数据融合法结合决策树优化东方草莓全草的地理溯源性研究

Data fusion method to improve geographical traceability of Fragaria orientalis Lozinsk.whole herb

目的:通过融合HPLC与ICP-MS的分析结果建立数据集,结合模式识别法对来自5个产地96批次的藏药东方草莓进行地理溯源性研究,以期为未知来源的东方草莓提供便捷有效的产地判别方法。方法:采用HPLC建立东方草莓指纹图谱,并以LC-MS对各色谱峰进行成分归属,色谱峰信息作为数据集1;采用ICP-MS对东方草莓中21种无机元素进行含量测定,并以对数法建立无机元素指纹图谱,无机元素信息作为数据集2;二者数据融合作为数据集3。结合CA、PCA、PCA-LDA及C5.0决策树算法,对比3个数据集的地理追溯结果。结果:数据融合法相较于单一技术所得数据集,CA、PCA和PCA-LDA均可使96批东方草莓成功归类。对数据融合法的PCA-LDA模型进行内部验证和外部验证,其正确分类率均大于88.3%,表明该模型可作为东方草莓的正确分类依据。C5.0决策树筛选出4个主要贡献变量,并获得树深度为3的分类规则,其十折交叉验证结果显示平均准确度为98.9%。结论:数据融合法可提高不同产地东方草莓地理溯源的准确性,且决策树算法得出的分类规则可降低实际操作难度。以数据融合法结合决策树规则所得出的东方草莓地理分类机制为未知样品的产地来源提供新的鉴别依据,且操作简便,准确率高,有利于东方草莓的质量控制,也为其他药材品种溯源提供参考。

Objective:To combine the data of HPLC and ICP-MS with the single technology data with the pattern recognition method to conduct a geographic tracing study of 96 batches of Fragaria orientalis Lozinsk.from 5 procinces,in order to provide a convenient and effective identification method for the unknown origin of Fragaria orientalis Lozinsk..Methods:The fingerprint of Fragaria orientalis Lozinsk.was established by HPLC,thecomponents of each chromatographic peak were assigned by LC-MS,and the chromatographic peak information was used as data set 1.ICP-MS was used to determine the contents of 21 inorganic elements in Fragaria orientalis Lozinsk.,the logarithm Inorganic element fingerprint map was established by the method,and the inorganic element information was used as data set 2;the fusion of the two data was used as data set 3.Combine CA,PCA,PCA-LDA and C5.0 decision tree algorithms to compare the geographic traceability results of the three data sets.Results:Compared with the data set obtained by a single technique,CA,PCA and PCA-LDA of the data fusion method can correctly classify 96 batches of Fragaria Orientalis Lozinsk..The PCA-LDA model of the data fusion method was verified internally and externally,and the correct classification rate was greater than 88.3%,The results showed that the model could be used as the correct basis for the classification of Fragaria Orientalis Lozinsk..The C5.0 decision tree screened out 4 main contributing variables and obtained a classification rule with a tree depth of 3.The 10-fold cross-validation result showed that the average accuracy was 98.9%.Conclusion:The data fusion method can improve the accuracy of the geographical traceability of Orientalis Lozinsk.from different origins,and the classification rules derived from the decision tree algorithm can reduce the difficulty of actual operation.The geographical classification mechanism of Orientalis Lozinsk.based on the data fusion method and decision tree rules provides a new identification basis for the unknown sample origin,with simple operation and high accuracy,which is beneficial to the quality control.This study also provided reference for other medicinal materials.