CN108520337B - Riadry risk assessment method based on network risk entropy difference - Google Patents

Riadry risk assessment method based on network risk entropy difference Download PDF

Info

Publication number
CN108520337B
CN108520337B CN201810239750.2A CN201810239750A CN108520337B CN 108520337 B CN108520337 B CN 108520337B CN 201810239750 A CN201810239750 A CN 201810239750A CN 108520337 B CN108520337 B CN 108520337B
Authority
CN
China
Prior art keywords
node
network
rumor
time
risk
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.)
Active
Application number
CN201810239750.2A
Other languages
Chinese (zh)
Other versions
CN108520337A (en
Inventor
肖喜
卞天
刘睿彤
郑海涛
江勇
夏树涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Graduate School Tsinghua University
Original Assignee
Shenzhen Graduate School Tsinghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shenzhen Graduate School Tsinghua University filed Critical Shenzhen Graduate School Tsinghua University
Priority to CN201810239750.2A priority Critical patent/CN108520337B/en
Publication of CN108520337A publication Critical patent/CN108520337A/en
Application granted granted Critical
Publication of CN108520337B publication Critical patent/CN108520337B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Economics (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • Educational Administration (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a rumor risk assessment method based on network risk entropy difference, which comprises the following steps: s1, establishing a rumor propagation model based on an SIR model for a network platform to be evaluated; s2, identifying rumor sources and propagation time by using a rumor source identification method based on the network structure of the rumor propagation model; s3, simulating a rumor forward propagation process in the network structure to obtain the probability that each node in the network is in different states at the current moment; s4, calculating the maximum risk entropy and the minimum risk entropy of the network at the current moment according to the probability that each node is in different states at the current moment; s5, calculating the network risk entropy at the current moment by using the maximum risk entropy and the minimum risk entropy of the network at the current moment; s6, calculating the network risk entropy difference according to the network risk entropy of the network at the current moment and the previous moment, and evaluating the potential risk of the rumor at the current moment on the network platform according to the network risk entropy difference. The invention can quantitatively and accurately evaluate the risk of the rumor on the network in real time.

Description

Riadry risk assessment method based on network risk entropy difference
Technical Field
The invention relates to the technical field of internet and information transmission, in particular to a risk assessment method for network public sentiment, network rumors or network false information and the like based on network risk entropy difference.
Background
With the popularity of the internet, more and more people are beginning to work, shop, study, and entertain with the internet. According to the 40 th statistical report of the development conditions of the Chinese Internet in the China Internet information center (CNNIC) in 8 and 4 months in 2017, the scale of Chinese netizens reaches 7.51 hundred million and accounts for one fifth of the total number of the global netizens when the number of the Chinese netizens reaches 6 months in 2017. Networks have become a medium and platform for people to acquire information, express ideas, and communicate with each other every day. The convenient network platform dredges a free channel for the netizens to express and exchange opinions, and people tend to publish opinion public opinions on the network more and more. Producers of internet public opinion content have been transformed by website producers into a mixed group of websites and vast network users acting together. This trend makes the internet the mainstream media for social public opinion dissemination beyond mass dissemination and interpersonal dissemination of newspapers, radio, television and the like. The network media as a novel media breaks through a plurality of limits of the traditional media on time and space, and widens the spreading range and depth. The network media can propagate information to every corner of the world for 24 hours through the internet. The rapid development of the network brings convenience to people in life, but also inserts wings to rumors. The rapid information dissemination, anonymization of netizens, emotion expression of opinions and extreme standpoints are hidden troubles for the sprouting of network rumors. The scale of the rapidly-growing netizens, the convenient internet surfing mode, the mobile internet intellectualization and various social negative emotions accumulated in the transition period of the economic society are combined with each other, and conditions are provided for breeding and spreading of rumors.
Network rumors refer to a false statement of no facts that are widely spread over a network. Compared with the traditional rumors, the network rumors, whether the propagation speed or the propagation range, are greatly expanded, particularly, the propagation speed and the propagation range are easy to grow and spread along with the generation of major events and emergencies, and meanwhile, the popularity of the internet also enables people on the network to become rumors, so that the network rumors tend to be inundated since the occurrence of the network rumors, and the tendency of the network rumors to be inundated seriously influences the stability and harmony of the society. The wide spread of network rumors is very easy to disturb normal social order, trigger social trust crisis and infringe the personal rights of citizens, so that the method for researching network rumors risk assessment is particularly important.
At present, related research of network rumors is still immature, most rumor risk assessment schemes provided by researchers are based on multi-criterion decision-making algorithms such as an analytic hierarchy process, and evaluation indexes are established in all aspects of rumors to comprehensively score to make risk assessment. The solution has large subjective factors, is difficult to collect data in real time according to the propagation dynamic state, and has inaccurate and timely evaluation results. Therefore, there is a need for a scheme for risk assessment of rumors accurately and in time.
The above background disclosure is only for the purpose of assisting understanding of the inventive concept and technical solutions of the present invention, and does not necessarily belong to the prior art of the present patent application, and should not be used for evaluating the novelty and inventive step of the present application in the case that there is no clear evidence that the above content is disclosed before the filing date of the present patent application.
Disclosure of Invention
Aiming at the defects of the conventional rumor risk assessment method, the invention provides a method which can monitor the propagation condition of network speeches (such as network rumors, public opinions, false information and the like) in real time and make risk quantitative assessment in time.
The technical scheme provided by the invention for overcoming the defects of the prior art is as follows:
a rumor risk assessment method based on network risk entropy difference comprises the following steps:
s1, establishing a rumor propagation model based on the SIR model for the network platform to be evaluated;
s2, based on the network structure of the rumor propagation model, identifying a rumor source and the propagation time thereof by using a rumor source identification method; the network structure comprises a plurality of network nodes, each node representing a network user;
s3, simulating the forward propagation process of the rumor in the network structure according to the rumor source and the propagation time thereof determined in the step S2, so as to obtain the probability that each node in the network structure is in different states at the current moment; wherein the states of a node include three: an infection-susceptible state, an infection state, and a recovery state;
s4, calculating the maximum risk entropy and the minimum risk entropy of the network structure at the current moment according to the probability that each node is in different states at the current moment;
s5, calculating the network risk entropy of the network structure at the current moment by using the maximum risk entropy and the minimum risk entropy of the network structure at the current moment;
s6, calculating the network risk entropy difference of the network structure according to the network risk entropy of the network structure at the current moment and the network risk entropy of the network structure at the previous moment, and evaluating the potential risk of the current moment rumor on the network platform according to the network risk entropy difference.
The invention provides a rumor propagation model by researching possible states of individuals and rumor propagation mechanisms among individuals in the rumor propagation process and based on an SIR model and factors influencing rumor propagation, and can effectively describe the process of rumor propagation in a social network from a rumor source; and the influence of various factors on rumor propagation in reality can be effectively reflected by adjusting the infection probability and the recovery probability in the model, so that the method provided by the invention is more practical and is easy to operate. Then, by calculating the network risk entropy difference analysis rumor risk, the influence of subjective factors in the analytic hierarchy process on the evaluation result is eliminated, and the rumor risk evaluation result is more accurate. Meanwhile, the entropy-based method can be used for analyzing the whole network rumor risk influence by combining the state of each node in the network, and the risk of the network rumor can be reflected comprehensively and objectively. Meanwhile, the model for evaluating the rumor risk based on the network risk entropy difference does not need to collect and construct a large amount of training set data or feature databases, obtains different network risk entropies aiming at different rumors, is not limited by model parameters, and enables the rumor risk evaluation result to be more accurate.
Drawings
Fig. 1 is a flowchart of a rumor risk assessment method based on network risk entropy differences according to the present invention;
FIG. 2 is a schematic diagram of a network structure of a rumor propagation model according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating the state transitions of individuals (nodes) in a rumor propagation model according to an embodiment of the present invention;
fig. 4 is a schematic diagram of finding rumor sources using a back propagation method in the network structure shown in fig. 2.
Detailed Description
The invention is further described with reference to the following figures and detailed description of embodiments.
In order to quantitatively evaluate the risk of rumors, the embodiment of the invention provides a method for evaluating the risk of rumors based on network risk entropy differences. Referring to fig. 1, the rumor risk assessment method of the present invention includes the following steps S1 to S6:
step S1, establishing a rumor propagation model based on the SIR model for the network platform to be evaluated.
The network platform can be, for example, a social network platform having mainstream functions of forwarding, commenting, praising and the like, and the network platform to be evaluated in the method is not limited to a certain platform, can be a platform group formed by a plurality of interconnected platforms, and can evaluate real-time risks possibly caused by rumors in a network formed by the platforms.
An exemplary network structure of the SIR model-based rumor propagation model is shown in fig. 2, which may be regarded as a social network model and is composed of a plurality of network nodes (small squares and small circles in the figure are network nodes, each node represents a network user) and a connection relationship between the nodes, where the connection relationship between two nodes means that network information can be directly propagated between the two nodes, and the network structure is represented by a connecting line.
In the rumor propagation model, three states are defined for each node itself according to the SIR model (typical infectious disease model): the infection status of the rumor is not received or the rumor is not believed to be infected, the rumor is received and the rumor is spread (i.e. the rumor is believed to be infected), and the rumor is transformed from the rumor believed to the recovery status of the rumor not believed to be spread (which is shown as stopping spreading the rumor) after the phase is heard. As shown in fig. 3, based on the SIR model, the individuals in the network are susceptible to infection before receiving the rumor or but not the rumor, and may transit to the infection state with a certain probability α at any time, and during the infection state, transit to the recovery state with a certain probability β. Therefore, the SIR model can effectively characterize the state transition process of each network user to the rumor, and in combination with the network structure shown in fig. 2, can characterize the dynamic process of rumors propagating in the network.
Step S2, based on the network structure of the rumor propagation model, using the rumor source identification method to identify the rumor source and its propagation time. The rumor source identification can be performed by back propagation, or other typical rumor source identification methods such as Jordan center method, effective distance method, monte carlo method, etc. The present invention is described in detail with reference to the following examples, which illustrate the use of back propagation method to search for rumor sources.
Rumors may propagate from any individual in the network, while rumors that propagate from different individuals may traverse different paths, with different effects. Therefore, finding the source of rumor propagation is the first task to study the rumor propagation process and understand the current status of each individual in the network. Referring to fig. 4, firstly, selecting some nodes from the network structure as observation nodes to form an observation node set S, and representing the selected observation nodes by "small square" nodes in the network structure shown in fig. 2; then, observation is carried out at a certain time t, and the observation nodes in the infection state at the time t form a set
Figure BDA0001604954930000041
Set of observation nodes in susceptible state
Figure BDA0001604954930000042
And recording the collection
Figure BDA0001604954930000043
Time T of day when each observation node (assuming that there are m observation nodes in the infected state at time T) is infected with rumor { T }1,t2,t3,…,tm}. To the collection
Figure BDA0001604954930000044
The observation nodes in the network structure are sequentially backward propagated along the network path of the network structure according to the time sequence of respective rumor infection from the node infected at the latest, and can be simultaneously gathered
Figure BDA0001604954930000045
All observed nodes in (1) will be regarded as possible rumor sources and added to suspected rumor source set U. As shown in FIG. 4, utilizing collections
Figure BDA0001604954930000046
The nodes in (1) perform backward propagation demonstration, initially (time τ is 0), backward propagation starts from the node a1 which is infected at the latest (for example, the infection time is at time t), backward propagation starts from the node a2 which is infected at time τ -1, backward propagation starts from the node A3 which is infected at time t-2, and so on. Can eventually be assembled at the same time
Figure BDA0001604954930000051
All observed nodes in (a) will be infected with nodes U as possible rumor sources, and added to the suspected rumor source set U. Finally, respectively calculating the maximum likelihood value at the time t for all the nodes U in the set U, wherein the node with the maximum likelihood value is the rumor source Uf. The maximum likelihood value of the node u at the time t is
Figure BDA0001604954930000052
Wherein, PI(h,th(ii) a u) is node h at thThe time instant is transmitted by the node U in the set UProbability of rumor being infected at infection state at time t, PS(k, t; U) is the probability that node k is in a susceptible state at time t without being infected by the rumor transmitted from node U in set U, and can be calculated by the following equations (2) and (5):
Figure BDA0001604954930000053
wherein, α (h, t)h) Indicates that the node h is at thProbability of being infected at a moment, PS(h,th-1) and PI(h,th-1) respectively represent that the node h is at (t)h-1) probability of being in a susceptible state and in an infected state at the moment, βhRepresents the probability of the node h in the infection state being converted into the recovery state, 0 < betah＜1；UhThe credit value of the user represented by the node h on the network platform can be obtained from the network platform; chIndicating the support of the user represented by node h on the network platform,
Figure BDA0001604954930000054
phthe content released on the network platform for the user represented by the node h contains the number of comments with favorable attitudes, nhAnd the content published on the network platform for the user represented by the node h contains the comment quantity with the objection attitude. And has:
Figure BDA0001604954930000055
Lhactivity of the user represented by node h in the last predetermined period of time, NhIs a set of adjacent nodes j, inf, of a node h in the network structurejInfluence value (e.g. number of microblog fans) P of user represented by node j on the network platformI(j,th-1) is node j at (t)h-1) probability of being in an infected state at a time; and has:
Figure BDA0001604954930000061
Figure BDA0001604954930000062
and
Figure BDA0001604954930000063
respectively representing the number of contents released on the network platform by the user represented by the node h in the last week and the last month,
Figure BDA0001604954930000064
and
Figure BDA0001604954930000065
respectively representing the number of praises of the user represented by node h on the network platform in the last week and the last month,
Figure BDA0001604954930000066
and
Figure BDA0001604954930000067
and respectively representing the number of comments replied to by the user represented by the node h on the network platform in the last week and the last month.
PS(k,t；u)＝(1-α(k,t))·PS(k,t-1) (5)
α (k, t) represents the probability that node k is infected at time t, PS(k, t-1) represents the probability that node k is in a susceptible state at time (t-1).
Meanwhile, for each node u which may be a rumor source, the corresponding rumor propagation time can be calculated by the following formula:
Figure BDA0001604954930000068
rumor source ufAnd its rumor propagation time tfComprises the following steps:
Figure BDA0001604954930000069
step S3, simulating the forward propagation process of the rumor in the network structure according to the rumor source and the propagation time thereof determined in step S2, so as to obtain the probability that each node in the network structure is in different states at the current moment. In finding the rumor source ufAnd calculating propagation time t of rumorfLater, we simulated the dynamic propagation process of rumors using the SIR model, and we source the rumors u at the initial timefAn infected state is defined, while other nodes in the network are all defined as susceptible to infection. According to the user characteristics of the individuals in the social network, the factors influencing the infection probability of the individuals are defined to be the recent activity and the influence of adjacent individuals, and the factors influencing the recovery probability of the individuals after infection are defined to be the user credit and the user support degree. Thus, at time t, the probability that node i in the network structure in the susceptible state is infected by rumor at time t is
Figure BDA0001604954930000071
In the formula (8), L is the same as in the formula (3)iActivity degree N of the user represented by the node i in the latest preset time period on the network platformiFor a set of adjacent nodes j, inf, of a node i in the network structurejIs the influence value of the user represented by the node j on the network platform, PI(j, t-1) is the probability that node j is in the infected state at time (t-1). Likewise, LiThe calculation of (2) is carried out by adopting a formula (4) and substituting the formula into corresponding parameters for calculation.
Next, the probability that each node is in a different state at time t is calculated. Calculating the probability P that the node i is in a susceptible state at the time t according to the probability alpha (i, t) that the node i is infected by the rumor at the time tS(i,t)：
PS(i,t)＝(1-α(i,t))·PS(i,t-1) (9)
Probability P that node i is in an infected state at time tI(i,t)：
Figure BDA0001604954930000072
And the probability P that node i is in the recovery state at time tR(i,t)：
Figure BDA0001604954930000073
βiRepresents the probability of the node i in the infection state to be converted into the recovery state, 0 < betai＜1；UiRepresenting a credit value of a user represented by a node i on the network platform; ciRepresenting the support of the user represented by node i on the network platform,
Figure BDA0001604954930000074
pithe content published on the network platform for the user represented by the node i contains the number of comments with favorable attitudes, niAnd the content published on the network platform for the user represented by the node i contains the comment quantity with the objection attitude.
After the probabilities of different states of each node in the network structure are obtained through the calculation, the method introduces the concept of network entropy to quantitatively analyze the magnitude of the rumor risks in the network. The concept of "entropy" was created in 1856 by the physicist clausius and thermodynamics measures the degree of energy depletion in matter systems using entropy. In subsequent studies, a large group of scientists such as Boltzmann and Einstein have made deeper knowledge of entropy. In the 40 th of the 20 th century, shannon introduced entropy into information theory to generate modern information science, and entropy is used for quantitatively describing information quantity, and the following formula is the definition of shannon on information entropy:
Figure BDA0001604954930000081
where H denotes the average information content of the source (the entity generating the information), PzRepresenting the probability of the occurrence of the z-th source.
For the qth performance indicator of the network, its network entropy value may be defined as:
Hq＝-log2Vq,q＝1,2,...,Q (13)
wherein VqAnd the normalized performance parameter is the q-th performance index.
The network entropy of the network system should be the weighted sum of the individual performance index entropies, namely:
Figure BDA0001604954930000082
wherein ω isqIs the weight of the q-th performance index.
Therefore, the "entropy difference" can be adopted "
Figure BDA0001604954930000083
The effect of attacks on a network system is described, where Δ Hq＝-log2(Vq2/Vq1) For the effect of the q-th performance index of the network being attacked, Vq1Normalized performance parameter, V, for the original qth performance index of the network systemq2The performance index is the normalized performance parameter of the q-th performance index after the network system is attacked. Obviously,. DELTA.HnetThe larger the size, the more the overall performance of the network system becomes worse, and the more obvious the attack effect becomes.
And step S4, calculating the maximum risk entropy and the minimum risk entropy of the network structure at the current time according to the probabilities of the nodes in different states at the current time obtained in the step S3.
The infection mechanism according to the SIR model is known: node x, which is susceptible to infection at time t, is represented by P at time t +1SThe probability of (x, t +1) remains susceptible while at the same time being PI(x, t +1) is infected into an infectious state; node y in the infected state at time t +1 is denoted by PRThe probability of (y, t +1) is converted into recoveryPlural states simultaneously with PIThe probability of (y, t +1) remains infected. Probability P that node x in susceptible state keeps susceptible state at t +1 momentSThe smaller (x, t +1), the probability P that the node y in the infected state transitions to the recovery state at time t +1RThe smaller (y, t +1), the greater the risk in the network. Therefore, we can define the maximum risk entropy H of the network at the moment tmax(t) is:
Figure BDA0001604954930000091
wherein:
Figure BDA0001604954930000092
GSand GIAnd respectively representing node sets which are susceptible to infection states and infection states in the network structure at the current moment.
Similarly, the probability P that the node x in the susceptible state at the time t is infected into the infected state at the time t +1IThe smaller (x, t +1), the probability P that the node y in the infected state remains in the infected state at time t +1IThe smaller (y, t +1), the less risk is posed by rumors in the network at this time. From this we can get the minimum risk entropy of the network at time t as
Figure BDA0001604954930000093
Wherein the content of the first and second substances,
Figure BDA0001604954930000094
thus, the network risk entropy h (t) of the network structure at time t is:
H(t)＝Hmax(t)-Hmin(t) (17)
the network risk entropy difference Δ h (t) of the network structure at time t is:
ΔH(t)＝H(t)-H(t-1)＝(Hmax(t)-Hmin(t))-(Hmax(t-1)-Hmin(t-1))。
therefore, when the maximum risk entropy value of the network at the time t is larger and the minimum risk entropy value is smaller, and the maximum risk entropy value at the time t-1 is smaller and the minimum risk entropy value is larger, the larger the value of the network risk entropy difference at the time t is, the larger the potential risk of the current network rumor is represented. At this time, the strength should be increased to control and inhibit further diffusion of rumors, so as to avoid influence on society.
The existing rumor propagation model needs to know the propagation source of the rumor in advance, and the dynamic propagation process is started from the propagation source, which is often difficult to acquire in advance in real life. Because in reality most rumors are not concerned until after they have been spread for some time, it is difficult to determine where in social networks the spread of rumors began.
Meanwhile, the method for evaluating the rumor risk based on the network risk entropy difference does not need to collect and construct a large amount of training set data or feature databases, obtains different network risk entropies aiming at different rumors, is not limited by model parameters, and enables the rumor risk evaluation result to be more accurate. The entropy-based method can analyze the whole network rumor risk in combination with the state of each node in the dynamic rumor propagation network, and can reflect the risk of the network rumor more comprehensively and quantitatively.
The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several equivalent substitutions or obvious modifications can be made without departing from the spirit of the invention, and all the properties or uses are considered to be within the scope of the invention.

Claims (5)

1. A rumor risk assessment method based on network risk entropy difference comprises the following steps:
s1, establishing a rumor propagation model based on the SIR model for the network platform to be evaluated;
s2, based on the network structure of the rumor propagation model, identifying a rumor source and the propagation time thereof by using a rumor source identification method; the network structure comprises a plurality of network nodes, each node representing a network user;
s3, simulating the forward propagation process of the rumor in the network structure according to the rumor source and the propagation time thereof determined in the step S2, so as to obtain the probability that each node in the network structure is in different states at the current moment; wherein the states of a node include three: an infection-susceptible state, an infection state, and a recovery state;
s4, calculating the maximum risk entropy and the minimum risk entropy of the network structure at the current moment according to the probability that each node is in different states at the current moment;
s5, calculating the network risk entropy of the network structure at the current moment by using the maximum risk entropy and the minimum risk entropy of the network structure at the current moment;
s6, calculating a network risk entropy difference of the network structure according to the network risk entropy of the network structure at the current moment and the network risk entropy of the network structure at the previous moment at the current moment, and evaluating the potential risk of the current moment rumor on the network platform according to the network risk entropy difference;
in step S3, the current time is defined as time t, and the probability that a node i in a susceptible state in the network structure is infected by a rumor at time t is defined as
Figure FDA0003215074760000011
Wherein L isiActivity degree N of the user represented by the node i in the latest preset time period on the network platformiFor a set of adjacent nodes j, inf, of a node i in the network structurejIs the influence value of the user represented by the node j on the network platform, PI(j, t-1) is the probability that node j is in the infection state at time (t-1);
Figure FDA0003215074760000012
wherein the content of the first and second substances,
Figure FDA0003215074760000013
and
Figure FDA0003215074760000014
respectively representing the number of contents released on the network platform by the user represented by the node i in the last week and the last month,
Figure FDA0003215074760000015
and
Figure FDA0003215074760000016
respectively representing the number of praises made on the network platform by the user represented by node i in the last week and the last month,
Figure FDA0003215074760000017
and
Figure FDA0003215074760000018
respectively representing the number of comments replied by the user represented by the node i on the network platform in the last week and the last month;
calculating the probability P of node i in a susceptible state at the time t according to the probability that the node i is infected by the rumor at the time tS(i,t)＝(1-α(i,t))·PS(i, t-1), probability that node i is in an infected state at time t
Figure FDA0003215074760000021
And, probability that node i is in a recovery state at time t
Figure FDA0003215074760000022
Wherein, betaiRepresents the probability of the node i in the infection state to be converted into the recovery state, 0 < betai＜1；UiRepresenting a credit value of a user represented by a node i on the network platform; ciRepresenting the support of the user represented by node i on the network platform,
Figure FDA0003215074760000023
pithe content published on the network platform for the user represented by the node i contains the number of comments with favorable attitudes, niThe content published on the network platform for the user represented by the node i contains the comment quantity with the objection attitude;
maximum risk entropy H of the network structure at the current momentmax(t) and minimum risk entropy Hmin(t) is calculated by the following formulas, respectively:
Figure FDA0003215074760000024
Figure FDA0003215074760000025
wherein the content of the first and second substances,
Figure FDA0003215074760000026
Figure FDA0003215074760000027
wherein G isSAnd GIRespectively representing node sets which are susceptible to infection states and infection states in the network structure at the current moment;
the network risk entropy H (t) of the network structure at time t is H (t) ═ Hmax(t)-Hmin(t)；
the network risk entropy difference Δ h (t) of the network structure at time t is:
ΔH(t)＝H(t)-H(t-1)＝(Hmax(t)-Hmin(t))-(Hmax(t-1)-Hmin(t-1))；
the larger the cyber risk entropy difference Δ h (t), the larger the potential risk posed by rumors; conversely, the less potential risk is posed by rumors.
2. A rumor risk assessment method according to claim 1, wherein: the rumor source identification method used in step S2 includes a back propagation method, a Jordan center method, an effective distance method, and a monte carlo method.
3. A rumor risk assessment method according to claim 1, wherein: in step S2, finding a rumor source by using a back propagation method includes:
selecting a plurality of nodes in the network structure as observation nodes to form a set S, observing at the time t, and forming a set by the observation nodes in the infection state at the time t
Figure FDA0003215074760000033
Set of observation nodes in susceptible state
Figure FDA0003215074760000034
Recording collection
Figure FDA0003215074760000035
The time when each observation node is infected by the rumor; to the collection
Figure FDA0003215074760000036
The observation nodes in the network structure are sequentially propagated along the path of the network structure in reverse directions according to the time sequence of respective rumor infection from the node infected at the latest, and can be simultaneously gathered
Figure FDA0003215074760000037
All nodes infected with the observation nodes in (1) are regarded as suspected rumors and added into the suspected rumorsRumor source set U;
respectively solving the maximum likelihood value at the time t for all the nodes U in the set U, wherein the node with the maximum likelihood value is the rumor source Uf。
4. A rumor risk assessment method according to claim 3, wherein: the maximum likelihood value of each node U in the set U at the time t is
Figure FDA0003215074760000031
Wherein, PI(h,th(ii) a u) is node h at thProbability of time being infected by rumors from nodes U in set U and being in infected state at time t, PS(k, t; U) is the probability that node k is in a susceptible state at time t without being infected by the rumor transmitted from node U in set U;
Figure FDA0003215074760000032
indicates that the node h is at thProbability of being infected at a moment, PS(h,th-1) and PI(h,th-1) respectively represent that the node h is at (t)h-1) probability of being in a susceptible state and in an infected state at the moment, βhRepresents the probability of the node h in the infection state being converted into the recovery state, 0 < betah＜1；UhRepresenting a credit value of a user represented by a node h on the network platform; chIndicating the support of the user represented by node h on the network platform,
Figure FDA0003215074760000041
phthe content released on the network platform for the user represented by the node h contains the number of comments with favorable attitudes, nhThe content published on the network platform for the user represented by the node h contains the number of comments with objection attitude;
Figure FDA0003215074760000042
Lhactivity of the user represented by node h in the last predetermined period of time, NhIs a set of adjacent nodes j, inf, of a node h in the network structurejIs the influence value of the user represented by the node j on the network platform, PI(j,th-1) is node j at (t)h-1) probability of being in an infected state at a time;
Figure FDA0003215074760000043
Figure FDA0003215074760000044
and
Figure FDA0003215074760000045
respectively representing the number of contents released on the network platform by the user represented by the node h in the last week and the last month,
Figure FDA0003215074760000046
and
Figure FDA0003215074760000047
respectively representing the number of praises of the user represented by node h on the network platform in the last week and the last month,
Figure FDA0003215074760000048
and
Figure FDA0003215074760000049
respectively representing the number of comments replied by the user represented by the node h on the network platform in the last week and the last month;
PS(k,t；u)＝(1-α(k,t))·PS(k, t-1), α (k, t) denotes the probability that node k is infected at time t, PS(k, t-1) represents the probability that node k is in a susceptible state at time (t-1).
5. A rumor risk assessment method according to claim 4, wherein: rumor propagation time of each node U in the set U
Figure FDA00032150747600000410
Rumor source ufAnd its rumor propagation time tfComprises the following steps:
Figure FDA00032150747600000411
CN201810239750.2A 2018-03-22 2018-03-22 Riadry risk assessment method based on network risk entropy difference Active CN108520337B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810239750.2A CN108520337B (en) 2018-03-22 2018-03-22 Riadry risk assessment method based on network risk entropy difference

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810239750.2A CN108520337B (en) 2018-03-22 2018-03-22 Riadry risk assessment method based on network risk entropy difference

Publications (2)

Publication Number Publication Date
CN108520337A CN108520337A (en) 2018-09-11
CN108520337B true CN108520337B (en) 2021-09-24

Family

ID=63432951

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810239750.2A Active CN108520337B (en) 2018-03-22 2018-03-22 Riadry risk assessment method based on network risk entropy difference

Country Status (1)

Country Link
CN (1) CN108520337B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109522464A (en) * 2018-10-22 2019-03-26 西南石油大学 Information source detection method and system
CN112597699B (en) * 2020-12-14 2023-03-14 新疆师范大学 Social network rumor source identification method integrated with objective weighting method
CN112667784B (en) * 2021-01-14 2022-04-05 浙江工商大学 Rumor restraining method based on weighted reverse sampling

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101853261A (en) * 2009-11-23 2010-10-06 电子科技大学 Network public-opinion behavior analysis method based on social network
CN103049643A (en) * 2012-11-22 2013-04-17 无锡南理工科技发展有限公司 Mobile ad hoc network security risk assessment method based on risk entropy method and markoff chain method
CN106411904A (en) * 2016-10-10 2017-02-15 华侨大学 Network risk control method based on microstate prediction

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101853261A (en) * 2009-11-23 2010-10-06 电子科技大学 Network public-opinion behavior analysis method based on social network
CN103049643A (en) * 2012-11-22 2013-04-17 无锡南理工科技发展有限公司 Mobile ad hoc network security risk assessment method based on risk entropy method and markoff chain method
CN106411904A (en) * 2016-10-10 2017-02-15 华侨大学 Network risk control method based on microstate prediction

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
SIR rumor spreading model in the new media age;Laijiun Zhao,et al.;《Physica A:Statistical Mechanics and its Applications》;20130215;第392卷(第4期);全文 *
基于信息熵的谣言信息度量方法;陈曦等;《华中科技大学学报(自然科学版)》;20140110;全文 *
基于新浪微博的金融谣言识别方法探索;黄烨;《金融科技时代》;20180310(第03期);全文 *

Also Published As

Publication number Publication date
CN108520337A (en) 2018-09-11

Similar Documents

Publication Publication Date Title
Chen et al. Social media visual analytics
CN106980692A (en) A kind of influence power computational methods based on microblogging particular event
Tomas et al. The effect of differential recruitment, non-response and non-recruitment on estimators for respondent-driven sampling
Skaza et al. Modeling the infectiousness of Twitter hashtags
CN102945279B (en) The appraisal procedure of microblog users influence power and device
CN110213164B (en) Method and device for identifying network key propagator based on topology information fusion
CN108520337B (en) Riadry risk assessment method based on network risk entropy difference
CN108229731B (en) System and method for predicting user behavior under hot topics under multi-message interaction
CN106789338B (en) Method for discovering key people in dynamic large-scale social network
Li et al. A trust evaluation scheme for complex links in a social network: a link strength perspective
Huang et al. Towards transdisciplinary impact of scientific publications: A longitudinal, comprehensive, and large-scale analysis on Microsoft Academic Graph
Zachreson et al. The effects of local homogeneity assumptions in metapopulation models of infectious disease
Ackland et al. Political homophily on the web
CN106156123A (en) Active value computational methods and device
Wu et al. Embedding funding consultation in library services: a co-occurrences network analysis of knowledge flow in scientific funding
CN106960401A (en) Emolument of obtaining employment predictor method, the method and device for recommending college entrance examination universities and colleges
CN105447196A (en) Key blogger tracking confirmation method and device
CN109492677A (en) Time-varying network link prediction method based on bayesian theory
Koseoglu et al. Co-Authorship networks in business ethics: A longitudinal study
Yin et al. The role of proactive behavior on COVID-19 infordemic in the Chinese Sina-Microblog: a modeling study
Greetham et al. Interventions in social networks: impact on mood and network dynamics
Luo et al. The correlation between social tie and reciprocity in social media
CN118035526B (en) Social media elite node identification method and system
Li et al. Construction of vaccination network and influencing factors: a case study of Chongqing, China
Cui et al. The study of customer involved service innovation under the crowdsourcing: A case study of MyStarbucksIdea. com

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
GR01 Patent grant
GR01 Patent grant