Multivariable classification of pumpkin seed oils using absorption spectra

Abstract

Classification models for edible oil samples which represent the most widespread brands of Austrian pumpkin seed oils were elaborated using several techniques of multivariate data analysis. Characterization of the samples was achieved by UV-VIS, Near IR and FT IR spectral methods and complemented by basic sensorial classification given by a panel of experts. Chemometrical treatment of measured data enabled to detect important spectral features, mostly suitable for categorization of investigated oils into two or three classes, defined according to their sensory quality. Thus classification was made into (a) two categories, containing good (fully satisfactory) or bad (not fully satisfactory) oils, (b) three categories, involving excellent, satisfactory and bad oil brands. The classifocation models, elaborated separately for each kind of spectra, enable to predict the category into which a hitherto unclassified oil sample belongs. It will perspectively facilitate determination of the chemical substances responsible for bad taste, improper odour and inconvenient colour of the respective oil brands, as well as selection of substances contributing to the excellent sensorial perception of the tested market products.