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

Title:	Fuzzy based meta-heuristic approaches for attribute reduction problem
Authors:	Majdi M.N. Mafarja (P51681)
Supervisor:	Salwani Abdullah, Assoc. Prof. Dr.
Keywords:	Soft computing Computer science Fuzzy logic Heuristic algorithms Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia
Issue Date:	3-Sep-2012
Description:	Attribute reduction represents a NP-hard problem that can be defined as the problem of locating a minimal subset of attributes from an original set. The key issue associated with feature selectors is the production of a minimal number of reducts that represents the original meaning of all features. Rough Set Theory has been used for attribute reduction with much success. This is due to the fact that rough set theory uses only the supplied data during the feature selection process and no more information is needed. The reduction method inside rough set theory is applicable only to small data sets because finding all possible reducts is a time-consuming process. However, no approach can ensure optimality when solving this problem. Some approaches are more efficient than others due to some of the characteristics of the algorithm such as the number of parameters involved. The aim of the research presented in this thesis is to provide effective approaches for finding the most informative and minimal attributes with least information loss. This has been achieved via a number of meta-heuristic approaches which mainly depend on two algorithms, i.e., the Record-to-Record Travel algorithm and the Great Deluge algorithm. Both algorithms are deterministic optimisation algorithms whose structures are inspired by and resemble the Simulated Annealing algorithm but differ in the acceptance of worse solutions. Moreover, they belong to the same family of meta-heuristic algorithms that are used to avoid the local optima by accepting non-improving neighbours. The research first highlights the use of the record-to-record travel algorithm in solving attribute reduction problem, and then examines the effects of enhancing the algorithm by incorporating a Fuzzy Logic Controller in order to intelligently control the parameter involved in the algorithm (called the Fuzzy Record-to-Record Travel algorithm). Next, two modifications of the Great Deluge algorithm are investigated, where the search space is divided into three regions. Instead of using a linear mechanism to update the water level (as in the original Great Deluge algorithm), the modified Great Deluge algorithm updates the water level for each region using a different scheme which is based on the quality of the trial solution. Then, a fuzzy logic controller is used to control the updated scheme of the water level (called the Fuzzy Great Deluge algorithm). This research further investigates the efficacy of the hybridisation approach between the aforementioned algorithms with the Genetic Algorithm (called Fuzzy Record-to-Record Travel with Genetic Algorithm and Fuzzy Great Deluge with Genetic Algorithm). Experimental results show that the fuzzy Great Deluge with Genetic Algorithm approach outperforms the other proposed approaches here and is effective for most of the University of California Irvine benchmark data sets when compared to other available approaches in the literature.,Certification of Master's / Doctoral Thesis" is not available
Pages:	165
Call Number:	QA76.9.A43M336 2012 3 tesis
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.