Comparación de herramientas quimiométricas de clasificación para la identificación de grasa perirrenal en corderos

Abstract

This study verifies the ability of the NIRS technique to correctly classify merino lamb perirenal fat according their feeding during fattening. Different spectral treatments (Absorbance-A, Reflectance-R, A-EMSC, R-EMSC), unsupervised (PCA-Projection) and supervised classification techniques (SIMCA, PLS-DA, LDA, C-SVM) are combined. The results indicate that untreated spectra contain valuable information in terms of classification, despite the existing light scattering. The simplest PCA-Projection and LDA classification methods are also the most effective. Of the other more sophisticated methods, C-SVM is more effective, although it is sensitive to the type of initiation mathematical seed used. It is concluded that NIRS spectroscopy combined with chemometric techniques is a valid method to classify lamb carcasses in slaughter room according to diet during fattening period.