A vector field visualization technique for self-organizing maps

Abstract

The Self-Organizing Map is one of most prominent tools for the analysis and visualization of high-dimensional data. We propose a novel visualization technique for Self-Organizing Maps which can be displayed either as a vector field where arrows point to cluster centers, or as a plot that stresses cluster borders. A parameter is provided that allows for visualization of the cluster structure at different levels of detail. Furthermore, we present a number of experimental results using standard data mining benchmark data. © Springer-Verlag Berlin Heidelberg 2005.