Vibrational spectroscopy data fusion for enhanced classification of different milk types

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Accurate classification of milk types is crucial for quality control and product differentiation.
• Vibrational spectroscopy (NIR, MIR, Raman) offers non-destructive methods for analyzing food products.
• Chemometric techniques are essential for extracting meaningful information from complex spectral data.

Purpose of the Study:

• To classify seven types of Irish milk using Near Infrared (NIR), Mid Infrared (MIR), and Raman spectroscopy.
• To evaluate the impact of spectral data fusion on the prediction accuracy of milk classification.
• To compare the effectiveness of Principal Component Analysis (PCA), Partial Least Square Discriminant Analysis (PLS-DA), and Sequential and Orthogonalized Partial Least Squares Linear Discriminant Analysis (SO-PLS-LDA).

Main Methods:

• Utilized vibrational spectroscopy techniques: NIR, MIR, and Raman spectroscopy.
• Applied chemometric methods including PCA, PLS-DA, and SO-PLS-LDA for classification.
• Investigated the effect of low-level data fusion on classification performance.

Main Results:

• PLS-DA on individual or fused spectral data achieved a maximum accuracy of 85.71%.
• The SO-PLS-LDA strategy, applied to fused spectral data, significantly enhanced prediction accuracy to 95%.
• This demonstrates the superiority of SO-PLS-LDA for milk type classification.

Conclusions:

• SO-PLS-LDA combined with spectral data fusion is a highly effective method for classifying different milk types.
• The developed classification model shows promise for industrial applications in milk quality control.
• Vibrational spectroscopy and advanced chemometrics provide a powerful analytical toolkit for food analysis.