A pattern recognition method for comprehensive evaluation of quality and storage time of fructus aurantii based on color, aroma and non-volatile organic compounds

Abstract

The application provides a pattern recognition method for comprehensively evaluating the quality of Chinese medicinal Exocarpium Citri Grandis based on color, aroma and non-volatile organic matter, and belongs to the technical field of drug quality control. The method comprises the following steps: collecting Chinese medicinal Exocarpium Citri Grandis samples with accurate source information, evaluating the color and aroma and collecting non-volatile organic matter data through crushing and brewing, establishing the color, aroma and non-volatile component fingerprint of the Exocarpium Citri Grandis sample, and obtaining the overall chemical information of the sample; then, the overall chemical information of the Exocarpium Citri Grandis sample is subjected to normalized data processing, combined with chemometrics, multivariate cluster analysis and factor analysis, and the characteristic sample data for establishing a discriminant function are screened out; a stepwise discriminant analysis method is used to extract characteristic variables from the screened sample data, a discriminant function is established, and meanwhile, the characteristic variables are used in combination with machine learning means to perform prediction analysis on unknown Exocarpium Citri Grandis samples, so that the pattern recognition of the quality of the Exocarpium Citri Grandis sample is completed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pharmaceutical quality control, and particularly relates to a pattern recognition method for establishing comprehensive evaluation of quality and storage time of Fructus Aurantii based on color, aroma and non-volatile organic matter. BACKGROUND

[0002] Fructus Aurantii is the dried immature or nearly mature epicarp of Citrus grandis 'Tomentosa' or Citrus grandis (L.) Osbeck, and has the effects of relieving phlegm, relieving cough and relieving asthma. There is a saying in folk that the longer the storage time of Fructus Aurantii is, the better the quality is. In order to study the quality of Fructus Aurantii and determine the storage time, the change rule of the components of Fructus Aurantii during storage should be understood first. Therefore, the primary task of quality control is to reveal the change rule of the components of Fructus Aurantii during storage. An effective change recognition technology of Fructus Aurantii must meet the following two requirements:

[0003] (1) meeting the large sample scale data screening;

[0004] (2) clearly showing that the content change of the characteristic components has a specific change rule with different storage time.

[0005] Traditionally, the storage time of Fructus Aurantii is identified by the label of the merchant and the'senses' of the sensory members, which not only has a long evaluation period, but also is easily interfered by various subjective and external factors, and the 'consistency' and 'quantization' of the data are insufficient, so that the actual quality cannot be objectively and stably represented, and there are phenomena such as false labeling of the storage time by the merchant.

[0006] Nowadays, the main means for scientifically evaluating the quality of medicines is to identify the quality of medicines by detecting the contents of volatile organic matter and flavonoids in the medicines. The commonly used detection technologies of volatile organic matter and flavonoids mainly include gas chromatography-mass spectrometry (GC-MS), high performance liquid chromatography-mass spectrometry (HPLC / MS), thin-layer chromatography (TLC), high performance liquid chromatography (HPLC), high speed countercurrent chromatography (HSCCC) and electronic tongue detection technology. However, these methods cannot comprehensively represent the quality of Fructus Aurantii, and the types and contents of volatile organic matter and flavonoids cannot be accurately represented.

[0007] Therefore, it is necessary to find a standardized and accurate method for monitoring the quality and storage time of Fructus Aurantii. SUMMARY

[0008] The application aims to provide a pattern recognition method for establishing comprehensive evaluation of quality and storage time of Fructus Aurantii based on color, aroma and non-volatile organic matter, solve the problem that the quality of Fructus Aurantii cannot be accurately characterized in the prior art, and realize rapid discrimination of Fructus Aurantii of different storage time.

[0009] In order to achieve the above-mentioned application purposes, the application provides the following technical solutions.

[0010] The application provides a pattern recognition method for establishing comprehensive evaluation of quality and storage time of Fructus Aurantii based on color, aroma and non-volatile organic matter, comprising the following steps:

[0011] (1) crushing and soaking and extracting Fructus Aurantii samples of different known storage time, collecting color difference data of the crushed materials and the soaking soup;

[0012] (2) extracting aroma components from the crushed materials of step (1) and analyzing the extracted aroma components to obtain volatile aroma component difference data of Fructus Aurantii;

[0013] (3) performing mass spectrum analysis on the soaking soup of step (1) to obtain non-volatile organic matter difference data;

[0014] (4) using discriminant analysis method on the color difference data, extracting difference metabolic components of characteristic variables, and establishing a discriminant function;

[0015] (5) performing normalization processing on the volatile aroma component difference data and the non-volatile organic matter difference data of Fructus Aurantii, dividing the normalized data into a training set and a test set according to a ratio of 7:3, performing data segmentation processing in combination with 10-fold cross-validation, using a supervised prediction machine learning method of multiple mathematical models, and adjusting to obtain the best prediction result of the quality and storage time of Fructus Aurantii.

[0016] Preferably, the particle size of the crushed materials in step (1) is 80-120 mesh; the crushed materials are mixed with water according to a mass-volume ratio of 1g:60-100mL, soaked for 20-40min, and the soaking soup is obtained.

[0017] Preferably, the method for extracting aroma components from the crushed materials is headspace solid-phase microextraction.

[0018] Preferably, the method for analyzing the extracted aroma components is gas chromatography-mass spectrometry / olfactometry, wherein the chromatography conditions are: the column type is a polar flexible quartz capillary column, the flow rate is 1 mL / min, the split ratio is 50:1, the injection port temperature is 260℃, and the column temperature is programmed to increase from 70℃ to 130℃ at a rate of 6℃ / min, then to 150℃ at a rate of 2℃ / min, and finally to 210℃ at a rate of 4℃ / min, and the retention time is 5 min; the mass spectrometry conditions are: the ion source is an electron impact source, the scanning mode is full scan 33-550 m / z, the electron energy is 70 eV, the ion source temperature is 230℃, the MS quadrupole rod temperature is 150℃, and the flow rate is 1-1.5 ml / min; and the olfactometry separation ratio is 7:3.

[0019] Preferably, the instrument for mass spectrometry analysis is UPLC-Q-TOF-MS, and the analysis conditions are: the flow rate is 0.2-0.4 mL / min, the mobile phase is methanol and water, the gradient change is: 0 min, 20% organic phase, 30 min, 90% organic phase, 40 min, 100% organic phase, the chromatography column is a C18 column, 5 cm x 2.1 mm, and the mass spectrometry scanning range is 100-2000 m / z.

[0020] Preferably, the method for extracting the differential metabolic components is: using PLS-DA and PCA unsupervised discrimination to infer the differential metabolic components of different ages of Fructus Aurantii, and using the VIP value and P-value to determine the differential metabolic components of different ages of Fructus Aurantii.

[0021] Preferably, the normalization processing is to uniformly convert all data to numbers between [-1, 1].

[0022] Preferably, the mathematical model includes KNN, SVM, ANN, and RF.

[0023] The application provides a pattern recognition method for comprehensively evaluating the quality of Chinese medicine Exocarpium Citri Grandis based on color, aroma and non-volatile organic matter, which comprises the following steps: collecting Chinese medicine Exocarpium Citri Grandis samples with accurate source information, evaluating the color and aroma and collecting non-volatile organic matter data through crushing and brewing, establishing the color, aroma and non-volatile component fingerprint of the Exocarpium Citri Grandis sample to obtain the overall chemical information of the sample; then performing normalized data processing on the overall chemical information of the Exocarpium Citri Grandis sample, combining chemometrics, multivariate cluster analysis and factor analysis to screen out characteristic sample data for establishing a discriminant function; using a stepwise discriminant analysis method, extracting characteristic variables from the screened sample data to establish a discriminant function, and simultaneously combining machine learning means to use the characteristic variables to perform predictive analysis on unknown Exocarpium Citri Grandi samples to complete the pattern recognition of the quality of the Exocarpium Citri Grandi sample.

[0024] The application provides a pattern recognition method for comprehensively evaluating the storage time of Chinese medicine Exocarpium Citri Grandis based on color, aroma and non-volatile organic matter, which obtains rules from the chemical component changes of Exocarpium Citri Grandi with a research span of 30 years, and determines the storage time of unknown Exocarpium Citri Grandi samples according to the color, aroma and non-volatile chemical component content changes. The application can be effectively used for identifying the storage time of Chinese medicine Exocarpium Citri Grandi and evaluating the grade of qualified products. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute the embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0026] Figure 1 Fig. 6 is a diagram of six Exocarpium Citri Grandi samples with storage times of 1998, 2000, 2005, 2010, 2015 and 2020 in Example 1;

[0027] Figure 2 Fig. 7 is a diagram of six Exocarpium Citri Grandi samples with storage times of 1998, 2000, 2005, 2010, 2015 and 2020 in Example 1, which are crushed into 80 meshes;

[0028] Figure 3 Fig. 8 is a diagram of six Exocarpium Citri Grandi samples with storage times of 1998, 2000, 2005, 2010, 2015 and 2020 in Example 1, which are soaked for 30 min with a brewing material liquid ratio of 1:60;

[0029] Figure 4The total ion chromatogram of volatile aroma components of Fructus Aurantii Immaturus with the age of 1998, 2000, 2005, 2010, 2015, 2020 in Example 2;

[0030] Figure 5 The mass spectrum data graph of non-volatile components in Example 4;

[0031] Figure 6 The graph of different components of Fructus Aurantii Immaturus with the age of 2020, 2015, 2010, 2005, 2000, 1988 in Example 4 by using PLS-DA unsupervised discriminant method, wherein A, B, C, D, E, F represent the age of 2020, 2015, 2010, 2005, 2000, 1988 respectively;

[0032] Figure 7 The graph of different components of Fructus Aurantii Immaturus with the age of 2020, 2015, 2010, 2005, 2000, 1988 in Example 4 by using PCA unsupervised discriminant method, wherein A, B, C, D, E, F represent the age of 2020, 2015, 2010, 2005, 2000, 1988 respectively;

[0033] Figure 8 The result graph of the characteristic difference components of Fructus Aurantii Immaturus and time spearman correlation analysis in Example 4. DETAILED DESCRIPTION

[0034] The technical solutions provided by the present application will be described in detail below in combination with examples, but they should not be understood as limiting the scope of protection of the present application.

[0035] Example 1

[0036] In order to meet the detection requirements of the color of Fructus Aurantii Immaturus, the crushing mesh number, the ratio of soaking material to liquid and the soaking time of Fructus Aurantii Immaturus are optimized in the present embodiment. Specifically as follows:

[0037] (1) Take Fructus Aurantii Immaturus with the age of 2020 with accurate source information, crush them into 40 mesh, 60 mesh, 80 mesh, 100 mesh and 120 mesh respectively, each group has three repeats, use CR-410 color difference meter to collect and analyze the color difference data of each powder, and the results are shown in Table 1.

[0038] Table 1 Influence of different crushing mesh numbers on the color difference of Fructus Aurantii Immaturus

[0039]

[0040]

[0041] As shown in Table 1, the best crushing mesh number is 80-120 mesh, and the color difference data is uniform and stable.

[0042] (2) Take the 2020 year's Fructus Clausiae powder of 80 mesh as mentioned above, and take the Fructus Clausiae whole fruit of 2015 year's as 80 mesh, and determine the best soaking liquid ratio. Soak the two kinds of Fructus Clausiae powder with 100℃ water according to the liquid ratio of 1:10, 1:20, 1:40, 1:60 for 30 min, and each group has three repeats. Use CR-410 colorimeter to collect and analyze the color, and the results are shown in Table 2.

[0043] Table 2 Influence of different liquid ratios on the color difference of Fructus Clausiae soaking soup

[0044]

[0045] From Table 2, the best brewing liquid ratio is 1:60.

[0046] (3) Take the 2020 year's Fructus Clausiae powder of 80 mesh as mentioned above, and determine the influence of different soaking time (15 min, 30 min, 1 h, 2 h) on the color difference of Fructus Clausiae soaking soup under the condition of limited brewing liquid ratio of 1:60, and each group has three repeats. Use CR-410 colorimeter to collect and analyze the color, and the results are shown in Table 3.

[0047] Table 3 Influence of different soaking time on the color difference of Fructus Clausiae soaking soup

[0048]

[0049] From Table 3, the best soaking time is 30 min.

[0050] Based on the above conclusions, take 6 Fructus Clausiae samples of 1998, 2000, 2005, 2010, 2015 and 2020 years as the source information (accurately) of different years, and crush them into 80 mesh (accurately), and soak them with 1:60 brewing liquid ratio at 100℃ for 30 min (accurately), and collect the color difference data of the crushed materials and the soaking soup, as shown in Tables 4 and 5. Figure 1 Figure 2 Figure 3

[0051] Table 4 Color difference data of Fructus Clausiae powder (80 mesh) of different storage years

[0052] Year L a b Delta E ab ]] 1988 34.6±0.27 28.06±0.27 20.99±0.16 12.46893 2000 31.64±0.29 32.43±1.26 26.04±1.56 15.30506 2005 39.61±0.47 26.97±0.61 22.88±0.27 7.222146 2010 37.19±0.27 28.5±0.22 26.79±1.53 8.501941 2015 41.77±0.46 22.2±0.38 18.54±0.43 9.641686 2020 44.72±0.16 24.6±0.36 27.4±1.05 0

[0053] Table 5 Color difference data of Fructus Clausiae soaking liquid of different storage years

[0054]

[0055]

[0056] Example 2 ​​​

[0057] Identification of volatile aroma components of Fructus aurantii

[0058] The powders (80 mesh) of Fructus aurantii samples of 6 different years in Example 1 were combined with the aroma extraction method of headspace solid-phase microextraction, and the gas chromatography-mass spectrometry / olfactometry (GC-MS / O) was used to determine the changes of volatile aroma organic matters in Fructus aurantii of different years, and to clarify the change rule of characteristic aroma components in Fructus aurantii of different storage periods. Specifically,

[0059] Aroma component extraction: the extraction head of solid-phase microextraction was aged at the injection port of gas chromatography, the aging temperature was 270℃, the aging time was 30 minutes, and the carrier gas volume flow rate was 1 mL / min. 1 g of Fructus aurantii sample powder (80 mesh) was taken, 0.5 g of sodium chloride was added, and shaken uniformly, 1 μL of 2-octanol was used as an internal standard. Place in a headspace bottle, heat and equilibrate on a magnetic stirrer, equilibrate at 60℃ for 15 min, then insert the extraction head into the sample bottle through a septum, push out the fiber head, and place the fiber head in the headspace of the sample bottle for adsorption, adsorption time 40 min, then take out the extraction head, insert it into the injection port, push out the fiber head, and analyze for 5 min.

[0060] Aroma component analysis (GC-MS / O method): chromatographic conditions: gas chromatograph, column type using polar flexible quartz capillary column. The carrier gas flow rate was 1 mL / min. Split ratio 50:1, injection port temperature 260℃, column temperature using programmed temperature: initial temperature 70℃, maintain for 2 min, then increase to 130℃ at 6℃ / min, maintain for 5 min, then increase to 150℃ at 2℃ / min, finally increase to 210℃ at 4℃ / min, maintain for 5 min.

[0061] Mass spectrometry conditions: electron energy 70 eV, ion source temperature 230℃, MS quadrupole temperature 150℃.

[0062] Olfactometry split ratio 7:3.

[0063] The identification of volatile aroma components of Fructus aurantii was obtained, as shown in Table 6, and the total ion chromatogram was as shown in Figure 4 .

[0064] Table 6 Identification of volatile aroma components in Fructus aurantii of different storage years

[0065]

[0066]

[0067]

[0068] Example 3

[0069] Non-volatile organic matter difference analysis of fructus aurantii immaturi

[0070] The 80-mesh powder of fructus aurantii immaturi of 6 different years in Example 1 was added into methanol solution (methanol: water = 1:1) at a solid-liquid ratio of 1 g: 60 mL, soaked and extracted for 30 min, and the soaking liquid was taken for mass spectrometric analysis.

[0071] Instrument: UPLC-Q-TOF-MS, USA / Agilent.

[0072] Mass spectrometric analysis conditions: liquid phase flow rate 0.2-0.4 mL / min, mobile phase methanol, water; gradient change: 0 min, 20% organic phase; 30 min, 90% organic phase, 40 min, 100% organic phase. Chromatographic column: C18 column, 5 cm x 2.1 mm. Mass spectrometric scanning range: 100-2000 m / z.

[0073] The mass spectrometric overlay of the obtained sample is shown in Figure 5 .

[0074] Figure 5 The middle ordinate from top to bottom is: x 10 6 , 2.7, 2.6, 2.5, 2.4, 2.3, 2.1, 2, 1.9, 1.8, 1.7, 1.6, 1.5, 1.4, 1.3, 1.2, 1.1, 1, 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1, 0; The abscissa from left to right is: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29; The letter below the abscissa is: Counts vs. Acquisiton Time (min); The upper left corner text is: +ESI TIC Scan Frag = 135.0V soy96-1.d.

[0075] Example 4

[0076] Establishment of machine learning data model

[0077] Based on the color difference data (Table 4), volatile aroma component data (Table 6) and non-volatile component mass spectrometric data (Table 7) obtained in Examples 1-3, first, the PLS-DA and PCA unsupervised discriminant methods were used to infer the different components of fructus aurantii immaturi of different years, as shown in Figure 5 , Figure 6 , 7VIP values and P-value were used to determine the different components of different storage years, and 35 potential differential metabolites with VIP values greater than 1 were screened out by PLS-DA model, of which 11 differential components (VIP > 1.5, p < 0.05) were shown in Table 7. Purple william, bicyclic squalene, alpha-copaene, juniper-1(6), 4-diene, delta-elemene, isocembrene, cis-molle-4(14), 5-diene, ylangene, beta-damascenone, alpha-cedrene, cadina-3, 5-diene were potential differential components. There were obvious differences in the absolute contents of these components in 6-year-old Fructus Aurantii, which might be important differential components for the formation of characteristics of different years of Fructus Aurantii.

[0078] Table 7 Differential metabolite analysis (VIP value > 1.5, p < 0.05)

[0079]

[0080]

[0081] Taking the data obtained in Example 1 (Table 4) as an example, the characteristic variables (the data in Table 4 are screened data, that is, the data of samples with larger errors are removed, and the values with a deviation of more than 3 times the standard deviation from the average value in the measured values are considered as data with larger errors) are extracted, and a discriminant function is established, and the established discriminant function is as follows:

[0082] y = -0.0009x3 + 5.224x2 - 10499x + 7E + 06, R 2 = 0.8204;

[0083] x represents the year, and y represents the L value.

[0084] y = 0.002x3 - 11.752x2 + 23578x - 2E + 07, R 2 = 0.7693;

[0085] x represents the year, and y represents the a value.

[0086] y = 0.0024x3 - 14.709x2 + 29485x - 2E + 07, R 2 = 0.4006;

[0087] x represents the year, and y represents the b value.

[0088] According to the data fitting of the corresponding discriminant function formula and the correlation analysis, the change rule of Fructus Aurantii was inferred, and the correlation analysis results were shown in Table 8. Figure 8

[0089] At the same time, the data obtained in Examples 2 and 3 (Table 6, Figure 5 ​) Normalization (unified data to [-1, 1]), the data is segmented, the overall test data is divided into training set and test set according to the proportion of 7:3, combined with 10-fold cross validation (randomly extracting the same data sample, forming 10 quantity sets, and taking turns to train 9 parts and test 1 part) to carry out data segmentation processing; supervised machine learning method is used to predict the mathematical model such as SVM, ANN, KNN, RF, etc. By adjusting the corresponding model parameters, the best prediction result can be obtained. Taking the SVM model as an example, the hyperparameter C is adjusted according to 1, 10, 50, 100, 500 and 1000, and the gamma parameter is adjusted according to 0.01, 0.001 and 0.0001. The best training and test results of different models are shown in Table 8.

[0090] Table 8 Best training and test results of different models

[0091]

[0092]

[0093] The data in Table 3 is divided into training set and test set according to the proportion of 7:3, the training set is mainly used to train the known data to the machine learning model, the training set accuracy is tested, and then the obtained result is compared with the original result, finally the ratio of the predicted correct value to the total number is calculated, which is the accuracy of the training set, and the SVM training effect is the best. The test set accuracy is to input unknown data samples in the above model to judge the year of the sample, and the test set accuracy is also the best in the SVM model.

[0094] From the above examples, the present application establishes a comprehensive evaluation method for the quality and storage life of Chinese herbal orange peel based on color, aroma and non-volatile organic matter, which can achieve good reproducibility and stability for Chinese herbal orange peel samples. Compared with the traditional sensory evaluation detection method, the present method has high objectivity and high detection accuracy, and the instrument is convenient to operate and use. The RF and SVM model analysis results show that the model has good stability, and the consistency with the actual result is higher than 80%, so the discrimination result has good practical value, and can well realize the identification and control of the quality of Chinese herbal orange peel with different storage life.

[0095] The above only describes the preferred embodiments of the present application, and it should be pointed out that for ordinary skilled persons in the technical field, some improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A pattern recognition method for establishing comprehensive evaluation of the quality and storage life of Fructus Aurantii based on color, aroma and non-volatile organic matter, characterized in that, The method comprises the following steps: (1) crushing and soaking different known age of Fructus Aurantii Immaturi samples, collecting color difference data of the crushed materials and the soaking soup; (2) extracting aroma components from the crushed materials of step (1) and analyzing the extracted aroma components to obtain volatile aroma component difference data of Fructus Aurantii Immaturi; (3) performing mass spectrometry on the soaking soup of step (1) to obtain non-volatile organic matter difference data; (4) using discriminant analysis method on the color difference data to extract characteristic variable difference metabolites and establish a discriminant function; (5) performing normalization processing on the volatile aroma component difference data and the non-volatile organic matter difference data, dividing the normalized data into a training set and a test set according to a ratio of 7:3, performing data segmentation processing by combining 10-fold cross-validation, and using a supervised prediction machine learning method of multiple mathematical models to adjust to obtain the best prediction result of the quality and storage age of Fructus Aurantii Immaturi.

2. The method of claim 1, wherein, In step (1), the particle size of the crushed materials is 80-120 mesh; the crushed materials are mixed with water at a mass-volume ratio of 1g:60-100mL, soaked for 20-40min, and the soaking soup is obtained.

3. The method of claim 2, wherein, The method for extracting aroma components from the crushed materials is headspace solid-phase microextraction.

4. The method of claim 3, wherein, The method for analyzing the extracted aroma components is gas chromatography-mass spectrometry / olfactometry, wherein the chromatographic conditions are as follows: the column type is a polar flexible quartz capillary column, the flow rate is 1mL / min, the split ratio is 50:1, the injection port temperature is 260℃, and the column temperature is programmed to be initially 70℃ for 2min, then increased to 130℃ at a rate of 6℃ / min, maintained for 5min, then increased to 150℃ at a rate of 2℃ / min, and finally increased to 210℃ at a rate of 4℃ / min and maintained for 5min; the mass spectrometry conditions are as follows: the ion source is an electron impact source, the scanning mode is full scan 33-550m / z, the electron energy is 70eV, the ion source temperature is 230℃, the MS quadrupole rod temperature is 150℃, and the flow rate is 1-1.5ml / min; and the olfactometry separation ratio is 7:

3.

5. The method of claim 4, wherein, The instrument for mass spectrometry is UPLC-Q-TOF-MS, and the analysis conditions are as follows: the flow rate is 0.2-0.4mL / min, the mobile phase is methanol and water, the gradient change is 0min, 20% organic phase, 30min, 90% organic phase, 40min, 100% organic phase, the chromatographic column is a C18 column with a size of 5cm×2.1mm, and the mass spectrometry scanning range is 100-2000m / z.

6. The method of claim 5, wherein, The method for extracting characteristic variable difference metabolites is as follows: using PLS-DA and PCA unsupervised discriminant method to infer different age Fructus Aurantii Immaturi difference metabolites from the color difference data, the volatile aroma component difference data and the non-volatile organic matter difference data of Fructus Aurantii Immaturi, and using VIP value and P-value to determine the difference metabolites of different age Fructus Aurantii Immaturi.

7. The method of claim 6, wherein, The normalization processing is to uniformly transform all data to a number between [-1, 1].

8. The method according to any one of claims 1 to 7, characterized in that, The mathematical models include KNN, SVM, ANN and RF.

Images