Selection of discriminant wavelength intervals in NIR spectrometry with genetic algorithms

Abstract

Abstract Spectra, obtained in visible and near infrared spectral regions, contain much information whose interest depends on the context. For instance, in a discrimination context, the only information of interest is the one, concerning the differences between groups. This is our problem and we want to determine zones in the spectra (intervals), allowing a good discrimination between groups defined in advance. In that goal, genetic algorithms (GAs) will be used as a method of feature (wavelengths) selection. The Fisher Linear Discriminant Analysis (or Factorial Discriminant Analysis denoted FDA) will be applied on the selected variables. The efficiency of the method is shown on the discrimination of tobacco products. Copyright © 2007 John Wiley & Sons, Ltd.