Home // BIOTECHNO 2014, The Sixth International Conference on Bioinformatics, Biocomputational Systems and Biotechnologies // View article

Experimental Verification of the Quality of Clusterings Produced by Hard Clustering Algorithms after the Removal of Unstable Data Elements

Authors:
Wim De Mulder
Zahra Zavareh
Konika Chawla
Martin Kuiper

Keywords: hard clustering; cluster quality; unstable elements; mutual information context; microarray data

Abstract:
Many different clustering algorithms have been developed to detect structure in data sets in an unsupervised way. As user intervention for these methods should be kept to a minimum, robustness with respect to user-defined initial conditions is of crucial importance. In a previous study, we have shown how the robustness of a hard clustering algorithm can be increased by the removal of what we called unstable data elements. Although robustness is a main characteristic of any clustering tool, the most important feature is still the quality of the produced clusterings. This paper experimentally investigates how the removal of unstable data elements from a data set affects the quality of produced clusterings, as measured by the mutual information index, using three biological gene expression data sets.

Pages: 63 to 69

Copyright: Copyright (c) IARIA, 2014

Publication date: April 20, 2014

Published in: conference

ISSN: 2308-4383

ISBN: 978-1-61208-335-3

Location: Chamonix, France

Dates: from April 20, 2014 to April 24, 2014