Premium
Multivariate strategies for classification of Eucalyptus globulus genotypes using carbohydrates content and NIR spectra for evaluation of their cold resistance
Author(s) -
Castillo Rosario,
Otto Matthias,
Freer Juanita,
Valenzuela Sofía
Publication year - 2008
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/cem.1126
Subject(s) - linear discriminant analysis , quadratic classifier , partial least squares regression , eucalyptus globulus , multivariate statistics , mathematics , near infrared spectroscopy , pattern recognition (psychology) , discriminant , artificial intelligence , statistics , chemometrics , eucalyptus , chemistry , chromatography , computer science , botany , biology , support vector machine , neuroscience
Different strategies of classification are tested and compared over near infrared (NIR) spectra and carbohydrates content data of genotypes of Eucalyptus globulus . Strategies used for the classification were regularized discriminant analysis (RDA) for carbohydrates content data and for the singular NIR spectral data, partial least squares (PLS) for reduction of variables to the scores matrix of NIR spectra plus RDA for classification (PLS/RDA on scores) and PLS‐discriminant analysis (PLS‐DA). Different types of discriminant functions were tested in PLS/RDA on scores method. Results obtained using NIR scores data outperformed those results obtained with carbohydrate content data. Comparison between specified PLS/RDA on scores strategies showed best classifications of the genotypes when parameters λ and γ pertaining to linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA) were used, with 0% risk of misclassification and 100% of correctly assigned samples in the external validation sets for classification of the genotypes by forest company and by cold chamber while external validation for classification of genotypes by cold resistance degree shows 70 and 90% of correctly assigned samples for sensible and tolerant genotypes. Copyright © 2008 John Wiley & Sons, Ltd.