In this paper, a new spectral variables selection method, induced mutation genetic algorithm (IMGA), is proposed for near-infrared (NIR) spectroscopy. Based on the idea of genetic algorithm (GA), the IMGA greatly simplifies the process of biological evolution, which not only inherits the advantages of global optimization of the GA, but also effectively improves the convergence speed. In this study, the IMGA is applied to the selection of characteristic spectral variables for green tea origin identification. After five times of genetic evolutions, 11 characteristic spectral variables are selected from 156 spectral variables. Based on the 11 characteristic spectral variables, the classification model is built by partial least squares (PLS), and both the sensitivity and specificity of classification model are raised to 1 for prediction set. The overall results indicate that the IMGA can be well applied to the selection of characteristic spectral variables and effectively improve the prediction accuracy and calculation speed of the near-infrared model.

中文翻译：

---

近红外光谱中光谱变量选择的新型诱变遗传算法

本文提出了一种新的光谱变量选择方法，即诱导突变遗传算法（IMGA），用于近红外光谱。IMGA基于遗传算法（GA）的思想，大大简化了生物进化的过程，不仅继承了GA全局优化的优点，而且有效提高了收敛速度。在这项研究中，IMGA被应用于特征光谱变量的选择，以用于绿茶产地鉴定。经过五次遗传进化，从156个光谱变量中选择了11个特征光谱变量。基于11个特征光谱变量，通过偏最小二乘（PLS）建立分类模型，并将分类模型的灵敏度和特异性都提高到1。