Exploring deep learning potential to authenticate vegetable oils and detect fraud using a micro-spectroscopic technique and chemometrics.

Authors

Juliana Opoku Yeboah, Roseline Love MacArthur, Francis Padi Lamptey, Jerry Opoku-Ansah, Charles Lloyd Yeboah Amuah, Robert Agbemafle, Caleb Mawuli Agbale, Ernest Teye

Published in

Analytical methods : advancing methods and applications. Jul 15, 2026. Epub Jul 15, 2026.

Abstract

Ensuring the authenticity of edible oils remains a critical challenge in food supply chains, particularly for high-value products susceptible to adulteration. This study investigated the potential of portable near-infrared (NIR) microspectroscopy (750-1050 nm) combined with chemometric and machine learning approaches for the authentication and fraud detection of virgin coconut oil (VCO), coconut oil (CO), palm kernel oil (PKO), and groundnut oil (GNO). NIR spectral fingerprints were acquired using an SCiO portable spectrometer and analysed using both classification and quantification models. For qualitative discrimination, several pattern recognition algorithms, including k-nearest neighbour (KNN), linear discriminant analysis (LDA), support vector machine (SVM), and artificial neural network (NN), were evaluated. The results demonstrated a clear spectral differentiation among the oil types. The NN model achieved the best classification performance, yielding 100% accuracy for the calibration set and 97.19% accuracy for the independent prediction set, outperforming KNN, LDA, and SVM models. For quantitative analysis, partial least squares regression (PLSR) models were developed to predict the level of adulteration of virgin coconut oil with coconut oil. Different variable selection and preprocessing strategies, including backward interval partial least squares (BiPLS), successive projections algorithm (SPA-PLS), and standard normal variate (SNV-PLS), were compared in this study. Among the evaluated models, the SNV-PLS approach provided the best predictive performance, achieving a coefficient of determination (R²) of 0.97, indicating an excellent agreement between the reference and predicted values. The findings demonstrate that portable NIR microspectroscopy combined with advanced chemometric and machine learning models offers a rapid, non-destructive, and reliable approach for both the classification and quantification of vegetable oil fraud. This approach shows strong potential for routine screening and on-site authenticity assessment of edible oils, particularly in resource-limited and decentralised food control environments.

PMID:
42454456
Bibliographic data and abstract were imported from PubMed on 15 Jul 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.