Quantitative detection of adulteration in avocado oil using laser-induced fluorescence and machine learning models

Abstract

The increasing demand and high market value of avocado oil (AO) have heightened the risk of adulteration, necessitating the development of rapid, non-destructive, and reliable detection methods. This study explores the application of Laser-Induced Fluorescence (LIF) spectroscopy, combined with advanced machine learning techniques, for detecting and quantifying adulteration of AO with sunflower oil (SO). LIF spectra were collected using a 405 nm laser from AO samples adulterated with SO at levels ranging from 0 % to 50 % (5 % increments) and trace levels below 1 % (as low as 0.1 %). Three machine learning models—1D-Convolutional Neural Network (1D-CNN), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost)—were employed to classify adulteration levels. Among these, the SVM model achieved the highest accuracy (99.06 %), followed closely by the 1D-CNN (98.75 %), while XGBoost demonstrated good performance with an accuracy of 88.75 %. Principal Component Analysis (PCA) revealed distinct clustering patterns corresponding to different adulteration levels, emphasizing the sensitivity of LIF spectroscopy. Additionally, the method demonstrated the ability to detect trace adulteration levels below 1 %, with a calculated minimum Limit of Detection (LOD) of 0.0288 %. These results validate the potential of LIF spectroscopy, coupled with machine learning, as a robust and efficient tool for ensuring the authenticity and quality of avocado oil, offering a portable, non-destructive solution for food quality control and fraud prevention.