DIRECT ANALYSIS OF VEGETABLE OILS BY ATMOSPHERIC PRESSURE LASER PLASMA IONIZATION COMBINED WITH MACHINE LEARNING METHODS

The atmospteric pressure laser plasma ionization (APLPI) method in combination with machine learning methods is investigated to solve the problem of classification of vegetable oils. Samples of olive oil, rapeseed oil, sunflower oil and linseed oil were studied. The samples were classified on the basis of mass spectrometric profiles of volatile organic compounds emitted by the oils. It was shown that when hierarchical cluster analysis (HCA) with pre-selection of features by analysis of variance (ANOVA) and reduction of the dimensionality of the response matrix by t-distributed stochastic neighbor embedding (t-SNE), each type of oil forms a distinct cluster. Using the example of olive and rapeseed oil blends analysis, it was demonstrated that the combination of the APLPI method with the multiple linear regression (MLR) method allows to quantify the share of oils in the studied blends. The developed approach allows for rapid, direct nondestructive analysis of vegetable oils without sample preparation and can be used for detection of adulterated products.