An evaluation of different bio-inspired feature selection techniques on multivariate calibration models in...

Area of Science:

• Chemometrics
• Computational Intelligence
• Spectroscopic Data Analysis

Background:

• Variable selection is crucial for building robust chemometric models.
• Swarm intelligence algorithms offer potential for optimizing complex data analysis tasks.
• Existing methods for variable selection in spectroscopy can be computationally intensive.

Purpose of the Study:

• To introduce and evaluate grey wolf optimization (GWO) and antlion optimization (ALO) as novel variable selection tools for spectroscopic data.
• To compare the performance of GWO and ALO against established algorithms like genetic algorithm (GA), particle swarm optimization (PSO), and firefly algorithm (FFA).
• To assess the effectiveness of these algorithms in constructing partial least squares (PLS) regression models.

Main Methods:

• Application of GWO, ALO, FFA, GA, and PSO algorithms for variable selection on UV and IR spectroscopic datasets.
• Development of PLS regression models using variables selected by each algorithm.
• Comparison of model performance using selected variables against models built with full spectral data.

Main Results:

• GWO and ALO algorithms successfully identified relevant variables for spectroscopic data analysis.
• These novel algorithms selected fewer variables compared to GA and PSO in most cases.
• The PLS regression models built using variables selected by GWO and ALO demonstrated comparable performance to those using full spectral data.

Conclusions:

• Grey wolf optimization (GWO) and antlion optimization (ALO) are effective and efficient tools for variable selection in spectroscopic data analysis.
• These algorithms offer a promising alternative to existing methods, potentially reducing model complexity without sacrificing predictive performance.
• The study validates the utility of GWO and ALO for advancing chemometric modeling techniques.