Background: The organoleptic profile of an olive oil is a fundamental quality parameter obtained by human sensory panels. In this work, a portable electronic nose was employed to predict the fruity aroma intensity of 199 olive oil samples from different Spanish regions and cultivar varieties ('Picual', 'Arbequina', and 'Cornicabra'), with special emphasis in testing the robustness of the predictions versus cultivar variety variability. The primary data given by the electronic nose were used to obtain two different feature vectors that were employed to fit ridge and lasso regressions models to two datasets: one consisting of all the samples and another just the cv. Picual samples.
Results: The results obtained showed mean average error (MAE) values below 0.88 in all cases, with an MAE of 0.67 for the 'Picual' model. These MAE values and the similarities in the model parameters fitted for the different data folds are in agreement with the results obtained in previous studies.
Conclusion: The large number of samples analyzed and the results obtained show the robustness of the approach and the applicability of the methods. Also, the results suggest that better performance can be obtained when specific models are fitted for particular cultivars. Overall, the proposed methods are capable of providing useful information for a fast screening of the fruity aroma intensity of olive oils. © 2023 The Authors. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
Keywords: e-nose; feature extraction; metal oxide semiconductor sensor; multivariate analysis; olive oil quality; organoleptic assessment.
© 2023 The Authors. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.