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.