论文部分内容阅读
目的:分析鉴别4个产地川东獐牙菜,并建立预测模型,预测产地区分准确性。方法:光谱数据导入UVProbe2.34,比较不同产地相同部位的紫外光谱图,将原始光谱数据以及经过8点平滑、一阶求导和二阶求导后的数据导入SIMCA-P11.5,进行主成分分析(PCA),比较三维得分图的产地鉴别效果。结果:主成分分析中以叶的原始数据以及8点平滑处理数据鉴别效果最佳,主成分累计贡献率均为98.8%,其余预处理方式无法取得较好的鉴别效果可能与主成分数累计值有关(一阶求导为83.9%,二阶求导为47.3%)。根部数据能将重庆、湖北的样品和湖南样品分开,但重庆和湖北的样品无法区分。建立偏最小二乘判别分析(PLS-DA)模型,检测鉴别模型的可靠性,并为预测更多产地的区分提供依据。将验证集带入训练集建立的模型进行偏最小二乘判别分析,能区分产地,证明该模型产地鉴别效果可行。PLS-DA中训练集的预测值和真实值相关系数为0.985,其评估均方差(RMSEE)为0.159,验证集导入训练集后其预测值与真实值的相关系数为0.927,预测均方差(RMSEP)为0.327,RMSEE与RMSEP两者相近,且都<0.500,该模型的预测可靠性高。结论:运用紫外光谱结合主成分分析和偏最小二乘判别分析能够较好的鉴别不同产地川东獐牙菜,构建模型预测效果较好,加入未知产地样品也能较好区分。
OBJECTIVE: To identify and discriminate Swertia mandshurica from four producing areas and establish a predictive model to predict the accuracy of producing areas. Methods: The spectral data were imported into UVProbe2.34. The UV spectra of the same region from different regions were compared. The original spectral data and the data after 8-point smoothing, first-order derivative and second-order derivative were introduced into SIMCA-P11.5. Component Analysis (PCA), comparing the three-dimensional scoring map of origin identification effect. Results: In the principal component analysis, the leaf original data and the 8-point smoothing data were the best, the main components cumulative contribution rates were 98.8%, the other pretreatment methods can not get better identification results may be related to the cumulative value of the principal component (The first derivative is 83.9% and the second derivative is 47.3%). Root data separates samples from Chongqing and Hubei from Hunan samples, but samples from Chongqing and Hubei are indistinguishable. The PLS-DA model was established to test the reliability of the discriminant model and provide a basis for the prediction of more producing areas. The verification set is introduced into the model set up by the training set for partial least-squares discriminant analysis, which can distinguish the provenance and prove that the model is feasible for discriminating the origin of the model. The correlation coefficient between the predicted value and the true value of the training set in PLS-DA was 0.985, and the RMSEE of the training set was 0.159. The correlation coefficient between the predicted value and the true value of the training set was 0.927, the mean square error of prediction (RMSEP) ) Is 0.327. RMSEE and RMSEP are both close to <0.500, and the prediction reliability of this model is high. Conclusion: UV-Vis spectroscopy combined with principal component analysis and partial least-squares discriminant analysis can better discriminate Swertia mandshurica from different origins. The prediction model is better and the samples with unknown origin can be well differentiated.