Rapid identification method of milk powder from different animals based on Raman spectroscopy

Abstract

This study developed an efficient method for identifying and quantitatively analyzing animal-origin milk powders using Raman spectroscopy combined with chemometrics. By employing the MultiClassClassifier model, the method achieved high accuracy in distinguishing various types of animal-origin milk powders, with sensitivity and specificity both exceeding 80% and an overall accuracy of 93%. Furthermore, the quantitative models based on Partial Least Squares Regression and Support Vector Machine Regression exhibited excellent linear correlations, with both Root Mean Square Error and Mean Relative Error below 0.2. These models successfully quantified adulteration in camel, mare, and donkey milk powders in comparison to goat and cow milk powders. The study's approach not only holds significant promise for detecting adulteration in specialty milk powders but also demonstrates wide applicability in analyzing other powdered adulterants.