CN115907165A - Derived topic propagation prediction method based on multi-topic emotion measurement - Google Patents
Derived topic propagation prediction method based on multi-topic emotion measurement Download PDFInfo
- Publication number
- CN115907165A CN115907165A CN202211489736.0A CN202211489736A CN115907165A CN 115907165 A CN115907165 A CN 115907165A CN 202211489736 A CN202211489736 A CN 202211489736A CN 115907165 A CN115907165 A CN 115907165A
- Authority
- CN
- China
- Prior art keywords
- topic
- user
- users
- probability
- calculating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of network topic propagation analysis, and particularly relates to a method for predicting propagation of derived topics based on multi-topic sentiment measurement; the method comprises the following steps: acquiring topic data; extracting message factor features and user factor features of the topic data; calculating message influence according to the topic heat, the user power, the user activity and the topic personal emotion matching; dividing a user network to obtain a classification result of a user; calculating the independent forwarding probability of the user according to the message influence; calculating conversion probability among different types of users according to the topic association degree and the independent forwarding probability of the users; constructing a dynamic equation according to the independent forwarding probability of the users and the conversion probability among different types of users; solving a dynamic equation to obtain a propagation prediction result of the user on the derived topic; the method can effectively predict the propagation situation of the derived topics in the social network, and has high practicability.
Description
Technical Field
The invention belongs to the field of network topic propagation analysis, and particularly relates to a derived topic propagation prediction method based on multi-topic emotion measurement.
Background
In recent years, with the popularization and the development scale of the internet being increasingly enlarged, people can conveniently and quickly acquire information from each social platform of the internet, and netizens tend to express their own ideas on social networks. With the development and application of technologies and software such as smart phones, microblogs, wechat and the like, a large amount of topic information fills the whole network, and the topic information has multiple propagation channels, high speed and wide range. Every time an emergency occurs, netizens always generate strong and various discussions, the discussion and forwarding promote the public opinion fermentation of the emergency in the network, and also generate one or more derived topics corresponding to the current event, thereby expanding the public opinion propagation range and accelerating the public opinion propagation speed.
In real life, the cognitive influence of network public sentiment on the public is increased day by day, and the public sentiment crisis of the social network can suddenly appear at any time, so that the public sentiment topic information is predicted to a certain degree, the public sentiment propagation situation and the derivative trend of the public sentiment topic are favorably integrally monitored and monitored in real time, and the public is guided to make relatively correct and objective judgment on the sudden event. In the process of topic propagation, the forwarding behavior of the user can be greatly influenced by the subjective cognition and the psychology of the user, and as an uncertain factor which is difficult to quantify, how to predict the psychology of the user becomes a great challenge. Therefore, accurate acquisition of network public opinion information and research on topics and derivative trends thereof have important significance for guiding and monitoring network public opinions.
In recent years, researchers have conducted a great deal of research on derived topic propagation models, mainly based on SIR infectious disease models, machine learning algorithm models, and deep learning algorithm models. The topic propagation prediction model based on the SIR infectious disease model mainly divides users into three states: susceptible (S), infected (I), immunized (R). Wherein, the susceptible person refers to a user who has not contacted with the topic, and the topic can be spread; the infected person refers to a user who has already contacted with a topic and can actively participate in the transmission of the topic; an immunized person refers to a user who has been exposed to a topic but does not participate in the dissemination of the topic. The prediction model based on the machine learning algorithm mainly extracts features influencing user propagation, changes prediction problems into classification or regression problems, can process massive data through the algorithm in machine learning, and is suitable for processing complex problems in a social network. Since the process of topic transmission is similar to that of infectious disease, SIR infectious disease models can be used to predict the transmission trends of the derived topics.
Analysis of what current researchers have done in topic propagation prediction can find that research on the propagation of derived topics has been fruitful, but there are still some challenges: 1. complexity of preposition of emotion. The preposition emotion is a non-negligible factor for the propagation of the derived topic, but the personal emotional preference has diversity and uncertainty for the user, and how to measure the personal emotional preference becomes a challenge. 2. Deriving a multilevel nature of the topic. In topics of different hierarchies, the situation and the features of message propagation are different, but correlation and conversion need to be performed between the hierarchies, and how to combine the difference between multiple layers of topics and the state conversion between the hierarchies becomes a challenge. 3. There is a continuing concern about psychological complexity. In the process of spreading multiple topics, in addition to the factors of the topics, the continuous attention psychology can influence the user to continuously pay attention to new topics derived from a certain topic, and how to measure the influence of the psychological factors on topic spreading becomes a challenge.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a method for predicting propagation of derived topics based on multi-topic emotion measurement, which comprises the following steps:
s1: obtaining topic data, wherein the topic data comprises user historical behavior information and topic participation information;
s2: extracting message factor features and user factor features of the topic data; the message factors comprise topic heat and topic relevance, and the user factors comprise user drive, user activity and topic personal emotion matching;
s3: calculating message influence according to the topic heat, the user power, the user activity and the topic personal emotion matching;
s4: dividing a user network to obtain a classification result of a user;
s5: calculating the independent forwarding probability of the user according to the message influence;
s6: calculating conversion probability among different types of users according to the topic association degree and the independent forwarding probability of the users;
s7: constructing a dynamic equation according to the independent forwarding probability of the users and the conversion probability among different types of users;
s8: and solving a dynamic equation to obtain a propagation prediction result of the user on the derived topic.
Preferably, the process of calculating the message influence comprises: calculating the influence of message factors according to the topic heat and the user drive; calculating the influence of the user factors according to the activity of the user and the topic personal emotion matching degree; and calculating the message influence according to the message factor influence and the user factor influence.
Further, the formula for calculating the message influence is as follows:
Eff(u i )＝β 0 +β 1 *fac user (u i )+β 2 *fac topic (u i )
wherein, eff (u) i ) Representing user u i Message influence of beta 0 Denotes the first partial regression coefficient, β 1 Represents the second partial regression coefficient, fac user (u i ) Representing user u i Influence of user factor, beta 2 Represents the third partial regression coefficient, fac topic (u i ) Representing user u i The message factor of (c) affects the power.
Preferably, the process of dividing the user network includes: dividing users into users who have participated in the predecessor topics and users who have not participated in the predecessor topics; and dividing the users who participate in the predecessor topics and the users who do not participate in the predecessor topics into susceptible users, infected users and immune users again to obtain user classification results.
Preferably, the formula for calculating the independent forwarding probability of the user is as follows:
where λ represents the user-independent forwarding probability, ψ (t) represents an intermediate parameter, and m represents the user u i N represents the number of neighbors of the forwarded topic, eff (u) i ) Representing user u i The message impact of (2).
Preferably, the process of calculating the conversion probability between different types of users according to the topic association degree and the independent forwarding probability of the users comprises the following steps:
calculating cross-model state transition probability according to topic association degree and independent forwarding probability of users, and converting immune users into susceptible users participating in derived topics according to the cross-model state transition probability;
calculating the transition probability of the susceptible users participating in the precursor topic according to the topic association degree and the independent forwarding probability of the users, and converting the susceptible users into infected users participating in the derived topic according to the transition probability;
calculating the transition probability of the susceptible users who do not participate in the precursor topic according to the topic association degree and the independent forwarding probability of the users, and converting the susceptible users into infected users who participate in the derived topic according to the transition probability;
infected users turn into immunized users with a fixed probability.
Further, the formula for calculating the cross-model state transition probability is:
wherein p is ab Represents the transition probability from the immune user who participates in the topic a to the susceptible user who participates in the topic b, and lambda represents the user independent forwarding probability Rlv (r) a ,r b ) The topic relevance degree of the topic a and the topic b is shown, and the beta is a threshold value of the topic relevance degree.
Further, the formula for calculating the transition probability of the susceptible users who participated in the predecessor topics is as follows:
λ p ＝λ+(1-λ)×Rlv(r a ,r b )
wherein λ is p Represents the transition probability of the susceptible user participating in the predecessor topic, λ represents the independent forwarding probability of the user, rlv (r) a ,r b ) The topic relevance between the topic a and the topic b is shown.
Further, the formula for calculating the transition probability of the susceptible users who have not participated in the predecessor topics is as follows:
λ q ＝(1-Rlv(r a ,r b ))×λ
wherein λ is q Denotes the transition probability of a susceptible user who has not participated in the predecessor topic, λ denotes the user's independent forwarding probability, rlv (r) a ,r b ) And the topic relevance degree of the topic a and the topic b is shown.
Preferably, the kinetic equation is expressed as:
wherein the content of the first and second substances,
representing the individual forwarding probabilities of the users; alpha represents the degree of association Rlv (r) of topic i with its predecessor topic i-1 ,r i )，G 0 User network, S, representing non-participation in propagation of predecessor topics i (t) and I i (t) shows participation in topic i at time tRatio of susceptible users to infected users, R i-1 (t) immune user ratio of predecessor topics participating in topic i at time t, G 1 Representing a network of users participating in the propagation of predecessor topics, p (i-1)i Representing the cross-model state conversion probability of the topic i and the predecessor topic thereof, beta representing the topic relevance threshold, S i 、I i 、R i Respectively, the susceptibility, infection and immune user ratios at any one time.
The invention has the beneficial effects that: the method and the device construct a multilayer iterative SIR model to simulate the mutual influence process of topics and derived topics in the transmission process, introduce psychological factors of continuous topic attention of users, quantify the influence of the users on topic transmission in the psychological states of different derived topics, construct a derived topic transmission dynamic equation by combining topics and user characteristics, and obtain the prediction result of the user on the derived topic transmission according to the equation result.
Drawings
FIG. 1 is a schematic structural diagram of a method for predicting propagation of a derived topic based on multi-topic sentiment measurement in the present invention;
FIG. 2 is a schematic diagram illustrating a process for quantifying message impact according to the present invention;
FIG. 3 is a schematic diagram of inter-user conversion in the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a method for predicting propagation of derived topics based on multi-topic emotion measurement, which comprises the following steps of:
s1: topic data is obtained, and the topic data comprises user historical behavior information and topic participation information.
Topic data can be obtained from public data websites or by utilizing mature social network public APIs. What needs to be obtained here is the user historical behavior information and topic participation information of all participants, namely users, of topics and derived topics in the life cycle of the topics. What the topic participation information needs to obtain is the time when the topic is forwarded and commented, the basic information of the participating users and the friend relationship information (including the concerned and concerned information) of the participating users; the user historical behavior information includes information that the user has historically forwarded and commented on.
S2: extracting message factor features and user factor features of the topic data; the message factors comprise topic heat and topic relevance, and the user factors comprise user drive, user activity and topic personal emotion matching.
Topic heat P (t):
the topic popularity P (t) is the traffic of the topic, i.e. the popularity of the topic. Each relatively independent topic undergoes a process of gradually increasing heat and then gradually fading, and the flow of the topic slowly fades with time and slowly reaches a low-level final value, which represents that the topic leaves the public visual field. The decay process is similar to the half-life property of physical elements, and the invention introduces a half-life function
To describe the decay process of topic traffic, the topic popularity P (t) is defined as follows:
wherein t and t 0 Respectively representing the current time and the starting time of the topic propagation heat attenuation process, wherein T is the average propagation period of the topic, and preferably, T can be set to 1000.
Topic relevance Rlv (r) 1 ,r 2 )：
Topic relevance (topic reservance) is defined as the degree of relevance or similarity between the contents of two topics. Extracting keywords or labels from the topic content, and performing normalization calculation by using the Jaccard similarity coefficient. The larger the value of the Jaccard coefficient is, the larger the degree of association between two topics is represented, whereas the smaller the value of the Jaccard coefficient is, the smaller the degree of association between two topics is represented. The invention defines the relevance of the questions as follows:
wherein, topic (r) 1 ) And Topic (r) 2 ) Respectively represent topic r 1 And topic r 2 The keyword or tag of (a).
User with power I (u) i )：
After the user publishes the topic information, the fan of the user browses, comments and forwards the topic information, so that the topic is gradually diffused in the social network; by using
Respectively represent users u i And average browsing amount, average comment number and average forwarding amount of all microblogs published a period of time before the topic is launched. Since the browsing volume is usually much larger than the other two items and the browsing volume does not have much influence on the propagation of the topic, a weakening coefficient mu is added before the average browsing volume. The driving force of the user in the information dissemination process can be defined as:
user activity Act (u) i )：
User activity represents user u i Spreading recent activity on the topic. It is generally considered that users with higher liveness are more likely to participate in topic dissemination, and the impetus for topic dissemination is greater. The definition of the user activity of the invention is as follows:
Act(u i )＝η×Num[orig(u i )]+Num[retw(u i )]
wherein, num [ orig (u) ] i )]And Num [ retw (u) i )]Respectively represent users u i The number of original microblogs released in a period of time before the topic is spread and the number of forwarded microblogs. The original microblog of a general user is far less than the forwarding microblog, so that eta epsilon [0,1 is defined]And as a weakening coefficient, the influence of the number of the original microblogs on the calculation of the activity of the user is reduced.
Topic personal emotion matching Match (r) i ,u i )：
Similar to the definition of the association degree between the topics, the emotional matching degree of the individual to the topics is defined to describe the interest degree of the user to the topics. The higher the matching degree is, the more likely the user is to participate in the forwarding of the topic, and the definition of the personal emotion matching degree of the topic is as follows:
wherein, topic (r) i ) Representing a topic r i Keyword or tag of (c), inter (u) i ) Representing user u i Interest preference tags of (1).
S3: and calculating the message influence according to the topic heat, the user driving force, the user activity and the topic personal emotion matching degree.
As shown in fig. 2, in the social network, the factors that determine the influence of the message are both the message itself and the user factor. The message factors such as topic popularity and relevance, and the user factors such as liveness, emotion matching, attention psychology and the like. The influence of the message can be measured by combining two factors; specifically, the method comprises the following steps:
calculating the influence of message factors according to the topic heat and the user driving force:
fac topic (u i )＝I(u i )*P(t)
calculating the influence of the user factors according to the activity of the user and the personal emotion matching degree of the topic:
fac user (u i )＝Act(u i )*Match(r i ,u i )
calculating the message influence according to the message factor influence and the user factor influence:
Eff(u i )＝β 0 +β 1 *fac user (u i )+β 2 *fac topic (u i )
wherein, eff (u) i ) Representing user u i The message impact of (1); beta is a beta 0 、β 1 And beta 2 Respectively representing a first partial regression coefficient, a second partial regression coefficient and a third partial regression coefficient obtained by training and fitting a multiple linear regression model, wherein beta 1 And beta 2 Respectively representing the weight of the user factor and the weight of the message factor, and describing the proportion of the two factors when the influence is formed; fac user (u i ) Representing user u i User factor influence of, fac topic (u i ) Representing user u i The message factor influences the power.
S4: and dividing the user network to obtain the classification result of the user.
The invention takes the precursor topic as the research background, and divides the whole topic network into two propagation networks which participate in the precursor topic propagation and do not participate in the precursor topic propagation. By G 0 ＝{V 0 ,E 0 And G 1 ＝{V 1 ,E 1 Denotes a user network which does not participate in the propagation of the predecessor topic and participates in the propagation of the predecessor topic, respectively (where V 0 Representing a set of users not participating in the propagation of predecessor topics, E 0 A set of user relationships that represent not participating in a predecessor topic propagation; v 1 And E 1 The same way), then G = (G) 0 ∪G 1 ) Representing the social network in the context of the entire predecessor topic.
And dividing the users who participate in the precursor topic and the users who do not participate in the precursor topic into a susceptible user S, an infected user I and an immune user R again to obtain user classification results.
S5: and calculating the independent forwarding probability of the user according to the message influence.
Considering the repeatability of the life cycle in the topic derivation process, the invention makes the SIR model according to different conversationsMultilayer coupling is carried out on the topic layer, and an Iterative-SIR (multiple topic iteration SIR) derived topic propagation model is constructed; due to the unidirectionality of information propagation of each derived topic layer, the state transitions of users also have unidirectionality in the same layer. That is, the user status can only go from susceptible status to infected status to immune status. Suppose a user node u i There are m neighbor nodes, where the probability of n neighbors forwarding the layer topic obeys a binomial distribution:
can obtain any user u i The probability of forwarding the layer topic at the time t is as follows:
and (3) combining an average field theory to obtain the infection rate at the time t, namely the independent forwarding probability of the user:
s6: and calculating the conversion probability among different types of users according to the topic association degree and the independent forwarding probability of the users.
Referring to the actual topic network propagation condition, the propagation rule of topics in the social network is defined as follows:
as topic transmission has the characteristics of short time and explosiveness, the invention considers that the number of user nodes in the social network is equal at any time in a fixed time period without considering population power factors such as birth, death, flow and the like, so that the sum S of the user ratios of different states in the same layer model at any time is i +I i +R i ＝1。
Topic transmission is similar to the transmission of infectious diseases, and a certain infection rate must exist when a new user contacts with a user who has transmitted a topic.
When the derived topic is born, the derived topic directly inherits the proportions of susceptible users, infected users and immune users under the predecessor topic.
When the infected node is contacted with the susceptible node on the same topic layer, the susceptible node is converted into the infected node with the probability of lambda under the condition that the influence of other factors is not considered. When psychological factors and prepositive emotion are considered, the probability becomes lambda p Or λ q 。
With the reduction of the heat of the topic and the advance of time, the infected node has a fixed probability
And converting into an immune node. But for a certain node, the immune node can still be converted into a susceptible node of another topic layer by the probability of p, and the other topic layer is interested again.
As shown in fig. 3, the present invention better predicts the forwarding process of the user by quantifying the continuous attention psychological factors. If the topic 0 (precursor topic), the topic 1 (derived topic 1) and the topic 2 (derived topic 2) exist; wherein topic 1 is derived from topic 0, topic 2 is derived from topic 1, topic 1 is more associated with topic 0, and topic 2 is less associated with topic 1. When topic 1 is born, the user network is distinguished by whether topic 0 is propagated or not, and there are two states when the user just contacts the derived topic, namely state X 1 ∈{G 0 S 1 ,G 1 S 1 Using G } 0 S 1 Indicating that the predecessor topic 0 has not been forwarded and is in a derived topic 1 susceptivity state, G 1 S 1 Indicating that predecessor topic 0 was forwarded and is in a derived topic 1 susceptive state. Similarly, if topic 2 is present, and if topic 1 is propagated or not, the user network is differentiated, and there is a state X 2 ∈{G 0 S 2 ,G 1 S 2 And, in the same sense as above.
Suppose for an independent topic, the infection rate is λ, i.e., the probability from the S state to the I state is λ. When the topic 2 is spread, because the association degree of the topic 2 and the topic 1 is small, no matter whether the user forwards the topic 0 or not, the probability of forwarding the topic 2 is not influenced, namely G 0 S 2 And G 1 S 2 State transition to I 2 The probability of (d) is still λ.
For transition probabilities between different classes of users, the following definitions apply:
calculating cross-model state transition probability according to topic relevance and independent forwarding probability of users, converting immune users into susceptible users participating in derived topics according to the cross-model state transition probability, and calculating the cross-model state transition probability according to the formula:
wherein p is ab Represents the cross-model state transition probability, namely the transition probability from the immune user who participated in the topic a to the susceptible user who participated in the topic b, and lambda represents the user independent forwarding probability, rlv (r) a ,r b ) The topic relevance degree of the topic a and the topic b is shown, and the beta is a threshold value of the topic relevance degree.
Because topic 1 is relatively associated with topic 0, if the user has forwarded topic 0, the probability of forwarding topic 1 is influenced by the psychological effect of continuous attention and is increased to λ p (ii) a Calculating the transition probability of the susceptible users participating in the precursor topic according to the topic association degree and the independent forwarding probability of the users, converting the susceptible users into infected users participating in the derived topic according to the transition probability, wherein the calculation formula is as follows:
λ p ＝λ+(1-λ)×Rlv(r a ,r b )
wherein λ is p Represents the transition probability of the susceptible user participating in the predecessor topic, λ represents the independent forwarding probability of the user, rlv (r) a ,r b ) Indicating the topic relevance of topic a and topic b, rlv (r) a ,r b )＝Rlv(r 0 ,r 1 )。
If the user does not forward the topic 0, and the topic 1 is similar to the topic 0, the probability of forwarding the topic 1 is reduced to lambda q Calculating the transition probability of the susceptible users which do not participate in the precursor topic according to the topic association degree and the independent forwarding probability of the users, and converting the susceptible users into infection participating in the derived topic according to the transition probabilityUser, calculating the transition probability G 0 S 1 Conversion to I 1 Probability of (c):
λ q ＝(1-Rlv(r a ,r b ))×λ
wherein λ is q Represents the transition probability of a susceptible user who has not participated in the predecessor topic, and λ represents the independent forwarding probability of the user, rlv (r) a ,r b ) Indicating the topic relevance of topic a and topic b, rlv (r) a ,r b )＝Rlv(r 0 ,r 1 )。
Infecting users with a fixed probability
Conversion into an immunized user, preferably>
S7: and constructing a dynamic equation according to the independent forwarding probability of the users and the conversion probability among different types of users.
Based on an Iterative-SIR model and the propagation rules, dynamic equations of k derived topic layers are constructed according to the independent forwarding probability of users and the conversion probability among different types of users, and the expression is as follows:
further, the kinetic equation is written as:
wherein the content of the first and second substances,
representing the individual forwarding probabilities of the users; alpha represents the relevance Rlv (r) of topic i to its predecessor topic i-1 ,r i )，G 0 Representing a network of users that have not participated in the propagation of predecessor topics,S i (t) and I i (t) shows the proportion of susceptible users and infected users participating in topic i at time t, R i-1 (t) immune user ratio of predecessor topics participating in topic i at time t, G 1 Representing a network of users participating in the propagation of predecessor topics, p (i-1)i Representing the cross-model state conversion probability of the topic i and the predecessor topic thereof, beta representing the topic relevance threshold, S i 、I i 、R i Respectively, the susceptibility, infection and immune user ratios at any one time.
S8: and solving a dynamic equation to obtain a propagation prediction result of the user on the derived topic.
Solving a dynamic equation to obtain a user state set { S) in different topic layers at different moments i (t)}{I i (t)}{R i (t) }, namely, the propagation prediction result of the user on the derived topic.
According to the output result of the derived topic propagation model, the derived topic propagation situation can be simulated and predicted, and the proportion of the user state in the network at each moment and the propagation trend chart of the whole topic are obtained. The public opinion supervision department can monitor the topic transmission situation in real time through the output results and take corresponding control measures according to the real situation and the change situation of the topic transmission, thereby effectively preventing and resolving the public opinion crisis.
The above-mentioned embodiments, which further illustrate the objects, technical solutions and advantages of the present invention, should be understood that the above-mentioned embodiments are only preferred embodiments of the present invention, and should not be construed as limiting the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for predicting propagation of derived topics based on multi-topic emotion measurement is characterized by comprising the following steps:
s1: obtaining topic data, wherein the topic data comprises user historical behavior information and topic participation information;
s2: extracting message factor features and user factor features of the topic data; the message factors comprise topic heat and topic relevance, and the user factors comprise user drive, user activity and topic personal emotion matching degree;
s3: calculating message influence according to topic heat, user power, user activity and topic personal emotion matching;
s4: dividing a user network to obtain a classification result of a user;
s5: calculating the independent forwarding probability of the user according to the message influence;
s6: calculating conversion probability among different types of users according to the topic association degree and the independent forwarding probability of the users;
s7: constructing a dynamic equation according to the independent forwarding probability of the users and the conversion probability among different types of users;
s8: and solving a dynamic equation to obtain a propagation prediction result of the user on the derived topic.
2. The method for predicting propagation of derived topics based on multi-topic emotion metrics as claimed in claim 1, wherein the process of calculating message influence comprises: calculating the influence of message factors according to the topic heat and the user drive; calculating the influence of the user factors according to the activity of the user and the personal emotion matching degree of the topic; and calculating the message influence according to the message factor influence and the user factor influence.
3. The method for predicting propagation of derived topics based on multi-topic emotion metrics as claimed in claim 2, wherein the formula for calculating the message influence is:
Eff(u i )＝β 0 +β 1 *fac user (u i )+β 2 *fac topic (u i )
wherein, eff (u) i ) Representing user u i Message influence of beta 0 Denotes the first partial regression coefficient, beta 1 Represents the second partial regression coefficient, fac user (u i ) Representing user u i To a userFactor influence, beta 2 Represents the third partial regression coefficient, fac topic (u i ) Representing user u i The message factor influences the power.
4. The method for predicting the propagation of the derived topics based on the multi-topic emotion measurement as claimed in claim 1, wherein the process of dividing the user network comprises: dividing users into users who have participated in the predecessor topics and users who have not participated in the predecessor topics; and dividing the users who participate in the predecessor topics and the users who do not participate in the predecessor topics into susceptible users, infected users and immune users again to obtain user classification results.
5. The method for predicting the propagation of the derived topics based on the multi-topic emotion measurement as claimed in claim 1, wherein the formula for calculating the independent forwarding probability of the user is as follows:
where λ represents the user-independent forwarding probability, ψ (t) represents an intermediate parameter, and m represents the user u i N represents the number of neighbors of the forwarded topic, eff (u) i ) Representing user u i The message impact of (2).
6. The method for predicting the propagation of the derived topics based on the multi-topic emotion measurement as claimed in claim 1, wherein the step of calculating the conversion probability between different types of users according to the topic relevance and the independent forwarding probability of the users comprises:
calculating cross-model state transition probability according to topic association degree and independent forwarding probability of users, and converting immune users into susceptible users participating in derived topics according to the cross-model state transition probability;
calculating the transition probability of the susceptible users participating in the precursor topic according to the topic association degree and the independent forwarding probability of the users, and converting the susceptible users into infected users participating in the derived topic according to the transition probability;
calculating the transition probability of the susceptible users which do not participate in the precursor topic according to the topic association degree and the independent forwarding probability of the users, and converting the susceptible users into infected users participating in the derived topic according to the transition probability;
infected users turn into immunized users with a fixed probability.
7. The method of claim 6, wherein the formula for calculating the cross-model state transition probability is as follows:
wherein p is ab Represents the transition probability from the immune user who participates in the topic a to the susceptible user who participates in the topic b, and lambda represents the user independent forwarding probability Rlv (r) a ,r b ) The topic relevance degree of the topic a and the topic b is shown, and the beta is a threshold value of the topic relevance degree.
8. The method for predicting propagation of derived topics based on multi-topic sentiment measurement as claimed in claim 6, wherein the formula for calculating the transition probability of the susceptible users who participated in the predecessor topics is:
λ p ＝λ+(1-λ)×Rlv(r a ,r b )
wherein λ is p Denotes the transition probability of a susceptible user who has participated in the predecessor topic, λ denotes the user's independent forwarding probability, rlv (r) a ,r b ) The topic relevance between the topic a and the topic b is shown.
9. The method for predicting propagation of derived topics based on multi-topic emotion measurement as claimed in claim 6, wherein the formula for calculating the transition probability of the susceptible users who have not participated in the predecessor topics is:
λ q ＝(1-Rlv(r a ,r b ))×λ
wherein λ is q Denotes the transition probability of a susceptible user who has not participated in the predecessor topic, λ denotes the user's independent forwarding probability, rlv (r) a ,r b ) And the topic relevance degree of the topic a and the topic b is shown.
10. The method for predicting the propagation of the derived topics based on the multi-topic emotion measurement as claimed in claim 1, wherein the dynamical equation is expressed as:
wherein the content of the first and second substances,
representing the individual forwarding probabilities of the users; alpha represents the degree of association Rlv (r) of topic i with its predecessor topic i-1 ,r i )，G 0 User network, S, representing non-participation in propagation of predecessor topics i (t) and I i (t) shows the proportion of susceptible users and infected users participating in topic i at time t, R i-1 (t) immune user ratio of predecessor topics participating in topic i at time t, G 1 Representing a network of users participating in the propagation of predecessor topics, p (i-1)i Representing the cross-model state conversion probability of the topic i and the predecessor topic thereof, beta representing the topic relevance threshold, S i 、I i 、R i Respectively, the susceptibility, infection and immune user ratios at any one time. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211489736.0A CN115907165A (en) | 2022-11-25 | 2022-11-25 | Derived topic propagation prediction method based on multi-topic emotion measurement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211489736.0A CN115907165A (en) | 2022-11-25 | 2022-11-25 | Derived topic propagation prediction method based on multi-topic emotion measurement |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115907165A true CN115907165A (en) | 2023-04-04 |
Family
ID=86495143
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211489736.0A Pending CN115907165A (en) | 2022-11-25 | 2022-11-25 | Derived topic propagation prediction method based on multi-topic emotion measurement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115907165A (en) |
-
2022
- 2022-11-25 CN CN202211489736.0A patent/CN115907165A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Shi et al. | mvn2vec: Preservation and collaboration in multi-view network embedding | |
| Xiao et al. | Rumor diffusion model based on representation learning and anti-rumor | |
| CN110795641B (en) | Network rumor propagation control method based on representation learning | |
| CN106651030B (en) | Improved RBF neural network hot topic user participation behavior prediction method | |
| CN105809554B (en) | Prediction method for user participating in hot topics in social network | |
| CN110825948B (en) | Rumor propagation control method based on rumor-splitting message and representation learning | |
| CN113807616B (en) | Information diffusion prediction system based on space-time attention and heterogeneous graph convolution network | |
| CN110336700B (en) | Microblog popularity prediction method based on time and user forwarding sequence | |
| CN109829114B (en) | Topic popularity prediction system and method based on user behaviors | |
| CN110807556B (en) | Method and device for predicting propagation trend of microblog rumors or/and dagger topics | |
| Xiao et al. | A rumor & anti-rumor propagation model based on data enhancement and evolutionary game | |
| CN108229731B (en) | System and method for predicting user behavior under hot topics under multi-message interaction | |
| Xiao et al. | User behavior prediction of social hotspots based on multimessage interaction and neural network | |
| CN113254719B (en) | Online social network information propagation method based on status theory | |
| CN115712772A (en) | Topic propagation prediction method based on topic association | |
| CN111353905B (en) | System for measuring and estimating individual influence of netizen users aiming at network community | |
| CN115470991A (en) | Network rumor propagation prediction method based on user short-time emotion and evolutionary game | |
| Li et al. | An information dissemination model based on the rumor and anti-rumor and stimulate-rumor and tripartite cognitive game | |
| Liu et al. | Information propagation prediction based on spatial–temporal attention and heterogeneous graph convolutional networks | |
| Wu et al. | Evolution prediction of multi-scale information diffusion dynamics | |
| CN113919440A (en) | Social network rumor detection system integrating dual attention mechanism and graph convolution | |
| Yan et al. | Rumor detection on social networks focusing on endogenous psychological motivation | |
| CN113850446B (en) | Information diffusion prediction method integrating space-time attention and heterogeneous graph convolution network | |
| CN115907165A (en) | Derived topic propagation prediction method based on multi-topic emotion measurement | |
| CN112256756B (en) | Influence discovery method based on ternary association diagram and knowledge representation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |