基于三种监督分类模型的中药材产地鉴别

Identification of Origin of Traditional Chinese Medicine Based on Three Supervised Classification Models

为快速鉴别中药材产地,基于中红外光谱数据,采用导数光谱法(一阶导数光谱法、二阶导数光谱法)、标准正态变量变换和多元散色校正等预处理方法并依据标准差法提取特征波段。利用线性判别分析、支持向量机、集成学习三种有监督分类模型与处理后的数据进行交叉组合鉴别。结果显示,二阶导数预处理方法与集成学习模型为最优组合,其训练集准确率达99.4%,检验集准确率达100%。该方法能准确地鉴别中药材产地,且速度快、成本低,为中药材的有效鉴别提供了借鉴。

In order to identify the origin of traditional Chinese medicine quickly,based on the mid infrared spectral data,derivative spectroscopy(first derivative spectroscopy,second derivative spectroscopy),standard normal variable transformation and multicolor dispersion correction were used as preprocessing methods,and the characteristic bands were extracted according to the standard deviation method.Linear discriminant analysis,support vector machine and ensemble learning was used to identify the cross combination between the three supervised classification models and the processed data.The results show that the second derivative preprocessing method and the ensemble learning model was the best combination,and the accuracy of the training set reached to 99.4%and the accuracy of the test set reached to 100%.This method could accurately identify the origin of traditional Chinese medicine with high speed and low cost,which provided a reference for the effective identification of traditional Chinese medicine.