White spirit grading method and device
By training the liquor grading model based on the loss function of the key compound characteristics, and optimizing the liquor preparation plan using an evolutionary algorithm, the problem of low efficiency of the existing liquor grading method is solved, and efficient and accurate liquor grading and preparation are achieved.
Patent Information
- Application Number
- CN202411733422.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-29
- Publication Date
- 2025-03-04
AI Technical Summary
The existing liquor grading method is relatively inefficient and depends on experience, making it difficult to achieve efficient and consistent grading.
The liquor grading model is trained based on the loss function of the key compound characteristics, the liquor grading is performed using neural networks, and the liquor preparation scheme is optimized through an evolutionary algorithm to obtain the liquor of the target level.
It improves the efficiency and accuracy of liquor grading, reduces the dependence on expert experience, and achieves faster and more reliable liquor grading and dispensing.
Smart Images
Figure CN119557790A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of wine brewing technology, and in particular to a method and device for grading liquor. Background Art
[0002] In the production of liquor, in order to ensure the consistency and reliability of liquor products, monitor and maintain the stability of the production process, and meet the needs of different consumers, liquor needs to be graded. The grading of liquor not only helps consumers choose products that suit them, but also ensures the quality and stability of the liquor. Specifically, through strict grading standards, high-quality liquor can be screened out to prevent inferior products from entering the market. Different grades of liquor are suitable for different consumer groups. High-end liquor is suitable for consumers who pursue high quality, while mid- and low-end liquor is more suitable for mass consumption. At present, in the traditional liquor production process, liquor grading mainly relies on experienced experts, which relies on expert experience and has low efficiency. Summary of the invention
[0003] The present application aims to solve the problem of low efficiency of the existing liquor grading method and proposes a liquor grading method and device.
[0004] The technical solution adopted by this application to solve the above technical problems is:
[0005] In a first aspect, the present application provides a liquor classification method, the method comprising:
[0006] The liquor classification model is obtained by training the key feature dataset of liquor samples and the loss function defined based on the key compound features;
[0007] Obtain key compound characteristic data of liquor to be classified;
[0008] Inputting the key compound characteristic data into the liquor classification model to obtain the classification probability of the liquor to be classified;
[0009] The liquor grade of the liquor to be classified is determined according to the classification probability.
[0010] Furthermore, the method further comprises:
[0011] A liquor blending scheme is generated based on key compound characteristic data of different liquor samples, and different liquor samples are blended according to the liquor blending scheme to obtain a target liquor of a target liquor grade.
[0012] Furthermore, the key feature data set includes a key compound feature data set and a liquor grade feature data set, and the key compound feature data in the key compound feature data set and the liquor grade feature data in the liquor grade feature data set are in one-to-one correspondence;
[0013] The method for acquiring key compound characteristic data comprises: performing key compound characteristic detection on a liquor sample according to predefined key compound characteristics to acquire key compound characteristic data of the liquor sample;
[0014] The method for acquiring liquor grade characteristic data comprises: performing expert rating on liquor samples according to predefined liquor grade characteristics to obtain liquor grade characteristic data of the liquor samples.
[0015] Furthermore, the key compound characteristic data of the liquor sample is the content of the key compound in the liquor sample;
[0016] The key compounds include ester compounds, alcohol compounds, acid compounds, aromatic compounds, aldehyde compounds, ketone compounds, furan compounds and pyrazine compounds; and also include sulfur-containing compounds and nitrogen-containing compounds;
[0017] The content of the key compounds in the liquor sample is obtained by mass spectrometry, hydrogen flame ionization detector or comprehensive two-dimensional gas chromatography.
[0018] Furthermore, the method further comprises: performing standardization processing on the key compound characteristic data, the formula is as follows:
[0019]
[0020] Among them, x represents the content of key compounds, x norm_1 represents the content of the compound after being standardized to the range of [0,1], x norm_2 represents the content of the compound after being standardized to the range of [-1,1], x min represents the minimum content of key compounds in all liquor samples, x max Indicates the maximum content of key compounds in all liquor samples.
[0021] Furthermore, the training method of the liquor classification model includes:
[0022] The key compound feature dataset is used as the input feature of the neural network model, and the liquor grade feature dataset is used as the output label of the neural network model;
[0023] Divide the key feature dataset into training and test sets;
[0024] Train the neural network model with the training set, compare the output features of the neural network model with the corresponding output labels, calculate the corresponding loss function, and optimize the model through the back propagation of the neural network;
[0025] The neural network model is verified with the test set. When the loss function is minimized and the accuracy of the neural network model is greater than the preset threshold, the corresponding neural network model is used as the liquor grading model.
[0026] Furthermore, the loss function is a cross entropy loss function, and the loss function is as follows:
[0027]
[0028] Where loss represents the loss function, N represents the number of liquor samples, and C represents the number of liquor grades; y ic represents a binary indicator. If the true grade of the i-th liquor sample is c, then y ic is 1, otherwise y ic is 0; p ic It represents the probability that the predicted grade of the i-th liquor sample is c.
[0029] Further, determining the liquor grade of the liquor to be graded according to the classification probability specifically includes:
[0030] The liquor grade corresponding to the maximum value of the classification probability is taken as the liquor grade of the liquor to be classified.
[0031] Furthermore, according to the key compound characteristic data of different liquor samples, a liquor blending plan is generated, which specifically includes:
[0032] Based on the known liquor grade and key compound characteristic data of liquor samples, an evolutionary algorithm is used to optimize and obtain the liquor blending ratio corresponding to the target liquor;
[0033] The evolutionary algorithm includes a genetic algorithm, a particle swarm algorithm, an ant colony algorithm, a simulated annealing algorithm and a differential evolution algorithm. The evolutionary algorithm evaluates the advantages and disadvantages of different blending ratios based on a fitness function, performs selective optimization operations, and continuously adjusts the blending ratios to gradually optimize the final result and generate a liquor blending plan. The optimization objectives of the evolutionary algorithm include one or more of the following combinations:
[0034] Generate target liquor with preset yield;
[0035] Using the liquor classification model to determine the liquor grade of the target liquor as the target liquor grade;
[0036] Based on the limited amount of liquor samples used for blending the target liquor, the amount of the target liquor produced is the largest;
[0037] When generating the target liquor, the amount of liquor sample used for blending the target liquor is minimal;
[0038] Generate target liquors of various target grades.
[0039] In a second aspect, the present application provides a liquor grading device, the device comprising:
[0040] A model training module is used to train a liquor classification model based on a key feature data set of liquor samples and a loss function defined based on key compound features;
[0041] Feature acquisition module, used to obtain key compound features of liquor to be classified;
[0042] A model analysis module, used to input the key compound characteristics into the liquor classification model to obtain the classification probability of the liquor to be classified;
[0043] The liquor grading module is used to determine the liquor grade of the liquor to be graded according to the grading probability.
[0044] The beneficial effects of the present application are as follows: the liquor grading method and device provided in the present application utilize the trained liquor grading model to grade liquor, thereby improving the efficiency of liquor grading, and the embodiment of the present application trains the liquor grading model based on the loss function defined by the key compound characteristics, thereby improving the efficiency of liquor grading model training and further improving the efficiency of liquor grading. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 A schematic diagram of a process for classifying liquor provided in an embodiment of the present application;
[0046] Figure 2 A schematic diagram of the structure of a neural network model provided in an embodiment of the present application;
[0047] Figure 3 A schematic diagram of the training process of a liquor classification model provided in an embodiment of the present application;
[0048] Figure 4 A schematic diagram of the structure of a liquor grading device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0049] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.
[0050] In some of the processes described in the specification of the present application and the above-mentioned figures, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel, and the sequence numbers of the operations are only used to distinguish different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel.
[0051] The technical solution of the embodiment of the present application is applicable to application scenarios where liquor needs to be graded. Since the current liquor grading method is usually performed by experienced experts to determine the liquor grade, this method is inefficient.
[0052] Based on this, the technical solution of the present application is proposed. In an embodiment of the present application, a liquor grading model is obtained by training based on a key feature data set of liquor samples and a loss function defined based on key compound features; the key compound features of the liquor to be graded are obtained; the key compound features are input into the liquor grading model to obtain the grading probability of the liquor to be graded; and the liquor grade of the liquor to be graded is determined based on the grading probability.
[0053] Specifically, the embodiment of the present application uses the trained liquor grading model to perform liquor grading, thereby improving the efficiency of liquor grading. In addition, the embodiment of the present application trains the liquor grading model based on the loss function defined by the key compound characteristics, thereby improving the efficiency of liquor grading model training and further improving the efficiency of liquor grading.
[0054] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.
[0055] Figure 1 A schematic diagram of a liquor classification method is shown, see Figure 1 , the method comprises the following steps:
[0056] S1. A liquor classification model is obtained by training based on the key feature dataset of liquor samples and the loss function defined based on key compound features.
[0057] In an embodiment of the present application, the key feature data set includes a key compound feature data set and a liquor grade feature data set, and the key compound feature data in the key compound feature data set and the liquor grade feature data in the liquor grade feature data set are one-to-one corresponding; that is, one liquor sample corresponds to one key compound feature data and one liquor grade feature data.
[0058] The method for acquiring key compound characteristic data includes: performing key compound characteristic detection on a liquor sample according to predefined key compound characteristics to acquire key compound characteristic data of the liquor sample.
[0059] The method for acquiring liquor grade characteristic data comprises: performing expert rating on liquor samples according to predefined liquor grade characteristics to obtain liquor grade characteristic data of the liquor samples.
[0060] In the embodiment of the present application, the key compound characteristic data of the liquor sample is the content of the key compound in the liquor sample, and the content of the key compound is obtained by mass spectrometry, hydrogen flame ionization detector or comprehensive two-dimensional gas chromatography. The key compounds include ester compounds, alcohol compounds, acid compounds, aromatic compounds, aldehyde compounds, ketone compounds, furan compounds and pyrazine compounds; and also include sulfur-containing compounds and nitrogen-containing compounds.
[0061] For example, the key compounds can select 62 compounds including acetic acid, propionic acid, 2-methylpropionic acid, butyric acid, 3-methylbutyric acid, valeric acid, 4-methylvaleric acid, hexanoic acid, octanoic acid, heptanoic acid, acetaldehyde, acetone, ethyl formate, ethyl acetate, acetal, ethyl 2-methylpropionate, propyl acetate, etc. as key compound characteristics, and their content is expressed by concentration.
[0062] The above-mentioned key compounds can comprehensively reflect the compounds that may affect the flavor and taste of liquor, thereby improving the accuracy of liquor classification.
[0063] In the embodiment of the present application, the characteristic data of key compounds are also standardized, and the formula is as follows:
[0064]
[0065] Among them, x represents the content of key compounds, x norm_1 represents the content of the compound after being standardized to the range of [0,1], x norm_2 represents the content of the compound after being standardized to the range of [-1,1], x min represents the minimum content of key compounds in all liquor samples, x max Indicates the maximum content of key compounds in all liquor samples.
[0066] For example, in this embodiment, the content of key compounds corresponding to a liquor sample is [679.09, 18.81, 8.37, 97.29, 11.87, 42.41, 0.53, 121.02, 8.21, 1.71, 35.04, 9.72, 91.54, 0.00, 1760.46, …], then after standardization, the standardized key compound characteristic data of the liquor sample is [0.2142, 0.0059, 0.0026, 0.0307, 0.0037, 0.0134, 0.0002, 0.0382, 0.0026, 0.0005, 0.0111, 0.0031, 0.0289, 0.0000, 0.5553, …].
[0067] By standardizing the characteristic data of key compounds, the accuracy of the liquor classification model can be further improved.
[0068] In the embodiment of the present application, the liquor grade feature data is represented by a one-hot vector, and the liquor grade feature data is divided into primary and secondary. That is, the liquor grade feature corresponding to the primary liquor is [1, 0], and the liquor grade feature corresponding to the secondary liquor is [0, 1].
[0069] The present application embodiment uses MLP (Multilayer Perceptron), RNN (Recurrent Neural Network) or LSTM (Long Short-Term Memory) neural network models that can be used for one-dimensional data classification to train the liquor classification model. Figure 2 , the neural network selected in the embodiment of the present application is MLP.
[0070] See also Figure 3 In the embodiment of the present application, the training method of the liquor classification model specifically includes the following steps:
[0071] S11, using the key compound feature dataset as the input feature of the neural network model, and using the liquor grade feature dataset as the output label of the neural network model;
[0072] S12, comparing the output features of the neural network model with the corresponding output labels, calculating the corresponding loss function, and optimizing the model through back propagation of the neural network;
[0073] S13. When the loss function is minimum and the accuracy of the neural network model is greater than a preset threshold, the corresponding neural network model is used as a liquor grading model.
[0074] In practical applications, the key feature data set can be divided into a training set and a test set. The training set is used to train the neural network model, and the verification set is used to calculate the accuracy of the neural network model. When the loss function is minimized and the accuracy of the neural network model is greater than the preset threshold, the liquor grading model can be obtained.
[0075] Among them, the loss function is the cross entropy loss function, and the loss function is as follows:
[0076]
[0077] Where loss represents the loss function, N represents the number of liquor samples, and C represents the number of liquor grades; y ic represents a binary indicator. If the true grade of the i-th liquor sample is c, then y ic is 1, otherwise y ic is 0; p icIt represents the probability that the predicted grade of the i-th liquor sample is c.
[0078] In this embodiment, the liquor sample key feature data set contains key feature data of 50 liquor samples, 20% of which are used as a test set, with 40 data used as a training set and 10 data used as a test set.
[0079] Since both level one and level two are taken into consideration, it can be considered that a liquor sample is either level one or level two, and the sum of the probabilities predicted to be level one and level two is 1. Therefore, the level with the highest probability of liquor sample grading is the level where the probability of liquor sample grading is greater than 0.5.
[0080] After training, the neural network model of this embodiment can achieve 100% accuracy in the test set. Therefore, the liquor grading model can effectively realize the grading of liquor. The prediction results of the test set in the liquor grading model are shown in Table 1.
[0081] Table 1 Prediction results of liquor classification model for test set data
[0082] Test set data The probability of predicting a first-class wine Predict whether it is a first-class wine Is the true value a first-class wine? 1 0.9979 yes yes 2 0.0075 no no 3 0.0021 no no 4 0.8983 yes yes 5 0.2860 no no 6 0.9998 yes yes 7 0.9964 yes yes 8 0.0037 no no 9 0.9719 yes yes 10 0.1672 no no
[0083] S2. Obtain the key compound characteristic data of the liquor to be classified.
[0084] After obtaining the liquor classification model, the liquor classification model can be used to classify the liquor to be classified. Specifically, first obtain the key compound feature data of the liquor to be classified. The key compound feature data of the liquor to be classified and the method for obtaining the key compound feature data are the same as the key compound feature data corresponding to the key compound feature data set, and the present embodiment of the application will not be repeated.
[0085] S3. Input the key compound characteristic data into the liquor classification model to obtain the classification probability of the liquor to be classified.
[0086] It can be understood that, since the key compound characteristic data in the key compound characteristic data set are standardized when the liquor classification model is trained in the embodiment of the present application, it is first necessary to standardize the key compound characteristic data of the liquor to be classified. The processing method is the same, and the embodiment of the present application will not be repeated. Then, the key compound characteristic data of the liquor to be classified after the standardized processing is input into the liquor classification model, and the classification probability of the liquor to be classified can be obtained, that is, the probability corresponding to each liquor grade.
[0087] S4. Determine the liquor grade of the liquor to be graded according to the classification probability.
[0088] In the embodiment of the present application, specifically, the liquor grade corresponding to the maximum value of the classification probability is used as the liquor grade of the liquor to be classified.
[0089] In an embodiment of the present application, the method further includes:
[0090] S5. Generate a liquor blending scheme based on key compound characteristic data of different liquor samples, and blend different liquor samples according to the liquor blending scheme to obtain target liquor of target liquor grade.
[0091] In the embodiment of the present application, a liquor blending scheme is generated based on the key compound characteristic data of different liquor samples, specifically including:
[0092] Based on the known liquor grade and key compound characteristic data of liquor samples, an evolutionary algorithm is used to optimize and obtain the liquor blending ratio corresponding to the target liquor;
[0093] The evolutionary algorithm includes a genetic algorithm, a particle swarm algorithm, an ant colony algorithm, a simulated annealing algorithm and a differential evolution algorithm. The evolutionary algorithm evaluates the advantages and disadvantages of different blending ratios based on a fitness function, performs selective optimization operations, and continuously adjusts the blending ratios to gradually optimize the final result and generate a liquor blending plan. The optimization objectives of the evolutionary algorithm include one or more of the following combinations:
[0094] Generate target liquor with preset yield;
[0095] Using the liquor classification model to determine the liquor grade of the target liquor as the target liquor grade;
[0096] Based on the limited amount of liquor samples used for blending the target liquor, the amount of the target liquor produced is the largest;
[0097] When generating the target liquor, the amount of liquor sample used for blending the target liquor is minimal;
[0098] Generate target liquors of various target grades.
[0099] Specifically, the embodiment of the present application is based on the liquor grades of different liquor samples determined by the above liquor classification model and their corresponding key compound characteristic data, and according to the combination of liquors of different grades, the combination ratio is calculated to blend liquors of different grades to make them become liquors of target grades. For example, second-grade liquor can be combined with an appropriate amount of first-grade liquor to become first-grade liquor.
[0100] In this embodiment, taking three known wine samples as an example, the above-mentioned liquor classification model has been used to determine that two of them are first-grade wines and the other is a second-grade wine, and a genetic algorithm is used to optimize the blending ratio.
[0101] In this embodiment, a genetic algorithm is used to optimize the liquor blending ratio, and the optimization goal is to "generate the maximum amount of target liquor based on a limited amount of liquor samples used for target liquor blending", specifically to generate as many first-class liquors as possible;
[0102] The pseudo code of the genetic algorithm is as follows:
[0103]
[0104]
[0105]
[0106] It should be noted that the above pseudo code is only used as an example. When solving specific problems, parameters such as selection, crossover, mutation method, population size, population generation, mutation rate, etc. in the genetic algorithm should be adjusted according to the specific problem.
[0107] In summary, the liquor grading method provided in the embodiment of the present application utilizes the trained liquor grading model to grade liquor, and optimizes the liquor blending ratio based on the evolutionary algorithm and the liquor grading model to obtain the target liquor of the target liquor grade, thereby improving the efficiency of liquor grading and liquor blending. In addition, the embodiment of the present application trains the liquor grading model based on the loss function defined by the key compound characteristics, thereby improving the efficiency of liquor grading model training and further improving the efficiency of liquor grading.
[0108] Based on the above technical solution, the present application also proposes a liquor grading device, see Figure 4 , the device comprises:
[0109] A model training module is used to train a liquor classification model based on a key feature data set of liquor samples and a loss function defined based on key compound features;
[0110] Feature acquisition module, used to obtain key compound features of liquor to be classified;
[0111] A model analysis module, used to input the key compound characteristics into the liquor classification model to obtain the classification probability of the liquor to be classified;
[0112] The liquor grading module is used to determine the liquor grade of the liquor to be graded according to the grading probability.
[0113] It can be understood that since the liquor grading device described in the embodiment of the present application is a device for implementing the liquor grading method described in the embodiment, for the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method, which will not be repeated here.
Claims
1. A liquor classification method, characterized in that: The method comprises: The liquor classification model is obtained by training the key feature dataset of liquor samples and the loss function defined based on the key compound features; Obtain key compound characteristic data of liquor to be classified; Inputting the key compound characteristic data into the liquor classification model to obtain the classification probability of the liquor to be classified; The liquor grade of the liquor to be classified is determined according to the classification probability.
2. The liquor classification method according to claim 1, characterized in that: The method further comprises: A liquor blending scheme is generated based on key compound characteristic data of different liquor samples, and different liquor samples are blended according to the liquor blending scheme to obtain a target liquor of a target liquor grade.
3. The liquor classification method according to claim 1, characterized in that: The key feature data set includes a key compound feature data set and a liquor grade feature data set, and the key compound feature data in the key compound feature data set and the liquor grade feature data in the liquor grade feature data set are in one-to-one correspondence; The method for acquiring key compound characteristic data comprises: performing key compound characteristic detection on a liquor sample according to predefined key compound characteristics to acquire key compound characteristic data of the liquor sample; The method for acquiring liquor grade characteristic data comprises: performing expert rating on liquor samples according to predefined liquor grade characteristics to obtain liquor grade characteristic data of the liquor samples.
4. The liquor classification method according to claim 3, characterized in that: The key compound characteristic data of the liquor sample is the content of the key compound in the liquor sample; The key compounds include ester compounds, alcohol compounds, acid compounds, aromatic compounds, aldehyde compounds, ketone compounds, furan compounds and pyrazine compounds; and also include sulfur-containing compounds and nitrogen-containing compounds; The content of the key compounds in the liquor sample is obtained by mass spectrometry, hydrogen flame ionization detector or comprehensive two-dimensional gas chromatography.
5. The liquor classification method according to claim 4, characterized in that: The method further includes: performing standardization processing on the key compound characteristic data, and the formula is as follows: Among them, x represents the content of key compounds, x norm_1 represents the content of the compound after being standardized to the range of [0,1], x norm_2 represents the content of the compound after being standardized to the range of [-1,1], x min represents the minimum content of key compounds in all liquor samples, x max Indicates the maximum content of key compounds in all liquor samples.
6. The liquor classification method according to claim 3, characterized in that: The training method of the liquor classification model includes: The key compound feature dataset is used as the input feature of the neural network model, and the liquor grade feature dataset is used as the output label of the neural network model; Divide the key feature dataset into training and test sets; Train the neural network model with the training set, compare the output features of the neural network model with the corresponding output labels, calculate the corresponding loss function, and optimize the model through the back propagation of the neural network; The neural network model is verified with the test set. When the loss function is minimized and the accuracy of the neural network model is greater than the preset threshold, the corresponding neural network model is used as the liquor grading model.
7. The liquor classification method according to claim 6, characterized in that: The loss function is a cross entropy loss function, and the loss function is as follows: Where loss represents the loss function, N represents the number of liquor samples, and C represents the number of liquor grades; ic represents a binary indicator. If the true grade of the i-th liquor sample is c, then y ic is 1, otherwise y ic is 0; p ic It represents the probability that the predicted grade of the i-th liquor sample is c.
8. The liquor classification method according to claim 1, characterized in that: Determining the liquor grade of the liquor to be classified according to the classification probability specifically includes: The liquor grade corresponding to the maximum value of the classification probability is taken as the liquor grade of the liquor to be classified.
9. The liquor classification method according to claim 1, characterized in that: Based on the key compound characteristic data of different liquor samples, a liquor blending plan is generated, including: Based on the known liquor grade and key compound characteristic data of liquor samples, an evolutionary algorithm is used to optimize and obtain the liquor blending ratio corresponding to the target liquor; The evolutionary algorithm includes a genetic algorithm, a particle swarm algorithm, an ant colony algorithm, a simulated annealing algorithm and a differential evolution algorithm. The evolutionary algorithm evaluates the advantages and disadvantages of different blending ratios based on a fitness function, performs selective optimization operations, and continuously adjusts the blending ratios to gradually optimize the final result and generate a liquor blending plan. The optimization objectives of the evolutionary algorithm include one or more of the following combinations: Generate target liquor with preset yield; Using the liquor classification model to determine the liquor grade of the target liquor as the target liquor grade; Based on the limited amount of liquor samples used for blending the target liquor, the amount of the target liquor produced is the largest; When generating the target liquor, the amount of liquor sample used for blending the target liquor is minimal; Generate target liquors of various target grades.
10. A liquor grading device, characterized in that: The device comprises: A model training module is used to train a liquor classification model based on a key feature data set of liquor samples and a loss function defined based on key compound features; Feature acquisition module, used to obtain key compound features of liquor to be classified; A model analysis module, used to input the key compound characteristics into the liquor classification model to obtain the classification probability of the liquor to be classified; The liquor grading module is used to determine the liquor grade of the liquor to be graded according to the grading probability.