A social topic propagation prediction method based on front and back face guided topics

Abstract

The application belongs to the field of network public opinion analysis, and particularly relates to a social topic propagation prediction method based on positive and negative guidance topics, which comprises the following steps: obtaining topic data by using an API interface of a social network, extracting message internal factor features and user internal factor features according to user topic data; calculating positive guidance message influence and negative guidance message influence; calculating positive guidance message driving force and negative guidance message driving force by using evolutionary game theory; calculating user independent forwarding probability; constructing propagation dynamics equation according to the user independent forwarding probability; solving the propagation dynamics equation to obtain propagation prediction results of users to different guidance type topics; and the application can more accurately predict the guidance topic propagation trend in the social network.

Description

Technical Field

[0001] This invention belongs to the field of online public opinion analysis, specifically involving a method for predicting the spread of social topics based on positive and negative guidance. Background Technology

[0002] With the rise of social media and online information dissemination, the spread of trending topics has become an important research subject in contemporary society. People share and disseminate various topics at an astonishing speed through channels such as social networking platforms, news websites, and forums, covering multiple areas from political opinions to popular trends, from social events to commercial promotions. However, topic dissemination is not always dominated by true and valuable information; its process is often influenced by false, misleading, and intentionally biased information. In recent years, scholars have been working to address a series of challenges in topic dissemination models and have published a number of authoritative articles in related fields. Therefore, guiding public opinion in the right direction is of great significance for maintaining social order.

[0003] Currently, research on topic propagation models on social networks mainly falls into three categories. First, there's the SIR model, based on infectious disease models. This model subdivides user nodes involved in topic propagation to construct state transition equations and explore the patterns of topic propagation. Second, there are evolutionary game theory models, which use evolutionary game theory to study the state transitions of different users during propagation. Third, there are machine learning methods for studying topic propagation models. These methods extract user features and input them into specific models to predict user behavior or propagation trends. Since the process of topic propagation on social networks is very similar to the spread of infectious diseases, the classic SIR propagation model can be used to predict the propagation trend of guiding topics.

[0004] While numerous scholars have conducted extensive research and achieved considerable success in the field of social network topic dissemination, several challenges remain: 1. The interactive influence of multiple topic features. When quantifying the driving force of multiple topic features, complex interrelationships and interactions may exist between features. These relationships may be non-linear, requiring appropriate analytical methods to capture and model them. 2. The competition and antagonism between different types of guiding information. The competition and antagonism between guiding information is a dynamic and diverse process, with different participants employing various strategies and means to promote and reinforce their viewpoints. Establishing a competition model to measure the competitive process between different types of guiding information is a challenge. 3. The dynamic nature of user states. The dynamic changes in user states are the result of multiple factors, including the user's own learning, experience, perception, and the external environment and communication. Users may develop different viewpoints after encountering guiding information due to differences in cognition. Summary of the Invention

[0005] To address the above problems, this invention provides a method for predicting the spread of social topics based on positive and negative guidance, comprising the following steps:

[0006] S1. Obtain topic data using the API interface of the social network, wherein the topic data includes user historical behavior information and topic participation information;

[0007] S2. Extract message internal factor features and user internal factor features based on user topic data; message internal factor features include topic popularity and neighbor influence, while user internal factor features include user dissemination activity, opinion consistency, and user driving force;

[0008] S3. Calculate the influence of positive guidance messages and negative guidance messages based on the characteristics of internal factors of the message and the characteristics of internal factors of the user;

[0009] S4. Based on the influence of positive and negative guiding messages, the driving force of positive and negative guiding messages is calculated using evolutionary game theory.

[0010] S5. Calculate the user's independent forwarding probability based on the driving force of positive and negative guiding messages;

[0011] S6. Construct the propagation dynamics equation based on the user's independent forwarding probability;

[0012] S7. Solve the propagation dynamics equation to obtain the propagation prediction results of users for topics of different guidance types.

[0013] The beneficial effects of this invention are:

[0014] This invention, considering the characteristics of real-world topic dissemination, introduces hesitation, positive propagation, and negative propagation states into the classic SIR model. This not only more accurately predicts the spread trends of guiding topics on social networks but also analyzes the impact of positive and negative guiding information on the entire topic dissemination process. By introducing evolutionary game theory, a game-driven mechanism based on user participation intention is designed to effectively analyze the competition and confrontation between different types of guiding information. A topic dissemination prediction model based on positive and negative guiding information is established by incorporating the classic infectious disease model. This model can accurately describe the dissemination trends of users after encountering different types of guiding information and can be applied to predicting and controlling public opinion. Regulatory authorities can choose when to release positive guiding information based on the prediction results, thereby guiding public opinion in a positive direction, reducing the negative impact of negative emotions and information dissemination, and promoting the healthy development of social networks. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention;

[0016] Figure 2 This is a schematic diagram illustrating the process of quantifying the influence of a message in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention provides a method for predicting the spread of social topics based on positive and negative guidance, such as... Figure 1 As shown, it includes the following steps:

[0019] S1. Use the API interface of the social network to obtain topic data, which includes user historical behavior information and topic participation information.

[0020] Specifically, topic data can be obtained from publicly available data websites or using mature social network public APIs. What needs to be obtained is the user history behavior information and topic participation information of all participants (users) throughout the lifecycle of the topic, positive-guiding topics, and negative-guiding topics. Topic participation information includes the time when the topic was forwarded and commented on, the basic information of participating users, and their friend relationship information (including following and being followed); user history behavior information includes the information on forwarded and commented content by the user in the past.

[0021] S2. Extract message internal factor features and user internal factor features based on user topic data; message internal factor features include topic popularity and neighbor influence, while user internal factor features include user dissemination activity, opinion consistency, and user driving force.

[0022] Specifically, the topic popularity includes both positively and negatively influenced topic popularity. The degree of discussion among the population during the spread of a topic determines its breadth and reach. Each relatively independent topic typically reaches its peak in the early stages of its spread, with its popularity decreasing over time; this decay process is similar to the half-life of an element. Therefore, this invention uses topic popularity to represent the degree of dissemination and attention a topic receives online within a certain timeframe. The definition for calculating topic popularity is as follows:

[0023] Positive guidance to boost topic popularity (Pop) p (t):

[0024]

[0025] Among them, t p and t p0 These represent the current moment and the beginning moment of the decline in the popularity of positively guided topics, respectively. p This indicates the average propagation cycle of a positively guided topic; Num p (topic) indicates the number of positive trending topics shared at the current moment; Num p (user) represents the number of users participating in the positive guidance topic discussion at the current moment; k1 and k2 represent the weights of the number of positive guidance topic forwards and the number of users participating in the discussion, respectively.

[0026] Negative influence on topic popularity (Pop) n (t):

[0027]

[0028] Among them, t n and t n0 These represent the current moment and the beginning moment of the decline in the popularity of negatively steered topics, respectively. n This indicates the average propagation cycle of a negatively oriented topic; Num n (topic) indicates the number of reposts of the negative trending topic at the current moment; Num n (user) represents the number of users participating in the discussion of the negative guidance topic at the current moment; k1 and k2 represent the weights of the number of forwards of the negative guidance topic and the number of users participating in the discussion, respectively.

[0029] The topic controversy level includes both positive and negative topic controversy level. Controversy level distinguishes between guiding topics and regular online hot topics. This invention uses topic controversy level to measure the impact of guiding topics on topic dissemination, wherein:

[0030] Positive guidance to increase the level of controversy Cont p :

[0031]

[0032] Among them, P i p This represents the sentiment polarity score of the i-th positively guiding topic; This represents the average content sentiment polarity score that positively guides the topic. This represents the number of discussions for the i-th positively guiding topic; N represents the total number of discussions for the topic, n p This indicates the total number of discussions that positively guide the discussion under this topic;

[0033] Cont's negative guidance on topic controversy n :

[0034]

[0035] Among them, P i n This represents the sentiment polarity score of the i-th negative topic-leading content; This represents the average content sentiment polarity score for topics that are presented in a negative light. This represents the number of discussions on the i-th opposing viewpoint; N represents the total number of discussions on the topic, n n This indicates the total number of discussions that lead to further discussion under this topic;

[0036] The potential for topic dissemination includes both positive and negative influences. A topic's ability to spread among users and their social circles is determined by the user's neighboring nodes. The behavior and attitudes of these neighboring nodes have a significant impact on topic dissemination. When a user shares a topic, their neighboring nodes may be inspired or influenced to further spread the topic. Therefore, a user's neighboring nodes can be considered a driving force, propelling the spread and diffusion of topics on social networks. This invention defines topic dissemination potential as follows:

[0037] Positive guidance on the potential for topic dissemination Inf p (v i ):

[0038] Inf p (v i )=b p ×(Num p [comt(v i )]+Num p [ret(v i )])

[0039] Among them, b p This represents a weakening factor. Since users may have posted similar topics before spreading positive guidance, a weakening factor is introduced to reduce the influence of irrelevant topics. Num p [comt(v i )] and Num p [ret(v i )] respectively represent user v i The number of discussions and shares of neighboring nodes that positively guide the topic;

[0040] Negative guidance on the potential for topic dissemination Inf n (v i ):

[0041] Inf n (v i )=b n ×(Num n[comt(v i )]+Num n [ret(v i )])

[0042] Where bn represents the weakening coefficient, Num n [comt(v i )] and Num n [ret(v i )] represent the number of discussions and shares of the negative guidance topic by user vi's neighboring nodes, respectively;

[0043] User dissemination activity includes both positive and negative topic dissemination activity. In the dissemination of guiding topics, user dissemination activity can be defined as the degree of user participation and activity on specific topic content. This activity can include user interactions, sharing, commenting, forwarding, and other behaviors, reflecting the user's level of attention and participation in the specific topic content. The definition of user dissemination activity in this invention is as follows:

[0044] Positive guidance to promote user engagement and activity in the topic. p (v i ):

[0045]

[0046] Activity p (v i = log(Num_p[orig(v i )])+Num_p[forw(v i )]+Num_p[comw(v i )]

[0047] Among them, Activity p (v i ) represents user v i To promote positive engagement and increase activity in the topic, Activity ave_p (net) represents the average activity level of all users on the network in spreading positive guiding topics; Num_p[orig(v i )]、Num_p[forw(v i )]、Num_p[comw(v i )] respectively represent user v iThe number of reposts, follows, and comments on the topic before the positive guidance topic spreads. Because users may have posted other topics online before the positive guidance topic spreads, the number of reposts will be much larger than the number of follows and comments. To reduce the impact of irrelevant topics on the estimated user activity, the logarithm of the original number of reposts is taken.

[0048] Negative guidance on topic user dissemination activity Act n (v i ):

[0049]

[0050] Activity n (v i = log(Num_n[orig(v i )])+Num_n[forw(v i )]+Num_n[comw(v i )]

[0051] Among them, Activity n (v i ) represents user v i To promote the activity of spreading negative guidance, Activity ave_n (net) represents the average activity level of all users on the network in spreading negative and guiding topics; Num_n[orig(v i )]、Num_n[forw(v i )]、Num_n[comw(v i )] respectively represent user v i The number of reposts, followers, and comments on the topic in the period leading up to the spread of the negative topic;

[0052] Opinion alignment includes both positive and negative topic guidance. User opinion alignment refers to the degree of match between a user's emotional response to content during the dissemination of a specific topic and the emotions the content disseminator or the content itself intends to evoke. In other words, it reflects the consistency between the user's attitude towards the topic and the attitude expected by the disseminator or the content itself. Opinion alignment is high when the user's attitude aligns with the content or its disseminator's, and low when they don't. This alignment encompasses not only agreement with the content but also emotional resonance and connection. This invention defines opinion alignment as follows:

[0053] Positive guidance on topic alignment and match p (v i ):

[0054]

[0055] Among them, idea(v i ) represents user v i High-frequency words in historical reposted topics p (msg p () indicates a positive message guiding the discussion. p Keywords appearing during the dissemination process;

[0056] Negative guidance of topic viewpoints match n (v i ):

[0057]

[0058] Among them, idea(v i ) represents user v i High-frequency words in historical reposted topics n (msg n () indicates a negative way of guiding the topic (msg) n Keywords appearing during the dissemination process;

[0059] In the process of topic dissemination, users with more friends and wider discussion platforms tend to have a greater influence on the spread of the topic. This invention defines user influence as follows:

[0060] User-driven Mot(v) i ):

[0061]

[0062] Among them, fol(v i ) represents user v i The number of followers on social networks, fol avg (net) represents the average number of users followed by all users on the social network.

[0063] S3. Calculate the influence of positive guidance messages and negative guidance messages based on the characteristics of internal factors of the message and the characteristics of internal factors of the user.

[0064] Specifically, step S3 calculates the influence of positive guidance messages and negative guidance messages based on the characteristics of internal factors of the message and the characteristics of internal factors of the user, such as... Figure 2 As shown, it includes:

[0065] S31. Calculate the influence of user factors based on user dissemination activity, opinion consistency, and user driving force; the influence of user factors includes the influence of user factors that positively guide the topic and the influence of user factors that negatively guide the topic.

[0066] Specifically, positively guiding the influence of user factors on topics. Represented as:

[0067]

[0068] Negative guidance of topic user factors influence Represented as:

[0069]

[0070] S32. Calculate the influence of message factors based on topic popularity, neighbor influence, and topic controversy; the influence of message factors includes the influence of message factors that positively guide the topic and the influence of message factors that negatively guide the topic.

[0071] Specifically, the positive guidance of topic news factors influences Represented as:

[0072]

[0073] Negative influence factors in guiding the topic Represented as:

[0074]

[0075] S33. Use a multiple linear regression model to fit the influence of user factors and the influence of message factors to obtain the influence of positive guidance messages and the influence of negative guidance messages.

[0076] Specifically, the positive guidance message influence Eff(pos_topic) is represented as:

[0077]

[0078] The influence of negative guidance messages, Eff(neg_topic), is represented as:

[0079]

[0080] Where α1 represents the first regression coefficient, α2 represents the second regression coefficient, and α3 represents the third regression coefficient. These three are regression coefficients obtained by fitting the multiple linear regression model through multiple training iterations. α2 and α3 represent the influence weights of user attributes and topic attributes in topic dissemination, respectively, describing the weights of user factors and message factors in constituting influence.

[0081] S4. Based on the influence of positive and negative guiding messages, the driving force of positive and negative guiding messages is calculated using evolutionary game theory.

[0082] Specifically, step S4 uses evolutionary game theory to calculate the message driving force based on message influence, including:

[0083] S41. Calculate user participation intention based on topic popularity and user dissemination activity; the user participation intention includes positive guidance topic participation intention and negative guidance topic participation intention.

[0084] Specifically, since users' willingness to spread topics is influenced by their own desire to do so, user participation intention is introduced to represent the influence of their own factors on their exposure to trending topics. User participation intention is influenced by multiple factors, including the user's activity level on the internet and the topic's popularity. Specifically, the more friends a user has, the more willing they are to share the topic; simultaneously, the higher a user's activity level on social networks, the more likely they are to forward the topic; furthermore, the higher the topic's popularity, the more likely users are to pay attention to and forward it. Therefore, this invention comprehensively considers factors of user characteristics and topic attributes to more accurately reflect the degree of user participation in topic dissemination. The definition of user participation intention in this invention is as follows:

[0085] Positive guidance on topic participation willingness Part p :

[0086] Part p (v i ) = Num(negber) × Act p (v i )×Pop p (t)

[0087] Anyway, guiding the topic and the willingness to participate (Part) n :

[0088] Part n (v i ) = Num(negber) × Act n (v i )×Pop n (t)

[0089] Num (negber) represents the number of a user's friends.

[0090] S42. Based on user participation willingness and message influence, define a payoff function using evolutionary game theory; the payoff function includes a payoff function for positively guiding the topic and a payoff function for negatively guiding the topic.

[0091] Specifically, when trending topics emerge, users often face the choice of forwarding information that guides the discussion positively or negatively. Considering the potential antagonism between these two types of information, we assume users will choose between them and will only forward one type of information. This choice process is influenced by user psychology and dissemination preferences. Users generally tend to forward information with greater influence and higher engagement, as this information is likely to be more persuasive and spread more widely. To address this, this invention defines two user forwarding game strategies based on the theoretical framework of evolutionary game theory. These strategies aim to describe the behavioral patterns and decision-making rules of users when faced with topics that guide the discussion positively or negatively. This invention defines two user forwarding game strategies, i.e., two payoff functions, based on evolutionary game theory:

[0092] Positive topic guidance benefit function Pro p (v i ):

[0093] Pro p (v i )=(1+β1)×P1×Eff(pos _ topic) × Part p (v i )

[0094] Negative topic guidance function Pro n (v i ):

[0095] Pro n (v i )=(1+β2)×P2×Eff(neg _ topic) × Part n (v i )

[0096] Where P1 and P2 represent user v respectively i The ratio of forwarding positive and negative guiding topics; to measure the guiding effect of guiding topics on users' forwarding behavior, two gain coefficients β1 and β2 are introduced to measure the guiding effect of positive and negative guiding topics, β1,β2∈[0,1].

[0097] S43. Calculate the driving force of positive and negative guiding messages based on the payoff function.

[0098] Specifically, users exhibit different game-theoretic behaviors during the dissemination process due to their own dissemination preferences. This paper, combining multiple types of topic drivers, introduces game theory to formalize the user's behavior of forwarding positive and negative topics as follows:

[0099] Positive guiding message drive Drf p (v i ):

[0100]

[0101] Among them, |U i | and U i These represent user node v respectively. i The number of upstream nodes and the set of upstream nodes; Represents user node v i Its upstream node v j The interplay between them regarding the gains from forwarding positively guiding topics; |H i |and H i These represent user node v respectively. i The number of recent historical communication topics and the collection of historical topics; Represents user node v i The game payoff between forwarding positive guidance topics and historical messages; W1 and W2 represent the weights of the game payoff of upstream nodes and historical messages on the final decision, respectively.

[0102] Negative guidance message driving force Drf n (v i ):

[0103]

[0104] in, Indicates the current user node v i Its upstream node v j The interplay between individuals regarding the gains from forwarding negatively trending topics; Represents user node v i The game payoff between forwarding negative messages and historical messages regarding the topic of guiding discussion; W1 and W2 represent the weights of the impact of the upstream node's game payoff and the historical message's game payoff on the final decision.

[0105] S5. Calculate the user's independent forwarding probability based on the driving force of positive and negative guiding messages.

[0106] Specifically, in a social network, each user can be viewed as an independent node, and the relationships and interactions between users can be represented by edges. Therefore, the topic propagation network in the entire social network can be defined as G = (V, E), where V represents the set of participating users in the social network, and E represents the set of edges used to show the connections between users. Furthermore, based on the SIR model, users are divided into five states: susceptible, hesitant, positive propagation, negative propagation, and immune.

[0107] Considering the selective nature of topic dissemination, users will only change from one state to another during the dissemination process, not multiple states simultaneously. For example, they might change from a susceptible state to a hesitant state; from a hesitant state to a positive dissemination state; and finally from a positive dissemination state to an immune state. The probability of a user forwarding a topic follows a binomial distribution.

[0108]

[0109] Where m represents user v i The number of neighbors, n represents the number of neighbors who forwarded a certain guiding topic, Drf(v i This indicates the driving force behind a particular topic.

[0110] The probability that a user will forward different types of guiding topics at a given time is as follows:

[0111]

[0112]

[0113] P h (t)=(1-P p (t)-P n (t))

[0114]

[0115]

[0116]

[0117] Among them, P p (t) represents the probability that a user forwards a positive, guiding topic, P. n P(t) represents the probability that a user forwards a topic with negative guidance. h (t) represents the probability that a user will not forward the guiding topic. This represents the independent probability of a user forwarding a positively guided topic at time t. This represents the independent forwarding probability of a user on a topic with negative guidance at time t. Let t represent the probability that a user does not forward the guiding topic at time t, and T represent the propagation time of the guiding topic.

[0118] S6. Construct the propagation dynamics equation based on the user's independent forwarding probability.

[0119] Specifically, based on the actual characteristics of topic dissemination on social networks, this invention defines the following topic dissemination rules:

[0120] Since trending topics on the entire network are characterized by rapid outbreaks and short cycles, this invention assumes that the number of user nodes remains constant throughout the entire topic propagation process, and the proportion of users in each state remains constant, i.e., S+H+Ip+In+R=1.

[0121] The entire process of spreading online topics is similar to an infectious disease infection model, where users in different states have a certain probability of infection after contact.

[0122] Topics have a certain lifespan. After a period of time, they will be forgotten and stop spreading on social networks. As the popularity of a topic gradually weakens, all people in the spreading state will eventually become immune, and all immune states will not continue to transform into other states.

[0123] Specifically, the propagation dynamics equations of this invention are constructed as follows:

[0124]

[0125] Where S(t) represents the percentage of users in a susceptible state at time t, H(t) represents the percentage of users in a hesitant state at time t, and I p (t) represents the percentage of users in a positive propagation state at time t, I n R(t) represents the percentage of users in a positive propagation state at time t, and R(t) represents the percentage of users in an immune state at time t.

[0126] S7. Solve the propagation dynamics equation to obtain the propagation prediction results of users for topics of different guidance types.

[0127] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "setting," "connection," "fixing," "rotation," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal connection of two components or the interaction between two components. Unless otherwise explicitly limited, those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0128] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for predicting social topic propagation based on leading topics on the front and back sides, characterized by, The method comprises the following steps: S1. obtaining topic data using an API interface of a social network, the topic data comprising user historical behavior information and topic participation information; S2. extracting message internal factor features and user internal factor features according to the user topic data; The message internal factor features comprise topic popularity, topic controversy, and topic propagation potential, and the user internal factor features comprise user propagation activity, viewpoint coincidence, and user driving force; S3. calculating positive guiding message influence and negative guiding message influence according to the message internal factor features and the user internal factor features, comprising: S31. calculating user factor influence according to the user propagation activity, the viewpoint coincidence, and the user driving force; the user factor influence comprising positive guiding topic user factor influence and negative guiding topic user factor influence; S32. calculating message factor influence according to the topic popularity, the topic propagation potential, and the topic controversy; the message factor influence comprising positive guiding topic message factor influence and negative guiding topic message factor influence; S33. fitting the user factor influence and the message factor influence by using a multiple linear regression model to obtain the positive guiding message influence and the negative guiding message influence; S4. calculating positive guiding message driving force and negative guiding message driving force based on the positive guiding message influence and the negative guiding message influence by using evolutionary game theory; S5. calculating the independent forwarding probability of a user based on the positive guiding message driving force and the negative guiding message driving force; S6. constructing a propagation dynamics equation according to the independent forwarding probability of the user; The propagation dynamics equation is expressed as: Wherein, S(t) represents the proportion of users in the susceptible state at time t, H(t) represents the proportion of users in the hesitant state at time t, I p (t) represents the proportion of users in the positive propagation state at time t, I n (t) represents the proportion of users in the positive propagation state at time t, R(t) represents the proportion of users in the immune state at time t. represents the independent forwarding probability of users to the positive guide topic at time t, represents the independent forwarding probability of users to the negative guide topic at time t, represents the probability that users do not forward the guide topic at time t. S7. solving the propagation dynamics equation to obtain a propagation prediction result of the user for different guiding types of topics. 2.The social topic propagation prediction method based on leading topics according to the front and back sides of a sheet according to claim 1, characterized in that, The topic popularity comprises positive guiding topic popularity and negative guiding topic popularity, wherein: Positive guidance topic popularity Pop p (t): wherein, t p and t p0 respectively represent the current time and the starting time of the positive guidance topic propagation heat attenuation process, w p represents the average propagation period of the positive guidance topic; represents the number of retweets of the positive guidance topic at the current time; represents the number of users participating in the discussion of the positive guidance topic at the current time; and respectively represent the weights of the number of retweets and the number of users participating in the discussion of the positive guidance topic. Counter-guide topic popularity Pop n (t): where t n and t n0 respectively represent the current time and the starting time of the negative guidance topic propagation heat decay process, w n represents the average propagation period of the negative guidance topic; represents the number of retweets of the negative guidance topic at the current time; represents the number of users participating in the discussion of the negative guidance topic at the current time; and respectively represent the weights of the number of retweets and the number of users participating in the discussion of the negative guidance topic. The topic controversy comprises positive guiding topic controversy and negative guiding topic controversy, wherein: Controversy degree of positive leading topic p : wherein, represents the content sentiment polarity score of the ith positive steering topic; represents the average content sentiment polarity score of the positive steering topics; represents the discussion quantity of the ith positive steering topic; N represents the total discussion quantity of the topics, n p represents the total discussion quantity of the positive steering topics under the topic; Cont's negative guidance on topic controversy n : wherein, represents the content sentiment polarity score of the ith negative steering topic; represents the average content sentiment polarity score of the negative steering topics; represents the discussion quantity of the ith negative steering topic; N represents the total discussion quantity of the topics, n n represents the total discussion quantity of the negative steering topics under the topic; The topic propagation potential comprises positive guiding topic propagation potential and negative guiding topic propagation potential, wherein: Positive guidance of topic propagation potential Inf p (v i ): where b p represents a weakening coefficient, and respectively represent the number of discussions and the number of shares of the user v i 's neighbor nodes for the positive guidance topic. Counter-Steering Topic Spread Potential Inf n (v i ): where b n represents a weakening coefficient, and respectively represent the number of discussions and the number of shares of the user v i 's neighbor nodes on the counter-guide topic. The user propagation activity comprises positive guiding topic user propagation activity and negative guiding topic user propagation activity, wherein: Positive guidance topic user propagation activity Act p (v i ) Among them, Activity p (v i ) represents user v i To promote positive engagement and increase activity in the topic, Activity ave_p (net) represents the average activity level of all users on the network in spreading positive guiding topics; Num_p[orig(v i )]、Num_p[forw(v i )]、Num_p[comw(v i )] respectively represent user v i The number of reposts, followers, and comments on the topic in the period leading up to the positive dissemination of the topic; Counter-guide topic user propagation activity Act n (v i ): where Activity n (v i ) denotes the activity of user v i , Activity ave_n (net) denotes the average activity of all users in the network to propagate the negative guidance topic; Num_n[orig(v i )], Num_n[forw(v i )], Num_n[comw(v i )] respectively denote the amount of retweet, the amount of follow, and the amount of comment of user v i in a certain period of time before the propagation of the negative guidance topic. The viewpoint coincidence comprises positive guiding topic viewpoint coincidence and negative guiding topic viewpoint coincidence, wherein: Match of positive leading topic view p (v i ): where idea(v i ) represents the idea of user v i The high-frequency words appearing in the history of forwarding topics, Words p (msg p ) represents the positive guiding topic msg p Key words appearing in the propagation process; Counter-leading topic match n (v i ): where idea(v i ) represents the idea of user v i The high-frequency words appearing in the history of forwarding topics, Words n (msg n ) represents the negative guidance topic msg n Key words appearing in the propagation process; User pushing force Mot(v i ): where fol(v i ) denotes the number of followers of user v i The number of followers of a user in a social network, fol avg (net) denotes the average number of followers of all users in a social network. 3.The social topic propagation prediction method based on front and back face guiding topics according to claim 1, characterized in that, The formula for calculating the message influence is: Positive guiding message influence Eff(pos_topic): Negative guiding message influence Eff(neg_topic): wherein, a1 represents a first regression coefficient, a2 represents a second regression coefficient, a3 represents a third regression coefficient, represents a positive guiding topic user factor influence of the user v i , represents a positive guiding topic message factor influence of the user v i , represents a negative guiding topic user factor influence of the user v i , represents a negative guiding topic message factor influence of the user v i . 4.The social topic propagation prediction method based on front and back face guiding topics according to claim 1, characterized in that, Step S4 uses evolutionary game theory to calculate the message driving force according to the message influence, comprising: S41. calculating user participation willingness according to the topic popularity and the user propagation activity; the user participation willingness comprising positive guiding topic participation willingness and negative guiding topic participation willingness; S42. defining a payoff function according to the user participation willingness and the message influence by using evolutionary game theory; the payoff function comprising positive guiding topic payoff function and negative guiding topic payoff function; S43. calculating the positive guiding message driving force and the negative guiding message driving force according to the payoff function.

5. The social topic propagation prediction method based on leading topics on the front and back sides according to claim 4, characterized in that, Participation willingness degree in positive direction of topic p (v i ): Part n (v i ): where Num(negher) represents the number of friends of user v i .

6. The social topic propagation prediction method based on leading topics on the front and back sides according to claim 5, characterized in that, According to the user participation willingness degree and the message influence, a benefit function is defined, wherein: Positive lead topic reward function Pro p (v i ): Conversational topic steering benefit function Pro n (v i ): Wherein, P1, P2 respectively represent the proportion of users forwarding positive guidance topics and negative guidance topics; is a gain coefficient.

7. The social topic propagation prediction method based on leading topics on the front and back sides according to claim 6, characterized in that, According to the benefit function, a positive guiding message driving force and a negative guiding message driving force are calculated, wherein: positive guidance message push force Drf p (v i ): wherein, and U i respectively represent the number of upstream nodes and the set of upstream nodes of user node ; represent the number of upstream nodes and the set of upstream nodes of user node v i ; j the game payoff of user node v j with its upstream nodes v j regarding forwarding a positively guided topic; and H i respectively represent the number of historical propagated topics and the set of historical topics of user node ; represent the game payoff of user node v i and historical messages regarding forwarding a positively guided topic; and represent weights; Counter steering message push force Drf n (v i ): wherein, represents the game payoff of the current user node v i with its upstream node v j regarding forwarding the counter-guide topic; represents the game payoff of the user node v i and the historical messages regarding forwarding the counter-guide topic.

Images