LAUSR

Abstract In this study the possibility to discriminate authentic honeys from adulterated honeys based on the FTIR spectra was investigated. Three types of authentic honeys (tilia, sunflower and acacia) were adulterated with agave, corn, inverted sugar, maple, and rice syrups in different percentages (5%, 10%, and 20%). The FTIR spectra was subjected to different spectral pre-treatment in order to improve the discrimination of authentic honeys from the adulterated ones using support vector machines (SVM) and partial least squares discriminant analysis (PLS-DA). SVM using 1st derivate spectra provided the most suitable model for the discrimination of the samples (in the calibration step from a total of 97 authentic honeys, 97 were correctly classified, while from a total of 210 adulterated honey, all were classified as adulterated ones; in the validation step from a total of 48 authentic honeys, 39 were correctly classified, while from 104 adulterated honeys 6 were classified as authentic). Partial least squares regression (PLS-R) was used for the prediction of the physicochemical parameters using the FTIR spectra.