Dandelion active component extraction process optimization method based on i-wo-SVR model

Abstract

The present invention relates to a dandelion active component extraction process optimization method based on an I-WO-SVR model. The method comprises: acquiring factors to be predicted, wherein said factors comprise the ethanol concentration, ultrasonic time, ultrasonic temperature and ultrasonic power; and inputting said factors into an I-WO-SVR model, so as to acquire the extraction rate of chlorogenic acid from dandelion flowers and stems, wherein the I-WO-SVR model is obtained by using a training set to train a support vector regression (SVR) prediction model and using an improved walrus optimizer (I-WO) to optimize a penalty parameter and a nuclear parameter of the SVR prediction model, and the training set comprises original factors and chlorogenic acid extraction rates corresponding to the original factors. The present invention combines modern artificial intelligence technology with the field of traditional Chinese medicine to design a Chinese medicinal material active component extraction optimization method, has broad application prospects, and can provide new ideas for Chinese medicine extraction and intelligent industrial production thereof.

Description

An Optimization Method for Extraction Process of Dandelion Active Components Based on I-WO-SVR Model Technical Field

[0001] This invention relates to the field of modern extraction technology of traditional Chinese medicine, and in particular to an optimization method for the extraction process of effective components from dandelion based on the I-WO-SVR model. Background Technology

[0002] Dandelion, also known as "Po Po Ding" or "Huang Hua Di Ding," is a perennial herb belonging to the genus Taraxacum in the family Asteraceae. The entire dried plant can be used medicinally, earning it the reputation of "Queen of Herbs" and "Natural Penicillin." Chlorogenic acid, a major component of dandelion leaves, is internationally recognized as "plant gold," possessing a wide range of functions including lowering blood lipids, antibacterial activity, antioxidation, antitumor effects, and liver protection and choleretic properties. Therefore, the extraction of chlorogenic acid from dandelion has become a hot topic in functional food research and development.

[0003] Traditional extraction process optimization methods typically use partial experiments instead of comprehensive experiments, thus failing to analyze the effects and interactions of each experiment individually as in a full-scale experiment. Machine learning methods, however, can effectively solve nonlinear problems and perform global optimization, reducing experimental costs and time. When applied to extraction process optimization, they can achieve excellent results.

[0004] Support Vector Regression (SVR) is a classic algorithm in machine learning that can provide technical support for optimizing the extraction of effective components from traditional Chinese medicine. SVR maps data to a high-dimensional space by introducing a kernel function, enabling regression analysis in this space and effectively handling nonlinear problems. However, the performance of SVR is highly dependent on the selection of its hyperparameters, and the walrus optimization algorithm is well-suited for optimizing SVR model parameters.

[0005] Therefore, there is an urgent need for an optimization method for the extraction process of dandelion effective components based on the I-WO-SVR model. By constructing a prediction model that integrates an improved walrus optimizer with SVR, the optimization method for the extraction process of dandelion effective components can be determined, which can improve the efficiency of extracting traditional Chinese medicine components. Summary of the Invention

[0006] To address the problems existing in the prior art, the present invention aims to provide an optimization method for the extraction process of dandelion active ingredients based on the I-WO-SVR model. This method combines artificial intelligence algorithms with the extraction process of traditional Chinese medicine, constructs a predictive model of process parameters, and optimizes the extraction process based on the model to obtain the optimal parameters.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] An optimization method for the extraction process of active ingredients from dandelion based on the I-WO-SVR model includes:

[0009] The factors to be predicted are obtained, including: ethanol concentration, ultrasound time, ultrasound temperature and ultrasound power;

[0010] The factors to be predicted are input into the I-WO-SVR model to obtain the chlorogenic acid extraction rate of dandelion flower stems; the I-WO-SVR model is trained using a training set to obtain an SVR prediction model, and the penalty parameters and kernel parameters of the SVR prediction model are optimized using the improved walrus optimizer I-WO; wherein, the training set includes: extracted influencing factors and corresponding chlorogenic acid extraction rates;

[0011] The SVR prediction model is optimized using the improved Walrus Optimizer I-WO, including:

[0012] Step 1: Initialize the walrus population using an infinitely folding iterative chaotic mapping;

[0013] Step 2: When the danger signal of the walrus's environment is too high, the walrus population is migrated. For the first time, adaptive weights are introduced to improve the walrus population migration process.

[0014] Step 3: When a walrus is attacked by a natural predator while foraging underwater, it emits the aforementioned danger signal and flees its current activity area. The adaptive weights are introduced for the second time to improve the walrus's escape process.

[0015] Step 4: Repeat steps 2 to 3 until the maximum number of iterations is reached, and output the optimal penalty parameters and kernel parameters.

[0016] Optionally, the penalty parameters and kernel parameters of the SVR prediction model are optimized using the improved walrus optimizer I-WO, including:

[0017] The SVR objective function is determined, and the radial basis function is selected as the kernel function of the SVR prediction model. The penalty factor C and kernel function parameter g of the SVR prediction model are determined using the improved walrus optimizer I-WO.

[0018] The objective function expression for SVR is:

[0019] ;

[0020] ;

[0021] in, b are SVR model parameters. , These represent the input and output variables of the model, respectively; n is the number of support vectors; and ε represents the tolerance bias.

[0022] The kernel function The expression is:

[0023] ;

[0024] in, For support vectors, For the original sample set, These are the parameters for the kernel function.

[0025] Optionally, the formula for initializing the walrus population using an infinitely folding iterative chaotic map is:

[0026] ;

[0027] in, For control parameters, and Let represent the values ​​of the i-th walrus and (i+1)-th walrus in the sequence generated by the infinitely folded iterative chaotic mapping, respectively.

[0028] Optionally, the formula for first introducing adaptive weights to improve the walrus population migration process is:

[0029] ;

[0030] in, This represents the new position of the i-th walrus in the j-th dimension during the (d+1)-th iteration of the migration process. This represents the current position of the i-th walrus in the j-th dimension during the d-th iteration of the migration process. Two patrolling walruses are set up during the migration process, with positions respectively... and , This indicates the first time the adaptive weights are introduced.

[0031] Optionally, the formula for obtaining the adaptive weights is:

[0032] ;

[0033] in, d represents the maximum number of iterations, and d represents the current number of iterations. It is a natural constant.

[0034] Optionally, the formula for improving the walrus escape process by introducing adaptive weights for the second time is:

[0035] ;

[0036] in, This indicates the distance between the current walrus and the optimal walrus. As a risk factor, This is the second time that adaptive weights have been introduced.

[0037] Optionally, obtaining the training set includes:

[0038] A standard curve was plotted, and the concentration of chlorogenic acid in dandelion flower stems was calculated using the standard curve. Then, the chlorogenic acid extraction rate was calculated.

[0039] Based on the single-factor experiments, a four-factor, three-level response surface experiment was conducted with chlorogenic acid extraction rate as the evaluation index to obtain experimental sample data.

[0040] Based on the experimental results, a sample set was constructed, and the sample set was screened and augmented. The expanded sample set was then standardized, and the sample data was set as the training set.

[0041] The beneficial effects of this invention are as follows:

[0042] This invention can determine the optimal process parameters for extracting the effective components of dandelion, improving production efficiency while ensuring product quality, and bringing new methods to the extraction of traditional Chinese medicine and its intelligent industrial production. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 is a schematic diagram of an optimization method for the extraction process of dandelion effective components based on the I-WO-SVR model according to an embodiment of the present invention. Detailed Implementation

[0045] 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] As shown in Figure 1, this embodiment discloses an optimization method for the extraction process of effective components from dandelion based on the I-WO-SVR model, including: obtaining the factors to be predicted, which include: ethanol concentration, ultrasonic time, ultrasonic temperature, and ultrasonic power; inputting the factors to be predicted into the I-WO-SVR model to obtain the chlorogenic acid extraction rate of dandelion; the I-WO-SVR model is obtained by training an SVR prediction model using a training set and optimizing the penalty parameters and kernel parameters of the SVR prediction model using an improved walrus optimizer I-WO; wherein, the training set includes: the original factors and the chlorogenic acid extraction rates corresponding to the original factors.

[0048] Furthermore, obtaining the training set includes:

[0049] Plot a standard curve, use the standard curve to calculate the chlorogenic acid concentration in dandelion flower stems, and then calculate the chlorogenic acid extraction rate;

[0050] Based on the single-factor experiments, a four-factor, three-level response surface experiment was conducted with chlorogenic acid extraction rate as the evaluation index to obtain experimental sample data.

[0051] Based on the experimental results, a sample set was constructed, and the sample set was screened and augmented. The expanded sample set was then standardized, and the sample data was divided into a training set.

[0052] Specifically:

[0053] Step (1) Constructing a data sample set:

[0054] Step (1.1) Plotting the chlorogenic acid standard curve:

[0055] Accurately weigh 5 mg of chlorogenic acid standard into a small beaker, add an appropriate amount of anhydrous ethanol, and sonicate until completely dissolved. Transfer the solution to a 100 mL volumetric flask and dilute to the mark with anhydrous ethanol to obtain the chlorogenic acid standard solution. Transfer 1, 3, 5, 7, and 9 mL of the chlorogenic acid standard solution into 10 mL volumetric flasks, dilute to the mark with anhydrous ethanol, and let stand for 15 min. Use anhydrous ethanol as a blank control group and measure the absorbance at 329 nm. Perform three parallel measurements for each group and calculate the average value. Plot a standard curve with the absorbance of the chlorogenic acid standard solution as the ordinate and the concentration of the chlorogenic acid standard solution as the abscissa.

[0056] Step (1.2) Extraction of chlorogenic acid from dandelion flower stems:

[0057] Fresh dandelion flower stems were collected, washed, and dried in a 60℃ oven to constant weight. The dried stems were then pulverized and passed through a 100-mesh sieve to obtain dandelion flower stem powder, which was stored in a brown ground-glass stoppered bottle for later use. 1.00 g of the dandelion flower stem powder was accurately weighed into an Erlenmeyer flask. A 30% (v / mL) ethanol aqueous solution was added at a material-to-liquid ratio of 1:20 (g / mL). After mixing, the mixture was ultrasonically extracted at 50℃ using a 300W ultrasonic cleaner for 1 hour. After cooling to room temperature, the mixture was centrifuged at 4000 rpm for 10 minutes. The supernatant was collected, diluted, and the absorbance was measured at 329 nm. The concentration of chlorogenic acid in the extract was calculated based on the standard curve. The chlorogenic acid extraction rate was calculated using the following formula:

[0058] (1)

[0059] In the formula: ρ represents the chlorogenic acid extraction rate; ρ is the chlorogenic acid concentration (μg / mL) calculated from the standard curve. This represents the volume (mL) of the supernatant. This is the dilution factor of the solution; The weight (g) is the weight of dandelion dry powder.

[0060] Step (1.3) Single-factor experiment:

[0061] Four single-factor experiments were conducted, using ethanol concentration, ultrasonic temperature, ultrasonic time, and ultrasonic power as the evaluation factors, with chlorogenic acid extraction rate as the evaluation index. Ethanol concentrations were 20%, 30%, 40%, 50%, and 60%; ultrasonic times were 30 min, 45 min, 60 min, 75 min, and 90 min; ultrasonic temperatures were 30℃, 40℃, 50℃, 60℃, and 70℃; and ultrasonic powers were 240W, 300W, 360W, 420W, and 480W. Extraction was performed twice in parallel, and the chlorogenic acid content in the extract was determined. Except for the variable single-factors, suitable conditions from the preliminary experiment were used: ethanol concentration 30%, ultrasonic time 60 min, ultrasonic temperature 50℃, and ultrasonic power 300W.

[0062] Step (1.4) Response surface methodology with 4 factors and 3 levels:

[0063] Based on the single-factor experiments, a response surface methodology with chlorogenic acid extraction rate as the evaluation index was designed with 4 factors and 3 levels to obtain experimental sample data, as shown in Table 1. Table 1 shows the experimental factors, levels and codes.

[0064] Table 1

[0065]

[0066] Step (1.5) Data Preprocessing:

[0067] Because of the differences in attributes among various input parameters in the sample data, the prediction accuracy of the extraction rate may not be accurate enough. Standardization can handle the differences among various input parameters and ensure the accuracy of SVR model prediction.

[0068] Step (1.5.1) Constructing the sample set: The experimental data is filtered, and sample data with outlier characteristics are deleted to obtain experimental sample data. The data sample set is expanded through data augmentation techniques.

[0069] Step (1.5.2) Standardization: Standardize the expanded sample dataset using the following formula to map the sample data to the range [0,1]:

[0070] (2)

[0071] In the above formula, xo is the standardized sample set, x is the original sample set, xmin is the minimum sample in the sample set, and xmax is the maximum sample in the sample set.

[0072] Step (1.5.3) uses K-fold cross-validation to divide the sample data into a training set and a validation set:

[0073] The original sample set N is divided into K equal parts, one of which is used as the model validation set, and the remaining K-1 parts are used as the model training set. This process of dividing the validation set needs to be repeated K times. Sample data from each subset will be selected sequentially. After the data for the validation set has been selected, the remaining data is used as the training set. In each round, the training set data is imported into the SVR model constructed in step (2) to train the model, and the results are applied to the validation set.

[0074] Furthermore, the SVR prediction model is optimized using the improved walrus optimizer, including:

[0075] Step 1: Initialize the walrus population using an infinitely folding iterative chaotic mapping;

[0076] Step 2: When the danger signal in the walrus's environment is too high, the walrus population is relocated, and adaptive weights are introduced to improve the walrus population migration process.

[0077] Step 3: When a walrus is attacked by a natural predator while foraging underwater, it sends out a danger signal and flees its current activity area. Adaptive weights are introduced again to improve the walrus's escape process.

[0078] Step 4: Repeat steps 2 to 3 until the maximum number of iterations is reached, and output the optimal penalty parameters and kernel parameters.

[0079] Specifically:

[0080] Step (2) Construct the I-WO-SVR model:

[0081] The SVR model is a branch of Support Vector Machines (SVM) used to solve nonlinear regression problems. SVR models have high prediction accuracy, effectively avoid overfitting, improve generalization ability, and exhibit good statistical regularity even with small sample data. Compared to traditional regression models, SVR models have a certain degree of tolerance; even if the predicted values ​​have some deviation, they can still be considered reasonable. This invention constructs an SVR model to solve the nonlinear relationship between different ethanol concentrations, ultrasonic time, ultrasonic temperature, ultrasonic power, and dandelion chlorogenic acid extraction rate.

[0082] Step (2.1) Construct the SVR prediction model:

[0083] Step (2.1.1) Establish the regression equation and objective function: assuming a training sample is given Let x and y represent the input variables (ethanol concentration, ultrasound time, ultrasound temperature, and ultrasound power) and output variable (dandelion chlorogenic acid extraction rate), respectively. For the training samples, the regression equation F(x) can be established as follows:

[0084] (3)

[0085] in, b are SVR model parameters. for The transpose operation. Let x be the feature vector after mapping.

[0086] The objective function of SVR can be expressed as:

[0087] (4)

[0088] (5)

[0089] Where C is the penalty factor, representing the tolerance for error. n is the number of support vectors, and ε represents the tolerance for bias.

[0090] Step (2.1.2) Determine the kernel function of the SVR model: When building a support vector machine model, the most important thing is to select the kernel function of the model.

[0091] The kernel function has a significant impact on the performance of Support Vector Machine (SVM) models. It maps input samples to a high-dimensional space, thereby making the problem easier to handle. Common kernel functions include linear kernel functions, polynomial kernel functions, radial basis function (RBF) kernel functions, and sigmoid kernel functions. The RBF kernel function is preferred due to its simplicity and powerful nonlinear modeling capabilities. Therefore, this invention selects the radial basis function (RBF) as the kernel function for the SVR prediction model, and its specific expression is as follows:

[0092] (6)

[0093] in, For support vectors, g is the parameter of the RBF kernel function.

[0094] Step (2.2) initializes the population using an infinitely folding iterative chaotic map, and improves the migration and escape processes of the walrus population through adaptive distribution. Improvements to the WO include:

[0095] WO is a swarm intelligence optimization method designed by simulating the migration, breeding, resting, foraging, gathering, and escape behaviors of walruses after receiving critical signals (danger and safety signals). However, WO still has room for improvement in terms of convergence accuracy and solution speed. Therefore, this invention improves WO through the following steps to solve for the optimal C and g parameters in the SVR model:

[0096] Step (2.2.1) Initializing the Population with Infinitely Folding Iterative Chaotic Mapping: Using chaotic variables for optimization search can evenly distribute the initial population in the search space, which is more advantageous than using random variables for random search. The ergodicity and randomness of chaos can prevent the search process from getting trapped in local optima. Through chaotic mapping, the diversity of the initial state of the walrus population is increased, effectively preventing getting trapped in local optima and premature convergence, and improving global convergence and accuracy. The expression for infinitely folding iterative chaotic mapping is as follows:

[0097] (7)

[0098] in, The parameter is a control parameter and takes a value between (0,1). Xi and Xi+1 represent the values ​​of the i-th and (i+1)-th walruses in the sequence generated by the infinitely folded iterative chaotic mapping, respectively.

[0099] Step (2.2.2) Adaptive Migration: When the environmental danger signal of walruses is too high, the walrus population will migrate to a more suitable area. In this stage, adaptive weights are introduced to improve the walrus population migration process, and the new walrus locations are updated as follows:

[0100] (8)

[0101] in, This represents the new position of the i-th walrus in the j-th dimension during the (d+1)-th iteration of the migration process. This represents the current position of the i-th walrus in the j-th dimension during the d-th iteration of the migration process. Two patrolling walruses are set up during the migration process, and their positions are... and . The first adaptive weights introduced are determined by the following formula:

[0102] (9)

[0103] in, d is the maximum number of iterations, and d is the current number of iterations.

[0104] Step (2.2.3) Adaptive Escape: Walruses are also vulnerable to attacks from natural predators while foraging underwater. They will flee their current activity area based on danger signals from their companions. This behavior occurs in the later stages of the WO (Walrus Exploration) iteration, and a certain degree of disturbance to the population helps walruses conduct global exploration. This invention chooses to use adaptive weights again to adjust the escape process, effectively coordinating the relationship between global and local searches. The replaced escape process is as follows:

[0105] (10)

[0106] in This indicates the distance between the current walrus and the optimal walrus. As a risk factor, This is the second time that adaptive weights have been introduced.

[0107] Step (2.3) Establish the I-WO-SVR model: Use the I-WO algorithm to optimize the penalty parameters and kernel parameters of SVR.

[0108] In the SVR model, the penalty factor C and the kernel function parameter g are two key indicators affecting the model's prediction results. A larger C indicates less tolerance for model errors, leading to overfitting; a smaller C indicates poorer generalization ability, resulting in underfitting. The size of the kernel function parameter g affects the number of support vectors; a larger g means fewer support vectors, and vice versa. Both excessively large and small C and g affect the estimation speed and accuracy. Furthermore, different combinations of parameters also affect model performance; therefore, determining the optimal parameters C and g is crucial for the model's predictive effectiveness. To more conveniently and quickly find the optimal parameters, this invention utilizes an improved walrus optimizer to find the optimal C and g parameters, further establishing an I-WO-SVR model to improve the model's prediction accuracy.

[0109] Step (3) determines the optimal extraction scheme for the effective components of dandelion: the training set is input into the I-WO-SVR model for training, and the trained model is applied to the validation set. The coefficient of determination (R²) is used to determine the optimal extraction scheme. 2 The accuracy of the SVR model is measured by three metrics: root mean square error (RMSE), mean absolute error (MAE), and mean square error (RMSE). These metrics are used to calculate the difference between the predicted value (the chlorogenic acid extraction rate predicted by the I-WO-SVR model) and the actual value (the chlorogenic acid extraction rate in the training set). The formulas for calculating these three metrics are as follows:

[0110] (11)

[0111] (12)

[0112] (13)

[0113] Where y is the predicted value, E[y] is the mean of the predicted values; z is the actual value, E[z] is the mean of the actual values.

[0114] Under normal circumstances, R 2 The higher the value of R, the better; the lower the values ​​of RMSE and MAE, the better. Because R... 2 An infinitely close value to 1 and an error infinitely close value to 0 indicate that the model's predicted value is closer to the true value, proving that the model has a better fit.

[0115] When R² is close to 1 and both RMSE and MAE are small, it indicates that the I-WO-SVR model has high prediction accuracy and can effectively predict the extraction rate of chlorogenic acid from dandelion flower stems. After the I-WO-SVR model is trained, adjusting different input factors to be predicted (ethanol concentration, ultrasonic time, ultrasonic temperature, and ultrasonic power) will output the corresponding extraction rate. The combination of process parameters that yields the highest extraction rate is the optimal extraction scheme.

[0116] This embodiment uses the I-WO-SVR model to couple the relationship between extraction process parameters and extraction rate, thereby determining the optimal extraction process parameters, for example:

[0117] The extraction rate was highest when the ethanol concentration was 30%, the ultrasonic time was 60 min, the ultrasonic temperature was 50℃, and the ultrasonic power was 300 W. This indicates that the optimal extraction process is ethanol concentration of 30%, ultrasonic time of 60 min, ultrasonic temperature of 50℃, and ultrasonic power of 300 W.

[0118] This invention can improve production efficiency while ensuring product quality, bringing a new method to the extraction of traditional Chinese medicine and its intelligent industrial production.

[0119] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for optimizing the extraction process of effective components of Taraxacum officinale based on an I-WO-SVR model, characterized in that, The method comprises the following steps: obtaining to-be-predicted factors, wherein the to-be-predicted factors comprise ethanol concentration, ultrasonic time, ultrasonic temperature and ultrasonic power; inputting the to-be-predicted factors into an I-WO-SVR model to obtain the extraction rate of green acid in dandelion flower stems; the I-WO-SVR model is obtained by training an SVR prediction model by using a training set and optimizing the penalty parameter and the kernel parameter of the SVR prediction model by using an improved walrus optimizer I-WO; wherein the training set comprises extraction influencing factors and corresponding green acid extraction rates; optimizing the SVR prediction model by using the improved walrus optimizer I-WO comprises the following steps: Step 1: initializing a walrus population by using infinite folding iterative chaotic mapping; Step 2: when the danger signal of the environment of the walrus is too high, migrating the walrus population, and introducing adaptive weight for the first time to improve the migration process of the walrus population; Step 3: when the walrus is attacked by a natural predator while foraging underwater, the danger signal is sent out, and the walrus escapes from the current activity area, and adaptive weight is introduced for the second time to improve the escape process of the walrus; Step 4: repeating steps 2 to 3 until the maximum number of iterations is reached, and outputting the optimal penalty parameter and kernel parameter.

2. The method for optimizing the extraction process of dandelion active ingredients based on the I-WO-SVR model according to claim 1, characterized in that, optimizing the penalty parameter and the kernel parameter of the SVR prediction model by using the improved walrus optimizer I-WO comprises the following steps: determining an SVR objective function, selecting a radial basis function as the kernel function of the SVR prediction model, and determining the penalty factor C and the kernel function parameter g of the SVR prediction model by using the improved walrus optimizer I-WO: the expression of the SVR objective function is: ； ； wherein and b are SVR model parameters, 、 respectively represent the input variable and the output variable of the model, n is the number of support vectors, and ε represents the tolerance deviation; The kernel function the expression is: ； wherein to support vectors, for the original sample set, is the kernel function parameter.

3. The method for optimizing the extraction process of dandelion active ingredients based on the I-WO-SVR model according to claim 1, characterized in that, The formula for initializing the walrus population by using infinite folding iterative chaotic mapping is: ； wherein, To control the parameters, and respectively represent the value of the i-th walrus and the (i+1)-th walrus in the sequence generated by the infinite folding iterative chaotic mapping. 4.The method for optimizing the extraction process of effective components of Taraxacum officinale based on the I-WO-SVR model according to claim 1, characterized in that, The formula for introducing adaptive weight for the first time to improve the migration process of the walrus population is: ； wherein represents the new position of the i-th seal in the j-th dimension at the d+1 iteration of the migration process, represents the current position of the i-th seal in the j-th dimension at the d-th iteration of the migration process, two sealers are set in the migration process, and their positions are and ， represents the adaptive weight introduced for the first time.

5. The I-WO-SVR model-based optimization method of extracting effective components from Taraxacum mongolicum, according to claim 4, characterized in that, The formula for obtaining adaptive weight is: ； wherein N is the maximum number of iterations, d is the current iteration number, is a natural constant. 6.The method for optimizing the extraction process of effective components of Taraxacum officinale based on an I-WO-SVR model according to claim 1, characterized in that, The formula for introducing adaptive weight for the second time to improve the escape process of the walrus is: ； wherein, represents the distance between the current sea picture and the best sea picture, for a risk factor, represents the adaptive weight introduced for the second time.

7. The method for optimizing the extraction process of dandelion active ingredients based on the I-WO-SVR model according to claim 1, characterized in that, obtaining the training set comprises the following steps: drawing a standard curve, calculating the concentration of green acid in dandelion flower stems by using the standard curve, and then calculating the extraction rate of green acid; based on single-factor experiments, performing a response surface experiment with 4 factors and 3 levels by taking the extraction rate of green acid as an evaluation index to obtain experimental sample data; according to the experimental results, constructing a sample set, screening and data enhancing the sample set, standardizing the expanded sample set, and setting the sample data as the training set.

Images