目的 利用稳定同位素质谱测定6个产区的竹节参Panax Japonicus中碳、氮、氢、氧4种同位素比率。方法 采用基于机器学习分类器的线性判别法（LD），高斯核支持向量机（SVM）和基于神经网络工具箱的模式识别反向传播学习算法（BPN）对竹节参的产地进行判别。结果 结果表明，稳定同位素碳（d¹³C）具有明显的地域特征，可有效地区分竹节参产地；LD法和BPN法均能区分6个产地的竹节参，判定准确率均达100%。结论 基于稳定同位素技术并结合LD和BPN法能有效进行竹节参的产地溯源。

Objective The isotopic ratios of light elements (C/N/H/O) in Panax japonicus from six producing areas were determined with isotope ratio mass spectrometry.Methods Three methods, including linear discrimination (LD), gaussian kernel support vector machine (SVM), and the back-propagation learning algorithm of pattern recognition based on neural network toolbox (BPN) were employed to establish a model for P. japonicus geographical origin discrimination. Results The results showed that stable isotope carbon (d ¹³C) had obvious regional characteristics, which will be used to effectively distinguish the origin of P. japonicus. The methods of LD and BPN could classify the geographical origin of P. japonicus from six producing areas, both of which showed that the accuracy rates were 100% using training dataset. Conclusion Therefore, the stable isotope technique combined with LD and BPN method can effectively trace the origin of P. japonicus.

R286

湖北省宜昌市应用基础研究项目（A18-302-a12）