Detection of ground coffee adulteration using FTIR coupled with pattern recognition techniques

Area of Science:

• Food Science
• Analytical Chemistry
• Spectroscopy

Background:

• Economically motivated adulteration of ground coffee is a significant concern.
• Robust analytical methods are essential for verifying coffee authenticity.

Purpose of the Study:

• To evaluate Fourier Transform Infrared (FTIR) spectroscopy combined with pattern recognition for detecting adulteration in ground coffee.
• To compare the performance of supervised and unsupervised machine learning models for this purpose.

Main Methods:

• Acquisition and preprocessing of spectral fingerprints from pure and adulterated coffee samples (barley, chickpea, date pit).
• Analysis using supervised (Random Forest, K-Nearest Neighbors, Decision Tree) and unsupervised (Principal Component Analysis, Hierarchical Cluster Analysis) learning models.

Main Results:

• Spectral preprocessing improved classification performance across all models.
• Random Forest achieved 94% training and 83% testing accuracy.
• Hierarchical Cluster Analysis (HCA) demonstrated superior performance with 96.8% accuracy, including perfect classification for pure coffee, date pit, and chickpea.

Conclusions:

• FTIR spectroscopy coupled with pattern recognition is effective for detecting coffee adulteration.
• Hierarchical Cluster Analysis (HCA) shows exceptional potential for rapid, label-free screening of coffee authenticity.
• This comparative study highlights HCA as a promising tool for food authenticity verification.