Research on the Quantitative Analysis Model of Ethyl Lactate in Raw Wine Based on Near Infrared Spectroscopy

Ethyl lactate is one of the important indicators of important aroma substances in liquor, and it is very important to quickly detection the content of ethyl lactate in the brewing process. Adopt partial least square method, support vector regression, bp neural network to establish the quantitative analysis model of ethyl lactate, The best model was selected through the comprehensive evaluation model of prediction determination coefficient(R2), and prediction root mean square error (RMSEP). Combined with different preprocessing methods to study the influence of preprocessing methods on the model. Then compare the Uninformative Variable Elimination (UVE) and Competitive Adaptive Reweighted Sampling (CARS) feature screening results to determine the best screening method. Finally, the best combination of preprocessing, feature screening, and modeling methods for ethyl lactate was determined to be SG13+UVE+SVR. R2 is 0.9745 and RMSEP is 0.0064, Experiments prove that the performance of each model can be effectively improved by preprocessing methods and feature wave screening. A rapid non-destructive detection method was provided for ethyl lactate in raw wine.