WIT Press


A Visual Tool For Mining Macroeconomics Data

Price

Free (open access)

Paper DOI

10.2495/DATA040231

Volume

33

Pages

11

Published

2004

Size

511 kb

Author(s)

D. Giordano & F. Maiorana

Abstract

Data mining environments need tools capable of aiding results comprehension, in particular by resorting to the power of visual perception. This paper presents a tool designed and implemented in MatLab™ to facilitate visual mining of a large set of macroeconomics data on the world import and export activity of seven countries, in order to extend an analysis already performed with the aim of building models of national specialization and identifying possible market outlets. A model of the \“market” is built, by using a Multilayer Perceptron; then it is fed to a graphical interface which supports queries such as: can country X expand towards country Y in one particular sector? Answer messages point out, for the selected countries, sectors in which the market could be expanded, and for what specific products. As an aid to making sense of a given suggestion, a detailed market analysis can be carried out by resorting to a set of tabular and graphical views that are able to map the productive structure of each country and also the dynamics of its evolution through the years. The advantages of MatLab™ as a development environment are pointed out. It is argued that the additional effort for developing ad-hoc task oriented interfaces is especially justified in the case of frequent and interactive mining to be performed by decision makers. Keywords: visual data mining, neural networks, MatLab, macroeconomics. 1 Introduction One of the issues of interest in developing usable data mining environments is to provide the analysts with tools that aid model comprehension, quick analysis, and results interpretation. Flexible tools for data exploration and results visualization are one of the most appropriate solutions towards this aim. In Kopanakis and Theodoulidis [1] visual data mining is defined as involving the

Keywords

visual data mining, neural networks, MatLab, macroeconomics.