Integrated surface-enhanced Raman spectroscopy and machine learning for identification and quantification of multi-target pesticide residues in fruit juices.

This study developed a novel SERS platform integrated with machine learning for rapid, sensitive, and discriminative analysis of multi-target pesticide residues in fruit juices. The core-shell Au@AgCNCs nanostructure markedly amplifies Raman signals through localized surface plasmon resonance and calcium-ion-induced aggregation, enabling ultrahigh-sensitivity detection of five pesticides with a limit of detection as low as 2.7 pM. When the Machine Learning (ML) algorithm was applied to multi-target pesticide residue analysis in juice, 99.56 % accuracy in identifying pesticide types within a complex juice matrix was achieved. Notably, the model was trained on pure pesticide spectra and tested on spiked juice samples. Using Support Vector Regression (SVR), recoveries for individual and mixed pesticides ranged from 94.2 % to 106.1 %, enabling quantitative analysis of multi-component pesticide residues. The integration of SERS with ML not only enhanced detection precision but also simplified analytical workflows, demonstrating significant potential for food safety monitoring and sustainable agricultural practices.