Método de classificação com seleção meta-heurística de variáveis para identificar adição de leite bovino em caprino

Abstract

Goat milk is a rich source of proteins and carbohydrates, with an international valorization, mainly in European an Asian countries. Because nutritional proprieties and economic value, this product may be adulterated after milking by the addition of different cattle's milk. These adulterants have similar chemical composition, but distinct quantity and qualitative aspects from goat milk. In this perspective, the main goat milk adulterant is cow milk, used to ensure a masked adulteration, and at the same time provide economic gains. According to Brazilian legislation, any adulteration of milk by adding substances that change it’s organoleptic proprieties without proper label indication is considered crime. In a way to assist the inspection of food control official organizations, researches are developed to obtain fast, low cost, accurate and exact methods that attend legislation criteria. Portable instruments are suitable in these studies because of their low cost compared to a benchtop, allowing remote inspections. Near-infrared Spectroscopy (NIR), in association with chemometrics provides interpretation of information that allows distinguish authentic and adulterated samples. From this view, this work proposes the goat milk added with cow milk identification using portable NIR spectrometer along with chemometric methods. For that it was associated metaheuristic algorithms and discriminant analysis. GA-LDA and ACO-LDA showed efficient models with 100% of accuracy, sensitivity and specificity to discriminate goat milk authentic samples from adulterated, even with only 5% of cow milk added. Therefore, the developed models recognized goat milk samples adulterated by cow milk using fast, non-destructive and accurate methods.