基于拉曼光谱的平性中药辨识研究
投稿时间:2022-04-18     点此下载全文
引用本文:纪徐维晟,梁浩,刘淑明,王献瑞,王耘.基于拉曼光谱的平性中药辨识研究[J].中国现代中药,2022,24(12):2364-2370
DOI:10.13313/j.issn.1673-4890.20220418005
摘要点击次数: 703
全文下载次数: 0
              
作者中文名作者英文名单位中文名单位英文名E-Mail
纪徐维晟 JI Xu-wei-sheng 北京中医药大学 中药学院 中药信息工程研究中心,北京 102488
北京中医药大学 生命科学学院,北京 102488
Research Center of TCM-Information Engineering, School of Chinese Materia Medica, Beijing University of Chinese Medicine, Beijing 102488, China
School of Life Sciences, Beijing University of Chinese Medicine, Beijing 102488, China
 
梁浩 LIANG Hao 北京中医药大学 中药学院 中药信息工程研究中心,北京 102488 Research Center of TCM-Information Engineering, School of Chinese Materia Medica, Beijing University of Chinese Medicine, Beijing 102488, China  
刘淑明 LIU Shu-ming 北京中医药大学 中药学院 中药信息工程研究中心,北京 102488 Research Center of TCM-Information Engineering, School of Chinese Materia Medica, Beijing University of Chinese Medicine, Beijing 102488, China  
王献瑞 WANG Xian-rui 北京中医药大学 中药学院 中药信息工程研究中心,北京 102488 Research Center of TCM-Information Engineering, School of Chinese Materia Medica, Beijing University of Chinese Medicine, Beijing 102488, China  
王耘* WANG Yun 北京中医药大学 中药学院 中药信息工程研究中心,北京 102488 Research Center of TCM-Information Engineering, School of Chinese Materia Medica, Beijing University of Chinese Medicine, Beijing 102488, China  
基金项目:国家自然科学基金项目(81973495)
中文摘要:目的 秉承中医药整体观念,开展平性药拉曼谱图与平性的相关性分析,实现平性辨识。方法 以中药拉曼光谱数据为分类指标,基于平均基尼系数降低度开展特征拉曼数据筛选,结合随机森林(RF)、支持向量机(SVM)、逻辑回归(LR)、人工神经网络(ANN)、贝叶斯网络(NN)算法建立平性辨识模型,同时采用交叉验证法对建立的辨识模型进行分析。结果 与SVM、LR、ANN、NN模型相比,以基尼系数排名前100的拉曼数据所建立的RF模型展现出最佳的辨识效果,准确度和精确度均大于0.93,受试者工作特征曲线下面积(AUC)>0.95。结论 平性药拉曼谱图与平性之间具有显著相关性,可作为药性表征,结合随机森林算法高效、准确地辨识平性中药,为平性的客观存在、可入性味提供有力依据。
中文关键词:中药  拉曼光谱  药性  随机森林  特征筛选
 
Identification of Neutral Chinese Medicine Based on Raman Spectroscopy
Abstract:Objective To identify neutral Chinese medicine based on the correlation analysis between Raman spectra and neutral properties from the holistic perspective of Chinese medicine.Methods The Raman spectral data of Chinese medicine were used as the classification index, and the quantified Raman data of Chinese medicines were subjected to feature screening by the Gini coefficient. Then the identification models of neutral properties were established based on random forest (RF), support vector machine (SVM), logistic regression (LR), artificial neural network (ANN), and Bayesian network (BN). At the same time, the established identification models were analyzed by the cross-validation method.Results Compared with SVM, LR, ANN, and BN, the RF model based on Raman data ranking top 100 of the Gini coefficient showed optimal results of identification, with accuracy and precision greater than 0.93 and receiver operating characteristic area under curve (AUC) greater than 0.95.Conclusion The correlation between Raman spectra of neutral Chinese medicine and neutral properties is significant, and it can be used as a characterization of medicinal properties. Combined with the RF algorithm for efficient and accurate identification of neutral Chinese medicine provides a strong basis for the objective existence of neutral property and the inclusion of neutral property in the property and flavor theory.
keywords:Chinese medicine  Raman spectroscopy  medicinal properties  random forest  feature selection
查看全文   查看/发表评论  下载PDF阅读器