Multilabel graph structure construction method based on conditional probability matrix and gcn

Abstract

The application provides a multi-label graph structure construction method based on a conditional probability matrix and a GCN, comprising: classifying label data in a multi-label data set, constructing a conditional probability matrix according to a co-occurrence frequency corresponding to each label combination, and constructing a corresponding mean-standard deviation dynamic threshold; the mean-standard deviation dynamic threshold is used for binary processing of the conditional probability matrix, a binary adjacency matrix is generated, binary information corresponding to each label combination is obtained, the binary adjacency matrix is input into a preset GCN for convolution processing, convolution information corresponding to each label combination is obtained, label association features corresponding to each label combination are generated based on the binary information and the convolution information corresponding to each label combination, and the multi-label graph structure of the multi-label data set is constructed by using the label association features, so that the problem of overfitting in the conventional technology is effectively solved, and the performance and quality of data detection are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of multi-label graph structure technology, and in particular to a method for constructing multi-label graph structures based on conditional probability matrices and GCNs. Background Technology

[0002] Image classification, as an important branch of computer vision, has been widely applied to applications such as target recognition and defect detection. Based on the number of labels in an image, there are two types of image classification: single-label image classification and multi-label image classification. With the development and widespread adoption of deep learning technology, the performance of single-label image classification methods has become sufficiently superior. Compared to single-label image classification, multi-label image classification meets more general needs. However, its implementation is more difficult and complex, making it more challenging in image processing tasks. Large-scale multi-label image classification requires determining the presence or absence of target objects in a large number of sample images. Due to the massive number of samples and varying levels of human awareness, this can lead to very low efficiency in multi-label image classification, especially for highly specialized and complex multi-label image sets. Therefore, ensuring the accuracy of image classification is particularly important.

[0003] However, existing multi-label image classification methods focus on the accuracy of label prediction while ignoring the structural information embedded in the hierarchical label space and lacking deep semantic associations, resulting in a decline in detection performance. Therefore, there is an urgent need for a method to solve the above problems.

[0004] Therefore, this invention provides a method for constructing multi-label graph structures based on conditional probability matrices and GCN. Summary of the Invention

[0005] This invention presents a multi-label graph structure construction method based on conditional probability matrices and GCN, which effectively solves the problem of overfitting that is prone to occur in traditional techniques, and also ensures the performance and quality of data detection.

[0006] This invention provides a method for constructing multi-label graph structures based on conditional probability matrices and GCNs, including:

[0007] Step 1: Classify the label data in the multi-label dataset to obtain the co-occurrence frequency corresponding to each label combination and construct the conditional probability matrix of the multi-label dataset;

[0008] Step 2: Calculate the mean and standard deviation of each of the labeled data, and construct the corresponding mean-standard deviation dynamic threshold;

[0009] Step 3: Use the mean-standard deviation dynamic threshold to binarize the conditional probability matrix to generate a binarized adjacency matrix and obtain the binary information corresponding to each of the label combinations;

[0010] Step 4: Input the binarized adjacency matrix into a preset GCN for convolution processing to obtain the convolution information corresponding to each label combination;

[0011] Step 5: Generate label association features corresponding to each label combination based on the binary information and convolution information corresponding to each label combination, and construct the multi-label graph structure of the multi-label dataset using the label association features.

[0012] In one feasible approach

[0013] Step 1 includes:

[0014] Step 11: Identify several label data contained in the multi-label dataset, perform binary classification training on each label data to obtain several labels contained in each label data, and identify the samples corresponding to the label data.

[0015] Step 12: Combine the labels according to the correspondence between samples and labels contained in the same label data to obtain several label combinations corresponding to each label data, and construct the combination conditions corresponding to each label combination according to the label features corresponding to each label.

[0016] Step 13: Perform a deep statistical search on the multi-label dataset using each of the aforementioned combination conditions to obtain the co-occurrence frequency corresponding to each of the aforementioned label combinations, construct corresponding label events based on the label combinations, and construct corresponding event probabilities based on the co-occurrence frequencies;

[0017] Step 14: Construct the conditional probability matrix of the multi-label dataset based on the labeled events and event probabilities.

[0018] In one feasible approach

[0019] Step 2 includes:

[0020] Step 21: Sample each of the label data to obtain several label values. Divide the label data into several data segments based on the sampling position corresponding to each label value. Filter the target data segments whose length is higher than the preset length and perform compensation sampling to obtain the corresponding label values.

[0021] Step 22: Calculate the mean and standard deviation of the corresponding tag data based on the tag values, determine the dynamic threshold range of the corresponding tag data using the mean and standard deviation, and establish the range ratio of the dynamic threshold range based on the ratio of the number of first tag values ​​falling within the dynamic threshold range to the number of second tag values ​​not falling within the dynamic threshold range;

[0022] Step 23: Using the range comparison, correct the corresponding mean and standard deviation to obtain several corrected mean and standard deviation, filter out several target corrected mean and target corrected standard deviation that fall within the corresponding dynamic threshold range, and generate the mean-standard deviation dynamic threshold corresponding to the label data.

[0023] In one feasible approach

[0024] Step 3 includes:

[0025] Step 31: Binarize the conditional probability matrix using each of the mean-standard deviation dynamic thresholds to obtain the assignment result of each of the mean-standard deviation dynamic thresholds to the conditional probability matrix;

[0026] Step 32: Calculate the assignment balance degree corresponding to each assignment result, and use the target assignment result with the highest assignment balance degree to perform binarization assignment on the conditional probability matrix to generate a binarized adjacency matrix.

[0027] Step 33: Identify the matrix element corresponding to each tag combination in the binary adjacency matrix, and determine the binary information corresponding to the tag combination based on the assigned value of the matrix element.

[0028] In one feasible approach

[0029] Step 4 includes:

[0030] Step 41: Obtain several neighboring nodes contained in the binarized adjacency matrix, input the binarized adjacency matrix into a preset GCN for convolution processing, and obtain the node relationships between different neighboring nodes;

[0031] Step 42: Match corresponding label combinations for each of the adjacent nodes, construct label combination relationships based on the node relationships, and generate external features for each of the label combinations;

[0032] Step 43: Capture the convolution features corresponding to each of the adjacent nodes in the convolution processing results, and combine them with the corresponding circumscribed features to establish the convolution information of the label combination.

[0033] In one feasible approach

[0034] Step 5 includes:

[0035] Step 51: Establish shadow distribution features corresponding to the label combination based on the binary information, and establish pixel distribution features corresponding to the label combination based on the convolution information;

[0036] Step 52: Obtain the presentation features of each of the label combinations in the binary neighborhood matrix, and construct the label-related features corresponding to each of the label combinations by combining the corresponding shadow distribution features and pixel distribution features;

[0037] Step 53: Draw the label image corresponding to the combined label according to the label-related features, fuse the label images according to the presentation features to generate the multi-label graph structure of the multi-label dataset, and display it.

[0038] In one feasible approach

[0039] Also includes:

[0040] Each tag name corresponding to each tag combination is marked in the multi-tag graph structure to obtain the graph domain corresponding to each tag combination and then displayed.

[0041] In one feasible approach

[0042] Also includes:

[0043] Identify the nodes and edges contained in the multi-label graph structure;

[0044] Identify several first labels corresponding to each node and several second labels corresponding to each edge;

[0045] Calculate the sparsity of the first label corresponding to each node and the sparsity of the second label corresponding to each edge to determine the key structural regions of the multi-label graph structure;

[0046] The key areas are highlighted in the multi-label graph structure.

[0047] In one feasible approach

[0048] Also includes:

[0049] The system retrieves relevant information contained in the multi-tab graph structure based on the execution command issued by the user, and transmits it to the corresponding terminal for display.

[0050] In one feasible approach

[0051] Also includes:

[0052] Based on the user-uploaded graph structure, the domain information contained in the multi-label graph structure is identified and displayed.

[0053] The beneficial effects of the above technical solution are as follows: To achieve image classification and structure determination, the multi-label dataset is first mechanically classified to determine the co-occurrence frequency of each label combination, thereby establishing a conditional probability matrix. Then, the mean and standard deviation of the label data are used to establish a mean-standard deviation dynamic threshold, which binarizes the conditional probability matrix and determines the binary information of each label combination. Next, a pre-defined GCN is used to convolve the binarized adjacency matrix, obtaining the convolution information of each label combination. Finally, the label association features of the label combinations are constructed based on the binary and convolution information of the label combinations. A multi-label graph structure is then constructed using the label association feature chain. This approach not only allows for the simultaneous analysis of multiple sets of label data but also significantly eliminates noise in the labels, improving the classification accuracy of the multi-label graph structure.

[0054] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0055] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0056] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0057] Figure 1 This is a schematic diagram illustrating the workflow of the multi-label graph structure construction method based on conditional probability matrix and GCN in an embodiment of the present invention.

[0058] Figure 2 This is a schematic diagram of the workflow of step 2 of the multi-label graph structure construction method based on conditional probability matrix and GCN in an embodiment of the present invention. Detailed Implementation

[0059] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0060] Example 1

[0061] This embodiment provides a method for constructing a multi-label graph structure based on conditional probability matrices and GCN, such as Figure 1 As shown, it includes:

[0062] Step 1: Classify the label data in the multi-label dataset to obtain the co-occurrence frequency corresponding to each label combination and construct the conditional probability matrix of the multi-label dataset;

[0063] Step 2: Calculate the mean and standard deviation of each of the labeled data, and construct the corresponding mean-standard deviation dynamic threshold;

[0064] Step 3: Use the mean-standard deviation dynamic threshold to binarize the conditional probability matrix to generate a binarized adjacency matrix and obtain the binary information corresponding to each of the label combinations;

[0065] Step 4: Input the binarized adjacency matrix into a preset GCN for convolution processing to obtain the convolution information corresponding to each label combination;

[0066] Step 5: Generate label association features corresponding to each label combination based on the binary information and convolution information corresponding to each label combination, and construct the multi-label graph structure of the multi-label dataset using the label association features.

[0067] In this example, the label combination represents the label group corresponding to data that contains two labels at the same time;

[0068] In this example, the mean-standard deviation dynamic threshold represents the result of determining the dynamic threshold of the label data based on the mean and standard deviation of the data.

[0069] In this example, binarization means adjusting the elements in the conditional probability matrix to contain only the two values ​​0 and 1.

[0070] In this example, the binary information represents the values ​​in the binary adjacency matrix associated with a combination of labels;

[0071] In this example, GCN is the default model used for deep learning;

[0072] In this example, the convolution information represents the result after convolving the label combination;

[0073] In this example, the label association feature represents the feature formed due to the label combination in a label combination.

[0074] The working principle and beneficial effects of the above technical solution are as follows: To achieve image classification and structure determination, the multi-label dataset is first mechanically classified to determine the co-occurrence frequency of each label combination, thereby establishing a conditional probability matrix. Then, the mean and standard deviation of the label data are used to establish a mean-standard deviation dynamic threshold, which binarizes the conditional probability matrix and determines the binary information of each label combination. Next, a pre-defined GCN is used to convolve the binarized adjacency matrix, obtaining the convolution information of each label combination. Finally, the label association features of the label combinations are constructed based on the binary and convolution information of the label combinations. A multi-label graph structure is then constructed using the label association feature chain. This approach not only allows for the simultaneous analysis of multiple sets of label data but also significantly eliminates noise in the labels, improving the classification accuracy of the multi-label graph structure.

[0075] Example 2

[0076] Based on Example 1, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN, step 1 includes:

[0077] Step 11: Identify several label data contained in the multi-label dataset, perform binary classification training on each label data to obtain several labels contained in each label data, and identify the samples corresponding to the label data.

[0078] Step 12: Combine the labels according to the correspondence between samples and labels contained in the same label data to obtain several label combinations corresponding to each label data, and construct the combination conditions corresponding to each label combination according to the label features corresponding to each label.

[0079] Step 13: Perform a deep statistical search on the multi-label dataset using each of the aforementioned combination conditions to obtain the co-occurrence frequency corresponding to each of the aforementioned label combinations, construct corresponding label events based on the label combinations, and construct corresponding event probabilities based on the co-occurrence frequencies;

[0080] Step 14: Construct the conditional probability matrix of the multi-label dataset based on the labeled events and event probabilities.

[0081] In this example, binary classification training represents the training process for recognizing two labels in the label data;

[0082] In this example, the sample represents the basic data of the label data, which consists of the features input by the user and the input labels;

[0083] In this example, the correspondence represents the relationship between samples and labels;

[0084] In this example, the combination condition represents the condition for generating a combination of tags.

[0085] The working principle and beneficial effects of the above technical solution are as follows: By performing binary classification training on the data labels in the multi-label dataset, the label and sample corresponding to each label data are determined. Based on the correspondence between the label and the sample, the combination conditions of the label combination are established, thereby calculating the co-occurrence frequency of each label combination, constructing the corresponding label events and determining the event probability corresponding to each label event. Finally, a conditional probability matrix is ​​constructed. The conditional probability matrix can simultaneously present the event probabilities corresponding to multiple label events, which is beneficial for subsequent analysis of the characteristics of the label data.

[0086] Example 3

[0087] Based on Example 1, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN, such as Figure 2 As shown, step 2 includes:

[0088] Step 21: Sample each of the label data to obtain several label values. Divide the label data into several data segments based on the sampling position corresponding to each label value. Filter the target data segments whose length is higher than the preset length and perform compensation sampling to obtain the corresponding label values.

[0089] Step 22: Calculate the mean and standard deviation of the corresponding tag data based on the tag values, determine the dynamic threshold range of the corresponding tag data using the mean and standard deviation, and establish the range ratio of the dynamic threshold range based on the ratio of the number of first tag values ​​falling within the dynamic threshold range to the number of second tag values ​​not falling within the dynamic threshold range;

[0090] Step 23: Using the range comparison, correct the corresponding mean and standard deviation to obtain several corrected mean and standard deviation, filter out several target corrected mean and target corrected standard deviation that fall within the corresponding dynamic threshold range, and generate the mean-standard deviation dynamic threshold corresponding to the label data.

[0091] In this example, the label value represents the value sampled from the label data;

[0092] In this example, the preset length is 5% of the tag data;

[0093] In this example, the dynamic threshold range represents the range within which the dynamic threshold of the label data may appear;

[0094] In this example, the quantity ratio represents the ratio between the first label value and the second label value;

[0095] In this example, the range ratio represents the range of values ​​used to adjust the dynamic threshold range within that range ratio.

[0096] The working principle and beneficial effects of the above technical solution are as follows: By sampling the label data, the corresponding label values ​​are obtained. The dynamic threshold range of the label data is determined based on the data mean and standard deviation. The data mean and standard deviation are corrected based on the proportion of label data falling outside the dynamic threshold range. The dynamic threshold of the label data mean-standard deviation is determined, which lays the foundation for subsequent binarization processing.

[0097] Example 4

[0098] Based on Example 1, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN, step 3 includes:

[0099] Step 31: Binarize the conditional probability matrix using each of the mean-standard deviation dynamic thresholds to obtain the assignment result of each of the mean-standard deviation dynamic thresholds to the conditional probability matrix;

[0100] Step 32: Calculate the assignment balance degree corresponding to each assignment result, and use the target assignment result with the highest assignment balance degree to perform binarization assignment on the conditional probability matrix to generate a binarized adjacency matrix.

[0101] Step 33: Identify the matrix element corresponding to each tag combination in the binary adjacency matrix, and determine the binary information corresponding to the tag combination based on the assigned value of the matrix element.

[0102] In this example, the assignment result represents the result of inputting the values ​​0 and 1 into the conditional probability matrix;

[0103] In this example, the assignment balance represents the balance ratio of 0 and 1 values ​​in each assignment result;

[0104] In this example, binarization assignment refers to the process of assigning 0 and 1 values ​​to the conditional probability matrix.

[0105] The working principle and beneficial effects of the above technical solution are as follows: The conditional probability matrix is ​​binarized using a mean-standard deviation dynamic threshold, which transforms the conditional probability matrix into a simple and efficient binary adjacency matrix. Then, the matrix elements corresponding to each label combination are identified in the binary adjacency matrix to construct the binary information of the label combination. In this way, the information of the label combination can be simplified, the binary information of the label combination can be determined, and the quality and efficiency of identifying the label combination information can be improved.

[0106] Example 5

[0107] Based on Example 1, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN, step 4 includes:

[0108] Step 41: Obtain several neighboring nodes contained in the binarized adjacency matrix, input the binarized adjacency matrix into a preset GCN for convolution processing, and obtain the node relationships between different neighboring nodes;

[0109] Step 42: Match corresponding label combinations for each of the adjacent nodes, construct label combination relationships based on the node relationships, and generate external features for each of the label combinations;

[0110] Step 43: Capture the convolution features corresponding to each of the adjacent nodes in the convolution processing results, and combine them with the corresponding circumscribed features to establish the convolution information of the label combination.

[0111] In this example, adjacent nodes represent nodes composed of elements that are adjacent to each other in the binary adjacency matrix;

[0112] In this example, node relationships represent the logical associations between different adjacent nodes;

[0113] In this example, the circumscribed features represent the features presented by the logical relationships between different combinations of labels;

[0114] In this example, the convolutional information represents the information presented after convolving the information of the label combination.

[0115] The working principle and beneficial effects of the above technical solution are as follows: By inputting the binarized adjacency matrix into GCN for convolution training, the node relationship between adjacent nodes is determined. Then, corresponding circumscribed features are set for the corresponding label combinations. At the same time, the convolution information is captured in the convolution processing result. Through convolution, noise in the original information can be eliminated, and the features of the label combinations at different levels can be captured, ensuring the integrity of the convolution information.

[0116] Example 6

[0117] Based on Example 1, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN, step 5 includes:

[0118] Step 51: Establish shadow distribution features corresponding to the label combination based on the binary information, and establish pixel distribution features corresponding to the label combination based on the convolution information;

[0119] Step 52: Obtain the presentation features of each of the label combinations in the binary neighborhood matrix, and construct the label-related features corresponding to each of the label combinations by combining the corresponding shadow distribution features and pixel distribution features;

[0120] Step 53: Draw the label image corresponding to the combined label according to the label-related features, fuse the label images according to the presentation features to generate the multi-label graph structure of the multi-label dataset, and display it.

[0121] In this example, the presentation feature represents the representation of a label combination in a binary neighborhood matrix.

[0122] The working principle and beneficial effects of the above technical solution are as follows: Binary information is used to establish the shadow distribution features of the label combination, and convolutional information is used to establish the pixel distribution features of the label combination. Combined with the presentation features of the label combination in the binary neighborhood matrix, the label-related features of each label combination are constructed. Then, label images are drawn based on these label-related features, thereby constructing a multi-label graph structure. This approach allows for in-depth analysis of the label combination, determining its features in various dimensions, and thus constructing a multi-label graph structure based on these features, thereby improving the accuracy of image classification.

[0123] Example 7

[0124] Based on Example 1, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN further includes:

[0125] Each tag name corresponding to each tag combination is marked in the multi-tag graph structure to obtain the graph domain corresponding to each tag combination and then displayed.

[0126] The working principle and beneficial effects of the above technical solution are as follows: By marking the label name in the multi-label graph structure, users can easily distinguish the information corresponding to different labels.

[0127] Example 8

[0128] Based on Example 7, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN further includes:

[0129] Identify the nodes and edges contained in the multi-label graph structure;

[0130] Identify several first labels corresponding to each node and several second labels corresponding to each edge;

[0131] Calculate the sparsity of the first label corresponding to each node and the sparsity of the second label corresponding to each edge to determine the key structural regions of the multi-label graph structure;

[0132] The key areas are highlighted in the multi-label graph structure.

[0133] In this example, nodes represent the smallest units that make up the multi-label graph structure, and edges represent the structural edges of the multi-label graph structure.

[0134] In this example, label sparsity refers to the density of the labels corresponding to a node or an edge.

[0135] The working principle and beneficial effects of the above technical solution are as follows: By analyzing the label sparsity of nodes and edges in a multi-label graph structure, key areas with high label sparsity are identified. To facilitate users in quickly locating key areas, these key areas are highlighted in the multi-label graph structure, thus improving the user experience.

[0136] Example 9

[0137] Based on Example 1, the method for constructing a multi-label graph structure based on a conditional probability matrix and GCN is characterized by further including:

[0138] The system retrieves relevant information contained in the multi-tab graph structure based on the execution command issued by the user, and transmits it to the corresponding terminal for display.

[0139] The working principle and beneficial effects of the above technical solution are as follows: When a user issues a command, the system responds promptly to the command and filters relevant information to transmit to the user's terminal for display.

[0140] Example 10

[0141] Based on Example 1, the method for constructing a multi-label graph structure based on conditional probability matrix and GCN further includes:

[0142] Based on the user-uploaded graph structure, the domain information contained in the multi-label graph structure is identified and displayed.

[0143] The working principle and beneficial effects of the above technical solution are as follows: Based on the graph structure domain uploaded by the user in advance, the relevant information contained in the multi-label graph structure is identified and displayed, which improves the accuracy of automatic identification of multi-label structures.

[0144] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multi-label graph structure construction method based on conditional probability matrix and GCN, used for image classification and structure determination, characterized in that, include: Step 1: Classify the label data in the multi-label dataset, obtain the co-occurrence frequency corresponding to each label combination, and construct the conditional probability matrix of the multi-label dataset; Step 2: Calculate the mean and standard deviation of each of the labeled data, and construct the corresponding mean-standard deviation dynamic threshold; Step 3: Use the mean-standard deviation dynamic threshold to binarize the conditional probability matrix to generate a binarized adjacency matrix and obtain the binary information corresponding to each of the label combinations; Step 4: Input the binarized adjacency matrix into a preset GCN for convolution processing to obtain the convolution information corresponding to each label combination; Step 5: Generate label association features corresponding to each label combination based on the binary information and convolution information corresponding to each label combination, and construct the multi-label graph structure of the multi-label dataset using the label association features; Step 3 includes: Step 31: Binarize the conditional probability matrix using each of the mean-standard deviation dynamic thresholds to obtain the assignment result of each of the mean-standard deviation dynamic thresholds to the conditional probability matrix; Step 32: Calculate the assignment balance degree corresponding to each assignment result, and use the target assignment result with the highest assignment balance degree to perform binarization assignment on the conditional probability matrix to generate a binarized adjacency matrix. Among them, the assignment balance degree represents the balance ratio of 0 and 1 values ​​in each assignment result; Step 33: Identify the matrix element corresponding to each tag combination in the binary adjacency matrix, and determine the binary information corresponding to the tag combination based on the assigned value of the matrix element; Step 5 includes: Step 51: Establish shadow distribution features corresponding to the label combination based on the binary information, and establish pixel distribution features corresponding to the label combination based on the convolution information; Step 52: Obtain the rendering features of each of the label combinations in the binary adjacency matrix, and construct the label association features corresponding to each of the label combinations by combining the corresponding shadow distribution features and pixel distribution features; Among them, the presentation feature represents the representation of a label combination in the binary adjacency matrix; Step 53: Draw a label image corresponding to the label combination based on the label association features, fuse the label images based on the presentation features to generate a multi-label graph structure of the multi-label dataset, and display it.

2. The method for constructing a multi-label graph structure based on conditional probability matrix and GCN as described in claim 1, characterized in that, Step 1 includes: Step 11: Identify several label data contained in the multi-label dataset, perform binary classification training on each label data to obtain several labels contained in each label data, and identify the samples corresponding to the label data. Step 12: Combine the labels according to the correspondence between samples and labels contained in the same label data to obtain several label combinations corresponding to each label data, and construct the combination conditions corresponding to each label combination according to the label features corresponding to each label. Step 13: Perform a deep statistical search on the multi-label dataset using each of the aforementioned combination conditions to obtain the co-occurrence frequency corresponding to each of the aforementioned label combinations, construct corresponding label events based on the label combinations, and construct corresponding event probabilities based on the co-occurrence frequencies; Step 14: Construct the conditional probability matrix of the multi-label dataset based on the labeled events and event probabilities.

3. The method for constructing a multi-label graph structure based on conditional probability matrix and GCN as described in claim 1, characterized in that, Step 2 includes: Step 21: Sample each of the label data to obtain several label values. Divide the label data into several data segments based on the sampling position corresponding to each label value. Filter the target data segments whose length is higher than the preset length and perform compensation sampling to obtain the corresponding label values. Step 22: Calculate the mean and standard deviation of the corresponding tag data based on the tag values, determine the dynamic threshold range of the corresponding tag data using the mean and standard deviation, and establish the range ratio of the dynamic threshold range based on the ratio of the number of first tag values ​​falling within the dynamic threshold range to the number of second tag values ​​not falling within the dynamic threshold range; Step 23: Using the range comparison, correct the corresponding data mean and standard deviation to obtain several corrected mean and standard deviation. Filter out several target corrected mean and target corrected standard deviation that fall within the corresponding dynamic threshold range to generate the dynamic threshold of mean-standard deviation for the corresponding label data.

4. The method for constructing a multi-label graph structure based on conditional probability matrix and GCN as described in claim 1, characterized in that, Step 4 includes: Step 41: Obtain several neighboring nodes contained in the binarized adjacency matrix, input the binarized adjacency matrix into a preset GCN for convolution processing, and obtain the node relationships between different neighboring nodes; Step 42: Match the corresponding label combination for each of the adjacent nodes, construct the label combination relationship according to the node relationship, and generate the external feature of each of the label combinations; Step 43: Capture the convolution features corresponding to each of the adjacent nodes in the convolution processing results, and combine them with the corresponding circumscribed features to establish the convolution information of the label combination.

5. The method for constructing a multi-label graph structure based on conditional probability matrix and GCN as described in claim 1, characterized in that, Also includes: Each tag name corresponding to each tag combination is marked in the multi-tag graph structure to obtain the graph domain corresponding to each tag combination and then displayed.

6. The method for constructing a multi-label graph structure based on conditional probability matrix and GCN as described in claim 5, characterized in that, Also includes: Identify the nodes and edges contained in the multi-label graph structure; Identify several first labels corresponding to each node and several second labels corresponding to each edge; Calculate the sparsity of the first label corresponding to each node and the sparsity of the second label corresponding to each edge to determine the key structural regions of the multi-label graph structure; The key areas are highlighted in the multi-label graph structure.

7. The method for constructing a multi-label graph structure based on conditional probability matrix and GCN as described in claim 1, characterized in that, Also includes: The system retrieves relevant information contained in the multi-tab graph structure based on the execution command issued by the user, and transmits it to the corresponding terminal for display.

8. The method for constructing a multi-label graph structure based on conditional probability matrix and GCN as described in claim 1, characterized in that, Also includes: Based on the user-uploaded graph structure, the domain information contained in the multi-label graph structure is identified and displayed.

Images