Multi-layer space rumor propagation prediction method and device based on user cognition

Abstract

The invention relates to the field of rumor propagation research, in particular to a rumor propagation prediction method and device based on user cognition and multi-layer space, and the method comprises the steps: obtaining user positive degree, user information perceptibility and user behavior data; analyzing data according to a multiple linear regression method to obtain an internal factor, an external factor and a message propagation driving force which influence the user; calculating the state probability of the user according to the internal factors and the external factors by adopting an SIDRR model; calculating information entropy according to the state probability of the user; and analyzing the message propagation driving force and the information entropy by adopting a long short-term memory network to obtain a rumor information propagation situation. According to the method, modeling is carried out on user interests of different space layers, internal and external factor conditions of the user are extracted, user cognition is analyzed in combination with the improved SIR model, the influence of subsequent user state conversion on rumor spreading is further considered, and therefore the spreading situation of rumor information is effectively predicted.

Description

Technical Field

[0001] The present invention relates to the field of rumor propagation research, and specifically relates to a method for predicting rumor propagation based on user cognition and multi-layer space. Background Art

[0002] In modern information society, the propagation speed and scope of rumors have increased significantly, posing a serious threat to social stability and public trust. In recent years, research has paid more attention to and proved the harm of rumor propagation to society, and many articles have been published in major journals. Further understanding the law of rumor propagation and controlling rumor propagation is of great significance to society.

[0003] Currently, the research on rumors in social networks mainly focuses on two major directions: one is to construct a rumor propagation model based on the classical epidemic model, and analyze the rumor propagation trend by improving and perfecting the SIR model; the other is to further study the law of rumor propagation by mining information features in different spatial layers based on machine learning and deep learning algorithms.

[0004] However, the existing technical solutions do not consider the differences in user cognition and the symbiosis and opposition of user awareness in different spatial layers. Summary of the Invention

[0005] In view of this, the present invention proposes a method and device for predicting rumor propagation based on user cognition and multi-layer space to solve the above problems, including:

[0006] S1. Obtain user enthusiasm, user information perception, user behavior data, and message popularity information;

[0007] S2. Use the multiple linear regression method to calculate user enthusiasm, user information perception, user behavior data, and message popularity information to obtain internal factors, external factors, and message propagation driving forces that affect users;

[0008] S3. Use the SIDRR model to calculate the state probability of each user according to internal factors and external factors;

[0009] S4. Calculate the information entropy according to the state probability of the user;

[0010] S5. Use the long short-term memory network to analyze the message propagation driving force and information entropy to obtain the rumor information propagation situation;

[0011] To achieve the above object, the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements any one of the above methods for predicting rumor propagation based on user cognition and multi-layer space.

[0012] To achieve the above object, the present invention further provides a prediction device for user cognition and multi-layer space rumor propagation, including a processor and a memory; the memory is used to store a computer program; the processor is connected to the memory and is used to execute the computer program stored in the memory, so that the prediction device for user cognition and multi-layer space rumor propagation executes any of the above prediction methods for user cognition and multi-layer space rumor propagation.

[0013] The beneficial effects of the present invention include:

[0014] By modeling the interests of users in different space layers, extracting internal and external factor conditions of users, and further analyzing user cognition by combining game theory with an improved SIR model, considering the impact of the transformation of subsequent user states on rumor propagation, the propagation trend of rumor information can be effectively predicted. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the prediction method for user cognition and multi-layer space rumor propagation of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] In order to make the object, technical solution, features and advantages of the present invention clearer and more understandable, the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] S1. Obtain user enthusiasm, user information perception ability, and user behavior data.

[0018] S2. Use the multiple linear regression method to calculate user enthusiasm, user information perception ability, user behavior data, and message popularity information to obtain internal factors, external factors, and message propagation driving forces that affect users.

[0019] Specifically, the calculation formulas for internal factors, external factors, and message propagation driving forces are:

[0020] fac in (v i ) = Eng(v i ) * Awa(v i )

[0021] fac out (v i ) = P(t) * Rlv(li, lj) * Dri(v i )

[0022]

[0023] Among them, Eng(v i ) is user enthusiasm, Awa(v i) is the user information perception ability, P(t) is the message popularity, Rlv(li, lj) is the correlation degree of different spatial layers, and Dri(v i ) represents the message propagation driving force of node v i . Drf j represents the message propagation driving force in the j-th spatial layer. represents the k-th state of the user in the j-th spatial layer, and α k is the influence factor of the k-th state.

[0024] Specifically, in the above formula, the message popularity P(t), the correlation degree of different spatial layers Rlv(li, lj), and the message propagation driving force Dri(v i ) are calculated as follows:

[0025]

[0026] where p 0 represents the value when the information popularity reaches the peak during the message propagation process, ΔT represents the duration of the message popularity decay, and T represents the half-life of the message popularity decay; spatial(li) and spatial(lj) respectively represent the feature points of two spatial layers, and Eng(v j ) represents the positivity of the neighbor node v j , and Awa(v j ) represents the information perception degree of the neighbor node v j , and N is the number of neighbor nodes of node v j .

[0027] S3. Adopt the SIDRR model to calculate the state probability of each user according to internal factors and external factors.

[0028] Specifically, the SIDRR model classifies users into five states: susceptible state S, rumor-spreading state I, skeptical state D, rejection state R 1 , and immune state R 2 to distinguish different attitudes of users towards rumors;

[0029] Further, through the dynamic equations of the SIDRR model, the probabilities of users in different states can be calculated, providing a basis for subsequent information entropy calculation and rumor propagation trend prediction.

[0030] The dynamic equations adopted by the SIDRR model are:

[0031]

[0032] S(t)+I(t)+D(t)+R 1 (t)+R 2 (t)=1

[0033] Among them, S(t), I(t), D(t), R 1 (t), R 2 (t) represent the susceptible state S, the rumor-spreading state I, the skeptical state D, and the rejecting state R 1 , and the immune state R 2 as functions of time, P 1 is the probability that a susceptible node transforms into a rumor-spreading node, P 2 is the probability that a susceptible node transforms into a skeptical node, P 3 is the probability that a susceptible node transforms into a rejecting node, P 4 is the probability that a rumor-spreading node transforms into a skeptical node, P 5 is the probability that a rumor-spreading node transforms into a rejecting node, where P 1 , P 2 , P 3 , P 4 , P 5 ∈[0,1], λ 1 is the immunization rate of rumor-spreading nodes, λ 2 is the immunization rate of skeptical nodes, λ 3 is the immunization rate of rejecting nodes.

[0034] Furthermore, according to the probabilities P 1 ~P 5 , the probability T 1 that users support the rumor and the probability T 2 that users support the refutation of the rumor are obtained. The calculation formula is:

[0035]

[0036] Furthermore, the formulas used to calculate the probabilities of different user states include:

[0037] Benefit Rumor (v i )=T 1 *(α 1 +α 2 *fac in (v i )+α 3 *fac′ out (v i ))

[0038] Benefit CountertRumor (v i )=T 2 *(α 1 +α 2 *fac in (v i )+α 3 *fac″out (v i ))

[0039]

[0040]

[0041] Among them, Benefit Rumor is the utility value for users to support rumors, and Benefit CountertRumor is the utility value for users to support rumor refutation information. α 1 , α 2 , α 3 are all regression coefficients of the multiple linear model. α 2 , α 3 respectively represent the proportion of internal and external factors in the information influence. T 1 represents the probability that users support rumors, and T 2 represents the probability that users support rumor refutation. σ Rumor (t) is the driving force for users to choose to spread rumor information, and σ CounterRumor (t) is the driving force for users to choose to spread rumor refutation information. represents the probability of the k-th node in the spatial layer j, and wj represents the weights of all states in the spatial layer, w k represents the weight of the specific state k in the spatial layer.

[0042] S4. Calculate the information entropy based on the state probability of the user.

[0043] Specifically, the formula for calculating the information entropy H(Drf j ) is:

[0044]

[0045] Among them, represents the probability of the k-th node in the spatial layer j.

[0046] S5. Analyze the message propagation driving force and the information entropy using a long short-term memory network to obtain the rumor information propagation trend.

[0047] Furthermore, LSTM is used to process the time series data of the message propagation driving force Drf t and the information entropy H to predict the propagation trend. The time series x t input and output [Drf t+1 , H(Drf t+1 )] formula is:

[0048] x t = [Drf t , H(Drf t+1)]

[0049] Drf t+1 = W d x t + b d

[0050] H(Drf t+1 ) = W h x t + b h

[0051] Among them, W d , W h respectively represent the weight matrix of the message propagation driving force Drf and the weight matrix of the information entropy H, b d , b h respectively represent the corresponding bias terms, Drf t+1 , H(Drf t+1 ) respectively represent the predicted message propagation driving force and information entropy at the next moment.

[0052] Furthermore, the analysis of user behavior data and information entropy includes: calculating the binomial distribution of the change in the propagation state with the number m of state changes in neighbor nodes. The calculation formula is:

[0053]

[0054] Among them, μ is the number of neighbor nodes of user v i , P(X = m) is the binomial distribution of state transitions, m is the number of binomial distributions that conform to state transitions, and the range is 1 to μ, Drf t+1 is the message propagation driving force.

[0055] Furthermore, draw the curve of the binomial distribution, that is, the rumor information propagation trend. The binomial distribution reflects the probability value P(X = m) under different numbers m of neighbor node state transitions. The curve has the following trends as m changes:

[0056] When m is small, that is, when the states of a small number of neighbor nodes change, P(X = m) usually shows an upward trend. This is because the driving force Drf(v i ) will cause the message to be propagated to more neighbor nodes, resulting in an initial increase in P(X = m).

[0057] The binomial distribution usually reaches a peak at a certain m value. This peak corresponds to the number of neighbor nodes that the propagation is most likely to affect. Further analysis is as follows:

[0058] The higher the driving force Drf(v i ), the more the peak will shift towards a larger m, that is, the higher the possibility of propagating to more neighbor nodes.

[0059] The larger the entropy value, the weight factor is smaller. When the uncertainty of the system is large, the propagation power is weakened, and the peak will shift to a smaller m, indicating that the uncertainty of the current information system is large, the weakening of the driving force is obvious, and the range of rumor propagation will decrease.

[0060] When m approaches μ, the curve will decline. This is because when most of the neighbor nodes have been covered by the propagation, the probability of further propagation gradually decreases.

[0061] By observing the peak position and width of the binomial distribution, the range of propagation of a certain rumor and the number of groups that may be affected can be analyzed: If Drf(v i ) is higher, and the peak corresponds to a larger m, it indicates that the rumor has a higher propagation ability and will affect more nodes.

[0062] The present invention further includes a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the above-mentioned prediction method for rumor propagation based on user cognition and multi-layer space is implemented.

[0063] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to a computer program. The foregoing computer program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: various media such as ROM, RAM, magnetic disk, or optical disc that can store program codes.

[0064] A prediction device for rumor propagation based on user cognition and multi-layer space includes a processor and a memory; the memory is used to store a computer program; the processor is connected to the memory and is used to execute the computer program stored in the memory, so that the prediction device for rumor propagation based on user cognition and multi-layer space executes any of the above-mentioned prediction methods for rumor propagation based on user cognition and multi-layer space.

[0065] Specifically, the memory includes: various media such as ROM, RAM, magnetic disk, USB flash drive, memory card, or optical disc that can store program codes.

[0066] Preferably, the processor may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0067] Finally, it should be noted that the above description only presents some embodiments of the present invention. For those skilled in the art, various changes, modifications, substitutions, and variations can be envisioned without departing from the principles and spirit of the present invention. The protection scope of the present invention is defined by the appended claims and their equivalents, and all of the above actions should be covered within the protection scope of the present invention.

Claims

1. A method for predicting rumor propagation based on user cognition and multi-layer space, characterized in that: include: S1. Obtain user activity, user information perception, user behavior data, and message popularity information; S2. Use the multivariate linear regression method to calculate user enthusiasm, user information perception, user behavior data, and message popularity information to obtain internal factors, external factors, and message dissemination driving forces that affect users; S3, using the SIDRR model to calculate the state probability of each user based on internal and external factors; S4. Calculate information entropy based on the user's state probability; S5. Use long short-term memory network to analyze the driving force of message propagation and information entropy to obtain the spread trend of rumor information.

2. The method for predicting rumor propagation based on user cognition and multi-layer space according to claim 1 is characterized in that: The formula for calculating the state probability of each user is: Benefit Rumor (v i )=T1*(α1+α2*fac in (v i )+α3*fac′ out (v i )) Benefit CounterRumor (v i )=T2*(α1+α2*fac in (v i )+α3*fac″ out (v i )) Among them, Benefit Rumor Benefit is the utility value of users supporting rumors. CountertRumor is the utility value of users supporting rumor-refuting information, α1, α2, and α3 are all regression coefficients of the multivariate linear model, where α2 and α3 represent the proportion of internal factors in information influence and the proportion of external factors in information influence, respectively, T1 represents the probability of users supporting rumors, T2 represents the probability of users supporting rumor-refuting, and σ Rumor (t) is the driving force for users to choose to spread rumor information, σ CounterRumor (t) It is the driving force for users to choose to spread rumor-refuting information. It is expressed as the probability of the kth state of spatial layer j, w j represents the weights of all states in the spatial layer, w k represents the weight of a specific state k in the spatial layer.

3. The method for predicting rumor propagation based on user cognition and multi-layer space according to claim 1 is characterized in that: The formula for calculating information entropy is: Among them, H(Drf j ) represents information entropy, represents the probability of the kth node in the jth spatial layer.

4. The method for predicting rumor propagation based on user cognition and multi-layer space according to claim 1 is characterized in that: The SIDRR model divides users into five states: susceptible state S, rumor spreading state I, suspicious state D, rejected state R1, and immune state R2.

5. The method for predicting rumor propagation based on user cognition and multi-layer space according to claim 4 is characterized in that: The kinetic equations used in the SIDRR model are: S(t)+I(t)+D(t)+R1(t)+R2(t)=1 Among them, S(t), I(t), D(t), R1(t), and R2(t) respectively represent the susceptible state S, rumor-spreading state I, suspicious state D, rejection state R1, and immune state R2, which are functions of time. P1 is the probability of a susceptible node transforming into a rumor-spreading node, P2 is the probability of a susceptible node transforming into a suspicious node, P3 is the probability of a susceptible node transforming into a rejection node, P4 is the probability of a rumor-promoting node transforming into a suspicious node, and P5 is the probability of a rumor-promoting node transforming into a rejection node. Among them, P1, P2, P3, P4, P5∈[0,1], λ1 is the immunity rate of rumor-spreading nodes, λ2 is the immunity rate of suspicious nodes, and λ3 is the immunity rate of rejection nodes.

6. The method for predicting rumor propagation based on user cognition and multi-layer space according to claim 1 is characterized in that: The analysis of the driving force of message propagation and information entropy includes: using LSTM to process the time series data of the driving force of message propagation and information entropy, calculating the binomial distribution of the change in the number of state changes m; and drawing the binomial distribution curve, that is, the rumor information propagation trend.

7. The method for predicting rumor propagation based on user cognition and multi-layer space according to claim 6, characterized in that: The formula for calculating the binomial distribution is: Where μ is the user v i The number of neighbor nodes, P(X=m) is the binomial distribution of state transition, m is the number of binomial distributions that conform to state transition, and the range of m is 1~μ, Drf t+1 Represents the driving force of message propagation, H(Drf t+1 ) represents information entropy, Represents weight.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the method for predicting rumor propagation based on user cognition and multi-layer space as described in any one of claims 1 to 7.

9. A method for predicting rumor propagation based on user cognition and multi-layer space, characterized in that: It includes a processor and a memory; the memory is used to store computer programs; the processor is connected to the memory and is used to execute the computer programs stored in the memory, so that the device for predicting the propagation of rumors based on user cognition and multi-layer space executes the method for predicting the propagation of rumors based on user cognition and multi-layer space as described in any one of claims 1 to 7.

Images