Two-Step Partial Least Squares-Discriminant Analysis Modeling for Accurate Classification of Edible Sea Salt Products...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Laser-induced breakdown spectroscopy (LIBS) is a key technique for material classification.
• Partial least squares-discriminant analysis (PLS-DA) is a common chemometric method for LIBS data analysis.
• PLS-DA model accuracy can be limited by the number and similarity of classes.

Purpose of the Study:

• To develop and validate a two-step PLS-DA modeling strategy.
• To improve classification accuracy for LIBS spectra, especially for similar classes.
• To address the sensitivity of PLS-DA to class characteristics.

Main Methods:

• A two-step PLS-DA modeling approach was proposed.
• The strategy involved initial sorting into broader classes and a subsequent focused classification of confusing classes.
• The method was applied to classify six commercial edible sea salts using LIBS spectra.

Main Results:

• The two-step PLS-DA significantly improved classification accuracy compared to a single-step approach.
• The strategy effectively maximized differences between highly similar classes.
• Accurate classification of six distinct sea salt origins (Japan, South Korea, France) was achieved.

Conclusions:

• The proposed two-step PLS-DA modeling is a robust strategy for enhancing LIBS-based material classification.
• This approach offers a significant improvement for datasets with challenging, similar classes.
• The method demonstrates practical utility in distinguishing geographically diverse edible sea salts.