基于多元分类判别方法对普洱生茶进行产地溯源研究

目的 探究多元分类判别方法对普洱茶产地溯源的有效性与可行性。方法 基于云南5个产地的60份普洱茶生茶样本,利用5种有机成分以及36种矿物元素的含量,结合主成分分析、判别分析和多层感知器分析等多元分类方法构建普洱茶产地溯源模型。结果 普洱茶(生茶)中总计33项指标具有与产地的显著相关性。不同产地的普洱茶样本在主成分上有差异,但分类的准确性较低。通过判别分析得到了普洱茶产地溯源的判别函数和组质心函数,该样本溯源正确率高于98%。通过多层感知器分析,该方法的分类正确率最高,生茶样本的训练集和测试集分类正确率分别为100.0%和95.5%。结论 普洱茶中有机成分与矿物元素受环境因素影响,在各自产地间形成各自的指纹特征,多维组合可用于建立普洱茶产地溯源模型,为普洱茶产地追溯提供一定的方法支持。

Origin traceability study of raw Pu-erh tea based on multivariate classification and discrimination method

Objective Exploring the effectiveness and feasibility of a multivariate classification and discrimination approach to trace the origin of Pu''er tea. Methods Based on 60 raw tea samples from five production areas in Yunnan, a multivariate classification method, including principal component analysis, discriminant analysis and multilayer perceptron analysis, was used to construct a traceability model for the origin of Pu-erh tea using the contents of five organic components and 35 mineral elements. Results The results showed that a total of 33 indicators in Pu-erh tea (raw tea) were significantly correlated with their origins. The samples of Pu-erh tea from different origins differed in terms of principal components, but the accuracy of classification was low. The discriminant function and group mass center function for the origin traceability of Pu-erh tea were obtained through discriminant analysis, and the correct traceability rate of this sample was higher than 98%. The method had the highest classification accuracy through multi-layer perceptron analysis, with 100% and 95% classification accuracy for the training and validation sets of raw tea samples, respectively. Conclusion Organic components and mineral elements in Pu-erh tea are influenced by environmental factors and form their own fingerprint characteristics among their respective origins, therefore, the multi-dimensional combination can be used to build a traceability model for the origin of Pu-erh tea. To provide some methodological support for tracing the origin of Pu-erh tea.