A review on hybrid strategy-based wavelength selection methods in analysis of near-infrared spectral data

• A review of the existing hybrid methods for analyzing near-infrared spectra is presented. • A systematic classification of developed variable selection algorithms based on hybrid strategy is presented. • Suggestions for connecting two or three different algorithms to create new hybrid methods are provided. • The proposals on the development of new hybrid methods are given. Wavelength (variable) selection is considered as a means of solving the curse of dimensionality, which is an important step in multivariate calibration of near-infrared spectra (NIR) and a topic of interest in NIR spectroscopy. A large number of variable selection methods have been developed. They possess diverse advantages and disadvantages because of their different principles and application scope. To overcome their drawbacks and fully utilize their merits, a hybrid strategy, which combines several methods was proposed, and has been proven effective in the selection of informative variables from high-dimensional NIR spectra. Based on the hybrid strategy, many variable selection methods have been developed. To aid people in learning about the various hybrid methods, this paper systematically classifies these methods and elucidates their principles and features. Further, two forms of developing new hybrid methods are suggested based on the existing single variable selection method. Every single method is classified into the stages of the suggested hybrid strategy based on its theory to enable more convenient and efficient utilization. Finally, several proposals are presented to better apply and promote the development of hybrid methods for analyzing NIR spectra in the future.