A machine learning proposal method to detect milk tainted with cheese whey

• DOI: 10.3168/jds.2021-21380
• Publication Date: 2022-10-04T23:56:18Z
• Authors: Juliana S. Lima; Daniela C.S.Z. Ri...; Habib Asseiss Neto; Sérgio V.A. Campos; Mônica O. Leite; Márcia E. de R. F...; Beatriz Pinho Mar...; Marcos Vinícius O...; Leorges M. Fonseca
• Keywords: Total dissolved solids; Dairy industry

Cheese whey addition to milk is a type of fraud with high prevalence and severe economic effects, resulting in low yield for dairy products, nutritional reduction milk-derived even some safety concerns. Nevertheless, methods detect fraudulent cheese are expensive time consuming, thus ineffective as screening methods. The Fourier-transform infrared (FTIR) spectroscopy technique promising alternative identify this because large number data generated, useful information might be extracted used by machine learning models. objective work was evaluate the use FTIR methods, such classification tree multilayer perceptron neural networks milk. A total 520 samples raw were added concentrations 1, 2, 5, 10, 15, 20, 25, 30%; 65 control. stored at 7, 30°C 0, 24, 48, 72, 168 h, analyzed using equipment. Complementary results authentic used. Selected components (fat, protein, casein, lactose, solids, solids nonfat) freezing point (°C) predicted then input features algorithms. Performance metrics included accuracy 96.2% CART (classification regression trees) 97.8% networks, precision, sensitivity, specificity above 95% both composition FTIR, associated techniques, highly efficient differentiate from whey. indicate that potential method high-performance process detected adulterated quality laboratories.