Construction method, system and equipment of network community structure and storage medium
By constructing a heterogeneous graph and matrix factorization model, the accuracy and completeness issues of community discovery in multi-view graph scenarios are solved, and more accurate community structure construction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA UNIV OF MINING & TECH
- Filing Date
- 2024-01-25
- Publication Date
- 2026-04-17
AI Technical Summary
Existing community detection methods neglect potential cross-view consistency features and the community structure characteristics of each view graph in multi-view graph scenarios, resulting in inaccurate and incomplete community structure of the obtained network data.
By constructing a heterogeneous graph and establishing a multi-view graph G, and combining the node adjacency matrix Mk, community structure matrix Xk, global consistent structure matrix C, and complementary residual structure matrix Ek, the matrix decomposition model is iterated to calculate the global community structure matrix Z, thus constructing the community structure of the network data.
It improves the community discovery performance of multi-view graph data, and the obtained community structure is more accurate and complete, integrating the community structure characteristics of various view graphs.
Smart Images
Figure CN121883196A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data analysis technology, and in particular to a method, system, device, and storage medium for constructing a network community structure. Background Technology
[0002] In the era of big data, we can obtain massive amounts of network data from various sources, such as data from social media, e-commerce, and internet searches. This network data reflects various interactive behaviors in people's lives and has significant value for industrial development in social, economic, and technological fields. Community discovery, as an effective data analysis method, can help us better understand the structure and function of networks, more accurately recognize the social phenomena behind people's behavioral data, and explore social development trends, thus having significant practical implications. The basic idea of community discovery is to divide network data into mutually separated subsets based on the connections between nodes; these subsets can be considered communities. Nodes within a community have denser connections, while connections between these nodes and nodes outside the community are relatively sparse.
[0003] Multi-view graphs are a novel data structure that can encompass multiple connections and attributes in network data, thus providing a more comprehensive description of network complexity. In multi-view graphs, we can perform community detection from different perspectives: based on node attributes, connection types, or even considering all these factors simultaneously.
[0004] Existing community detection methods involve constructing algorithms to merge nodes with similar attributes in a multi-view graph, forming multiple sub-clusters to constitute a community structure. The specific process is as follows: transforming network data into a form suitable for algorithm processing; calculating the similarity between each pair of nodes in the network; merging nodes with high similarity to form sub-clusters; and obtaining the community structure after all nodes have been merged.
[0005] The drawback of the above technical solution is that it ignores the potential cross-view graph consistency characteristics and the community structure characteristics of each view graph in the multi-view graph scenario, resulting in the community structure corresponding to the obtained network data being inaccurate and incomplete. Summary of the Invention
[0006] Therefore, it is necessary to provide a method, system, device, and storage medium for constructing a network community structure to address the aforementioned technical problems.
[0007] This invention provides a method for constructing a network community structure, comprising:
[0008] Obtain the heterogeneous graph corresponding to the network data. The heterogeneous graph includes: multiple nodes representing user terminals with different types of attributes and edges representing the connection relationships between user terminals with different types of attributes. Based on the nodes and edges in the heterogeneous graph, establish a multi-view graph G for each node. The multi-view graph G includes: a central node and edges connected to the central node.
[0009] Based on the multi-view graph G of each central node, construct the adjacency matrix M representing the self-circling network of nodes from each viewpoint. k And the community structure matrix X of nodes with similar attribute features. k And the adjacency matrix M representing the self-circling network of nodes k Combined with the community discovery model;
[0010] Based on the community detection model, a matrix decomposition model is obtained by combining the globally consistent structure matrix and the complementary residual structure matrix that represent the constraint terms of the relationship between nodes.
[0011] The matrix factorization model is iterated through multiple rounds of parameter value calculation. The parameter value in the (t+1)th round is calculated based on the parameter value obtained in the tth round. The iteration ends when the parameter value converges or the set number of iterations is reached.
[0012] The global community structure matrix Z of the multi-view graph G is calculated based on the iteration results, and then... The corresponding community tag serves as the community to which the central node belongs;
[0013] The community structure of network data is constructed based on the relationship between the central node and its affiliated community.
[0014] Additionally, the community detection model is as follows:
[0015]
[0016] Among them, D k For the community-to-community correlation matrix, X k For the community structure matrix, M k This is the node adjacency matrix. It represents the square of the F-norm of the matrix.
[0017] Furthermore, the globally consistent structure matrix is:
[0018]
[0019] Among them, X k Let C be the community structure matrix, and let C be the community consistency structure matrix across different attribute perspectives. It represents the square of the F-norm of the matrix.
[0020] Furthermore, the complementary residual structure matrix is:
[0021]
[0022] Where C is the community consistency structure matrix across different attribute perspectives, E k D is the complementary residual structure matrix from the perspective of attribute k. k M is the community-to-community correlation matrix. k This is the node adjacency matrix. It represents the square of the F-norm of the matrix.
[0023] Alternatively, the matrix decomposition model is as follows:
[0024]
[0025]
[0026] Among them, X k For the community structure matrix, D k Let C be the community-to-community correlation matrix, and E be the community consistency structure matrix across different attribute perspectives. k M is the complementary residual structure matrix from the perspective of attribute k. k Let Q be the node adjacency matrix. k As the normalization factor, α k It is a structural weight adjustment parameter. It represents the square of the F-norm of the matrix.
[0027] Additionally, the community structure matrix X k The calculation formula is:
[0028]
[0029] The community-community correlation matrix D k The calculation formula is:
[0030]
[0031] f-(D k )=β k C T M k C+(1-β k E k T M k E k
[0032] f + (D k )=β k 2 (C T CD kC T C+E k T E k D k E k T E k )+β k (1-β k (E) k T CD k C T E k +C T E k D k E k T C)
[0033] The formula for calculating the community consensus structure matrix C across different attribute perspectives is as follows:
[0034]
[0035] The complementary residual structure matrix E from the perspective of attribute k k The calculation formula is:
[0036]
[0037] The normalization factor Q k The calculation formula is:
[0038]
[0039] Structural weight adjustment parameter α k The calculation formula is:
[0040]
[0041] Where dig() represents a diagonal matrix, α k These are the structural weight adjustment parameters, γ and β. k p k q k D is the weight hyperparameter of the corresponding term. k For the community-to-community correlation matrix, X k For the community structure matrix, M k Let E be the node adjacency matrix. k Let Q be the complementary residual structure matrix from the perspective of attribute k. k is the normalization factor, and C is the community consistency structure matrix across different attribute perspectives.
[0042] Furthermore, the formula for calculating the global community structure matrix Z of the multi-view graph G based on the iteration results is as follows:
[0043]
[0044] Where C is the community consistency structure matrix across different attribute perspectives, E k Let γ and β be the complementary residual structure matrices from the perspective of attribute k. k a k b k This is a hyperparameter.
[0045] Another type of community discovery system for online data includes:
[0046] The acquisition module is used to acquire the heterogeneous graph corresponding to the network data. The heterogeneous graph includes: multiple nodes representing user terminals with different types of attributes and edges representing the connection relationships between user terminals with different types of attributes. Based on the nodes and edges in the heterogeneous graph, a multi-view graph G is established for each node. The multi-view graph G includes: a central node and multiple edges connected to the central node.
[0047] The community discovery module is used to construct an adjacency matrix M representing the self-circling network of nodes from each perspective, based on the multi-view graph G of each central node. k And construct a community structure matrix X with nodes having similar attribute features. k And the adjacency matrix M representing the self-circling network of nodes k Community detection model;
[0048] The matrix factorization module, based on the community detection model, combines the globally consistent structure matrix and the complementary residual structure matrix representing the constraint terms of the relationship between nodes to obtain the matrix factorization model.
[0049] The iteration module is used to perform multiple rounds of parameter value iteration on the matrix factorization model. The parameter value in the (t+1)th round is calculated based on the parameter value obtained in the tth round. The iteration ends when the parameter value converges or the set number of iterations is reached.
[0050] The community relations module is used to calculate the global community structure matrix Z of the multi-view graph G based on the iteration results. The corresponding community tag serves as the community to which the central node belongs;
[0051] The structure building module is used to construct the community structure of network data based on the relationship between the central node and its affiliated communities.
[0052] Another computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described network community structure construction method.
[0053] Alternatively, a storage medium may be provided on which a computer program is stored, which, when executed by a processor, implements the steps of the aforementioned network community structure construction method.
[0054] The method, system, device, and storage medium for constructing a network community structure provided in this invention have the following advantages compared to the prior art:
[0055] By acquiring a heterogeneous graph of network data, which contains multiple nodes representing different types of objects, a multi-view graph G is constructed for each node based on the relationships between these different object types. Then, an adjacency matrix M is built from each viewpoint based on the multi-view graph G. k Constructing the community structure matrix X k With the node adjacency matrix M k A community detection model; based on the community detection model, a globally consistent structure matrix C and a complementary residual structure matrix E are added. k The matrix factorization model is obtained, and the global community structure matrix of the multi-view graph is calculated based on the iterative results of the matrix factorization model. The corresponding community tags serve as the affiliated communities, and the community structure of network data is established based on the relationship between the central node and its affiliated communities.
[0056] Compared to existing technologies, the community detection model, combined with the global consistent structure matrix and the complementary residual structure matrix, forms a matrix decomposition model. The global consistent structure matrix and the complementary residual structure matrix, which represent the constraint terms of the relationship between nodes, can integrate the communities of each view and enhance the community detection effect for multi-view graph data. This achieves a comprehensive consideration of the scene effects of the multi-view graph itself and the community structure characteristics under each view graph, making the obtained network data community structure more accurate and complete. Attached Figure Description
[0057] Figure 1 A flowchart illustrating a method for constructing a network community structure in one embodiment;
[0058] Figure 2 This is a general block diagram of a method for constructing a network community structure provided in one embodiment. Detailed Implementation
[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0060] Example 1:
[0061] Embodiment 1 of this invention provides a method for constructing a network community structure, which can be used to analyze data containing social media. Since social media users can establish attribute relationships at different levels, such as friend relationships, preference relationships, and colleague relationships, utilizing the multi-perspective hierarchical relationships of social media can more comprehensively and effectively discover group relationships among users. Based on this, the potential group structure of users can be mined, and a three-dimensional user group profile can be built for product recommendations and advertising.
[0062] It can also be used to analyze data containing scientific collaboration networks. Current scientific research is rapidly evolving, with many studies being interdisciplinary and cross-professional. Different scientists can establish scientific collaboration networks based on project and paper partnerships. By examining different specialties and research themes, these networks can be decomposed into hierarchical structures. Based on this, multi-perspective analysis can be conducted to gain a deeper understanding of scientists' research communities, enhance the understanding of potential relationships between various scientific research directions, and construct academic knowledge graphs.
[0063] For example, it can be used to analyze data containing financial transaction networks. Transaction networks in the financial field can be abstracted as hierarchical attribute graphs from different perspectives. Each level represents a different transaction type, such as stock trading, bond trading, and foreign exchange trading. Different trading entities are connected through these different transaction types. Community building and node relationship analysis based on financial transaction networks help identify financial risks, monitor market fluctuations, and assist financial institutions in risk assessment and management to maintain the stability and sustainable development of the financial system.
[0064] Furthermore, it can be applied to analyze transportation network data. Each layer in the network represents the characteristics of the transportation network from different perspectives, such as road structure networks, road flow networks, and transportation mode networks. Locations are connected through attribute relationships from different perspectives. Based on this, the clustering relationships between locations can be explored and abstracted into community structures. By optimizing traffic flow, reducing congestion, and assisting in the planning of urban transportation infrastructure and traffic management strategies, the efficiency and sustainability of urban transportation systems can be improved.
[0065] Furthermore, it can be used to analyze internet communication network data. Terminal nodes are connected through relationships from different perspectives, such as router networks, switch networks, traffic networks, protocol-type networks, and terminal-type networks. Based on these layered networks, community structures can be discovered, thereby identifying and preventing network attacks, detecting abnormal behavior, and providing real-time threat intelligence and security decision support to protect the security and privacy of networks and information systems.
[0066] One embodiment provides a method for constructing a network community structure, such as... Figure 1 As shown, the method includes:
[0067] Let G be the multi-view graph, then G k ∈G represents the network set under different perspectives k, and U represents the node set.
[0068] Define the node adjacency matrix as M k Then the adjacency matrix M k Specifically refers to a network in which nodes form their own loops.
[0069] Define the community structure matrix as X k Then matrix X k This represents the community structure matrix formed from the perspective of attribute k in the network, where nodes with similar attribute characteristics are formed. The row element values in the matrix represent the user's community membership degree.
[0070] Let Z denote the global community structure matrix of the multi-view graph G, and E represent it. k Let C represent the complementary residual structure matrix from the perspective of attribute k, and let C represent the community consensus structure matrix across different attribute perspectives.
[0071] Step 1: Obtain the heterogeneous graph corresponding to the network data. The heterogeneous graph includes multiple nodes representing user terminals with different attribute types and edges representing the connection relationships between user terminals with different attribute types. Based on the nodes and edges in the heterogeneous graph, construct a multi-view graph G for each node. The multi-view graph G includes a central node and multiple edges connected to the central node. Perform numerical processing, and construct an adjacency matrix M representing the self-contained loop network of the nodes from each perspective based on the multi-view graph G of each central node. k .like Figure 2 As shown, multi-view graphs can exhibit a hierarchical structure. Each view k can generate a graph structure, with nodes forming an adjacency matrix M representing a self-contained loop network of nodes. k In this invention, to reduce the impact of differences in the distribution of adjacency matrix element values from different perspectives, ||M| is taken as... k ||1 = 1 performs norm normalization on the adjacency matrix.
[0072] Step 2: At each attribute k-view level, conduct community discovery from that perspective. The specific optimization objectives are as follows:
[0073]
[0074] Among them, D k This represents the community-to-community correlation matrix. As an intermediate factor in a three-factor nonnegative matrix factorization, it can increase the degrees of freedom for the values of other factors during the matrix factorization process. It represents the square of the F-norm of the matrix.
[0075] Based on the optimization objective of the community discovery model, a globally consistent structure matrix C representing the constraints on the relationships between nodes is added to the optimization objective. This matrix is used to fuse the communities from different perspective graphs. The calculation process for the constraints is as follows:
[0076]
[0077] This yields a fusion factor that reflects the structure of each perspective, and its value will approach the community structure matrix X of each perspective graph as closely as possible. k , It represents the square of the F-norm of the matrix.
[0078] We further introduce the complementary residual structure matrix E to represent the constraint terms of the relationships between nodes. k To enhance community discovery for multi-view graph data, the model components are as follows:
[0079]
[0080] Where C is the community consistency structure matrix across different attribute perspectives, E k D is the complementary residual structure matrix from the perspective of attribute k. k M is the community-to-community correlation matrix. k This is the node adjacency matrix. It represents the square of the F-norm of the matrix.
[0081] Combining the above model components, the matrix factorization model is as follows:
[0082]
[0083]
[0084] Among them, X k For the community structure matrix, D k Let C be the community-to-community correlation matrix, and E be the community consistency structure matrix across different attribute perspectives. k M is the complementary residual structure matrix from the perspective of attribute k. k Let Q be the node adjacency matrix. k As the normalization factor, α k It is the structural weight adjustment parameter, p k q k γ and γ are the weight hyperparameters of the corresponding terms. It represents the square of the F-norm of the matrix.
[0085]
[0086] Where dig() represents a diagonal matrix, α kThese are the structural weight adjustment parameters, γ and β. k p k q k These are the weight hyperparameters of the corresponding items.
[0087] Step 3: Train the parameters of the community detection model using a variable iteration strategy. Specifically, based on the community detection optimization objective obtained in Step 2, construct the Lagrangian function. Under the condition of satisfying the KKT conditions, the parameters X of each model can be obtained. k D k C, E k and α k The calculation process for each parameter is as follows.
[0088]
[0089]
[0090] in,
[0091] f - (D k )=β k C T M k C+(1-β k E k T M k E k
[0092] f + (D k )=β k 2 (C T CD k C T C+E k T E k D k E k T E k )+β k (1-β k (E) k T CD k C T E k +C T E k D k E k T C)
[0093]
[0094]
[0095]
[0096] Where dig() represents a diagonal matrix, α k These are the structural weight adjustment parameters, γ and β. k p k q k D is the weight hyperparameter of the corresponding term. k For the community-to-community correlation matrix, X k For the community structure matrix, M k Let E be the node adjacency matrix. k Let Q be the complementary residual structure matrix from the perspective of attribute k. k is the normalization factor, and C is the community consistency structure matrix across different attribute perspectives.
[0097] After initializing the model parameters, the parameter values are iterated through multiple rounds. The parameter values in the (t+1)th round are calculated based on the parameter values obtained in the tth round. The iteration ends when the parameter values converge or after the set number of iterations, and the parameter values of the model can be obtained.
[0098] Step 4: Calculate the global community structure matrix Z of the multi-view graph based on the iteration results. The process is as follows.
[0099]
[0100] Finally, based on the element values of the community structure matrix Z, take... The corresponding community serves as the subordinate community of the central node, thus revealing the community relationships between nodes. The hyperparameters γ and β in the model... k a k b k The value can be determined through experimental results. Generally, γ∈[0.01,10] and β can be taken as follows. k ∈(0,1), p k =q k =1.
[0101] Step 5: Construct the community structure of network data based on the relationship between the central node and its affiliated community.
[0102] In one embodiment, a community discovery system for network data is provided, the system comprising:
[0103] The acquisition module is used to acquire the heterogeneous graph corresponding to the network data. The heterogeneous graph includes: multiple nodes representing user terminals with different types of attributes and edges representing the connection relationships between user terminals with different types of attributes. Based on the nodes and edges in the heterogeneous graph, a multi-view graph G is established for each node. The multi-view graph G includes: a central node and multiple edges connected to the central node.
[0104] The community discovery module is used to construct an adjacency matrix M representing the self-circling network of nodes from each perspective, based on the multi-view graph G of each central node. k And construct a community structure matrix X with nodes having similar attribute features. k And the adjacency matrix M representing the self-circling network of nodes k Community detection model;
[0105] The matrix factorization module, based on the community detection model, combines the globally consistent structure matrix and the complementary residual structure matrix representing the constraint terms of the relationship between nodes to obtain the matrix factorization model.
[0106] The iteration module is used to perform multiple rounds of parameter value iteration on the matrix factorization model. The parameter value in the (t+1)th round is calculated based on the parameter value obtained in the tth round. The iteration ends when the parameter value converges or the set number of iterations is reached.
[0107] The community relations module is used to calculate the global community structure matrix Z of the multi-view graph G based on the iteration results. The corresponding community tag serves as the community to which the central node belongs;
[0108] The structure building module is used to construct the community structure of network data based on the relationship between the central node and its affiliated communities.
[0109] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described network community structure construction method.
[0110] A storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described network community structure construction method.
[0111] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.
Claims
1. A method of constructing a network community structure, characterized by, include: Obtain the heterogeneous graph corresponding to the network data. The heterogeneous graph includes: multiple nodes representing user terminals with different types of attributes and edges representing the connection relationships between user terminals with different types of attributes. Based on the nodes and edges in the heterogeneous graph, establish a multi-view graph G for each node. The multi-view graph G includes: a central node and multiple edges connected to the central node. According to the multi-view graph G of each center node, an adjacency matrix M representing a node self-loop network under each view is constructed k , and a community structure matrix X of nodes with similar attribute characteristics k is combined with the adjacency matrix M representing the node self-loop network k to obtain a community discovery model; On the basis of the community discovery model, a global consistent structure matrix C and a complementary residual structure matrix E are combined to represent the relationship constraints between nodes k A matrix decomposition model is obtained; The matrix factorization model is iterated through multiple rounds of parameter value calculation. The parameter value in the (t+1)th round is calculated based on the parameter value obtained in the tth round. The iteration ends when the parameter value converges or the set number of iterations is reached. According to the iteration result, a global community structure matrix Z of the multi-view graph G is calculated, and the community label corresponding to the center node is taken as the membership community of the center node. the community label corresponding to the center node is taken as the membership community of the center node. The community structure of network data is constructed based on the relationship between the central node and its affiliated community.
2. The method of claim 1, wherein the network community structure is constructed by: The community detection model is as follows: where D k is the community-community association matrix, X k is the community structure matrix, M k is the node adjacency matrix, denotes the square of the F-norm of a matrix, which is the sum of the squares of all elements in the matrix.
3. The method for constructing a network community structure as described in claim 1, characterized in that, The globally consistent structure matrix is: Among them, X k Let C be the community structure matrix, and let C be the community consistency structure matrix across different attribute perspectives. It represents the square of the F-norm of the matrix.
4. The method of claim 1, wherein the network community structure is constructed by: The complementary residual structure matrix is: where C is the community-consistent structure matrix across different attribute perspectives, E k is the complementary residual structure matrix under attribute k perspective, D k is the community-community relevance matrix, M k is the node adjacency matrix, denotes the square of the F-norm of a matrix.
5. The method of claim 1, wherein the network community structure is constructed by: The matrix factorization model is as follows: Among them, X k For the community structure matrix, D k Let C be the community-to-community correlation matrix, and E be the community consistency structure matrix across different attribute perspectives. k M is the complementary residual structure matrix from the perspective of attribute k. k Let Q be the node adjacency matrix. k As the normalization factor, α k It is the structural weight adjustment parameter, p k q k γ and γ are the weight hyperparameters of the corresponding terms. It represents the square of the F-norm of the matrix.
6. The method for constructing a network community structure as described in claim 5, characterized in that, The community structure matrix X k The calculation formula is: The community-community correlation matrix D k The calculation formula is: f - (D k )=β k C T M k C+(1-β k )E k T M k E k f + (D k )6β k 2 (C T CD k C T C+E k T E k D k E k T E k )+β k (1-β k )(E k T CD k C T E k +C T E k D k E k T C) The formula for calculating the community consensus structure matrix C across different attribute perspectives is as follows: The attribute k perspective complementary residual structure matrix E k The calculation formula is: The normalization factor Q k The calculation formula is: The structural weight adjustment parameter α k The calculation formula is: Where dig() represents a diagonal matrix, α k These are the structural weight adjustment parameters, γ and β. k p k q k D is the weight hyperparameter of the corresponding term. k For the community-to-community correlation matrix, X k For the community structure matrix, M k Let E be the node adjacency matrix. k Let Q be the complementary residual structure matrix from the perspective of attribute k. k is the normalization factor, and C is the community consistency structure matrix across different attribute perspectives.
7. The method of claim 1, wherein the network community structure is constructed by: The formula for calculating the global community structure matrix Z of the multi-view graph G based on the iteration results is as follows: where C is the community-consistent structure matrix across different attribute perspectives, E k is the complementary residual structure matrix under attribute k perspective, γ, β k , a k , b k are hyperparameters.
8. A community discovery system of network data, characterized by, include: The acquisition module is used to acquire the heterogeneous graph corresponding to the network data. The heterogeneous graph includes: multiple nodes representing user terminals with different types of attributes and edges representing the connection relationships between user terminals with different types of attributes. Based on the nodes and edges in the heterogeneous graph, a multi-view graph G is established for each node. The multi-view graph G includes: a central node and multiple edges connected to the central node. a community discovery module, configured to construct an adjacency matrix M representing a self-loop network of nodes in each perspective according to the multi-perspective graph G of each center node k , and construct a community structure matrix X of nodes with similar attribute characteristics k , and the community discovery model of the adjacency matrix M representing the self-loop network of nodes k The matrix factorization module, based on the community detection model, combines the globally consistent structure matrix and the complementary residual structure matrix representing the constraint terms of the relationship between nodes to obtain the matrix factorization model. The iteration module is used to perform multiple rounds of parameter value iteration on the matrix factorization model. The parameter value in the (t+1)th round is calculated based on the parameter value obtained in the tth round. The iteration ends when the parameter value converges or the set number of iterations is reached. The community relations module is used to calculate the global community structure matrix Z of the multi-view graph G based on the iteration results. The corresponding community tag serves as the community to which the central node belongs; The structure building module is used to construct the community structure of network data based on the relationship between the central node and its affiliated communities.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the network community structure construction method according to any one of claims 1-7.
10. A storage medium having stored thereon a computer program, characterized in that When the computer program is executed by a processor, it implements the steps of the network community structure construction method according to any one of claims 1-7.