Chemometric tools combined with excitation-emission matrix fluorescence for the identification and quantification of...

Pequi oil, extracted from the pulp of the pequi fruit, is a high-value edible oil native to the Brazilian Cerrado. It is frequently targeted for adulteration with lower-cost vegetable oils due to its high economic value. This study presents a novel, low-complexity strategy for processing excitation-emission matrix (EEM) fluorescence data that avoids the use of advanced tensor decomposition techniques. Instead, it employs dimensionality reduction through basic vector operations and first-order derivatives to extract key spectral features in a compact 2D format. Combined with chemometric tools, this approach was applied to detect and quantify adulteration in pequi oil with soybean, corn, and canola oils. Principal component analysis (PCA) allowed clear separation between pure and adulterated samples. Classification models, including support vector machines (SVM) and linear discriminant analysis (LDA), achieved accuracies of up to 87 % in training and 77 % in the independent test set. Partial least squares regression (PLSR) models developed for quantifying soybean, corn, and canola oils yielded R² values of 0.9749, 0.9706, and 0.9423, with root mean square error of prediction (RMSE) of 1.74 %, 1.81 %, and 2.23 %, respectively. The corresponding mean square error in 5-fold cross-validation (MSECV) values were 0.17 %, 0.32 %, and 0.25 %. On the independent test set, the predicted versus actual R² values were 0.8518 (soybean), 0.7866 (canola), and 0.8150 (corn), indicating strong predictive performance. These findings highlight the proposed method's simplicity, sensitivity, and suitability for rapid screening, offering a practical alternative to more complex fluorescence or chromatographic methods in oil authentication and quality control.