Rapid Identification Methods For Anoectochilus Roxburgii Based On Electrochemical Fingerprinting And Machine Learning

Anoectochilus roxburghii(A.roxburghii)is a valuable herb with antioxidant,anti-inflammatory,and diabetic properties.Currently,there is mainly a wide range of species and substitution of fakes for A.roxburghii in the market.Therefore,accurate identification of A.roxburghii plays an important role in safeguarding its medicinal value and regulating the market economy.Compared to other identification methods,electrochemical fingerprinting is ideal for the identification of A.roxburghii because of its high sensitivity,low price,simplicity of operation,and ability to record a variety of chemical markers.However,the electrochemical fingerprinting of the same species fluctuates dynamically due to the different content of each component.In addition,the electrochemical fingerprints of the different species may be similar due to the similarity in composition between the closely related species.This all limits the electrochemical fingerprinting method can not be directly used for the identification of A.roxburghii.Surprisingly,machine learning(ML)can effectively solve the problems in electrochemical fingerprinting and enable the accurate identification of A.roxburghii,which is attributed to its ability to autonomously learn information from the data and find hidden relationships between data and categories.Therefore,this dissertation developed an intelligent and accurate method for the identification of A.roxburghii using a machine learning-based electrochemical fingerprinting technology.The main contents were as follows:1.The electrochemical sensor was constructed to collect electrochemical fingerprints of real and fake A.roxburghii and the collected data were preprocessed.Firstly,the ethanol was used as an extractant to pretreat plant samples and the glassy carbon electrode was modified by the drop coating method.Subsequently,the electrochemical fingerprints of samples were collected by differential pulse voltammetry in phosphate and acetic acid buffer solutions.Afterward,the Kruskal-Wallis method was employed to test the significance of the electrochemical fingerprints of each species.The results showed there were significant differences among different samples of the same species,which validated the necessity of introducing ML.Finally,the collected data were preprocessed by the MinMax method and the smoothing derivative filtering method,which was applied to prepare for the construction of the models.2.Species identification models based on the electrochemical fingerprinting of A.roxburghii were constructed.Two feature extraction methods,including principal component analysis(PCA)and linear discriminant analysis(LDA),were first investigated to eliminate redundant information and speed up the operation of the classification models.Subsequently,K-nearest neighbor(KNN)and support vector machine(SVM)classification models were constructed based on different feature extraction methods.The results showed that the KNN and SVM models constructed based on LDA feature extraction had better performance than the classification models constructed by PCA feature extraction.When LDA extracted the data to 5 dimensions,the SVM model exhibited the best performance for the identification of A.roxburghii and the accuracy rate reached 99.4%.This indicated that accurate identification of species can be achieved based on the electrochemical fingerprinting of A.roxburghii and the ML method.3.Authenticity identification models based on the electrochemical fingerprinting of real and fake A.roxburghii were constructed.As with species recognition models,PCA and LDA were first investigated to extract features,where the results of LDA feature extraction demonstrated large differences between different classes of feature data.Subsequently,classification models of PCA+KNN,PCA+SVM,and LDA+LDA were constructed.The results showed that the LDA models constructed based on LDA feature extraction outperformed the KNN and SVM models constructed by PCA feature extraction.When LDA extracted the data to 2 dimensions,the LDA classification model showed the best identification ability for the authenticity of A.roxburghii with an accuracy of100%.This indicated that authenticity identification can be achieved based on the electrochemical fingerprinting of real and fake A.roxburghii and the ML method.