Rapid Detection of Adulterants in Palm Oil Using A-TEEM and Advanced Predictive Models

Economically-motivated adulteration (EMA) of palm oil with cheaper oils like corn, soybean, and rapeseed oil, is a growing concern in the food industry; it compromises product quality and poses potential health risks to consumers. Adulteration also affects the accuracy of traceability, thereby undermining broader sustainability efforts such as carbon accounting or responsible sourcing. Thus, there is an urgent need for fast and accurate adulteration detection. This study employs A-TEEM (Absorbance, Transmittance, and Emission-Excitation Matrix) spectroscopy to provide a solution for screening adulterants in palm oil. Various commercial palm oil samples were tested, which were artificially adulterated with 10% (w/w) of unwanted oils. The difference between control and adulterated samples were confirmed by fatty acid profiling via GC-FID analysis.  The sample preparation was only one step of dilution with hexane at a 1:30 ratio and detailed EEM and absorbance data were generated within seconds, which enabled high-throughput screening. A robust dataset comprising 420 samples, 379 randomly assigned samples were used to train predictive models for binary classification, and 41 samples were used for external validation. The training process incorporated five-fold cross-validation repeated over 10 iterations to ensure model reliability and robustness. Machine learning techniques, including Random Forest, Support Vector Machines, K-nearest neighbors, Gradient Boosting Machines, Single-Layer Neural Networks, Stochastic Gradient Boosting, and Extreme Gradient Boosting Machines, were applied to classify adulterated samples. Among these, models based on Random Forest, Stochastic Gradient Boosting, and Extreme Gradient Boosting Machines demonstrated superior performance, achieving high accuracy ( >98%), sensitivity (100%), and specificity ( >97%). This combined approach of A-TEEM spectroscopy and machine learning offers a fast and reliable solution for detecting adulterants in palm oil. By making more efficient the detection process, the method has significant potential to be used for in-bound tests and enhance quality and food safety in the food industry.