Characteristic Wavelength Selection Based on Multi-Strategy Fusion Zebra Optimization Algorithm for Plsr

Abstract

Wavelength selection plays a key role in near-infrared spectral analysis, because it can significantly enhance the generalization capability of multivariate models while simultaneously reducing their complexity. The zebra optimization algorithm is a simple and novel algorithm, which iteratively searches for the optimal solution by mimicking zebra foraging and defense behavior. In this work, the zebra optimization algorithm is applied to the wavelength selection of near-infrared spectra for the first time. And a wavelength selection method based on multi-strategy fusion of zebra optimization algorithm (MFZOA) is proposed for the shortcomings of the ZOA algorithm which converges slowly and easily falls into local optimum. Partial Least Squares Regression (PLSR) models are built using Gray Wolf Optimization Algorithm (GWO), Sparrow Search Algorithm (SSA), Whale Optimization Algorithm (WOA), ZOA algorithms and MFZOA algorithm for four NIR datasets. The experimental results show that MFZOA has better prediction performance compared with other wavelength selection algorithms.

Keywords: Near-infrared spectral analysis, Zebra optimization algorithm, Wavelength selection, Partial least squares regression