拉曼光谱结合机器学习算法分类中药药性的动物研究  

Animal research on classifying the properties of traditional Chinese medicine using Raman spectroscopy combined with machine learning methods

作　　者：陈子任 张硕[2] 徐丛剑 CHEN Zi-ren;ZHANG Shuo;XU Cong-jian(Department of Integrated Traditional Chinese and Western Medicine,Obstetrics and Gynecology Hospital,Fudan University,Shanghai 200011,China;Department of Reproductive and Endocrinology,Obstetrics and Gynecology Hospital,Fudan University,Shanghai 200011,China)

机构地区：[1]复旦大学附属妇产科医院中西医结合科,上海200011 [2]复旦大学附属妇产科医院生殖与内分泌科,上海200011

出　　处：《复旦学报（医学版）》2024年第5期795-799,共5页Fudan University Journal of Medical Sciences

摘　　要：目的探究中药温、凉、寒、热、平5种药性分类模型的构建及验证。方法选择小鼠服用不同性质中药后的尿液标本作为研究对象,利用拉曼光谱相关技术进行检测。将得到的数据集分为训练集和测试集,用随机森林、极端梯度提升、支持向量机、逻辑回归4种机器学习方法构建分类模型,使用精确率、召回率、F1分数以及准确率来评估模型性能。结果本研究共收集到4888组光谱,其中3910组(80%)光谱数据用以构建模型,其余978组(20%)光谱数据用以测试模型性能。随机森林模型、极端梯度提升模型、支持向量机模型和逻辑回归模型的准确率分别为92%、87%、83%和75%。参与分类权重最高的拉曼位移分别为872、1012、1108、1190和1668 cm-1。结论拉曼光谱结合机器学习算法可以用于中药5种药性的分类,为中药药性分类提供新的方法,其中随机森林模型效果最佳。Objective To explore the construction and verification of the classification model for the five properties of traditional Chinese medicine:warm,cool,cold,hot,and neutral.Methods Urine samples of mice after taking Chinese medicine of different properties were selected as research objects,and Raman spectroscopy-related technology was used for detection.The obtained data set was classified into training set and test set,and the classification model was constructed using four machine learning methods:random forest,extreme gradient boosting,support vector machine,and logistic regression.The model performance was evaluated using precision,recall,F1 score,and accuracy.Results A total of 4888 sets of spectra were collected in this study,of which 80%,totaling 3910 sets of spectral data,were used to build the model,and the remaining 20%,totaling 978 sets of spectral data,were used to test model performance.The accuracy of the random forest model was 92%,the extreme gradient boosting model was 87%,the support vector machine model was 83%,and the logistic regression model was 75%.The Raman shifts with the highest classification weights were 872,1012,1108,1190 and 1668 cm-1.Conclusion Raman spectroscopy combined with machine learning algorithms can be used to classify the five medicinal properties of traditional Chinese medicine,among which the random forest model has the best effect.

关 键 词：拉曼光谱 机器学习 中药 药性 小鼠 

分 类 号：R285.1[医药卫生—中药学] O657.37[医药卫生—中医学]