201. Discrimination of Radix Paeoniae varieties on the basis of their geographical origin by a novel method combining high-performance liquid chromatography and Fourier transform infrared spectroscopy measurements
- Author
-
Baohui Li, Yongnian Ni, and Serge Kokot
- Subjects
Analyte ,Chromatography ,Basis (linear algebra) ,business.industry ,Chemistry ,General Chemical Engineering ,General Engineering ,Pattern recognition ,Linear discriminant analysis ,High-performance liquid chromatography ,Least squares ,Analytical Chemistry ,Support vector machine ,Radix ,Artificial intelligence ,Fourier transform infrared spectroscopy ,business - Abstract
Performance of a novel analytical method for complex samples, based on combined or fused high-performance liquid chromatography (HPLC) and Fourier transform infrared spectroscopy (FT-IR) data, was compared with that based on measurements from either of the techniques. The analytes were different varieties of Radix Paeoniae (Shaoyao), traditional Chinese medicines (TCM), from various regions. When the HPLC and FT-IR datasets (L and R) were separately submitted to either PCA or several supervised classification and prediction methods, the recognition rates were unsatisfactory. Thus, the potential of combined data of these two techniques was investigated, and two statistical methods for reducing the dimensions of the fused data were developed based on: (i) four PCs from the L and R matrices, and (ii) selection of optimum variables by the genetic algorithm-partial least squares (GA-PLS). PCA of these two fused sets indicated improved discrimination of the different sample clusters. Subsequent application of the supervised classification and prediction methods—linear discriminant analysis (LDA), least squares-support vector machine (LS-SVM), and radial basis function neural network (RBF-ANN)—indicated significantly higher recognition rates. Thus, the best discrimination of the complex Shaoyao TCM samples on the basis of their variety and geographical origin, was obtained with the use of non-linear RBF-ANN and LS-SVP models.
- Published
- 2012