Social network key node identification method and system based on discrete attractor search

Abstract

The invention relates to the technical field of information processing, and provides a social network key node identification method and system based on discrete attractor search. The social network key node identification method based on discrete attractor search comprises the following steps: constructing a social network by taking users as nodes and taking false information propagation paths among the users as edges; based on the social network, a graph neural network encoder is adopted to obtain a user node embedding vector; introducing a binary gating mechanism to discretize continuous user node embedded vectors into a user node binary state; based on the discretized user node binary state, using attractor search to iteratively update the user node state, when iteration stops, obtaining all attractors and attraction basin information thereof, and mapping the attractors and the attraction basin information thereof back to the user nodes in the original social network; key scores of all user nodes are calculated, and M user nodes with high key scores are selected as key nodes. The search efficiency of the key nodes in the social network is improved.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a method and system for identifying key nodes in social networks based on discrete attractor search. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In the field of combating misinformation on social networks, accurately identifying key nodes in the dissemination chain of misinformation is a core technological step in blocking its spread and reducing the harm to public opinion. Currently, social networks, with their massive user base, rapid information dissemination, and complex interactive relationships, have become the main carriers of misinformation. Misinformation can quickly infiltrate different circles through user forwarding, commenting, and community sharing. The failure or control of key nodes that play a decisive role in the spread of misinformation (such as the initial account for the misinformation, highly active forwarding nodes, and cross-community dissemination hubs) can directly sever the dissemination chain, curb the global spread of misinformation, and is of great significance for maintaining public opinion order on social networks and protecting public information security.

[0004] However, existing technical solutions for identifying key nodes of misinformation on social networks still have significant limitations and are difficult to meet the needs of precise and real-time governance: 1. Methods based on scoring rules, such as heuristic metrics like degree centrality, proximity centrality, and betweenness centrality, assess importance solely through local connectivity features of nodes (e.g., number of friends, distance from other nodes). In social networks, some key nodes for spreading misinformation (e.g., core sharers within niche interest communities) may have low local connectivity, but they can drive misinformation to specific highly sensitive circles. These nodes are easily missed by such methods. Furthermore, these methods ignore the global topology of social networks (e.g., relationships between communities, hierarchical paths of information dissemination), failing to accurately pinpoint nodes that control the overall dissemination of misinformation, resulting in insufficiently targeted governance measures.

[0005] 2. Methods based on influence propagation models, such as PageRank and its variants, iteratively calculate the influence weight of nodes in information propagation. However, social networks often have hundreds of millions of users, and the spread of misinformation is dynamic (such as a large number of users forwarding information in a short period of time due to sudden trending events). These methods are computationally intensive and have long iteration cycles, making them difficult to adapt to the real-time analysis needs of large-scale social networks. They also fail to identify key nodes in sudden misinformation propagation events in a timely manner, easily missing the best opportunity for governance.

[0006] 3. While graph neural network-based methods can integrate social network topology (such as user follow relationships and community affiliation) with node attributes (such as user activity, frequency of forwarding misinformation, and historical speech characteristics) and learn node representations through neighborhood aggregation mechanisms to predict key nodes, they suffer from two major drawbacks: First, the oversmoothing problem—as the network depth increases, the representations of different nodes tend to be similar, making it impossible to effectively distinguish between core nodes and ordinary participating nodes in the misinformation propagation chain; second, poor dynamic adaptability—user relationships and information propagation paths in social networks often dynamically adjust with hot events and changes in user behavior. Such methods struggle to quickly update node representations and re-identify key nodes, resulting in a significant decrease in the timeliness and accuracy of key node identification, making them unable to adapt to the dynamic changes in the spread of misinformation on social networks. Summary of the Invention

[0007] To address the technical problems mentioned above, this invention provides a method and system for identifying key nodes in social networks based on discrete attractor search. This invention maps network nodes to the state space of a discrete dynamical system, identifies key nodes with stable attractor properties, and improves the search efficiency for key nodes in social networks.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of the present invention provides a method for identifying key nodes in a social network based on discrete attractor search.

[0009] A method for identifying key nodes in a social network based on discrete attractor search includes: A social network is constructed using users as nodes and the paths through which false information spreads between users as edges. Based on social networks, a graph neural network encoder is used to obtain user node embedding vectors; A binary gating mechanism is introduced to discretize the continuous user node embedding vector into a binary state of the user node; based on the discretized binary state of the user node, attractor search is used to iteratively update the state of the user node. When the iteration stops, all attractors and their attractor basin information are obtained and mapped back to the user nodes in the original social network. Calculate the criticality score of all user nodes, and select the M user nodes with high criticality scores as critical nodes.

[0010] Furthermore, the graph neural network encoder is constructed by using a graph convolutional neural network based on neighborhood aggregation as the basic architecture, introducing skip connections and attention mechanisms.

[0011] Furthermore, based on social networks, a graph neural network encoder is used to obtain user node embedding vectors; the method includes: using a graph neural network encoder, calculating a user node embedding matrix based on the user node feature matrix and the false information propagation path attribute feature matrix; the user node embedding matrix includes several user node embedding vectors.

[0012] Furthermore, a binary gating mechanism is introduced to discretize the continuous user node embedding vectors into binary user node states; this is expressed by the following formula:

[0013] in, Represents the binary state of the discretized user node; This indicates a binary gating mechanism; Represents the user node embedding vector; when , In other cases, ; The weight matrix is ​​trainable. Indicates bias; This represents the input signal transmitted from the neuron node.

[0014] Furthermore, attractor search is used to iteratively update the user node state. When the iteration stops, all attractors and their attractor basin information are obtained and mapped back to the user nodes in the original social network. The method includes: randomly generating a set of initial states. ; Preset state transition rules, current state of user node, calculate the new state of user node in the next moment; when the system reaches a stable state or reaches the maximum number of iterations, the iteration stops, and all attractors and their attractor basin information are obtained. The lowest point of attractor potential energy corresponds to the user node of the original social network.

[0015] Furthermore, the calculation of the criticality score for all user nodes is expressed using the following formula:

[0016] in, This represents the key score of the user node. It is the total number of attractors. For nodes The size of the attraction basin of the attractor.

[0017] Furthermore, the continuous user node embedding vectors are discretized into binary user node states; based on the discretized binary user node states, attractor search is used to iteratively update the user node states; the method includes: processing the continuous-time dynamic network through a sliding window mechanism, discretizing the continuous user node embedding vectors in each time window into binary user node states, applying attractor search in each time window, and tracking the evolution trajectory of the attractors to identify key nodes in different time periods.

[0018] Furthermore, during the training of the model, a contrastive learning strategy is adopted to maximize the similarity of key nodes and minimize the similarity of non-key nodes by constructing sample pairs of key nodes and non-key nodes.

[0019] A second aspect of the present invention provides a social network key node identification system based on discrete attractor search.

[0020] A key node identification system for social networks based on discrete attractor search includes: The social network building module is configured to construct a social network with users as nodes and the paths of false information propagation between users as edges. The encoder module is configured to: obtain user node embedding vectors using a graph neural network encoder based on a social network; The discrete attractor search module is configured to: introduce a binary gating mechanism to discretize the continuous user node embedding vector into a binary state of the user node; based on the discretized binary state of the user node, use attractor search to iteratively update the state of the user node; when the iteration stops, obtain all attractors and their attractor basin information, and map them back to the user nodes in the original social network. The critical node output module is configured to calculate the criticality score of all user nodes and select the M user nodes with high criticality scores as critical nodes.

[0021] A third aspect of the present invention provides a computer device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the method for identifying key nodes in a social network based on discrete attractor search as described in the first aspect above.

[0022] Compared with the prior art, the beneficial effects of the present invention are: This invention transforms the critical node identification problem into an attractor discovery problem in discrete dynamical systems using discrete attractor search technology. This process can be achieved through local iterative computation without the need for global optimization. Specifically, the time complexity of the attractor search is O(log n). ,in The number of edges is reduced, significantly lowering the computational burden and making the method applicable to ultra-large-scale networks (such as social networks or transportation networks with millions of nodes).

[0023] The discrete attractor search technique employed in this invention has the inherent characteristic of adapting to dynamic changes. When local changes occur in the network (such as the addition / deletion of a small number of nodes or edges), it is only necessary to re-perform the attractor search on the nodes in the affected region, without having to retrain the entire model.

[0024] This invention avoids reliance on deep GNNs by introducing discrete attractor search. Node embeddings are obtained only through a shallow GNN encoder (typically 2-3 layers), and key nodes are then identified through discrete dynamical system transformation and attractor search. Because discrete dynamical systems maintain state diversity, the node representations corresponding to different attractors differ significantly, thus effectively overcoming the oversmoothing problem. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0026] Figure 1 This is a flowchart illustrating a method for identifying key nodes in a social network based on discrete attractor search, as shown in an embodiment of the present invention. Figure 2 This is a structural diagram of a social network key node identification system based on discrete attractor search, as shown in an embodiment of the present invention. Figure 3 This is a structural diagram of a computer device shown in an embodiment of the present invention. Detailed Implementation

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0030] To facilitate understanding of the technical solutions of this invention, some technical terms involved in this invention will be introduced below.

[0031] Discrete attractors are a class of entities in discrete-time dynamical systems that can attract surrounding states to themselves and remain stable after perturbation. They are a core concept for describing the long-term evolution trend of a system.

[0032] Graph Convolutional Neural Networks (GNNs) are a deep learning-based algorithm framework for processing graph-structured data. By extracting features from nodes, edges, and the graph as a whole, they perform tasks such as classification, prediction, and generation. This framework can transform non-Euclidean graph data into a normalized representation and can be applied to various fields such as social networks, transportation networks, and molecular structures.

[0033] As introduced in the background section, identifying key nodes in the spread of misinformation on social networks (such as initial accounts and highly active forwarding nodes) is crucial for preventing the spread of misinformation. However, existing technologies have limitations. Methods based on scoring rules (such as degree centrality) only consider local node features, easily overlooking key nodes in niche communities and ignoring the global topology. Methods based on influence propagation models (such as PageRank) are computationally intensive and time-consuming, making them difficult to adapt to the real-time needs of large-scale social networks. While graph neural network-based methods integrate topology and node attributes, they suffer from over-smoothing and poor dynamic adaptability, making it difficult to meet the requirements of precise and real-time governance. Therefore, this invention provides a method and system for identifying key nodes in social networks based on discrete attractor search. The following examples illustrate the solution in detail.

[0034] Figure 1 This is a flowchart illustrating a method for identifying key nodes in a social network based on discrete attractor search, as shown in an embodiment of the present invention; see also... Figure 1 The method includes: A social network is constructed using users as nodes and the paths through which false information spreads between users as edges. Based on social networks, a graph neural network encoder is used to obtain user node embedding vectors; A binary gating mechanism is introduced to discretize the continuous user node embedding vector into a binary state of the user node; based on the discretized binary state of the user node, attractor search is used to iteratively update the state of the user node. When the iteration stops, all attractors and their attractor basin information are obtained and mapped back to the user nodes in the original social network. Calculate the criticality score of all user nodes, and select the M user nodes with high criticality scores as critical nodes.

[0035] This invention proposes a key node identification method based on the combination of discrete attractor search technology and graph neural network. By mapping network nodes to the state space of discrete dynamic system, key nodes with stable attractor properties are identified. It aims to solve the problems of oversmoothing, low computational efficiency and poor adaptability to dynamic networks in existing key node identification methods for fake information in social networks, and achieve more accurate, efficient and robust key node identification.

[0036] The key node identification method based on the combination of discrete attractor search technology and graph neural network described in this embodiment is described in detail below. The method includes: A graph convolutional neural network (GNN) based on neighborhood aggregation is used as the basic architecture. Skip connections and attention mechanisms are introduced to avoid oversmoothing issues, and a graph neural network encoder is constructed. Specifically, the user node feature matrix of the social network is input into the input layer. The feature matrix E representing the propagation path of false information is obtained and preprocessed. A two-layer graph attention network (GAT) is used to obtain the user node embedding matrix. ,in For the number of user nodes, For the embedded dimension.

[0037] Specifically, in the first layer of the graph attention network, based on the user node feature matrix... and the first weight matrix The product of these components yields the feature matrix of the first user node. Based on the feature matrix E of the false information propagation path and the second weight matrix Calculate the first false information propagation path attribute feature matrix by multiplying the first user node feature matrix; Obtain the feature vectors of user a and user b; calculate the original attention score based on the feature vector of user a, the first false information propagation path attribute feature matrix between user a and user b, and the feature vector of user b, and then normalize it. Introduce the normalized original attention score based on the first user node feature matrix. The features of the first false information propagation path attribute feature matrix are fused using neighborhood aggregation and skip connections. This aggregation of neighborhood features yields the output of the first-layer graph attention network. Similarly, the output of the second-layer graph attention network is obtained. The user node feature matrix is ​​then used... The outputs of the first-layer graph attention network and the second-layer graph attention network are fused to obtain the user node embedding matrix. ,in For the number of user nodes, For the embedded dimension.

[0038]

[0039]

[0040] in, It is a graph neural network embedding matrix The The embedding vector of the i-th user node is derived from the i-th user node. User nodes The transformation yields, For attention weights, For trainable weight matrix, This is the activation function.

[0041] In this embodiment, the nodes of the social network are user nodes, and the edges are the paths for the spread of false information.

[0042] After obtaining the user node embedding vector, the user node embedding vector output by the graph neural network encoder is converted into the state of the discrete dynamic system. Specifically, by introducing a binary gating mechanism, the continuous user node embedding vector is discretized into binary states (0 or 1) to form a discrete state space.

[0043]

[0044] in, Represents the binary state of the discretized user node; This indicates a binary gating mechanism; Represents the user node embedding vector; when , In other cases, . This is a trainable weight matrix. In the neuron node model, neurons receive input signals transmitted from other neuron nodes. Each neuron contains weights. and bias The weight The importance of controlling the input signal, and It controls how easily neurons are activated.

[0045] After obtaining the discretized binary state of the user node, an attractor search is used for iteration to obtain all attractors and their attractor basin information when the iteration stops. The specific process is as follows: Initialization: Randomly generate a set of initial states. ; Iterative Update: Based on the state transition rule (T), a predefined function calculates the new state of the user node at the next time step, according to the current state of the user node. .

[0046] Termination condition: When the system reaches a steady state (i.e., (or until the maximum number of iterations is reached, record all attractors and their attractor bowl information.)

[0047] In discrete state space, the attractor search algorithm is used to identify groups of states with stable attracting properties (i.e., attractors) and map them back to user nodes in the original social network (the lowest point of the attractor potential corresponds to a user node in the original social network). Key nodes are identified based on their frequency of participation in the attractor and the size of the attractor basin (i.e., the number of initial states that converge to the attractor).

[0048] Furthermore, this invention can process continuous-time dynamic networks through a sliding window mechanism, apply attractor search within each time window, and track the evolution trajectory of the attractors, thereby identifying key nodes in different time periods.

[0049] Traditional key node identification methods based on global optimization (such as betweenness centrality) often have a time complexity of up to [missing information]. ,in Due to the large number of nodes, this method is difficult to apply to large-scale networks. Even GNN-based methods require handling all nodes and edges in the entire network during training and inference. This invention transforms the key node identification problem into an attractor discovery problem in discrete dynamical systems using discrete attractor search technology. This process can be achieved through local iterative computation without global optimization. Specifically, the time complexity of attractor search is O(log n). ,in The number of edges is reduced, significantly lowering the computational burden and making the method applicable to ultra-large-scale networks (such as social networks or transportation networks with millions of nodes).

[0050] After mapping all attractors and their attractor basin information back to the original social network nodes, the criticality scores of all user nodes are calculated, and the criticality scores of user nodes are sorted from largest to smallest. The top M critical scores are selected and the corresponding user node list is output, i.e., the critical user node list.

[0051] In this embodiment, the criticality score of a user node is represented by the following formula:

[0052] in, This represents the key score of the user node. It is the total number of attractors. For nodes The size of the attraction basin of the attractor.

[0053] In this embodiment, a contrastive learning strategy can be used to train the entire model without requiring a large amount of labeled data. By constructing positive and negative sample pairs (key nodes and non-key nodes), the similarity of positive sample pairs is maximized, and the similarity of negative sample pairs is minimized. This invention achieves unsupervised or semi-supervised training through contrastive learning, reducing the dependence on labeled data.

[0054] Networks are often dynamically evolving, with nodes and edges changing over time. Traditional GNN methods, when faced with dynamic networks, typically require retraining the entire model or performing complex incremental learning, resulting in high computational costs and poor real-time performance. The discrete attractor search technique employed in this invention naturally adapts to dynamic changes. When local changes occur in the network (such as the addition / deletion of a small number of nodes or edges), only the nodes in the affected region need to be re-searched for attractors, without retraining the entire model. This is due to the locality of the attractor search process—local changes in the network only affect a limited range of dynamic system behavior.

[0055] Traditional GNN methods suffer from oversmoothing (i.e., the representations of all nodes tend to become similar as the network depth increases), making it difficult to capture multi-layered topological structures. While some improvements (such as skip connections and attention mechanisms) partially alleviate this problem, they do not fundamentally solve it. This invention avoids reliance on deep GNNs by introducing discrete attractor search. Node embeddings are obtained only through a shallow GNN encoder (typically 2-3 layers), and key nodes are then identified through discrete dynamical system transformation and attractor search. Because discrete dynamical systems maintain state diversity, the node representations corresponding to different attractors differ significantly, thus effectively overcoming the oversmoothing problem.

[0056] The above combination Figure 1 The method for identifying key nodes in a social network based on discrete attractor search provided in the embodiments of the present invention has been described in detail. Next, the system for identifying key nodes in a social network based on discrete attractor search provided in the embodiments of the present invention will be described in conjunction with the accompanying drawings.

[0057] Figure 2 This is a schematic diagram illustrating the structure of a social network key node identification system based on discrete attractor search, as shown in an embodiment of the present invention. Figure 2 The system described in this invention includes: The social network building module is configured to construct a social network with users as nodes and the paths of false information propagation between users as edges. The encoder module is configured to: obtain user node embedding vectors using a graph neural network encoder based on a social network; The discrete attractor search module is configured to: introduce a binary gating mechanism to discretize the continuous user node embedding vector into a binary state of the user node; based on the discretized binary state of the user node, use attractor search to iteratively update the state of the user node; when the iteration stops, obtain all attractors and their attractor basin information, and map them back to the user nodes in the original social network. The critical node output module is configured to calculate the criticality score of all user nodes and select the M user nodes with high criticality scores as critical nodes.

[0058] In some embodiments, the graph neural network encoder is constructed by using a graph convolutional neural network based on neighborhood aggregation as the basic architecture, introducing skip connections and attention mechanisms.

[0059] In some embodiments, based on a social network, a graph neural network encoder is used to obtain user node embedding vectors; this includes: using a graph neural network encoder to calculate a user node embedding matrix based on a user node feature matrix and a false information propagation path attribute feature matrix; the user node embedding matrix includes several user node embedding vectors.

[0060] In some embodiments, a binary gating mechanism is introduced to discretize the continuous user node embedding vectors into binary user node states; expressed by the following formula:

[0061] in, Represents the binary state of the discretized user node; This indicates a binary gating mechanism; Represents the user node embedding vector; when , In other cases, ; The weight matrix is ​​trainable. Indicates bias; This represents the input signal transmitted from the neuron node.

[0062] In some embodiments, attractor search is used to iteratively update the user node state. When the iteration stops, all attractors and their attractor basin information are obtained and mapped back to the user nodes in the original social network; this includes: randomly generating a set of initial states. ; Preset state transition rules, current state of user node, calculate the new state of user node in the next moment; when the system reaches a stable state or reaches the maximum number of iterations, the iteration stops, and all attractors and their attractor basin information are obtained. The lowest point of attractor potential energy corresponds to the user node of the original social network.

[0063] In some embodiments, the calculation of the critical scores for all user nodes is expressed using the following formula:

[0064] in, This represents the key score of the user node. It is the total number of attractors. For nodes The size of the attraction basin of the attractor.

[0065] In some embodiments, continuous user node embedding vectors are discretized into binary user node states; based on the discretized binary user node states, attractor search is used to iteratively update the user node states; including: processing continuous-time dynamic networks through a sliding window mechanism, discretizing continuous user node embedding vectors into binary user node states within each time window, applying attractor search within each time window, tracking the evolution trajectory of the attractors, and identifying key nodes in different time periods.

[0066] In some embodiments, during the training of the model, a contrastive learning strategy is adopted to maximize the similarity of key nodes and minimize the similarity of key nodes by constructing sample pairs of key nodes and non-key nodes.

[0067] The social network key node identification system based on discrete attractor search according to embodiments of the present invention can correspond to the execution of the method described in the embodiments of the present invention, and the above and other operations and / or functions of each module of the social network key node identification system based on discrete attractor search are respectively for implementing Figure 1 For the sake of brevity, the corresponding processes of each method in the code will not be elaborated here.

[0068] See Figure 3The diagram shows the structure of a computer device, which includes a processor, a communication interface, and a computer-readable storage medium. The processor, communication interface, and computer-readable storage medium are connected via a bus or other means. The communication interface is used to receive and send data. The computer-readable storage medium can be stored in the computer device's memory. The computer-readable storage medium stores computer programs, including program instructions, and the processor executes the program instructions stored in the computer-readable storage medium. The processor (or CPU, Central Processing Unit) is the computing and control core of the computer device, adapted to implement one or more instructions, specifically adapted to load and execute one or more instructions to implement the corresponding steps in the embodiment of the method for identifying key nodes in a social network based on discrete attractor search.

[0069] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0070] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0073] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0074] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying key nodes in a social network based on discrete attractor search, characterized in that, include: A social network is constructed using users as nodes and the paths through which false information spreads between users as edges. Based on social networks, a graph neural network encoder is used to obtain user node embedding vectors; A binary gating mechanism is introduced to discretize the continuous user node embedding vector into a binary state of the user node; based on the discretized binary state of the user node, attractor search is used to iteratively update the state of the user node. When the iteration stops, all attractors and their attractor basin information are obtained and mapped back to the user nodes in the original social network. Calculate the criticality score of all user nodes, and select the M user nodes with high criticality scores as critical nodes.

2. The method for identifying key nodes in a social network based on discrete attractor search according to claim 1, characterized in that, The graph neural network encoder is constructed by using a graph convolutional neural network based on neighborhood aggregation as the basic architecture, and introducing skip connections and attention mechanisms.

3. The method for identifying key nodes in a social network based on discrete attractor search according to claim 2, characterized in that, Based on social networks, a graph neural network encoder is used to obtain user node embedding vectors. The method includes: using a graph neural network encoder, calculating a user node embedding matrix based on the user node feature matrix and the false information propagation path attribute feature matrix; the user node embedding matrix includes several user node embedding vectors.

4. The method for identifying key nodes in a social network based on discrete attractor search according to claim 1, characterized in that, A binary gating mechanism is introduced to discretize the continuous user node embedding vector into a binary state of the user node. Expressed using the following formula: in, Represents the binary state of the discretized user node; This indicates a binary gating mechanism; Represents the user node embedding vector; when , In other cases, ; The weight matrix is ​​trainable. Indicates bias; This represents the input signal transmitted from the neuron node.

5. The method for identifying key nodes in a social network based on discrete attractor search according to claim 1, characterized in that, Attractor search is used to iteratively update the user node state. When the iteration stops, all attractors and their attractor basin information are obtained and mapped back to the user nodes in the original social network. The method includes: randomly generating a set of initial states. ; Preset state transition rules, current state of user node, calculate the new state of user node in the next moment; when the system reaches a stable state or reaches the maximum number of iterations, the iteration stops, and all attractors and their attractor basin information are obtained. The lowest point of attractor potential energy corresponds to the user node of the original social network.

6. The method for identifying key nodes in a social network based on discrete attractor search according to claim 1, characterized in that, The critical score for all user nodes is calculated using the following formula: in, This represents the key score of the user node. It is the total number of attractors. For nodes The size of the attraction basin of the attractor.

7. The method for identifying key nodes in a social network based on discrete attractor search according to claim 1, characterized in that, Discretize the continuous user node embedding vectors into binary user node states; Based on the discretized binary state of user nodes, attractor search is used to iteratively update the user node state. The method includes: processing the continuous-time dynamic network through a sliding window mechanism, discretizing the continuous user node embedding vectors in each time window into the binary state of user nodes, applying attractor search in each time window, and tracking the evolution trajectory of the attractor to identify key nodes in different time periods.

8. The method for identifying key nodes in a social network based on discrete attractor search according to claim 1, characterized in that, During the training of the model, a contrastive learning strategy is adopted. By constructing sample pairs of key nodes and non-key nodes, the similarity of key nodes is maximized and minimized.

9. A social network key node identification system based on discrete attractor search, characterized in that, include: The social network building module is configured to construct a social network with users as nodes and the paths of false information propagation between users as edges. The encoder module is configured to: obtain user node embedding vectors using a graph neural network encoder based on a social network; The discrete attractor search module is configured to: introduce a binary gating mechanism to discretize the continuous user node embedding vector into a binary state of the user node; based on the discretized binary state of the user node, use attractor search to iteratively update the state of the user node; when the iteration stops, obtain all attractors and their attractor basin information, and map them back to the user nodes in the original social network. The critical node output module is configured to calculate the criticality score of all user nodes and select the M user nodes with high criticality scores as critical nodes.

10. A computer device, characterized in that, A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps of the method for identifying key nodes in a social network based on discrete attractor search as described in any one of claims 1-8.

Images