Equivalent and Complementary Variables Screening for the Optimization of Wavelengths in Spectral Multivariate...

Area of Science:

• Chemometrics
• Spectroscopy
• Data analysis

Background:

• Multivariate calibration relies on effective variable selection.
• Identifying equivalent and complementary variables can enhance model robustness and predictive power.

Purpose of the Study:

• To introduce and validate a novel strategy for variable selection using equivalent variables (EVs) and complementary variables (CVs).
• To assess the effectiveness of this strategy in improving multivariate calibration models across various spectral data types (NIR, MIR, UV-vis).

Main Methods:

• Utilized three variable selection algorithms: Stability Competitive Adaptive Reweighted Sampling (SCARS), Competitive Adaptive Reweighted Sampling (CARS), and Monte Carlo and Uninformative Variable Elimination (MC-UVE).
• Screened for EVs and CVs from NIR, MIR, and UV-vis spectral datasets.
• Evaluated model performance using Root Mean Square Error of Calibration (RMSEC) and Root Mean Square Error of Prediction (RMSEP).

Main Results:

• Identified 54 EVs from corn NIR spectra using SCARS, demonstrating comparable modeling performance to basic variables (BVs) with minimal prediction error deviation (<0.003 RMSEP).
• Screened 15 CVs from the EVs of CARS and MC-UVE, which significantly improved SCARS models when combined with BVs (RMSEC decreased from 0.0207 to 0.0109, RMSEP from 0.0290 to 0.0136).
• Consistent performance improvements were observed across other spectral datasets.

Conclusions:

• The developed strategy of using EVs and CVs offers an effective approach to optimize variable selection in multivariate calibration.
• Screening CVs from EVs of different algorithms and combining them with BVs demonstrably enhances model performance and predictive accuracy.