Rapid discrimination of different primary processing Arabica coffee beans using FT-IR and machine learning

Area of Science:

• Food Science
• Analytical Chemistry
• Spectroscopy

Background:

• Primary processing significantly impacts coffee bean characteristics.
• Rapid and accurate methods are needed to differentiate coffee varieties.
• Understanding structural and chemical differences is crucial for quality control.

Purpose of the Study:

• To develop and validate a rapid analytical method for distinguishing three primary processed Arabica coffee beans.
• To compare the effectiveness of FT-IR spectroscopy and machine learning with other analytical techniques.
• To identify key analytical markers for coffee bean processing differentiation.

Main Methods:

• Fourier transform infrared (FT-IR) spectroscopy coupled with machine learning algorithms.
• Colorimetry, low-field nuclear magnetic resonance (NMR) spectroscopy, scanning electron microscopy (SEM), and two-dimensional correlation spectroscopy (2D-COS).
• Multivariate statistical analysis, including orthogonal partial least squares-discriminant analysis (OPLS-DA).

Main Results:

• Sun-exposed processed beans (SPB) showed higher color difference and larger pore size.
• Wet-processed beans (WPB) retained more bound and immobilized water.
• FT-IR indicated similar functional groups but different structural characteristics via 2D-COS.
• OPLS-DA effectively distinguished coffee bean types.
• Machine learning models, particularly SNV-Voting, achieved high accuracy (88.67% precision, recall, F1-score) in classification.

Conclusions:

• The integrated FT-IR and machine learning approach provides a rapid and effective method for differentiating primary processed Arabica coffee beans.
• Specific physical and chemical properties correlate with different processing methods.
• This methodology holds potential for quality assessment and authentication in the coffee industry.