Classification of oils and margarines by FTIR spectroscopy in tandem with machine learning

This paper presents a comprehensive approach to food authentication using Fourier-transform infrared (FTIR) spectroscopy combined with machine learning algorithms. The research focuses on identifying and classifying pure njangsa seed oil, palm kernel oil, coconut oil, and their corresponding margarines, while also detecting and quantifying adulteration with sunflower oil and canola-flaxseed oil margarine.

The study employed six different machine learning models including K-nearest neighbors (KNN), support vector machines (SVM), decision trees, and ensemble methods to achieve highly accurate classification results. KNN emerged as the most effective classifier for pure oils (97% accuracy) and the best predictor for quantifying adulterant concentrations.

This research addresses critical food safety and authenticity concerns in the edible oil industry, where economic adulteration is a significant problem. The developed FTIR-ML methodology offers a rapid, non-destructive alternative to traditional chromatographic methods, making it particularly valuable for quality control applications in food processing and regulatory settings.

The collaborative effort involved researchers from multiple institutions and represents an important advancement in applying spectroscopic techniques and chemometrics for food authentication, particularly focusing on underutilized African seed oils like njangsa. The methodology successfully distinguished between pure and adulterated samples with detection limits as low as 1% (v/v) or 1% (w/w), demonstrating its potential for practical implementation in food quality control laboratories.

Recommended citation: Tachie, C. Y. E., Obiri-Ananey, D., Alfaro-Cordoba, M., Tawiah, N. A., & Aryee, A. N. A. (2024). "Classification of oils and margarines by FTIR spectroscopy in tandem with machine learning." Food Chemistry, 431, 137077.
Download Paper