基于灰狼算法的近红外光谱变量选择方法研究

基于群体智能的灰狼优化(GWO)算法具有参数少、结构简单、易于实现的优点,但在光谱领域的应用较少。该研究将GWO算法引入近红外光谱的变量筛选中,以玉米数据为例,考察了GWO算法中狼群性能、迭代次数、狼群数量及运算效率,并建立了偏最小二乘(PLS)模型对玉米样品中蛋白质、脂肪、水分以及淀粉含量的测定。结果显示,GWO算法运算效率很高,经过参数调优后建立PLS模型,其蛋白质、脂肪、水分及淀粉的保留变量数分别为19、19、14、34,预测均方根误差(RMSEP)从全波长PLS建模的0.245 8、0.122 4、0.339 8、1.105 8分别下降到0.147 7、0.080 1、0.176 2、0.739 8,分别下降了40%、35%、48%、33%,相关系数也相应地提高。因此,GWO算法不仅优化速度快,选择变量数少,还可以显著提高PLS模型的预测精度,是一种近红外光谱变量选择的有效方法。

A Variable Selection Method for Near Infrared Spectroscopy Based on Gray Wolf Optimizer Algorithm

Gray wolf optimizer(GWO) algorithm, which is based on swarm intelligence, is easy to implement due to its few parameters and simple structure. However, to our knowledge, few studies used GWO for the spectral analysis. In this study, the GWO was introduced into the variable selection of NIR spectra.Taking corn dataset as an example,the performance,numbers of iterationsnumbers of wolves and efficiency of GWO algorithm were investigated.Based on this,a partial least squares(PLS) model was established to determine the protein,fat,moisture and starch contents in corn samples.Results showed that GWO algorithm was very efficient.With optimized parameters,the retention variable numbers of GWO algorithm for protein,fat,moisture and starch were 19,19,14 and 34,respectively.Compared with root mean square error of prediction(RMSEP) values of the full wavelength PLS model for the four components,those of the GWO-PLS model decreased from 0.245 8,0.122 4,0.339 8 and 1.105 8 to 0.147 7,0.080 1,0.176 2 and 0.739 8,with their decreasing percentages of 40%,35%,48% and 33%,respectively.Meanwhile,the correlation coefficients were increased accordingly.Therefore,GWO algorithm could improve the prediction accuracy of the PLS model apparently with high efficiency and fewer selected variables.It is a promising method for variable selection of NIR spectroscopy.