Online viewpoint evolution simulation method and device based on offline event excitation model

Abstract

The invention provides an online viewpoint evolution simulation method and device based on an offline event excitation model, computing equipment and a storage medium, and the method comprises the steps: obtaining social media data related to a public opinion event, and extracting a user interaction network structure and viewpoint values of a user in different time periods in a public opinion development process from the social media data; according to the public opinion event data, modeling a real event occurrence sequence into a Horkes point process to obtain a Horkes function of a real event; constructing a viewpoint dynamic model based on offline event excitation according to the model for expressing the user viewpoint interaction mechanism, the social specification regulation and control factor and the Horkes function, and adjusting parameters by using a Monte Carlo method; and inputting user initial viewpoints obtained in real data into the viewpoint dynamic model, and carrying out iterative calculation until viewpoint values of all users in the network converge to a stable state. According to the method, an influence mechanism of an offline event on online public opinion development can be accurately modeled to a great extent, and analogue simulation of social media platform user viewpoint evolution is realized.

Description

Technical Field

[0001] The present invention relates to the field of social network application technology, and in particular to a method, apparatus, computing device, and storage medium for simulating the evolution of social media opinions based on an offline event incentive model. Background Art

[0002] With the rapid development of the Internet, people are increasingly interacting and communicating on social media platforms. Events that occur in the real world can be easily posted on social media, sparking widespread discussion and dissemination, and influencing the evolution of public opinion.

[0003] Existing research on opinion evolution simulation primarily focuses on opinion dynamics, which aims to understand and explain the mechanisms by which opinions form, evolve, and spread among people. Existing studies of opinion dynamics on social media primarily consider specific nodes within the network and the unique interaction mechanisms of social media, simulating and explaining phenomena such as echo chambers and spirals of silence. However, these approaches overlook the role of real-world events in incentivizing and shaping the evolution of online opinions, as well as the impact of platform regulations on limiting the extremism of opinions, resulting in inaccurate simulation results. Summary of the Invention

[0004] In view of this, the present invention is proposed to provide an online opinion evolution simulation method based on an offline event incentive model to overcome the above-mentioned problems or at least partially solve the above-mentioned problems, which includes:

[0005] Step S101: Acquire social media data about public opinion events and construct a user interaction network structure; quantify the opinion tendencies in users' historical posts and obtain the opinion values ​​of users at different time periods during the development of public opinion;

[0006] Step S102: Based on the public opinion event data, the real event sequence is modeled as a Hawkes point process to obtain the Hawkes function of the real event;

[0007] Step S103: Based on the model representing the user opinion interaction mechanism, the social norm regulation factor that regulates opinions to align with public opinion, and the Hawkes function of the real event, a viewpoint dynamics model based on offline event incentives is constructed; based on the user interaction network structure and user opinion values ​​in the public opinion event, the parameters of the viewpoint dynamics model are determined using the Monte Carlo method;

[0008] Step S104: Input the initial user opinion values ​​obtained from the real data into the opinion dynamics model with determined parameters, and iterate the calculation until the opinion values ​​of all users in the user interaction network structure converge to a stable state, thereby completing the online opinion evolution simulation.

[0009] Preferably, the quantified opinion tendencies in the user's historical posts are used to obtain the user's opinion values ​​at different time periods during the development of public opinion:

[0010] The user's social media speech text is encoded into a vector, and the opinion value is obtained by normalizing it by measuring the distance between it and the pre-selected extreme opinion vector.

[0011] Preferably, the vector distance between two texts is obtained by using the average cosine similarity between each word vector in the text.

[0012] Preferably, the step 102 is: modeling the real time series as a self-excitation point process, and the excitation function adopts an exponential decay function, then the Hawkes function of the real event is expressed as:

[0013]

[0014] Among them, λ base >0 is the basic intensity, t is the current time, t p are different times in the development of the situation before t, and the constants a, b>0.

[0015] Preferably, in step 103, in the process of determining the parameters of the opinion dynamics model using the Monte Carlo method, taking into account that the user opinion interaction obeys the confidence interval mechanism, the confidence interval is further used as a parameter adjustment object to adjust the interaction weights h between the interaction nodes in the user interaction network structure required to be substituted into the opinion dynamics model. ij Perform a secondary zero judgment: If the difference in opinions between two interacting nodes is not within the confidence interval, no interaction will occur. Let the interaction weight h between the two interacting nodes be ij =0, and then substitute it into the opinion dynamics model.

[0016] Preferably, the step 103 is:

[0017] First, based on the FJ model that represents the interaction mechanism of user opinions and the social norm regulation factor that regulates opinions to move closer to public opinion, a model of opinion evolution under social norm constraints is constructed as follows:

[0018]

[0019] Among them, x i (t) represents the opinion value of user i, which changes with time t; K i represents the social norm regulation coefficient of user i; Ω represents the neutral opinion value; α is the regulation parameter, the larger α is, the better the regulation effect is; μ i is the opinion influence coefficient of user i; h ij Represents the weight of the edge between user i and user j in the user interaction network structure;

[0020] The expression of the opinion dynamics model is:

[0021] dx(t)=[(I n -Λ-K-ω)x(0)+Ku+Λh(x(t))]dt+ωdλ * (t)

[0022] Among them, I n Represents the unit matrix, Λ represents the diagonal matrix composed of sensitivity coefficient μ; represents the social media platform's feedback control vector for each user's opinion, and the feedback control factor y for user k k (t) = -tanh(αx k (t)); K=diag(K1, K2,...,K n ) represents the social norm regulation coefficient matrix; ω=diag(ω1,ω2,...,ω n ) represents the influence matrix of events in real space on different users. For any coefficient ω i ∈[0, 1]; λ * (t) represents the intensity function of the conditions for offline events in real space, that is, the Hawkes function expression of real events; dλ * (t) is the conditional intensity variable in the time window [t, t+dt); h(x) represents h ij The edge weight matrix.

[0023] Preferably, the neutral opinion value Ω is the average of the maximum and minimum opinions of each user in the user interaction network structure.

[0024] The present invention provides an online opinion evolution simulation device based on an offline event incentive model, comprising:

[0025] The public opinion event data processing and extraction module is used to obtain social media data about public opinion events and build a user interaction network structure; quantify the opinion tendencies in users' historical posts and obtain the user's opinion values ​​at different time periods during the development of public opinion;

[0026] The offline event incentive function module is used to model the real event sequence as a Hawkes point process based on public opinion event data to obtain the Hawkes function of the real event;

[0027] A dynamic model parameter adjustment module is used to construct an opinion dynamics model based on offline event incentives based on a model that represents the user opinion interaction mechanism, a social norm regulation factor that regulates opinions to align with public opinion, and the Hawkes function of the real event; and to determine the parameters of the opinion dynamics model using the Monte Carlo method based on the user interaction network structure and user opinion values ​​in the public opinion event;

[0028] The opinion evolution simulation module is used to input the user's initial opinion values ​​obtained from real data into the opinion dynamics model with determined parameters, iteratively calculate until the opinion values ​​of all users in the user interaction network structure converge to a stable state, and complete the online opinion evolution simulation.

[0029] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0030] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned online opinion evolution simulation method based on the offline event incentive model.

[0031] According to another aspect of the present invention, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned viewpoint simulation evolution method based on an offline event incentive model.

[0032] Beneficial effects:

[0033] (1) The present invention combines the characteristics of social media platforms being affected by the development of offline events, uses the Hawkes process to describe the sequence of offline events, and establishes a model of opinion evolution that conforms to the characteristics of social platforms. It can predict the trend of users' opinions in social media public opinion events, facilitate the control of social media public opinion dissemination, and help build a healthier and more harmonious online community space. In addition, the present invention also captures the characteristics of extreme content in social media being restricted by platform regulations, and adds social norm regulatory factors to the opinion evolution model, thereby capturing the characteristics of opinion evolution in user interactions on social media platforms from both online and offline aspects. Compared with existing methods, the present invention increases the dimension of influencing factors in the opinion evolution process, retains the core mechanism of opinion interaction, and achieves accurate fitting of the development process of social media public opinion events.

[0034] (2) Online social media public opinion will change with the occurrence and development of offline events. In order to express this influence, the influence of real events on social media platforms will gradually decay. In a preferred embodiment, the excitation function of the Hawkes process adopts an exponential decay function. The exponential decay function can describe the increase in intensity of offline events and the decrease over time, which is consistent with the changing trend of public opinion discussion on social media platforms.

[0035] (3) In a preferred embodiment, a hyperbolic tangent function is used to model the social norm regulation factor. The nonlinear shape of this function ensures that individuals will adjust in the direction of regulation and gradually approach a neutral point of view, which can describe the influence of social norms on extreme views.

[0036] (4) In a preferred embodiment, the user opinion interaction is further considered to obey the confidence threshold mechanism, and the user interaction weight h ij Make a secondary judgment. If the difference between the opinions of the two interacting nodes is not within the confidence interval, no interaction will occur. Then h ij It should be 0, which describes that it is difficult for two users with too large a gap in opinions to communicate and influence each other, reproduces the discussion atmosphere in the social media public opinion environment, and more accurately fits the development process of social media public opinion events. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flowchart of a method for simulating and evolving opinions based on an offline event incentive model provided by an embodiment of the present invention is shown;

[0038] Figure 2 A schematic diagram of a text opinion value extraction process according to an embodiment of the present invention is shown;

[0039] Figure 3 A schematic structural diagram of a viewpoint simulation evolution device based on an offline event incentive model provided by an embodiment of the present invention is shown;

[0040] Figure 4 A schematic structural diagram of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0041] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0042] Example 1

[0043] Figure 1 The flowchart of the present invention shows a method for simulating the evolution of social media opinions based on an offline event incentive model. Figure 1 As shown, the method includes the following steps:

[0044] Step S101: Obtain relevant social media data of the public opinion event and extract the user interaction network structure from it; quantify the opinion tendencies in the user's historical posts and obtain the user's opinion values ​​in different time periods during the development of public opinion.

[0045] This example uses an opinion dynamics model based on offline event incentives to simulate the evolution of user opinion values ​​based on a real-world interactive network structure. Specifically, a real-world interactive network structure is constructed based on relationships such as comments, reposts, and follows. Each node represents a user, and each edge represents an interaction relationship. Edges have a full-featured nature, and their weights are determined by factors such as interaction frequency and interaction type (e.g., reposts, comments).

[0046] The idea behind quantifying user opinion values ​​is to encode a user's social media post into a vector, normalize it, and measure its distance to a pre-selected set of extreme opinion vectors to obtain the opinion value. The vector distance between two texts can be calculated using the average cosine similarity between each word vector in the text.

[0047] like Figure 2 The specific process of quantifying user opinion values ​​is shown. As shown in the figure, Internet information text data is often unstructured data, and the data content contains special characters such as punctuation, symbols, and emoticons, so the text needs to be preprocessed. First, based on the stop word library, noise vocabulary data is eliminated to improve the accuracy of subsequent processing. Secondly, the text is segmented based on the corpus to find the key words that make up the semantics of the sentence. Then the key words are encoded, and the distributed representation of words in a large corpus is learned through word2vec (a mapping model from words to word vectors) to obtain semantic and syntactic information, and then the word vector is obtained. The text of the extreme opinion content that has been manually selected in advance is subjected to the above operation to obtain the embedding vectors of all words after segmentation, and the average embedding vector of the polarized opinion content is obtained by averaging. After the other content in the relevant topic data of the social media platform is processed in the same way to obtain the average vector, the cosine similarity is calculated to quantify the vector distance between it and the average embedding vector of the polarized opinion content. For two vectors u and v, the cosine similarity is calculated as follows:

[0048]

[0049] Where u·v represents the inner product of u and v, and ||u|| and ||v|| represent the moduli of u and v, respectively.

[0050] The vector distance between texts can be obtained by calculating the average cosine similarity between each word vector. Assume that text A and text B are represented by a1, a2, ..., a N and b1, b2, ..., b N′ Where N and N′ are the number of words in the two texts respectively, then the average cosine similarity between text A and text B is:

[0051]

[0052] Finally, these distances are normalized to obtain the sentiment value of each piece of content. The sentiment value range is set to -1 to 1, where more positive sentiment is closer to 1, neutral sentiment is closer to 0, and negative sentiment is closer to -1.

[0053] In step S102, based on the public opinion event data, the real event sequence is modeled as a Hawkes point process, and the values ​​of relevant parameters in the modeling are determined to obtain the Hawkes function of the real event.

[0054] Online social media public opinion can change as offline events occur and progress. To illustrate this influence, this paper describes offline events as an event sequence {t1, t2, ...}, modeled using a Hawkes point process.

[0055] Specifically, the number of events that occurred before time t is expressed as:

[0056]

[0057] Among them, t k represents the time when event p occurs, Is an indicator function, indicating that event i occurred before t. Historical events The counting process N(t) satisfies:

[0058]

[0059] in, is a historical event, m is the number of new events between the previous and next moments, o(h) represents the minimum term of probability; λ * (t) is the conditional intensity function, which is in the form of:

[0060]

[0061] Among them, λ base >0 is the basic strength, φ:(0,∞)→[0,∞) is the excitation function, which represents the impact of the triggering event on the occurrence of subsequent events.

[0062] For an observed sequence of events {t1, t2, ...}, the conditional intensity function of the Hawkes process is:

[0063]

[0064] In order to model the gradual attenuation of the influence of real events on social media platforms, the incentive function φ(·) adopts an exponential decay function, with φ(t)=ae -bt , where constants a, b>0, then:

[0065]

[0066] The above equation shows that the system strength increases by an amount a each time, and then decays at a rate b with each event. Real time series are modeled as a self-exciting point process, where each event increases the likelihood of another similar event.

[0067] From a series of related real events, important time nodes that have a significant role in promoting online public opinion are manually extracted and expressed as a time series point process. Based on likelihood estimation, the Hawkes form expression of the real events is obtained.

[0068] Specifically, the likelihood function of the Hawkes process is:

[0069]

[0070] Since the conditional function of the Hawkes process used in this method is an exponential decay function, for [0, t p ] event, set The likelihood function of the interval Hawkes process is written as:

[0071]

[0072] [0, t p ] discretize, that is, divide it into [0, t1], [t1, t2], ..., [t p-1 , t p ], so we have:

[0073]

[0074] Then the likelihood function of the exponential Hawkes process is:

[0075]

[0076] The parameters of the Hawkes function can be calculated using the above formula based on real data.

[0077] Step S103: construct an opinion dynamics model based on offline event incentives based on a model that represents the user opinion interaction mechanism, the social norm regulation factor that regulates opinions to be closer to public opinion, and the Hawkes function of the real event; and use the Monte Carlo method to determine the parameters of the opinion dynamics model based on the user interaction network structure and user opinion values ​​in the public opinion event.

[0078] This embodiment constructs an opinion dynamics model based on three considerations: user opinion interaction mechanisms, social norm regulation factors, and the occurrence and development of offline events. Furthermore, during the model parameter adjustment process, consideration was further added regarding the confidence threshold mechanism for user opinion interactions, resulting in a comprehensive opinion dynamics model based on offline event incentives.

[0079] (1) User opinion interaction mechanism

[0080] The interaction mechanism of user opinions is based on the Friedkin-Johnsen (FJ) model. After the user interacts, they will adjust their own opinions based on the opinions of others. The differential equation form of the FJ model is:

[0081]

[0082] Among them, x i represents the opinion value of individual i, which changes with time t; μ i is the opinion influence coefficient of individual i; h ij represents the weight of the edge between user i and user j in the user interaction network structure. ∈ represents random perturbation, which generally conforms to the Poisson distribution.

[0083] (2) Social norm regulatory factors

[0084] The evolution of opinions on social media platforms is not only influenced by offline events but also by platform regulations. The existence of phenomena like echo chambers makes it easy for opinion polarization to occur on social media platforms, but the expression of extreme views is restricted by the platforms. Platform policies tend to favor more rational and neutral speech, so the evolutionary mechanism of opinion dynamics includes a mechanism to regulate extreme views, which can be called a social norm-based regulatory factor.

[0085] The presence of a social norms control factor reduces group polarization. The control factor and opinion deviation are in opposite directions, and under the constraints of this factor, the opinion evolution process still satisfies the opinion value range. Let the control intensity of social norms be y(t) and the control amplitude be K∈(0,1). The control mechanism is described using the hyperbolic tangent function tanh(·), where α>0 is the control coefficient of the S-type control function. Therefore, the equation for the control factor is:

[0086]

[0087] This nonlinear shape ensures that individuals will adjust according to the regulated direction of opinion evolution. The present invention sets the opinion evolution direction to gradually approach the neutral opinion Ω. Therefore, based on (1) the user opinion interaction mechanism and (2) the social norm constraint, the opinion evolution model under social norm constraint is:

[0088]

[0089] Among them, K i represents the social norm control coefficient of user i; Ω represents the neutral opinion value, taking the average of the maximum and minimum opinions. i ∈[-1, 1], Ω=0; the larger the control parameter α is, the better the control effect is.

[0090] (3) Occurrence and progress of offline events

[0091] The occurrence and progress of offline events are expressed using Hawkes functions.

[0092] Combining (1)(2)(3), the differential equation form of the opinion dynamics model based on offline event incentives is obtained as follows:

[0093] dx(t)=[(I n -Λ-K-ω)x(0)+Ku+Λh(x(t))]dt+ωdλ * (t);

[0094] Among them, I n Represents the unit matrix, Λ represents the diagonal matrix composed of sensitivity coefficient μ; represents the social media platform's feedback control vector for each user's opinion, and the feedback control factor y for user k k (t) = -tanh(αx k (t)); K=diag(K1, K2,...,K n ) represents the social norm regulation coefficient matrix; ω=diag(ω1,ω2,...,ω n ) represents the influence matrix of events in real space on different users. For any coefficient ω i ∈[0, 1]; λ * (t) represents the intensity function of the offline event occurrence condition in the real space, dλ * (t) is the conditional intensity variable within the time window [t, t+dt); h(x) represents the edge weight matrix.

[0095] In the opinion dynamics model based on offline event incentives, on the one hand, the social norm regulatory factor is used to reflect the influence of the public average opinion on users, so that user opinions are closer to the public opinion. On the other hand, the influence of offline events in the real space on the evolution of opinions is also incorporated.

[0096] After determining the model form, the Monte Carlo method based on the minimum mean error is used to adjust the parameters according to the network structure and user opinion values ​​in the public opinion event. In the above formula, Λ, K, ω, and α are all parameter adjustment objects. When adjusting the parameters, according to the user interaction network structure in the public opinion event (the weight hij ) and user opinion value x i , substituted into the model, and the Monte Carlo method was used to determine the parameters of the opinion dynamics model.

[0097] In the parameter adjustment process, the determination of h(x) value is further considered. Considering that the user opinion interaction obeys the confidence interval mechanism, if the opinion difference between the two interacting nodes is not within the confidence interval, no interaction will occur, then h ij It should be 0. Then we further set the confidence interval as the parameter object and substitute the interaction weight h into the opinion dynamics model. ij Perform a secondary zero judgment: If the difference in opinions between two interacting nodes is not within the confidence interval, no interaction will occur. Let the interaction weight h between the two interacting nodes be ij = 0, and then substitute it into the opinion dynamics model. Through the parameter adjustment process, the upper and lower limits of the confidence interval obeyed by the user opinion interaction are determined at the same time.

[0098] The parameter adjustment process includes two stages: training and testing. The early user opinion values ​​in the real data are used as the training set to adjust the model parameters. The later opinion values ​​are used as the test set to check the accuracy of the model. Specifically, first, according to the parameter value range defined in the model, a set of parameter combinations are randomly generated in the parameter space. For each set of parameters, the data of the training set is used for calculation, and the mean absolute error (MAE) of the calculation model is repeatedly verified. The mean absolute error is a commonly used regression model performance evaluation indicator, which is used to measure the average difference between the model prediction value and the true value. Suppose the true value of n samples in the training set is y i , the model's prediction value for these samples is Then the mean absolute error is defined as:

[0099]

[0100] A smaller MAE indicates a better fit for the model. This method optimizes parameters over multiple iterations until the maximum number of iterations is reached or a parameter combination with a smaller MAE is found on the validation set. The parameter combination with the smallest MAE is selected as the model's calculation parameters.

[0101] In step S104, the user initial state (opinion value at the first moment) obtained from the real data is input into the opinion dynamics model based on offline event incentives, and iterative calculation is performed until the opinion values ​​of all users in the network converge to a stable state.

[0102] The interactive network structure of the public opinion event and the user's opinion values ​​at different times are obtained from step S101. The edge weight matrix in the model is determined based on the interactive network structure, and the user's opinion value at the initial time is used as the initial value of the model. If there is a user whose opinion value data does not exist at the initial time, the opinion value data is supplemented using interpolation to obtain the data of all users at the initial time.

[0103] In step S102, the Hawkes excitation function of the real event is obtained. In step S103, the optimal parameters of the opinion dynamics model based on offline event excitation are obtained. The edge weight matrix and initial opinion values ​​determined in S101 are input into the opinion dynamics model. Iterative calculations are performed until the opinion values ​​of all users in the network converge to a stable state. The stabilized opinion value is used as the final value of the model simulation.

[0104] By adopting the opinion evolution simulation method based on the offline event incentive model provided in this embodiment, combined with the characteristics of social media platforms being affected by the progress of offline events, and using the Hawkes process to describe the sequence of offline events, an opinion evolution model that conforms to the characteristics of social platforms is established. This can predict the trend of users' opinions in social media public opinion events, facilitate the control of social media public opinion dissemination, and contribute to the construction of a healthier and more harmonious online community space. Furthermore, the present invention also captures the characteristics of extreme content in social media being restricted by platform regulations. Based on the hyperbolic tangent function, the social norm regulation factor is modeled, thereby capturing the characteristics of opinion evolution in user interactions on social media platforms from both online and offline aspects. Compared with existing methods, this method increases the dimension of influencing factors in the opinion evolution process, retains the core mechanism of opinion interaction, and achieves an accurate fit for the development process of social media public opinion events.

[0105] Example 2

[0106] Figure 3 FIG. 1 shows a schematic diagram of a social media opinion evolution simulation device embodiment based on an offline event incentive model. Figure 3 As shown, the device includes: a public opinion event data processing and extraction module 301, an offline event incentive function module 302, a dynamic model parameter adjustment module 303, and an opinion evolution simulation module 304.

[0107] The public opinion event data processing and extraction module 301 is used to obtain relevant social media data of public opinion events, extract the user interaction network structure from it, and quantify the opinion tendencies in the user's historical posts to obtain the user's opinion values ​​in different time periods during the development of public opinion.

[0108] In an optional manner, the public opinion event data processing and extraction module 301 is further used to: construct a network structure based on the relationships such as comments, forwarding, and attention based on the acquired social media data related to the public opinion event.

[0109] In an optional manner, the public opinion event data processing and extraction module 301 is further used to: encode the user's social media speech text into a vector, and normalize it to obtain the opinion value by measuring the distance between it and the pre-selected extreme opinion vector.

[0110] The offline event incentive function module 302 is used to model the real event occurrence sequence as a Hawkes point process based on the public opinion event data and determine the relevant parameters of the event incentive function.

[0111] In an optional manner, the offline event excitation function module 302 is further used to model the real time series as a self-excitation point process, where the occurrence of each event increases the possibility of another similar event occurring.

[0112] In an optional manner, the offline event excitation function module 302 is further used to extract important time nodes from a series of related real events, express them as time series point processes, and obtain a Hawkes form expression of the real events based on likelihood estimation.

[0113] The dynamic model parameter adjustment module 303 is used to construct an opinion dynamics model based on offline event incentives based on a model that represents the user opinion interaction mechanism, a social norm regulation factor that regulates opinions to be closer to public opinion, and the Hawkes function of the real event; and use the Monte Carlo method to determine the parameters of the opinion dynamics model based on the user interaction network structure and user opinion value in the public opinion event.

[0114] In an optional manner, the dynamics model parameter adjustment module 303 is further used to: combine the social norm regulation factors and the confidence threshold mechanism existing in the evolution mechanism of opinion dynamics to comprehensively form an opinion dynamics model based on offline event incentives.

[0115] In an optional manner, the dynamic model parameter adjustment module 303 is further used to: adjust parameters based on the network structure and user opinion values ​​in the public opinion event using the Monte Carlo method based on the minimum mean error to determine the optimal parameter combination.

[0116] The opinion evolution simulation module 304 is used to input the opinion dynamics model based on offline event incentives according to the user initial state and dynamic parameters obtained from real data, and iteratively calculate until the opinion values ​​of all users in the network converge to a stable state.

[0117] In an optional embodiment, the opinion evolution simulation module 304 is further configured to input the edge weight matrix and the initial opinion value into the opinion dynamics model and perform iterative calculations until the opinion values ​​of all users in the network converge to a stable state. The stabilized opinion value is then used as the final simulation value of the model.

[0118] Example 3

[0119] An embodiment of the present invention provides a non-volatile computer storage medium storing at least one executable instruction. The computer executable instruction can execute a social media opinion evolution simulation method based on an offline event incentive model in any of the above method embodiments.

[0120] The executable instructions can be used to cause the processor to perform the following operations:

[0121] Obtain relevant social media data on public opinion events, extract the user interaction network structure from it, and quantify the opinion tendencies in users' historical posts to obtain the user's opinion values ​​at different time periods during the development of public opinion.

[0122] Based on the public opinion event data, the real event sequence is modeled as a Hawkes point process, and the relevant parameters of the event incentive function are determined.

[0123] According to the network structure, user opinion value, and Hawkes function of real events extracted from the data, the parameters of the dynamic model based on offline event incentives are determined based on the Monte Carlo method.

[0124] The user initial state and dynamic parameters obtained from real data are input into the opinion dynamics model based on offline event incentives, and the iterative calculation is carried out until the opinion values ​​of all users in the network converge to a stable state.

[0125] Example 4

[0126] Figure 4 The schematic diagram of the structure of the computing device embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0127] like Figure 4 As shown, the computing device may include:

[0128] Processor, Communications Interface, Memory, and Communication Bus.

[0129] The processor, communication interface, and memory communicate with each other via a communication bus. The communication interface is used to communicate with other devices, such as clients or other server network elements. The processor is used to execute programs, specifically, to perform the steps described in the aforementioned embodiment of a method for simulating social media opinion evolution based on an offline event-driven model.

[0130] Specifically, the program may include program codes including computer operation instructions.

[0131] The processor may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the server may be processors of the same type, such as one or more CPUs, or processors of different types, such as one or more CPUs and one or more ASICs.

[0132] Memory is used to store programs. The memory may include high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk storage.

[0133] The program can be specifically used to cause the processor to perform the following operations:

[0134] Obtain relevant social media data on public opinion events, extract the user interaction network structure from it, and quantify the opinion tendencies in users' historical posts to obtain the user's opinion values ​​at different time periods during the development of public opinion;

[0135] Model the public opinion event as a Hawkes point process and obtain the first processing result;

[0136] Processing the network structure and opinion value training set data based on the Monte Carlo method to obtain a second processing result;

[0137] According to the network structure and the initial opinion value, the first processing result and the second processing result, the user opinion value after reaching a stable state is predicted.

[0138] The algorithm or demonstration provided herein are not inherently relevant to any particular computer, virtual system or other equipment. Various general-purpose systems may also be used together with the teachings based on this. According to the above description, it is apparent that the structure required for constructing this type of system. In addition, the embodiment of the present invention is not directed to any specific programming language yet. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of the present invention.

[0139] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0140] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the invention, various features of the embodiments of the invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all the features of the individual embodiments previously disclosed. Accordingly, the claims that follow the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of the invention.

[0141] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition may be divided into multiple submodules or subunits or subcomponents. All features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed herein may be combined in any combination, except that at least some of such features and / or processes or units are mutually exclusive. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0142] Furthermore, those skilled in the art will appreciate that although some embodiments herein include certain features included in other embodiments but not other features, combinations of features from different embodiments are intended to be within the scope of the present invention and to form different embodiments. For example, any of the claimed embodiments may be used in any combination.

[0143] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It will be appreciated by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components according to an embodiment of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for executing a part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0144] It should be noted that the above embodiments illustrate rather than limit the invention, and that alternative embodiments may be devised by a person skilled in the art without departing from the scope of the appended claims. In the claims, any reference signs placed between brackets should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising several different elements and by means of appropriately programmed computers. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments should not be understood as limiting the order of execution unless otherwise specified.

Claims

1. A method for simulating online opinion evolution based on an offline event incentive model, characterized in that: include: Step S101: obtaining social media data about public opinion events, constructing a user interaction network structure, and obtaining user opinion values ​​at different time periods during the development of public opinion; Step S102: Based on the public opinion event data, the real event occurrence sequence is modeled as a Hawkes point process to obtain the Hawkes function of the real event; Step S103: constructing an opinion dynamics model based on offline event incentives according to the model representing the user opinion interaction mechanism, the social norm regulation factor that regulates opinions to be closer to public opinions, and the Hawkes function of the real event; According to the user interaction network structure and user opinion values ​​in the public opinion event, the Monte Carlo method is used to determine the parameters of the opinion dynamics model; Step S104: Input the initial opinion values ​​of users obtained from real data into the opinion dynamics model with determined parameters, iterate and calculate until the opinion values ​​of all users in the user interaction network structure converge to a stable state, and complete the online opinion evolution simulation.

2. The method according to claim 1, characterized in that The quantified opinion tendencies in the user's historical posts are used to obtain the user's opinion values ​​at different time periods during the development of public opinion: The user's social media speech text is encoded into a vector, and the opinion value is obtained by normalizing it by measuring the distance between it and the pre-selected extreme opinion vector.

3. The method according to claim 2, characterized in that The vector distance between two texts is obtained by the average cosine similarity between each word vector in the text.

4. The method according to claim 1, characterized in that The step 102 is: modeling the real time series as a self-excitation point process, and the excitation function adopts an exponential decay function, then the Hawkes function of the real event is expressed as: Among them, λ base >0 is the base intensity, t is the current time, t p are different times in the development of public opinion before t, and the constants a, b>0.

5. The method according to claim 1, characterized in that In step 103, in the process of determining the parameters of the opinion dynamics model using the Monte Carlo method, considering that the user opinion interaction obeys the confidence interval mechanism, the confidence interval is further used as a parameter adjustment object to adjust the interaction weights h between the interaction nodes in the user interaction network structure to be substituted into the opinion dynamics model. ij Perform a secondary zeroing judgment: If the difference in opinions between two interacting nodes is not within the confidence interval, no interaction will occur. Let the interaction weight h between the two interacting nodes be ij =0, and then substitute it into the opinion dynamics model.

6. The method according to any one of claims 1 to 5, characterized in that: The step 103 is: First, based on the FJ model that represents the interaction mechanism of user opinions and the social norm regulatory factors that regulate opinions to move closer to public opinions, the opinion evolution model under social norm constraints is constructed as follows: Among them, x i (t) represents the opinion value of user i, which changes with time t; K i represents the social norm regulation coefficient of user i; Ω represents the neutral opinion value; α is the regulation parameter, the larger α is, the better the regulation effect is; μ i is the opinion influence coefficient of user i; h ij Represents the weight of the edge between user i and user j in the user interaction network structure; The expression of the opinion dynamics model is: dx(t)=[(I n -Λ-K-ω)x(0)+Ku+Λh(x(t))]dt+ωdλ * (t) Among them, I n represents the unit matrix, Λ represents the diagonal matrix composed of the sensitivity coefficient μ; represents the social media platform's feedback control vector for each user's opinion, and the feedback control factor y of user k k (t) = -tanh(αx k (t)); K = diag(K1, K2, …, K n ) represents the social norm regulation coefficient matrix; ω=diag(ω1,ω2,…,ω n ) represents the influence matrix of events in real space on different users. For any coefficient ω i ∈[0,1];λ * (t) represents the intensity function of the conditions for offline events in real space, that is, the Hawkes function expression of real events; dλ * (t) is the conditional intensity variable in the time window [t, t+dt); h(x) represents h ij The edge weight matrix.

7. The method according to any one of claim 6, characterized in that: The neutral opinion value Ω is the average of the maximum and minimum opinions of each user in the user interaction network structure.

8. An online opinion evolution simulation device based on an offline event incentive model, characterized in that: include: The public opinion event data processing and extraction module is used to obtain social media data about public opinion events and build a user interaction network structure; Quantify the opinion tendencies in users' historical posts and obtain the opinion values ​​of users at different time periods during the development of public opinion; The offline event incentive function module is used to model the real event sequence as a Hawkes point process based on the public opinion event data and obtain the Hawkes function of the real event; The dynamic model parameter adjustment module is used to construct an opinion dynamic model based on offline event incentives according to the model that expresses the user opinion interaction mechanism, the social norm regulation factor that regulates opinions to be closer to public opinions, and the Hawkes function of the real event; according to the user interaction network structure and user opinion value in the public opinion event, the Monte Carlo method is used to determine the parameters of the opinion dynamic model; The opinion evolution simulation module is used to input the user's initial opinion values ​​obtained from real data into the opinion dynamics model with determined parameters, iteratively calculate until the opinion values ​​of all users in the user interaction network structure converge to a stable state, and complete the online opinion evolution simulation.

9. A computing device, characterized in that include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the online opinion evolution simulation method based on the offline event incentive model as described in any one of claims 1-7.

10. A computer storage medium, characterized in that: The storage medium stores at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the online opinion evolution simulation method based on the offline event incentive model as described in any one of claims 1-7.

Images