Home // ADAPTIVE 2013, The Fifth International Conference on Adaptive and Self-Adaptive Systems and Applications // View article

A Gravitational Approach for Enhancing Cluster Visualization in Self-Organizing Maps

Authors:
Leonardo Enzo Brito da Silva
José Alfredo Ferreira Costa

Keywords: self-organizing maps; gravitational clustering; visualization techniques

Abstract:
This paper presents a modified gravitational clustering algorithm applied to the neurons of self-organizing maps in order to enhance the visualization of clusters through the U-matrix technique. For a given neuron, the proposed method considers the attraction among its k nearest neighbors in the data space, where k decreases monotonically over time and its value may vary according to the local pattern density. The attraction between neurons that are considered as not belonging to the same close-knit group is penalized. The results obtained for some synthetic and real world data sets are presented.

Pages: 48 to 54

Copyright: Copyright (c) IARIA, 2013

Publication date: May 27, 2013

Published in: conference

ISSN: 2308-4146

ISBN: 978-1-61208-274-5

Location: Valencia, Spain

Dates: from May 27, 2013 to June 1, 2013