A rumor propagation prediction method based on rumor refutation and rumor promotion information

Abstract

The present invention relates to a rumor propagation prediction method based on rumor-refuting and rumor-promoting information, comprising the following steps: obtaining target topic data and user information data; calculating user activity based on the number of times a user forwards a blog post and the blog posts published by the user; utilizing a TF-IDF algorithm to extract high-frequency words from the blog posts published by the user, keywords from the rumor blog posts, keywords from the rumor-refuting blog posts, and keywords from the rumor-promoting blog posts, and calculating the degree of interest consistency between the user and the rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts; utilizing a multiple linear regression model to calculate the influence of the rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts on the user; utilizing game theory principles and binomial distribution to calculate the probability of a user forwarding a rumor blog post, a rumor-refuting blog post, and a rumor-promoting blog post; and utilizing mean field theory to construct a propagation dynamics model based on the probability of a user forwarding a rumor blog post, a rumor-refuting blog post, and a rumor-promoting blog post to predict the propagation trend of the rumor blog posts, and thus monitor and control online rumors.

Description

Technical Field

[0001] The present invention belongs to the field of network public opinion analysis, and specifically relates to a rumor propagation prediction method based on rumor-refuting and rumor-promoting information. Background Art

[0002] With the rapid development of internet technology, the internet has replaced traditional media as a crucial medium for information dissemination. Due to the complexity and convenience of social networks, rumors can easily spread like a butterfly effect, causing rumors to spread rapidly in a short period of time, exponentially increasing the harm to public society. Traditional rumor propagation research has not considered the impact of rumor-promoting information and the interaction of multiple messages within a rumor topic on the spread of rumors. Therefore, studying the rumor propagation process in social networks, establishing appropriate online rumor propagation models, and better exploring the factors that influence the spread of various types of rumor information are of great significance for effectively controlling online public opinion.

[0003] In recent years, scholars have conducted extensive research on rumor propagation models, primarily based on the SIR epidemic model and machine learning algorithm models. The SIR epidemic prediction model primarily categorizes user status into three categories: susceptible (S), infected (I), and immune (R). The S status represents users who are unaware of the rumor and easily infected by it; the I status represents users who have been exposed to the rumor and actively spread it; and the R status represents users who have been exposed to the rumor but do not spread it. Based on the machine learning algorithm model, starting from user behavior patterns, the problem is transformed into a classification or regression problem by extracting user characteristics in the propagation network and constructing a rumor propagation prediction model.

[0004] Since rumor propagation on social networks can be viewed as a special type of infectious disease, the SIR infectious disease model is highly effective in predicting rumor propagation. Liu et al. (Liu F, Buss M. Optimal control for heterogeneous node-based information epidemics over social networks[J]. IEEE Transactions on Control of Network Systems, 2020, 7(3): 1115-1126.) proposed a SIRS model based on heterogeneous nodes. This model takes into account the heterogeneity of network structure and the heterogeneity of individual characteristics, and proposes an optimal control framework for preventing rumor propagation.

[0005] Scholars have conducted a series of studies on rumor propagation prediction models and have achieved considerable success, but some technical problems still exist:

[0006] 1. The coexistence and conflict of multiple types of rumor information. The coexistence and conflict of rumors, debunking rumors, and rumor-promoting information on the same topic during their dissemination inevitably affect the speed and momentum of rumor spread. Measuring the competitive and cooperative relationships between these messages is a difficult issue.

[0007] 2. Differences in user cognition. On social networks, users will experience different dissemination states when exposed to various types of rumor information due to differences in cognition. Therefore, it is necessary to consider the impact of rumor-debunking and rumor-promoting information on user status characteristics. Summary of the Invention

[0008] In order to solve the problems existing in the background technology, the present invention provides a rumor propagation prediction method based on rumor-refuting and rumor-promoting information, comprising:

[0009] S1: Obtain target topic data and user information data participating in the target topic through the API interface provided by the social network; the target topic data includes: rumor blog post data, rumor-refuting blog post data, rumor-promoting blog post data, regular blog post data, the publishing time of rumor blog posts, the publishing time of rumor-refuting blog posts, and the publishing time of rumor-promoting blog posts; the user information data includes: friendship relationships between users, the number of times and the forwarding time of rumor blog posts, rumor-refuting blog posts, rumor-promoting blog posts, and regular blog posts by users, blog post data posted by users, and the time when users posted blog posts;

[0010] S2: Calculate the user's activism based on the number of times the user forwarded a regular blog post and the number of blog posts published by the user before the rumor blog post was published;

[0011] S3: Use the TF-IDF algorithm to extract high-frequency words from user-posted blogs, keywords from rumor blogs, keywords from rumor-refuting blogs, and keywords from rumor-promoting blogs, and calculate the degree of interest alignment between the user and the rumor blogs, rumor-refuting blogs, and rumor-promoting blogs.

[0012] S4: Based on the activity of the user's friends; the degree of interest alignment between the user and rumor blogs, rumor-refuting blogs, and rumor-promoting blogs; and the number and time of forwarding of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs by the user, a multivariate linear regression model is used to calculate the influence of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs on the user;

[0013] S5: Based on the influence of rumor blogs, rumor-debunking blogs, and rumor-promoting blogs on users and the number of users among the user's friends who forward rumor blogs, rumor-debunking blogs, and rumor-promoting blogs, we use game theory principles and binomial distribution to calculate the probability that a user will forward a rumor blog, rumor-debunking blog, or rumor-promoting blog.

[0014] S6: Based on the probability of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts, a propagation dynamics model is constructed using mean field theory to predict the spread trend of rumor blog posts.

[0015] The present invention has at least the following beneficial effects

[0016] The present invention proposes a rumor propagation prediction method based on rumor-debunking and rumor-promoting information. By introducing rumor-debunking and rumor-promoting information into the rumor propagation process, the method can not only more accurately predict the rumor propagation trend, but also analyze the impact of rumor-debunking and rumor-promoting information on the overall rumor topic propagation trend. By constructing a rumor-debunking and rumor-promoting three-party game-driven mechanism, the method characterizes the psychological game process generated by users in the cognitive process and quantifies the driving force of user state transitions. The method comprehensively considers the influence of users' friends and calculates the final rumor, rumor-debunking, and rumor-promoting forwarding probabilities. The method establishes a rumor propagation prediction model based on rumor-debunking and rumor-promoting information through an infectious disease model. The model represents the propagation trend of multiple types of rumor information and whether users participate in the forwarding of rumor topics. The method can be applied to rumor propagation prediction and control in social networks. Public opinion departments can monitor and control online rumors more timely and accurately, and guide and suppress them at a reasonable time. The method can also be used to promote corporate products and services, helping to quickly promote and spread advertisements among target groups, increase advertising exposure and brand awareness, and thus achieve good economic and social benefits.

[0017] Figures in the specification

[0018] Figure 1 is a flow chart of the method of the present invention;

[0019] Figure 2 This is a schematic diagram of the influence of rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts on users in the present invention;

[0020] Figure 3 Schematic diagram of the driving force behind users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts in the present invention. Specific implementation methods

[0021] In order to better illustrate the technical solution of the present invention and make its advantages more concise and clear, the problem to be solved by the present invention will be specifically explained below, and then the specific implementation methods of the present invention will be further described in detail with reference to the accompanying drawings.

[0022] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0023] See also Figure 1 The present invention provides a rumor propagation prediction method based on rumor-refuting and rumor-promoting information, comprising:

[0024] S1: Obtain target topic data and user information data participating in the target topic through the API interface provided by the social network; the target topic data includes: rumor blog post data, rumor-refuting blog post data, rumor-promoting blog post data, regular blog post data, the publishing time of rumor blog posts, the publishing time of rumor-refuting blog posts, and the publishing time of rumor-promoting blog posts; the user information data includes: friendship relationships between users, the number of times and the forwarding time of rumor blog posts, rumor-refuting blog posts, rumor-promoting blog posts, and regular blog posts by users, blog post data posted by users, and the time when users posted blog posts; the target topics include: military, sports, entertainment, etc.

[0025] For example, a rumor blog post refers to a post published by a user under a military topic that is inconsistent with the facts; a rumor-refuting post refers to a post published by a user under a military topic that clarifies a rumor blog post, i.e., confirms that the rumor blog post is inconsistent with the facts; a rumor-inciting post refers to a post published by a user under a military topic that is neither consistent with the rumor blog post nor the rumor-refuting post, but is related;

[0026] Perform simple data cleaning on the acquired target topic data, including deleting null values ​​or outliers in the original rumor data. Usually, the acquired data is unstructured and cannot be used directly for data analysis. Most of the unstructured data can be structured through simple data cleaning, such as unifying the time format of the data and deleting some data that are null values ​​or outliers.

[0027] The cleaned original rumor data is stored in the local database, and the data is uniformly and standardizedly stored and named through the table structure to improve the retrieval and reuse efficiency of the original rumor data and the mapping of relationships between tables.

[0028] S2: Calculate the user's activism based on the number of times the user forwarded a regular blog post and the number of blog posts published by the user before the rumor blog post was published;

[0029]

[0030]

[0031] Active(v i )=α*Num[orig(v i )]+Num[forw(v i )]

[0032] Among them, Act(v i ) indicates user v i Active(vi ) indicates user v i Active index, ave (net) represents the average positive index of all users, N is the number of users, Num[orig(v i )] indicates user v i Before the rumor blog post was published, user v i The number of blog posts, Num[forw(v i )] indicates user v i Before the rumor blog post was published, user v i The number of forwarded blog posts, α represents the weakening factor, α∈[0,1].

[0033] S3: Use the TF-IDF algorithm to extract high-frequency words from user-posted blogs, keywords from rumor blogs, keywords from rumor-refuting blogs, and keywords from rumor-promoting blogs, and calculate the degree of interest alignment between the user and the rumor blogs, rumor-refuting blogs, and rumor-promoting blogs.

[0034]

[0035]

[0036]

[0037] Among them, A represents the high-frequency words in all blog posts published by users, B1 represents the keywords in rumor blog posts, B2 represents the keywords in rumor-refuting blog posts, B3 represents the keywords in rumor-promoting blog posts, Mat1(v i ) indicates user v i The degree of interest matching with the rumor blog post, Mat2(v i ) indicates user v i The degree of interest matching with the rumor-refuting blog post, Mat3(v i ) indicates user v i The degree of consistency with the interests of the rumor-promoting blog post.

[0038] S4: Based on the activity of the user's friends; the degree of interest alignment between the user and rumor blogs, rumor-refuting blogs, and rumor-promoting blogs; and the number and time of forwarding of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs by the user, a multivariate linear regression model is used to calculate the influence of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs on the user;

[0039] S41: Calculate the user's motivation for rumor blogs, rumor-refuting blogs, and rumor-promoting blogs based on the user's friends' enthusiasm and the degree of interest compatibility between the user and the rumor blogs, rumor-refuting blogs, and rumor-promoting blogs;

[0040] Push1(v i )=Mat1(v i)*ΣAct(v r )

[0041] Push2(v i )=Mat2(v i )*ΣAct(v a )

[0042] Push3(v i )=Mat3(v i )*ΣAct(v p )

[0043] Among them, Mat1(v i ) indicates user v i The degree of interest matching with the rumor blog post, Mat2(v i ) indicates user v i The degree of interest matching with the rumor-refuting blog post, Mat3(v i ) indicates user v i The degree of interest matching with the rumor-promoting blog post, v r Indicates user v i Friends who forwarded the rumor blog post, v a Indicates user v i Friends who forwarded the rumor-refuting blog post, v p Indicates user v i Friends who forward rumor-mongering blog posts, Act(v r ) represents the friend node v r Activeness, Act(v a ) represents the friend node v a Activeness, Act(v p ) represents the friend node v p The degree of positivity, Push1(v i ) indicates user v i Driven by the rumor blog post, Push2 said user v i Driven by the rumor-busting blog post, Push3 (v i ) indicates user v i Driven by rumor-mongering blog posts.

[0044] S42: Calculate the popularity of the rumor blog post, rumor-refuting blog post, and rumor-promoting blog post based on the number of times and forwarding time of the rumor blog post, rumor-refuting blog post, and rumor-promoting blog post by users;

[0045]

[0046]

[0047]

[0048] Among them, RumNum(t) represents the number of forwardings of the rumor blog post at time t, RumNum(t-1) represents the number of forwardings of the rumor blog post at time t-1, AntNum(t) represents the number of forwardings of the rumor-refuting blog post at time t, AntNum(t-1) represents the number of forwardings of the rumor-refuting blog post at time t-1, ProNum(t) represents the number of forwardings of the rumor-promoting blog post at time t, ProNum(t-1) represents the number of forwardings of the rumor-promoting blog post at time t-1, t0 represents the publishing time of the rumor blog post, Pop1(t) is the popularity of the rumor blog post, Pop2(t) is the popularity of the rumor-refuting blog post, Pop3(t) is the popularity of the rumor-promoting blog post, and w is the regularization factor, which is 1000.

[0049] See also Figure 2 S43: Calculate the influence of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs on users using a multiple linear regression model based on user activity, user motivation for rumor blogs, user motivation for rumor-refuting blogs, user motivation for rumor-promoting blogs, popularity of rumor blogs, popularity of rumor-refuting blogs, and popularity of rumor-promoting blogs:

[0050] Eff1(v i )=ρ1+ρ2×Act(v i )×Mat1(v i )+ρ3×Push1(v i )×Pop1(t)

[0051] Eff2(v i )=ρ1+ρ2×Act(v i )×Mat2(v i )+ρ3×Push2(v i )×Pop2(t)

[0052] Eff3(v i )=ρ1+ρ2×Act(v i )×Mat3(v i )+ρ3×Push3(v i )×Pop3(t)

[0053] Among them, ρ1, ρ2, ρ3 are partial regression coefficients obtained from the training of the multivariate linear regression model, Act(v i ) indicates user v i The positivity of Mat1(v i ) indicates user v i The degree of interest matching with the rumor blog post, Mat2(v i ) indicates user v i The degree of interest matching with the rumor-refuting blog post, Mat3(v i ) indicates user vi The degree of interest matching with the rumor-promoting blog post, Push1(v i ) indicates user v i Driven by the rumor blog, Push2 (v i ) indicates user v i Driven by the rumor-busting blog post, Push3 (v i ) indicates user v i Driven by the rumor-promoting blog post, Pop1(t) is the popularity of the rumor blog post, Pop2(t) is the popularity of the rumor-refuting blog post, Pop3(t) is the popularity of the rumor blog post, Eff1(v i ) represents the influence of rumor blog posts on users, Eff2(v i ) indicates the influence of rumor-refuting blog posts on users, Eff3(v i ) represents the influence of rumor-mongering posts on users. ρ1 is 0.3, ρ2 is 0.12, and ρ3 is 0.3. This step can explore the drivers of user behavior from multiple dimensions and effectively measure the influence of rumor, rumor-debunking, and rumor-mongering posts on users.

[0054] S5: Based on the influence of rumor blogs, rumor-debunking blogs, and rumor-promoting blogs on users and the number of users among the user's friends who forward rumor blogs, rumor-debunking blogs, and rumor-promoting blogs, we use game theory principles and binomial distribution to calculate the probability that a user will forward a rumor blog, rumor-debunking blog, or rumor-promoting blog.

[0055] S51: Based on the influence of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs on users and the number of users among the user's friends who forward rumor blogs, rumor-refuting blogs, and rumor-promoting blogs, the benefits of forwarding rumor blogs, rumor-refuting blogs, and rumor-promoting blogs are calculated using game theory principles;

[0056]

[0057]

[0058]

[0059] Where N is the user v i Num1 is the number of users among the user’s friends who forwarded rumor posts, Num2 is the number of users among the user’s friends who forwarded rumor-refuting posts, and Num3 is the number of users among the user’s friends who forwarded rumor-promoting posts. Eff1(v i ) represents the influence of rumor blog posts on users, Eff2(v i ) indicates the influence of rumor-refuting blog posts on users, Eff3(v i ) indicates the influence of rumor-mongering blog posts on users, Pro1(v i) is the income of users forwarding rumor blog posts, Pro2(v i ) is the income for users to forward rumor-refuting blog posts, Pro3(v i ) is the income from users forwarding rumor-mongering blog posts.

[0060] See also Figure 3 ,S52: Calculate the driving force of users forwarding rumor blogs, rumor-refuting blogs, and rumor-promoting blogs based on the benefits of users forwarding rumor blogs, rumor-refuting blogs, and rumor-promoting blogs using game theory principles;

[0061]

[0062]

[0063]

[0064] Among them, Pro1(v i ) is the income of users forwarding rumor blog posts, Pro2(v i ) is the income for users to forward rumor-refuting blog posts, Pro3(v i ) is the profit of users forwarding rumor-promoting blog posts, w1, w2, w3, w4 are adjustable parameters used to measure the interaction between rumor-promoting blog posts, rumor blog posts and rumor-promoting blog posts, w1, w2, w3, w4∈[0, 1], Drf1(v i ) represents the driving force behind users forwarding rumor blog posts, Drf2(v i ) represents the driving force behind users forwarding rumor-refuting blog posts, Drf3(v i ) represents the driving force behind users forwarding rumor-promoting blog posts. In this invention, w1, w2, w3, and w4 are set to 0.4, 0.3, 0.1, and 0.1, respectively. This step takes into account the interaction between rumor posts, rumor-debunking posts, and rumor-promoting posts, effectively measuring the information game process through the rumor-debunking-rumor-promoting three-way game-driven mechanism.

[0065] S53: Based on the user's driving force for rumor blogs, rumor-refuting blogs, and rumor-promoting blogs, and the number of users among the user's friends who forward rumor blogs, rumor-refuting blogs, and rumor-promoting blogs, the probability of the user forwarding rumor blogs, rumor-refuting blogs, and rumor-promoting blogs is calculated using binomial distribution:

[0066]

[0067]

[0068]

[0069] Among them, m1 is user v i The number of users who forwarded the rumor blog among their friends, m2 is the number of users v iThe number of users who forwarded the rumor-refuting blog post among their friends, m3 is the number of users v i The number of users who forwarded rumor-mongering blog posts among their friends, θ R (t) represents user v i The probability of forwarding a rumor blog post at time t, θ A (t) represents user v i The probability of forwarding a rumor-refuting blog post at time t, θ P (t) represents user v i The probability of forwarding a rumor-mongering blog post at time t.

[0070] S6: Based on the probability of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts, a propagation dynamics model is constructed using mean field theory to predict the spread trend of rumor blog posts.

[0071] S61: Based on the forwarding status of blog posts in social networks, users are divided into susceptible state, rumor-spreading state, rumor-refuting state, rumor-promoting state, and immune state; when susceptible users are exposed to rumor blog posts, there is a probability that they will become rumor-spreading state, when susceptible users are exposed to rumor-refuting blog posts, there is a probability that they will become rumor-refuting state, when susceptible users are exposed to rumor-promoting blog posts, there is a probability that they will become rumor-promoting state, rumor-promoting users have a probability of becoming rumor-spreading state and rumor-refuting state, and rumor-spreading users, rumor-refuting users, and rumor-promoting users have a probability of being immune state. The change probability can be set according to actual conditions and technical personnel in this field, and users can be divided according to the proportion.

[0072] S62: Constructing a propagation dynamics model based on mean field theory;

[0073] S63: Based on the probability of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts, a propagation dynamics model is used to predict the current proportion of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts to obtain the propagation trend of rumor blog posts:

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] Among them, S(t) represents the proportion of users in the immune state at time t, I(t) represents the proportion of users in the rumor-spreading state at time t, A(t) represents the proportion of users in the rumor-refuting state at time t, P(t) represents the proportion of users in the rumor-promoting state at time t, R(t) represents the proportion of users in the immune state at time t, and m1 is the user v i The number of users who forwarded the rumor blog among their friends, m2 is the number of users v i The number of users who forwarded the rumor-refuting blog post among their friends, m3 is the number of users v i The number of users who forwarded rumor-mongering blog posts among their friends, θ R (t) represents user v i The probability of forwarding a rumor blog post at time t, θ A (t) represents user v i The probability of forwarding a rumor-refuting blog post at time t, θ P (t) represents user v i The probability of forwarding a rumor blog post at time t, where T represents the propagation time of the rumor blog post, represents the average probability of a user forwarding a rumor blog post, represents the average probability of a user forwarding a rumor-refuting blog post, Represents the average probability of a user forwarding a rumor-mongering blog post.

[0082] Through the output results of the prediction model of the present invention, the propagation trend of rumors, rumor-refuting and rumor-promoting information under the current rumor topic can be predicted, and the status ratio of users participating in the topic at each moment can be obtained.

[0083] Relevant public opinion departments can understand the spread trend of rumor topics through predicted situation maps, and choose the appropriate time to release rumor-refuting information based on the proportion of user status at different times, or limit the forwarding of information by users with greater influence in releasing rumor-promoting information to suppress the spread of rumors.

[0084] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A rumor propagation prediction method based on rumor-refuting and rumor-promoting information, characterized in that: include: S1: Obtain target topic data and user information data participating in the target topic through the API interface provided by the social network; The target topic data includes: rumor blog post data, rumor-refuting blog post data, rumor-promoting blog post data, regular blog post data, the publishing time of rumor blog posts, the publishing time of rumor-refuting blog posts, and the publishing time of rumor-promoting blog posts; the user information data includes: the friendship between users, the number of times and the forwarding time of rumor blog posts, rumor-refuting blog posts, rumor-promoting blog posts, and regular blog posts by users, the blog post data posted by users, and the time when users posted the blog posts; S2: Calculate the user's activism based on the number of times the user forwarded a regular blog post and the number of blog posts published by the user before the rumor blog post was published; S3: Use the TF-IDF algorithm to extract high-frequency words from user-posted blogs, keywords from rumor blogs, keywords from rumor-refuting blogs, and keywords from rumor-promoting blogs, and calculate the degree of interest alignment between the user and the rumor blogs, rumor-refuting blogs, and rumor-promoting blogs. S4: Based on the activity of the user's friends; the degree of interest alignment between the user and rumor blogs, rumor-refuting blogs, and rumor-promoting blogs; and the number and time of forwarding of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs by the user, a multivariate linear regression model is used to calculate the influence of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs on the user; S5: Based on the influence of rumor blogs, rumor-debunking blogs, and rumor-promoting blogs on users and the number of users among the user's friends who forward rumor blogs, rumor-debunking blogs, and rumor-promoting blogs, we use game theory principles and binomial distribution to calculate the probability that a user will forward a rumor blog, rumor-debunking blog, or rumor-promoting blog. S6: Based on the probability of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts, a propagation dynamics model is constructed using mean field theory to predict the spread trend of rumor blog posts.

2. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 1 is characterized in that: The user's activity level includes: Active(v i )=α*Num[orig(v i )]+Num[forw(v i )] Among them, Act(v i ) indicates user v i Active(v i ) indicates user v i Active index, ave (net) represents the average positive index of all users, N is the number of users, Num[orig(v i )] indicates user v i Before the rumor blog post was published, user v i The number of blog posts, Num[forw(v i )] indicates user v i Before the rumor blog post was published, user v i The number of forwarded blog posts, α represents the weakening factor, α∈[0,1].

3. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 1 is characterized in that: The degree of interest compatibility between the user and rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts includes: Among them, A represents the high-frequency words in all blog posts published by users, B1 represents the keywords in rumor blog posts, B2 represents the keywords in rumor-refuting blog posts, B3 represents the keywords in rumor-promoting blog posts, Mat1(v i ) indicates user v i The degree of interest matching with the rumor blog post, Mat2(v i ) indicates user v i The degree of interest matching with the rumor-refuting blog post, Mat3(v i ) indicates user v i The degree of consistency with the interests of the rumor-promoting blog post.

4. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 1, characterized in that: The calculation process of the influence of rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts includes: S41: Calculate the user's motivation for rumor blogs, rumor-refuting blogs, and rumor-promoting blogs based on the user's friends' enthusiasm and the degree of interest compatibility between the user and the rumor blogs, rumor-refuting blogs, and rumor-promoting blogs; S42: Calculate the popularity of the rumor blog post, rumor-refuting blog post, and rumor-promoting blog post based on the number of times and forwarding time of the rumor blog post, rumor-refuting blog post, and rumor-promoting blog post by users; S43: The influence of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs on users is calculated using a multivariate linear regression model based on user activity, user motivation for rumor blogs, user motivation for rumor-refuting blogs, user motivation for rumor-promoting blogs, popularity of rumor blogs, popularity of rumor-refuting blogs, and popularity of rumor-promoting blogs.

5. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 4 is characterized in that: The popularity of rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts includes: Among them, RumNum(t) represents the number of forwardings of the rumor blog post at time t, RumNum(t-1) represents the number of forwardings of the rumor blog post at time t-1, AntNum(t) represents the number of forwardings of the rumor-refuting blog post at time t, AntNum(t-1) represents the number of forwardings of the rumor-refuting blog post at time t-1, ProNum(t) represents the number of forwardings of the rumor-promoting blog post at time t, ProNum(t-1) represents the number of forwardings of the rumor-promoting blog post at time t-1, t0 represents the publishing time of the rumor blog post, Pop1(t) is the popularity of the rumor blog post, Pop2(t) is the popularity of the rumor-refuting blog post, Pop3(t) is the popularity of the rumor blog post, and w is the regularization factor.

6. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 4 is characterized in that: The influence of rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts on users includes: Eff1(v i )=ρ1+ρ2×Act(v i )×Mat1(v i )+ρ3×Push1(v i )×Pop1(t) Eff2(in i )=ρ1+ρ2×Act(in i )×Mat2(in i )+ρ3×Push2(v i )×Pop2(t) Eff3(v i )=ρ1+ρ2×Act(v i )×Mat3(v i )+ρ3×Push3(v i )×Pop3(t) Among them, ρ1, ρ2, ρ3 are partial regression coefficients obtained from the training of the multivariate linear regression model, Act(v i ) indicates user v i The positivity of Mat1(v i ) indicates user v i The degree of interest matching with the rumor blog post, Mat2(v i ) indicates user v i The degree of interest matching with the rumor-refuting blog post, Mat3(v i ) indicates user v i The degree of interest consistency with the rumor-promoting blog post, Push1(vi) indicates that user v i Driven by the rumor blog post, Push2(vi) indicates that user v i Driven by the rumor-busting blog post, Push3 (v i ) indicates user v i Driven by the rumor-promoting blog post, Pop1(t) is the popularity of the rumor blog post, Pop2(t) is the popularity of the rumor-refuting blog post, Pop3(t) is the popularity of the rumor blog post, Eff1(v i ) represents the influence of rumor blog posts on users, Eff2(v i ) indicates the influence of rumor-refuting blog posts on users, Eff3(v i ) represents the influence of rumor-mongering blog posts on users.

7. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 1 is characterized in that: The probability of the user forwarding a rumor post, a rumor-refuting post, or a rumor-promoting post includes: S51: Based on the influence of rumor blogs, rumor-refuting blogs, and rumor-promoting blogs on users and the number of users among the user's friends who forward rumor blogs, rumor-refuting blogs, and rumor-promoting blogs, the benefits of forwarding rumor blogs, rumor-refuting blogs, and rumor-promoting blogs are calculated using game theory principles; S52: Based on the benefits of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts, the driving force behind users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts is calculated using game theory principles; S53: Based on the user's driving force for rumor blogs, rumor-refuting blogs, and rumor-promoting blogs, and the number of users among the user's friends who forward rumor blogs, rumor-refuting blogs, and rumor-promoting blogs, the probability of the user forwarding rumor blogs, rumor-refuting blogs, and rumor-promoting blogs is calculated using binomial distribution.

8. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 7 is characterized in that: The step S6 comprises: S61: Based on the forwarding status of the blog post in the social network, the users are divided into susceptible state, rumor spreading state, rumor refuting state, rumor promoting state, and immune state; S62: Constructing a propagation dynamics model based on mean field theory; S63: Based on the probability of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts, a communication dynamics model is used to predict the proportion of users forwarding rumor blog posts, rumor-refuting blog posts, and rumor-promoting blog posts at the current moment to obtain the spread trend of rumor blog posts.

9. The rumor propagation prediction method based on rumor-refuting and rumor-promoting information according to claim 8 is characterized in that: The propagation dynamics model is used to predict the current proportion of users forwarding rumor blog posts, refuting rumor blog posts, and promoting rumor blog posts, including: Among them, S(t) represents the proportion of users in the immune state at time t, I(t) represents the proportion of users in the rumor-spreading state at time t, A(t) represents the proportion of users in the rumor-refuting state at time t, P(t) represents the proportion of users in the rumor-promoting state at time t, R(t) represents the proportion of users in the immune state at time t, Drf1(v i ) represents the driving force behind users forwarding rumor blog posts, Drf2(v i ) represents the driving force behind users forwarding rumor-refuting blog posts, Drf3(v i ) represents the driving force for users to forward rumor-inducing blog posts, and m1 is the user v i The number of users who forwarded the rumor blog among their friends, m2 is the number of users v i The number of users who forwarded the rumor-refuting blog post among their friends, m3 is the number of users v i The number of users who forwarded rumor-mongering blog posts among their friends, θ R (t) represents user v i The probability of forwarding a rumor blog post at time t, θ A (t) represents user v i The probability of forwarding a rumor-refuting blog post at time t, θ P (t) represents user v i The probability of forwarding a rumor blog post at time t, where T represents the propagation time of the rumor blog post, represents the average probability of a user forwarding a rumor blog post, represents the average probability of a user forwarding a rumor-refuting blog post, Represents the average probability of a user forwarding a rumor-mongering blog post.

Images