Data sample reduction for classification of interval information using neural network sensitivity analysis

Abstract

The aim of this paper is present a novel method of data sample reduction for classification of interval information. Its concept is based on the sensitivity analysis, inspired by artificial neural networks, while the goal is to increase the number of proper classifications and primarily, calculation speed. The presented procedure was tested for the data samples representing classes obtained by random generator, real data from repository, with clustering also being used. © 2010 Springer-Verlag.