Linear and non-linear quantification of extra virgin olive oil, soybean oil, and sweet almond oil in blends to assess their commercial labels

In this work, a method to quantify binary and ternary mixtures composed of extra virgin olive oil (protected designation of origin “Siurana”), soybean oil, and sweet almond oil is presented. The analytical approach is based on visible spectroscopy coupled with algorithmic tools of diverse natures (partial least squares and artificial neural networks (ANNs)). In general terms, the combination of absorption spectroscopy and ANNs result in better models than the linear ones tested. Specifically, when using ANN models, the estimation error of the concentration of every component in ternary blends leads to a 28% better accuracy than linear models. Due to the mean absolute error values achieved, this technique is suitable to rapidly verify the composition of mixtures of edible oils and, therefore, it could be used to validate the labels of extra virgin olive oil bottles.