Authority element recommendation method based on adaptive association, medium and equipment

Abstract

The invention discloses a syndrome element recommendation method based on adaptive association, a medium and equipment, and the method comprises the steps: building a learnable initialization feature vector, and enabling the features of the initialization feature vector to comprise a symptom feature vector and a syndrome element item feature vector; performing information aggregation on the symptom feature vector and the syndrome element item feature vector by using a graph convolutional neural network to obtain an aggregated feature vector; calculating the aggregated feature vector to obtain a recommendation relation matrix; taking the recommendation relation matrix as input, optimizing relevant parameters of the prediction model in the feature learning stage, and obtaining a trained prediction model; the symptom features and the syndrome element features which need to be predicted are input into the trained prediction model, a first result is output, and the first result comprises the relevance between the symptom features and the syndrome element features. According to the scheme, the initial features of the symptoms and the syndrome element items are optimized by using the graph convolutional neural network, personalized information can be provided for different symptoms, and then a more distinctive mode is generated.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a certificate element recommendation method, medium and device based on adaptive association. Background Art

[0002] TCM recommendation is an intelligent system based on TCM theory, modern data technology and artificial intelligence. It is a personalized solution designed for target users. This system can recommend reasonable and accurate syndrome elements to users based on the actual symptoms provided by users. With the research and development of deep learning, intelligent recommendation systems have achieved remarkable success, which has laid the foundation for TCM syndrome element recommendation systems. The core of the syndrome element recommendation system is to predict whether users will adopt syndrome element recommendations combined with their own symptoms.

[0003] With the development of graph neural networks, recommendation systems are gradually using graph structure data to improve the accuracy and personalization of recommendations. By modeling the complex relationship between users and items, graph neural networks can effectively capture user preferences and the mutual influence between items. The combination of graph neural networks and recommendation systems enables the syndrome recommendation system to learn the preference pattern of symptoms from the graph structure, thereby providing more accurate and personalized recommendations. In addition, improving the feedback efficiency of the recommendation system has opened up new research perspectives, which can deeply explore complex relationship networks and achieve more accurate and efficient recommendations.

[0004] Although existing recommendation algorithms based on graph neural networks have shown promising results, the design of these methods is quite cumbersome and burdensome. Many operations directly follow the design of graph convolutional neural networks, lack a reasonable basis, and are not suitable for TCM prescription scenarios. Therefore, it is necessary to redesign the model in combination with TCM scenarios. Moreover, most existing methods use relations as input and perform multi-layer nonlinear feature transformations. This will not only not bring any benefits, but will increase the difficulty of model training. In addition, existing recommendation systems are inefficient when the data scale is large. Summary of the invention

[0005] In view of the above problems, the present application provides a technical solution for syndrome factor recommendation based on adaptive association, which is used to solve the problems that the existing recommendation algorithm based on graph neural network is not suitable for TCM prescription scenarios, the model training is difficult, and the efficiency is low when the data scale is large.

[0006] To achieve the above objectives, in a first aspect, the present application provides a certificate element recommendation method based on adaptive association, the method comprising:

[0007] S1: establishing a learnable initialization feature vector, wherein the initialization vector features include a symptom feature vector and a syndrome item feature vector;

[0008] S2: Use graph convolutional neural network to aggregate information of symptom feature vectors and syndrome feature vectors to obtain aggregated feature vectors;

[0009] S3: Calculate the aggregated feature vector to obtain a recommendation relationship matrix;

[0010] S4: taking the recommendation relationship matrix as input, optimizing the relevant parameters of the prediction model in the feature learning stage, and obtaining a trained prediction model;

[0011] S5: Input the symptom characteristics and syndrome characteristics to be predicted into the trained prediction model, and output a first result, which includes the correlation between the symptom characteristics and the syndrome characteristics.

[0012] Further, step S2 includes:

[0013] Define the relationship network between symptom feature vectors as the first relationship network, the relationship network between syndrome element feature vectors and syndrome element feature vectors as the second relationship network, and the relationship network between symptom feature vectors and syndrome element feature vectors as the third relationship network;

[0014] Integrate the first relationship network, the second relationship network, and the third relationship network into the same relationship matrix to obtain a relationship matrix between symptom feature vectors and syndrome item feature vectors;

[0015] The symptom feature vector and the syndrome item feature vector are vertically spliced ​​to obtain the spliced ​​feature matrix;

[0016] A graph convolutional neural network is used to aggregate information of the concatenated feature matrix through a relationship matrix between the symptom feature vector and the syndrome item feature vector to obtain an aggregated feature vector.

[0017] Furthermore, a graph convolutional neural network is used to aggregate information of the concatenated feature matrix through a relationship matrix between the symptom feature vector and the syndrome item feature vector, and the obtained aggregated feature vector includes:

[0018] A graph convolutional neural network is used to encode the symptom feature vector and the syndrome item feature vector, and the embedded representations of the encoded symptom feature vector and syndrome item feature vector are obtained at different network layers. The embedded representations obtained at each network layer are aggregated to obtain an aggregated feature vector.

[0019] Furthermore, the first relationship network, the second relationship network, and the third relationship network are integrated into the same relationship matrix to obtain a relationship matrix between symptom feature vectors and syndrome element feature vectors, including:

[0020] The relationship matrix between the symptom feature vector and the syndrome item feature vector is obtained in the following way:

[0021]

[0022] Among them, U is the first relationship network, I is the second relationship network, R represents the third relationship network, and A represents the relationship matrix between the symptom feature vector and the syndrome item feature vector.

[0023] Furthermore, the symptom feature vector and the syndrome item feature vector are vertically spliced ​​to obtain a spliced ​​feature matrix including:

[0024] Define the symptom feature vector as x u , the eigenvector of the certificate term is x i ;

[0025] x u With x i Perform vertical splicing to obtain the splicing feature matrix X;

[0026] The graph convolutional neural network is used to aggregate information of the concatenated feature matrix through the relationship matrix between the symptom feature vector and the syndrome element feature vector, and the aggregated feature vector obtained includes:

[0027] E = GCN(A,X);

[0028] Among them, GCN represents a graph convolutional neural network, which is used to aggregate information on the concatenated feature matrix X through the relationship matrix A, and E represents the aggregated feature vector;

[0029] E contains the aggregated symptom features e u Characteristics of the syndrome element i , symptoms and features u Characteristics of the syndrome element i The relationship is as follows:

[0030]

[0031] Among them, k represents the number of network layers, N u Indicates the number of symptoms.

[0032] Furthermore, the method further comprises:

[0033] The obtained E is normalized by:

[0034]

[0035] Among them, N i Indicates the number of certificate items.

[0036] Furthermore, the symptom features after standardization of each network layer are obtained by weighted average and information aggregation is performed on the syndrome item features. The calculation formula is as follows:

[0037]

[0038] Among them, α k Represents the weight of each network layer feature.

[0039] Further, step S3 includes:

[0040] The recommended relationship matrix is ​​obtained by:

[0041]

[0042] in, Represents the predicted recommendation relationship matrix;

[0043] Taking the recommendation relationship matrix as input, optimizing the relevant parameters of the prediction model in the feature learning stage includes:

[0044] The relevant parameters of the prediction model in the feature learning stage are optimized by the following formula:

[0045]

[0046] Among them, σ represents the sigmoid activation function, λ represents the relevant parameters of the adjustable prediction model, and E (0) Represents the initially established learnable initialization feature vector.

[0047] In a second aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the certificate element recommendation method based on adaptive association as described in the first aspect of the present application.

[0048] In a third aspect, the present application provides an electronic device having a computer program stored thereon, comprising a processor and a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by the processor, the method for recommending certificate elements based on adaptive association as described in the first aspect of the present application is implemented.

[0049] Different from the prior art, the above technical solution involves a syndrome factor recommendation method, medium and device based on adaptive association, which includes: S1: establishing a learnable initialization feature vector, the initialization vector features include a symptom feature vector and a syndrome factor item feature vector; S2: using a graph convolutional neural network to aggregate information on the symptom feature vector and the syndrome factor item feature vector to obtain an aggregated feature vector; S3: calculating the aggregated feature vector to obtain a recommendation relationship matrix; S4: using the recommendation relationship matrix as input to optimize the relevant parameters of the prediction model in the feature learning stage to obtain a trained prediction model; S5: inputting the symptom features and syndrome factor features to be predicted into the trained prediction model, and outputting a first result, which includes the correlation between the symptom features and the syndrome factor features. The above solution uses a graph convolutional neural network to optimize the initial features of symptoms and syndrome factor items, so that it can provide personalized information for different symptoms, thereby generating more distinguishing patterns. This optimization allows the recommendation system to better identify the degree of correlation of recommendations.

[0050] The above-mentioned records related to the invention content are only an overview of the technical solution of the present application. In order to enable ordinary technicians in the field to more clearly understand the technical solution of the present application, and then implement it according to the text of the specification and the contents recorded in the drawings, and to make the above-mentioned purpose and other purposes, features and advantages of the present application easier to understand, the following is an explanation in combination with the specific implementation mode and drawings of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The drawings are only used to illustrate the principles, implementation methods, applications, characteristics and effects of the specific embodiments of the present invention and other related contents, and shall not be considered as limiting the present application.

[0052] In the drawings of the specification:

[0053] Figure 1 This is a flowchart of the method for recommending certificate elements based on adaptive association as described in the first exemplary embodiment of the present application;

[0054] Figure 2 This is a flowchart of the method for recommending certificate elements based on adaptive association as described in the second exemplary embodiment of the present application;

[0055] Figure 3 This is a flow chart of the method for recommending certificate elements based on adaptive association as described in the third exemplary embodiment of the present application;

[0056] Figure 4 This is a flow chart of the method for recommending certificate elements based on adaptive association as described in the fourth exemplary embodiment of the present application;

[0057] Figure 5 A schematic diagram of an electronic device according to an exemplary embodiment of the present application;

[0058] The reference numerals in the above drawings are described as follows:

[0059] 10. Electronic equipment;

[0060] 101. Processor;

[0061] 102. Storage medium. DETAILED DESCRIPTION

[0062] In order to explain in detail the possible application scenarios, technical principles, specific schemes that can be implemented, and the purposes and effects that can be achieved, the following is a detailed description of the specific embodiments listed in conjunction with the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0063] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The term "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, the various technical features mentioned in the embodiments can be combined in any way to form a corresponding implementable technical solution.

[0064] Unless otherwise defined, the technical terms used in this document have the same meanings as those generally understood by those skilled in the art to which this application belongs; the use of relevant terms in this document is only for describing specific embodiments and is not intended to limit this application.

[0065] In the description of this application, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" logical relationship.

[0066] In the present application, terms such as “first” and “second” are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship of quantity, priority or sequence between these entities or operations.

[0067] Without further limitations, in this application, the words "include", "comprises", "has" or other similar open-ended expressions used in sentences are intended to cover non-exclusive inclusion. These expressions do not exclude the presence of additional elements in the process, method or product including the elements, so that the process, method or product including a series of elements may include not only those limited elements, but also other elements not explicitly listed, or also include elements inherent to such process, method or product.

[0068] In this application, expressions such as "greater than", "less than", "exceed" and the like are understood to exclude the number itself; expressions such as "above", "below", "within" and the like are understood to include the number itself. In addition, in the description of the embodiments of this application, "multiple" means more than two (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups", "multiple times", etc., unless otherwise clearly and specifically limited.

[0069] In the description of the embodiments of the present application, space-related expressions used, such as "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "vertical", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicate the orientation or position relationship based on the orientation or position relationship shown in the specific embodiments or drawings, and are only for the convenience of describing the specific embodiments of the present application or facilitating the reader's understanding, and do not indicate or imply that the referred device or component must have a specific position, a specific orientation, or be constructed or operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present application.

[0070] Unless otherwise expressly specified or limited, in the description of the embodiments of the present application, the terms such as "install", "connect", "connect", "fix", "set", etc. used should be understood in a broad sense. For example, the "connection" can be a fixed connection, a detachable connection, or an integrated setting; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For technicians in the technical field to which the present application belongs, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0071] In a first aspect, the present application provides a certificate element recommendation method based on adaptive association, such as Figure 1 As shown, the method includes:

[0072] S1: establishing a learnable initialization feature vector, wherein the initialization vector features include a symptom feature vector and a syndrome item feature vector;

[0073] S2: Use graph convolutional neural network to aggregate information of symptom feature vectors and syndrome feature vectors to obtain aggregated feature vectors;

[0074] S3: Calculate the aggregated feature vector to obtain a recommendation relationship matrix;

[0075] S4: taking the recommendation relationship matrix as input, optimizing the relevant parameters of the prediction model in the feature learning stage, and obtaining a trained prediction model;

[0076] S5: Input the symptom characteristics and syndrome characteristics to be predicted into the trained prediction model, and output a first result, which includes the correlation between the symptom characteristics and the syndrome characteristics.

[0077] The above scheme uses graph convolutional neural networks to optimize the initial features of symptoms and syndromes, so that it can provide personalized information for different symptoms, thereby generating more discriminative patterns. This optimization allows the recommendation system to better identify the degree of relevance of recommendations.

[0078] In some embodiments, Figure 2 As shown, step S2 includes:

[0079] Step S201: define the relationship network between symptom feature vectors and symptom feature vectors as a first relationship network, the relationship network between syndrome element feature vectors and syndrome element feature vectors as a second relationship network, and the relationship network between symptom feature vectors and syndrome element feature vectors as a third relationship network;

[0080] Step S202: Integrate the first relationship network, the second relationship network, and the third relationship network into the same relationship matrix to obtain a relationship matrix between symptom feature vectors and syndrome item feature vectors;

[0081] Step S203: vertically concatenate the symptom feature vector and the syndrome item feature vector to obtain a concatenated feature matrix;

[0082] Step S204: Use a graph convolutional neural network to aggregate information on the concatenated feature matrix through a relationship matrix between the symptom feature vector and the syndrome item feature vector to obtain an aggregated feature vector.

[0083] In this embodiment, the first relationship network, the second relationship network, and the third relationship network are integrated into the same relationship matrix to obtain a relationship matrix between symptom feature vectors and syndrome element feature vectors, including:

[0084] The relationship matrix between the symptom feature vector and the syndrome item feature vector is obtained in the following way:

[0085]

[0086] Among them, U is the first relationship network, I is the second relationship network, R represents the third relationship network, and A represents the relationship matrix between the symptom feature vector and the syndrome item feature vector.

[0087] For example, given m symptoms and n syndrome items, the size of U is an m*m relationship matrix, and I is an n*n relationship matrix.

[0088] In some embodiments, the symptom feature vector and the syndrome item feature vector are vertically spliced ​​to obtain a spliced ​​feature matrix including:

[0089] Define the symptom feature vector as x u , the eigenvector of the certificate term is x i ;

[0090] x u With x i Perform vertical splicing to obtain the splicing feature matrix X;

[0091] The graph convolutional neural network is used to aggregate information of the concatenated feature matrix through the relationship matrix between the symptom feature vector and the syndrome element feature vector, and the aggregated feature vector obtained includes:

[0092] E = GCN(A,X);

[0093] Among them, GCN represents a graph convolutional neural network, which is used to aggregate information on the concatenated feature matrix X through the relationship matrix A, and E represents the aggregated feature vector;

[0094] E contains the aggregated symptom features e u Characteristics of the syndrome element i , symptoms and features u Characteristics of the syndrome element i The relationship is as follows:

[0095]

[0096] Among them, k represents the number of network layers. The output result of the kth layer is used as the input parameter of the k+1th layer to participate in the calculation of the next layer. N u Indicates the number of symptoms.

[0097] The size of A is (m+n)*(m+n), which means there are m+n node features in total. For each node, aggregating the information of its directly connected neighbors can obtain more comprehensive and richer context information, that is, each item in the concatenated feature matrix X tends to be consistent with the features between adjacent nodes. This way of fusing information can enhance local structural information and make X more discriminative. At the same time, moderate information smoothing (i.e., information tends to be consistent) will also make the model generalizable.

[0098] In some embodiments, the method further comprises:

[0099] The obtained E is normalized by:

[0100]

[0101] Among them, N i Indicates the number of certificate items.

[0102] Through standardization, the symptom characteristics can be u Characteristics of the syndrome element i The corresponding values ​​are within a specific range, which is convenient for further processing.

[0103] In some embodiments, the method further comprises:

[0104] The symptom features after standardization of each network layer are obtained by weighted average and information aggregation is performed on the syndrome item features. The calculation formula is as follows:

[0105]

[0106]

[0107] Among them, α k Represents the weight of each network layer feature.

[0108] In some embodiments, step S3 includes:

[0109] The recommended relationship matrix is ​​obtained by:

[0110]

[0111] in, Represents the predicted recommendation relationship matrix;

[0112] pass (The inner product of the characteristic matrix of symptoms and the characteristic matrix of syndrome elements) can get a relationship matrix between symptoms and syndrome elements. This relationship matrix represents the relationship coefficient between each predicted symptom and each syndrome factor. The larger the value, the more correlated the two items are.

[0113] Taking the recommendation relationship matrix as input, optimizing the relevant parameters of the prediction model in the feature learning stage includes:

[0114] The relevant parameters of the prediction model in the feature learning stage are optimized by the following formula:

[0115]

[0116] Among them, σ represents the sigmoid activation function, λ represents the relevant parameters of the adjustable prediction model, and E (0) Represents the initially established learnable initialization feature vector.

[0117] It represents the predicted relationship value between symptoms and syndromes in the prediction matrix that have no direct relationship in the training data. The relevant parameters of the prediction model in the feature learning stage are to adjust ||E (0) || 2 The larger the parameter value is, the greater the change will be during each optimization. (0) || 2 The 2 norm is used, that is, E (0) The sum of the squares of all elements in , after continuous optimization, the value of the overall loss function will continue to decrease, resulting in E (0) The sum of the squares of all elements in E becomes smaller. (0) That is, the symptom feature vector mentioned above is x u The eigenvector of the sum-of-certificate term is x i The initial value of .

[0118] like Figure 3 As shown, in some embodiments, a graph convolutional neural network is used to aggregate information of the concatenated feature matrix through a relationship matrix between the symptom feature vector and the syndrome item feature vector, and the aggregated feature vector obtained includes:

[0119] Step S301: Use a graph convolutional neural network to encode the symptom feature vector and the syndrome element feature vector, and obtain the respective embedded representations of the encoded symptom feature vector and syndrome element feature vector at different network layers:

[0120] Step S302: Aggregate information on the embedded representations obtained at each network layer to obtain an aggregated feature vector.

[0121] The training process of the present invention is divided into two stages: information aggregation and feature learning. In the information aggregation stage, the graph convolutional neural network encodes the features of the learnable symptoms and syndrome elements, and obtains their respective embedded representations at different network layers, and then performs information aggregation on the embedded representations to enable the model to obtain a more accurate embedded representation. In the feature learning stage, the embedded representation after information aggregation is used as input, and the embedded representations of the symptoms and syndrome elements are inner-producted to obtain the final relationship matrix. Finally, the final recommendation prediction result is obtained through continuous optimization of the objective function.

[0122] like Figure 4 As shown, in some embodiments, the method includes:

[0123] Step S401: establishing a learnable initialization feature vector, wherein the initialization vector features include a symptom feature vector and a syndrome item feature vector.

[0124] Step S402: construct a common relationship matrix between feature vectors, including a first relationship network between symptom feature vectors and symptom feature vectors, a second relationship network between syndrome element feature vectors and syndrome element feature vectors, and a third relationship network between symptom feature vectors and syndrome element feature vectors.

[0125] Step S403: Obtain the aggregated feature vectors of each layer through a multi-layer graph convolutional neural network.

[0126] Step S404: performing a standardization operation on the aggregated feature vector.

[0127] Step S405: Obtaining the predicted recommendation relationship matrix

[0128] Step S406: Determine whether the optimization of the relevant parameters of the prediction model is completed?

[0129] If the judgment result of step S406 is yes, then proceed to step S407: output the correlation between the symptom characteristics and the syndrome characteristics through the prediction model;

[0130] If the determination result of step S406 is negative, the process returns to step S401.

[0131] This application proposes a syndrome factor recommendation method based on adaptive association, which uses a lightweight graph convolutional neural network to aggregate feature information and finally obtains a predicted relationship matrix for recommendation. By initializing a learnable vector as the initial feature of the symptom and syndrome factor items, and continuously optimizing the constraints of the objective function to obtain accurate original information features, the recommendation system has the ability to identify association relationships.

[0132] In a second aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for recommending certificate elements based on adaptive association as described in the first aspect of the present invention is implemented.

[0133] The computer-readable storage medium may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories.

[0134] The non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a ferromagnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD ROM); the magnetic surface memory can be a disk memory or a tape memory.

[0135] The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), direct memory bus random access memory (DRRAM). The computer-readable storage medium described in the embodiments of the present invention is intended to include these and any other suitable types of memory.

[0136] like Figure 5 As shown, in a third aspect, the present invention provides an electronic device 10, comprising a processor 101 and a storage medium 102, wherein the storage medium stores a computer program, and when the computer program is executed by the processor, the method for recommending certificate elements based on adaptive association as described in the first aspect of the present invention is implemented.

[0137] In some embodiments, the processor can be implemented by software, hardware, firmware or a combination thereof, and can use circuits, single or multiple application specific integrated circuits (Application Specific Integrated Circuit, ASIC), digital signal processors (Digital Signal Processor, DSP), digital signal processing devices (Digital Signal Processing Device, DSPD), programmable logic devices (Programmable Logic Device, PLD), field programmable gate arrays (Field Programmable Gate Array, FPGA), central processing units (Central Processing Unit, CPU), controllers, microcontrollers, at least one of microprocessors, so that the processor can execute some or all of the steps or any combination of the steps in the adaptive association-based certificate recommendation method described in the various embodiments of the present application.

[0138] Finally, it should be noted that although the above embodiments have been described in the specification and drawings of this application, this does not limit the scope of patent protection of this application. All technical solutions generated by replacing or modifying equivalent structures or equivalent processes based on the essential concept of this application using the contents recorded in the specification and drawings of this application, as well as directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, are included in the scope of patent protection of this application.

Claims

1. A certificate factor recommendation method based on adaptive association, characterized in that: The method comprises: S1: establishing a learnable initialization feature vector, wherein the initialization vector features include a symptom feature vector and a syndrome item feature vector; S2: Use graph convolutional neural network to aggregate information of symptom feature vectors and syndrome feature vectors to obtain aggregated feature vectors; S3: Calculate the aggregated feature vector to obtain a recommendation relationship matrix; S4: taking the recommendation relationship matrix as input, optimizing the relevant parameters of the prediction model in the feature learning stage, and obtaining a trained prediction model; S5: Input the symptom characteristics and syndrome characteristics to be predicted into the trained prediction model, and output a first result, which includes the correlation between the symptom characteristics and the syndrome characteristics.

2. The certificate element recommendation method based on adaptive association as claimed in claim 1, characterized in that: Step S2 includes: Define the relationship network between symptom feature vectors as the first relationship network, the relationship network between syndrome element feature vectors and syndrome element feature vectors as the second relationship network, and the relationship network between symptom feature vectors and syndrome element feature vectors as the third relationship network; Integrate the first relationship network, the second relationship network, and the third relationship network into the same relationship matrix to obtain a relationship matrix between symptom feature vectors and syndrome item feature vectors; The symptom feature vector and the syndrome item feature vector are vertically spliced ​​to obtain the spliced ​​feature matrix; A graph convolutional neural network is used to aggregate information of the concatenated feature matrix through a relationship matrix between the symptom feature vector and the syndrome item feature vector to obtain an aggregated feature vector.

3. The method for recommending certificate elements based on adaptive association as claimed in claim 2, characterized in that: The graph convolutional neural network is used to aggregate information of the concatenated feature matrix through the relationship matrix between the symptom feature vector and the syndrome element feature vector, and the aggregated feature vector obtained includes: A graph convolutional neural network is used to encode the symptom feature vector and the syndrome item feature vector, and the embedded representations of the encoded symptom feature vector and syndrome item feature vector are obtained at different network layers. The embedded representations obtained at each network layer are aggregated to obtain an aggregated feature vector.

4. The method for recommending certificate elements based on adaptive association as claimed in claim 2, characterized in that: The first relationship network, the second relationship network, and the third relationship network are integrated into the same relationship matrix to obtain a relationship matrix between symptom feature vectors and syndrome element feature vectors, including: The relationship matrix between the symptom feature vector and the syndrome item feature vector is obtained in the following way: Among them, U is the first relationship network, I is the second relationship network, R represents the third relationship network, and A represents the relationship matrix between the symptom feature vector and the syndrome item feature vector.

5. The certificate element recommendation method based on adaptive association as claimed in claim 4, characterized in that: The symptom feature vector and the syndrome feature vector are vertically spliced ​​to obtain a spliced ​​feature matrix including: Define the symptom feature vector as x u , the eigenvector of the certificate term is x i ; x u With x i Perform vertical splicing to obtain the splicing feature matrix X; The graph convolutional neural network is used to aggregate information of the concatenated feature matrix through the relationship matrix between the symptom feature vector and the syndrome element feature vector, and the aggregated feature vector obtained includes: E = GCN(A,X); Among them, GCN represents a graph convolutional neural network, which is used to aggregate information on the concatenated feature matrix X through the relationship matrix A, and E represents the aggregated feature vector; E contains the aggregated symptom features e u Characteristics of the syndrome element i , symptoms and features u Characteristics of the syndrome element i The relationship is as follows: Among them, k represents the number of network layers, N u Indicates the number of symptoms.

6. The method for recommending certificate elements based on adaptive association according to claim 5, characterized in that: The method further comprises: The obtained E is normalized by: Among them, N i Indicates the number of certificate items.

7. The method for recommending certificate elements based on adaptive association according to claim 6, characterized in that: The method further comprises: The symptom features after standardization of each network layer are obtained by weighted average and information aggregation is performed on the syndrome item features. The calculation formula is as follows: Among them, α k Represents the weight of each network layer feature.

8. The certificate element recommendation method based on adaptive association as claimed in claim 7, characterized in that: Step S3 includes: The recommended relationship matrix is ​​obtained by: in, Represents the predicted recommendation relationship matrix; Taking the recommendation relationship matrix as input, optimizing the relevant parameters of the prediction model in the feature learning stage includes: The relevant parameters of the prediction model in the feature learning stage are optimized by the following formula: Among them, σ represents the sigmoid activation function, λ represents the relevant parameters of the adjustable prediction model, and E (0) Represents the initially established learnable initialization feature vector.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for recommending certificate elements based on adaptive association as described in any one of claims 1 to 8 is implemented.

10. An electronic device having a computer program stored thereon, characterized in that: It comprises a processor and a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by the processor, the certificate element recommendation method based on adaptive association as described in any one of claims 1 to 8 is implemented.

Images