Hypoglycemic solid beverage formula generation method and device based on deep learning model

Abstract

The invention provides a method and a device for generating a blood sugar reducing solid beverage formula based on a deep learning model. The method comprises the steps that user input data are acquired, and the user input data comprise raw material information; searching for blood sugar reducing solid beverage formula data corresponding to the raw material information from a blood sugar reducing solid beverage formula database, and generating the blood sugar reducing solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; the formula design model passes the hypoglycemic effect verification after the training is finished; and displaying the searched formula data of the hypoglycemic solid beverage and the formula data of the hypoglycemic solid beverage generated by the model under the condition that the formula data of the hypoglycemic solid beverage are found. According to the technical scheme provided by the invention, specific requirements of different users can be met.

Description

Technical Field

[0001] The embodiments of the present application relate to the fields of food science and artificial intelligence technology, and in particular to a method and device for generating a blood sugar-lowering solid beverage formula based on a deep learning model. Background Art

[0002] With the improvement of people's living standards and the enhancement of health awareness, the demand for functional foods and beverages is growing. In particular, the development of foods and beverages with specific health benefits has become a hot topic in research, especially for the special needs of patients with chronic diseases, such as diabetics. Diabetes is a common metabolic disease, and patients need to strictly control their diet to maintain stable blood sugar levels. Therefore, it is of great significance to develop solid beverage formulas that can effectively lower blood sugar.

[0003] At present, the design method of hypoglycemic solid beverage formula mainly relies on traditional trial and error and experience accumulation. This method is not only time-consuming and labor-intensive, but also difficult to ensure the hypoglycemic effectiveness and safety of solid beverage formulas. In recent years, with the development of artificial intelligence and machine learning technologies, some studies have begun to try to use these technologies to assist in the design of solid beverage formulas. For example, rule-based expert systems and machine learning algorithms (such as decision trees, support vector machines, etc.) have improved the efficiency and accuracy of solid beverage formula design to a certain extent.

[0004] However, these methods still have some limitations. For example, diabetic patients may have other complications, and the physical conditions of different diabetic patients vary greatly, so their needs are highly personalized, and existing methods are difficult to meet the specific needs of different users. Summary of the invention

[0005] The embodiment of the present application provides a method for generating a blood sugar-lowering solid beverage formula based on a deep learning model, so as to solve the problem that existing methods are difficult to meet the specific needs of different users.

[0006] In a first aspect, the present application provides a method for generating a hypoglycemic solid beverage formula based on a deep learning model, comprising:

[0007] Acquiring user input data, wherein the user input data includes raw material information;

[0008] Searching for hypoglycemic solid beverage formula data corresponding to the raw material information from a hypoglycemic solid beverage formula database, and generating hypoglycemic solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; the formula design model has been verified for hypoglycemic effect after training;

[0009] If found, the found blood sugar lowering solid beverage formula data and the blood sugar lowering solid beverage formula data generated by the model are displayed.

[0010] Optionally, the training process of the formulation design model includes:

[0011] Obtain data on plant active ingredients that have been confirmed to have blood sugar lowering function and known blood sugar lowering solid beverage formula data;

[0012] Preprocessing the plant active ingredient data and the solid beverage formula data respectively to form a functional raw material database and a blood sugar-lowering solid beverage formula database;

[0013] A formula design model is constructed using deep learning technology, and the formula design model is trained based on the functional raw material database and the blood sugar-lowering solid beverage formula database until a preset stop training condition is met to obtain a pre-trained formula design model.

[0014] Optionally, training a formulation design model based on the functional raw material database and the hypoglycemic solid beverage formulation database until a preset stop training condition is met to obtain a pre-trained formulation design model, including:

[0015] Determine a main task and a sub-task; the main task is to generate known hypoglycemic solid beverage formula data corresponding to the plant active ingredient data; the sub-task includes at least one of the following: determine the formula composition structure and raw material ratio relationship in different known hypoglycemic solid beverage formula data, determine the formula mode corresponding to the raw material information in the plant active ingredient data, and determine the associated raw material information corresponding to the raw material information in the plant active ingredient data;

[0016] Based on the functional raw material database and the blood sugar-lowering solid beverage formula database, a formula design model is trained by multi-task learning until a preset stop training condition is met to obtain a pre-trained formula design model.

[0017] Optionally, during the training process of the formulation design model, the method further includes:

[0018] Introduce regularization techniques, including L1 regularization and L2 regularization.

[0019] Optionally, during the training process of the formulation design model, the method further includes:

[0020] Hyperparameters such as learning rate, number of training rounds, and batch size are initialized, and the parameters of the recipe design model are updated using the stochastic gradient descent optimizer, and the mean squared error loss is used as the loss function of the recipe design model.

[0021] Optionally, during the training process of the formulation design model, the method further includes:

[0022] A data balancing technique is introduced to cope with the difference in data distribution between different tasks. The data balancing technique includes a data enhancement strategy and a specific weight adjustment strategy.

[0023] Optionally, the structure of the deep learning model includes an input layer, a hidden layer and an output layer, wherein the input layer receives user input data and generates an input feature vector based on the user input data; the user input data includes at least one input parameter; the hidden layer gradually processes the input feature vector through two fully connected layers to obtain an intermediate feature vector; and the output layer generates blood sugar lowering solid beverage formula data corresponding to the raw material information based on the intermediate feature vector.

[0024] Optionally, the two fully connected layers include a first fully connected layer and a second fully connected layer; the hidden layer includes a first fully connected layer and a second fully connected layer connected in sequence; the first fully connected layer includes a first linear transformation module and a first ReLU activation function connected in sequence, and the second fully connected layer includes a second linear transformation module and a second ReLU activation function connected in sequence;

[0025] The hidden layer processes the input feature vector step by step through two fully connected layers, including:

[0026] Performing a linear transformation on the input feature vector through a first linear transformation module to obtain a first linear feature vector;

[0027] Using a first ReLU activation function to convert the first linear feature vector into a nonlinear feature vector;

[0028] Performing a linear transformation on the nonlinear feature vector by a second linear transformation module to obtain a second linear feature vector;

[0029] The second linear feature vector is converted into an intermediate feature vector using a second ReLU activation function.

[0030] Optionally, the hidden layer includes a residual connection module located between the first fully connected layer and the second fully connected layer;

[0031] After converting the first linear feature vector into a nonlinear feature vector using a first ReLU activation function, the method further includes:

[0032] The input feature vector and the nonlinear feature vector are added to obtain the feature vector after residual connection.

[0033] In a second aspect, the embodiment of the present application provides a device for generating a hypoglycemic solid beverage formula based on a deep learning model, comprising:

[0034] An acquisition module, used for acquiring user input data, wherein the user input data includes raw material information;

[0035] A search and generation module, used to search for hypoglycemic solid beverage formula data corresponding to the raw material information from a hypoglycemic solid beverage formula database, and generate hypoglycemic solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; the formula design model has passed the hypoglycemic effect verification after the training;

[0036] The display module is used to display the found blood sugar lowering solid beverage formula data and the blood sugar lowering solid beverage formula data generated by the model when found.

[0037] The embodiment of the present application provides a method for generating a hypoglycemic solid beverage formula based on a deep learning model, the method comprising: obtaining user input data, the user input data including raw material information; searching for hypoglycemic solid beverage formula data corresponding to the raw material information from a hypoglycemic solid beverage formula database, and generating hypoglycemic solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; the formula design model has been verified by hypoglycemic effect at the end of training; in the case of finding, displaying the hypoglycemic solid beverage formula data found and the hypoglycemic solid beverage formula data generated by the model. In the embodiment of the present application, the user can input different raw material information according to his or her needs, in the case of searching, the hypoglycemic solid beverage formula data found and the hypoglycemic solid beverage formula data generated by the model are provided to the user, and in the case of not finding, the hypoglycemic solid beverage formula data generated by the model are provided to the user, so the embodiment of the present application can generate hypoglycemic solid beverage formula data that meets the personalized needs of different users through the formula design model.

[0038] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 A flow chart of a method for generating a blood sugar-lowering solid beverage formula based on a deep learning model provided in an embodiment of the present application;

[0041] Figure 2A flowchart of another method for generating a blood sugar-lowering solid beverage formula based on a deep learning model provided in an embodiment of the present application;

[0042] Figure 3 A flowchart of a deep learning model training provided in an embodiment of the present application;

[0043] Figure 4 A flowchart of another method for generating a blood sugar-lowering solid beverage formula based on a deep learning model provided in an embodiment of the present application;

[0044] Figure 5 A schematic diagram of the experimental results of the hypoglycemic function verification provided in the embodiments of the present application;

[0045] Figure 6 A schematic diagram of the structure of a device for generating a formula for a hypoglycemic solid beverage based on a deep learning model provided in an embodiment of the present application;

[0046] Figure 7 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] 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.

[0048] Diabetes is a metabolic disease characterized by high blood sugar. Long-term high blood sugar can lead to a variety of complications, seriously affecting the quality of life of diabetic patients. Diabetes treatment and management require comprehensive diet control, drug therapy and lifestyle adjustment. At present, although there are some hypoglycemic drugs on the market, long-term use may bring some side effects. Therefore, the development of a natural food with blood sugar-lowering function has become an urgent need for people. In diabetes management, diet control is one of the key links, and hypoglycemic solid beverages, as a convenient food form, have huge market potential and practical application value.

[0049] However, the traditional formula design of hypoglycemic solid beverages mainly relies on experience and trial and error, lacks systematicity and scientificity, and is difficult to meet the needs of diabetic patients for personalization and scientific verification.

[0050] In order to solve the problem that the traditional hypoglycemic solid beverage formula design is difficult to meet the needs of diabetic patients for personalization and scientific verification, the embodiment of the present application proposes a method for generating a hypoglycemic solid beverage formula based on a deep learning model. Different user input data corresponds to different pre-trained formula design models. The embodiment of the present application can be based on the user input data (such as raw material information) to personalize the solid beverage formula that effectively lowers blood sugar. At the same time, the embodiment of the present application can also verify the hypoglycemic effect of the hypoglycemic solid beverage formula data designed by the formula design model, thereby proving the practicality and effectiveness of the formula design model.

[0051] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear 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 article or executed in parallel. The sequence numbers of the operations, such as 11, 12, etc., are only used to distinguish between 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. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to different types.

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0053] Figure 1 A flow chart of a method for generating a hypoglycemic solid beverage formula based on a deep learning model provided in an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0054] S11. Obtain user input data, where the user input data includes raw material information.

[0055] The user input data may include raw material information. The raw material information may be raw material identification such as raw material name, raw material number, etc. The raw material information is used to reflect user needs.

[0056] S12. Searching for blood sugar lowering solid beverage formula data corresponding to the raw material information from a blood sugar lowering solid beverage formula database, and generating blood sugar lowering solid beverage formula data corresponding to the raw material information through a pre-trained formula design model, wherein the formula design model has been verified for blood sugar lowering effect at the end of training.

[0057] It should be understood that the hypoglycemic solid beverage formula database contains pre-treated hypoglycemic solid beverage formula data, and different hypoglycemic solid beverage formula data may correspond to the same or different raw material information. The hypoglycemic solid beverage formula may be referred to as a formula, which is a formula containing corresponding raw materials and having a hypoglycemic function. The hypoglycemic solid beverage formula data generated by the model is the specific data of the generated hypoglycemic solid beverage formula. The generated hypoglycemic solid beverage formula can be understood as a formula generated by the model, a new formula, etc.

[0058] Optionally, the hypoglycemic effect verification can be a zebrafish experiment verification. After the model training is completed, the embodiment of the present application can obtain experimental data, and then analyze the experimental data to obtain experimental results, and then obtain the result of whether the formula design model verification is passed.

[0059] S13. If found, display the found blood sugar lowering solid beverage formula data and the blood sugar lowering solid beverage formula data generated by the model; or, if not found, display the blood sugar lowering solid beverage formula data generated by the model.

[0060] The embodiment of the present application provides a method for generating a hypoglycemic solid beverage formula based on a deep learning model, which can improve the efficiency and accuracy of hypoglycemic solid beverage formula design through a deep learning model. The core is to use a deep learning algorithm, combined with big data and food science knowledge, to design an effective hypoglycemic solid beverage formula for diabetic patients or people who want to control blood sugar. In the embodiment of the present application, the user can input different raw material information according to their needs, and the application embodiment can provide the user with the hypoglycemic solid beverage formula data found and the hypoglycemic solid beverage formula data generated by the model when searching, and provide the user with the hypoglycemic solid beverage formula data generated by the model when not found, so the present embodiment can generate hypoglycemic solid beverage formula data that meets the personalized needs of different users through the formula design model. At the same time, the embodiment of the present application can also perform zebrafish experiment verification on the formula design model after the model training ends, based on the hypoglycemic solid beverage formula data designed by the formula design model, and then prove the hypoglycemic effect of the hypoglycemic solid beverage formula data, so as to prove the practicality and effectiveness of the formula design model.

[0061] The above process is the application process of the recipe design model. For the recipe design model, reasonable parameter design can improve the accuracy of the recipe design model. Therefore, before executing the application process of the recipe design model, this embodiment can design the training process of the recipe design model. It should be understood that the training process of the recipe design model mainly includes data collection, data processing and model construction. This embodiment takes the training process of an exemplary recipe design model as an example for the following analysis:

[0062] In some optional embodiments, the training process of the formulation design model includes:

[0063] Step 21: Acquire data on plant active ingredients that have been confirmed to have blood sugar lowering function and known formula data on blood sugar lowering solid beverages.

[0064] It should be understood that acquisition can be understood as collection, and function can be understood as efficacy. The plant active ingredient data that has been determined to have the function of lowering blood sugar can refer to the plant active ingredient data with the function of lowering blood sugar that can be collected by consulting literature, the Internet or patent documents; the known blood sugar-lowering solid beverage formula data can refer to the blood sugar-lowering solid beverage formula data that can be collected by consulting literature, the Internet or patent documents.

[0065] Among them, plant active ingredient data include raw material information, active ingredient information in raw materials that play a hypoglycemic function, data sources such as literature, websites or patent documents that provide raw material information and active ingredient information (including: title, author, publication time, patent number, etc.), other functional information of raw materials (such as antioxidant, anti-tumor, immunomodulatory, etc.), etc. The formula data of hypoglycemic solid beverages include the raw materials in the formula, the amount of each raw material added in the formula, other functional information of the formula, and data sources such as literature, websites or patent documents that provide the formula (including: title, author, publication time, patent number, etc.).

[0066] Step 22: pre-process the plant active ingredient data and the solid beverage formula data respectively to form a functional raw material database and a blood sugar-lowering solid beverage formula database.

[0067] It should be understood that preprocessing may include: cleaning, integration, etc. Specifically, the present embodiment can clean the various collected data to remove outliers and noise data, and convert different types of data into a unified data format and range to facilitate the training and application of the formulation design model. Among them, data cleaning involves checking errors, inconsistencies and missing values ​​in the data, and its purpose is to ensure that the database is accurate and complete for subsequent analysis. For example, delete duplicate records, correct erroneous entries, or fill in missing data. Outliers refer to data that is significantly different from other data in the database, and this embodiment removes these data. Noise data refers to data containing random errors or inaccuracies. Removing these data helps to reduce interference in the model training process and improve the performance of the formulation design model. Unifying the data format is a key step to ensure that the formulation design model can correctly understand and process data. Data preprocessing can ensure that the model training process is carried out effectively and ultimately obtain high-quality prediction results.

[0068] It should also be understood that the pre-processed data can be integrated to form a complete database. Therefore, this embodiment can establish a functional raw material database and a hypoglycemic solid beverage formula database through the integration step.

[0069] Exemplary, the functional raw material database includes plant active ingredients with hypoglycemic function, and the plant active ingredients include polysaccharides, polyphenols, alkaloids, flavonoids, saponins and triterpenoids. These plant active ingredients are common in various plant foods, and the data sources can include scientific literature, patent libraries and experimental studies. The hypoglycemic solid beverage formula database includes information such as the name, formula composition structure, raw material ratio relationship and other functions of the solid beverage formula with hypoglycemic function. Establishing a functional raw material database and a hypoglycemic solid beverage formula database not only helps to organize and store a large amount of information, but also can improve the efficiency of data retrieval and analysis.

[0070] Step 23: Use deep learning technology to build a formula design model, and train the formula design model based on the functional raw material database and the blood sugar-lowering solid beverage formula database until the preset stop training condition is met to obtain a pre-trained formula design model. Among them, the functional raw material database can be referred to as the raw material database, and the blood sugar-lowering solid beverage formula database can be understood as the formula library.

[0071] It should be understood that in step 23, the present embodiment trains the formula design model for the development objectives, which include: formula composition structure and raw material ratio relationship, formula mode, synergy between raw materials in the formula, etc. In other words, in order to better consider the solid beverage formula composition structure and raw material ratio relationship, formula mode, and synergy between raw materials in the formula, the present embodiment can use deep learning technology to obtain the formula design model.

[0072] In this embodiment, the formula design model is also called a deep learning model. The deep learning model uses deep learning technology, especially the powerful data processing ability of neural networks, to analyze and learn a large amount of plant active ingredient data and hypoglycemic solid beverage formula data to reflect the complex relationship between the data, which can improve the accuracy of the formula design model.

[0073] It should also be understood that Figure 2 As shown in FIG. 1 , in the training process of the recipe design model, step 21 realizes data collection, step 22 realizes data processing and the establishment of a database (i.e., a raw material library and a recipe library), and step 23 realizes model construction. Figure 2After the deep learning model in the formula design model is trained, the hypoglycemic effect of the solid beverage formula generated by the formula design model is verified by zebrafish experiments, thereby realizing the verification of the practicality and effectiveness of the formula design model. After the verification is passed, the formula design model is deployed on the server, and the application process of the formula design model is executed by the server. During the application process of the formula design model, in S11, the user demand is sent to the server deployed with the deep learning model through the network, so that the server generates a formula based on the user demand through the deep learning model in S13, that is, according to the user input data (including raw material information), the formula design model is used to generate a formula containing the corresponding raw materials and having the hypoglycemic function. Therefore, compared with the prior art, the embodiment of the present application has the following positive effects:

[0074] The embodiment of the present application utilizes deep learning technology in machine learning to study and analyze a large amount of plant active ingredient data and blood sugar-lowering solid beverage formula data to improve the effectiveness and accuracy of formula design; through zebrafish experiment verification, the practicality and effectiveness of the formula design model are guaranteed; according to user needs, personalized blood sugar-lowering solid beverage formulas are provided to users. In other words, the formula design model can provide reliable blood sugar-lowering solid beverages for diabetic patients or people who want to control blood sugar.

[0075] It should be noted that the development goal can be one or more, and the embodiment of the present application does not specifically limit this. Based on the design of different development goals, this embodiment can use multi-task learning to train the formula design model to improve the adaptability of the formula design model in different scenarios.

[0076] In some optional embodiments, in step 23, the formula design model is trained based on the functional raw material database and the hypoglycemic solid beverage formula database until a preset stop training condition is met to obtain a pre-trained formula design model, including:

[0077] Step 231, determine the main task and the slave task; the main task is to generate known blood sugar lowering solid beverage formula data corresponding to the plant active ingredient data; the slave task includes at least one of the following: determining the formula composition structure and raw material ratio relationship in different known blood sugar lowering solid beverage formula data, determining the formula pattern corresponding to the raw material information in the plant active ingredient data, and determining the associated raw material information corresponding to the raw material information in the plant active ingredient data.

[0078] Step 232: Based on the functional raw material database and the blood sugar-lowering solid beverage formula database, a formula design model is trained by multi-task learning until a preset stop training condition is met to obtain a pre-trained formula design model.

[0079] In this embodiment, a multi-task learning method is adopted to train a formula design model. Compared with the prior art, in the model training process, the embodiment of the present application can simultaneously process multiple related tasks through multi-task learning, specifically including: Task 1, analyzing the hypoglycemic solid beverage formula database, understanding the formula composition structure and raw material ratio relationship of different hypoglycemic solid beverage formulas; Task 2: Determine a common formula mode, such as a raw material with a main function of hypoglycemic as the core, and match other raw materials (or auxiliary raw materials) to enhance the hypoglycemic effect or improve the taste, etc.; Task 3: Study the synergy between the raw materials in the hypoglycemic solid beverage formula, such as some raw materials combined together can produce a better hypoglycemic effect; Task 4: Based on the analysis of raw materials and formulas, generate a hypoglycemic solid beverage formula containing user input raw materials. By sharing the underlying fully connected neural network structure, knowledge sharing can be achieved between different tasks, and the learning efficiency and generalization ability of the formula design model can be improved. In contrast, the traditional single task model cannot make full use of the correlation between tasks. In formula design, plant active ingredient data and hypoglycemic solid beverage formula data can be represented as vectors of fixed length, and the fully connected neural network structure can effectively learn the complex relationship between these vectors.

[0080] In some optional embodiments, during the training process of the formulation design model, there is a limitation of easy fitting, and the method further includes: introducing regularization technology, including L1 regularization and L2 regularization. In this embodiment, the regularization technology can overcome the limitation of easy overfitting in the training process.

[0081] In some optional embodiments, during the training process of the recipe design model, the method also includes: initializing hyperparameters such as learning rate, number of training rounds and batch size, using a stochastic gradient descent optimizer to update the parameters of the recipe design model, and using the mean square error loss as the loss function of the recipe design model.

[0082] Among them, the parameters of the recipe design model can be referred to as model parameters. Define hyperparameters such as learning rate lr, number of training rounds epochs and batch size batch_size. The learning rate determines the step size of the model parameter update, the number of training rounds determines the number of times the deep learning model learns the data, and the batch size affects the amount of data processed in each training. After creating the model instance model = SimpleModel(), define the loss function criterion as the mean square error loss, which is a loss function for regression problems. It measures the average square difference between the model output and the true label. In this embodiment, a stochastic gradient descent optimizer can be selected to update the model parameters. During the training process, this embodiment passes the model parameters and learning rate as input to the above-mentioned stochastic gradient descent optimizer. Use the DataLoader class provided by the PyTorch framework to create a data loader train_loader, divide the data set into batches of size batch_size, and set the data order to be randomly disrupted to improve the training effect and generalization ability of the recipe design model. The data set consists of a functional raw material database and a hypoglycemic solid beverage formula database. Each piece of data in the data set includes a pre-processed plant active ingredient data in the functional raw material database and a pre-processed hypoglycemic solid beverage formula data corresponding to the pre-processed plant active ingredient data in the hypoglycemic solid beverage formula database. Model evaluation needs to comprehensively consider multiple task evaluation indicators, and more hyperparameter optimization is required during adjustment, including hyperparameters unique to multi-task learning.

[0083] In some optional embodiments, in terms of the training process, this embodiment can simultaneously optimize the loss functions of multiple tasks, use task relevance for regularization, and consider the differences in data distribution between different tasks, using data balancing technology. Therefore, in the training process of the formulation design model, the method also includes: introducing data balancing technology to cope with the differences in data distribution between different tasks, and the data balancing technology includes data enhancement strategy and specific weight adjustment strategy. The embodiment of the present application uses a more complex data enhancement strategy to cope with data quality and diversity requirements, and the use of a specific weight adjustment strategy can balance the learning progress of different tasks.

[0084] In some optional embodiments, the structure of the deep learning model includes an input layer, a hidden layer and an output layer, wherein, during the model application process, the input layer receives user input data and generates an input feature vector based on the user input data; the user input data includes at least one input parameter; the hidden layer gradually processes the input feature vector through two fully connected layers to obtain an intermediate feature vector; the output layer generates blood sugar lowering solid beverage formula data corresponding to the raw material information based on the intermediate feature vector.

[0085] In some optional embodiments, the hidden layer includes a fully connected layer fc1, a ReLU activation function, and a fully connected layer fc2. During the model training process, the deep learning model can be trained under the PyTorch framework. First, the simpleModel is defined to provide an infrastructure for subsequent training and application, aiming to perform regression prediction on given input data. During the deep learning model processing process, it gradually processes the corresponding input data through two fully connected layers fc1 and fc2. First, the fully connected layer fc1 maps the specific attributes (i.e., input features) in the user input data to the first linear feature vector. The input feature is obtained through selection and extraction, and can better represent the relationship between the input data and the model output. Secondly, the fully connected layer fc2 further maps the first linear feature vector to an output, thereby completing the entire data processing and prediction process. This hierarchical structure allows the deep learning model to gradually learn the complex mapping relationship from input data to output. Specifically, in the forward propagation function forward of the deep learning model, the user input data is first linearly transformed through the fully connected layer fc1, and then the ReLU activation function is applied to increase the nonlinear expression ability of the deep learning model, and finally the final output result is obtained through the fully connected layer fc2.

[0086] During model training, the number of outer loops corresponds to the number of training rounds (epochs). The number of initial training rounds is also called the set epochs. At the beginning of each round, the cumulative loss is initialized to zero. The inner loop traverses each batch of data in the data loader, such as Figure 3 As shown. For each batch, the gradient zeroing operation is first performed on the stochastic gradient descent optimizer to prevent the gradient of the previous round from having an adverse effect on the current model parameter update. Then forward propagation is performed to pass the input data to the deep learning model to obtain the output result. Subsequently, the loss is calculated by backpropagation (calculating the loss value between the output of the deep learning model and the true label) to determine the gradient of the loss relative to the model parameters. After that, the stochastic gradient descent optimizer is used to update the model parameters based on the calculated gradient. Finally, the loss of the current batch is accumulated, and the training information is periodically printed, which includes indicators such as the current round number, average loss, and perplexity, in order to observe the progress and stability of the training process.

[0087] In other words, throughout the model training process, the deep learning model is committed to minimizing the mean square error loss function. By continuously adjusting its own parameters, the output of the deep learning model is made as close to the true label value as possible. In each training round, after obtaining batch data from the data loader, the gradient is first cleared to prevent gradient accumulation, and then forward propagation is performed to calculate the loss through forward propagation, so that the input data passes through two fully connected layers and the ReLU activation function in the middle to obtain the output. The loss is then calculated, and the gradient of the loss function with respect to the model parameters is determined through back propagation. Finally, the stochastic gradient descent optimizer is used to update the model parameters according to the gradient. By repeating this process continuously, the deep learning model gradually learns the mapping relationship between the input data and the output label, and can then accurately predict the new user input data.

[0088] In some other optional embodiments, the two fully connected layers include a first fully connected layer and a second fully connected layer; the hidden layer includes a first fully connected layer and a second fully connected layer connected in sequence; the first fully connected layer includes a first linear transformation module and a first ReLU activation function connected in sequence, and the second fully connected layer includes a second linear transformation module and a second ReLU activation function connected in sequence. Combined with the structure of the hidden layer, the function of the hidden layer is refined. Specifically, during the model application process, the hidden layer gradually processes the input feature vector through two fully connected layers, including:

[0089] Step a1: linearly transform the input feature vector through the first linear transformation module to obtain a first linear feature vector. Step a2: convert the first linear feature vector into a nonlinear feature vector using the first ReLU activation function. Step a3: linearly transform the nonlinear feature vector through the second linear transformation module to obtain a second linear feature vector. Step a4: convert the second linear feature vector into an intermediate feature vector using the second ReLU activation function.

[0090] The first linear transformation module and the second linear transformation module are linear transformation modules with the same structure or different structures, and are located in different fully connected layers. Similarly, the first ReLU activation function and the second ReLU activation function correspond to ReLU activation functions with the same formula or different formulas, and are located in different fully connected layers.

[0091] In this embodiment, through step-by-step linear transformation and nonlinear conversion, the deep learning model can automatically select and extract the most useful features for the task, reducing the risk of overfitting. At the same time, each fully connected layer is divided into a linear transformation module and a ReLU activation function, making the design of the deep learning model more modular. This modular design facilitates the expansion and optimization of the deep learning model, and more layers can be added or the structure of the layer can be adjusted according to specific task requirements.

[0092] In some optional embodiments, the hidden layer may further include a residual connection module located between the first fully connected layer and the second fully connected layer; after step a2, converting the first linear feature vector into a nonlinear feature vector using the first ReLU activation function, the method further includes:

[0093] Step a5: Add the input feature vector and the nonlinear feature vector to obtain a feature vector after residual connection.

[0094] Residual connections enable deep learning models to utilize both input feature vectors and nonlinear feature vectors. This feature fusion mechanism enhances the expressiveness of the model, allowing deep learning models to better capture complex patterns and relationships in the data. In addition, the introduction of the residual connection module makes the design of deep learning models more modular, and residual connections can be easily added or removed between different layers without significantly affecting the overall performance of the deep learning model. This modular design improves the scalability and flexibility of deep learning models.

[0095] After the model training process is completed, the present embodiment can deploy the deep learning model to the server. In the deployment code of the server, a model instance is first created, and then the model parameters are saved based on the PyTorch framework, and the deep learning model is set to the evaluation mode to ensure that training operations such as gradient calculation are not performed during the reasoning process. Further, the present embodiment can adopt Web service deployment or containerized deployment. Among them, the Web service deployment creates a Web service with the help of a Web framework (such as Flask or Django). The deep learning model is loaded into the Web service, and the user input data is received through the Hypertext Transfer Protocol (HTTP) request based on the network, and the result is returned after model reasoning. For example, when using Flask, the present embodiment can create a route to receive a POST request, obtain input data from the POST request, and use the function provided by the PyTorch framework to convert the preprocessed user input data into a tensor format. Then the user input data in tensor format is passed into the deep learning model to obtain the output, and finally the result is returned in JSON format. Containerized deployment refers to using Docker to package the deep learning model and its dependencies into a container image, and by creating a Docker file, determining the base image, installing dependencies, copying the model file, and setting the startup command. Build container images and run containers to ensure the consistency and portability of deep learning models in different environments. This embodiment can also use tools such as Postman or curl to send requests to deployed services to verify whether the deep learning model can respond correctly and return accurate results. At the same time, monitor the resource usage of the server to ensure the performance and stability of the deep learning model.

[0096] like Figure 4 As shown, another method for generating a hypoglycemic solid beverage formula based on a deep learning model provided in this embodiment includes obtaining user needs, using a deep learning model to perform formula design to obtain a formula generated by the model, and then verifying the formula generated by the model. After the formula passes the verification, it is used as a hypoglycemic solid beverage formula. Specifically, Figure 4 As shown in the figure, the model application process is illustrated as follows:

[0097] Step 1: Obtain user input data for reflecting user needs. The user input data may be the name of the raw material that needs to be formulated by the user. In order to efficiently obtain the raw material name, this embodiment may develop a user-friendly interface or tool that may integrate automatic completion and error checking functions.

[0098] Step 2: Send the acquired user input data to the target server where the deep learning model is deployed. First, the user input data (such as the name of the raw material) is encapsulated. At this time, the user input data is in the network interface layer. User input data usually exists in binary form, but in order to transmit on the network, the user input data can be encapsulated into specific data packets (or data frames). Each data packet contains the data itself, the source address, the destination address, and some necessary control information. Subsequently, the data packet can enter the transport layer, which is responsible for establishing a communication connection between the source and the destination. The Transmission Control Protocol (TCP) ensures the integrity and sequential transmission of the data packet to the target server through confirmation mechanisms, retransmission mechanisms, and flow control. When the data packet enters the network layer, the network layer is responsible for assigning an Internet Protocol Address (IP) to the data packet and sending the data packet from the source network to the target network based on the information in the routing table. At the network layer, the data packet may be split and reassembled multiple times to adapt to the bandwidth and transmission capacity of different network links. The data link layer encapsulates the data packet into a frame and adds necessary control information, such as the source Media Access Control (MAC) address and the destination MAC address, so that it can be transmitted in a local area network or a wide area network. At the same time, the data link layer is also responsible for error detection and correction to ensure the accuracy of the data frame during transmission. When the data packet arrives at the target server, the network interface card (NIC) on the target server will perform a receiving operation and pass the data packet to the operating system of the target server for processing. In this embodiment, the source server and the target server can both refer to computers.

[0099] Step 3, generate a recipe. Specifically, in this embodiment, the target server can process the raw material information contained in the received user input data, first retrieve the existing information in the blood sugar lowering solid beverage recipe database, if the raw material information exists, then call the recipe data of all existing recipes in the blood sugar lowering solid beverage recipe database (including: the recipe composition structure containing the corresponding raw materials and the raw material ratio relationship, the literature or patent documents of the recipe, etc.); secondly, through the deep learning model, take the raw material information as input, and use its combination with other raw materials and the raw material ratio relationship as output, and provide one or more model-generated recipes. On the contrary, if there is no recipe containing the raw material information in the blood sugar lowering solid beverage recipe database, the recipe generated by the model is directly used. Finally, all the retrieved existing recipes (if not, they will not be displayed) and the recipes generated by the model are recommended to the user.

[0100] Exemplarily, when the user inputs "corn silk powder", the formula generated by the model is made up of the following components in parts by mass: 8 parts of corn silk powder, 2 parts of astragalus powder, 10 parts of barley grass juice powder, and 10 parts of burdock powder.

[0101] Step 4: Verify the function. In order to evaluate the hypoglycemic function of the formula generated by the model involved in the embodiment of the present application, and then determine the practicality of the deep learning model, this embodiment conducts a zebrafish experiment on the hypoglycemic function of the formula generated by the model, and the experimental results are as follows:

[0102] Corn silk powder, astragalus powder, barley grass juice powder and burdock seed powder are weighed in a raw material ratio of 4:1:5:5 and mixed evenly to obtain a solid beverage.

[0103] In the present embodiment, zebrafish animals are grouped and administered. For example, 100 adult zebrafish (4 months old, body weight 0.5 ± 0.1g, AB wild type, male-female ratio 1: 1) are from the same growth environment. Zebrafish are raised in a desktop aquatic system, and the feeding environment is 14h light and 10h dark alternation, and the temperature is 27 ± 2 ° C. These zebrafish are adapted to the feeding environment for 7 days, keep filtering and bursting oxygen, and feed with ordinary feed once a day. In order to induce type 2 diabetes (Type 2 Diabetes Mellitus, T2DM) model, zebrafish are randomly divided into 5 groups, 20 in each group. Normal control group: blank control group (Normal Control, NC). Diabetes model group: model control group (Model Control, MC), low concentration solid beverage treatment group (Low Solid Beverage, LSB), medium concentration solid beverage treatment group (Medium Solid Beverage, MSB) and high concentration solid beverage treatment group (High Solid Beverage, HSB). The zebrafish in the diabetic model group were placed in a 111 mM glucose solution for 10 hours a day and fed a high-fat diet containing 10% cholesterol for 28 days. The blank control group was raised in water without glucose.

[0104] During the experiment, zebrafish were fed once a day. Each fish was fed 20 mg of feed per day. After 28 days, the glucose soaking was stopped. Subsequently, all zebrafish were placed in a clean water environment for feeding. The NC group was fed with ordinary feed, and the MC group was fed with the same feed. The LSB, MSB, and HSB groups were fed with mixed feeds with 10%, 30%, and 50% solid drinks added to the ordinary feed, respectively. The process lasted for 28 days.

[0105] At the end of the experiment (e.g., day 56), the zebrafish were fasted for 12 hours and then soaked in 2-4°C cold water for 15 minutes. At the same time, when collecting samples, the surface moisture of the zebrafish was wiped off with a paper towel, and their body length and weight were measured, and their body mass index (BMI) was calculated based on the formula: body weight (g) divided by the square of body length (cm 2 ).

[0106] The sample weight (g) was weighed, 9 volumes of physiological saline were added, and the fish tissue was treated with a high-speed handheld homogenizer for 1 min to ensure complete tissue lysis. Centrifugation was performed at 2500r / min for 10 min at 4°C, and the supernatant was collected for biochemical analysis. The assay was performed according to the procedures and steps specified in the instructions provided with the kit. According to the manufacturer's instructions, the biochemical kit was used to detect the levels of glucose (GLU), triglycerides (TG), protein quantification (TP), total cholesterol (T-CHO), high-density lipoprotein (HDL-C), and low-density lipoprotein (LDL-C) in the whole body samples. In all experiments, results with P < 0.05 were considered significant. Among them, * indicates P < 0.05, ** indicates P < 0.01, *** indicates P < 0.001, and **** indicates P < 0.0001.

[0107] Figure 5 As shown, the BMI, GLU, TG, T-CHO, and LDL-C of the diabetes model group were significantly higher than those of the normal group, and the HDL-C was significantly lower than that of the normal group, which shows that feeding with 111mM glucose solution combined with cholesterol feed can induce glucose and lipid metabolism disorders in adult zebrafish, causing them to suffer from diabetes, and this phenomenon is similar to the test results of the four clinical blood lipids of diabetes. After stopping the high-sugar and high-fat diet, BMI, GLU, TG, T-CHO, and LDL-C remained at a relatively high level, and HDL-C remained at a relatively low level, indicating that the zebrafish type 2 diabetes model constructed by the embodiment of the present application is stable and effective. Based on this, a model formula intervention experiment was carried out on sick zebrafish. Solid beverage intervention can reduce the obesity level, glucose level, triglyceride level, cholesterol level, and low-density lipoprotein level of zebrafish, and increase the high-density lipoprotein level, which shows that solid beverage intervention can improve the symptoms of type 2 diabetes in zebrafish.

[0108] The above experimental data can show that the solid beverage formula obtained by the method described in the examples of the present application has a significant blood sugar lowering function.

[0109] The use of the above-mentioned blood sugar lowering effect verification method can improve the reliability of the formula generated by the model.

[0110] Figure 6 A schematic diagram of a device for generating a blood sugar-lowering solid beverage formula based on a deep learning model provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, the device comprises:

[0111] The acquisition module 61 is used to acquire user input data, where the user input data includes raw material information.

[0112] The search generation module 62 is used to search for blood sugar lowering solid beverage formula data corresponding to the raw material information from the blood sugar lowering solid beverage formula database, and generate blood sugar lowering solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; the formula design model has passed the blood sugar lowering effect verification at the end of training.

[0113] The display module 63 is used to display the found blood sugar lowering solid beverage formula data and the blood sugar lowering solid beverage formula data generated by the model when found.

[0114] Figure 6 The device for generating a hypoglycemic solid beverage formula based on a deep learning model can be executed Figure 1 The implementation principle and technical effects of the method for generating a hypoglycemic solid beverage formula based on a deep learning model described in the illustrated embodiment will not be described in detail. The specific manner in which each module and unit performs operations in the hypoglycemic solid beverage formula generation device based on a deep learning model in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0115] In one possible design, Figure 6 The blood sugar lowering solid beverage formula generation device based on the deep learning model of the illustrated embodiment can be implemented as a computing device, such as Figure 7 As shown, the computing device may include a storage component 71 and a processing component 72 .

[0116] The storage component 71 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 72 .

[0117] The processing component 72 is used to: obtain user input data, which includes raw material information; search for blood sugar lowering solid beverage formula data corresponding to the raw material information from a blood sugar lowering solid beverage formula database, and generate blood sugar lowering solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; the formula design model has passed the blood sugar lowering effect verification at the end of training; if found, display the found blood sugar lowering solid beverage formula data and the blood sugar lowering solid beverage formula data generated by the model.

[0118] The processing component 72 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0119] The storage component 71 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0120] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0121] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0122] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0123] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0124] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is a method for generating a blood sugar lowering solid beverage formula based on a deep learning model.

[0125] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0126] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0127] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for generating a hypoglycemic solid beverage formula based on a deep learning model, characterized in that: include: Acquiring user input data, wherein the user input data includes raw material information; Searching for hypoglycemic solid beverage formula data corresponding to the raw material information from a hypoglycemic solid beverage formula database, and generating hypoglycemic solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; The formula design model has been verified by the blood sugar lowering effect at the end of the training; If found, the found blood sugar lowering solid beverage formula data and the blood sugar lowering solid beverage formula data generated by the model are displayed.

2. The method according to claim 1, characterized in that The training process of the formulation design model includes: Obtain data on plant active ingredients that have been confirmed to have blood sugar lowering function and known blood sugar lowering solid beverage formula data; Preprocessing the plant active ingredient data and the solid beverage formula data respectively to form a functional raw material database and a blood sugar-lowering solid beverage formula database; A formula design model is constructed using deep learning technology, and the formula design model is trained based on the functional raw material database and the blood sugar-lowering solid beverage formula database until a preset stop training condition is met to obtain a pre-trained formula design model.

3. The method according to claim 2, characterized in that The formula design model is trained based on the functional raw material database and the hypoglycemic solid beverage formula database until a preset stop training condition is met to obtain a pre-trained formula design model, including: Determine a main task and a sub-task; the main task is to generate known hypoglycemic solid beverage formula data corresponding to the plant active ingredient data; the sub-task includes at least one of the following: determine the formula composition structure and raw material ratio relationship in different known hypoglycemic solid beverage formula data, determine the formula mode corresponding to the raw material information in the plant active ingredient data, and determine the associated raw material information corresponding to the raw material information in the plant active ingredient data; Based on the functional raw material database and the blood sugar-lowering solid beverage formula database, a formula design model is trained by multi-task learning until a preset stop training condition is met to obtain a pre-trained formula design model.

4. The method according to claim 2, characterized in that: During the training process of the formulation design model, the method further includes: Introduce regularization techniques, including L1 regularization and L2 regularization.

5. The method according to claim 2, characterized in that: During the training process of the formulation design model, the method further includes: Hyperparameters such as learning rate, number of training rounds, and batch size are initialized, and the parameters of the recipe design model are updated using the stochastic gradient descent optimizer, and the mean squared error loss is used as the loss function of the recipe design model.

6. The method according to claim 3, characterized in that During the training process of the formulation design model, the method further includes: A data balancing technique is introduced to cope with the difference in data distribution between different tasks. The data balancing technique includes a data enhancement strategy and a specific weight adjustment strategy.

7. The method according to claim 1, characterized in that The structure of the deep learning model includes an input layer, a hidden layer and an output layer, wherein the input layer receives user input data and generates an input feature vector based on the user input data; the user input data includes at least one input parameter; the hidden layer gradually processes the input feature vector through two fully connected layers to obtain an intermediate feature vector; the output layer generates blood sugar-lowering solid beverage formula data corresponding to the raw material information based on the intermediate feature vector.

8. The method according to claim 7, characterized in that The two fully connected layers include a first fully connected layer and a second fully connected layer; the hidden layer includes a first fully connected layer and a second fully connected layer connected in sequence; the first fully connected layer includes a first linear transformation module and a first ReLU activation function connected in sequence, and the second fully connected layer includes a second linear transformation module and a second ReLU activation function connected in sequence; The hidden layer processes the input feature vector step by step through two fully connected layers, including: Performing a linear transformation on the input feature vector through a first linear transformation module to obtain a first linear feature vector; Using a first ReLU activation function to convert the first linear feature vector into a nonlinear feature vector; Performing a linear transformation on the nonlinear feature vector by a second linear transformation module to obtain a second linear feature vector; The second linear feature vector is converted into an intermediate feature vector using a second ReLU activation function.

9. The method according to claim 8, characterized in that The hidden layer includes a residual connection module between the first fully connected layer and the second fully connected layer; After converting the first linear feature vector into a nonlinear feature vector using a first ReLU activation function, the method further includes: The input feature vector and the nonlinear feature vector are added to obtain the feature vector after residual connection.

10. A device for generating a formula for hypoglycemic solid beverage based on a deep learning model, characterized in that: include: An acquisition module, used for acquiring user input data, wherein the user input data includes raw material information; A search and generation module, used to search for hypoglycemic solid beverage formula data corresponding to the raw material information from a hypoglycemic solid beverage formula database, and generate hypoglycemic solid beverage formula data corresponding to the raw material information through a pre-trained formula design model; The formula design model has been verified by the blood sugar lowering effect at the end of the training; The display module is used to display the found blood sugar lowering solid beverage formula data and the blood sugar lowering solid beverage formula data generated by the model when found.

Images