灰度关联分析结合支持向量机用于近红外光谱研究

灰度关联分析是通过关联度的计算来理清系统中各因素之间的主次关系，找出影响较大的因素。简述了灰度关联分析的基本原理，并利用其对180个烟草样品的近红外谱进行了谱区优化，选取其中120个样品用于建模，另外60个样品用于模型检验。进一步利用偏最小二乘法和径向基支持向量机法分别建立了烟草样品的总糖、还原糖、烟碱及总氮的定量分析模型。结果表明，将灰度关联分析与支持向量机法联合用于烟草近红外光谱四个组分的定量分析，其模型的泛化能力和预测精度均有较明显的提高，从而能够有效地提高建模效率。

Application of Grey Correlation Analysis with Support Vector Machine in Near-Infrared Spectroscopy

The foundation of grey relational analysis is to clarify the primary and secondary relationship among different factors in the system through calculating correlation degree, and find out the influential factors. In the present study, the near-infrared spectra of 180 tobacco samples were determined. Among them, 120 samples were used for modeling and 60 samples were used for model checking. Then the quantitative analysis models of the tobacco samples, corresponding to total sugar, reducing sugar, nicotine and total nitrogen, were established using the partial least squares method and radial basis of support vector machine method. The experimental results show that, the grey correlation analysis with support vector machine method was used in the quantitative analysis of four tobacco components by near infrared spectroscopy, the generalization ability of the models and the prediction precision are obviously improved, which can effectively enhance the modeling efficiency.