Enhanced data point importance for efficient data splitting in classification models: application to olive oil authentication.

In multivariate data analysis, selecting a representative subset of samples is crucial for constructing reliable one-class classification models. This study evaluates the application of the Enhanced Data Point Importance (EDPI) method for sample selection in DD-SIMCA modeling, in comparison with the classical Kennard-Stone (KS) approach. EDPI ranks samples based on their structural significance using a DPI-driven layered convex hull strategy, thereby identifying the most informative points in the dataset. For a dataset of 70 pure olive oil samples, the DD-SIMCA model built with 50 EDPI-selected samples achieved 100% sensitivity on the training set. These results are comparable to those of the KS model constructed with 60 samples, which yielded 98.33% sensitivity. Moreover, EDPI provided specificity levels of 100% for canola, hazelnut, and sunflower oils, and above 96% for soya oil and its mixtures comparable to or better than the Kennard-Stone method while reducing both the number of required samples and computational effort. Overall, the findings highlight EDPI as an efficient strategy for representative sample selection, offering practical applications in chemometrics and food authenticity verification.