A novel approach to olive oil sensory profiling: Predicting key attributes using near-infrared spectroscopy and open-source software

The official classification of olive oils into commercial categories relies on the Panel Test, a standardized method conducted by trained tasters. While essential for regulatory purposes, this approach is constrained by limited sample throughput, high cost, and dependence on specialized personnel. This study explores the use of near-infrared spectroscopy (NIRS) to predict sensory attributes of olive oil and to support their classification into official commercial categories. A total of 488 olive oil samples were analysed using three near-infrared (NIR) spectrometers—two portable devices and one benchtop instrument. Spectral data were processed using both qualitative and quantitative modelling approaches in an open-source environment to ensure transparency and reproducibility. Classification algorithms—partial least squares discriminant analysis (PLS-DA) and random forest (RF) classifier—were initially employed to detect fruitiness and sensory defects. Partial least squares regression (PLSR) models were subsequently used to predict the intensity of positive attributes: fruitiness, bitterness, and pungency. Model outputs enabled sample assignment to commercial categories. Classification models demonstrated strong performance in validation, achieving correct classification rates exceeding 94% and 82% for fruitiness and sensory defects, respectively. Quantitative predictions were moderate (residual predictive deviation for prediction, RPD p , between 1.12 and 1.57); however, a low-cost portable device performed comparably to the benchtop instrument, highlighting its potential for on-site quality control and broader accessibility for small and medium-sized producers. By integrating NIRS with sensory modelling, this work provides a practical, transparent, and cost-effective tool to complement official methods and expand access to reliable sensory quality control across the olive oil sector. • NIRS and open-source tools applied to olive oil sensory categorization • Quantitative and qualitative sensory attribute predictions using NIRS technology • PLS-DA accurately detects fruitiness and sensory defects in olive oils • Low-cost portable NIR device enables accessible, on-site olive oil quality analysis • NIRS may complement official Panel Test with more feasible olive oil quality control