Determination of Chlorpyrifos in Pears by Raman Spectroscopy with Random Forest Regression Analysis

Pesticide detection has been a long-running concern within the agricultural industry. Therefore, rapid and accurate detection techniques are required. Raman spectroscopy is favored by scientists due to the rapid detection without the need for the pretreatment or destruction of samples. However, the accuracy of the quantitative analysis that is based on Raman spectra is related to the detection model and analysis algorithm. In this study, random forest regression (RFR) was employed to construct a quantitative detection model that correlated the Raman spectral intensity of chlorpyrifos at 341 cm⁻¹ and the chlorpyrifos concentration remaining on the pear surface. RFR performs better than partial least square regression (PLSR). The correlation coefficient (R²) of RFR was 0.9003 and 0.8495 for the training and test sets, respectively. In particular, the R² value of the test sets was significantly higher than for PLSR (R² of 0.6985). The intensity of the Raman spectra at 341 cm⁻¹ was improved by four-fold using 100 nm gold nanoparticles. The results show that Raman spectroscopy combined with RFR achieves the rapid and accurate quantitative determination of pesticide residues.