LAUSR

Abstract The influence of temperature, frequency, composition and physicochemical characteristics of sheep milk on its dielectric properties was evaluated as well as the performance of different chemometric methods to predict the dielectric properties and classify the type of sheep milk (whole, semi-skimmed and skimmed). Of the chemometric methods evaluated, artificial neural network exhibited the best performance for prediction of the dielectric properties, while sensitivity analysis showed temperature, electrical conductivity, and fat and calcium content as variables with the most impact. All pattern recognition techniques showed 100% for recognition and prediction ability to classify correctly the type of milk. Although the approach used in this study is limited to the specific operating conditions and sheep milk studied, chemometric methods have proven to be promising tools because of accuracy and suitability for both prediction of the dielectric properties of sheep milk and monitoring quality control parameters of milk and dairy products.