses with different skills and working contracts, subject to hard and soft constraints. Harmony Search Algorithm (HSA) is a recent evolutionary computing technique, mimicking the musical improvisation process where a group of musicians play the pitches of their musical instruments. Recently, HSA has been used for NRP, with promising results. This paper extends HSA to HHSA by adding two powerful concepts to HSA: (i) hybridization with hill climbing optimizer to improve the exploitation ability, and (ii) hybridization with global-best concept of particle swarm optimization to improve the speed of convergence. The proposed HHSA is evaluated against a dataset provided by INRC2010. The results show that it is a powerful technique for INRC2010 dataset. A comparative analysis with five competitive methods is conducted. HHSA outperforms the other competitive methods in three instances and obtained the best results in 29 others out of 69 instances. The efficiency of our method lends further support to the previous theory based on hybridizing the local search within evolutionary computing technique for hard combinatorial optimization problems.