This study presents a novel method for the quantitative detection of residual chlorpyrifos in corn oil through Raman spectroscopy using a combined long short-term memory network (LSTM) and convolutional neural network (CNN) architecture. The QE Pro Raman+ spectrometer was employed to collect Raman spectra of corn oil samples with varying concentrations of chlorpyrifos residues. A deep-learning model based on LSTM combined with a CNN structure was designed to realize feature self-learning and model training of Raman spectra of corn oil samples. In the study, it was discovered that the LSTM-CNN model has superior generalization performance compared to both the LSTM and CNN models. The root-mean-square error of prediction (RMSEP) of the LSTM-CNN model is 12.3 mg·kg-1, the coefficient of determination (RP2) is 0.90, and the calculation of the relative prediction deviation (RPD) results in a value of 3.2. The study demonstrates that the deep-learning network based on an LSTM-CNN structure can achieve feature self-learning and multivariate model calibration for Raman spectra without preprocessing. The results of this study present an innovative approach for chemometric analysis using Raman spectroscopy.
Keywords: Raman spectroscopy; chlorpyrifos; corn oil; deep learning; quantitative analysis.