Abstract

Visualization is a key issue for multivariate data analysis. Multivariate visualization is an active research topic and many efforts have been made in order to find suitable and meaningful visual representations. In this paper we present a technique for data projection in multivariate datasets, named Target Data Projection (TOP). Through this technique a vector is created for each multivariate data item considering a subset of the available variables. A new scalar variable is generated projecting those vectors over a target vector that defines the direction of interest for visual analysis. End-users set up target vectors in order to explore particular relationships by means of application meaningful projections. Hence, it is possible to map a combination of multivariate data into one scalar variable for graphical representation and interaction. This technique has proved to be very flexible and useful in mine planning providing valuable information for decision making. © Springer-Verlag 2004.