A rapid determination method of effective components of dandelion medicinal materials based on color chroma value

Abstract

The application discloses a dandelion medicinal material effective component rapid determination method based on color chroma values, and relates to the technical field of traditional Chinese medicinal material quality detection. * , a * , b * Chroma values of dandelion medicinal material powder are collected, and the contents of chicoric acid, monocaffeoyl tartaric acid and chlorogenic acid are determined by using high performance liquid chromatography as true values; subsequently, a neural network prediction model with chroma values as input and effective component contents as output is constructed and trained, preferably a single hidden layer structure; finally, the model is used for fast and nondestructive content prediction of unknown samples; compared with a traditional multiple linear regression model, the neural network model used in the application has significantly higher prediction accuracy, and the test set R² is all more than 0.86; the application realizes fast, nondestructive, low-cost and high-precision detection of dandelion medicinal material effective components, and is suitable for multiple scenes such as raw material quality control, market supervision, optimal harvest period judgment and fast pricing grading.

Description

Technical Field

[0001] This invention belongs to the field of quality testing technology for traditional Chinese medicinal materials, and in particular relates to a method for determining the effective components of dandelion medicinal materials based on colorimetric values. Background Technology

[0002] Color is an important trait indicator of the quality of Chinese medicinal materials, and can reflect their intrinsic quality to a certain extent. During the drying process, enzymatic and non-enzymatic browning reactions usually lead to L... * The level decreases, a * and b * The value increased. Dandelion samples dried by different drying methods and under different environments using the same method showed increased L. * a * and b * The values ​​show significant differences, reflecting changes in color characteristics.

[0003] As a functional food, dandelion is widely used in wine, confectionery, energy drinks, and other health products. Dandelion possesses various biological activities, including anti-inflammatory, antioxidant, antibacterial, and gastroprotective effects. These properties are primarily attributed to its phenolic acids and flavonoids, such as chicoric acid, chlorogenic acid, monocaffeoyl tartaric acid, caffeic acid, ferulic acid isoquercitrin, and luteolin-7-O-glucoside. Dandelion medicinal materials are relatively high in chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid, while the content of other phenolic acids is relatively low.

[0004] Currently, the detection of active ingredients in medicinal materials mainly relies on chemical analysis methods, such as high-performance liquid chromatography (HPLC) and gas chromatography-mass spectrometry (GC-MS). While these methods are accurate, they require complex sample pretreatment, specialized equipment, and lengthy detection times, resulting in high costs and potential for destructive processes. Therefore, a rapid and non-destructive alternative method is needed in this field. Color, as an appearance characteristic of medicinal materials, may be related to their internal chemical composition, but there is currently no systematic method in the art to predict the range of active ingredients using colorimetric values. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a simple, fast and non-destructive method for determining the effective components of dandelion medicinal materials, which addresses the shortcomings of the prior art.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] A method for determining the effective components of dandelion medicinal materials based on colorimetric values, specifically including the following steps:

[0008] Step 1, Sample Library Establishment and Data Acquisition: Collect dandelion samples from different sources, prepare them into powder, and simultaneously measure their L...* a * b * The color value and the actual content of at least one active ingredient, wherein the active ingredient is chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid; wherein L * Represents brightness, a * Represents red-green values, b * Represents the yellow-blue value;

[0009] Step 2, Model Construction: Using the L * a * b * The chromaticity value is used as the input variable, and the actual content of the effective ingredient is used as the output variable to construct and train a neural network prediction model.

[0010] Step 3, Predictive Application: Measure the L of the dandelion powder to be tested. * a * b * The chromaticity value is then input into the trained neural network prediction model, and the predicted content of the effective ingredient is output.

[0011] As a further preferred embodiment of the method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to the present invention, in step 2, data preprocessing and model construction specifically include the following steps;

[0012] (1) Data set partitioning: The data is divided into training set, validation set and test set;

[0013] (2) Constructing a multiple linear regression model: Using the training set data, with the contents of chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid as dependent variables, and L... * a * b * With chromaticity as the independent variable, a linear regression equation is fitted using the least squares method.

[0014] Y 菊苣酸 =-1.760+0.094L * -0.154a * -0.09b * ;

[0015] Y 单咖啡酰酒石酸 =-0.310+0.023L * -0.047a * -0.02b * ;

[0016] Y 绿原酸 = -0.152 + 0.006L * -0.005a * -0.004b* ;

[0017] (3) Construct and train a neural network prediction model, optimize the model parameters using the validation set, evaluate the model performance using the test set, and select the neural network as the final prediction model.

[0018] As a further preferred embodiment of the method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to the present invention, in step 2, the neural network prediction model is a single hidden layer feedforward neural network with 10 hidden layer neurons, the hidden layer activation function is a hyperbolic tangent sigmoid function, and the output layer activation function is a linear function.

[0019] As a further preferred embodiment of the method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to the present invention, in step 2, the performance of the trained neural network prediction model is evaluated using independent test set data. When the test set determination coefficient R... 2 A value greater than 0.85 indicates the model is acceptable.

[0020] As a further preferred embodiment of the method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to the present invention, the model construction, training, and verification are as follows: MATLAB's neural network fitting toolbox is used to construct and train a nonlinear mapping model from color features to effective component content; to obtain the test set R... 2 A qualified model with a value >0.85 can be achieved through the following operational steps:

[0021] (1) Data preparation and partitioning:

[0022] Import the sample dataset containing L*, a*, b* values ​​and the corresponding measured values ​​of effective components into the MATLAB workspace; when calling the neural network tool, use its built-in random partitioning function to automatically divide the data into three parts: 80% training set, used for model parameter learning; 10% validation set, used to monitor model performance and prevent overfitting during training; and 10% test set, used as completely independent unknown data for the final evaluation of the model's generalization ability.

[0023] (2) Network structure and key parameter settings:

[0024] Network type: Select a standard feedforward-backpropagation neural network;

[0025] Input and output layers: The number of neurons in the input layer is 3, corresponding to L*, a*, b*, and the number of neurons in the output layer is 1, corresponding to the effective component prediction value;

[0026] Hidden layer configuration: One hidden layer is used; the number of neurons in the hidden layer is determined by trying different preset values ​​in MATLAB tools for comparative training; the selection criterion is that, under the same training conditions, it maximizes the model's performance on the validation set, i.e., minimizing the mean squared error (MSE) or R0. 2 The highest number of neurons;

[0027] Training function and algorithm: The default MATLAB function trainlm is used, which is the Levenberg-Marquardt backpropagation algorithm;

[0028] (3) Training process and early stop optimization:

[0029] Start network training: During training, the toolbox will automatically execute the key anti-overfit tuning strategy of early stopping: After each training cycle, the system calculates the error of the model on the validation set; when the validation set error no longer decreases but starts to increase within several consecutive cycles of the default value of 6, the training will automatically stop and backtrack to the network state when the validation set error is the lowest.

[0030] (4) Model performance evaluation:

[0031] Calculation of performance metrics: After model training, a final performance evaluation is performed using independent test set data; the main metrics output by the MATLAB Neural Network Toolbox on the test set are the mean squared error (MSE) and the correlation coefficient R between the predicted and actual values; the coefficient of determination R is used. 2 As a quantitative evaluation metric for model performance; R 2 The correlation coefficient R is directly calculated from the output of MATLAB. The calculation formula is as follows:

[0032] R 2 =(R) 2

[0033] R 2 The value range is [0,1]. The closer the value is to 1, the stronger the model's ability to explain the data variance and the higher the prediction accuracy.

[0034] By analyzing the R test set 2 The value is used to objectively judge the predictive reliability of the model; the R-value of the neural network model trained with the aforementioned default parameters is calculated on an independent test set. 2 The value is greater than 0.85.

[0035] As a further preferred embodiment of the method for determining the effective components of dandelion medicinal materials based on colorimetric values ​​according to the present invention, in step 1, the actual content of the effective components is determined by high performance liquid chromatography.

[0036] As a further preferred embodiment of the method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to the present invention, in step 1, the dandelion medicinal material powder needs to pass through a No. 4 sieve with a pore size of 250μm.

[0037] As a further preferred embodiment of the method for determining the effective components of dandelion medicinal materials based on colorimetric values ​​according to the present invention, in step 1, a CM-5 colorimeter is used to measure the L... * a * b * Colorimetric value.

[0038] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method for determining the effective components of dandelion medicinal material based on color chromaticity values.

[0039] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method for determining the effective components of dandelion medicinal material based on color chromaticity values.

[0040] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects:

[0041] This invention discloses a method for determining the effective components of dandelion medicinal materials based on colorimetric values. This method involves collecting L values ​​from dandelion medicinal powder. * a * b * The colorimetric values ​​were used to determine the contents of chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid using high-performance liquid chromatography (HPLC) as the true values. Subsequently, a neural network prediction model was constructed and trained, with colorimetric values ​​as input and effective ingredient content as output, preferably a single-hidden-layer structure. Finally, this model was used to quickly and non-destructively predict the content of unknown samples. Compared with traditional multiple linear regression models, the neural network model used in this invention has significantly higher prediction accuracy. The test set R... 2 All values ​​exceeded 0.85; This invention enables rapid, non-destructive, low-cost, and high-precision detection of the effective components of dandelion medicinal materials, and is applicable to multiple scenarios such as raw material quality control, market supervision, determination of the optimal harvest period, and rapid pricing and grading. Attached Figure Description

[0042] Figure 1 This is a heatmap showing the correlation between the active ingredients of this invention and colorimetric values;

[0043] Figure 2 This is a fitting graph of the chicoric acid multiple linear regression model of the present invention;

[0044] Figure 3This is the multiple linear regression fitting diagram of monocaffeoyl tartaric acid in this invention;

[0045] Figure 4 This is the chlorogenic acid multiple linear regression fitting diagram of the present invention;

[0046] Figure 5 This is the neural network fitting diagram of chicoric acid in this invention;

[0047] Figure 6 This is the neural network fitting diagram of single coffee acid tartaric acid of the present invention;

[0048] Figure 7 This is the chlorogenic acid neural network fitting diagram of the present invention. Detailed Implementation

[0049] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings:

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. The present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. The purpose and effects of the present invention will become clearer. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0051] A method for determining the effective components of dandelion medicinal materials based on colorimetric values ​​includes the following steps:

[0052] Step 1, Sample Library Establishment and Data Acquisition: Collect dandelion samples from different sources, prepare them into powder, and simultaneously measure their L... * a * b * The color value and the actual content of at least one active ingredient, wherein the active ingredient is chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid; wherein L * Represents brightness, a * Represents red-green values, b * Representing the yellow-blue value; establishing a sample library and determining the L value of dandelion powder. * a * b * The color value was determined by high performance liquid chromatography, and the contents of chicoric acid, monocaffeoyl tartaric acid and chlorogenic acid were used as the actual values. The dandelion powder was passed through a No. 4 sieve (pore size 250μm) and the color value was detected by a CM-5 colorimeter.

[0053] Step 2: Perform correlation analysis on the data to confirm the significant correlation between color characteristics and the content of effective ingredients; data preprocessing includes Pearson correlation analysis using SPSS Pro software.

[0054] Step 3: Construct and train a multiple linear regression model and a neural network model, and optimize the model parameters using a validation set; the neural network model is a single hidden layer feedforward neural network containing 10 neurons, and uses the hyperbolic tangent sigmoid function as the activation function.

[0055] Data preprocessing and model building specifically include the following steps;

[0056] (1) Data set partitioning: The data is divided into training set, validation set and test set;

[0057] (2) Constructing a multiple linear regression model: Using the training set data, with the contents of chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid as dependent variables, and L... * a * b * With chromaticity as the independent variable, a linear regression equation is fitted using the least squares method.

[0058] Y 菊苣酸 =-1.760+0.094L * -0.154a * -0.09b * ;

[0059] Y 单咖啡酰酒石酸 =-0.310+0.023L * -0.047a * -0.02b * ;

[0060] Y 绿原酸 = -0.152 + 0.006L * -0.005a * -0.004b * ;

[0061] Step 4: Evaluate the model performance using the test set and select the neural network as the final prediction model;

[0062] Step 5: Input the chromaticity value of the new sample into the trained neural network model to predict its effective ingredient content.

[0063] This invention involves collecting L from dandelion powder. * a * b *The colorimetric values ​​were used to determine the contents of chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid using high-performance liquid chromatography (HPLC) as the true values. Subsequently, a neural network prediction model was constructed and trained, with colorimetric values ​​as input and effective ingredient content as output, preferably a single-hidden-layer structure. Finally, this model was used to quickly and non-destructively predict the content of unknown samples. Compared with traditional multiple linear regression models, the neural network model used in this invention has significantly higher prediction accuracy. The test set R... 2 All values ​​exceeded 0.90; This invention enables rapid, non-destructive, low-cost, and high-precision detection of the effective components of dandelion medicinal materials, and is applicable to multiple scenarios such as raw material quality control, market supervision, determination of the optimal harvest period, and rapid pricing and grading.

[0064] Model building, training, and validation:

[0065] This invention uses the Neural Net Fitting App (Neural Net Fitting App) of MATLAB (https: / / matlab.mathworks.com / ) to implement the fitting of color features (L... * a * b * Construction and training of a nonlinear mapping model from the content of effective components to the test set R. 2 A qualified model with a value >0.85 can be achieved through the following operational steps:

[0066] (1) Data preparation and partitioning:

[0067] The sample dataset (containing L) * a * b * Import the data (including the measured values ​​of the effective components) into the MATLAB workspace. When calling the neural network tool, use its built-in random partitioning function to automatically divide the data into three parts: training set (80%, used for model parameter learning), validation set (10%, used to monitor model performance and prevent overfitting during training), and test set (10%, as completely independent "unknown data" used for the final evaluation of the model's generalization ability). This partitioning is the basis for ensuring the objectivity of the evaluation.

[0068] (2) Network structure and key parameter settings:

[0069] Network type: Select standard feedforward backpropagation neural network.

[0070] Input and Output Layers: The input layer has 3 neurons (corresponding to L...) * a * b * The number of neurons in the output layer is 1 (corresponding to the predicted value of the effective component).

[0071] Hidden layer configuration: One hidden layer is used. The number of neurons in the hidden layer was determined through comparative training using different preset values ​​(e.g., 8, 10, 12, 15) in MATLAB. The selection criterion was that, under the same training conditions, the model would perform optimally on the validation set (i.e., the validation mean squared error MSE or R0). 2 The highest number of neurons.

[0072] Training function and algorithm: The default MATLAB function trainlm (Levenberg-Marquardt backpropagation algorithm) is used because it usually provides a faster convergence speed.

[0073] Other parameters: Activation function, performance function, etc., all use the default settings of this MATLAB toolbox.

[0074] (3) Training process and "early stop method" optimization:

[0075] Start network training. During training, the toolbox automatically executes the crucial overfitting prevention and tuning strategy of "early stopping": after each training epoch, the system calculates the model's error on the validation set. When the validation set error stops decreasing and begins to increase for several consecutive epochs (default value is 6), training will automatically stop and revert to the network state when the validation set error was lowest. This process effectively avoids the model over-memorizing the training data and improves its predictive ability on unknown data.

[0076] (4) Model performance evaluation:

[0077] Calculation of performance metrics: After model training, a final performance evaluation is performed using independent test set data. The main metrics output by the MATLAB Neural Network Toolbox on the test set are the mean squared error (MSE) and the correlation coefficient (R) between the predicted and actual values. This invention uses the coefficient of determination R. 2 As a quantitative evaluation metric for model performance. R 2 The correlation coefficient R is directly calculated from the output of MATLAB. The calculation formula is as follows:

[0078] R 2 =(R) 2

[0079] R 2 The value range is [0,1]. The closer the value is to 1, the stronger the model's ability to explain the data variance and the higher the prediction accuracy.

[0080] Interpretation of Model Performance and Implementation Results: Analyzing the R-values ​​of the test set 2The value can objectively determine the predictive reliability of the model. In a specific embodiment of the present invention, the R-value calculated on an independent test set is obtained by training a neural network model using the aforementioned default parameters. 2 The value is greater than 0.85. This result indicates that the model can be based on color features (L... * a * b * The method enables high-precision prediction of the effective component content of dandelion medicinal materials, meeting the accuracy requirements for practical applications.

[0081] Method Generality Description: This invention provides an effective component prediction method based on color features and neural networks. Those skilled in the art should understand that the final performance (R²) of the model is subject to change. 2 The value (R) is affected by sample quality, data size, and specific network parameters. The network configuration given in the above embodiments and the resulting R... 2 The value (>0.85) is merely an example to demonstrate the technical feasibility of this method. In practical applications, technicians can determine the highest possible test set R based on specific needs. 2 The target value is used to train and validate the model.

[0082] This invention can be widely applied to: (1) raw material quality control of Chinese medicinal materials production enterprises; (2) market supervision by drug regulatory authorities; (3) judgment of the optimal harvesting period of medicinal material planting bases; (4) rapid pricing and quality grading of Chinese medicinal materials trading markets; and (5) quality monitoring during the storage of Chinese medicinal materials.

[0083] This invention also provides the application of the method in the quality control, market supervision, determination of the optimal harvest period, rapid pricing and quality grading of dandelion medicinal materials.

[0084] Correlation analysis: Data preprocessing removes outliers. Outliers are sample points that significantly deviate from other observations in the dataset, possibly due to experimental errors or random interference. In this experiment, a method based on standard deviation is used for determination: if the L of a sample... * a * b * If the measured value of the active ingredient exceeds the range of ±3 standard deviations of the variable's mean, it is considered an outlier and should be removed; correlation analysis (Pearson) and heatmap visualization are performed to confirm L. * a * b * Significant correlation with the active ingredient. Results are as follows. Figure 1 As shown, correlation coefficients (ranging from +1 to -1) are used to quantify the relationship between variables. Positive coefficients indicate a direct correlation, while negative coefficients indicate an inverse correlation. The correlation between phenolic acids and the color parameters of dried dandelion were analyzed. Figure 1To determine their relationship. * It showed a strong positive correlation with chicoric acid (r = 0.74, p ≤ 0.001), monocaffeoyl tartaric acid (r = 0.71, p ≤ 0.001), chlorogenic acid (r = 0.81, p ≤ 0.001), ferulic acid (r = 0.79, p ≤ 0.001), luteolin-7-O-glucoside (r = 0.76, p ≤ 0.001), isoquercitrin (r = 0.25, p ≤ 0.001), and total phenolic acids (r = 0.59, p ≤ 0.001). Similarly, b * It was positively correlated with chicoric acid (r = 0.57, p ≤ 0.001), monocaffeoyl tartaric acid (r = 0.54, p ≤ 0.001), chlorogenic acid (r = 0.59, p ≤ 0.001), ferulic acid (r = 0.59, p ≤ 0.001), luteolin-7-O-glucoside (r = 0.65, p ≤ 0.001), isoquercitrin (r = 0.25, p ≤ 0.001), and total phenolic acids (r = 0.54, p ≤ 0.001). Conversely, a * Strong negative correlations were observed with chicoric acid (r = -0.79, p ≤ 0.001), monocaffeoyl tartaric acid (r = -0.78, p ≤ 0.001), chlorogenic acid (r = -0.68, p ≤ 0.001), ferulic acid (r = -0.78, p ≤ 0.001), luteolin-7-O-glucoside (r = -0.75, p ≤ 0.001), isoquercitrin (r = -0.35, p ≤ 0.001), and total phenolic acids (r = -0.69, p ≤ 0.001). These results indicate that the content of phenolic acids can be inferred from the color changes in dried dandelion samples. * and b * The higher the value, the higher the phenolic acid content. * The lower the value, the higher the phenolic acid content. Therefore, the content can be optimized by adjusting the L value. * a * and b * Choose the best drying method based on color parameters. Figure 2 This is a fitting graph of the chicoric acid multiple linear regression model of the present invention; Figure 3 This is the multiple linear regression fitting diagram of monocaffeoyl tartaric acid in this invention; Figure 4 This is the chlorogenic acid multiple linear regression fitting diagram of the present invention; Figure 5 This is the neural network fitting diagram of chicoric acid in this invention; Figure 6 This is the neural network fitting diagram of single coffee acid tartaric acid of the present invention; Figure 7 This is the chlorogenic acid neural network fitting diagram of the present invention.

[0085] Table 1 shows a comparison of the prediction performance of different models for the effective components of dandelion.

[0086] Table 1

[0087]

[0088] In all three component predictions, the neural network model demonstrated significantly better performance than traditional linear regression. (Test set R) 2 All values ​​exceeded 0.86, with the highest reaching 0.9305, demonstrating that the model has extremely high prediction accuracy, as shown in Table 1.

[0089] Based on standardized coefficient analysis, the coefficients of the multiple linear regression model for each component are shown in Table 2.

[0090] Table 2

[0091]

[0092]

[0093] The contribution of each color feature to component prediction is as follows:

[0094] ① Chicoric acid: a * (-0.657)>L * (0.643)>b * (-0.346);

[0095] ② Monocaffeoyl tartaric acid: a * (-0.692)>L * (0.555)>b * (-0.339);

[0096] ③Chlorogenic acid: L * (0.750)>a * (-0.425)>b * (-0.253);

[0097] This discovery provides an important theoretical basis for optimizing detection equipment and feature selection.

[0098] The high consistency in performance across the training, validation, and test sets indicates that the model possesses good generalization ability and stability, making it suitable for practical applications.

[0099] Example 1: Sample Bank Establishment and Data Acquisition

[0100] (1) Sample preparation: A total of 192 batches of dandelion medicinal material samples were collected from major producing areas such as Shaanxi, Shanxi, and Jiangsu. The dried samples with a moisture content of less than 13% were obtained by using 10 different drying methods, including sun drying, shade drying, oven drying, and freeze drying, and 5 different drying temperatures. The dried samples were pulverized using a Chinese medicine pulverizer, passed through a No. 4 sieve, and obtained uniform fine powder. The powder was then sealed and stored at -20℃ for later use.

[0101] (2) Colorimetric value measurement: A CM-5 colorimeter was used for measurement. An appropriate amount of powder was placed in a sample dish, flattened, and the L value of each sample was measured. * value, a * value, b * The colorimetric value of each sample is calculated by measuring it three times in parallel and taking the average value as the colorimetric value of that sample.

[0102] (3) Determination of effective ingredient content (true value calibration): High performance liquid chromatography was used for determination.

[0103] Preparation of mixed reference solution: Accurately weigh appropriate amounts of each reference standard, dissolve in methanol and dilute to volume to prepare a solution containing 160 μg / mL chicoric acid. -1 100 μg·mL monocaffeoyl tartaric acid -1 10 μg / mL of chlorogenic acid -1 Caffeic acid 1 μg·mL -1 ferulic acid 10 μg·mL -1 Luteolin glycoside 10 μg·mL -1 Isoquercetin 5 μg·mL -1 The mixed reference standard stock solution should be stored at 4°C protected from light.

[0104] Preparation of test solution: Weigh approximately 0.5 g of dried sample powder accurately, place it in a stoppered conical flask, add 20 mL of 80% methanol accurately, weigh the mass, sonicate (400 W, 40 kHz) for 20 min, cool to room temperature, weigh again, replenish the lost mass with 80% methanol, shake well, filter through a 0.22 μm microporous membrane, and take the filtrate as the test solution.

[0105] Chromatographic conditions: Determined according to General Chapter 0512 of the 2020 edition of the Chinese Pharmacopoeia. A Thermo Ultimate 3000 HPLC system was used with a Thermo Hypersil GOLD C18 column (4.6 mm × 250 mm, 5 μm); the mobile phase was 0.1% formic acid aqueous solution (A)-methanol (B), with gradient elution (0–7 min, 13%–20% A; 7–18 min, 20%–30% A; 18–28 min, 30%–41% A; 28–35 min, 41%–45% A; 35–38 min, 45%–62% A; 38–45 min, 62%–69% A; 45–50 min, 69%–95% A; 50–52 min, 95%–13% A; 52–60 min, 13% A). Flow rate 1.0 mL·min⁻¹; column temperature 30℃; detection wavelength 327 nm; injection volume 10 μL.

[0106] Example 2: Data Preprocessing and Model Building

[0107] (1) Data set partitioning: The 192 valid data sets were randomly divided into training set (approximately 80%), validation set (approximately 10%) and test set (approximately 10%).

[0108] (3) Model building and training:

[0109] Multiple linear regression model: Using training set data, with chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid content as dependent variables, and L... * a * b * The values ​​are used as independent variables, and the least squares method is used to fit a linear regression equation.

[0110] The parameter comparisons for the chicoric acid model are shown in Table 3:

[0111] Table 3

[0112]

[0113] The comparison of the model parameters for monocaffeoyl tartaric acid is shown in Table 4:

[0114] Table 4

[0115]

[0116] The parameter comparisons for the chlorogenic acid model are shown in Table 5:

[0117] Table 5

[0118]

[0119] Neural Network Model: Modeling was performed using the Neural Network Toolbox in MATLAB R2021a. The neural network model fitting used a two-layer feedforward neural network with sigmoid hidden neurons and linear output neurons, suitable for regression tasks. A 3-10-1 network structure was constructed, consisting of 3 nodes in the input layer, 10 nodes in the hidden layer (tansig activation function), and 1 node in the output layer (purelin activation function). The training algorithm was set to the Levenberg-Marquardt algorithm, with a maximum training iteration count of 1000 and a learning rate of 0.01. The network was trained using the training set, and the training process was monitored using the validation set to prevent overfitting.

[0120] Example 3: Model Performance Evaluation and Application Example:

[0121] (1) Model Performance Evaluation: The performance of the two models trained in Example 2 was evaluated using independent test set data. Their coefficients of determination and mean squared errors were calculated, and the results are shown in Table 3. The neural network model performed well in predicting the three effective components, with R... 2All values ​​were above 0.86, and significantly better than the multiple linear regression model.

[0122] (2) Application Example: A new sample of dandelion medicinal material with unknown content was taken, and powder was prepared according to the method in Example 1. Its color value was measured, and the following result was obtained: L * =65.4,a * =13.5,b * =41.8.

[0123] Input this color value into the trained neural network model as in Example 2.

[0124] The model output predicted the following results: chicoric acid content 1.65%, monocaffeoyl tartaric acid content 0.42%, and chlorogenic acid content 0.36%.

[0125] Based on the comparison between the current version of the Pharmacopoeia of the People's Republic of China and the predicted results (chicoric acid content > 0.45%), this batch of dandelion medicinal materials was determined to be "qualified".

[0126] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A rapid method for identifying the effective components of dandelion medicinal materials based on colorimetric values, characterized in that: Specifically, the following steps are included: Step 1, Sample Library Establishment and Data Acquisition: Collect dandelion samples from different sources, prepare them into powder, and simultaneously measure their L... * a * b * The color value and the actual content of at least one active ingredient, wherein the active ingredient is chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid; wherein L * Represents brightness, a * Represents red-green values, b * Represents the yellow-blue value; Step 2, Model Construction: Using the L * a * b * The chromaticity value is used as the input variable, and the actual content of the effective ingredient is used as the output variable to construct and train a neural network prediction model. Step 3, Predictive Application: Measure the L of the dandelion powder to be tested. * a * b * The chromaticity value is then input into the trained neural network prediction model, and the predicted content of the effective ingredient is output.

2. The method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to claim 1, characterized in that, In step 2, data preprocessing and model building specifically include the following steps; (1) Dataset partitioning: Divide the data into training set, validation set and test set; (2) Constructing a multiple linear regression model: Using the training set data, with the contents of chicoric acid, monocaffeoyl tartaric acid, and chlorogenic acid as dependent variables, and L... * a * b * With chromaticity as the independent variable, a linear regression equation is fitted using the least squares method. Y 菊苣酸 =-1.760+0.094L * -0.154a * -0.09b * ; Y 单咖啡酰酒石酸 =-0.310+0.023L * -0.047a * -0.02b * ; Y 绿原酸 =-0.152+0.006L * -0.005a * -0.004b * ; (3) Construct and train a neural network prediction model, optimize the model parameters using the validation set, evaluate the model performance using the test set, and select the neural network as the final prediction model.

3. The method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to claim 1, characterized in that, In step 2, the neural network prediction model is a single-hidden-layer feedforward neural network with 10 hidden layer neurons, a hyperbolic tangent sigmoid activation function for the hidden layer, and a linear activation function for the output layer.

4. The method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to claim 2, characterized in that, In step 2, the performance of the trained neural network prediction model is evaluated using independent test set data. When the test set determination coefficient R... 2 A value greater than 0.85 indicates the model is acceptable.

5. The method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to claim 4, characterized in that, Model building, training, and validation: MATLAB's neural network fitting toolbox was used to build and train a nonlinear mapping model from color features to effective component content; to obtain the test set R... 2 A qualified model with a value >0.85 can be achieved through the following operational steps: (1) Data preparation and partitioning: Include the sample dataset in L * a * b * Import the values ​​and corresponding measured values ​​of effective components into the MATLAB workspace; when calling the neural network tool, use its built-in random partitioning function to automatically divide the data into three parts: 80% training set, used for model parameter learning; 10% validation set, used to monitor model performance and prevent overfitting during training; and 10% test set, used as completely independent unknown data for the final evaluation of model generalization ability. (2) Network structure and key parameter settings: Network type: Select a standard feedforward-backpropagation neural network; Input and Output Layers: The input layer has 3 neurons, corresponding to L... * a * b * The number of neurons in the output layer is 1, which corresponds to the predicted value of the effective component. Hidden layer configuration: One hidden layer is used; the number of neurons in the hidden layer is determined by trying different preset values ​​in MATLAB tools for comparative training; the selection criterion is that, under the same training conditions, it maximizes the model's performance on the validation set, i.e., minimizing the mean squared error (MSE) or R0. 2 The highest number of neurons; Training function and algorithm: The default MATLAB function trainlm is used, which is the Levenberg-Marquardt backpropagation algorithm; (3) Training process and early stop optimization: Start network training: During training, the toolbox will automatically execute the key anti-overfit tuning strategy of early stopping: After each training cycle, the system calculates the error of the model on the validation set; when the validation set error no longer decreases but starts to increase within several consecutive cycles of the default value of 6, the training will automatically stop and backtrack to the network state when the validation set error is the lowest. (4) Model performance evaluation: Calculation of performance metrics: After model training, a final performance evaluation is performed using independent test set data; the main metrics output by the MATLAB Neural Network Toolbox on the test set are the mean squared error (MSE) and the correlation coefficient R between the predicted and actual values; the coefficient of determination R is used. 2 As a quantitative evaluation metric for model performance; R 2 The correlation coefficient R is directly calculated from the output of MATLAB. The calculation formula is as follows: R 2 =(R) 2 R 2 The value range is [0,1]. The closer the value is to 1, the stronger the model's ability to explain the data variance and the higher the prediction accuracy. By analyzing the R test set 2 The value is used to objectively judge the predictive reliability of the model; the R-value of the neural network model trained with the aforementioned default parameters is calculated on an independent test set. 2 The value is greater than 0.

85.

6. The method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to claim 1, characterized in that, In step 1, the actual content of the active ingredient is determined by high performance liquid chromatography.

7. According to claim 1, a method for determining the effective components of dandelion medicinal material based on colorimetric values, wherein in step 1, the dandelion medicinal material powder needs to pass through a No. 4 sieve with a pore size of 250 μm.

8. The method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​according to claim 1, characterized in that, In step 1, the L is measured using a CM-5 colorimeter. * a * b * Colorimetric value.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for determining the effective components of dandelion medicinal materials based on color chromaticity values ​​as described in any one of claims 1 to 8.

Images