Proxy opinion prediction method and device in social network and electronic equipment

Abstract

The invention provides an agent opinion prediction method and device in a social network and electronic equipment. The method comprises the following steps: acquiring current information of at least one information source at a current moment; obtaining agency opinions of the current opinion leader based on the current information and the information preference of the opinion leader for the information source; determining a current common agency opinion based on the current opinion leader agency opinion; iteratively executing the step of obtaining the agency opinion of the current opinion leader based on the current information and the information preference of the opinion leader for the information source, and the step of determining the current common agency opinion based on the agency opinion of the current opinion leader until the obtained agency opinion of the current opinion leader reaches the opinion state steady state; the opinion state is stable until the opinion state of the current common agent is stable; a steady-state opinion leader agent opinion is predicted based on the current opinion leader agent opinion, and a steady-state common agent opinion is predicted based on the current common agent opinion. Accurate prediction of opinion leader agency opinions and common agency opinions is realized.

Description

Technical Field

[0001] This invention relates to the field of proxy opinion prediction technology, and in particular to a method, apparatus and electronic device for proxy opinion prediction in social networks. Background Technology

[0002] According to relevant technologies, the information dissemination process in real social networks generally follows a two-step dissemination model of "information source → opinion leader → ordinary agent".

[0003] However, existing classical opinion dynamics models have significant limitations in their architectural design. Most models do not clearly distinguish the heterogeneous roles of opinion leaders and ordinary agents, and assume that information can be applied directly to all agent nodes without discrimination. This results in a lack of characterization of the special function of opinion leaders as information filters and amplifiers in the two-step propagation path, which in turn limits the model's ability to predict agent opinions. Summary of the Invention

[0004] This invention provides a method, apparatus, and electronic device for predicting proxy opinions in social networks, which accurately depicts the hierarchical structure of information dissemination and opinion exchange in real social networks, and ensures accurate prediction of proxy opinions of opinion leaders and ordinary proxy opinions.

[0005] This invention provides a method for predicting proxy opinions in a social network. The method includes: acquiring current information from at least one information source at the current moment, wherein the information source is an information source about a preset social event; obtaining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the information preference of the opinion leader for the information source, wherein the information preference is used to characterize the confirmation bias of the opinion leader; determining the current ordinary proxy opinion of a common agent at the current moment based on the current opinion leader proxy opinion; iteratively executing the step of obtaining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the information preference of the opinion leader for the information source, up to the step of determining the current ordinary proxy opinion of a common agent at the current moment based on the current opinion leader proxy opinion, until the obtained current opinion leader proxy opinion reaches a steady state of opinion, and the current ordinary proxy opinion reaches a stable state of opinion; predicting the steady-state opinion leader proxy opinion of the opinion leader for the preset social event based on the current opinion leader proxy opinion, and predicting the steady-state ordinary proxy opinion of the common agent for the preset social event based on the current ordinary proxy opinion.

[0006] According to a method for predicting proxy opinions in a social network provided by the present invention, before obtaining the current proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, the method further includes: obtaining the initial proxy opinion of the opinion leader for the preset social event, and the opinion leader's opinion leader characteristic parameters, wherein the opinion leader characteristic parameters are used to describe the degree of adherence of the opinion leader to its own opinion; obtaining the current proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source includes: obtaining the current proxy opinion of the opinion leader at the current moment based on the current information, the initial proxy opinion, the opinion leader characteristic parameters, and the opinion leader's information preference for the information source.

[0007] According to a method for predicting proxy opinions in a social network provided by the present invention, the information sources for a preset social event include multiple information sources; before obtaining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information, the initial opinion leader proxy opinion, the opinion leader characteristic parameters, and the opinion leader's information preference for the information sources, the method further includes: determining each information preference of the opinion leader for each of the information sources; for any one of the information sources, normalizing the information preference of the information source based on each information preference to obtain a normalized information preference; using the normalized information preference of the information source as the influence weight of the information source on the opinion leader; the method further includes: determining the current opinion leader proxy opinion of the opinion leader based on the current information, the initial opinion leader proxy opinion, the opinion leader characteristic parameters, and the opinion leader's information preference for the information sources; and determining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information, the initial opinion leader proxy opinion, the opinion leader characteristic parameters, and the opinion leader's information preference for the information sources. The method involves taking the opinion leader's opinion, the opinion leader's characteristic parameters, and the opinion leader's information preference for the information source, to obtain the current opinion leader's proxy opinion at the current moment. This includes: multiplying the influence weight of any information source on the opinion leader with the current information; obtaining multiple parameters after multiplication; summing the multiple parameters after multiplication to obtain a weighted average opinion information; determining a first weight for the initial opinion leader's proxy opinion and a second weight for the weighted average opinion information based on the opinion leader's characteristic parameters; and obtaining the current opinion leader's proxy opinion at the current moment based on the first weight, the initial opinion leader's proxy opinion, the second weight, and the weighted average opinion information.

[0008] According to a method for predicting agent opinions in a social network provided by the present invention, before determining the current ordinary agent opinion of an ordinary agent at the current moment based on the current opinion leader agent opinion, the method further includes: obtaining the initial ordinary agent opinion of the ordinary agent on the preset social event, the previous moment's neighbor ordinary agent opinion of the neighbor ordinary agent at the previous moment, the ordinary agent characteristic parameters of the ordinary agent, the influence ratio of the neighbor ordinary agent, and the influence ratio of the opinion leader, wherein the ordinary agent characteristic parameters are used to describe the degree of adherence of the ordinary agent to its own opinion; the neighbor ordinary agent is an ordinary agent that has information interaction with the ordinary agent; the sum of the ordinary agent characteristic parameters, the influence ratio of the neighbor ordinary agent, and the influence ratio of the opinion leader is 1; determining the current ordinary agent opinion of an ordinary agent at the current moment based on the current opinion leader agent opinion includes: performing a weighted summation of the initial ordinary agent opinion, the ordinary agent characteristic parameters, the previous moment's neighbor ordinary agent opinion, the influence ratio of the neighbor ordinary agent, the current opinion leader agent opinion, and the influence ratio of the opinion leader to obtain the current ordinary agent opinion of the ordinary agent at the current moment.

[0009] According to the method for predicting agent opinions in a social network provided by the present invention, the neighbor ordinary agent is a plurality of neighbor ordinary agents, and correspondingly, the previous moment neighbor ordinary agent opinion is a plurality of previous moment neighbor ordinary agent opinions; the previous moment neighbor ordinary agent opinion is determined by the following method: for any one of the neighbor ordinary agents, the influence weight of the neighbor ordinary agent on the ordinary agent is determined, and the previous moment neighbor ordinary agent opinion and the influence weight of the neighbor ordinary agent on the ordinary agent are weighted and summed to obtain the previous moment neighbor ordinary agent opinion.

[0010] According to a method for predicting proxy opinions in a social network provided by the present invention, before obtaining the current proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, the method further includes: obtaining the opinion leader's proxy opinion at the previous moment; determining the opinion leader's information preference for the information source based on the distance between the previous proxy opinion and the current information; iteratively executing the step of obtaining the current proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, up to the step of determining the current ordinary proxy opinion of the ordinary proxy at the current moment based on the current proxy opinion of the opinion leader, until obtaining the current proxy opinion of the opinion leader at the current moment. The process continues until the current opinion leader's proxy opinion reaches a stable opinion state, and the current ordinary proxy opinion reaches a stable opinion state. This includes: taking the opinion leader's proxy opinion at the current moment as the previous opinion leader's proxy opinion at the next moment, and repeatedly iterating the step of obtaining the previous opinion leader's proxy opinion at the previous moment until the step of determining the current ordinary proxy opinion at the current moment is completed. This continues until the difference between the current opinion leader's proxy opinion at the current moment and the previous opinion leader's proxy opinion at the previous moment is less than a threshold, and the difference between the current ordinary proxy opinion at the current moment and the previous ordinary proxy opinion at the previous moment is less than a threshold.

[0011] This invention also provides a device for predicting proxy opinions in a social network. The device includes: an acquisition module for acquiring current information from at least one information source at the current moment, wherein the information source is an information source about a preset social event; a generation module for obtaining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, wherein the information preference is used to characterize the opinion leader's confirmation bias; a determination module for determining the current ordinary proxy opinion of the ordinary proxy at the current moment based on the current opinion leader proxy opinion; and a processing module for iteratively executing the current information... The system obtains the current opinion leader's proxy opinion steps at the current moment, based on the current opinion leader's proxy opinion, and determines the current ordinary agent's proxy opinion steps at the current moment, until the obtained current opinion leader's proxy opinion reaches a steady state and the current ordinary agent's proxy opinion reaches a stable state; the prediction module is used to predict the steady-state opinion leader's proxy opinion on the preset social event based on the current opinion leader's proxy opinion, and to predict the steady-state ordinary agent's proxy opinion on the preset social event based on the current ordinary agent's proxy opinion.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the agent opinion prediction method in the social network as described above.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the agent opinion prediction method in a social network as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the agent opinion prediction method in a social network as described above.

[0015] This invention provides a method, apparatus, and electronic device for predicting proxy opinions in a social network. The method involves acquiring current information from at least one information source, where the information source is about a preset social event. Based on the current information and the opinion leader's information preference for the information source, the method obtains the current opinion leader's proxy opinion at the current moment. Based on the current opinion leader's proxy opinion, it determines the current ordinary proxy opinion of ordinary agents at the current moment. The method iteratively executes the steps of obtaining the current opinion leader's proxy opinion based on the current information and the opinion leader's information preference for the information source, up to the step of determining the current ordinary proxy opinion of ordinary agents at the current moment, until the obtained current opinion leader's proxy opinion reaches a steady state and the current ordinary proxy opinion reaches a stable state. Based on the current opinion leader's proxy opinion, the method predicts the steady-state opinion leader's proxy opinion for the preset social event, and based on the current ordinary proxy opinion, it predicts the steady-state ordinary proxy opinion of ordinary agents for the preset social event. This method accurately depicts the hierarchical structure of information dissemination and opinion exchange in real social networks, ensuring accurate prediction of opinion leader's proxy opinion and ordinary proxy opinion. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is one of the flowcharts illustrating the proxy opinion prediction method in social networks provided by this invention.

[0018] Figure 2 This is a schematic diagram of the process provided by the present invention for obtaining the current opinion leader's proxy opinion at the current moment based on the current information, the initial opinion leader's proxy opinion, the opinion leader's characteristic parameters, and the opinion leader's information preference for the information source.

[0019] Figure 3 This is the second flowchart of the proxy opinion prediction method in social networks provided by the present invention.

[0020] Figure 4 This is a schematic diagram of the structure of the proxy opinion prediction device in the social network provided by the present invention.

[0021] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] This invention provides a method for predicting proxy opinions in social networks, aiming to address the core technical problem that existing opinion dynamics models struggle to accurately depict the hierarchical structure of information dissemination and opinion exchange in real social networks, as well as the prevalent confirmation bias behavior among proxies. This invention constructs a three-layer network model comprising information sources, opinion leaders, and ordinary proxies, and introduces a mathematically tractable information preference model based on the beta distribution kernel function to quantify confirmation bias. This allows for precise analysis of information flow and opinion evolution during the two-step dissemination process, providing effective theoretical tools and technical means for understanding and guiding public opinion in social networks. A social network refers to a network structure composed of proxies (such as agents or institutions) and their interactive relationships (such as following and forwarding), used to describe the dynamic process of information dissemination and opinion exchange.

[0024] Figure 1 This is one of the flowcharts illustrating the proxy opinion prediction method in social networks provided by this invention.

[0025] The following will combine Figure 1 The process of the proxy opinion prediction method in social networks provided by this invention is described.

[0026] In an exemplary embodiment of the present invention, combined with Figure 1 As can be seen, the proxy opinion prediction method in social networks may include steps 110 to 150, which will be described in detail below.

[0027] In step 110, current information at the current moment from at least one information source is obtained, wherein the information source is an information source about a preset social event.

[0028] In one embodiment, the preset social event can be a hot social issue. A message source refers to an entity (such as a media organization or government department) in a social network that generates and publishes information; this information is a random variable used to influence opinion leaders. Message sources can be preset to news media, expert blogs, or authoritative platforms that are related to the issue and have different stances. For example, message source A tends to support the technology, while message source B tends to hold a conservative or critical attitude. At the start of the simulation or prediction (t=0), the system captures or simulates initial text, data, or opinion information from these preset message sources as the current information.

[0029] In step 120, based on the current information and the opinion leader's information preference for the information source, the current opinion leader's proxy opinion at the current moment is obtained, wherein the information preference is used to characterize the opinion leader's confirmation bias.

[0030] In one embodiment, the current opinion leader proxy opinion of the opinion leader at the current moment can be obtained based on the current information (e.g., the quantitative value of the opinions published by each information source) and the information preferences of the opinion leader. The current opinion leader proxy opinion is formed after the opinion leader has absorbed the opinions of each information source at the current moment and filtered by its own confirmation bias.

[0031] Among them, an opinion leader is an agent (such as an expert or public figure) with significant influence in a social network who can interpret information and guide the formation of opinions among ordinary agents; a normal agent is an agent in a social network whose opinions are influenced by opinion leaders, neighbors, and friends; and confirmation bias refers to the agent's tendency to accept information that is similar to their own opinions while ignoring or rejecting inconsistent information, which is manifested as a limited willingness to accept information during the opinion updating process.

[0032] In step 130, based on the opinion leader's opinion, the current opinion of the ordinary agent at the current moment is determined.

[0033] In one embodiment, each ordinary agent has a follow or trust link with one or more opinion leaders. The current agent opinions of all opinion leaders can be aggregated based on the network connection strength between the ordinary agent and the opinion leaders to obtain the current ordinary agent opinion formed at the current moment. Ordinary agents themselves do not have strong pre-set information preferences; their opinions are mainly influenced by the opinion leaders they follow, and they do not have their own confirmation bias.

[0034] In step 140, the process iteratively executes the current opinion leader proxy opinion step based on the current information and the opinion leader's information preference for the information source, and then determines the current ordinary agent opinion step based on the current opinion leader proxy opinion, until the current opinion leader proxy opinion reaches a steady state and the current ordinary agent opinion reaches a stable state.

[0035] In step 150, based on the current opinion leader proxy opinion, the steady-state opinion leader proxy opinion of the opinion leader on the preset social event is predicted, and based on the current ordinary proxy opinion, the steady-state ordinary proxy opinion of the ordinary agent on the preset social event is predicted.

[0036] In one embodiment, the current opinion of ordinary agents, along with the currently updated information from the information source, can be used as input for the next time step (t=1), and steps S120 to S130 can be re-executed. That is, at time t=1, opinion leaders will form new opinions based on the current information and information preferences; ordinary agents will then update their own opinions based on the new opinions of the opinion leaders at time t=1. This process is repeated continuously.

[0037] Furthermore, after each iteration, the average change in the opinions of all opinion leaders and all ordinary agents in the entire network can be calculated. When both changes are below a preset minimum threshold for several consecutive iterations, the network opinion state is considered to have reached a steady state, and the iteration stops. The final opinion of each opinion leader is predicted as their steady-state opinion on a preset social event. The final opinion of each ordinary agent is predicted as their steady-state opinion on a preset social event. Through these steps, this embodiment completes the prediction of the final opinions of each agent in the social network regarding a specific event.

[0038] In this embodiment, a closed-loop iterative mechanism—information acquisition, leader preference filtering, ordinary agent following, and iterative feedback—can simulate the complete dynamic process of opinion propagation, collision, and evolution in social networks, rather than performing static snapshot analysis. Ultimately, mathematical judgment is used to determine convergence to a steady state, thereby enabling a more scientific and accurate prediction of the long-term equilibrium state of public opinion, providing a forward-looking basis for public opinion analysis and decision-making.

[0039] This invention provides a method for predicting proxy opinions in social networks. The method involves acquiring current information from at least one information source, where the information source is about a preset social event. Based on the current information and the opinion leader's information preference for the information source, the method obtains the current opinion leader's proxy opinion at the current moment. Based on the current opinion leader's proxy opinion, the method determines the current ordinary proxy opinion of ordinary agents at the current moment. The method iteratively executes the steps of obtaining the current opinion leader's proxy opinion based on the current information and the opinion leader's information preference for the information source, up to the step of determining the current ordinary proxy opinion of ordinary agents at the current moment, until both the current opinion leader's proxy opinion and the current ordinary proxy opinion reach a stable opinion state. Based on the current opinion leader's proxy opinion, the method predicts the steady-state opinion leader's proxy opinion regarding the preset social event, and predicts the steady-state ordinary proxy opinion of ordinary agents regarding the preset social event. This method accurately depicts the hierarchical structure of information dissemination and opinion exchange in real social networks, ensuring accurate prediction of opinion leader's proxy opinion and ordinary proxy opinion.

[0040] In yet another exemplary embodiment of the present invention, continuing with the previously described embodiments as an example, before obtaining the current opinion leader proxy opinion of the opinion leader at the current moment (corresponding to step 120) based on the current information and the opinion leader's information preference for the information source, the proxy opinion prediction method in the social network may further include the following steps: Obtain the initial opinion leaders' proxy opinions on a pre-defined social event, as well as the opinion leaders' characteristic parameters, whereby the opinion leaders' characteristic parameters are used to describe the degree to which opinion leaders adhere to their own opinions. The current opinion leader's proxy opinion at the current moment, based on current information and the opinion leader's information preference for the information source, can be obtained in the following way: Based on current information, initial opinion leader proxy opinions, opinion leader characteristic parameters, and opinion leaders' information preferences for information sources, we obtain the current opinion leader proxy opinions at the current moment.

[0041] In one embodiment, the initial opinion leader proxy opinion can be set as an initial opinion value for each opinion leader proxy regarding the preset social event. This value can be quantified based on the opinion leader's historical statements, public stance, or background information. Alternatively, a characteristic parameter can be set for each opinion leader, denoted as... (0 ≤ α ≤ 1). This parameter is used to quantify the degree to which the opinion leader p adheres to his or her existing opinions. The higher the value, the more stubborn the opinion leader is, and the less likely they are to change their views due to new external information; The lower the value, the more open the individual is, and the easier it is for them to adjust their views based on new information. Indicates that he has not changed his opinion at all. Opinions are entirely shaped by information.

[0042] In another embodiment, at the current moment, the opinion leader's current proxy opinion = ×Initial opinion leader's proxy opinion+ × (Information preference vector × Current information vector). Where, That is, the opinion leader characteristic parameters (degree of adherence); the initial opinion leader proxy opinion represents the leader's preset initial position; This indicates the leader's receptiveness to external information; (information preference vector × current information vector) can represent the influence value of external information formed after filtering external information based on information preference. By introducing "initial opinion" and "feature parameter α", this embodiment upgrades the opinion leader from a simple information filter to a cognitive subject with memory and personality traits. This is more in line with the fact that an individual's existing views and personality stubbornness are key factors influencing their acceptance of new information, thereby ensuring the accuracy of the final predicted steady-state opinion leader proxy opinion and steady-state ordinary proxy opinion.

[0043] Figure 2 This is a schematic diagram of the process provided by the present invention for obtaining the current opinion leader's proxy opinion at the current moment based on the current information, the initial opinion leader's proxy opinion, the opinion leader's characteristic parameters, and the opinion leader's information preference for the information source.

[0044] The following will combine Figure 2 The present invention describes the process by which the current opinion leader's proxy opinion is obtained at the current moment based on the current information, the initial opinion leader's proxy opinion, the opinion leader's characteristic parameters, and the opinion leader's information preference for the information source.

[0045] In an exemplary embodiment of the present invention, the information sources regarding a preset social event may include multiple information sources. (Combined with...) Figure 2 It can be seen that, based on the current information, the initial opinion leader proxy opinion, the opinion leader characteristic parameters, and the opinion leader's information preference for the information source, obtaining the current opinion leader proxy opinion at the current moment may include steps 210 to 270, and each step will be described below.

[0046] In step 210, the opinion leaders’ information preferences for each information source are determined.

[0047] In one embodiment, for a preset social event, there are N different information sources, wherein the information generated by the information sources at time t is modeled as a random vector. Let N be the number of information sources. Scalars 0 and 1 represent two extremely opposing pieces of information supporting a particular issue. Assume... At each time t, it follows a time-invariant probability distribution (called the information distribution), whose mean is denoted as . To simplify the analysis, it is assumed that information from different sources is independent at different times. The opinion leader's opinion at time t is modeled as a random vector. , where P is the number of opinion leaders.

[0048] For each opinion leader, an information preference for each information source can be assigned to them. This value is a non-negative real number, representing the leader's inherent trust in or tendency to focus on the information source.

[0049] Among them, the p-th opinion leader's information preference for the n-th information source at time t It can be expressed by the following formula (1): (1) in, This indicates the opinion leader's opinion at the previous moment. With current information The square of the distance between them. and It is an information preference parameter that quantifies the degree of confirmation bias. smaller and The larger the value, the stronger the confirmation bias (i.e., the more inclined one is to accept information that is similar to one's own opinion).

[0050] In step 220, for any information source, the information preferences of the information source are normalized based on each information preference to obtain the normalized information preferences.

[0051] In step 230, the normalized information preferences of the information sources are used as the influence weight of the information sources on opinion leaders.

[0052] In one embodiment, information preferences can be normalized to obtain the influence weight of the nth information source on the pth opinion leader. Among them, the influence weight of information sources on opinion leaders This can be expressed as formula (2): (2) in, Indicates the information source's ID; influence weight. This reflects the actual proportion of influence of different information sources on opinion leaders under the influence of confirmation bias.

[0053] In step 240, the influence weight of any information source on the opinion leader is multiplied by the current information to obtain multiple multiplied parameters.

[0054] In step 250, the parameters after multiple multiplication are summed to obtain the weighted average information of opinions.

[0055] In step 260, based on the opinion leader characteristic parameters, the first weight of the initial opinion leader proxy opinion and the second weight of the opinion weighted average information are determined respectively.

[0056] In step 270, based on the first weight, the initial opinion leader proxy opinion, the second weight, and the opinion weighted average information, the current opinion leader proxy opinion at the current moment is obtained.

[0057] In one embodiment, the opinion of the p-th opinion leader at time t It is a linear combination of its initial opinion and weighted average information, where the opinion leader's current opinion at the current moment can be expressed as formula (3): (3) in, It is the degree of stubbornness of this opinion leader. Indicates that he has not changed his opinion at all. Opinions are entirely shaped by information. This is their initial opinion, that is, the opinion of the initial opinion leader acting on their behalf.

[0058] Understandable, This represents the parameters obtained after multiple multiplication processes; This represents a weighted average of the opinions received. Indicates the first weight; This indicates the second weight.

[0059] In this embodiment, the abstract concept of opinion leader confirmation bias is explicitly transformed into influence weights in the form of a probability distribution through normalization. A complete and deterministic mathematical operation chain is provided, from multiplying the weights by the information values ​​to summing them, and then to weighted synthesis with the opinions themselves. This eliminates ambiguity in algorithm implementation and ensures the accuracy of the final predicted steady-state opinion leader proxy opinions and steady-state ordinary proxy opinions.

[0060] In yet another exemplary embodiment of the present invention, continuing with the previously described embodiments as an example, before determining the current ordinary agent's opinion at the current moment based on the current opinion leader's agent opinion (corresponding to step 130), the agent opinion prediction in the social network may further include the following steps: The system acquires the initial opinion of ordinary agents on a preset social event, the opinion of neighboring ordinary agents at the previous time step, the ordinary agent characteristic parameters of the ordinary agents, the influence ratio of neighboring ordinary agents, and the influence ratio of opinion leaders. The ordinary agent characteristic parameters describe the degree to which ordinary agents adhere to their own opinions. Neighboring ordinary agents are ordinary agents that interact with ordinary agents. The sum of the ordinary agent characteristic parameters, the influence ratio of neighboring ordinary agents, and the influence ratio of opinion leaders is 1. Among them, determining the current ordinary agent's opinion at the current moment based on the opinion leader's opinion can be achieved in the following way: The initial ordinary agent opinion, ordinary agent characteristic parameters, the neighbor ordinary agent opinion at the previous time step, the neighbor ordinary agent influence ratio, the current opinion leader agent opinion, and the opinion leader influence ratio are weighted and summed to obtain the current ordinary agent opinion at the current time step.

[0061] In another exemplary embodiment of the present invention, there can be multiple neighbor ordinary agents, and correspondingly, there can be multiple neighbor ordinary agent opinions at the previous moment; the neighbor ordinary agent opinions at the previous moment are determined in the following manner: For any neighboring ordinary agent, determine the influence weight of the neighboring ordinary agent on the ordinary agent, and The neighbor general agent opinions of each previous time step and the influence weights of the neighbor general agents on the general agent are weighted and summed to obtain the neighbor general agent opinions of the previous time step.

[0062] Among them, the opinion of the neighbor's ordinary agent at the previous moment can also be expressed as ,in, It is the influence weight of the q'th ordinary agent on the qth ordinary agent, that is, the influence weight of the neighboring ordinary agents on the ordinary agent.

[0063] In another exemplary embodiment of the present invention, there can be multiple opinion leaders, and correspondingly, the proxy opinion of the current opinion leader can be multiple proxy opinions of the current opinion leader; the proxy opinion of the current opinion leader can be determined in the following way: For any one of the aforementioned opinion leaders, determine the influence weight of the opinion leader on ordinary agents, and The current opinion leader's proxy opinion is obtained by weighting and summing the influence weights of each current opinion leader's proxy and the opinion leader's influence on ordinary proxies.

[0064] Among them, the proxy opinion of current opinion leaders can also be expressed as ,in, It is the influence weight of the p-th opinion leader on the q-th ordinary agent, that is, the influence weight of the opinion leader on the ordinary agent.

[0065] In yet another embodiment, the opinion of a regular agent at time t can be modeled as a random vector. , where Q is the number of ordinary agents. Its opinion updates are based on the Friedkin-Johnsen model, incorporating the influence of opinion leaders. The opinion of the q-th ordinary agent at time t... It is a linear combination of its initial opinion, the opinion of the neighboring ordinary agent at the previous moment, and the current opinion of the opinion leader. It can be understood that the current ordinary agent opinion at the current moment can be expressed as formula (4): (4) in, These are the degree of stubbornness of the ordinary agent, the proportion of influence of the ordinary agent, and the proportion of influence of the opinion leader, respectively, and they meet the following conditions: . It is the first The influence weight of each ordinary agent on the q-th ordinary agent. It is the influence weight of the p-th opinion leader on the q-th ordinary agent. This is its initial opinion, also known as the initial ordinary agency opinion.

[0066] These can be the characteristic parameters of a regular proxy; It could be the proportion influenced by ordinary agents in the neighborhood; This could be the proportion of influence of opinion leaders; It could be the opinion of a neighbor's ordinary agent from the previous moment; This indicates a neighbor's ordinary agent.

[0067] In this embodiment, by clearly distinguishing and quantifying the three core dimensions of social influence—self-conservatism, peer influence, and authority influence—it completely transcends the crude model that treats ordinary agents as simple followers. It accurately portrays the complex result of individual opinions being the combined effect of vertical authority influence and horizontal peer contagion within a social network, weighed against one's own cognitive inertia, greatly enhancing the model's realism and explanatory power.

[0068] Figure 3 This is the second flowchart of the proxy opinion prediction method in social networks provided by the present invention.

[0069] The following will combine Figure 3 The process of another proxy opinion prediction method in social networks provided by the present invention will be described.

[0070] In yet another exemplary embodiment of the present invention, combined with Figure 3 As can be seen, the proxy opinion prediction method in social networks may include steps 310 to 350, which will be described in detail below.

[0071] In step 310, the opinion leader's proxy opinion at the previous time step is obtained.

[0072] In step 320, the opinion leader's information preference for the information source is determined based on the distance between the opinion leader's proxy opinion at the previous moment and the current information.

[0073] In one embodiment, the opinion leader's proxy opinion at the previous time can be obtained, and the opinion leader's information preference for the information source can be determined based on the distance between the previous opinion leader's proxy opinion and the current information. The information preference can be determined using the formula (1) mentioned above.

[0074] In step 330, based on the current information and the opinion leader's information preference for the information source, the current opinion leader's proxy opinion at the current moment is obtained.

[0075] In one embodiment, the current opinion leader's proxy opinion at the current moment can be obtained based on the current information and the opinion leader's information preference for the information source. The current opinion leader's proxy opinion at the current moment can be determined using the formula (3) mentioned above.

[0076] In step 340, based on the opinion leader's opinion, the current opinion of the ordinary agent at the current moment is determined.

[0077] In another embodiment, the current ordinary agent opinion at the current moment can be determined based on the current opinion leader agent opinion, wherein the current ordinary agent opinion at the current moment can be determined using the formula (4) above.

[0078] In step 350, the opinion leader's proxy opinion at the current moment is taken as the opinion leader's proxy opinion at the previous moment in the next moment. The step of obtaining the opinion leader's proxy opinion at the previous moment is repeatedly and iteratively executed until the current ordinary agent's opinion at the current moment is determined. This continues until the difference between the current opinion leader's proxy opinion at the current moment and the opinion leader's proxy opinion at the previous moment is less than a threshold, and the difference between the current ordinary agent's opinion at the current moment and the ordinary agent's proxy opinion at the previous moment is less than a threshold.

[0079] In one embodiment, the opinion leader's proxy opinion at the current moment can be used as the opinion leader's proxy opinion at the previous moment in the next moment. The step of obtaining the opinion leader's proxy opinion at the previous moment (i.e., step 310) is repeatedly executed until the current ordinary agent's opinion at the current moment is determined (i.e., step 340), until the obtained current opinion leader's proxy opinion reaches a steady state and the current ordinary agent's opinion reaches a stable state.

[0080] Among them, the steady-state condition of opinion leaders can be that for each opinion leader in the network, the absolute value of the difference between the current opinion leader proxy opinion calculated at the current time t and the opinion leader proxy opinion at the previous time t-1 is less than a preset threshold. The steady-state condition for a normal agent can be that for each normal agent in the network, the absolute value of the difference between the current normal agent opinion calculated at the current time t and the normal agent opinion at the previous time t-1 is less than a preset threshold.

[0081] The entire iterative cycle terminates and the system reaches steady state only when the changes in opinions of all agents, including all opinion leaders and all ordinary agents, simultaneously satisfy their respective threshold conditions.

[0082] This embodiment successfully embeds the classic selective exposure and confirmation bias mechanisms from social psychology into a computational model by defining information preference as a function related to the distance between one's existing opinions and external information. Opinion leaders no longer hold fixed source preferences, but dynamically become more inclined to information sources that are similar to their current views. This more profoundly simulates the formation process of the echo chamber effect and self-reinforcing bias, thereby ensuring accurate prediction of opinion leader proxy opinions and ordinary proxy opinions.

[0083] As another variation of the present invention, an important contribution of the present invention is that when the number of information sources N is large enough, it is possible to derive approximate analytical solutions for the steady-state opinions of opinion leaders and ordinary agents, thereby supporting in-depth theoretical analysis.

[0084] Among them, the steady-state opinions of opinion leader p It can be approximated by formula (5): (5) in, It is a composite parameter that takes into account the degree of stubbornness. and information preference parameters The impact. It is a constant obtained through curve fitting, which ensures that the approximate solution maintains high accuracy under different information distributions. This indicates that convergence is almost certain. The formula shows that the steady-state opinion of opinion leaders is a weighted average of their initial opinions and the mean of the information distribution, with the weights... It is determined by both the degree of stubbornness and the confirmation bias.

[0085] Steady-state opinion of ordinary agent q This can be expressed as formula (6): (6) in, It is the sample mean of the initial opinions of all ordinary agents. It is the sample mean of the steady-state opinions of all opinion leaders. This indicates that convergence is almost certain. The formula shows that the steady-state opinion of a normal agent is its own initial opinion. Average initial opinion of the general agent group and the average steady-state opinion of opinion leaders The weighted average of the three. The weights are determined by their respective degrees of stubbornness. and the proportion of influence of opinion leaders A joint decision.

[0086] Based on the analytical form of the above steady-state solution, this invention can theoretically and rigorously analyze the impact of various system parameters on group opinions. The sample mean reflects the average opinion tendency, and the sample variance reflects the degree of consensus. Information distribution mean It is positively correlated with the mean of steady-state opinion samples of opinion leaders and ordinary agents; Initial opinion Its sample mean and sample variance positively influence the sample mean and sample variance of steady-state opinions; Stubbornness The higher the degree of stubbornness, the more the steady-state opinion tends to be based on the initial opinion rather than external influences (information or opinion leaders), which leads to an increase in the sample variance of the steady-state opinion; Information preference parameters When the level of stubbornness is moderate, increase or reduce (That is, reducing confirmation bias) will reduce the sample variance of opinion leaders' steady-state opinions, making their opinions more consistent.

[0087] Through approximate analytical solutions of steady-state opinions and theoretical analysis of the impact of system parameters, it can be seen that, given a sufficiently large number of information sources, the steady-state opinions of opinion leaders can be derived. The approximate analytical solution, where the key composite parameters A combination of stubbornness and information preference parameters The method also includes determining constants through curve fitting. and The specific implementation method is designed to ensure the universality and accuracy of the approximate solution under different information distributions. A general agent steady-state opinion is derived. The analytical solution is derived. Based on the steady-state analytical solutions for opinion leaders and ordinary agents, a complete theoretical analysis system is established to quantitatively analyze the impact of system parameters such as information distribution mean, initial opinion, degree of obstinacy, and confirmation bias parameter on the statistical characteristics of group opinion mean, variance, etc.

[0088] As described above, this invention provides a method for predicting proxy opinions in social networks. Based on a hierarchical network modeling method using a two-step propagation process, it constructs a social network model containing three heterogeneous roles: information source, opinion leader, and ordinary agent. This model accurately corresponds to the two-stage information flow process of "information source → opinion leader → ordinary agent" in the real world. This modeling method is the foundation for accurately simulating the evolution of opinions in real social networks. The confirmation bias quantification model based on the beta distribution kernel function innovatively introduces a parameterized information preference function to quantify the confirmation bias of opinion leaders. This model utilizes the characteristics of the beta distribution kernel function to flexibly and interpretably characterize confirmation bias behavior of different intensities through two parameters with clear physical meaning. In particular, when the distance between opinion and information is zero, the function value tends to infinity, thus more realistically simulating the "echo chamber" effect.

[0089] The following describes the proxy opinion prediction device in a social network provided by the present invention. The proxy opinion prediction device in a social network described below can be referred to in correspondence with the proxy opinion prediction method in a social network described above.

[0090] Figure 4 This is a schematic diagram of the structure of the proxy opinion prediction device in the social network provided by the present invention.

[0091] The following will combine Figure 4 The structure of the proxy opinion prediction device in the social network provided by the present invention will be described.

[0092] In an exemplary embodiment of the present invention, combined with Figure 4 As can be seen, the proxy opinion prediction device in a social network may include an acquisition module 410, a generation module 420, a determination module 430, a processing module 440, and a prediction module 450. Each module will be described below.

[0093] The acquisition module 410 can be configured to acquire current information at the current moment from at least one information source, wherein the information source is an information source about a preset social event; The generation module 420 can be configured to obtain the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, wherein the information preference is used to characterize the opinion leader's confirmation bias; The determination module 430 can be configured to determine the current ordinary agent opinion at the current moment based on the current opinion leader agent opinion; Processing module 440 can be configured to iteratively execute steps based on the current information and the opinion leader's information preference for the information source to obtain the current opinion leader's proxy opinion at the current moment, and to determine the current ordinary agent's opinion at the current moment based on the current opinion leader's proxy opinion, until the obtained current opinion leader's proxy opinion reaches a steady state and the current ordinary agent's opinion reaches a stable state. The prediction module 450 can be configured to predict, based on the current opinion leader proxy opinion, the steady-state opinion leader proxy opinion of the opinion leader regarding the preset social event, and based on the current ordinary proxy opinion, the steady-state ordinary proxy opinion of the ordinary proxy regarding the preset social event.

[0094] In an exemplary embodiment of the present invention, the generation module 420 may further be configured to: The initial opinion leader proxy opinion of the opinion leader on the preset social event is obtained, as well as the opinion leader characteristic parameters of the opinion leader, wherein the opinion leader characteristic parameters are used to describe the degree to which the opinion leader adheres to his / her own opinion; The generation module 420 can obtain the current opinion leader's proxy opinion at the current moment based on the current information and the opinion leader's information preference for the information source in the following way: Based on the current information, the initial opinion leader proxy opinion, the opinion leader characteristic parameters, and the opinion leader's information preference for the information source, the current opinion leader proxy opinion of the opinion leader at the current moment is obtained.

[0095] In an exemplary embodiment of the present invention, the information sources regarding the preset social event include multiple information sources; the generation module 420 may also be configured to: Determine the information preferences of opinion leaders for each of the aforementioned information sources; For any of the aforementioned information sources, the information preferences of the information sources are normalized based on each information preference to obtain the normalized information preferences; The information preference after normalization of the information source is used as the influence weight of the information source on the opinion leader. The generation module 420 can obtain the current opinion leader's proxy opinion at the current moment based on the current information, the initial opinion leader's proxy opinion, the opinion leader's characteristic parameters, and the opinion leader's information preference for the information source, in the following manner: For the influence weight of any of the aforementioned information sources on the opinion leader, multiply it by the current information to obtain multiple parameters after multiplication. The parameters obtained after multiplication are summed to obtain a weighted average of opinions. Based on the opinion leader characteristic parameters, a first weight of the initial opinion leader proxy opinion and a second weight of the opinion weighted average information are determined respectively. Based on the first weight, the initial opinion leader proxy opinion, the second weight, and the weighted average opinion information, the current opinion leader proxy opinion of the opinion leader at the current moment is obtained.

[0096] In an exemplary embodiment of the present invention, the determining module 430 may further be configured to: The system acquires the initial opinion of the ordinary agent regarding the preset social event, the opinion of the neighboring ordinary agent at the previous time step, the ordinary agent characteristic parameters of the ordinary agent, the influence ratio of the neighboring ordinary agent, and the influence ratio of the opinion leader. The ordinary agent characteristic parameters describe the degree to which the ordinary agent adheres to its own opinion. The neighboring ordinary agent is the ordinary agent that interacts with the ordinary agent. The sum of the ordinary agent characteristic parameters, the influence ratio of the neighboring ordinary agent, and the influence ratio of the opinion leader is 1. The determination module 430 can determine the current ordinary agent's opinion at the current moment based on the current opinion leader's opinion: The initial ordinary agent opinion, the ordinary agent characteristic parameters, the neighbor ordinary agent opinion at the previous time, the influence ratio of the neighbor ordinary agent, the current opinion leader agent opinion, and the influence ratio of the opinion leader are weighted and summed to obtain the current ordinary agent opinion at the current time.

[0097] In an exemplary embodiment of the present invention, the neighbor general agent is a plurality of neighbor general agents, and correspondingly, the previous time-of-flight neighbor general agent opinion is a plurality of previous time-of-flight neighbor general agent opinions; the determining module 430 may determine the previous time-of-flight neighbor general agent opinion in the following manner: For any of the neighboring ordinary agents, determine the influence weight of the neighboring ordinary agent on the ordinary agent, and The previous neighbor ordinary agent opinion is obtained by weighted summing of the opinions of each neighbor ordinary agent at the previous time step and the influence weight of each neighbor ordinary agent on the ordinary agent.

[0098] In an exemplary embodiment of the present invention, the opinion leader is a plurality of opinion leaders, and correspondingly, the current opinion leader proxy opinion is a plurality of current opinion leader proxy opinions; the determining module 430 may determine the current opinion leader proxy opinion in the following manner: For any one of the opinion leaders, determine the influence weight of the opinion leader on the ordinary agent, and The opinions of the current opinion leaders are obtained by weighted summing of the influence weights of each current opinion leader agent and the opinion leaders on the ordinary agents.

[0099] In yet another exemplary embodiment of the present invention, the generation module 420 may also be configured to: Obtain the opinion leader's proxy opinion from the previous moment; Based on the distance between the opinion leader's proxy opinion at the previous moment and the current information, determine the opinion leader's information preference for the information source; Processing module 440 can iteratively execute the following steps: based on the current information and the opinion leader's information preference for the information source, to obtain the current opinion leader's proxy opinion at the current moment; based on the current opinion leader's proxy opinion, to determine the current ordinary agent's opinion at the current moment; until the obtained current opinion leader's proxy opinion reaches a steady state and the current ordinary agent's opinion reaches a stable state. The opinion leader's proxy opinion at the current moment is taken as the opinion leader's proxy opinion at the previous moment in the next moment. The step of obtaining the opinion leader's proxy opinion at the previous moment is repeatedly and iteratively executed until the step of determining the current ordinary agent's opinion at the current moment is determined. This continues until the difference between the current opinion leader's proxy opinion at the current moment and the opinion leader's proxy opinion at the previous moment is less than a threshold, and the difference between the current ordinary agent's opinion at the current moment and the ordinary agent's opinion at the previous moment is less than a threshold.

[0100] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communications bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other through the communications bus 540. The processor 510 can invoke logical instructions in the memory 530 to execute a method for predicting proxy opinions in a social network. This method includes: acquiring current information from at least one information source at the current moment, wherein the information source is an information source about a preset social event; obtaining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, wherein the information preference is used to characterize the opinion leader's confirmation bias; determining the current ordinary proxy opinion of a common agent at the current moment based on the current opinion leader proxy opinion; iteratively executing the steps of obtaining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, up to the step of determining the current ordinary proxy opinion of a common agent at the current moment based on the current opinion leader proxy opinion, until the obtained current opinion leader proxy opinion reaches a steady state and the current ordinary proxy opinion reaches a stable opinion state; predicting the steady-state opinion leader proxy opinion of the opinion leader regarding the preset social event based on the current opinion leader proxy opinion, and predicting the steady-state ordinary proxy opinion of the common agent regarding the preset social event based on the current ordinary proxy opinion.

[0101] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the proxy opinion prediction method in a social network provided by the above methods. The method includes: acquiring current information of at least one information source at the current moment, wherein the information source is an information source about a preset social event; obtaining the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the information preference of the opinion leader for the information source, wherein the information preference is used to characterize the confirmation bias of the opinion leader; and determining the general opinion based on the current opinion leader proxy opinion. The process involves iteratively executing the following steps: First, determining the current ordinary agent's opinion based on the current information and the opinion leader's information preference for the information source. Second, determining the current ordinary agent's opinion based on the current opinion leader's opinion, continuing until both the current opinion leader's opinion and the current ordinary agent's opinion reach a stable state. Third, predicting the steady-state opinion leader's opinion on the preset social event based on the current opinion leader's opinion, and predicting the steady-state ordinary agent's opinion on the preset social event based on the current ordinary agent's opinion.

[0103] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the agent opinion prediction method in a social network provided by the methods described above. The method includes: acquiring current information from at least one information source at the current moment, wherein the information source is an information source about a preset social event; obtaining the current agent opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, wherein the information preference is used to characterize the opinion leader's confirmation bias; and determining the current ordinary agent's opinion at the current moment based on the current agent opinion. The process involves iteratively executing steps to obtain the current opinion leader's proxy opinion at the current moment based on the current information and the opinion leader's information preference for the information source, and then determining the current ordinary agent's proxy opinion at the current moment based on the current opinion leader's proxy opinion, until the obtained current opinion leader's proxy opinion reaches a steady state and the current ordinary agent's proxy opinion reaches a stable state; based on the current opinion leader's proxy opinion, the steady-state opinion leader's proxy opinion for the preset social event is predicted, and based on the current ordinary agent's proxy opinion, the steady-state ordinary agent's proxy opinion for the preset social event is predicted.

[0104] The device embodiments described above are merely illustrative. 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 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.

[0105] 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., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting agent opinions in social networks, characterized in that, The method includes: Obtain current information at the current moment from at least one information source, wherein the information source is an information source about a preset social event; Based on the current information and the opinion leader's information preference for the information source, the current opinion leader's proxy opinion at the current moment is obtained, wherein the information preference is used to characterize the opinion leader's confirmation bias; Based on the opinions of the current opinion leaders, determine the current opinion of the ordinary agents at the current moment; Iteratively execute the current opinion leader proxy opinion steps based on the current information and the opinion leader's information preference for the information source, until the current opinion leader proxy opinion steps are determined based on the current opinion leader proxy opinion, until the current opinion leader proxy opinion reaches a steady state and the current opinion of the ordinary agent reaches a stable state. Based on the current opinion leader's proxy opinion, the steady-state opinion leader's proxy opinion on the preset social event is predicted, and based on the current ordinary proxy opinion, the steady-state ordinary proxy opinion on the preset social event is predicted.

2. The method for predicting proxy opinions in social networks according to claim 1, characterized in that, Before obtaining the current opinion leader's proxy opinion at the current moment based on the current information and the opinion leader's information preference for the information source, the method further includes: The initial opinion leader proxy opinion of the opinion leader on the preset social event is obtained, as well as the opinion leader characteristic parameters of the opinion leader, wherein the opinion leader characteristic parameters are used to describe the degree to which the opinion leader adheres to his / her own opinion; The process of obtaining the current opinion leader's proxy opinion at the current moment based on the current information and the opinion leader's information preference for the information source includes: Based on the current information, the initial opinion leader proxy opinion, the opinion leader characteristic parameters, and the opinion leader's information preference for the information source, the current opinion leader proxy opinion of the opinion leader at the current moment is obtained.

3. The method for predicting proxy opinions in social networks according to claim 2, characterized in that, The information sources for the preset social event include multiple information sources; before obtaining the current opinion leader's proxy opinion at the current moment based on the current information, the initial opinion leader's proxy opinion, the opinion leader's characteristic parameters, and the opinion leader's information preference for the information sources, the method further includes: Determine the information preferences of opinion leaders for each of the aforementioned information sources; For any of the aforementioned information sources, the information preferences of the information sources are normalized based on each information preference to obtain the normalized information preferences; The information preference after normalization of the information source is used as the influence weight of the information source on the opinion leader. The step of obtaining the current opinion leader's proxy opinion at the current moment based on the current information, the initial opinion leader's proxy opinion, the opinion leader's characteristic parameters, and the opinion leader's information preference for the information source includes: For the influence weight of any of the aforementioned information sources on the opinion leader, multiply it by the current information to obtain multiple parameters after multiplication. The parameters obtained after multiplication are summed to obtain a weighted average of opinions. Based on the opinion leader characteristic parameters, a first weight of the initial opinion leader proxy opinion and a second weight of the opinion weighted average information are determined respectively. Based on the first weight, the initial opinion leader proxy opinion, the second weight, and the weighted average opinion information, the current opinion leader proxy opinion of the opinion leader at the current moment is obtained.

4. The method for predicting proxy opinions in social networks according to claim 1, characterized in that, Before determining the current ordinary agent's opinion at the current moment based on the current opinion leader's opinion, the method further includes: The system acquires the initial opinion of the ordinary agent regarding the preset social event, the opinion of the neighboring ordinary agent at the previous time step, the ordinary agent characteristic parameters of the ordinary agent, the influence ratio of the neighboring ordinary agent, and the influence ratio of the opinion leader. The ordinary agent characteristic parameters describe the degree to which the ordinary agent adheres to its own opinion. The neighboring ordinary agent is the ordinary agent that interacts with the ordinary agent. The sum of the ordinary agent characteristic parameters, the influence ratio of the neighboring ordinary agent, and the influence ratio of the opinion leader is 1. The process of determining the current ordinary agent's opinion at the current moment based on the current opinion leader's opinion includes: The initial ordinary agent opinion, the ordinary agent characteristic parameters, the neighbor ordinary agent opinion at the previous time, the influence ratio of the neighbor ordinary agent, the current opinion leader agent opinion, and the influence ratio of the opinion leader are weighted and summed to obtain the current ordinary agent opinion at the current time.

5. The method for predicting proxy opinions in social networks according to claim 4, characterized in that, The neighbor general agent refers to multiple neighbor general agents, and correspondingly, the previous time-lapse neighbor general agent opinion refers to multiple previous time-lapse neighbor general agent opinions; the previous time-lapse neighbor general agent opinion is determined in the following way: For any of the neighboring ordinary agents, determine the influence weight of the neighboring ordinary agent on the ordinary agent, and The previous neighbor ordinary agent opinion is obtained by weighted summing of the opinions of each neighbor ordinary agent at the previous time step and the influence weight of each neighbor ordinary agent on the ordinary agent.

6. The method for predicting proxy opinions in social networks according to claim 4, characterized in that, The opinion leaders are multiple opinion leaders, and correspondingly, the proxy opinions of the current opinion leaders are multiple proxy opinions of the current opinion leaders; the proxy opinions of the current opinion leaders are determined in the following way: For any one of the opinion leaders, determine the influence weight of the opinion leader on the ordinary agent, and The opinions of the current opinion leaders are obtained by weighted summing of the influence weights of each current opinion leader agent and the opinion leaders on the ordinary agents.

7. The method for predicting proxy opinions in social networks according to claim 1, characterized in that, Before obtaining the current opinion leader's proxy opinion at the current moment based on the current information and the opinion leader's information preference for the information source, the method further includes: Obtain the opinion leader's proxy opinion from the previous moment; Based on the distance between the opinion leader's proxy opinion at the previous moment and the current information, determine the opinion leader's information preference for the information source; The iterative execution, based on the current information and the opinion leader's information preference for the information source, obtains the current opinion leader's proxy opinion step at the current moment, and determines the current ordinary agent's opinion step at the current moment based on the current opinion leader's proxy opinion, until the obtained current opinion leader's proxy opinion reaches a steady state and the current ordinary agent's opinion reaches a stable state, including: The opinion leader's proxy opinion at the current moment is taken as the opinion leader's proxy opinion at the previous moment in the next moment. The step of obtaining the opinion leader's proxy opinion at the previous moment is repeatedly and iteratively executed until the step of determining the current ordinary agent's opinion at the current moment is determined. This continues until the difference between the current opinion leader's proxy opinion at the current moment and the opinion leader's proxy opinion at the previous moment is less than a threshold, and the difference between the current ordinary agent's opinion at the current moment and the ordinary agent's opinion at the previous moment is less than a threshold.

8. A device for predicting proxy opinions in a social network, characterized in that, The device includes: The acquisition module is used to acquire current information at the current moment from at least one information source, wherein the information source is an information source about a preset social event; The generation module is used to obtain the current opinion leader proxy opinion of the opinion leader at the current moment based on the current information and the opinion leader's information preference for the information source, wherein the information preference is used to characterize the opinion leader's confirmation bias; The determination module is used to determine the current ordinary agent's opinion at the current moment based on the current opinion leader's agent opinion; The processing module is used to iteratively execute the steps of obtaining the current opinion leader's proxy opinion at the current moment based on the current information and the opinion leader's information preference for the information source, and the steps of determining the current ordinary agent's opinion at the current moment based on the current opinion leader's proxy opinion, until the obtained current opinion leader's proxy opinion reaches the opinion state steady state and the current ordinary agent's opinion reaches the opinion state stable state. The prediction module is used to predict, based on the current opinion leader proxy opinion, the steady-state opinion leader proxy opinion of the opinion leader regarding the preset social event, and based on the current ordinary proxy opinion, to predict the steady-state ordinary proxy opinion of the ordinary proxy regarding the preset social event.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the agent opinion prediction method in a social network as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the agent opinion prediction method in a social network as described in any one of claims 1 to 7.

Images