Community discovery method and device based on binary symbol network, and computing equipment

Abstract

The invention belongs to the technical field of complex network analysis, and particularly relates to a community discovery method and device based on a binary symbol network and computing equipment. The method specifically comprises the following steps: step S101, establishing node and edge connection relation data of a bipartite symbol network according to evaluation relation data of a user and a commodity; s102, for two nodes with common secondary nodes in the main node set, according to the edge connection relation of the network and the weight of the secondary nodes, the edge connection weight value of the two nodes is calculated, and a main node set network adjacency matrix W is constructed; step S103, according to the adjacency matrix W of the main set node network, calculating to obtain an embedding matrix Y of the nodes and a community membership matrix C; and S104, adjusting the network adjacency matrix W according to the calculated node community membership matrix C, then re-optimizing the node embedding matrix Y and the node community membership matrix C to form iterative optimization, and finally obtaining a node embedding and community division result.

Description

Technical Field

[0001] The present invention belongs to the technical field of complex network analysis, and specifically relates to a community discovery method, device and computing equipment based on a bipartite symbolic network. Background Art

[0002] A bipartite network consists of two heterogeneous nodes, and edges only exist between different types of nodes. Many complex networks in the real world can be represented as bipartite networks, such as the collaboration network between authors and articles, the transaction network between buyers and sellers, and the evaluation network of movie websites. Therefore, community discovery for bipartite networks has important practical value for complex network research and can effectively analyze the social structure and function in real-world networks. However, current research on bipartite networks mainly focuses on unsigned networks. Real networks often have both negative and positive edges, forming a bipartite signed network with the characteristics of both bipartite and signed networks.

[0003] Since the edges in a bipartite network only exist between two node sets, it is necessary to reconstruct the relationship between nodes in the same node set based on the common neighbors of the nodes to form a network within the node set. However, the above method cannot distinguish different types of neighbors in a bipartite symbolic network, and the process of reconstructing the network will lose the secondary set nodes in the network, ignoring the balanced structural information formed by the difference in edge signs, resulting in an inability to accurately divide the community. In addition, the existing methods only consider low-order information such as first-order neighbors and second-order neighbors, ignoring the global community information of the network, making it difficult to learn the low-dimensional representation vector of the node and accurately divide the network community. Summary of the invention

[0004] In view of the above problems, the present invention is proposed to provide a community discovery method, apparatus and computing device based on a bipartite symbolic network that overcomes the above problems or at least partially solves the above problems.

[0005] The technical solution for implementing the present invention is as follows:

[0006] A community discovery method based on a bipartite symbolic network, comprising:

[0007] Step S101: Based on the evaluation relationship data of users and products, establish the node and edge relationship data of the bipartite symbol network, wherein the user group is represented by the main node set in the network, and the products are represented by the secondary node set in the network. When the user evaluates the product positively, the edge is positive, and when the user evaluates the product negatively, the edge is negative; based on the positive and negative degrees of the network secondary node set and the butterfly structure, calculate the weight of each node in the secondary node set;

[0008] Step S102: for two nodes in the main node set that have a common secondary node, calculate the edge weight values ​​of the two nodes according to the network edge relationship and the secondary node weight; and construct the main node set network adjacency matrix W based on the edge weight values ​​of all nodes in the main node set;

[0009] Step S103: Calculate the node embedding matrix Y and community membership matrix C according to the adjacency matrix W of the main set node network;

[0010] Step S104: The network adjacency matrix W is adjusted according to the calculated node community membership matrix C, and then the node embedding matrix Y and the node community membership matrix C are re-optimized to form an iterative optimization, and finally the node embedding and community division results are obtained.

[0011] Optionally, in step S101, the weight of each node in the secondary node set is calculated according to the positive and negative degrees of the network secondary node set and the number of butterfly structures:

[0012] Each secondary node set in the bipartite signed network is analyzed, and the number of positive edges and negative edges of the node is counted, which is recorded as the positive and negative degree of the node. The number of balanced structures and the number of unbalanced structures in the butterfly structure containing the node in the network are counted. The weight of each node in the secondary node set is calculated based on the positive and negative degree and the number of balanced and unbalanced structures.

[0013] Optionally, the weight z of each node k in the secondary node set is k Specifically:

[0014]

[0015] in, represents the positive degree of node k, represents the negative degree of node k, U k represents the number of unbalanced structures containing node k, B k Represents the number of balanced structures containing node k.

[0016] Optionally, step S102 calculates the edge weight values ​​of the two nodes having a common secondary node (neighbor) in the primary node set according to the edge relationship of the network and the secondary node weights, and the specific process is:

[0017] Assume that there is a common secondary node k between node i and node j in the main node set, then the edge size corresponding to node k is the weight of node k, and the positive or negative edge is determined as follows: when nodes i and j have edges with the same sign as node k, the edge is positive, otherwise, it is negative; calculate the edges corresponding to the remaining common secondary nodes of node i and node j according to the above process;

[0018] Based on the edges corresponding to all common subnodes of node i and node j, calculate the edges between node i and node j.

[0019] Optionally, the edge W between the computing node i and the node j is ij The specific formula is:

[0020]

[0021] Among them, A ik A represents the edge symbol between node i and node k. jk Represents the edge symbol between node j and node k.

[0022] Optionally, the step S103 includes:

[0023] The corresponding network adjacency matrix W is constructed according to the main set node network, and the node embedding matrix and community membership matrix are calculated using the augmented Lagrange multiplier method.

[0024] Optionally, the node embedding matrix Y and the community membership matrix C are specifically:

[0025]

[0026] C=(μQ-λ4)(μI-2J+2R)-1

[0027] Among them, matrices F and H are auxiliary matrices related to community membership and network adjacency matrix, J and R are auxiliary matrices related to node embedding, matrix D is the degree matrix of the network, matrices P and Q are intermediate variable matrices of Y and C, and their initial values ​​are the same; λ1, λ3, λ4 are Lagrange multipliers, and μ is the penalty coefficient.

[0028] Optionally, the step S104 adjusts the network adjacency matrix W according to the calculated node community membership matrix C to:

[0029]

[0030] Among them, α is the setting parameter, c i represents the community membership of node i, c j represents the community membership of node j.

[0031] According to another aspect of the present invention, a community discovery device for a binary symbol network is provided, comprising:

[0032] The network construction and sub-node set weighting module is used to establish the node and edge relationship data of the binary symbol network based on the evaluation relationship data of users and products, where the user group is represented by the main node set in the network, and the products are represented by the sub-node set in the network. When the user evaluates the product positively, the edge is positive, and when the user evaluates the product negatively, the edge is negative; based on the positive and negative degrees of the network sub-node set and the butterfly structure, the weight of each node in the sub-node set is calculated;

[0033] The main node set network construction module is used to calculate the edge weight values ​​of the two nodes with a common secondary node in the main node set according to the edge relationship of the network and the secondary node weight; based on the edge weight values ​​of all nodes in the main node set, construct the main node set network adjacency matrix W;

[0034] The node embedding and membership calculation module calculates the node embedding matrix Y and community membership matrix C according to the adjacency matrix W of the main set node network;

[0035] The optimization module adjusts the network adjacency matrix W according to the calculated node community membership matrix C, and then re-optimizes the node embedding matrices Y and C to form an iterative optimization, and finally obtains the node embedding and community division results.

[0036] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0037] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned community discovery method for a binary symbolic network.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] First, compared with the existing method of using bipartite network for community discovery, which does not consider the positive or negative user evaluation of products, the present invention further considers the positive and negative evaluations when using the evaluation relationship data of users and products, establishes positive edges for positive evaluations, and negative edges for negative evaluations. Therefore, the established bipartite network can better reflect the relationship characteristics between users and products, thereby making the community discovery results more accurate.

[0040] Second, the present invention calculates the weight of each node in the secondary node set based on the positive and negative degrees and butterfly structure of the network secondary node set, and then constructs the network adjacency matrix W of the main node set based on the network's edge relationship and the secondary node weight; compared to the existing binary network, which cannot form a network within the node set, the present invention constructs a network adjacency matrix in the main node set, thereby being able to distinguish different types of neighbors and ensure the accuracy of the community discovery results.

[0041] Third, the present invention reconstructs the edge symbols between the main set nodes based on the structural balance theory, and weights the secondary set nodes based on the proportion of the butterfly unbalanced structure in the network, and then maps the edge weights between the main set nodes, thereby realizing the construction of the main set network; according to the edge weights of the main set network, the node embedding and community membership are optimized, and then the network edge weights are dynamically adjusted according to the optimization results during the iteration process, realizing the cyclic optimization of network edge weights and node embedding. Different from the existing models, the present invention adds the optimization of community information to the network embedding, realizes the coupling of network embedding and community discovery, accelerates the iterative process of node embedding, makes the network representation results more discriminative and effective, and the community discovery results more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0043] Figure 1 A flow chart of a community discovery method for a binary symbolic network provided by an embodiment of the present invention is shown;

[0044] Figure 2 A schematic diagram of the main node network construction process provided by an embodiment of the present invention is shown;

[0045] Figure 3 A flowchart of updating rules for node embedding matrix and community membership matrix provided by an embodiment of the present invention is shown;

[0046] Figure 4 A schematic diagram of the structure of a community discovery device for a binary symbol network provided by an embodiment of the present invention is shown;

[0047] Figure 5 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0048] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0049] Example 1

[0050] Figure 1 A flowchart of an embodiment of a method for discovering communities in a binary symbolic network according to the present invention is shown. Figure 1 As shown, the method comprises the following steps:

[0051] Step S101: Based on the evaluation relationship data of users and products, the node and edge relationship data of the bipartite symbol network are established, wherein the user group is represented by the main node set in the network, and the products are represented by the secondary node set in the network. When the user evaluates the product positively, the edge is positive, and when the user evaluates the product negatively, the edge is negative; based on the positive and negative degrees and the butterfly structure of the network secondary node set, the weight of each node in the secondary node set is calculated.

[0052] In this embodiment, the evaluation relationship network of users and commodities with polarity is represented by a binary symbolic network graph structure data G=(U, V, ε + ,ε - ), where the user group is represented by the main set node U in the bipartite symbolic network, the product is represented by the secondary set node V in the bipartite symbolic network, and the user's evaluation relationship on the product is represented by the edges between the nodes. In particular, the positive relationship is represented by the positive edge, the negative relationship is represented by the negative edge, and the non-existent relationship or the neutral relationship is represented by the edge with a weight of 0.

[0053] Specifically in this embodiment, and Represents the primary node set and the secondary node set respectively, n U and n V Respectively represent the number of nodes in each set. + and ε - Represents the set of positive and negative edges between two sets of nodes. The edges describe the relationship between nodes, which can also be expressed as an adjacency matrix. Indicates that A ij Represents node u i With v j The relationship between A ij =1 indicates node u i With v j There is a positive edge between them, A ij =0 indicates node u i With v jThere is no edge between them. ij =-1 indicates node u i With v j There are negative edges between them.

[0054] Count the positive degree of node k in the secondary node set V and negative The positive degree and negative degree of a node represent the number of positive edges and negative edges of the node, respectively. The number of butterfly structures containing node k in the statistical network is a unique balanced structure in a binary signed network. It contains two main set nodes and two secondary set nodes. The edges between the nodes form a butterfly-shaped loop structure with a length of 4. Based on the structural balance theory, the sign of the edge can be used to determine whether the butterfly structure is balanced. The nodes are weighted according to the ratio of the positive and negative degrees of node k and the ratio of the unbalanced structure in the butterfly structure. The node weight z k The specific formula is as follows:

[0055]

[0056] Among them, U k represents the number of unbalanced structures containing node k, B k Represents the number of balanced structures containing node k.

[0057] Step S102: For two nodes in the main node set that have a common secondary node (neighbor), calculate the edge weight values ​​of the two nodes according to the network's edge relationship and the secondary node weights; based on the edge weight values ​​of all nodes in the main node set, construct the main node set network adjacency matrix W.

[0058] For a bipartite signed network, there are no edges between nodes in the same node set. Therefore, based on the structural balance theory, according to the two edges between different node sets, a third edge in the same node set is constructed to obtain a signed network composed of nodes of the same type. Figure 2 As shown in the figure, when nodes i and j in the primary node set U have edges with the same sign as node k in the secondary node set V, a positive edge is constructed between them, and when nodes i and j have edges with different signs, a negative edge is constructed between them. At the same time, the weight of node k is used as the weight of the constructed edge.

[0059] Consider all common neighbors of nodes i and j, build new edges based on the edge signs and weights of each common neighbor, and then determine the final edge W based on the signs and weight ratios of all new edges ij , the specific formula is as follows:

[0060]

[0061] Among them, A ikA represents the edge symbol between node i and node k. jk Represents the edge symbol between node j and node k.

[0062] When all the edges between two nodes with a common secondary node in the main node set are calculated, all the edges together constitute the main node set network adjacency matrix W.

[0063] Step S103, constructing a corresponding network adjacency matrix W according to the main set node network, and using the augmented Lagrange multiplier method to calculate the node embedding matrix Y and the community membership matrix C.

[0064] The community membership of a node is used to represent the community information of the network. The community membership is represented by any value between 0 and 1. A node can belong to different communities with different memberships. The community membership matrix of the network is represented by C = {c ui} l×n , where c ui represents the membership between community u and node i, n is the number of nodes, l is the number of network communities, and there are the following constraints:

[0065]

[0066] According to the above definition of community membership, each column of the matrix represents the affiliation between the node and all network communities. Usually, the community with the largest affiliation represents the community label of the node.

[0067] Through node embedding, the nodes in the network are mapped to a low-dimensional vector space, capturing the structure and feature information of the nodes as the representation vector of the nodes, and obtaining the node embedding matrix Y. By converting the original features of the nodes into more expressive continuous vectors, the distance and similarity of the nodes in the vector space can better reflect the relationship between the nodes.

[0068] The following convex optimization problem is established based on the node embedding matrix Y and the community membership matrix C:

[0069]

[0070] stYTDY=I,CT1l×n=1n×n

[0071] Among them, c i represents the community membership of node i, and matrix D is the degree matrix of the graph. α>0 is a parameter used to balance the influence between the network edge weight and the error of the node community membership.

[0072] Since the objective function is convex, the above optimization problem has a stable solution. The augmented Lagrange multiplier method is used to determine its optimal solution. In order to avoid the partial differential terms of the constraints as well as Intermediate variables P and Q are introduced in the optimization, and the optimization problem is rewritten into the following matrix form.

[0073]

[0074] stYTDP=I,QT1l×n=1n×n, Y=P, C=Q

[0075] in

[0076]

[0077] By calculating the stationary point of the augmented Lagrangian function, we can obtain the optimal solution that minimizes it. We get the matrices Y and P:

[0078]

[0079] P=(μDYYTD+μI)-1 ( DYλ1+λ3+μDY+μY )

[0080] The objective function is converted into a matrix form about the matrix C as follows:

[0081]

[0082] in

[0083]

[0084]

[0085] The matrices C and Q are obtained by calculating the stationary points of the augmented Lagrangian function:

[0086] C=(μQ-λ4)(μI-2J+2R)-1

[0087]

[0088] Step S104, the network adjacency matrix W is adjusted according to the calculated node community membership matrix C, and then the matrices Y and C are re-optimized to form an iterative optimization, and finally the node embedding vector is obtained according to the matrix Y, and the community discovery result is obtained according to the matrix C.

[0089] The network adjacency matrix is ​​adjusted according to the calculated community membership matrix as follows:

[0090]

[0091] According to the adjusted adjacency matrix, the node embedding matrix and community membership matrix are re-optimized, and the related auxiliary matrices H, F, J and R, as well as the Lagrangian function and Lagrangian multipliers λ1, λ2, λ3, λ4 are updated. When the Lagrangian function reaches the convergence condition, the update stops, and the node embedding and community division results are obtained. The specific process is as follows Figure 3 shown.

[0092] The community discovery method for the binary symbol network provided in this embodiment is used to perform node weighting and edge reconstruction in combination with the balanced structure and unbalanced structure in the network, and the node information and local network structure information in the network are retained while distinguishing the edges of different symbols in the network. In addition, the optimization of community information is added to the network embedding, so that the network edge weight is dynamically adjusted according to the optimization results during the iteration process, thereby realizing the cyclic optimization of the network edge weight and node embedding, accelerating the iterative process of node embedding, making the nodes in the same community closer, and the nodes in different communities farther away, making the network representation result more distinguishable; at the same time, the two independent steps of network embedding and community discovery are coupled together to make the community division result more accurate.

[0093] Example 2

[0094] Figure 4 FIG. 2 shows a schematic diagram of a community discovery device for a binary symbol network according to an embodiment of the present invention. Figure 4 As shown, the device includes: a network construction and secondary node set weighting module 401, a main node set network construction module 402, a node embedding and membership calculation module 403, and an optimization module 404.

[0095] The network construction and secondary node set weighting module 401 is used to establish the node and edge relationship data of the bipartite symbolic network based on the evaluation relationship data of users and products, wherein the user group is represented by the main node set in the network, and the products are represented by the secondary node set in the network. When the user evaluates the product positively, the edge is positive, and when the user evaluates the product negatively, the edge is negative; based on the positive and negative degrees and the butterfly structure of the network secondary node set, the weight of each node in the secondary node set is calculated.

[0096] In an optional manner, the network construction and sub-node set weighting module 401 is further used to: analyze each sub-node set in the bipartite symbolic network, count the number of positive edges and negative edges of the node, record them as the positive and negative degrees of the nodes, count the number of balanced structures and the number of unbalanced structures in the butterfly structure containing the node in the network, and calculate the weight of each node in the sub-node set based on the positive and negative degrees and the number of balanced and unbalanced structures.

[0097] The main node set network construction module 402 is used to calculate the edge weight values ​​of two nodes with a common secondary node (neighbor) in the main node set according to the network edge relationship and the secondary node weight; based on the edge weight values ​​of all nodes in the main node set, construct the main node set network adjacency matrix W.

[0098] In an optional manner, the main node set network construction module 402 is further used to: assume that there is a common secondary node k between node i and node j in the main node set, then the size of the edge corresponding to node k is the weight of node k, and the positive or negative edge is determined as follows: when nodes i and j have edges with the same sign as node k, the edge is positive, otherwise, it is negative; calculate the edges corresponding to the remaining common secondary nodes of node i and node j according to the above process; calculate the edges between node i and node j based on the edges corresponding to all common secondary nodes of node i and node j.

[0099] The node embedding and membership calculation module 403 calculates the node embedding matrix Y and the community membership matrix C according to the adjacency matrix of the main set node network.

[0100] The optimization module 404 is used to dynamically adjust the network adjacency matrix W according to the calculated node community membership matrix C, and then re-optimize the node embedding matrix Y and C to form an iterative optimization, and finally obtain the node embedding and community division results.

[0101] In an optional manner, the optimization module 404 is further used to: stop after the Lagrangian functions of the matrices Y and C reach a convergence condition, and finally obtain the embedding and community division results of the nodes.

[0102] Example 3

[0103] An embodiment of the present invention provides a non-volatile computer storage medium, wherein the computer storage medium stores at least one executable instruction, and the computer executable instruction can execute a community discovery method for a binary symbolic network in any of the above method embodiments.

[0104] The executable instructions may be specifically used to cause the processor to perform the following operations:

[0105] According to the evaluation relationship data of users and products, the node and edge relationship data of the binary signed network are established, in which the user group is represented by the main node set in the network, and the products are represented by the secondary node set in the network. When the user evaluates the product positively, the edge is positive, and when the user evaluates the product negatively, the edge is negative; based on the positive and negative degrees and butterfly structure of the network secondary node set, the weight of each node in the secondary node set is calculated;

[0106] For two nodes in the main node set that have a common secondary node, the edge weight values ​​of the two nodes are calculated according to the edge relationship of the network and the secondary node weight; based on the edge weight values ​​of all nodes in the main node set, the main node set network adjacency matrix W is constructed;

[0107] According to the adjacency matrix W of the main set node network, the node embedding matrix Y and community membership matrix C are calculated;

[0108] The network adjacency matrix W is adjusted according to the calculated node community membership matrix C, and then the node embedding matrices Y and C are re-optimized to form an iterative optimization, and finally the node embedding and community division results are obtained.

[0109] Example 4

[0110] Figure 5 The schematic diagram of the structure of the computing device embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0111] like Figure 5 As shown, the computing device may include:

[0112] Processor, Communications Interface, Memory, and Communications Bus.

[0113] Wherein: the processor, the communication interface, and the memory communicate with each other via a communication bus. The communication interface is used to communicate with other devices such as a client or other server network elements. The processor is used to execute a program, specifically, to execute the relevant steps in the above-mentioned embodiment of a community discovery method for a binary symbol network.

[0114] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A user community discovery method based on a bipartite symbolic network, characterized in that: The steps include: Step S101: Based on the evaluation relationship data of users and products, establish the node and edge relationship data of the bipartite symbol network, wherein the user group is represented by the main node set in the network, and the products are represented by the secondary node set in the network. When the user evaluates the product positively, the edge is positive, and when the user evaluates the product negatively, the edge is negative; based on the positive and negative degrees of the network secondary node set and the butterfly structure, calculate the weight of each node in the secondary node set; Step S102: for two nodes in the primary node set that have a common secondary node, calculate the edge weight values ​​of the two nodes according to the network edge relationship and the secondary node weights; Based on the edge weight values ​​of all nodes in the main node set, the main node set network adjacency matrix W is constructed; Step S103: Calculate the node embedding matrix Y and community membership matrix C according to the adjacency matrix W of the main set node network; Step S104: The network adjacency matrix W is adjusted according to the calculated node community membership matrix C, and then the node embedding matrix Y and the node community membership matrix C are re-optimized to form an iterative optimization, and finally the node embedding and community division results are obtained.

2. The community discovery method for a binary symbolic network according to claim 1, characterized in that: In step S101, the weight of each node in the secondary node set is calculated according to the positive and negative degrees of the network secondary node set and the number of butterfly structures: Each secondary node set in the bipartite signed network is analyzed, and the number of positive edges and negative edges of the node is counted, which is recorded as the positive and negative degree of the node. The number of balanced structures and the number of unbalanced structures in the butterfly structure containing the node in the network are counted. The weight of each node in the secondary node set is calculated based on the positive and negative degree and the number of balanced and unbalanced structures.

3. The community discovery method for a binary symbolic network according to claim 2, characterized in that: The weight z of each node k in the secondary node set k Specifically: in, represents the positive degree of node k, represents the negative degree of node k, U k represents the number of unbalanced structures containing node k, B k Represents the number of balanced structures containing node k.

4. The community discovery method for a binary symbolic network according to claim 1, characterized in that: The step S102 calculates the edge weight values ​​of the two nodes in the primary node set that have a common secondary node according to the edge relationship of the network and the secondary node weights. The specific process is as follows: Assume that there is a common secondary node k between node i and node j in the main node set, then the edge size corresponding to node k is the weight of node k, and the positive or negative edge is determined as follows: when nodes i and j have edges with the same sign as node k, the edge is positive, otherwise, it is negative; calculate the edges corresponding to the remaining common secondary nodes of node i and node j according to the above process; Based on the edges corresponding to all common subnodes of nodes i and j, calculate the edges between nodes i and j.

5. The community discovery method for a binary symbolic network according to claim 4, characterized in that: Calculate the edge W between node i and node j ij The specific formula is: Among them, A ik A represents the edge symbol between node i and node k. jk Represents the edge symbol between node j and node k.

6. The community discovery method for a binary symbolic network according to claim 1, characterized in that: The step S103 includes: The corresponding network adjacency matrix W is constructed according to the main set node network, and the node embedding matrix and community membership matrix are calculated using the augmented Lagrange multiplier method.

7. The community discovery method for a binary symbolic network according to claim 6, characterized in that: The node embedding matrix Y and the community membership matrix C are specifically: C=(μQ-λ4)(μI-2J+2R) -1 Among them, matrices F and H are auxiliary matrices related to community membership and network adjacency matrix W, J and R are auxiliary matrices related to node embedding, matrix D is the degree matrix of the network, matrices P and Q are intermediate variable matrices of Y and C, and their initial values ​​are the same; λ1, λ3, λ4 are Lagrange multipliers, and μ is the penalty coefficient.

8. The community discovery method for a binary symbolic network according to claim 1, characterized in that: The step S104 adjusts the network adjacency matrix W according to the calculated node community membership matrix C to: Among them, α is the setting parameter, c i represents the community membership of node i, c j represents the community membership of node j.

9. A user community discovery device based on a binary symbol network, characterized in that: include: The network construction and sub-node set weighting module is used to establish the node and edge relationship data of the binary symbol network based on the evaluation relationship data of users and products, where the user group is represented by the main node set in the network, and the products are represented by the sub-node set in the network. When the user evaluates the product positively, the edge is positive, and when the user evaluates the product negatively, the edge is negative; based on the positive and negative degrees of the network sub-node set and the butterfly structure, the weight of each node in the sub-node set is calculated; The main node set network construction module is used to calculate the edge weight value of two nodes with a common secondary node in the main node set according to the network edge relationship and the secondary node weight; Based on the edge weight values ​​of all nodes in the main node set, the main node set network adjacency matrix W is constructed; The node embedding and membership calculation module calculates the node embedding matrix Y and community membership matrix C according to the adjacency matrix W of the main set node network; The optimization module adjusts the network adjacency matrix W according to the calculated node community membership matrix C, and then re-optimizes the node embedding matrices Y and C to form an iterative optimization, and finally obtains the node embedding and community division results.

10. A computing device, characterized in that include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the community discovery method for a binary symbolic network according to any one of claims 1-8.

Images