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

Title:	Hybrid water cycle algorithm for attribute reduction problems
Authors:	Ahmed Sattar Jabbar (P62466)
Supervisor:	Suhaila Zainudin, Dr.
Keywords:	Hybrid water Data mining
Issue Date:	8-Jan-2014
Description:	This thesis is concerned with the attribute reduction problem that is a well-known stage in data mining process. Attribute reduction seeks for selecting a subset of attribute form the given set in the way that the information in selected subset is same as the whole set of attributes. Rough Set Theory (RST) has been used to generate a subset of attributes and then remove the redundant or irrelevant ones. However, due to size of the problem, which is considered as NP-hard problem and the greedy method used by the basic RST, a basic RST can be used to solve a small problem only. For this reasons, researchers have used heuristic and meta-heuristic algorithms to generate the subset of attribute within the RST. However, since different dataset has different number of attributes, no one method can be the best for all dataset and there is still a room for improvement to get better results. In this study, a new optimization approach, known as the Water Cycle Algorithm (WCA), has been used for attribute reduction where RST is employed as a mathematical tool to assess the quality of the solutions that are produced by the WCA. The idea of the WCA as an optimization algorithm was derived from nature, after examining the whole water cycle process which involves the flow of rivers and streams to the sea in the natural world. To further improve the performance of the WCA, WCA was hybridized with the hill climbing algorithm in order to improve the exploitative process of the WCA. Our aim is to provide effective approaches for finding the most informative and minimal attributes with least information loss. The proposed method is tested on the public domain datasets that are obtainable in UCI. From the results of the experiments, it has been shown that the proposed method performs equally well or even better than other methods that have been proposed in the literature for attribute selection problems, we obtained a new three best results compared to the literature, i.e. heart, vote and credit datasets.,Master/Sarjana
Pages:	122
Call Number:	QA76.9.D343J334 2014 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.