Development of the POLIVAR model using neural networks as a tool to predict and identify monovarietal olive oils

Abstract

The growing consumer demand for varietal olive oils of differentiated properties necessitates implementing systematic procedures and tools that serve to detect fraud or alterations. The olive variety is, furthermore, a factor that guarantees and certifies an oil belongs to one of the registered geographical indications established in the European Union. The aim of this work was, by means of deploying artificial neural networks (ANNs), to develop a model that enables monovarietal olive oils to be predicted and identified, using a minimal number of chemical parameters, ensuring highly accurate quality control. To develop the model, we use data from 124 plots in areas assigned to protected designations of origin and quality marks in the region of Castilla-La Mancha (Spain). The results are influenced by the different content of the chemical parameters and show that with the analysis of only one parameter (margaroleic acid), the overall accuracy achieved (OAA) is 62.35%, rising to 91.50% when three parameters are included (stearic and oleic acids and concentration of campesterol) and 98.79% with five parameters (palmitic, stearic, linoleic and arachidic acids and concentration of campesterol). The development of the POLIVAR (Prediction of OLIve oil VARieties) model enables some of the most important monovarietal olive oils to be identified, thus being a useful tool to safeguard product quality and consumer interest.