A social propagation risk mitigation and robust group selection method based on extreme value dependent network and maximum independent set

Abstract

This invention discloses a method for mitigating social propagation risk and selecting robust groups based on extreme value dependency networks and maximally independent sets. The method includes: acquiring user interaction data from a social network and constructing a graph structure based on the user interaction data; obtaining target node pairs from the graph structure for a preset propagation activity period and calculating the extreme value dependency of the target node pairs; constructing propagation networks with target nodes as vertices and extreme value dependency as the edge construction criterion under different dependency thresholds, evaluating the propagation networks, and determining the extreme value dependency propagation network based on the evaluation results; dividing the extreme value dependency propagation network into communities and determining central nodes from the divided communities; constructing a minimum extreme value dependency group based on the extreme value dependency propagation network; and identifying high-risk node groups in the minimum extreme value dependency group using a propagation loss index to obtain the identification result. This invention can achieve the identification, isolation, and control of tail propagation linkage relationships.

Description

Technical Field

[0001] This invention relates to the field of social network technology, specifically to a method for mitigating social propagation risk and selecting robust groups based on extreme value dependent networks and maximally independent sets. Background Technology

[0002] With the rapid development of social networks and online platforms, information dissemination is exhibiting exponential diffusion characteristics. Individual nodes, such as opinion leaders, media accounts, and bot groups, can trigger cascading effects through multi-level and multi-channel paths, leading to the risk of public opinion outbreaks or the systemic spread of false information.

[0003] In scenarios where user behavior distribution exhibits a heavy tail and extreme propagation events are frequent, traditional methods based on linear correlation or average influence are insufficient to effectively characterize the tail resonance effect. The Pearson correlation coefficient is calculated using the following formula: In the formula, Cov(X, Y) represents the covariance of random variables X and Y. These represent the standard deviations of X and Y, respectively. Correlation coefficients can become distorted in extreme cases, leading to biases in the identification of propagation chains and key bridging nodes.

[0004] Extreme value statistics and multivariate regularization theory provide a rigorous mathematical framework for describing the common extreme propagation among nodes. Extreme value dependence measures can reflect the collaborative strength of nodes under extreme propagation conditions, and can be used to determine the asymptotic propagation independence or extreme dependence of public opinion. However, most existing research remains at the level of propagation correlation measurement, lacking integration with network structure analysis, node selection, and propagation risk mitigation.

[0005] Complex network methods can describe the topology, community division, and propagation paths among groups, providing a foundation for identifying bridging nodes and critical propagation paths. However, existing network construction is mostly based on general similarity or frequency indicators, which are difficult to cope with extreme propagation scenarios; the threshold lacks a systematic optimization mechanism, resulting in networks that are too dense or too sparse, thus masking potential propagation chains.

[0006] Therefore, there is an urgent need for a method that can remain effective even in extreme situations and accurately locate potential transmission chains. Summary of the Invention

[0007] This invention provides a method for mitigating social propagation risk and selecting robust groups based on extreme value dependent networks and maximally independent sets, in order to solve the above-mentioned technical problems.

[0008] In a first aspect, the present invention provides a method for mitigating social propagation risk and selecting robust groups based on extreme value dependency networks and maximally independent sets, the method comprising:

[0009] Acquire user interaction data from social networks and construct a graph structure based on the user interaction data. The graph structure includes nodes and edges, where the nodes are users or topic tags and the edges are propagation behaviors between users.

[0010] Obtain target node pairs with a preset propagation period from the graph structure, and calculate the extreme value dependence of the target node pairs, wherein the target node pairs include two target nodes;

[0011] Construct a propagation network with the target node as the vertex and the extreme value dependency as the edge construction criterion under different dependency thresholds, evaluate the propagation network, and determine the extreme value dependency propagation network based on the evaluation results;

[0012] The extreme value dependency propagation network is divided into communities, and central nodes are determined from the divided communities;

[0013] Based on the extreme value dependency propagation network, by selecting unconnected nodes and filtering nodes with betweenness numbers below the threshold, a minimum extreme value dependency group can be constructed.

[0014] The high-risk node group in the minimum extreme value dependent group is identified using the propagation loss index, and the identification results are obtained.

[0015] In some embodiments of the present invention, the step of using the propagation loss index to identify high-risk node groups in the minimum extreme value dependent group and obtaining identification results includes:

[0016] Based on the public opinion influence of node i at time t The magnitude of public opinion influence at time t+Δt To obtain within the propagation window Changes in internal propagation;

[0017] When the propagation change is greater than the propagation risk value, the expected value of the propagation change is calculated; and based on the expected value, the average positive influence gain, and the mean propagation noise, the influence Sharpe ratio is calculated.

[0018] When the propagation change is less than the propagation risk value, the conditional risk propagation value of the network as a whole is obtained, and the conditional risk propagation increment is calculated based on the conditional risk propagation value of the whole and the propagation risk value.

[0019] In some embodiments of the present invention, evaluating the propagation network and determining the extreme value dependent propagation network based on the evaluation results includes:

[0020] Calculate the mean, average path length, network diameter, graph density, and clustering coefficient for each propagation network;

[0021] Based on the mean, average path length, network diameter, graph density, and clustering coefficient, the optimal threshold and the propagation network corresponding to the optimal threshold are determined from a pre-set candidate set of extreme value dependency thresholds as the extreme value dependency propagation network.

[0022] In some embodiments of the present invention, the step of dividing the extreme value dependency propagation network into communities and determining central nodes from the divided communities includes:

[0023] The extreme value dependency propagation network is divided into communities to obtain multiple communities.

[0024] The central nodes are determined based on the betweenness centrality and eigenvector centrality of each node in the divided community.

[0025] In some embodiments of the present invention, the step of constructing a minimum extreme value dependency group based on the extreme value dependency propagation network by selecting unconnected nodes and filtering nodes with betweenness numbers below a threshold includes:

[0026] Solve for the maximum independent set on the extreme value dependency propagation network to obtain the set of mutually unconnected nodes;

[0027] A low betweenness priority rule is introduced, which prioritizes non-core nodes and nodes with shorter diffusion paths from the set of unconnected nodes in order to construct a minimum extreme value dependency group.

[0028] In some embodiments of the present invention, obtaining target node pairs with a preset propagation activity period from the graph structure and calculating the extreme value dependence of the target node pairs includes:

[0029] For any pair of nodes in the graph structure, extract the joint sample set of the arbitrary node pair during a preset propagation activity period;

[0030] The joint sample set is filtered according to a preset high quantile threshold to obtain joint propagation events, which include the target node pairs.

[0031] The extreme value dependency of the target node pair is calculated using extreme value theory.

[0032] In some embodiments of the present invention, after obtaining the graph structure, the method further includes:

[0033] The propagation intensity sequence is obtained from the graph structure, and the propagation intensity sequence is subjected to fluctuation filtering. The filtered sequence is then verified.

[0034] In some embodiments of the present invention, the method further includes:

[0035] We employ a rolling window or event-triggered mechanism to periodically update network filtering, extreme value dependency estimation, social network topology reconstruction, and network structure and maximum independent set results.

[0036] Secondly, the present invention also provides a social propagation risk mitigation and robust group selection system based on extreme value dependency networks and maximally independent sets, the system comprising:

[0037] The graph structure acquisition module is used to acquire user interaction data in social networks and construct a graph structure based on the user interaction data. The graph structure includes nodes and edges. The nodes are users or topic tags, and the edges are propagation behaviors between users.

[0038] The extreme value dependency calculation module is used to obtain target node pairs with a preset propagation activity period from the graph structure and calculate the extreme value dependency of the target node pairs, wherein the target node pairs include two target nodes;

[0039] The propagation network screening module is used to construct propagation networks with different dependency thresholds, using the target node as the vertex and the extreme dependency as the edge construction criterion, and to evaluate the propagation network and determine the extreme dependency propagation network based on the evaluation results.

[0040] The centrality node determination module is used to divide the extreme value dependency propagation network into communities and determine the central nodes from the divided communities.

[0041] The minimum extreme value dependency group construction module is used to construct a minimum extreme value dependency group based on the extreme value dependency propagation network by selecting unconnected nodes and filtering nodes with betweenness numbers lower than the threshold.

[0042] The identification module is used to identify high-risk node groups in the minimum extreme value dependent group using the propagation loss index, and obtain the identification results.

[0043] In the social media propagation risk mitigation and robust group selection method based on extreme value dependency networks and maximally independent sets provided by this invention, the extreme value dependency measure of node pairs is estimated based on multivariate regular variation theory and generalized polar coordinate representation, thereby characterizing the potential resonance relationship under extreme propagation conditions. An extreme value dependency propagation network based on thresholds is constructed using users as nodes and extreme value correlations as edges. The maximally independent set of this network is solved using graph theory algorithms to obtain the robust node group with the weakest systematic tail linkage. Propagation loss or public opinion spillover intensity is used as a tail risk indicator, and the risk-adjusted propagation efficiency is evaluated through a robustness index driven by propagation loss. The systematic diffusion suppression effect is verified by combining a conditional propagation value increment heatmap. The method provided by this invention can effectively identify key propagation communities and bridging nodes in multiple types of social media datasets. The constructed maximally independent set node group exhibits lower systematic diffusion risk and higher robustness across different platforms and public opinion environments. This invention can be widely applied to social network risk management, public opinion monitoring and group behavior analysis, and can also be extended to complex systems with heavy-tailed dependency structures and cascading propagation risks, such as communication networks, network security and climate networks. It has good feasibility and broad application prospects. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a flowchart illustrating the method for mitigating social propagation risk and selecting robust groups based on extreme value dependent networks and maximally independent sets provided in this embodiment of the invention.

[0046] Figure 2 This is a structural diagram of the social topic extremum dependency network provided in an embodiment of the present invention;

[0047] Figure 3 This is a community structure diagram of a user-associated extreme value dependency network provided in an embodiment of the present invention;

[0048] Figure 4 This is a topic relevance heatmap based on the semantic level provided in this embodiment of the invention;

[0049] Figure 5 This is a topic relevance heatmap based on user behavior provided in an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0052] "A and / or B" includes the following three combinations: A only, B only, and a combination of A and B.

[0053] The use of "applies to" or "configured to" in this invention implies an open and inclusive language, which does not exclude the applicability to or configuration to devices performing additional tasks or steps. Additionally, the use of "based on" implies openness and inclusivity, because processes, steps, calculations, or other actions "based on" one or more conditions or values ​​may in practice be based on additional conditions or values ​​beyond those conditions.

[0054] In this invention, the term "exemplary" is used to mean "serving as an example, illustration, or description." Any embodiment described as "exemplary" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0055] The following describes, with reference to the accompanying drawings, a method for mitigating social propagation risk and selecting robust groups based on extreme value dependent networks and maximum independent sets, provided by embodiments of the present invention.

[0056] like Figure 1 As shown, this embodiment of the invention provides a method for mitigating social propagation risk and selecting robust groups based on extreme value dependency networks and maximally independent sets. The method includes the following steps:

[0057] S101, acquire user interaction data from the social network, and construct a graph structure based on the user interaction data.

[0058] User interaction data is collected from social network platforms, including likes, comments, reposts, and message transmissions. By arranging this data in chronological order to form a unified time series, the consistency of the data over time can be ensured.

[0059] The graph structure includes nodes and edges, where nodes are users or topic tags, and edges represent propagation behavior between users.

[0060] S102, obtain target node pairs with a preset propagation activity period from the graph structure, and calculate the extreme value dependence of the target node pairs, wherein the target node pairs include two target nodes.

[0061] S103, construct a propagation network with the target node as the vertex and the extreme value dependency as the edge construction criterion under different dependency thresholds, evaluate the propagation network, and determine the extreme value dependency propagation network based on the evaluation results.

[0062] S104, perform community division on the extreme value dependency propagation network, and determine the central node from the divided communities.

[0063] S105, based on the extreme value dependency propagation network, by selecting unconnected nodes and filtering nodes with betweenness numbers lower than the threshold, a minimum extreme value dependency group is constructed.

[0064] S106, use the propagation loss index to identify high-risk node groups in the minimum extreme value dependent group and obtain the identification results.

[0065] This invention provides a method for mitigating social media propagation risk and selecting robust groups based on extreme value dependency networks and maximally independent sets. It estimates the extreme value dependency measure of node pairs using multivariate regular variation theory and generalized polar coordinate representation, thus characterizing potential resonance relationships under extreme propagation conditions. A threshold-based extreme value dependency propagation network is constructed using users as nodes and extreme value correlations as edges. A graph theory algorithm is used to solve for the maximally independent set of this network, obtaining a robust node group with the weakest systemic tail linkage. Propagation loss or public opinion spillover intensity is used as a tail risk indicator. A robustness index driven by propagation loss is used to evaluate the risk-adjusted propagation efficiency, and a heatmap of conditional propagation value increments is used to verify the systemic diffusion suppression effect. The method provided by this invention can effectively identify key propagation communities and bridging nodes in various social media datasets. The constructed maximally independent set node group exhibits lower systemic diffusion risk and higher robustness across different platforms and public opinion environments. This invention can be widely applied to social network risk management, public opinion monitoring and group behavior analysis, and can also be extended to complex systems with heavy-tailed dependency structures and cascading propagation risks, such as communication networks, network security and climate networks. It has good feasibility and broad application prospects.

[0066] In some embodiments of the present invention, the step of using the propagation loss index to identify high-risk node groups in the minimum extreme value dependent group and obtaining identification results includes:

[0067] Based on the public opinion influence of node i at time t The magnitude of public opinion influence at time t+Δt To obtain within the propagation window Changes in internal propagation;

[0068] When the propagation change is greater than the propagation risk value, the expected value of the propagation change is calculated; and based on the expected value, the average positive influence gain, and the mean propagation noise, the influence Sharpe ratio is calculated.

[0069] When the propagation change is less than the propagation risk value, the conditional risk propagation value of the network as a whole is obtained, and the conditional risk propagation increment is calculated based on the conditional risk propagation value of the whole and the propagation risk value.

[0070] In some examples, let the sentiment influence of node i in the minimum extremum dependency group at time t be . The change within a propagation window Δt is defined as:

[0071]

[0072] Given a confidence level q, the propagation risk of node i is denoted as . That is, satisfying:

[0073]

[0074] Propagation loss is defined as:

[0075]

[0076] Used to characterize the expected loss of node i in an extreme negative propagation scenario.

[0077] The Sharpe ratio of influence based on expected propagation loss is calculated and defined as:

[0078]

[0079] In the formula, This represents the average positive influence gain of node i. The Sharpe ratio represents the baseline public opinion intensity or the mean of background propagation noise for the system. A higher Sharpe ratio indicates that, under the premise of controlling tail risk, the node provides higher effective reach efficiency.

[0080] Calculate and compare the conditional propagation increments to quantify the systematic spillover effects of specific nodes or subgroups on the overall network. This is done under the condition that node i is in an extremely negative propagation state, i.e.

[0081]

[0082] The overall network condition under risk propagation value is:

[0083]

[0084] Furthermore, the condition for node i in terms of risk propagation increment is:

[0085]

[0086] When necessary, construct and visualize the spillover effect matrix to verify whether the spillover intensity within key subgroups is significantly lower than the network-wide average, thereby identifying relatively stable isolated communities. Verify that the minimum extreme value dependent group is stable and highly effectively reachable compared to the overall average, and output periodic evaluation reports to provide decision support for risk control and reconfiguration. (Illustratively, as shown...) Figure 4 The heatmap shown is based on semantic-level topic relevance, where darker colors indicate stronger relevance between topics. Figure 5 The heatmap shown is based on user behavior and topic relevance. The darker the color, the stronger the relevance between users.

[0087] In some embodiments of the present invention, evaluating the propagation network and determining the extreme value dependent propagation network based on the evaluation results includes:

[0088] Calculate the mean, average path length, network diameter, graph density, and clustering coefficient for each propagation network;

[0089] Based on the mean, average path length, network diameter, graph density, and clustering coefficient, an optimal threshold and the propagation network corresponding to the optimal threshold are determined from multiple dependency thresholds as the extreme value dependency propagation network. Figure 3 As shown, in this extreme value-dependent propagation network, each user is a node, and its size represents the user's influence.

[0090] In some examples, a propagation network with different dependency thresholds is constructed, using the target node as the vertex and the extreme value dependency as the edge construction criterion. Specifically, the network is generated with the node as the vertex and the extreme value dependency as the edge construction criterion, resulting in a propagation network with different thresholds.

[0091] Within a pre-defined candidate set of extreme value dependency thresholds, the average degree, average path length, network diameter, graph density, and clustering coefficient of the extreme value dependency propagation networks constructed under different thresholds are calculated. The specific calculation methods are as follows:

[0092] Average path length L:

[0093]

[0094] In the formula, N represents the number of nodes. This represents the number of edges connecting node i and node j.

[0095] Image density:

[0096]

[0097] In the formula, M represents the total number of edges in the network.

[0098] Clustering coefficient:

[0099]

[0100] In the formula, This represents the number of edges between node i and its neighboring nodes. This represents the number of neighboring nodes of node i.

[0101] The aforementioned metrics characterize the network's overall connectivity, extreme dependency propagation efficiency, maximum propagation distance, connection sparsity, and local clustering structure features. Threshold selection follows two principles: firstly, avoiding excessively low thresholds that lead to overly dense networks and the introduction of numerous weak dependencies, thus weakening the interpretability of the propagation structure; secondly, avoiding excessively high thresholds that cause high network fragmentation, resulting in the loss of key propagation paths and community structures. The optimal threshold maximizes both the average path length and network diameter while maintaining overall network connectivity. This determines the optimal threshold, which is then used to fix the network edge set, resulting in the final extreme dependency propagation network for subsequent community identification and node selection.

[0102] In some embodiments of the present invention, the step of dividing the extreme value dependency propagation network into communities and determining central nodes from the divided communities includes:

[0103] The extreme value dependency propagation network is divided into communities to obtain multiple communities.

[0104] The central nodes are determined based on the betweenness centrality and eigenvector centrality of each node in the divided community.

[0105] In some examples, community partitioning is performed on the constructed extreme value dependency propagation network using the Girvan-Newman (GN) or Louvain algorithm to obtain multiple partitioned communities, such as... Figure 2 As shown, the extreme value dependency propagation network is divided into multiple communities, with nodes of different colors representing different communities, vertex colors representing topic affiliation, their size representing the popularity of the topic, and edges between clustered nodes representing the association between topics.

[0106] Highly cohesive communities are identified based on their internal connectivity density, and cross-community bridging nodes are determined by combining the cross-community edge characteristics and betweenness centrality of nodes. The main information diffusion path is characterized by the bridging nodes and their connections.

[0107] Further calculations are performed on the betweenness centrality of the main path nodes to identify key propagation nodes. The betweenness centrality of node v is defined as:

[0108]

[0109] In the formula, It is the number of shortest paths between node i and node j. It is the number of paths passing through node v in the shortest path between i and j. The normalized betweenness centrality is:

[0110]

[0111] Highly central nodes are defined as high-risk sources of propagation or propagation bridges.

[0112] In some embodiments of the present invention, based on the extreme value dependency propagation network, a minimum extreme value dependency group is constructed by selecting a set of mutually unconnected nodes that avoid high centrality nodes, including:

[0113] Solving for the maximum independent set on an extreme value dependency propagation network yields the set of nodes that are not connected to any two nodes.

[0114] Since the solution to the maximum independent set may not be unique, a high centrality avoidance rule is introduced to prioritize the maximum independent set that does not contain core nodes among multiple feasible solutions in order to construct the minimum extreme value dependency group.

[0115] In some examples, a greedy algorithm is used to solve for the maximum independent set of the extreme value dependency propagation network, which yields the set of nodes with the largest number of nodes among all sets that satisfy the condition of being mutually unconnected.

[0116] Due to the non-uniqueness of the maximum independent set solution, a high centrality avoidance rule is introduced to preferentially select the maximum independent set solution that does not contain core nodes.

[0117] Construct a minimum extreme value dependency group to achieve network-level isolation of tail propagation risk.

[0118] In some embodiments of the present invention, obtaining target node pairs with a preset propagation activity period from the graph structure and calculating the extreme value dependence of the target node pairs includes:

[0119] For any pair of nodes in the graph structure, extract the joint sample set of the arbitrary node pair during a preset propagation activity period.

[0120] The joint sample set is filtered according to a preset high quantile threshold to obtain joint propagation events, which include the target node pairs.

[0121] The extreme value dependency of the target node pair is calculated using extreme value theory.

[0122] In some examples, for node pairs in joint propagation events, the extreme value dependence of the node pairs is estimated using multivariate canonical variation theory and polar coordinate transformation in extreme value theory:

[0123]

[0124] In the formula, They are independent and identically distributed random vectors. , These are the j-th sample data points from nodes 1 and 2, respectively. and x is R j 95% quantile.

[0125] In some embodiments of the present invention, after obtaining the initial graph structure, the method further includes the following data preprocessing steps:

[0126] In some examples, the graph structure is first cleaned to obtain a cleaned graph structure, such as removing zombie accounts, duplicate samples, and abnormal propagation peaks, and unifying the time granularity. Based on the cleaned graph structure, the propagation intensity sequence of each node or node pair is extracted.

[0127] A conditional heteroscedasticity model, such as EGARCH, is used to perform fluctuation filtering on the propagation intensity sequence to obtain the corresponding propagation residual sequence. The Ljung-Box test and ARCHLM test are used to verify the residual sequence to confirm that autocorrelation and fluctuation clustering have been effectively eliminated. When the verification passes, the corresponding propagation residual sequence is used as input data in the subsequent extreme value dependence calculation step; when the verification fails, the node sequence is deleted.

[0128] In some embodiments of the present invention, risk warning rules are set so that when tail risk or systemic spillover indicators exceed the threshold, automatic intervention measures are triggered, such as limiting the flow of high-risk accounts, adjusting the information recommendation weight, and starting a manual review queue.

[0129] Maintain audit records of historical versions and parameters to support traceable management and compliance review.

[0130] This invention also provides a social propagation risk mitigation and robust group selection system based on extreme value dependency networks and maximally independent sets, the system comprising:

[0131] The graph structure acquisition module is used to acquire user interaction data in social networks and construct a graph structure based on the user interaction data. The graph structure includes nodes and edges. The nodes are users or topic tags, and the edges are propagation behaviors between users.

[0132] The extreme value dependency calculation module is used to obtain target node pairs with a preset propagation activity period from the graph structure and calculate the extreme value dependency of the target node pairs, wherein the target node pairs include two target nodes;

[0133] The propagation network screening module is used to construct propagation networks with different dependency thresholds, using the target node as the vertex and the extreme dependency as the edge construction criterion, and to evaluate the propagation network and determine the extreme dependency propagation network based on the evaluation results.

[0134] The centrality node determination module is used to divide the extreme value dependency propagation network into communities and determine the central nodes from the divided communities.

[0135] The minimum extreme value dependency group construction module is used to construct a minimum extreme value dependency group based on the extreme value dependency propagation network by selecting unconnected nodes and filtering nodes with betweenness numbers lower than the threshold.

[0136] The identification module is used to identify high-risk node groups in the minimum extreme value dependent group using the propagation loss index, and obtain the identification results.

[0137] It should be noted that the social propagation risk mitigation and robust group selection system based on extreme value dependent networks and maximum independent sets provided in this embodiment of the invention corresponds to the above-mentioned social propagation risk mitigation and robust group selection method based on extreme value dependent networks and maximum independent sets, and will not be elaborated further here.

[0138] Furthermore, the social media propagation risk mitigation and robust group selection system based on extreme value dependency networks and maximum independent sets provided in this invention supports batch processing and online computation, and offers parameterized configuration, including window length, threshold quantiles, edge construction thresholds, risk confidence levels, community size, and sensitive topic restrictions. It supports result export, report generation, and graphical display, facilitating integration and application in scenarios such as public opinion monitoring, content governance, and platform compliance.

[0139] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0141] The foregoing has provided a detailed description of a social propagation risk mitigation and robust group selection method based on extreme value dependent networks and maximum independent sets, as provided in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for mitigating social propagation risk and selecting robust groups based on extreme value dependency networks and maximally independent sets, characterized in that, The method includes: Acquire user interaction data from social networks and construct a graph structure based on the user interaction data. The graph structure includes nodes and edges, where the nodes are users or topic tags and the edges are propagation behaviors between users. Obtain target node pairs with a preset propagation period from the graph structure, and calculate the extreme value dependence of the target node pairs, wherein the target node pairs include two target nodes; Construct a propagation network with the target node as the vertex and the extreme value dependency as the edge construction criterion under different dependency thresholds, evaluate the propagation network, and determine the extreme value dependency propagation network based on the evaluation results; The extreme value dependency propagation network is divided into communities, and central nodes are determined from the divided communities; Based on the extreme value dependency propagation network, by selecting unconnected nodes and filtering nodes with betweenness numbers below the threshold, a minimum extreme value dependency group can be constructed. The high-risk node group in the minimum extreme value dependent group is identified using the propagation loss index, and the identification results are obtained.

2. The method for social propagation risk mitigation and robust group selection based on extreme value dependency networks and maximally independent sets according to claim 1, characterized in that, The process of identifying high-risk node groups in the minimum extreme value dependency group using the propagation loss index and obtaining the identification results includes: Based on the public opinion influence of node i at time t The magnitude of public opinion influence at time t+Δt To obtain within the propagation window Changes in internal propagation; When the propagation change is greater than the propagation risk value, the expected value of the propagation change is calculated; and based on the expected value, the average positive influence gain, and the mean propagation noise, the influence Sharpe ratio is calculated. When the propagation change is less than the propagation risk value, the conditional risk propagation value of the network as a whole is obtained, and the conditional risk propagation increment is calculated based on the conditional risk propagation value of the whole and the propagation risk value.

3. The method for social propagation risk mitigation and robust group selection based on extreme value dependency networks and maximally independent sets according to claim 1, characterized in that, The step of evaluating the propagation network and determining the extreme value dependent propagation network based on the evaluation results includes: Calculate the mean, average path length, network diameter, graph density, and clustering coefficient for each propagation network; Based on the mean, average path length, network diameter, graph density, and clustering coefficient, the optimal threshold and the propagation network corresponding to the optimal threshold are determined from a pre-set candidate set of extreme value dependency thresholds as the extreme value dependency propagation network.

4. The method for mitigating social propagation risk and selecting robust groups based on extreme value dependency networks and maximally independent sets according to claim 1, characterized in that, The step of dividing the extreme value dependency propagation network into communities and determining central nodes from the divided communities includes: The extreme value dependency propagation network is divided into communities to obtain multiple communities. The central nodes are determined based on the betweenness centrality and eigenvector centrality of each node in the divided community.

5. The method for social propagation risk mitigation and robust group selection based on extreme value dependency networks and maximally independent sets according to claim 1, characterized in that, Based on the aforementioned extreme value dependency propagation network, the process involves selecting unconnected nodes and filtering nodes with betweenness numbers below a threshold to construct a minimum extreme value dependency group, including: Solve for the maximum independent set on the extreme value dependency propagation network to obtain the set of mutually unconnected nodes; A low betweenness priority rule is introduced, which prioritizes non-core nodes and nodes with shorter diffusion paths from the set of unconnected nodes in order to construct a minimum extreme value dependency group.

6. The method for social propagation risk mitigation and robust group selection based on extreme value dependency networks and maximally independent sets according to claim 1, characterized in that, The step of obtaining target node pairs with a preset propagation activity period from the graph structure and calculating the extreme value dependence of the target node pairs includes: For any pair of nodes in the graph structure, extract the joint sample set of the arbitrary node pair during a preset propagation activity period; The joint sample set is filtered according to a preset high quantile threshold to obtain joint propagation events, which include the target node pairs. The extreme value dependency of the target node pair is calculated using extreme value theory.

7. The method for social propagation risk mitigation and robust group selection based on extreme value dependency networks and maximally independent sets according to claim 1, characterized in that, After obtaining the graph structure, the method further includes: The propagation intensity sequence is obtained from the graph structure, and the propagation intensity sequence is subjected to fluctuation filtering. The filtered sequence is then verified.

8. A social propagation risk mitigation and robust group selection system based on extreme value dependency networks and maximal independent sets, characterized in that, The system includes: The graph structure acquisition module is used to acquire user interaction data in social networks and construct a graph structure based on the user interaction data. The graph structure includes nodes and edges. The nodes are users or topic tags, and the edges are propagation behaviors between users. The extreme value dependency calculation module is used to obtain target node pairs with a preset propagation activity period from the graph structure and calculate the extreme value dependency of the target node pairs, wherein the target node pairs include two target nodes; The propagation network screening module is used to construct propagation networks with different dependency thresholds, using the target node as the vertex and the extreme dependency as the edge construction criterion, and to evaluate the propagation network and determine the extreme dependency propagation network based on the evaluation results. The centrality node determination module is used to divide the extreme value dependency propagation network into communities and determine the central nodes from the divided communities. The minimum extreme value dependency group construction module is used to construct a minimum extreme value dependency group based on the extreme value dependency propagation network by selecting unconnected nodes and filtering nodes with betweenness numbers lower than the threshold. The identification module is used to identify high-risk node groups in the minimum extreme value dependent group using the propagation loss index, and obtain the identification results.

Images