A Key Personnel Analysis Method and System Based on Graph Embedding Relationship Graph

Abstract

This invention discloses a method and system for analyzing key personnel in a relationship graph based on graph embedding. The method includes the following steps: constructing a relationship graph based on social media data; analyzing each node in the relationship graph using a graph embedding algorithm to obtain the embedding vector of each node; generating key node seeds for the relationship graph based on pre-defined relevant indicators; and analyzing the key node seeds using a clustering algorithm based on the embedding vectors of each node to identify key personnel nodes. This invention fully utilizes the topological properties of the relationship graph, possessing learnability, and eliminates the need for manually setting parameter values ​​or defining degree gain calculation rules, thereby eliminating the adverse effects of unreasonable human rule settings on the results. Furthermore, by calculating the entire graph, it integrates the isomorphism and heterogeneity of nodes, resulting in more accurate key personnel analysis results.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph analysis technology, specifically to a method and system for analyzing key personnel in a relational graph based on graph embedding. Background Technology

[0002] A relationship graph is a knowledge graph constructed around the entities of "people" and the social, kinship, and emotional relationships between them. According to the "six degrees of separation" theory, in interpersonal communication, any two strangers can be connected through a maximum of five friends. To some extent, everyone in the world can be connected through their personal networks. Due to the complexity of the real world, the types of people and relationships involved in the construction of relationship graphs are increasing. In several subgraphs of a relationship graph, often only one or a few people play a major role. Especially in public opinion analysis, administrative management, risk control, and recommendation systems, the identification of key personnel plays a decisive role in business operations and has become an important technology for knowledge graph analysis and application.

[0003] Key personnel identification in relationship graphs relies heavily on manual qualitative analysis or simple static numerical calculations due to limited learning methods. For example, Chinese patent CN113032607A discloses a key personnel analysis method, which includes: "obtaining a member relationship graph, obtaining initial member weights, obtaining member interaction information, calculating and updating the member's full value based on the interaction information and initial full value, and if the sum of the weight differences between two adjacent values ​​corresponding to each node is less than a preset weight threshold, then the node with the largest updated weight is extracted as the target node." This approach has the following shortcomings: 1) The node information, interaction information values, and node weight update methods in the relationship graph are all manually set by rules, lacking learnability. 2) When adding or deleting nodes and relationships, or when performing cross-domain business migrations, manual intervention is required to provide corresponding business rules, lacking scalability. 3) The weight update of node personnel only includes local structural and personnel information, failing to utilize the global topological structure, resulting in low accuracy. These problems prevent intelligent key personnel analysis in relationship graphs and severely limit its application.

[0004] For example, Chinese patent CN 112269922 A discloses a method for identifying key figures in community public opinion based on network representation learning. The method includes: "Inputting a social network relationship graph into a community structure and structural hole node discovery model to obtain a community partition set and structural hole nodes; inputting the social network relationship graph and community partition set into a network embedding model containing social influence and community structure to obtain the social influence of nodes in the community network graph and node network embedding representation vectors; performing visualization analysis based on structural hole nodes, social influence, and network embedding representation vectors to obtain key figures in public opinion." This approach still has the following shortcomings: 1) The direct and indirect modularity gains of the relationship graph, until the target matrix of the network embedding vector is obtained through eigenvalue decomposition, are vectorized by rules throughout the process, still belonging to manual feature selection rather than adaptive learning. This method heavily relies on the rule definition of direct and indirect modularity gains. If the rule definition cannot reflect the network structure, the method will be greatly affected, reducing the accuracy of key figure discovery. Summary of the Invention

[0005] In view of the above technical problems, the purpose of this invention is to provide a method and system for analyzing key personnel in relational graphs based on graph embedding, which solves the problems that traditional key personnel mining methods lack scalability, or that the weight update of node personnel only includes local structural and personnel information, resulting in low accuracy, or that the accuracy rate is low due to the reliance on rules of direct and indirect module degree gain.

[0006] The present invention adopts the following technical solution:

[0007] A key personnel analysis method based on graph embedding relational graphs includes the following steps:

[0008] Building a relationship graph based on social media data;

[0009] A graph embedding algorithm is used to analyze each node in the relationship graph to obtain the embedding vector of each node;

[0010] Generate key node seeds for the character relationship graph based on pre-defined relevant indicators;

[0011] Based on the embedding vectors of each node, a clustering algorithm is used to analyze the key node seeds and identify key personnel nodes.

[0012] Optionally, the construction of the relationship graph based on social media data includes:

[0013] By mining the entities and relationships of people from news data throughout the entire cycle of the events that trigger public opinion, a relationship graph of people can be generated.

[0014] Optionally, the step of mining entities and relationships from news data throughout the entire cycle of the public opinion event to generate a relationship graph includes:

[0015] Web crawling technology is used to filter news reports and social media data published during a specified public opinion period on an online platform by keywords. This yields text and social media content related to the public opinion event in the news reports during the public opinion period, as well as the interaction relationships between entities. Text structuring technology is then used to generate corresponding relationship graphs.

[0016] Optionally, the step of using a graph embedding algorithm to analyze each node in the character relationship graph to obtain the embedding vector of each node includes:

[0017] For each node, a random walk method is used to obtain the nearest neighbor nodes, resulting in a set of neighbor nodes. The skip-gram model is used to train the set of neighbor nodes, and each neighbor node is used to predict the current node, so that the probability of the current node appearing is maximized. The neighbor nodes in the set of neighbor nodes are trained in turn to obtain the embedding vector of each node.

[0018] Optionally, generating key node seeds based on pre-defined relevant indicators for nodes includes:

[0019] A graph adjacency matrix is ​​generated based on preset relevant indicators, and the adjacency matrix is ​​subjected to feature decomposition to obtain eigenvalues ​​and eigenvectors.

[0020] Obtain the eigenvector corresponding to the largest eigenvalue among the eigenvalues ​​of each node. The centrality of the i-th node is the i-th element in the eigenvector corresponding to the largest eigenvalue. Generate key node seeds based on the centrality of each node.

[0021] Optionally, the step of analyzing the embedding vectors of each node using a clustering algorithm based on the key node seeds to identify key personnel nodes includes:

[0022] Based on the key node seeds, a clustering algorithm is used to classify the embedding vectors to obtain several clustering categories;

[0023] Calculate c for each cluster category i The calculated cluster centers are used as the updated cluster centers, and the updated cluster centers are used as key personnel nodes.

[0024] Optionally, the clustering algorithm is used to classify the embedding vectors to obtain several cluster categories, including:

[0025] Using the key node seed as the initial cluster center, calculate the distance from each embedding vector to each initial cluster center, and obtain the initial cluster center with the shortest distance to each embedding vector. Then, classify each node into the cluster category to which the initial cluster center with the shortest distance to it belongs.

[0026] A key personnel analysis system based on graph embedding relational graphs includes:

[0027] Graph construction unit, used to build relationship graphs based on social media data;

[0028] The graph analysis unit is used to analyze each node in the character relationship graph using a graph embedding algorithm to obtain the embedding vector of each node.

[0029] A key node seed generation unit is used to generate key node seeds for the character relationship graph based on pre-defined relevant indicators.

[0030] The identification unit is used to analyze the key node seeds using a clustering algorithm based on the embedding vectors of each node, and identify key personnel nodes.

[0031] An electronic device includes: at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the graph embedding-based relationship graph key personnel analysis method.

[0032] A computer storage medium storing a computer program, which, when executed by a processor, implements the graph embedding-based key personnel analysis method.

[0033] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0034] This invention constructs a relationship graph based on social media data and uses a graph embedding algorithm to analyze each node in the graph, obtaining the embedding vector for each node. This fully utilizes the topological properties of the relationship graph and possesses learnability. The network embedding representation and node vectorization are determined by random walk control and corresponding machine learning methods, respectively, eliminating the need for manually setting parameter values ​​or defining degree gain calculation rules, thus avoiding the adverse effects of unreasonable human rule settings on the results. Furthermore, this invention constructs a relationship graph based on social media data, relying only on the network topology. When adding or deleting nodes and relationships, or migrating across domains, the network can be quickly trained without additional knowledge injection. Key node seeds for the relationship graph are generated based on pre-defined relevant indicators. Based on the embedding vectors of each node, a clustering algorithm is used to analyze the key node seeds, identifying key personnel nodes. During the identification of key personnel nodes, calculations are performed on the entire graph, integrating the isomorphism and heterogeneity of nodes, resulting in more accurate key personnel analysis results.

[0035] Furthermore, a random walk method is used to obtain nearest neighbor nodes, resulting in a set of neighboring nodes. Each neighboring node is used to predict the current node, maximizing the probability of the current node appearing. Each neighboring node in the set of neighboring nodes is trained sequentially to obtain the embedding vector of each node. A graph embedding method based on random walk is used for analysis, eliminating the need for manually setting parameter values ​​or specifying the degree gain calculation rules, thereby further improving the high accuracy of identifying key personnel nodes. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating a key personnel analysis method based on graph embedding relationship graphs, provided in an embodiment of the present invention.

[0037] Figure 2 This is a schematic diagram of a random walk sampling of nearest neighbor nodes provided in an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of a key personnel analysis system based on graph embedding relationship graphs, provided in an embodiment of the present invention.

[0039] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0040] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments:

[0041] Example 1:

[0042] The following is an explanation of the technical terms used in this invention:

[0043] Graph embedding (also called network embedding) is a process of mapping graph data (usually high-dimensional dense matrices) to low-density vectors, which can effectively solve the problem that graph data is difficult to efficiently input into machine learning algorithms.

[0044] An adjacency matrix is ​​a matrix that represents the adjacency relationships between vertices. The logical structure of an adjacency matrix consists of two parts: sets V and E, where V represents vertices and E represents edges. Therefore, a one-dimensional array stores all vertex data in the graph, and a two-dimensional array stores the data of the relationships (edges or arcs) between vertices; this two-dimensional array is called the adjacency matrix.

[0045] Centrality is used to measure the importance of a node in a network. Centrality can be defined for a single node or a group of nodes. Eigenvector centrality combines the centrality of a node's neighbors as its centrality.

[0046] The embedding vector of a node refers to the vector representation of a vertex in the network obtained through the connection relationship in the network structure. It is used as a basic feature for tasks such as clustering and classification.

[0047] Please refer to Figure 1 As shown, Figure 1 This invention presents a key personnel analysis method based on graph embedding relationship graphs, comprising the following steps:

[0048] Step S1: Construct a relationship graph based on social media data;

[0049] Specifically, the construction of a relationship graph based on social media data includes:

[0050] By mining the entities and relationships of people from news data throughout the entire cycle of the events that trigger public opinion, a relationship graph of people can be generated.

[0051] In practice, web crawling technology can be used to filter news reports and social media data published during a specified public opinion period on a web platform by keywords, so as to obtain the text and social media content related to the public opinion event in the news reports during the public opinion period, as well as the interaction relationship between entities, and generate the corresponding relationship map using text structuring technology.

[0052] In practice, the construction of a person-relationship graph can be achieved through knowledge triple extraction technology, dynamic knowledge graph generation technology that evolves over time, relationship mining technology, and domain knowledge-based transfer learning technology.

[0053] Step S2: Analyze each node in the relationship graph using a graph embedding algorithm to obtain the embedding vector of each node;

[0054] Optionally, step S2 includes:

[0055] For each node, a random walk method is used to find its nearest neighbors, resulting in a set of neighboring nodes; for details, please refer to [link to documentation]. Figure 2 As shown, Figure 2 This diagram illustrates a random walk sampling method for nearest neighbor nodes according to an embodiment of the present invention; wherein, given the current vertex v, the probability of going to vertex x is:

[0056]

[0057] Where, π vx Z represents the unnormalized transition probability between vertices, which is the probability that a random walk will reach node v after passing through node t, and then reach node x; Z is the normalization constant.

[0058] Specifically, to control the direction of the random walk and express our preferences, suppose the current random walk passes through node t and reaches node v. The probability of walking to node x at this point is π. vx Satisfy the following formula:

[0059] ω vx =α pq (t, x)·ω vx ;ω vx Here, p is the edge weight, q is the distance parameter, and d is the edge weight. tx The shortest path distance; coefficient α pq(t,x) Satisfy the following formula:

[0060]

[0061] If q > 1, the random walk tends to visit nodes that are close to the previous node; if q < 1, the random walk tends to visit nodes that are far away from the previous node.

[0062] In the above implementation process, the present invention is based on a vectorized method of random walk, which is different from the non-vectorized method of updating the value of interactive information and node weights in Chinese patent CN113032607 A, and also different from the rule-based method of modularity gain in Chinese patent CN112269922 A. It has learnability and adaptability.

[0063] Then, the skip-gram model is used to train the neighboring node set, and each neighboring node is used to predict the current node, so that the probability of the current node appearing is maximized. The neighboring nodes in the neighboring node set are trained in turn to obtain the embedding vector of each node.

[0064] In the above implementation process, a relationship graph of individuals is constructed based on social media data. Graph embedding algorithms are used to analyze each node in the relationship graph. For example, by mining the entities and relationships of individuals from news data throughout the entire cycle of a public opinion event, a relationship graph of individuals is generated. A graph embedding machine learning method based on random walks is used to analyze the graph and obtain node vectors. The entire graph is directly vectorized, resulting in more comprehensive acquisition of feature information. By calculating the entire graph, the isomorphism and heterogeneity of nodes are combined, making the analysis results of key personnel more accurate.

[0065] In practical implementation, when using the technique of predicting the current node using neighboring nodes, derivative methods of word2vec such as CBOW, as well as training optimization methods based on negative sampling or Huffman trees, can also be used to help predict the current node.

[0066] Specifically, we obtain the set of neighboring nodes of the current node, denoted as N. S (u) First, train the skip-gram model on each neighboring node, then use the neighboring nodes to predict the current node, maximizing the probability of the current node appearing. The maximum probability is... Then, each neighboring node is trained sequentially to obtain the embedding vector.

[0067] Step S3: Generate key node seeds for the character relationship graph based on pre-defined relevant indicators;

[0068] Optionally, step S3 includes:

[0069] A graph adjacency matrix is ​​generated based on preset relevant indicators, and the adjacency matrix is ​​subjected to feature decomposition to obtain eigenvalues ​​and eigenvectors.

[0070] Obtain the eigenvector corresponding to the largest eigenvalue among the eigenvalues ​​of each node. The centrality of the i-th node is the i-th element in the eigenvector corresponding to the largest eigenvalue. Generate key node seeds based on the centrality of each node.

[0071] Specifically, a graph adjacency matrix A can be generated based on relevant indicators such as network density, reachability, clustering coefficient, and centrality measure. The adjacency matrix is ​​then decomposed into eigenvalues, i.e., Ax = λx, to obtain eigenvalues ​​and eigenvectors. In the eigenvector corresponding to the largest eigenvalue, the centrality of the i-th node is equal to the i-th element in the eigenvector.

[0072] In specific implementation, small sample annotation methods such as manual annotation, pre-trained model annotation, and remote unsupervised annotation can also be used to perform annotation first. The centrality can also include importance metrics such as degree centrality, betweenness centrality, and compact centrality.

[0073] Step S4: Based on the embedding vectors of each node, a clustering algorithm is used to analyze the key node seeds and identify key personnel nodes.

[0074] Optionally, step S4 specifically includes:

[0075] The key node seeds are used as initial cluster centers, and the initial cluster centers are α1, α2, ..., α3 respectively. k The initial cluster centers form an initial cluster center set α = α1, α2, ... α k ;

[0076] Clustering algorithms are used to classify the embedding vectors, resulting in several cluster categories; c is calculated for each cluster category. i The cluster centers are calculated and used as key personnel nodes.

[0077] In the above implementation process, the present invention directly classifies the vectorization method of graph embedding, without relying on strong assumptions, which is different from the community structure and social influence assumptions of Chinese patent CN 112269922 A, and has universality.

[0078] The steps for classifying the embedding vectors using a clustering algorithm include:

[0079] Calculate each embedding vector x i Calculate the distance to each initial cluster center, and obtain the initial cluster center α that has the shortest distance to each embedding vector. i Each node is assigned to the initial cluster center α that is closest to it. i Cluster category c i Where 1≤i≤k, and i and k are both natural numbers;

[0080] Specifically, the method used to calculate the cluster centers is as follows:

[0081]

[0082] Among them, |c i | indicates the number of nodes in the cluster category. The algorithm for cluster centering is iterated repeatedly until a certain termination condition is reached. The class of the seed node of the key node is called the key node class.

[0083] In this embodiment, machine learning methods are used to analyze vectorized nodes and identify key nodes. Specifically, the algorithm used to identify key personnel nodes can be a supervised or semi-supervised machine learning classification algorithm.

[0084] Please refer to Figure 3 As shown, Figure 3 This invention illustrates a key personnel analysis system based on graph embedding relationship graphs, comprising:

[0085] Graph construction unit, used to build relationship graphs based on social media data;

[0086] The graph analysis unit is used to analyze each node in the character relationship graph using a graph embedding algorithm to obtain the embedding vector of each node.

[0087] A key node seed generation unit is used to generate key node seeds for the character relationship graph based on pre-defined relevant indicators.

[0088] The identification unit is used to analyze the key node seeds using a clustering algorithm based on the embedding vectors of each node, and identify key personnel nodes.

[0089] Example 3:

[0090] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. In this application, it can be... Figure 4 The schematic diagram shown illustrates an electronic device 100 for implementing the present invention's graph embedding-based relationship graph key personnel analysis method according to embodiments of this application.

[0091] Figure 4 shows a schematic diagram of an electronic device 100. The electronic device 100 includes one or more processors 102 and one or more storage devices 104. These components are interconnected via a bus system and / or other forms of connection mechanisms (not shown). It should be noted that... Figure 4 The components and structure of the electronic device 100 shown are merely exemplary and not limiting; the electronic device may have, as needed. Figure 4 The components shown may also have Figure 4 Other components and structures not shown.

[0092] The processor 102 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 100 to perform desired functions.

[0093] The storage device 104 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 102 may execute the program instructions to implement the functions (implemented by the processor) in the embodiments of this application described below, and / or other desired functions. Various application programs and various data may also be stored in the computer-readable storage medium, such as various data used and / or generated by the application programs.

[0094] This invention also provides a computer storage medium on which a computer program is stored. If the method of this invention is implemented as a software functional unit and sold or used as an independent product, it can be stored in this computer storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer storage medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the contents of the computer storage medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer storage medium does not include electrical carrier signals and telecommunication signals.

[0095] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

Claims

1. A method for analyzing key personnel in a relational graph based on graph embedding, characterized in that, Includes the following steps: Building a relationship graph based on social media data; A graph embedding algorithm is used to analyze each node in the relationship graph to obtain the embedding vector of each node; The graph embedding algorithm is used to analyze each node in the character relationship graph to obtain the embedding vector of each node, including: For each node, a random walk method is used to obtain the nearest neighbors, resulting in a set of neighboring nodes; The skip-gram model is used to train the set of neighboring nodes, and each neighboring node is used to predict the current node so that the probability of the current node appearing is maximized. The neighboring nodes in the set of neighboring nodes are trained in turn to obtain the embedding vector of each node. Generate key node seeds for the character relationship graph based on pre-defined relevant indicators; the generation of key node seeds based on pre-defined relevant indicators for nodes includes: A graph adjacency matrix is ​​generated based on preset relevant indicators, and the adjacency matrix is ​​subjected to feature decomposition to obtain eigenvalues ​​and eigenvectors. Obtain the eigenvector corresponding to the largest eigenvalue among the eigenvalues ​​of each node. The centrality of the i-th node is the i-th element in the eigenvector corresponding to the largest eigenvalue. Generate key node seeds based on the centrality of each node. Based on the embedding vectors of each node, a clustering algorithm is used to analyze the key node seeds and identify key personnel nodes. This process includes: Based on the key node seeds, a clustering algorithm is used to classify the embedding vectors to obtain several clustering categories; Calculate c for each cluster category i The calculated cluster centers are used as the updated cluster centers, and the updated cluster centers are used as key personnel nodes.

2. The method for analyzing key personnel in a relational graph based on graph embedding according to claim 1, characterized in that, The construction of a relationship graph based on social media data includes: By mining the entities and relationships of people from news data throughout the entire cycle of the events that trigger public opinion, a relationship graph of people can be generated.

3. The method for analyzing key personnel in a relational graph based on graph embedding according to claim 2, characterized in that, The process of mining individuals and relationships from news data throughout the entire cycle of a public opinion event to generate a relationship graph includes: Web crawling technology is used to filter news reports and social media data published during a specified public opinion period on an online platform by keywords. This yields text and social media content related to the public opinion event in the news reports during the public opinion period, as well as the interaction relationships between entities. Text structuring technology is then used to generate corresponding relationship graphs.

4. The method for analyzing key personnel in a relational graph based on graph embedding according to claim 1, characterized in that, The clustering algorithm is used to classify the embedding vectors to obtain several cluster categories, including: Using the key node seed as the initial cluster center, calculate the distance from each embedding vector to each initial cluster center, and obtain the initial cluster center with the shortest distance to each embedding vector. Then, classify each node into the cluster category to which the initial cluster center with the shortest distance to it belongs.

5. A key personnel analysis system based on graph embedding relationship graphs, characterized in that, include: Graph construction unit, used to build relationship graphs based on social media data; The graph analysis unit is used to analyze each node in the character relationship graph using a graph embedding algorithm to obtain the embedding vector of each node. The step of using a graph embedding algorithm to analyze each node in the character relationship graph and obtain the embedding vector of each node includes: for each node, using a random walk method to obtain the nearest neighbor nodes and obtain a set of neighbor nodes; using a skip-gram model to train the set of neighbor nodes, using each neighbor node to predict the current node so that the probability of the current node appearing is maximized, and sequentially training each neighbor node in the set of neighbor nodes to obtain the embedding vector of each node. A key node seed generation unit is used to generate key node seeds for the character relationship graph based on pre-defined relevant indicators. The step of generating key node seeds for nodes based on pre-defined relevant indicators includes: generating a graph adjacency matrix based on pre-defined relevant indicators; performing eigenvalue decomposition on the adjacency matrix to obtain eigenvalues ​​and eigenvectors; obtaining the eigenvector corresponding to the largest eigenvalue among the eigenvalues ​​of each node, wherein the centrality of the i-th node is the i-th element in the eigenvector corresponding to the largest eigenvalue; and generating key node seeds based on the centrality of each node. The identification unit is used to analyze the key node seeds using a clustering algorithm based on the embedding vectors of each node to identify key personnel nodes. The step of analyzing the embedding vectors of each node using a clustering algorithm based on the key node seeds to identify key personnel nodes includes: classifying the embedding vectors using a clustering algorithm based on the key node seeds to obtain several cluster categories; and calculating c for each cluster category. i The calculated cluster centers are used as the updated cluster centers, and the updated cluster centers are used as key personnel nodes.

6. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the graph embedding-based key personnel analysis method for relational graphs according to any one of claims 1-4.

7. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the key personnel analysis method based on graph embedding relationship graphs as described in any one of claims 1-4.

Images