Machine learning-assisted FT-IR spectroscopy for identification of pork oil adulteration in tuna fish oil

Abstract

Tuna fish oil (TO) is a valuable source of omega fatty acids and polyunsaturated fatty acids required for human growth and development. Triggered by economic reasons, TO can potentially be adulterated with pork oil (PO), which has a lower price. The adulteration is a serious problem because PO is a non-halal oil, which is truly prohibited to be consumed, especially for Muslim. This research aimed to develop an effective and efficient analytical technique for detecting PO adulteration in TO using Fourier transform infrared (FT-IR) spectroscopy aided by machine learning techniques. Various machine learning techniques were developed, including linear regression, support vector machine (SVM), k-nearest neighbor (kNN), artificial neural network (ANN), and gradient boosting. The result showed that SVM at the fingerprint region (1400–900 cm⁻¹) demonstrated the best model to detect and predict PO in TO with the highest R² (0.993) and the lowest root mean square error (RMSE) of 2.719 %. All levels of PO contained in TO could be accurately predicted, as indicated by the closeness between the actual value and predicted value of PO levels predicted by the model. In conclusion, machine learning could be a promising tool for detecting adulterants in fish oil samples. Further research on method standardization is important to propose the method as the method of choice for fish oil authentication, including halal authentication.