Abstract: Setting the 5 Gentiana species-G. straminea, G. dahurica, G. macrophylla, G. crassicaulis, G. officinalis-as the research objects. Near-infrared combined with chemometrics were used to build two discriminant models. Model 1 was to discriminant authentic Radix Gentiana Macrophyllae from falsify;model 2 was to distinguish 5 different species. And the cluster analysis was also proceeded to make a comparison. Model 1 was taken to optimize modeling approaches and spectral processing methods. Finally, two models were established by discriminant analysis, with no-smoothing first-derivative spectra. And the recognition rates and prediction rates of two models were all 100%. The NIR absorbance was analyzed by cluster analysis. The result showed that cluster analysis could discriminant authentic gentiana from adulterants exactly, but could not identify 5 different species. The analysis indicated that discriminant models have more practicability and accuracy. They can provide a rapid and scientific method on clinical medication safety and effectiveness of Radix Gentiana Macrophyllae.

Key words: G. straminea, G. dahurica, G. macrophylla, G. crassicaulis, G. officinalis, near-infrared discriminate model