Harmony Search with Greedy Shuffle for Nurse Rostering

In this paper, a hybridization of Harmony Search Algorithm (HSA) with a greedy shuffle move is proposed for Nurse Rostering Problem (NRP). NRP is a combinatorial optimization problem that is tackled by assigning a set of nurses with different skills and contracts to different types of shifts, over a pre-determined scheduling period. HSA is a population-based method which mimics the improvisation process that has been successfully applied for a wide range of optimization problems. The performance of HSA is enhanced by hybridizing it with a greedy shuffle move. The proposed method is evaluated using a dataset defined in first International Nurse Rostering Competition (INRC2010). The hybrid HSA obtained the best results of the comparative methods in four datasets. DOI: 10.4018/jncr.2012040102 International Journal of Natural Computing Research, 3(2), 22-42, April-June 2012 23 Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited. Genetic Algorithm (Cai & Li, 2000; Tsai & Li, 2009), Ant Colony Optimization (Gutjahr & Rauner, 2007), Electromagnetic (Maenhout & Vanhoucke, 2007), Tabu Search (Burke, De Causmaecker, & Vanden Berghe, 1999; Dowsland, 1998), Simulated Annealing (Brusco & Jacobs, 1995), Variable Neighbourhood Search (Bilgin, De Causmaecker, Rossie, & Vanden Berghe, 2011; Burke, Curtois, Post, Qu, & Veltman, 2008), and Scatter Search (Burke, Curtois, Qu, & Berghe, 2009). For more details see the surveys of Burke, De Causmaecker, Berghe, and Van Landeghem, (2004), and Cheang, Li, Lim, and Rodrigues (2003). Burke et al. (1999) hybridized the tabu search with a greedy shuffling move to solve NRP. The hybrid tabu search was tested using real dataset sampled from Belgian hospitals. The proposed method was better than the manual methods in terms of solution quality and computational time. Burke et al. (2001) hybridized the tabu search optimizer with genetic algorithm for NRP. The performance of the hybrid algorithm was compared against tabu search using real dataset sampled from Belgian hospitals. In comparison, the hybrid genetic algorithm provided the best results. In another study, Bellanti et al. (2004) separately applied two methods including tabu search and iterated local search for NRP. The performance of their methods was measured using real dataset from an intensive care unit of a hospital in Turin, Italy. Özcan (2005) hybridized the hill climbing optimizer within the genetic algorithm to tackle NRP. The author obtained satisfactory results using real dataset from the Memorial Hospital, Turkey. Gutjahr and Rauner (2007) investigated the applicability of ant colony optimization for NRP using dataset from Vienna Hospital, Austria. The ant colony optimization obtained high quality solutions in a reasonable computational time in comparison with simulated annealing method. Burke et al. (2008) hybridized the variable neighbourhood search with heuristic ordering technique to tackle NRP. The results of hybrid algorithm outperformed commercial genetic algorithm software applied over 40 hospitals in Belgium. An efficient genetic algorithm was presented by Tsai and Li (2009) for NRP. Their method obtained suitable results when tested using a dataset from the Otolaryngology Hospital in Taiwan. Burke et al. (2009) hybridized the scatter search with hill climbing to solve NRP. The performance of the hybrid method is evaluated on real-world benchmark dataset, called ‘BCV,’ sampled from Belgian hospitals. Their method is compared with the previous methods using ‘BCV.’ They concluded that their hybrid method is robust and effective on a wide range of real-world dataset. Based on the above, it is difficult to make comparison between the different methods used previously in tackling NRP, because each method was implemented for specific hospital constraints. These constraints are varied in number or type from hospital to hospital and region to region. In order to make comparison between these methods, the researchers in the domain needs a public standard dataset for NRP that was made available in 2010. This dataset was introduced during the first International Nurse Rostering Competition (INRC2010) for the researchers to compare their methods. This competition was organized by CODeS research group at Katholieke Universiteit Leuven in Belgium, SINTEF Group in Norway and the University of Udine in Italy. The INRC2010 dataset was divided into three tracks: sprint, medium, and long datasets which are different in complexity and size. Each track is categorized into four types in accordance with the publication time in the competition: early, late, hidden, and hint. Briefly, the methods used in tackling the INRC2010 dataset during the competition are described as follows: Valouxis et al. (2010) used Integer Programming to tackle INRC2010 dataset. The solution method includes two steps for all tracks. Firstly, assigning different nurses to working days, while the second step schedules the nurses assigned to working days and certain shifts. Furthermore, the authors used three additional neighbourhood structures in the first step for medium and long track datasets. These neighbourhood structures 19 more pages are available in the full version of this document, which may be purchased using the "Add to Cart" button on the product's webpage: www.igi-global.com/article/harmony-search-greedy-shufflenurse/73012?camid=4v1 This title is available in InfoSci-Journals, InfoSci-Journal Disciplines Medicine, Healthcare, and Life Science. Recommend this product to your librarian: www.igi-global.com/e-resources/libraryrecommendation/?id=2