Rapid Classification of Milk Using a Cost-Effective Near Infrared Spectroscopy Device and Variable Cluster-Support...

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Food fraud, including fat removal and water addition, compromises milk quality.
• Deviations in fat content are key indicators of milk adulteration.
• Infrared spectroscopy, particularly portable near-infrared spectroscopy (NIRS), is effective for milk analysis.

Purpose of the Study:

• To develop and implement a rapid milk classification method using FT-NIR spectroscopy and a VC-SVM model.
• To assess the model's effectiveness in identifying milk fat content variations without pre-treatment.
• To comply with EU Regulation EC No. 1308/2013 for milk classification and fraud detection.

Main Methods:

• Utilized a compact, low-cost Fourier Transform near-infrared (FT-NIR) spectrometer.
• Employed a variable cluster-support vector machine (VC-SVM) hybrid model for data analysis.
• Performed external validation to assess classification accuracy for different milk types.

Main Results:

• Achieved perfect classification (100% sensitivity, 100% specificity) for whole vs. not-whole and skimmed vs. not-skimmed milk using the VC-SVM model with a radial basis function (RBF) kernel.
• Demonstrated strong classification (94.4% sensitivity, 100% specificity) for semi-skimmed vs. not-semi-skimmed milk.
• Validated the method's efficacy without any sample pre-treatment.

Conclusions:

• The FT-NIR spectroscopy combined with the VC-SVM model offers a practical, simple, and efficient solution for the dairy industry.
• This approach enables rapid identification of skimmed, semi-skimmed, and whole milk, aiding in the detection of potential food fraud.
• The method provides reliable analytical performance comparable to benchtop instruments for in situ analysis.