Detection and classification of lambda-cyhalothrin and iprodione residues in tobacco leaves by SERS combined with supervised machine learning

Surface-enhanced Raman spectroscopy (SERS) is a promising technique for the rapid detection of trace pesticide residues in agricultural produce. Currently, research is predominantly centered on utilizing SERS to discern residues characterized by non-overlapping spectral peaks. Yet, the technology encounters challenges in accurately identifying substances when their Raman spectral peaks overlap. In this paper, a silver nanoparticles (Ag Nps) modified SiO₂ nanoparticles (SiO₂ Nps) arrays composite SERS substrate (SiO₂@Ag Nps-SERS substrate) was prepared for detecting lambda-cyhalothrin (LCT) and iprodione (IPR) residues in tobacco leaves, with detection limits of 0.1 ppm and 1 ppm respectively. And then, supervised machine learning models were used to identify and classify LCT and IPR residues with overlapped Raman characteristic peaks at 1000 cm⁻¹. Support Vector Machine (SVM), Partial Least Squares Discriminant Analysis (PLS-DA) and Decision Tree (DT) models could all accurately classify the spectral data of LCT and IPR residues, and the classification accuracy of SVM model was the highest, reaching 99.74 %. This capability is a testament to the innovative strength of our approach, highlighting its potential for universal applicability in spectral data analysis.