基于机器学习的中药材鉴别方法  

Identification Methods for Traditional Chinese Medicine Based on Machine Learning

作　　者：陈丹 陈伟 CHEN Dan;CHEN Wei(Public Course Teaching Department,Changjiang Institute of Technology,Wuhan 430200,Hubei;Graduate School,University of Baguio,Baguio 26000,Philippines)

机构地区：[1]长江工程职业技术学院公共课部,湖北武汉430200 [2]菲律宾碧瑶大学研究生院,菲律宾碧瑶26000

出　　处：《济源职业技术学院学报》2023年第2期65-70,共6页Journal of Jiyuan Vocational and Technical College

摘　　要：就2021年“高教社杯”全国大学生数学建模竞赛E题“中药材的鉴别”的第1、2问给出了可行的解法。针对问题1,使用极差和主成分分析方法将数据进行降维,利用平均轮廓法和肘部法则来确定最佳的聚类个数,使用K-Means聚类的方法将中药品聚类分为3类。针对问题2,分别使用支持向量机、BP神经网络、Logistic回归方法构建了药材产地分类模型,三个模型在训练集和测试集的准确率、精确率、召回率和F1值都分别均在0.8及0.7以上。特别地,Logistic回归模型在训练集和测试集的F1值高达0.866、0.789。结合三个分类模型为待鉴别的15个产品找到了合适的产地。这样的药材鉴别方法分析速度快、分类效果好,可为其他红外光谱数据分类鉴别问题提供借鉴。A feasible solution is provided for the first and second questions of Part E“Identification of Traditional Chinese Medicine”in China Undergraduate Mathematical Contest in Modeling of 2021 High Education Club Cup.For Question 1,use range and principal component analysis methods to reduce the dimensionality of the data,use the mean profile and Elbow methods to determine the optimal number of clusters,and use K-Means clustering method to classify the traditional Chinese medicine products into three categories.As to Question 2,support vector machines,BP neural networks,and Logistic regression methods are adopted respectively to construct classification models of the Chinese medicinal materials.The accuracy,precision,recall,and F1 values of the three models in the training and testing sets are all above 0.8 and 0.7.Specifically,the F1 values of the Logistic regression model in the training and testing sets are as high as 0.866 and 0.789,respectively.Based on the three classification models,suitable places of origin were found for the 15 products to be identified.The identification method for medicinal materials has the effect of fast analysis and good classification,which can provide reference for other infrared spectral data classification and identification problems.

关 键 词：红外光谱图 K-MEANS聚类 中草药鉴别 BP神经网络 LOGISTIC回归 

分 类 号：R282[医药卫生—中药学]