Icon and geometric data visualization with a self-organizing map grid

Abstract

Data Visualization is an important tool for tasks related to Knowledge Discovery in Databases (KDD). Often the data to be visualized is complex, have multiple dimensions or features and consists of many individual data points, making visualization with traditional icon- and pixel-based and geometric techniques difficult. In this paper we propose a combination of icon-based and geometric-based visualization techniques backed up by a Self-Organizing Map, which allows dimensionality reduction and topology preservation. The technique is applied to some datasets of simple and intermediate complexity, and the results shows that it is possible to reduce clutter and facilitate identification of associations, clusters and outliers. © 2014 Springer International Publishing.