View : 15 Download: 0

Heterogeneous clustering ensemble method for combining different cluster results

Title
Heterogeneous clustering ensemble method for combining different cluster results
Authors
Yoon H.-S.Ahn S.-Y.Lee S.-H.Cho S.-B.Kim J.H.
Ewha Authors
이상호
SCOPUS Author ID
이상호scopus
Issue Date
2006
Journal Title
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN
0302-9743JCR Link
Citation
vol. 3916 LNBI, pp. 82 - 92
Indexed
SCOPUS WOS scopus
Abstract
Biological data set sizes have been growing rapidly with the technological advances that have occurred in bioinformatics. Data mining techniques have been used extensively as approaches to detect interesting patterns in large databases. In bioinformatics, clustering algorithm technique for data mining can be applied to find underlying genetic and biological interactions, without considering prior information from datasets. However, many clustering algorithms are practically available, and different clustering algorithms may generate dissimilar clustering results due to bio-data characteristics and experimental assumptions. In this paper, we propose a novel heterogeneous clustering ensemble scheme that uses a genetic algorithm to generate high quality and robust clustering results with characteristics of bio-data. The proposed method combines results of various clustering algorithms and crossover operation of genetic algorithm, and is founded on the concept of using the evolutionary processes to select the most commonly-inherited characteristics. Our framework proved to be available on real data set and the optimal clustering results generated by means of our proposed method are detailed in this paper. Experimental results demonstrate that the proposed method yields better clustering results than applying a single best clustering algorithm. © Springer-Verlag Berlin Heidelberg 2006.
DOI
10.1007/11691730_9
ISBN
3540331042; 9783540331049
Appears in Collections:
엘텍공과대학 > 컴퓨터공학과 > Journal papers
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

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

BROWSE