Spectra-Based Multivalued Fingerprints as Predictive Vectors for Partial Least Squares Regressions Processes (SI-CMMSE-2006)

Miguel Ángel Gómez-Nieto (1) , Irene Luque Ruiz (2) , Manuel Urbano Cuadrado (3)

1 Computing and Numerical Analysis
2 Computing and NUmerical Analysis
3 ICIQ - Institute of Chemical Research of Catalonia

A new method for transforming spectra into multivalued fingerprints is here presented and applied to multivariate regression. The method, aimed at enlarging differences between long-dimensional vectors showing a high degree of similarity, is based on: spectral outliers removal; data normalization aimed at transforming the spectral matrix into a new data set within the [0,1] range; and selection of threshold values for assigning significance values to each variable. A study case is described: the processing of mid infrared spectra in partial least squares regression processes for predicting total acidity and content of reducing sugars in wines. The original spectra matrix consisted of 156 samples and 1142 columns predictors -a wavelenght range of 3000-800 cm^-1 with a spectral resolution slightly greater than 2 cm-1. The fact of using the here proposed method yielded better predictions than those obtained by means of both classical treatments and spectral data without any processing.

Mots clés: Physical Sciences

Miguel Ángel Gómez-Nieto, Irene Luque Ruiz, Manuel Urbano Cuadrado. Spectra-Based Multivalued Fingerprints as Predictive Vectors for Partial Least Squares Regressions Processes (SI-CMMSE-2006). International Journal of Computer Mathematics, 2008, 85 (03-04), pp.691-702. ⟨10.1080/00207160601161436⟩. ⟨hal-00545345⟩