Mining multi-dimensional data with visualization techniques

Abstract

This paper describes a method to generate classification rules by using an interactive multidimensional data visualization and classification tool, called PolyCluster. PolyCluster is a system that adopts state-of-the-art algorithms for data visualization and integrates human domain knowledge into the construction process of classification rules. In addition, PolyCluster proposes a pair of novel and robust measurements, called the Average External Connecting Distance and the Average Internal Connecting Distance to evaluate the quality of the induced clusters. Experimental evaluation shows that PolyCluster is a visual-based approach that offers numerous improvements over previous visual-based techniques. © Springer-Verlag Berlin Heidelberg 2004.