Rapid Detection of Major Components in Shuanghuanglian Oral Liquid by UV-Vis Spectroscopy Combined with Chemometrics

OBJECTIVE To study on a rapid method for determining the quality of Shuanghuanglian oral liquid based on UV-Vis spectroscopy and chemometrics. METHODS The UV-Vis spectrum data of Shuanghuanglian oral liquid were collected, and the principal component analysis was used to screen out the abnormal sample data. All samples were segmented into training sets and test sets in a 7:3 ratio using the Kennard-Stone algorithm. The screened data was pre-processed by first derivation and conventional normal variate transform. Then, data feature extraction was performed by competitive adaptive reweighted sampling. Consequently, support vector regression(SVR), least squares to support vector regression(LS-SVR) and back propagation(BP) neural network was used to establish a quantitative analysis model of soluble solids content(SSC) and total flavones(TF). RESULTS All the models achieved good prediction effect, R² was ≥ 0.816 8, RMSE was ≤ 4.378 2. After comparing the SSC model and the TF model, it was revealed that compared with BP neural network and LS-SVR, the SVR model obtained the highest R² and the lowest RMSE. The R² of the SVR-SSC model was 0.999 8, RMSE was 0.260 3, and R² of SVR-TF model was 0.998 3, RMSE was 0.543 3. CONCLUSION UV-Vis spectrum combined with SVR can provide a high-precision and rapid on-site detection method for the quality of Shuanghuanglian oral liquid.