近红外光谱和支持向量机用于凌霄花产地鉴别

收集6个产地凌霄花样品的近红外光谱，构建支持向量机(SVM)模型进行产地鉴别.运用竞争自适应重加权采样(CARS)算法提取特征波长变量，在此基础上建立CARS-SVM产地判别模型.将该判别模型与线性判别分析、偏最小二乘判别分析和簇类独立软模式法3种模型进行比较.结果表明，SVM模型对不同产地凌霄花样品的鉴别结果良好，经CARS提取特征波长后，波长变量数从1 557减小至52,所构建的CARS-SVM模型对6个产地样品的判别准确率较高，明显优于上述3种模型.因此，近红外光谱技术可快速准确判别凌霄花的产地，为凌霄花的产地鉴别与质量评价提供一种新的方法.

Geographical origin discrimination of Campsis grandiflora by near-infrared spectroscopy coupled with support vector machine

Campsis grandiflora (Thunb.) K. Schum is a significant traditional Chinese medicine with efficacy in cerebrovascular diseases, gynecological diseases and inflammation. However, the chemical components of Campsis grandiflora from different geographical origins show significant differences which affect the efficacy greatly. Firstly, the near infrared (NIR) spectra of 271 samples from six geographical origins were collected and analyzed in this study. Support vector machine (SVM), was used to establish a model. Secondly, competitive adaptive reweighted sampling (CARS) was applied to extract characteristic wavelength variables. Finally, the CARS-SVM model was established and compared with the linear discriminant analysis (LDA), partial least squares discriminant analysis (PLS-DA) and soft independent modelling of class analogy (SIMCA) to determine the optimal model. The results showed that SVM model had a good discriminant effect on Campsis grandiflora from different origins. The discriminant rate of training set was 98.36%, and that of prediction set was 96.30%. When the characteristic wavelength variables were sharply reduced from 1557 to 52 by CARS, the constructed CARS-SVM model can achieve the best classification accuracy (100%), which was obviously superior to that of other three models. Therefore, NIR spectroscopy can quickly and accurately identify the origin of Campsis grandiflora, and provide a certain reference for the geographical origin discrimination and quality assessment of Campsis grandiflora.