An efficient swarm intelligence approach to feature selection based on invasive weed optimization: Application to...

Area of Science:

• Chemometrics
• Machine Learning
• Optimization Algorithms

Background:

• Variable selection is crucial for accurate classification and multivariate calibration.
• Swarm intelligence optimization methods, inspired by nature, are increasingly respected for their adaptability.
• Existing methods may not always be optimal for complex datasets.

Purpose of the Study:

• To propose a novel and simple variable selection algorithm based on Invasive Weed Optimization (IWO).
• To demonstrate the first application of IWO for variable selection in chemometrics.
• To evaluate the performance of IWO in classification and multivariate calibration tasks.

Main Methods:

• Development of a new variable selection algorithm inspired by the ecological behavior of invasive weeds.
• Application of the Invasive Weed Optimization (IWO) algorithm to select relevant variables.
• Integration of IWO with Linear Discriminant Analysis (LDA) and Partial Least Squares (PLS) for classification and calibration, respectively (IWO-LDA and IWO-PLS).

Main Results:

• The proposed IWO-based algorithm effectively selects relevant variables for classification and calibration.
• Demonstrated successful application of IWO to FTIR and NIR experimental datasets.
• Achieved robust performance in both classification and multivariate calibration tasks.

Conclusions:

• Invasive Weed Optimization (IWO) is a simple, powerful, and adaptive metaheuristic for variable selection.
• IWO-LDA and IWO-PLS offer effective solutions for multivariate classification and calibration problems.
• This work highlights the potential of bio-inspired algorithms in chemometric data analysis.