Moving window smoothing on the ensemble of competitive adaptive reweighted sampling algorithm.

Area of Science:

• Chemometrics
• Spectroscopy
• Analytical Chemistry

Background:

• Multivariate calibration relies on effective spectral variable selection.
• Existing methods like Variable Importance Projection (VIP), Uninformative Variables Elimination (UVE), and Genetic Algorithms (GA) have limitations.

Purpose of the Study:

• To introduce and evaluate a novel spectral variable selection method, Moving Window Smoothing-Ensemble of Competitive Adaptive Reweighted Sampling (MWS-ECARS).
• To assess the performance of MWS-ECARS in improving prediction accuracy for multivariate calibration models.

Main Methods:

• Developed MWS-ECARS by combining Moving Window Smoothing (MWS) with an ensemble of Competitive Adaptive Reweighted Sampling (CARS).
• Applied MWS-ECARS to select variables from mid-infrared (MIR) spectra of pesticide active ingredients, near-infrared (NIR) spectra of soil organic matter, and NIR spectra of total nitrogen in Solanaceae plants.
• Optimized MWS-ECARS by identifying the variable subset with the lowest standard error of cross-validation (SECV) and the optimal moving window width.

Main Results:

• MWS-ECARS effectively selected informative spectral variables.
• The method demonstrated improved prediction performance compared to VIP, UVE, and GA.
• Successful application across diverse spectral datasets (MIR, NIR) and sample types (pesticides, soil, plants).

Conclusions:

• MWS-ECARS is a promising and effective method for spectral variable selection in multivariate calibration.
• The proposed method offers enhanced prediction accuracy over traditional techniques.
• MWS-ECARS provides a robust approach for analyzing complex spectral data.