A method and system for long-term intervention in online rumors

Abstract

This invention relates to the field of online social network information governance, and more particularly to a method and system for long-term intervention in online rumors. Addressing the weaknesses of traditional models in dynamic rumor debunking mechanisms, this invention introduces an "evolutionary factor" mechanism and categorizes network nodes into four classes: relatives, friends, strangers, and authorities. A multi-level priority processing mechanism is designed, and a modified Giza Pyramid Construction Algorithm (GPC) is used to perform global source detection. The Crow Search Algorithm (CSA) is employed for pseudo-extreme adaptive avoidance, and Particle Swarm Optimization (PSO) is combined to achieve group collaborative consensus optimization. This effectively overcomes the shortcomings of single intelligent optimization algorithms that are prone to getting trapped in local optima, significantly improving the accuracy of rumor identification and the timeliness of rumor debunking.

Description

Technical Field

[0001] This invention relates to the field of online social network information governance, and in particular to a method and system for long-term intervention in online rumors. Background Technology

[0002] Against the backdrop of rapid development in information technology and highly mature infrastructure, online social networks have deeply penetrated all areas of social governance, while also providing fertile ground for the breeding and spread of online rumors. The evolution of modern online rumors exhibits new characteristics such as strong concealment, precise targeting, and lasting negative impact. Their dissemination process is often accompanied by information distortion and group polarization, seriously threatening national cybersecurity and social stability.

[0003] In existing technologies, academia has conducted extensive research on the mechanisms of rumor propagation, such as propagation dynamics analysis based on the SIR model and its variant, the SEIR model, as well as research on authoritative media intervention mechanisms. However, the above methods have significant shortcomings in dealing with the rapidly changing public opinion environment: First, traditional passive rumor debunking models struggle to track and cut off the exponentially growing flow of false information; second, existing models mostly focus on macro-level propagation dynamics analysis, leaving considerable room for optimization in terms of the execution efficiency of underlying algorithms and long-term rumor resistance in complex networks; third, single-swarm intelligent optimization algorithms generally face the dilemma of an imbalance between global exploration and local development capabilities when dealing with the highly nonlinear and strongly coupled problem of optimizing rumor debunking parameters.

[0004] Specifically, the Giza Pyramid Construction Algorithm (GPC) is prone to search stagnation in the later stages of optimization; the Crow Search Algorithm (CSA) may suffer from loss of population diversity and get stuck in local optima due to its blind following behavior; and the Particle Swarm Optimization (PSO) algorithm is prone to premature convergence when facing high-dimensional and complex solution spaces. The inherent limitations of these single operators make it difficult for them to meet the extremely high timeliness and accuracy requirements of debunking strategies in the post-truth era.

[0005] Therefore, it is necessary to develop a dynamic rumor-debunking mechanism that can achieve high efficiency, continuous and real-time operation, and to build a precise identification and intervention system to effectively curb the spread of online rumors. Summary of the Invention

[0006] This invention provides a method and system for long-term intervention in online rumors, aiming to solve the problems of low convergence efficiency of single optimization algorithms and insufficient accuracy of debunking strategies in existing rumor propagation intervention methods, and to achieve efficient and real-time debunking of online rumors.

[0007] Firstly, this application provides a method for long-term intervention in online rumors, the method comprising:

[0008] User nodes in an online social network are initialized, and an online social network simulation model containing debunker nodes and audience nodes is constructed. Each node is assigned an initial fitness value. The fitness value is used to measure the degree to which a node grasps the truth, and the value range is [0,1]. At the initial moment, the fitness of all nodes is at a low level.

[0009] Based on a trust-based hierarchical mechanism, audience nodes are categorized into four types according to their social relationship attributes: family member nodes, friend nodes, colleague / acquaintance nodes, and stranger nodes. to Four decreasing levels of trust;

[0010] The OGPC-CSA-PSO hybrid optimization algorithm is initiated, and the algorithm iteratively evolves according to the following three-stage mechanism: First, a global source detection is performed using an improved Giza pyramid construction algorithm, by introducing... The flight operator and transfer function complete the position evolution of individuals in the solution space of the social network; secondly, the crow search algorithm is used to perform pseudo-extreme adaptive avoidance, and the memory mechanism and awareness probability are used to determine whether individuals should make forced displacements in order to escape the local optimum trap; finally, the particle swarm optimization algorithm is used to perform group collaborative consensus optimization, and the evolution direction of all individuals is fine-tuned through the velocity update formula to guide the group to gather towards the global optimal rumor-refuting parameter region.

[0011] Perform boundary constraint processing and memory updates, and determine whether the termination condition is met. If it is met, output the optimal intervention strategy parameters.

[0012] In a preferred embodiment, the initialization of user nodes in the online social network specifically involves: subdividing audience nodes into authority nodes, family member nodes, friend nodes, colleague nodes, and stranger nodes based on differences in trust levels, and assigning corresponding dissemination influence weights to nodes of different levels as the priority basis for subsequent debunking information intervention.

[0013] In a preferred embodiment, the step of performing global source detection using the improved Giza pyramid construction algorithm specifically involves: simulating the Mantegna algorithm. flight stride ,Right now:

[0014] ,in and Follows a normal distribution. For exponential parameters;

[0015] Using the step size Generate a random search distance and combine it with an adjustment coefficient. The formula for updating individual positions is:

[0016] ,in Current position This represents the displacement.

[0017] In a preferred embodiment, the evolution of an individual's position in the solution space of a social network is nonlinearly mapped using a transfer function. Specific steps include:

[0018] Obtain the worker position update vector calculated in a continuous coordinate space by the improved Giza pyramid construction algorithm. The worker position update vector represents the evolution trend of the intervention strategy in the parameter space.

[0019] use Shape transfer function or The shape transfer function maps continuous displacements to intervals. The probability value within;

[0020] Based on the probability value, the audience nodes in the social network are discretized and their states are determined to determine whether the corresponding node accepts the intervention of the debunking information flow in the current iteration.

[0021] By using a nonlinear mapping mechanism, the originally continuous search step size is converted into the jump probability between discrete nodes in the social network, so as to solve the problems of low optimization efficiency and discontinuity that traditional optimization algorithms are prone to when dealing with discrete social network topologies.

[0022] In a preferred embodiment, the pseudo-extreme adaptive avoidance is performed using the crow search algorithm, and the specific steps include:

[0023] Build location memory for each debunking agent in social networks This is used to store the historical best intervention points found by the corresponding agent individual in the rumor dissemination path;

[0024] Through the preset probability of perception Simulate the ability of individuals debunking rumors on social networks to identify false information and generate random numbers;

[0025] If random number The strategy involves implementing a location-following approach, which updates the current search direction based on the experience of other excellent debunkers, in order to quickly approach the core source of the rumor.

[0026] If random number The system implements an adaptive avoidance strategy, which determines that the current area is a rumor trap or a local pseudo-extreme point. It forces individuals to perform random flight within the solution space of the social network, simulating the debunking information actively jumping out of the information cocoon or echo barrier effect area, so as to ensure the global robustness of the optimization process.

[0027] In a preferred embodiment, the particle swarm optimization algorithm is used to perform group cooperative consensus optimization, and the specific steps include:

[0028] The evolution of group cognition in social networks is abstracted as the motion process of a particle swarm, defining that the speed update of each individual is affected by individual cognitive factors. Social experience factors Driven by both; among them, This corresponds to the independent judgment of facts by social network nodes based on their own trust logic. Consensus offset caused by the influence of authoritative nodes and the chain of trust on the corresponding nodes;

[0029] By using the velocity update formula to adjust the evolution speed and displacement direction of all nodes in the cognitive space in real time, the audience in the social network is guided to converge toward the global optimal solution under the interference of multi-level trust weights.

[0030] Through iterative evolution, social network nodes in a chaotic state gradually reach a collective consensus, thereby completely blocking the chain of rumor propagation.

[0031] In a preferred embodiment, the calculation of an individual fitness value includes the following steps:

[0032] A multidimensional fitness function based on the social network background is constructed, which considers the coverage depth of the debunking information, the propagation attenuation factor, and the trust difference of the audience nodes.

[0033] An evolutionary factor mechanism is introduced to dynamically assess the probability of each node transforming from a victim of rumors into a disseminator of facts, and this probability is quantified into an interval. The value within;

[0034] Calculate the average fitness of all individuals globally. When the fitness value of the audience node monotonically increases and converges to the threshold 1, it is determined that the social network has formed a long-term rumor immunity mechanism.

[0035] If the maximum number of iterations or error convergence accuracy is met, then the current node weight distribution and intervention parameters are locked as the final optimal intervention strategy output.

[0036] In a preferred embodiment, the fitness of each individual node after each iteration is evaluated. When the average fitness of all nodes in the network exceeds the 0.99 threshold, the algorithm is determined to have successfully converged, and the optimal rumor-refuting strategy parameters are output.

[0037] Based on the optimal output parameters, implement differentiated debunking intervention strategies according to the identity attributes of the disseminators, and evaluate the debunking effect and iterate the strategies.

[0038] Secondly, this application provides a long-term intervention system for online rumors, the system comprising:

[0039] The data acquisition and preprocessing module is used to capture information flow data from online social networks in real time, identify the main propagators and audience nodes in the rumor propagation chain, and complete the attribute labeling and classification of node trust level based on the interaction frequency and historical trust endorsement between nodes.

[0040] The initialization and parameter configuration module is used to load the search solution space range of the intervention strategy, set the population size and maximum number of iterations of the preset hybrid intelligent optimization model, and configure the awareness probability of the crow search algorithm according to the urgency of the current public opinion. ;

[0041] The intervention strategy output module is used to generate precise intervention plans for nodes with different trust levels based on the global optimal solution output by the preset hybrid intelligent optimization model. These plans include intervention timing, information weight distribution, and key propagation path recommendations, in order to achieve long-term blocking of online rumors.

[0042] The preset hybrid intelligent optimization model integrates:

[0043] A global detection unit is used in conjunction with an improved Giza pyramid construction algorithm. Flight operators perform a wide-ranging search for debunking intervention points in discrete topologies of social networks;

[0044] The adaptive avoidance unit utilizes the memory mechanism of the crow search algorithm to monitor and forcibly avoid local pseudo-extreme points in the solution space in real time, so as to prevent the intervention strategy from getting trapped in an information cocoon.

[0045] The group consensus coordination unit uses the speed update logic of the particle swarm optimization algorithm to simulate the consensus shift process of a social group toward the truth, guiding the overall nodes to converge toward the optimal cognitive state.

[0046] The convergence judgment unit is used to perform boundary constraint processing and memory update, and to determine whether the termination condition is met. If it is met, the optimal intervention strategy parameters are output.

[0047] The present invention has the following beneficial effects:

[0048] (1) The OGPC-CSA-PSO hybrid optimization algorithm proposed in this invention effectively overcomes the imbalance between global exploration and local development of a single algorithm through a three-stage hierarchical fusion mechanism. Simulation experiments show that the algorithm can achieve high-quality convergence in only 20 iterations, which is significantly better than the GWO algorithm (25 iterations) and the WOA algorithm (poor convergence effect), and demonstrates excellent convergence speed and global search accuracy in the optimization of rumor-refuting strategy parameters.

[0049] (2) This invention innovatively introduces a node classification mechanism based on trust level, classifying social network users into four categories of social relationships: family members, friends, colleagues / acquaintances, and strangers, and assigning them... to Differentiated trust levels enable an upgrade in debunking strategies, moving from full-network coverage to precise targeting, significantly enhancing the social acceptance and dissemination of debunking information.

[0050] (3) This invention introduces an evolutionary factor mechanism and constructs a systematic network public opinion dynamic governance system through three major control modules: full-domain source detection, pseudo-extreme adaptive avoidance, and group collaborative consensus. It can effectively characterize the herd immunity effect generated by netizens after long-term exposure to rumors and debunking information, and provides an efficient calculation scheme for long-term and real-time intervention of network rumors. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the steps of the long-term intervention method for online rumors provided in the embodiments of the present invention;

[0052] Figure 2 This is a schematic diagram showing the convergence comparison experiment results of the OGPC-CSA-PSO hybrid algorithm, GWO algorithm, and WOA algorithm provided in the embodiments of the present invention.

[0053] Figure 3 This is a schematic diagram comparing the convergence trajectories of the CSA algorithm and the PSO algorithm when running alone in the ablation experiment provided in this embodiment of the invention;

[0054] Figure 4 This is a schematic diagram of the structure of the long-term intervention system for online rumors provided in an embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] Example 1

[0057] This embodiment is a flowchart illustrating a long-term intervention method for online rumors, based on the OGPC-CSA-PSO hybrid algorithm. (See also...) Figure 1 As shown, it includes the following steps:

[0058] The user nodes in the online social network are initialized and divided into debunkers and audiences. The population size and maximum number of iterations are loaded to construct an online social network simulation model.

[0059] Network nodes are classified based on a trust level grading mechanism, and a mapping between social relationship attributes and trust levels is established.

[0060] The OGPC-CSA-PSO hybrid optimization algorithm is launched, and a three-stage iterative optimization is performed, which includes full-domain source detection, pseudo-extreme value avoidance, and group collaborative consensus.

[0061] The fitness of nodes after each iteration is evaluated to determine whether the algorithm meets the convergence condition and output the optimal debunking parameters.

[0062] Implement differentiated rumor-debunking intervention strategies based on node attributes according to optimal parameters, evaluate the effects, and conduct feedback iterations.

[0063] Specifically, in step 1, a structure containing An online social network (OSN) simulation environment with an initial number of nodes. The nodes are divided into... Refuters and The listeners are named. The fitness (truth-knowing degree) of all nodes is strictly limited to a range of values. Within this framework, the mathematical equivalent value of ultimate truth is defined as 1. At the initial moment the debunking mechanism is activated ( All nodes initially have low fitness, simulating the initial state of social network users being misled by false information.

[0064] Specifically, in step 2, network nodes are divided into four categories based on a trust level mechanism: family member nodes correspond to the highest trust level. ; Friend nodes correspond to a high level of trust. Colleague / acquaintance nodes correspond to general trust levels. Stranger nodes correspond to a lower level of trust. The four trust levels satisfy an ordinal relation. Furthermore, by establishing authoritative nodes (government accounts, mainstream media, and industry experts), their credibility is similar to... Levels serve as a core source of information for debunking rumors.

[0065] Specifically, in step 3, the OGPC-CSA-PSO hybrid optimization algorithm is started, and the execution process is as follows:

[0066] S31: System initialization and parameter loading. Initialize population size. Maximum number of iterations The oblique angle coefficient of the GPC algorithm CSA perception probability and flight step length coefficient Individual learning factor of PSO Social learning factors In the solution space Internal random generation The initial individual position.

[0067] S32: Fitness Evaluation and Role Assignment. Calculate the fitness value of each individual in the current population, sort them in descending order, and divide them into three levels: Level 1 (Pharaoh) is the individual with the best fitness, serving as a global guide; Level 2 (Engineers) is the second-best performing individual, responsible for local searches; Level 3 (Workers) consists of the remaining individuals, performing global exploration tasks.

[0068] S33: Global Source Detection (Improved GPC Evolution). Perform position updates for individual workers: calculate initial displacement based on pyramid slope. Worker movement ; Introduction Flight operator, step size simulated according to Mantegna algorithm ( For exponential parameters, , (Following a normal distribution); the improved worker position update formula is: Finally utilize Shape transfer function or V-shaped transfer function This maps continuous displacement to nonlinear position changes, thus completing the position evolution.

[0069] S34: Adaptive Avoidance of False Extrema (CSA Interference Check). Each search individual randomly selects another individual from the population for observation: if the random number... Then, the historical best memory matrix is ​​used for strong correction, forcing individuals to turn to unexplored areas; if If the current update trajectory is not detected, then the current update trajectory is maintained. The CSA individual location update formula is: when not detected... When being aware For a random location within the search space. The crow's memory update formula is: when hour Otherwise keep constant.

[0070] S35: Group Collaborative Consensus Optimization (PSO Precision Refinement). The current globally optimal individual is used as the global extreme value of PSO. Using the speed update formula and the position update formula This drives all search agents to quickly converge on the region with the optimal debunking parameters.

[0071] S36: Boundary constraint handling and memory update. Check if all individuals exceed the limit. The parameter value range is used to perform boundary bounce or re-initialization processing on individuals that exceed the range, and to update the historical best memory and global best position of each individual.

[0072] S37: Termination Check. Determine if the current iteration count has been reached. Alternatively, the convergence accuracy requirement must be met, i.e., the average fitness of the entire network ≥ 0.99. If this is met, the iteration stops, and the optimal rumor-refuting strategy parameters are output; otherwise, return to S32 to continue the evolution.

[0073] Specifically, in step 4, when the average fitness of all nodes in the network exceeds the 0.99 threshold, the algorithm is considered to have successfully converged to the ultimate truth point. Simulation experiments show that the OGPC-CSA-PSO algorithm achieves high-quality convergence in only 20 iterations in a social network model with 500 nodes (100 debunkers and 400 audience members). See [link to relevant documentation]. Figure 2 As shown in the diagram. Comparative experiments show that the GWO algorithm requires 25 iterations to converge; the WOA algorithm has poor convergence performance, with some user fitness failing to reach the convergence threshold even after 95 iterations. Ablation experiments (see...) Figure 3 This further shows that both the CSA and PSO algorithms, when run independently, require more than 40 iterations to reach a steady state.

[0074] Specifically, in step 5, based on the optimal parameters output by the algorithm, a differentiated rumor-refuting intervention strategy is implemented according to the node trust level: authoritative nodes prioritize issuing official announcements and providing authoritative fact verification; family member nodes ( Level 1) indirectly reminds users via private messages or family groups, leveraging high trust levels to increase acceptance of debunking claims; friend nodes ( Level 1) spreads debunking content through friends' circles or friend groups, achieving secondary penetration of debunking information; stranger nodes ( (Level) It achieves lateral topological diffusion through large-scale public broadcasting. The system continuously evaluates the blocking effect of each strategy and dynamically iterates and optimizes the strategies based on feedback results, thereby curbing the repeated outbreaks of rumors over a long period of time.

[0075] Example 2

[0076] This embodiment describes a long-term intervention system for online rumors. (See also...) Figure 4 As shown, it includes:

[0077] The data acquisition and preprocessing module 100 is used to capture information flow data from online social networks in real time, identify the main propagators and audience nodes in the rumor propagation chain, and complete the attribute labeling and classification of node trust level based on the interaction frequency and historical trust endorsement between nodes.

[0078] The initialization and parameter configuration module 200 is used to load the search solution space range of the intervention strategy, set the population size and maximum number of iterations of the preset hybrid intelligent optimization model 300, and configure the awareness probability of the crow search algorithm according to the urgency of the current public opinion. ;

[0079] The intervention strategy output module 400 is used to generate precise intervention plans for nodes with different trust levels based on the global optimal solution output by the preset hybrid intelligent optimization model. These plans include intervention timing, information weight distribution, and key propagation path recommendations, in order to achieve long-term blocking of online rumors.

[0080] The pre-defined hybrid intelligent optimization model 300 specifically includes:

[0081] Global detection unit 301 is used in conjunction with an improved Giza pyramid construction algorithm. Flight operators perform a wide-ranging search for debunking intervention points in discrete topologies of social networks;

[0082] The adaptive avoidance unit 302 utilizes the memory mechanism of the crow search algorithm to monitor and forcibly avoid local pseudo-extreme points in the solution space in real time, so as to prevent the intervention strategy from getting trapped in an information cocoon.

[0083] The group consensus coordination unit 303 uses the speed update logic of the particle swarm optimization algorithm to simulate the consensus shift process of the social group towards the truth of the facts, and guides the overall nodes to converge towards the optimal cognitive state.

[0084] The convergence judgment unit 304 is used to perform boundary constraint processing and memory update, and to determine whether the termination condition is met. If it is met, the optimal intervention strategy parameters are output.

[0085] It should be noted that the explanation of the above-mentioned embodiments of the long-term intervention method for online rumors also applies to the long-term intervention system for online rumors described in the above embodiments, and will not be repeated here.

[0086] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0087] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0088] Those skilled in the art will recognize that the units and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0089] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0090] The units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0091] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0092] The present invention is not limited to the description in the specification and embodiments, and other advantages and modifications can be readily realized by those skilled in the art. Therefore, the present invention is not limited to the specific details, representative devices and illustrated examples shown and described herein without departing from the spirit and scope of the general concept as defined by the claims and their equivalents.

Claims

1. A method for long-term intervention in online rumors, characterized in that, The method includes: Initialize user nodes in online social networks and divide them into debunkers and audiences, and load the population size and maximum number of iterations; The fitness value of each individual in the population is calculated, and the individuals are sorted in descending order based on their fitness values ​​and divided into three execution levels to achieve multi-level priority processing. A global source detection algorithm is performed using an improved Giza pyramid construction algorithm, by introducing... Flight operators and transfer functions are used to determine the positional evolution of an individual within the solution space of a social network. The crow search algorithm is used to perform pseudo-extreme adaptive avoidance. Based on the memory mechanism and the perception probability, it is determined whether the individual should make a forced displacement in order to escape the local optimum trap. The particle swarm optimization algorithm is used to perform group collaborative consensus optimization. The evolution direction of all individuals is fine-tuned through the velocity update formula, guiding the group to converge towards the globally optimal rumor-refuting parameter region. Perform boundary constraint processing and memory updates, and determine whether the termination condition is met. If it is met, output the optimal intervention strategy parameters.

2. The method for long-term intervention in online rumors according to claim 1, characterized in that, The initialization of user nodes in the online social network specifically involves: Based on differences in trust levels, audience nodes are subdivided into authority nodes, family member nodes, friend nodes, colleague nodes, and stranger nodes. Corresponding dissemination influence weights are assigned to nodes of different levels, serving as the priority basis for subsequent intervention in debunking information.

3. The method for long-term intervention in online rumors according to claim 1, characterized in that, The specific method for performing global source detection using the improved Giza pyramid construction algorithm is as follows: The Mantegna algorithm is used to simulate... flight stride ,Right now: ,in and Follows a normal distribution. For exponential parameters; Using the step size Generate a random search distance and combine it with an adjustment coefficient. The formula for updating individual positions is: ,in Current position This represents the displacement.

4. The method for long-term intervention in online rumors according to claim 3, characterized in that, The evolution of an individual's position within the solution space of a social network involves a nonlinear mapping using a transfer function. The specific steps include: Obtain the worker position update vector calculated in a continuous coordinate space by the improved Giza pyramid construction algorithm. The worker position update vector represents the evolution trend of the intervention strategy in the parameter space. use Shape transfer function or The shape transfer function maps continuous displacements to intervals. The probability value within; Based on the probability value, the audience nodes in the social network are discretized and their states are determined to determine whether the corresponding node accepts the intervention of the debunking information flow in the current iteration. By using a nonlinear mapping mechanism, the originally continuous search step size is converted into the jump probability between discrete nodes in the social network, so as to solve the problems of low optimization efficiency and discontinuity that traditional optimization algorithms are prone to when dealing with discrete social network topologies.

5. The method for long-term intervention in online rumors according to claim 4, characterized in that, The pseudo-extreme adaptive avoidance using the crow search algorithm includes the following steps: Build location memory for each debunking agent in social networks This is used to store the historical best intervention points found by the corresponding agent individual in the rumor dissemination path; Through the preset probability of perception Simulate the ability of individuals debunking rumors on social networks to identify false information and generate random numbers; If random number The strategy involves implementing a location-following approach, which updates the current search direction based on the experience of other excellent debunkers, in order to quickly approach the core source of the rumor. If random number The system implements an adaptive avoidance strategy, which determines that the current area is a rumor trap or a local pseudo-extreme point. It forces individuals to perform random flight within the solution space of the social network, simulating the debunking information actively jumping out of the information cocoon or echo barrier effect area, so as to ensure the global robustness of the optimization process.

6. The method for long-term intervention in online rumors according to claim 1, characterized in that, The specific steps for performing group collaborative consensus optimization using the particle swarm optimization algorithm include: The evolution of group cognition in social networks is abstracted as the motion process of a particle swarm, defining that the speed update of each individual is affected by individual cognitive factors. Social experience factors Driven by both; among them, This corresponds to the independent judgment of facts by social network nodes based on their own trust logic. Consensus offset caused by the influence of authoritative nodes and the chain of trust on the corresponding nodes; By using the velocity update formula to adjust the evolution speed and displacement direction of all nodes in the cognitive space in real time, the audience in the social network is guided to converge toward the global optimal solution under the interference of multi-level trust weights. Through iterative evolution, social network nodes in a chaotic state gradually reach a collective consensus, thereby completely blocking the chain of rumor propagation.

7. The method for long-term intervention in online rumors according to claim 1, characterized in that, The specific steps for calculating an individual's fitness value include: A multidimensional fitness function based on the social network background is constructed, which considers the coverage depth of the debunking information, the propagation attenuation factor, and the trust difference of the audience nodes. An evolutionary factor mechanism is introduced to dynamically assess the probability of each node transforming from a victim of rumors into a disseminator of facts, and this probability is quantified into an interval. The value within; Calculate the average fitness of all individuals globally. When the fitness value of the audience node monotonically increases and converges to the threshold 1, it is determined that the social network has formed a long-term rumor immunity mechanism. If the maximum number of iterations or error convergence accuracy is met, then the current node weight distribution and intervention parameters are locked as the final optimal intervention strategy output.

8. A system for implementing the long-term intervention method for online rumors according to any one of claims 1 to 7, characterized in that, The system includes: The data acquisition and preprocessing module is used to capture information flow data from online social networks in real time, identify the main propagators and audience nodes in the rumor propagation chain, and complete the attribute labeling and classification of node trust level based on the interaction frequency and historical trust endorsement between nodes. The initialization and parameter configuration module is used to load the search solution space range of the intervention strategy, set the population size and maximum number of iterations of the preset hybrid intelligent optimization model, and configure the awareness probability of the crow search algorithm according to the urgency of the current public opinion. ; The intervention strategy output module is used to generate precise intervention plans for nodes with different trust levels based on the global optimal solution output by the preset hybrid intelligent optimization model. These plans include intervention timing, information weight distribution, and key propagation path recommendations, in order to achieve long-term blocking of online rumors. The preset hybrid intelligent optimization model integrates: A global detection unit is used in conjunction with an improved Giza pyramid construction algorithm. Flight operators perform a wide-ranging search for debunking intervention points in discrete topologies of social networks; The adaptive avoidance unit utilizes the memory mechanism of the crow search algorithm to monitor and forcibly avoid local pseudo-extreme points in the solution space in real time, so as to prevent the intervention strategy from getting trapped in an information cocoon. The group consensus coordination unit uses the speed update logic of the particle swarm optimization algorithm to simulate the consensus shift process of a social group toward the truth, guiding the overall nodes to converge toward the optimal cognitive state. The convergence judgment unit is used to perform boundary constraint processing and memory update, and to determine whether the termination condition is met. If it is met, the optimal intervention strategy parameters are output.

Images