Answers are grounded in this paper and cite their sources.
Cymbopogon citratus (lemongrass) essential oil (EO), widely used in food and pharmaceutical sectors, is often adulterated with low‑cost vegetable oils and synthetic chemicals, compromising quality and safety. Raman spectroscopy integrated with variable selection methods and machine learning algorithms was employed to detect adulterants including castor, groundnut, sunflower oils, polyethylene glycol‑400, and diethyl phthalate. Variable selection methods-Monte Carlo Uninformative Variable Elimination (MC‑UVE), Random Frog, and Successive Projection Algorithm (SPA) were applied. Principal Component Analysis (PCA) revealed that SPA efficiently identified variables capturing 90.1% variance with improved interpretability. Random Forest classification achieved consistently high accuracy, with SPA yielding the best independent test set accuracy (97.6%) and balanced accuracy (97.9%), significantly outperforming MC‑UVE and Random Frog. Extreme Gradient Boosting (XGBoost) further confirmed the importance of variable selection, with SPA‑XGBoost achieving the highest independent test accuracy (98.8%) and balanced accuracy (92.5%), demonstrating superior per‑class generalisation even for challenging adulterants such as groundnut and sunflower oils. Support Vector Machine Regression (SVR) provided excellent quantitative prediction of adulterant concentrations, with calibration, validation, and prediction R 2 values consistently above 0.98 and low RMSEP values (0.03-0.04). Random Frog and MC‑UVE offered marginal improvements in reducing prediction errors, confirming SVR’s robustness and precision. Overall, this study establishes a proof‑of‑concept framework demonstrating that Raman spectroscopy combined with chemometric and machine learning approaches enables rapid, non‑destructive, and reproducible authentication of Cymbopogon citratus EO. While models showed excellent stability under controlled conditions, future work should incorporate broader sample heterogeneity and multi‑adulterant systems to enhance industrial applicability.
No takes yet. Share an insight, caveat, or question.
Jayasekher et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: