A machine learning proposal method to detect milk tainted with cheese whey
Juliana S Lima,Daniela C S Z Ribeiro,Habib Asseiss Neto,Sérgio V A Campos,Mônica O Leite,Márcia E de R Fortini,Beatriz Pinho Martins de Carvalho,Marcos Vinícius Oliveira Almeida,Leorges M Fonseca
DOI: https://doi.org/10.3168/jds.2021-21380
Abstract: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 of milk and milk-derived products, and even some safety concerns. Nevertheless, methods to detect fraudulent addition of cheese whey to milk are expensive and time consuming, and are thus ineffective as screening methods. The Fourier-transform infrared (FTIR) spectroscopy technique is a promising alternative to identify this type of fraud because a large number of data are generated, and useful information might be extracted to be used by machine learning models. The objective of this work was to evaluate the use of FTIR with machine learning methods, such as classification tree and multilayer perceptron neural networks to detect the addition of cheese whey to milk. A total of 520 samples of raw milk were added with cheese whey in concentrations of 1, 2, 5, 10, 15, 20, 25, and 30%; and 65 samples were used as control. The samples were stored at 7, 20, and 30°C for 0, 24, 48, 72, and 168 h, and analyzed using FTIR equipment. Complementary results of 520 samples of authentic raw milk were used. Selected components (fat, protein, casein, lactose, total solids, and solids nonfat) and freezing point (°C) were predicted using FTIR and then used as input features for the machine learning algorithms. Performance metrics included accuracy as high as 96.2% for CART (classification and regression trees) and 97.8% for multilayer perceptron neural networks, with precision, sensitivity, and specificity above 95% for both methods. The use of milk composition and freezing point predicted using FTIR, associated with machine learning techniques, was highly efficient to differentiate authentic milk from samples added with cheese whey. The results indicate that this is a potential method to be used as a high-performance screening process to detected milk adulterated with cheese whey in milk quality laboratories.