Multilevel LASSO-based NIR temperature-correction modeling for viscosity measurement of bisphenol-A

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Near-infrared (NIR) spectroscopy is valuable for chemical analysis.
• Temperature fluctuations can significantly degrade the accuracy of NIR predictive models.
• Accurate viscosity measurement of bisphenol-A is crucial in industrial applications.

Purpose of the Study:

• To develop a robust method for correcting temperature effects in NIR viscosity prediction models.
• To enhance the stability and interpretability of NIR models for bisphenol-A viscosity measurement.
• To address the limitations of existing methods in handling temperature variations.

Main Methods:

• A novel multilevel least-absolute shrinkage and selection operator (LASSO) algorithm was proposed.
• The method integrates LASSO with multilevel simultaneous component analysis (MLSCA).
• MLSCA decomposes spectral data into temperature-dependent and concentration-dependent components, while LASSO performs variable selection and regularization.

Main Results:

• The proposed multilevel LASSO model demonstrated superior robustness against temperature variations.
• Experimental results confirmed the effectiveness of the method in accurately predicting bisphenol-A viscosity.
• The new approach provided more stable prediction results compared to models without temperature correction.

Conclusions:

• The multilevel LASSO method effectively corrects for temperature-induced errors in NIR viscosity prediction.
• This approach enhances the reliability and interpretability of NIR analytical models.
• The developed technique offers a significant improvement for bisphenol-A viscosity measurements under varying conditions.