Quantitative Detection of Acid Value During Edible Oil Storage by Raman Spectroscopy: Comparison of the Optimization Effects of BOSS and VCPA Algorithms on the Characteristic Raman Spectra of Edible Oils

Acid value is one of the crucial indicators reflecting the quality of edible oil during storage. This study proposed a quantitative detection method for edible oil acid value during storage based on reduced Raman spectral characteristics. First, the Raman spectra of edible oil samples from different storage periods were collected, and the original spectra were preprocessed using Savitzky-Golay smoothing (SG smoothing), multiple scattering correction (MSC), adaptive iteratively reweighted penalized least squares (airPLS) combined with normalization. Then, bootstrap soft shrinkage (BOSS) and variable combination population analysis (VCPA) methods based on model population analysis (MPA) strategy were introduced to optimize Raman spectral characteristics. Finally, the SVM prediction model based on the optimized characteristic Raman shifts was established to realize the quantitative detection of edible oil acid value during storage. The results showed that compared with the fidl-spectrum SVM model, the BOSS-SVM and VCPA-SVM models had no overfitting during the training process, and the VCPA-SVM model obtained the best generalization performance, and its RI and RMSEP in the prediction set were 0.932 and 0.153, respectively. This study sufficiently demonstrated that the SVM model integrated with the characteristic Raman shifts which were optimized by the BOSS and VCPA was a promising tool to monitoring the acid value in edible oil during the storage, and the characteristic Raman shifts of Raman spectra optimized by the VCPA was of more interpretation and pertinence.