Machine learning-enabled high-throughput industry screening of edible oils

Area of Science:

• Food Science
• Analytical Chemistry
• Spectroscopy

Background:

• Edible oil adulteration poses significant global public health risks.
• Complex supply chains often allow fraudulent practices to go undetected.
• Consumer confidence requires robust traceability and accountability in the food industry.

Purpose of the Study:

• To develop a high-throughput, non-destructive, and label-free method for edible oil authentication.
• To assess the feasibility of machine learning-assisted spectroscopy for large-scale industrial screening.
• To enable rapid analysis and detection of adulteration in edible oils.

Main Methods:

• Utilized machine learning (ML) algorithms with windowed spectroscopy (visible and infrared bands).
• Applied the method to a large-scale dataset (approximately 11,500 samples) for olive oil adulteration detection.
• Focused on spectral fingerprint analysis for oil authentication.

Main Results:

• Achieved high discriminant performance (Area Under the Curve > 0.96) in detecting olive oil adulteration.
• Demonstrated the effectiveness of ML in analyzing spectral data for authentication.
• Showcased high clustering fidelity of spectral fingerprints, enabling database compilation.

Conclusions:

• Machine learning-assisted windowed spectroscopy provides a feasible solution for rapid, large-scale edible oil authentication.
• The developed method enhances traceability and accountability within the edible oil supply chain.
• This approach facilitates the creation of self-sustaining, hypothesis-free databases for oil quality monitoring.