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

Title:	K-means with modified simulated annealing and great deluge algorithms for clustering problems
Authors:	Kittaneh Rawnaq Raji Nemer. (P53656)
Supervisor:	Salwani Abdullah, Prof.
Keywords:	Algorithms. Data mining
Issue Date:	10-Jun-2011
Description:	Clustering problem is one of type of optimization problems, which is considered as a critical area of Data Mining, as it depends on unsupervised mechanisms. Clustering refers to the problem of selecting input features that are most predictive of a given outcome. These problems are encountered in many areas, such as machine learning, pattern recognition, and signal processing. In the recent years clustering medical problems has attracted the attention of the Operational Research and Artificial Intelligence communities. A wide variety of existing approaches for constructing the clusters have been reviewed described and discussed in the literature. In this thesis, we have developed meta-heuristic approaches i.e. simulated annealing and great deluge algorithms over K-Means clustering, to identify the common diagnosis of a specific medical disease. Three objective functions are considered in this thesis such as minimizing intra cluster distance “between objects”, maximizing inter distance “between centres” and to minimizing the intra and maximizing the inter distance between clusters “combination between objects and centres”. The K-Means algorithm is employed as a constructive algorithm to obtain an initial solution. The initial solution was further improved in the improvement stage, where three variations on improvement approaches have been introduced namely (i) iterative simulated annealing, (ii) iterative great deluge, and (iii) hybridization between iterative simulated annealing with iterative great deluge algorithms. The proposed approaches have been tested on six benchmark medical datasets that are available in UCI Machine Learning Repository, and evaluated these approaches over other standard benchmark datasets. Experimental results show that the proposed methods are capable of obtaining high quality results,Master
Pages:	88
Call Number:	QA9.58 .K538 2011
Publisher:	UKM, Bangi
URI:	https://ptsldigital.ukm.my/jspui/handle/123456789/476320
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.