Random forest as one-class classifier and infrared spectroscopy for food adulteration detection

Area of Science:

• Food Science
• Analytical Chemistry
• Machine Learning

Background:

• Food adulteration poses significant risks to consumer health and economic integrity.
• Accurate and efficient detection methods are crucial for ensuring food quality and safety.
• Existing methods may have limitations in sensitivity or applicability across diverse matrices.

Purpose of the Study:

• To develop and validate a novel random forest-based approach for detecting food adulteration.
• To compare the performance of the proposed method against established techniques like PLS-DA and SIMCA.
• To assess the robustness of the random forest model in real-world food analysis scenarios.

Main Methods:

• Utilized the random forest algorithm combined with artificial outlier generation from authentic samples.
• Applied the method to two distinct food matrices: evening primrose oil (ATR-FTIR spectroscopy) and ground nutmeg (NIR diffuse reflectance spectroscopy).
• Validated the models using various pure and adulterated samples, including common adulterants and diverse oil/spice types.

Main Results:

• The random forest method demonstrated superior performance compared to PLS-DA for both evening primrose oil and ground nutmeg.
• For evening primrose oil, the random forest showed performance similar to SIMCA.
• In both applications, the random forest model achieved excellent results without excluding any samples from external validation.

Conclusions:

• The proposed random forest approach is a highly effective and robust method for food adulteration detection.
• This technique offers a promising alternative or enhancement to existing analytical methods in food safety.
• The successful application across different spectroscopic techniques and food types highlights its versatility.