Application of machine-learning methods to milk mid-infrared spectra for discrimination of cow milk from pasture or total mixed ration diets

摘要

The prevalence of “grass-fed” labeled food productson the market has increased in recent years, often commandinga premium price. To date, the majority ofmethods used for the authentication of grass-fed sourceproducts are driven by auditing and inspection of farmrecords. As such, the ability to verify grass-fed sourceclaims to ensure consumer confidence will be importantin the future. Mid-infrared (MIR) spectroscopy is widelyused in the dairy industry as a rapid method for theroutine monitoring of individual herd milk compositionand quality. Further harnessing the data from individualspectra offers a promising and readily implementablestrategy to authenticate the milk source at both farmand processor levels. Herein, a comprehensive comparisonof the robustness, specificity, and accuracy of11 machine-learning statistical analysis methods weretested for the discrimination of grass-fed versus nongrass-fed milks based on the MIR spectra of 4,320 milksamples collected from cows on pasture or indoor totalmixed ration–based feeding systems over a 3-yr period.Linear discriminant analysis and partial least squaresdiscriminant analysis (PLS-DA) were demonstrated tooffer the greatest level of accuracy for the prediction ofcow diet from MIR spectra. Parsimonious strategies forthe selection of the most discriminating wavelengthswithin the spectra are also highlighted.