XGBoost algorithm assisted multi-component quantitative analysis with Raman spectroscopy.

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Raman spectroscopy is crucial for analyzing chemical mixtures.
• Preprocessing is vital to reduce artifacts and improve prediction accuracy in spectral data.
• Existing preprocessing methods may not fully address the complexities of multicomponent spectra.

Purpose of the Study:

• To develop and evaluate an eXtreme Gradient Boosting (XGBoost) preprocessing method for Raman spectra.
• To quantitatively analyze glucose, glycerol, and ethanol concentrations in mixed solutions.
• To assess the performance enhancement of XGBoost preprocessing when combined with Linear Regression (LR) and Multilayer Perceptron (MLP) models.

Main Methods:

• Developed an eXtreme Gradient Boosting (XGBoost) preprocessing technique for Raman spectra.
• Integrated XGBoost preprocessing with Linear Regression (X-LR) and Multilayer Perceptron (X-MLP) models.
• Utilized hyperparameter tuning via proportion maps and evaluated models using R², RMSEC, and RMSEP metrics.

Main Results:

• The XGBoost preprocessing method significantly improved prediction performance and reduced artifacts in Raman spectra.
• Both X-LR and X-MLP models demonstrated enhanced accuracy and generalization capabilities.
• XGBoost preprocessing outperformed other tested methods for both LR and MLP models.

Conclusions:

• The developed XGBoost preprocessing method offers a robust and reliable approach for Raman spectral analysis.
• This technique effectively enhances the quantitative analysis of glucose, glycerol, and ethanol mixtures.
• XGBoost preprocessing represents a significant advancement in spectral data analysis for chemical sensing applications.