1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Enhanced and hybridized harmony search algorithm for nurse rostering problems
Authors:	Mohammed Abdo Saleh Hadwan (P39774)
Supervisor:	Masri Ayob, Assoc. Prof. Dr.
Keywords:	Enhanced harmony search algorithm Hybridized harmony search algorithm Nurse rostering problems Computer algorithms
Issue Date:	12-Jan-2012
Description:	Due to the worldwide shortage in nursing workforce, building up the nurses' duty roster is one of the main issues that are faced by hospital management. High workloads, inflexible schedules and unfair distribution of duty shift slots are among the key issues that cause nurses' job dissatisfaction. Hence, this work focuses on nurse rostering problem (NRP), an NP-hard problem, which is difficult to solve to optimality. Our research aims at adapting the Harmony Search Algorithm (HSA) for NRP, utilizing its advantages and overcoming its limitation by hybridizing it with other metaheuristic algorithms. HSA has proved its ability to solve difficult optimization problems such as university course timetabling, vehicle routing and other optimization problems. However, not many work reported in applying HSA to solve NRP. Therefore, we first apply a classical HSA for NRP and investigate the suitable parameter settings for HSA. Although promising results have been obtained, slow convergence is noticed. To overcome this weakness, we enhanced the harmony search algorithm (EHSA) by employing a semi-cyclic shift pattern approach to construct the harmony memory instead of the random mechanism in classical HSA. Furthermore, instead of employing fixed values for Harmony Memory Considering Rate (HMCR) and Pitch Adjustment Rate (PAR), EHSA utilizes a dynamic HMCR and PAR. EHSA produces better results than classical HSA in terms of time and solutions quality. In order to further improve the exploitation mechanism in EHSA, we hybridize EHSA with three local search-based algorithms. These algorithms are: Hill Climbing (HC), Simulated Annealing (SA) and Great Deluge (GD). The results reveal that EHSA with SA (EHSA-SA) yield the best results in all the datasets compared to EHSA-HC and EHSA-GD. As a comparative study, the proposed algorithms have been compared to classical genetic algorithm and two other variants of HSA from the scientific literature, namely, adaptive harmony search and harmony annealing search. These approaches are tested on real-world NRP datasets that we have collected from Universiti Kebangsaan Malaysia Medical Centre (UKMMC) and also against benchmark datasets. Results demonstrate that the combination of EHSA with SA outperforms other methods tested on these datasets.,PhD
Pages:	206
Call Number:	QA76.9.A43 .H33 2012
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.