Title: Some studies on mapping methods

Authors: Parikshit Dutta, Dilip Kumar Pratihar

Addresses: Department of Mechanical Engineering, Indian Institute of Technology, Kharagpur-721 302, India. ' Department of Mechanical Engineering, Indian Institute of Technology, Kharagpur-721 302, India

Abstract: A human being can visualise only up to 3-dimensions. A mapping tool is essential to map the higher dimensional data to a lower dimension for visualisation. Both linear as well as non-linear mapping methods have been used by various researchers for the said purpose. In the present work, a non-linear mapping method based on a genetic algorithm has been developed and its performance is compared to that of other methods, namely Sammon|s NLM, VISOR and SOM, in terms of accuracy in mapping, visibility of the mapped data and computational complexity, for solving Schaffer|s and DeJong|s test functions. The proposed GA-like approach and VISOR algorithm are found to be the best and worst, respectively, in terms of accuracy in mapping, ease of visualization. Moreover, the GA-like approach and VISOR algorithm are seen to be the slowest and fastest, respectively, of all the methods.

Keywords: mapping methods; visualisation; NLM; VISOR; SOM; nonlinear mapping; genetic algorithms; mapping accuracy; modelling precision.

DOI: 10.1504/IJBIDM.2006.009139

International Journal of Business Intelligence and Data Mining, 2006 Vol.1 No.3, pp.347 - 370

Published online: 06 Mar 2006 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article