Traditional Chinese medicine data classification processing method and system based on graph loop network, and medium

Abstract

The invention provides a traditional Chinese medicine data classification processing method and system based on a graph loop network, and a medium. The method comprises the following steps: collecting traditional Chinese medicine data; preprocessing the collected traditional Chinese medicine data; constructing a graph loop network; and inputting the preprocessed traditional Chinese medicine data into the graph loop network to obtain a traditional Chinese medicine data classification result. In the application, the first-level graph convolution layer focuses on local feature extraction and enhancement of the perception ability of the model, and the high-level graph convolution layer integrates global information, captures detail information around nodes and enhances the understanding ability of the model. And the feedback mechanism layer adaptively adjusts the mechanism, monitors data in real time, adjusts network parameters, and reduces the over-fitting risk. The traditional Chinese medicine knowledge is fused into the graph structure construction process, multi-dimensional features can be extracted from the traditional Chinese medicine data, local and global information can be comprehensively captured, the classification and screening performance of the unstructured traditional Chinese medicine data is improved, and rapid data selection is provided for new medicine research and development or medicine compatibility and the like aiming at diseases.

Description

Technical Field

[0001] The present application relates to the technical field of processing and classification of traditional Chinese medicine data, and particularly relates to a method, system and medium for classifying and processing traditional Chinese medicine data based on a graph recurrent network. Background Art

[0002] With the rapid development of information technology, the combination of traditional Chinese medicine and modern technology has become increasingly deep. In the process of disease diagnosis, treatment and new drug research and development in traditional Chinese medicine, the compatibility of drugs for specific diseases is particularly important. Due to the complexity (including various types of text data and image data) and professionalism of traditional Chinese medicine data, how to effectively extract information, identify correlations and perform rapid classification has become an urgent problem to be solved.

[0003] Traditional Chinese medicine data is usually unstructured and diverse. Traditional data processing and analysis methods often seem powerless when dealing with these unstructured, two-dimensional or even multi-dimensional data, resulting in low classification efficiency and insufficient accuracy, which in turn affects the speed of drug selection and the progress of new drug research and development.

[0004] In addition, the dynamic nature of traditional Chinese medicine data causes data characteristics to change over time. Traditional models lack the ability to adapt to new data and rely mostly on manual feature selection, which makes it difficult to automatically mine deep features in the data. At the same time, a large amount of domain knowledge is contained in traditional Chinese medicine theory, and existing technologies often cannot effectively incorporate this knowledge, further reducing the accuracy of analysis and classification.

[0005] Therefore, there is an urgent need for a real-time classification processing method that can quickly and efficiently perform medicinal material selection and drug compatibility for different structured and dynamically changing traditional Chinese medicine data, so as to improve the treatment effect for specific diseases and the progress of new drug research and development. Summary of the Invention

[0006] The present application provides a method, system and medium for classifying and processing traditional Chinese medicine data based on a graph recurrent network to solve the technical problem that existing traditional Chinese medicine data processing and analysis technologies cannot effectively process unstructured and multi-dimensional data.

[0007] In a first aspect, the present application provides a method for classifying and processing traditional Chinese medicine data based on a graph recurrent network, including:

[0008] Collect traditional Chinese medicine data, where the traditional Chinese medicine data is text data or image data including one or more of the name of medicinal materials, properties, place of origin, chemical composition and pharmacological effects;

[0009] Preprocess the collected traditional Chinese medicine data;

[0010] Construct a graph recurrent network, and the expression of the graph recurrent network is:

[0011]

[0012] Among them, M represents the graph recurrent network, C (1) represents the graph convolution factor of the first-level graph convolution layer, B i represents the i-th traditional Chinese medicine data after preprocessing, γ represents the model hyperparameter factor, Z (L) represents the output feature of the L-th feedback mechanism layer, V (L) represents the output feature of the L-th advanced graph convolution layer, G (L) represents the dilation convolution factor of the L-th advanced graph convolution layer, B max represents the maximum value of the traditional Chinese medicine data after preprocessing, B min represents the minimum value of the traditional Chinese medicine data after preprocessing, V (L-1) represents the output feature of the (L - 1)-th advanced graph convolution layer, μ L-1 represents the mean value V of the output feature of the (L - 1)-th advanced graph convolution layer (L-1) of, σ L-1 represents the output feature V of the (L - 1)-th advanced graph convolution layer (L-1) of the standard deviation, W j (L) represents the j-th weight factor of the L-th advanced graph convolution layer, W j (1) represents the j-th weight factor of the first-level graph convolution layer;

[0013] Input the preprocessed traditional Chinese medicine data into the graph recurrent network to obtain the classification result of the traditional Chinese medicine data.

[0014] Optionally, in the preprocessing of the collected traditional Chinese medicine data, the following formula is used to preprocess the traditional Chinese medicine data:

[0015]

[0016] Among them, B represents the traditional Chinese medicine data after preprocessing, A represents the traditional Chinese medicine data before preprocessing, m represents the number of traditional Chinese medicine data, A i is the i-th traditional Chinese medicine data before preprocessing, μ is the mean value of the traditional Chinese medicine data A before preprocessing, σ is the standard deviation of the traditional Chinese medicine data A before preprocessing, A max represents the maximum value of the traditional Chinese medicine data A before preprocessing, A min represents the minimum value of the traditional Chinese medicine data A before preprocessing, δ represents the chi-square distribution density function.

[0017] Optionally, the graph recurrent network includes a first-level graph convolution layer, multiple advanced graph convolution layers and multiple feedback mechanism layers,

[0018] Among them, the first-level graph convolution layer performs local feature extraction based on the preprocessed traditional Chinese medicine data,

[0019] The multiple advanced graph convolutional layers integrate global information based on the convolutional features of the hidden layer of the preprocessed traditional Chinese medicine data.

[0020] The multiple feedback mechanism layers monitor data flow and graph structure changes in real time based on the adaptive adjustment mechanism.

[0021] Optionally, the expression of the first-level graph convolutional layer is:

[0022]

[0023] where V (1) represents the output feature of the first-level graph convolutional layer, C (1) represents the graph convolution factor of the first-level graph convolutional layer, n represents the feature scaling factor, W j (1) represents the j-th weight factor of the first-level graph convolutional layer, D represents the graph convolution degree factor, μ1 represents the mean of the preprocessed traditional Chinese medicine data B, σ1 represents the standard deviation of the preprocessed traditional Chinese medicine data B, N i represents the i-th preprocessed traditional Chinese medicine data, B i-1 represents the (i - 1)-th preprocessed traditional Chinese medicine data.

[0024] Optionally, the expression of the L-th advanced graph convolutional layer in the multiple advanced graph convolutional layers is:

[0025]

[0026] where V (L) represents the output feature of the L-th advanced graph convolutional layer, G (l) represents the dilated convolution factor of the L-th advanced graph convolutional layer, C (1) represents the graph convolution factor of the first-level graph convolutional layer, D represents the graph convolution degree factor, V (L-1) represents the output feature of the (L - 1)-th advanced graph convolutional layer, B i represents the i-th preprocessed traditional Chinese medicine data, W j (L) represents the j-th weight factor of the L-th advanced graph convolutional layer, μ L-1 represents the mean of the output feature V (L-1) of the (L - 1)-th advanced graph convolutional layer, σ L-1 represents the standard deviation of the output feature V (L-1) of the (L - 1)-th advanced graph convolutional layer.

[0027] Optionally, the expression of the L-th feedback mechanism layer in the multiple feedback mechanism layers is:

[0028]

[0029] Among them, Z (L) represents the output feature of the L-th feedback mechanism layer, and θ1, θ2 represent feature adjustment factors, satisfying 0 < θ1 + θ2 < 1, V (1) represents the output feature of the first-level graph convolutional layer, V (L) represents the output feature of the L-th high-level graph convolutional layer, V (L-1) represents the output feature of the (L - 1)-th high-level graph convolutional layer, μ L-1 represents the output feature V of the (L - 1)-th high-level graph convolutional layer (L-1) mean, σ L-1 represents the output feature V of the (L - 1)-th high-level graph convolutional layer (L-1) standard deviation.

[0030] In a second aspect, the present application also provides a traditional Chinese medicine data classification processing system based on a graph recurrent network, including:

[0031] A data acquisition unit configured to acquire traditional Chinese medicine data, where the traditional Chinese medicine data is text data or image data including one or more of the name of traditional Chinese medicine materials, properties, origin, chemical components, and pharmacological effects;

[0032] A preprocessing unit configured to preprocess the acquired traditional Chinese medicine data;

[0033] A graph recurrent network unit configured to provide and update a graph recurrent network, and the expression of the graph recurrent network is:

[0034]

[0035] Among them, M represents the graph recurrent network, C (1) represents the graph convolution factor of the first-level graph convolutional layer, B i represents the i-th preprocessed traditional Chinese medicine data, γ represents the model hyperparameter factor, Z (L) represents the output feature of the L-th feedback mechanism layer, V (L) represents the output feature of the L-th high-level graph convolutional layer, G (L) represents the dilation convolution factor of the L-th high-level graph convolutional layer, B max represents the maximum value of the preprocessed traditional Chinese medicine data, B min represents the minimum value of the preprocessed traditional Chinese medicine data, V (L-1) represents the output feature of the (L - 1)-th high-level graph convolutional layer, μ L-1 represents the output feature V of the (L - 1)-th high-level graph convolutional layer (L-1) mean, σ L-1 represents the output feature V of the (L - 1)-th high-level graph convolutional layer (L-1) standard deviation, W j (L)denotes the j-th weight factor of the L-th high-level graph convolutional layer, W j (1) denotes the j-th weight factor of the first-level graph convolutional layer;

[0036] a classification unit, which obtains a classification result of traditional Chinese medicine data based on the traditional Chinese medicine data input into the graph recurrent network.

[0037] Optionally, the preprocessing unit preprocesses the traditional Chinese medicine data using the following formula:

[0038]

[0039] where B represents the preprocessed traditional Chinese medicine data, A represents the traditional Chinese medicine data before preprocessing, m represents the number of traditional Chinese medicine data, A i is the i-th traditional Chinese medicine data before preprocessing, μ is the mean of the traditional Chinese medicine data A before preprocessing, σ is the standard deviation of the traditional Chinese medicine data A before preprocessing, A max represents the maximum value of the traditional Chinese medicine data A before preprocessing, A min represents the minimum value of the traditional Chinese medicine data A before preprocessing, and δ represents the chi-square distribution density function.

[0040] In a third aspect, the present application further provides an electronic device, including:

[0041] a memory configured to store instructions

[0042] a processor configured to call the instructions from the memory to execute the above-mentioned method for classifying traditional Chinese medicine data based on a graph recurrent network.

[0043] In a fourth aspect, the present application further provides a computer-readable storage medium, on which instructions are stored, and the instructions are used to cause a machine (or a processor) to execute the above-mentioned method for classifying traditional Chinese medicine data based on a graph recurrent network.

[0044] By collecting and preprocessing traditional Chinese medicine (TCM) data, this application can extract multi-dimensional features from TCM data (traditional classification methods often focus on single features or fewer feature dimensions, easily overlooking the diversity and complexity of information). Through the design of the graph recurrent network in this application, it can simultaneously focus on local features and global features, thereby capturing richer data information. This multi-dimensional feature extraction process enables the model to provide more accurate classification results when facing complex TCM data. The graph recurrent network consists of a first-level graph convolutional layer combined with multiple high-level graph convolutional layers and multiple feedback mechanism layers. Among them, the first-level graph convolutional layer focuses on local feature extraction, enhancing the model's perception ability of local structures. This design enables the model to deeply understand the detailed information of each node in the data. Multiple high-level graph convolutional layers integrate global information, capturing the detailed information around nodes through the feature aggregation of adjacent nodes, and enhancing the model's ability to understand complex interrelationships. The integration of this local and global information enhances the model's understanding ability, making it show higher accuracy and robustness in classification tasks. Through the adaptive adjustment mechanism of multiple feedback mechanism layers, it can monitor the data flow and changes in the graph structure in real time, adjust network parameters to reduce the risk of overfitting, ensure the generalization ability of the model, and cope with the ever-changing TCM data environment. Therefore, the graph recurrent network can extract multi-dimensional features from TCM data by using multi-level graph convolutional layers, comprehensively capture local and global information, and significantly improve the classification performance. Thus, by integrating the knowledge in the field of traditional Chinese medicine into the graph structure construction process, the model can better capture the features in traditional Chinese medicine theory, not only improving the classification performance of the model, but also enhancing the credibility of the results, making the results of TCM data analysis more valuable in practice.

[0045] This application can also achieve: quickly and real-time obtain effective TCM data information from a large amount of TCM data, enabling researchers to deeply analyze TCM theory, drug compatibility, and disease treatment plans, etc. Through the analysis of TCM databases, classic TCM literature, clinical cases, and drug ingredients, researchers can obtain new medical insights and promote the modernization and scientific process of traditional Chinese medicine. This method also has important application value in the field of drug research and development. Through the analysis of TCM ingredients and effects, etc., this method can quickly classify TCM data, obtain one or more TCM materials that need to be compatible or treated, thus reducing the selection time of TCM materials. This can not only increase the research and development and compatibility speed of new drugs or combined drugs, but also help to quickly obtain a more reasonable treatment plan. In addition, this method can also provide an in-depth understanding of the action mechanism and potential side effects of existing drugs. This process can accelerate drug discovery and optimize the use of existing drugs. Brief Description of the Drawings

[0046] To more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0047] Figure 1 is a schematic flowchart of the traditional Chinese medicine data classification and processing method based on the graph recurrent network provided by the present application;

[0048] Figure 2 is a schematic diagram of the traditional Chinese medicine data classification and processing system based on the graph recurrent network provided by the present application. Detailed implementation manners

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application. In addition, it should be understood that the specific implementation manners described herein are only used to illustrate and explain the present application, and are not used to limit the present application. In the present application, unless otherwise stated, the orientation words such as "upper", "lower", "left", and "right" generally refer to the upper, lower, left, and right in the actual use or working state of the device, specifically the drawing direction in the accompanying drawings.

[0050] The present application provides a traditional Chinese medicine data classification and processing method based on the graph recurrent network, which will be described in detail below. It should be noted that the description order of the following embodiments does not limit the preferred order of the embodiments of the present application. And in the following embodiments, each embodiment is described with emphasis. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0051] Please refer to Figure 1 , the present application provides a traditional Chinese medicine data classification and processing method based on the graph recurrent network. By combining domain knowledge, the traditional theory of traditional Chinese medicine is integrated into the process of constructing the graph structure, and professional knowledge of traditional Chinese medicine is introduced into the node representation to enhance the model's understanding ability of traditional Chinese medicine data.

[0052] As an emerging deep learning technology, the graph recurrent network has received extensive attention due to its excellent graph structure analysis ability. The graph recurrent network can automatically learn the features of nodes and edges, and through modeling the relationships between data, it can better process data with graph structures.

[0053] The flexibility and adaptability of graph recurrent networks enable them to excel in fields such as social networks, recommendation systems, drug classification and discovery, etc., and can effectively capture complex relationships and dynamic changes. In the field of healthcare, graph recurrent networks are gradually being applied to drug discovery and genomic data analysis, showing great potential in processing complex biomedical data. Graph-based analysis methods are particularly suitable for dealing with the complex relationships in traditional Chinese medicine (TCM) data, can quickly classify TCM data with different structures, and can quickly mine the connections between drug compatibility and symptoms, diseases, and treatment plans, thus providing a new technical path for TCM data analysis.

[0054] Please refer to Figure 1 , a method for classifying and processing TCM data based on graph recurrent networks, specifically including the following steps:

[0055] S100. Collect TCM data, where the TCM data is text data or image data including one or more of the name of traditional Chinese medicine materials, properties, origin, chemical components, and pharmacological effects;

[0056] TCM data usually includes TCM medical data and TCM pharmaceutical data. The clinical data involved in TCM medical data (the data type can be text data, image data, etc., but is not limited to this) is usually unstructured and diverse, including patient medical records (including physiological indicators such as height, weight, blood pressure, medical history records, medical images, etc.), symptom descriptions, pulse conditions, tongue conditions, drug use, medical CT data, etc. (including text data, image data, etc.). TCM pharmaceutical data is usually also unstructured and diverse, and can include the name of traditional Chinese medicine materials, properties, origin, chemical components, pharmacological effects, etc. The data type can be text data, image data, etc., but is not limited to this.

[0057] TCM data can usually be collected from multiple channels such as TCM clinical cases, drug databases (such as the Chinese TCM Periodical Literature Database, TCM Pharmacology Literature Database, TCM Chemistry Literature Database, TCM Clinical Diagnosis and Treatment Literature Database, Pharmaron Cloud - TCM Database Group, NMPA TCM Database, HIT 2.0 SymMap, TCMSID, LTM - TCM, ETCM, etc.), and literature materials (databases such as CNKI data, data publicly disclosed by the patent office, etc.) to collect relevant data (the data type can be text data, image data, etc., but is not limited to this), ensuring the diversity and representativeness of the data.

[0058] This application takes TCM pharmaceutical data as an example for illustration. However, this application can be equivalently applied to at least the above - mentioned TCM medical data.

[0059] It should be noted that this application only describes the classification of traditional Chinese medicine data and can realize the rapid acquisition of new drug research and development or drug compatibility schemes for specific diseases or symptoms. However, the present invention can synchronously analyze traditional Chinese medicine data and traditional Chinese medicine data, so that while analyzing diseases or symptoms, new drug research and development or drug compatibility schemes can be synchronously matched, which will not be elaborated here.

[0060] S200. Preprocess the collected traditional Chinese medicine data;

[0061] Preprocess the collected traditional Chinese medicine data to ensure data quality for subsequent analysis. A = {A1, A2,..., A i}, perform data preprocessing through the following formula. In the preprocessing, perform logarithmic transformation on the data, which can effectively handle the long-tailed distribution in the data and reduce the influence of extreme values. Introduce a non-linear transformation, which can better capture the complex relationships in the data. Process the outliers in the data through an exponential decay term to reduce the influence of outliers on subsequent analysis. Through preprocessing, the analysis results can be made more reliable, reducing the noise and interference of traditional Chinese medicine data. The formula for the above preprocessing is:

[0062]

[0063] Among them, B represents the preprocessed traditional Chinese medicine data, A represents the traditional Chinese medicine data before preprocessing, m represents the number of traditional Chinese medicine data, A i is the i-th traditional Chinese medicine data before preprocessing, μ is the mean of the traditional Chinese medicine data A before preprocessing, σ is the standard deviation of the traditional Chinese medicine data A before preprocessing, A max represents the maximum value of the traditional Chinese medicine data A before preprocessing, A min represents the minimum value of the traditional Chinese medicine data A before preprocessing, and δ represents the chi-square distribution density function.

[0064] S300. Construct a graph recurrent network, and the expression of the graph recurrent network is:

[0065] Construct a graph recurrent network, and the expression of the graph recurrent network is:

[0066]

[0067] Among them, M represents the graph recurrent network, C (1) represents the graph convolution factor of the first-level graph convolution layer, B i represents the i-th preprocessed traditional Chinese medicine data, γ represents the model hyperparameter factor, Z (L) represents the output feature of the L-th feedback mechanism layer, V (L) represents the output feature of the L-th high-level graph convolution layer, G (L) represents the dilation convolution factor of the L-th high-level graph convolution layer, Bmax Represents the maximum value of the preprocessed traditional Chinese medicine data, B min Represents the minimum value of the preprocessed traditional Chinese medicine data, V (L-1) Represents the output features of the (L - 1)-th high-level graph convolutional layer, μ L-1 Represents the output features V of the (L - 1)-th high-level graph convolutional layer (L-1) The mean value of, σ L-1 Represents the output features V of the (L - 1)-th high-level graph convolutional layer (L-1) The standard deviation of, W j (L) Represents the j-th weight factor of the L-th high-level graph convolutional layer, W j (1) Represents the j-th weight factor of the first-level graph convolutional layer;

[0068] The graph recurrent network consists of a first-level graph convolutional layer combined with multiple high-level graph convolutional layers and multiple feedback mechanism layers. Among them, the first-level graph convolutional layer focuses on local feature extraction, enhancing the model's perception ability of local structures. Multiple high-level graph convolutional layers integrate global information, capturing detailed information around nodes through feature aggregation of adjacent nodes, enhancing the model's ability to understand complex interrelationships. Through multiple feedback mechanism layers' adaptive adjustment mechanism, it monitors the changes in data flow and graph structure in real time, adjusts network parameters to reduce the risk of overfitting, and ensures the generalization ability of the model.

[0069] The expression of the first-level graph convolutional layer is as follows:

[0070]

[0071] Among them, V (1) Represents the output features of the first-level graph convolutional layer, C (1) Represents the graph convolution factor of the first-level graph convolutional layer, n represents the feature scaling factor, W j (1) Represents the j-th weight factor of the first-level graph convolutional layer, D represents the graph convolution degree factor, μ1 represents the mean value of the preprocessed traditional Chinese medicine data B, σ1 represents the standard deviation of the preprocessed traditional Chinese medicine data B, N i Represents the i-th preprocessed traditional Chinese medicine data, B i-1 Represents the (i - 1)-th preprocessed traditional Chinese medicine data.

[0072] The first-level graph convolutional layer performs local feature extraction based on the preprocessed traditional Chinese medicine data. The input of the first-level graph convolutional layer is the preprocessed traditional Chinese medicine data, including text data or image data such as the names of medicinal materials, traits, origins, chemical components, and pharmacological effects.

[0073] The first-level graph convolutional layer mainly focuses on local feature extraction, capturing detailed information around nodes by aggregating the features of adjacent nodes. This layer is designed to process the local structure in graph data to enhance the understanding and representation of information. For example, in drug research and development, the first-level graph convolutional layer can collect local information of traditional Chinese medicines from a traditional Chinese medicine database. For instance, it can extract the common features (such as pharmacological effects, properties, corresponding diseases, etc.) of specific Chinese medicines (nodes) and their adjacent Chinese medicines and avoid the possible effects of drug incompatibility. In this way, the model can identify combinations of Chinese medicines with similar effects.

[0074] The expression of the L-th high-level graph convolutional layer among multiple high-level graph convolutional layers is:

[0075]

[0076] Among them, V (L) represents the output feature of the L-th high-level graph convolutional layer, G (L) represents the dilation convolution factor of the L-th high-level graph convolutional layer, C (1) represents the graph convolution factor of the first-level graph convolutional layer, D represents the graph convolution degree factor, V (L-1) represents the output feature of the (L - 1)-th high-level graph convolutional layer, B i represents the i-th preprocessed traditional Chinese medicine data, W j (L) represents the j-th weight factor of the L-th high-level graph convolutional layer, μ L-1 represents the mean of the output feature V (L-1) of the (L - 1)-th high-level graph convolutional layer, σ L-1 represents the standard deviation of the output feature V (L-1) of the (L - 1)-th high-level graph convolutional layer.

[0077] The high-level graph convolutional layer integrates global information based on the features of the hidden layer convolution of the preprocessed traditional Chinese medicine data. The input of the high-level graph convolution is the features of the hidden layer convolution of the preprocessed traditional Chinese medicine data.

[0078] The high-level graph convolutional layer integrates global information, summarizes the features of multi-layer nodes, and enhances the model's ability to understand complex interrelationships. This layer can handle deeper relationships, considering not only directly adjacent nodes but also the multi-level structure in the entire graph. For example, in the process of drug research and development, the interactions between drugs are crucial. The high-level graph convolutional layer can integrate the complex relationships between drugs, analyze the effects of different drug combinations, avoid drug incompatibilities, and ensure safety and effectiveness. For example, for a specific disease, the model can comprehensively consider the comprehensive treatment effects of multiple drugs alone or in combination on the disease.

[0079] The expression of the L-th feedback mechanism layer among multiple feedback mechanism layers is:

[0080]

[0081] Among them, Z (L) represents the output feature of the L-th feedback mechanism layer, and θ1, θ2 represent feature adjustment factors, satisfying 0 < θ1 + θ2 < 1. V (1) represents the output feature of the first-level graph convolutional layer, V (L) represents the output feature of the L-th high-level graph convolutional layer, V (L-1) represents the output feature of the (L - 1)-th high-level graph convolutional layer, μ L-1 represents the mean value of the output feature V of the (L - 1)-th high-level graph convolutional layer (L-1) σ L-1 represents the standard deviation of the output feature V of the (L - 1)-th high-level graph convolutional layer (L-1) of the output feature V of the (L - 1)-th high-level graph convolutional layer

[0082] The feedback mechanism layer monitors the data flow and graph structure changes in real time based on the adaptive adjustment mechanism. The feedback mechanism layer performs weighted combination on the outputs (local features and global features) of different levels through the feature adjustment factors θ1 and θ2. If the drug screening results can be repeated and stable within a certain range, then the output feature Z (L) will be more consistent. This means that the model has a stronger ability to capture the features of effective drugs. The exponential decay term and the fourth-power exponent of the feature value included in the feedback mechanism can suppress the outputs with large deviations. When an extreme value or instability appears in a certain screening result, the feedback mechanism will suppress these outliers, thereby reducing their impact on the overall model output. By monitoring the stability and consistency of the output of the feedback layer, it can be judged whether the model is accurately identifying and screening relevant drugs. When the output feature Z (L) of the feedback mechanism layer reflects a high degree of uncertainty or inconsistency, real-time parameter adjustment can be performed. For example, by adjusting the weights W j (1) and W j (L) , the model can dynamically adapt to the complex relationship between different drugs and diseases, thereby improving the accuracy of drug screening. If the stability of the screening results is improved, it can be considered that the model is more accurate after adjustment.

[0083] The graph recurrent network utilizes multi-level graph convolutional layers and can extract multi-dimensional features from traditional Chinese medicine data to capture information at different levels. The first-level layer provides a detailed analysis of specific drugs (in another embodiment, it can also provide an analysis of cases), while the high-level layer ensures that the model can understand the complex relationships between nodes in the overall network, providing support and guarantee for the research and development of new drugs or drug compatibility. The feedback mechanism layer adaptively adjusts network parameters by monitoring the relationship between the output features and input features of the model in real time, reducing the risk of overfitting and enhancing the generalization ability of the model. Thus, during the process of traditional Chinese medicine research and development, it can effectively screen and match drugs, promoting the development of personalized medicine. It can also provide services for the classification of traditional Chinese medicine data.

[0084] The advantages of the graph recurrent network are also reflected in its effective integration ability of domain knowledge. Traditional Chinese medicine contains a large amount of professional knowledge and experience. By constructing a knowledge graph and combining it with the graph recurrent network, it is possible to better understand and analyze traditional Chinese medicine data. This combination can not only improve the performance of the model but also provide more accurate data support for clinical decision-making in traditional Chinese medicine, promoting the development of personalized medicine. In short, with the continuous progress of information technology, especially the application of advanced technologies such as graph recurrent networks, the modernization and intelligentization of traditional Chinese medicine are gradually being realized. This process not only injects new impetus into the inheritance and development of traditional Chinese medicine but also provides more accurate and personalized medical services for patients, marking the transformation and innovation of traditional Chinese medicine in the new era background.

[0085] S400. Input the preprocessed traditional Chinese medicine data into the graph recurrent network to obtain the classification result of traditional Chinese medicine data;

[0086] The classification result of traditional Chinese medicine data by the graph recurrent network can be the classification score, thereby realizing the classification and screening of Chinese medicinal materials, providing rapid screening and matching of drugs for the traditional Chinese medicine research and development process. Through the classification process of traditional Chinese medicine data by analyzing traditional Chinese medicine data, through automated classification processing, traditional Chinese medicine drugs can be classified quickly in a large amount of traditional Chinese medicine data, improving the efficiency of new drug research and development or drug compatibility combination. The classified traditional Chinese medicine data can be used as part of a clinical decision support system.

[0087] By integrating domain knowledge into the graph structure construction process, the graph recurrent network can better capture the features in traditional Chinese medicine theory and improve the ability to understand and interpret traditional Chinese medicine data. The graph recurrent network utilizes multi-level graph convolutional layers and can extract multi-dimensional features from traditional Chinese medicine data to capture information at different levels. The convolutional layer can focus on local feature extraction, while the high-level convolutional layer can integrate global information, thereby enhancing the classification performance. It can extract multi-dimensional features from traditional Chinese medicine data, comprehensively capture local and global information, and greatly improve the classification performance.

[0088] By introducing a dynamic learning mechanism, the graph recurrent network can adapt to data flow and structural changes in real time, reduce the risk of overfitting, and improve the generalization ability of the model to cope with the ever-changing traditional Chinese medicine data environment.

[0089] Please refer to Figure 2 , the traditional Chinese medicine data classification and processing system based on the graph recurrent network includes:

[0090] Data acquisition unit 1, the data acquisition unit 1 is configured to acquire traditional Chinese medicine data, and the traditional Chinese medicine data is text data or image data including one or more of the name of traditional Chinese medicine materials, properties, production areas, chemical components, and pharmacological effects;

[0091] Preprocessing unit 2, the preprocessing unit 2 is configured to preprocess the acquired traditional Chinese medicine data;

[0092] Graph recurrent network unit 3, the graph recurrent network unit 3 is configured to provide and update the graph recurrent network:

[0093]

[0094] Among them, M represents the improved graph recurrent network, C (1) represents the graph convolution factor of the first-level graph convolution layer, B i represents the i-th traditional Chinese medicine data after preprocessing, γ represents the model hyperparameter factor, Z (L) represents the output feature of the L-th feedback mechanism layer, V (L) represents the output feature of the L-th advanced graph convolution layer, G (L) represents the dilation convolution factor of the L-th advanced graph convolution layer, B max represents the maximum value of the traditional Chinese medicine data after preprocessing, B min represents the minimum value of the traditional Chinese medicine data after preprocessing, V (L-1) represents the output feature of the (L - 1)-th advanced graph convolution layer, μ L-1 represents the mean value of the output feature V (L-1) of the (L - 1)-th advanced graph convolution layer, σ L-1 represents the output feature V (L-1) of the (L - 1)-th advanced graph convolution layer, W j (L) represents the j-th weight factor of the L-th advanced graph convolution layer, W j (1) represents the j-th weight factor of the first-level graph convolution layer;

[0095] Classification unit 4, the classification unit 4 obtains the classification result of the traditional Chinese medicine data based on the traditional Chinese medicine data input into the graph recurrent network.

[0096] During the construction of the graph recurrent network, the professional knowledge of traditional Chinese medicine is integrated into the node representation design to enhance the adaptability of the model to traditional Chinese medicine theory and practical applications. Combining with the multi-level graph convolutional layer structure can effectively extract local and global features and improve the model's understanding and processing ability of traditional Chinese medicine data.

[0097] The preprocessing unit 2 preprocesses the traditional Chinese medicine data using the following formula:

[0098]

[0099] where B represents the preprocessed traditional Chinese medicine data, A represents the traditional Chinese medicine data before preprocessing, m represents the number of traditional Chinese medicine data, A i is the i-th traditional Chinese medicine data before preprocessing, μ is the mean of the traditional Chinese medicine data A before preprocessing, σ is the standard deviation of the traditional Chinese medicine data A before preprocessing, A max represents the maximum value of the traditional Chinese medicine data A before preprocessing, A min represents the minimum value of the traditional Chinese medicine data A before preprocessing, and δ represents the chi-square distribution density function.

[0100] The graph recurrent network includes a first-level graph convolutional layer, multiple high-level graph convolutional layers, and multiple feedback mechanism layers.

[0101] The expression of the first-level graph convolutional layer is:

[0102]

[0103] where V (1) represents the output feature of the first-level graph convolutional layer, C (1) represents the graph convolution factor of the first-level graph convolutional layer, n represents the feature scaling factor, W j (1) represents the j-th weight factor of the first-level graph convolutional layer, D represents the graph convolution degree factor, μ1 represents the mean of the preprocessed traditional Chinese medicine data B, σ1 represents the standard deviation of the preprocessed traditional Chinese medicine data B, B i represents the i-th preprocessed traditional Chinese medicine data, B i-1 represents the (i - 1)-th preprocessed traditional Chinese medicine data.

[0104] The expression of the L-th high-level graph convolutional layer in multiple high-level graph convolutional layers is:

[0105]

[0106] where V (L) represents the output feature of the L-th high-level graph convolutional layer, G (L) represents the dilated convolution factor of the L-th high-level graph convolutional layer, C (1) represents the graph convolution factor of the first-level graph convolutional layer, D represents the graph convolution degree factor, V (L-1)Denote the output features of the (L - 1)-th advanced graph convolutional layer, B i Denote the i-th preprocessed traditional Chinese medicine data, W k (L) Denote the j-th weight factor of the L-th advanced graph convolutional layer, μ L-1 Denote the output features V of the (L - 1)-th advanced graph convolutional layer (L-1) The mean value of, σ L-1 Denote the output features V of the (L - 1)-th advanced graph convolutional layer (L-1) The standard deviation of.

[0107] The expression of the L-th feedback mechanism layer in multiple feedback mechanism layers is:

[0108]

[0109] Among them, Z (L) Denote the output features of the L-th feedback mechanism layer, θ1, θ2 denote feature adjustment factors, satisfying 0 < θ1 + θ2 < 1, V (1) Denote the output features of the first-level graph convolutional layer, V (L) Denote the output features of the L-th advanced graph convolutional layer, V (L-1) Denote the output features of the (L - 1)-th advanced graph convolutional layer, μ L-1 Denote the output features V of the (L - 1)-th advanced graph convolutional layer (L-1) The mean value of, σ L-1 Denote the output features V of the (L - 1)-th advanced graph convolutional layer (L-1) The standard deviation of.

[0110] Among them, the first-level graph convolutional layer focuses on local feature extraction, enhancing the model's perception ability of local structures. Multiple advanced graph convolutional layers integrate global information, capturing detailed information around nodes through feature aggregation of adjacent nodes. Enhance the model's ability to understand complex interrelationships. Through the adaptive adjustment mechanism of multiple feedback mechanism layers, monitor the changes in data flow and graph structure in real time, and adjust network parameters to reduce the risk of overfitting and ensure the generalization ability of the model.

[0111] The graph recurrent network aggregates multi-dimensional data features through the design of multi-layer graph convolution, improving the accuracy and efficiency of data classification processing, especially suitable for complex traditional Chinese medicine data. Introduce an adaptive feedback adjustment mechanism, enabling the model to adjust parameters in real time according to data flow and structural changes, thereby reducing the risk of overfitting and improving the generalization ability of the model.

[0112] In summary, at least the following technical effects can be achieved through this application:

[0113] 1: Multi-dimensional Feature Extraction: The ability to extract multi-dimensional features through graph recurrent networks is very important. This process can capture the complexity and diversity of traditional Chinese medicine data and overcome the limitations of traditional classification methods;

[0114] 2: Fusion of Local and Global Features: The design of local feature extraction and global information integration enables the model to understand the data more comprehensively. The combination of such local and global features can indeed improve the accuracy and robustness of the model in classification tasks;

[0115] 3: Feedback Mechanism Layer: Introducing multiple feedback mechanism layers to adjust network parameters in real time, reducing the risk of overfitting and ensuring the generalization ability of the model, especially when dealing with the dynamically changing traditional Chinese medicine data environment;

[0116] 4: Fast Real-time Information Acquisition: The ability to quickly and real-time obtain effective information from a large amount of traditional Chinese medicine data can help researchers deeply analyze traditional Chinese medicine theories, drug compatibility, and disease treatment plans;

[0117] 5: Application Value in Drug R & D: Emphasize the important application value of this method in drug R & D, which can quickly classify traditional Chinese medicinal materials and reduce the selection time, improving the speed of new drug R & D and compatibility;

[0118] 6: In-depth Understanding of Drug Mechanisms: Provide an in-depth understanding of the action mechanisms and potential side effects of existing drugs, promoting the ability of drug discovery and optimized use, which reflects the practicality of this method.

[0119] It should be noted that this application can also be equivalently applied to the classification and screening of raw materials in the process of Western medicine R & D, which will not be elaborated here.

[0120] The above has introduced in detail a method for classifying and processing traditional Chinese medicine data based on graph recurrent networks. Specific examples are used in this article to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A method for classifying and processing TCM data based on graph recurrent networks, including: Collecting TCM data, wherein the TCM data is text data or image data including one or more of the name, properties, origin, chemical composition and pharmacological effects of medicinal materials; Preprocess the collected TCM data; Construct a graph recurrent network, the expression of which is: Among them, M represents the graph recurrent network, C (1) represents the graph convolution factor of the first-level graph convolution layer, B i represents the i-th TCM data after preprocessing, γ represents the model hyperparameter factor, Z (L) represents the output feature of the Lth feedback mechanism layer, V (L) represents the output features of the Lth high-level graph convolutional layer, G (L) represents the dilated convolution factor of the Lth high-level graph convolutional layer, B max represents the maximum value of the preprocessed TCM data, B min Represents the minimum value of the preprocessed TCM data, V (L-1) represents the output features of the L-1th high-level graph convolutional layer, μ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The mean value, σ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The standard deviation of j (L) represents the jth weight factor of the Lth high-level graph convolutional layer, W j (1) Represents the jth weight factor of the first-level graph convolutional layer; The preprocessed TCM data is input into the graph recurrent network to obtain the TCM data classification results.

2. According to the method for classifying and processing TCM data based on graph recurrent network in claim 1, In preprocessing the collected TCM data, the following formula is used to preprocess the TCM data: Among them, B represents the TCM data after preprocessing, A represents the TCM data before preprocessing, m represents the number of TCM data, and A i is the ith TCM data before preprocessing, μ is the mean of TCM data A before preprocessing, σ is the standard deviation of TCM data A before preprocessing, A max represents the maximum value of TCM data A before preprocessing, A min represents the minimum value of the TCM data A before preprocessing, and δ represents the chi-square distribution density function.

3. According to the method for classifying and processing TCM data based on graph recurrent network in claim 1, The graph recurrent network includes a primary graph convolution layer, multiple advanced graph convolution layers, and multiple feedback mechanism layers. The first-level graph convolution layer extracts local features based on the preprocessed TCM data. The multiple advanced graph convolutional layers integrate global information based on the features of the hidden layer convolution of the preprocessed TCM data. The multiple feedback mechanism layers monitor data flow and graph structure changes in real time based on an adaptive adjustment mechanism.

4. According to claim 3, the TCM data classification and processing method based on graph recurrent network, The expression of the first-level graph convolution layer is: Among them, V (1) represents the output features of the first-level graph convolutional layer, C (1) represents the graph convolution factor of the first-level graph convolution layer, n represents the feature scaling factor, and W j (1) represents the jth weight factor of the first-level graph convolution layer, D represents the graph convolution degree factor, μ1 represents the mean of the preprocessed TCM data B, σ1 represents the standard deviation of the preprocessed TCM data B, B i represents the ith TCM data after preprocessing, B i-1 Represents the i-1th TCM data after preprocessing.

5. According to the method for classifying and processing TCM data based on graph recurrent network in claim 3, The expression of the Lth high-level graph convolution layer among the multiple high-level graph convolution layers is: Among them, V (L) represents the output features of the Lth high-level graph convolutional layer, G (L) represents the dilated convolution factor of the Lth high-level graph convolutional layer, C (1) represents the graph convolution factor of the first-level graph convolution layer, D represents the graph convolution degree factor, V (L-1) represents the output features of the L-1th high-level graph convolutional layer, B i represents the ith TCM data after preprocessing, W j (L) represents the jth weight factor of the Lth high-level graph convolutional layer, μ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The mean value, σ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The standard deviation of .

6. The TCM data classification and processing method based on graph recurrent network according to claim 3, wherein: The expression of the Lth feedback mechanism layer among the multiple feedback mechanism layers is: Among them, Z (L) represents the output feature of the Lth feedback mechanism layer, θ1, θ2 represent feature adjustment factors, satisfying 0<θ1+θ2<1, V (1) Represents the output features of the first-level graph convolutional layer, V (L) represents the output features of the Lth high-level graph convolutional layer, V (L-1) represents the output features of the L-1th high-level graph convolutional layer, μ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The mean value, σ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The standard deviation of .

7. A TCM data classification and processing system based on graph recurrent networks, including: A data collection unit, wherein the data collection unit is configured to collect TCM data, wherein the TCM data is text data or image data including the name, properties, origin, chemical composition, and pharmacological effects of medicinal materials; A preprocessing unit, wherein the preprocessing unit is configured to preprocess the collected traditional Chinese medicine data; A graph recurrent network unit, wherein the graph recurrent network unit is configured to provide and update a graph recurrent network, wherein the expression of the graph recurrent network is: Among them, M represents the graph recurrent network, C (1) represents the graph convolution factor of the first-level graph convolution layer, B i represents the i-th TCM data after preprocessing, γ represents the model hyperparameter factor, Z (L) represents the output feature of the Lth feedback mechanism layer, V (L) represents the output features of the Lth high-level graph convolutional layer, G (L) represents the dilated convolution factor of the Lth high-level graph convolutional layer, B max represents the maximum value of the preprocessed TCM data, B min Represents the minimum value of the preprocessed TCM data, V (L-1) represents the output features of the L-1th high-level graph convolutional layer, μ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The mean value, σ L-1 Represents the output feature V of the L-1th high-level graph convolutional layer (L-1) The standard deviation of j (L) represents the jth weight factor of the Lth high-level graph convolutional layer, W j (1) Represents the jth weight factor of the first-level graph convolutional layer; A classification unit, wherein the classification unit obtains a classification result of the TCM data based on the TCM data input into the graph recurrent network.

8. The TCM data classification and processing system based on graph recurrent network according to claim 7, wherein: The preprocessing unit preprocesses the TCM data using the following formula: Among them, B represents the TCM data after preprocessing, A represents the TCM data before preprocessing, m represents the number of TCM data, and A i is the ith TCM data before preprocessing, μ is the mean of TCM data A before preprocessing, σ is the standard deviation of TCM data A before preprocessing, A max represents the maximum value of TCM data A before preprocessing, A min represents the minimum value of the TCM data A before preprocessing, and δ represents the chi-square distribution density function.

9. A medium having instructions stored thereon, wherein the instructions are used to enable a machine to execute the method for classifying and processing traditional Chinese medicine data based on a graph recurrent network according to any one of claims 1 to 6.

Images