Ethanol Concentration Determination in Baijiu by Graph-Regularized PCA and Random Forest-Based Raman Spectroscopy

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Baijiu, a traditional Chinese spirit, has significant cultural and economic value.
• Accurate ethanol concentration determination is crucial for baijiu quality control.
• Current Raman spectroscopy methods face limitations in accuracy and robustness for biochemical analysis.

Purpose of the Study:

• To develop an advanced method for precise ethanol concentration determination in baijiu using Raman spectroscopy.
• To enhance the flexibility and robustness of chemometric tools for spectral data analysis.
• To improve the industrial quality control process for baijiu manufacturing.

Main Methods:

• Integration of graph-regularized principal component analysis (graph-regularized PCA) for dimensionality reduction and data stabilization.
• Application of a random forest ensemble learning framework for improved spectral data representation.
• Development of a single regression model protocol using diverse ethanol concentration training sets for multi-type baijiu analysis.

Main Results:

• The proposed method achieved a mean average percentage error (MAPE) of 0.415% for ethanol concentration determination across three types of baijiu.
• The combined graph-regularized PCA and random forest approach significantly reduced spectral data instability.
• The developed protocol demonstrated superior accuracy and robustness compared to existing methods.

Conclusions:

• The novel method offers a highly accurate and robust solution for ethanol concentration measurement in baijiu.
• This technique holds significant potential for advancing quality control in the baijiu industry.
• The findings validate the effectiveness of combining advanced chemometric and machine learning techniques for spectral analysis.