Research article Special Issues

Dynamic analysis and optimal control of rumor propagation models considering different education levels and hesitation mechanisms

  • Received: 08 April 2024 Revised: 21 May 2024 Accepted: 07 June 2024 Published: 20 June 2024
  • MSC : 34D20, 49J15, 37D35

  • The spread of rumors has an important impact on the production and life of human society. Moreover, in the process of rumor propagation, individuals with different educational levels show different degrees of trust and ability to spread rumors. Therefore, a new rumor propagation model was established, which considers the influence of education level on rumor propagation. Initially, the basic reproduction number of the model was calculated. Then, we analyzed the existence and stability of the rumor equilibrium point. Next, based on the principle of Pontryagin's maximum value, we obtained a control strategy, which effectively reduced the spread of rumors. Numerical simulations verified the results of theoretical analysis. The results showed that the higher the education level of the population, the slower the spread of rumors to a certain extent, but it could not prevent the spread of rumors. In addition, through the support of the government and the propaganda of the official media, strengthening education can improve people's education level to a certain extent, and then minimize the speed of rumor propagation.

    Citation: Hongshuang Wang, Sida Kang, Yuhan Hu. Dynamic analysis and optimal control of rumor propagation models considering different education levels and hesitation mechanisms[J]. AIMS Mathematics, 2024, 9(8): 20089-20117. doi: 10.3934/math.2024979

    Related Papers:

    [1] Xintong Wang, Sida Kang, Yuhan Hu . Dynamic analysis of a rumor propagation model considering individual identification ability. AIMS Mathematics, 2025, 10(2): 2295-2320. doi: 10.3934/math.2025107
    [2] Ying Yu, Jiaomin Liu, Jiadong Ren, Qian Wang, Cuiyi Xiao . Minimize the impact of rumors by optimizing the control of comments on the complex network. AIMS Mathematics, 2021, 6(6): 6140-6159. doi: 10.3934/math.2021360
    [3] Liuqin Huang, Jinling Wang, Jiarong Li, Tianlong Ma . Analysis of rumor spreading with different usage ranges in a multilingual environment. AIMS Mathematics, 2024, 9(9): 24018-24038. doi: 10.3934/math.20241168
    [4] Xuefeng Yue, Weiwei Zhu . The dynamics and control of an ISCRM fractional-order rumor propagation model containing media reports. AIMS Mathematics, 2024, 9(4): 9721-9745. doi: 10.3934/math.2024476
    [5] Rongrong Yin, Ahmadjan Muhammadhaji . Dynamics in a delayed rumor propagation model with logistic growth and saturation incidence. AIMS Mathematics, 2024, 9(2): 4962-4989. doi: 10.3934/math.2024241
    [6] Hui Wang, Shuzhen Yu, Haijun Jiang . Rumor model on social networks contemplating self-awareness and saturated transmission rate. AIMS Mathematics, 2024, 9(9): 25513-25531. doi: 10.3934/math.20241246
    [7] Yanchao Liu, Pengzhou Zhang, Deyu Li, Junpeng Gong . Dynamic analysis and optimum control of a rumor spreading model with multivariate gatekeepers. AIMS Mathematics, 2024, 9(11): 31658-31678. doi: 10.3934/math.20241522
    [8] Youming Guo, Tingting Li . Dynamics and optimal control of an online game addiction model with considering family education. AIMS Mathematics, 2022, 7(3): 3745-3770. doi: 10.3934/math.2022208
    [9] Sayed Saber, Azza M. Alghamdi, Ghada A. Ahmed, Khulud M. Alshehri . Mathematical Modelling and optimal control of pneumonia disease in sheep and goats in Al-Baha region with cost-effective strategies. AIMS Mathematics, 2022, 7(7): 12011-12049. doi: 10.3934/math.2022669
    [10] Xuechao Zhang, Yuhan Hu, Shichang Lu, Haomiao Guo, Xiaoyan Wei, Jun li . Dynamic analysis and control of online information dissemination model considering information beneficiaries. AIMS Mathematics, 2025, 10(3): 4992-5020. doi: 10.3934/math.2025229
  • The spread of rumors has an important impact on the production and life of human society. Moreover, in the process of rumor propagation, individuals with different educational levels show different degrees of trust and ability to spread rumors. Therefore, a new rumor propagation model was established, which considers the influence of education level on rumor propagation. Initially, the basic reproduction number of the model was calculated. Then, we analyzed the existence and stability of the rumor equilibrium point. Next, based on the principle of Pontryagin's maximum value, we obtained a control strategy, which effectively reduced the spread of rumors. Numerical simulations verified the results of theoretical analysis. The results showed that the higher the education level of the population, the slower the spread of rumors to a certain extent, but it could not prevent the spread of rumors. In addition, through the support of the government and the propaganda of the official media, strengthening education can improve people's education level to a certain extent, and then minimize the speed of rumor propagation.



    Rumors are unconfirmed information that is spread through public or private channels [1]. Since ancient times, rumors have had a significant impact on the development of the country and society. Now, with the development of society, rumors can spread rapidly through platforms such as WeChat, QQ and Microblogging. This may lead to economic losses and psychological panic. Therefore, in today's society, it is of great practical significance to study the spread of rumors.

    Due to the resemblance between the rumor propagation and infectious disease propagation models, we frequently apply the infectious disease propagation principle when examining the former. Initially, in 1965, Daley and Kendall assigned the rumor propagation predicament to a mathematical model, which they named the D-K model [2]. Subsequently, in 1973, Maki and Thomson refined the propagation law using the D-K model and introduced the M-T model [3].

    After people's in-depth exploration and research on rumor propagation, many researchers have incorporated the forgetting mechanism [4,5,6], hesitation mechanism [7,8] and counterattack mechanism [9], which is considered by early psychologists into the establishment of the rumor propagation model. Gu et al. [4] proposed a method to study the transmission process by considering factors of forgetting mechanism and memory mechanism. They used basic mathematical functions to characterize the evolution of forgetting and memory over time, and found that the saturation of rumor propagation is influenced by the "forgetting and remembering" mechanism, and may even lead to the end of rumor propagation under certain conditions. Zhao et al. [5] introduced the memory mechanism and forgetting mechanism into a homogeneous network and established a SHIR model to study the effects of propagation rate, forgetting rate and network average degree on rumor propagation. They found that the forgetting and memory mechanisms that occur in hibernators delay the end of the rumor and somewhat reduce the maximum rumor impact. Wang et al. [6] proposed a new SIR model, which mainly studied the influence of forget-recall mechanism and loss-benefit mechanism on rumor propagation in complex networks, considering that people would forget and lose interest in rumors during the propagation process. Xia et al. [7] added a new factor of hesitation mechanism to the classical SIR model and proposed a SEIR model. Hong et al. [8] added the factor of the actual situation of the investor network to the SEIR model, and then proposed a SE2IR rumor propagation model with hesitation mechanism. Zan et al. [9] added a counterattack mechanism to the classical SIR model, and then proposed two new models: SCIR model and the upgraded SCIR model.

    Currently, researchers primarily focus on various aspects including cost control [10], feedback mechanisms [11], multi-channel network interaction [12], and competitive information propagation [13]. Jain et al. [10] studied the optimal control of rumor propagation in uniformly mixed populations by introducing the thinker's influence delay. They established an SEI model. It has been concluded that the delay in influence of thinkers ought not to surpass the threshold for controlling the instability of social network systems. The system can be stabilized when the thinker has more time to react, and when experts in the media and interpersonal organizations disseminate information at the optimal speed. The optimal control system is capable of effectively suppressing the propagation of anti-news during emergency situations, at a minimized cost. Yao et al. [11] employed an epidemiological theory-based diffusion model to study the propagation of rumors amongst investors. The model incorporated feedback mechanisms and time-lag factors. The research shows that increasing the intensity of information supervision and the proportion coefficient related to the infected population in the short time lag is helpful to control the spread of rumors more effectively. Zhang et al. [12] proposed a new interaction model for rumor propagation and behavioural propagation in multiplexed networks-the S1I1R1S2I2R2 model. The results of their analysis and calculations indicated that the final size of the rumor spread was larger than the size of the behavioural spread. Nevertheless, coupled intensification results in the opposite range of final size changes. This shows that there is a complex interaction between rumor and behavior propagation in the multiplex network. Yang et al. [13] proposed a simple linear threshold model with competing generalised versions that takes into account the competition between rumor and truth in the same network. They employ three distinct heuristics to arrive at our conclusions: Both the diffusion dynamics-based approach (ContrId) and the centrality based approach (PageRank) possess the same computational complexity, with said complexity increasing linearly with the number of nodes n, rendering them scalable to large networks. Upon introducing the proximity effect, ContrId performs equally to MinGreedy, but at a significantly faster pace. These findings offer significant insights into the investigation of the rivalry between the propagation of rumors and the propagation of truth.

    When discussing the influencing factors of rumor propagation, in addition to the factors discussed previously, some scholars such as Komi Afassinou [14] introduced the two key variables of population education rate and forgetting mechanism based on the classical SIR model. They demonstrated that the greater the proportion of educated individuals in a group, the smaller the eventual scale of rumor propagation. This highlights the role of education in the process of rumor propagation. Hu et al. [15] examined the impact of alterations in the proportion of wise individuals in a population on rumor propagation. They demonstrated that an increase in the proportion of wise individuals in a population has a systematic effect in resolving the issue of rumor propagation. Furthermore, it is important to consider the role of social media in rumor propagation. Furthermore, Li et al. [16] developed an enhanced rumor propagation model, with a particular focus on the impact of knowledge education and intervention strategies on the reduction of rumor propagation. Their results show that strengthening rumor recognition education and timely refutation of false information are very effective in controlling rumor propagation. In the above studies, Afassinou and Li et al. both involved the influence of education level on rumor propagation, but their classification of education level was relatively simple, only divided into the educated and the uneducated. Furthermore, the Hu et al. study only looked at the influence of the proportion of wise men on rumor propagation and did not classify the population. Therefore, based on the existing literature, we provide a more comprehensive analysis of the mechanism of rumor propagation. This study comprehensively considers several factors such as education level, hesitation mechanism and forgetting mechanism, and divides education level into three categories, namely less educated, more educated and uneducated categories. This comprehensive classification enables a more nuanced understanding of the role of groups with varying educational backgrounds in the process of rumor propagation. This, in turn, facilitates a more comprehensive understanding of the mechanisms underlying rumor propagation.

    The rest of this paper is organized as follows: In Section 2, a rumor propagation model is constructed. In Section 3, the basic reproduction number and the rumor equilibrium point of the model are calculated, and the stability of the rumor equilibrium point is analyzed. In Section 4, the optimal control strategy is proposed using the Pontryagin's maximum principle. In Section 5, the feasibility of the conditions proposed above is verified by numerical simulation. In Section 6, we draw the conclusion of the paper.

    We discuss the influence of individuals with different levels of education on rumor propagation. First of all, we divide the total population into three categories according to the level of education: the first category is the highly educated individuals; the second category is the less educated individuals; the third category is the uneducated individual. Then, according to the classical rumor propagation model, we use N(t) to represent the change in the total number of people at any time t. Then, using the method of the chamber model, we divided the population into five categories: (1) Rumor unknowers (people who have never been exposed to rumors); (2) rumor hesitators with less education level (the less educated rumor unknowers contact the rumor, but do not directly believe the rumor, but think about the content of the spread); (3) rumor hesitators with higher education (the more educated rumor unknowers contact the rumors, but do not believe them straight away, but think about them); (4) rumor spreaders (people who believe and spread rumors); and (5) rumor immunizers (people who will not believe and will not spread rumors). The above five categories of people are respectively represented by S(t),D(t),E(t),I(t),R(t).

    In the society, the flow of people is a universal phenomenon. Therefore, the model established in this paper is based on the open system, and we make the following assumptions:

    (1) Assume that individuals enter the rumor propagation system with a constant number of Λ. Then, consider that the population of each chamber has moved out of the system for specific reasons such as relocation, natural death, etc. Therefore, in order to facilitate the study, we assume that these five types of chambers have the same removal rate, which is defined as μ.

    (2) When rumor unknowers with less education come into contact with rumors spreaders, the contact rate between them is defined as α1; the proportion of rumor unknowers who received a less level of education in the total number of rumor unknowers was defined as β1; moreover, when rumor unknowers with less education are contacted by rumors spreaders, generally speaking, these people are more likely to hesitate about the information spread by rumors spreaders than directly deny it. Therefore, we think that these people will transform into less educated rumor hesitators with a probability of α1(1+β1).

    (3) When rumor unknowers with higher education come into contact with rumors spreaders, the contact rate between them is defined as α2; the proportion of rumor unknowers who have received higher education in the total number of rumor unknowers is defined as β2; furthermore, when rumor unknowers with higher education are contacted by rumor spreaders, generally speaking, these people tend to directly deny the rumor rather than hesitate to the information spread by rumor spreaders. Therefore, we believe that these people will transform into highly educated rumor hesitators with a probability of α2(1β2).

    (4) When uneducated rumor unknowers come into contact with rumor spreaders, generally speaking, these rumor unknowers will choose to directly believe the information spread by the rumor spreaders without their own thinking. We define the probability of these people turning into rumor spreaders as λ.

    (5) In addition, rumor hesitators with less education and rumor hesitators with higher education will be removed from the system as mentioned above, and the following two situations may occur. In the first case, after thinking about it, they feel that the information they received before is correct and choose to continue believing and spreading it. Then, we define the probabilities of them turning into rumor spreaders as γ1 and γ2 respectively; in the second case, after thinking about it, they feel that the information they received before is wrong and choose not to continue believing and spreading it. Thus, we define the probabilities of them turning into rumor immunizers as ϕ1 and ϕ3, respectively. Furthermore, because of the rumor spreading process, the timeliness of the rumor spreaders may lose interest in the rumor spreading process, and people have the characteristic of forgetting and the awakening of people's own consciousness (people find out that the information they believe is a rumor and then automatically become rumor immunizers) and other reasons, so that rumor spreaders will spontaneously become rumor immunizers. Let us define the probability of its transformation as ϕ2.

    The parameters of model SDEIR are summarized in Table 1.

    Table 1.  Description of parameters in the model.
    S(t) the number of rumors unknowers at time t
    D(t) the number of rumor hesitators with less education at time t
    E(t) the number of rumor hesitators with higher educated at time t
    I(t) the number of rumor spreaders at time t
    R(t) the number of rumor immunizers at time t
    Λ the number of people who enter the rumor unknowers per unit time
    μ the probability of moving out of the rumor propagation system
    α1 the natural contact rate between less educated rumor unknowers and the rumor spreaders
    β1 the proportion of less educated rumor unknowers in the total number of rumor unknowers
    α1(1+β1) the rate that S(t) transforms into D(t)
    α2 the natural contact rate between higher educated rumor unknowers and rumor spreaders
    β2 the proportion of higher educated rumor unknowers in the total number of rumor unknowers
    α2(1β2) the rate that S(t) transforms into E(t)
    λ the probability of uneducated rumor unknowers transforming into rumor spreaders
    γ1 the probability that D(t) transforms into I(t)
    γ2 the probability that E(t) transforms into I(t)
    ϕ1 the probability that D(t) transforms into R(t)
    ϕ2 the probability that I(t) transforms into R(t)
    ϕ3 the probability that E(t) transforms into R(t)

     | Show Table
    DownLoad: CSV

    On this basis, we drew the model flow chart, as shown in Figure 1, and constructed the dynamic equation, such as Eq (2.1).

    Figure 1.  Consider the rumor propagation diagram of the education level.
    {dS(t)dt=Λα1(1+β1)SIλSIα2(1β2)SIμSdD(t)dt=α1(1+β1)SIγ1Dϕ1DμDdE(t)dt=α2(1β2)SIγ2Eϕ3EμEdI(t)dt=λSI+γ1D+γ2Eϕ2IμIdR(t)dt=ϕ1D+ϕ2I+ϕ3EμR (2.1)

    where,

    Λ,α1,α2,β1,β2,γ1,γ2,ϕ1,ϕ2,ϕ3,μ>0;S(t)+D(t)+E(t)+I(t)+R(t)=N(t).

    Lemma 3.1. The closed set Γ={(S,D,E,I,R)R5+|0S+D+E+I+RΛμ} is the positive definite invariant set of the model dynamics equation (2.1).

    Proof of Lemma 3.1. According to the model dynamics equation, we can easily know dNdt=ΛμN. Thus, we can derive N(t)=Λμ(ΛμN0)eμt, where N(0)=N0 and when t, limtN(t)=Λμ. This proves that the positive definite invariant set of the model dynamics equation (2.1) is Γ={(S,D,E,I,R)R5+|0S+D+E+I+RΛμ}.

    The basic reproduction number R0 represents the number of people that a rumor spreader can infect during the average period of rumor propagation when all people are rumor unknowers [17]. In general, R0=1 is used as a threshold to determine whether a rumor is dead or not. In other words, when R0<1, the number of people a rumor spreader can infect during the average rumor propagation period is less than 1, which indicates that rumors will disappear in the future with time. When R0>1, a rumor spreader can infect more than 1 person during the average rumor propagation period, indicating that rumors do not disappear over time, but will always exist. This paper uses the method in previous literature [18], that is, the next generation matrix method, to calculate the basic reproduction number R0 of the model dynamic equation (2.1).

    Let X=(I,D,E,R,S)T, then the model dynamics equation (2.1) can be written as

    dN(t)dt=F(X)V(X), (3.1)

    where,

    F(X)=[λSI+γ1D+γ2Eα1(1+β1)SIα2(1β2)SI00],V(X)=[(ϕ2+μ)I(ϕ1+γ1+μ)D(ϕ3+γ2+μ)DμRϕ1D+ϕ2I+ϕ3Eα1(1+β1)SI+λSI+α2(1β2)SI+μSΛ]. (3.2)

    Thus, we can get

    f=[λΛμγ1γ2α1(1+β1)Λμ00α2(1β2)Λμ00],v=[μ+ϕ2000ϕ1+γ1+μ000ϕ3+γ2+μ]. (3.3)

    And by calculation, we can get

    v1=[1μ+ϕ20001ϕ1+γ1+μ0001ϕ3+γ2+μ]. (3.4)

    Then, fv1 is used to represent the next generation matrix

    fv1=[Λλμ(μ+ϕ2)γ1μ+γ1+ϕ1γ2μ+γ2+ϕ3α1(1+β1)Λμ(μ+ϕ2)00α2(1β2)Λμ(μ+ϕ2)00]. (3.5)

    Therefore, according to reference [19], the basic reproduction number of model dynamic equation (2.1) is the spectral radius of matrix :

    R0=λΛcd+(λΛcd)2+4μecdΛ(bcγ2+adγ1)2μecd, (3.6)

    where,

    {a=α1(1+β1)b=α2(1β2)c=ϕ1+γ1+μd=ϕ3+γ2+μe=ϕ2+μ. (3.7)

    According to the model dynamics equation (2.1), we can find two equilibrium points, which are:

    E0=(Λμ,0,0,0,0)

    and

    E=(S,D,E,I,R)
    S=ecdadγ1+bcγ2+cdλ;I=Λ(λcd+adγ1+bcγ2)μecdecd(a+d+λ)
    D=aΛ(λcd+adγ1+bcγ2)aμecdc(λcd+adγ1+bcγ2)(a+d+λ);E=bΛ(λcd+adγ1+bcγ2)bμecdd(λcd+adγ1+bcγ2)(a+d+λ)
    R=(Λadγ1+Λbcγ2+Λcdλcdeμ)(cdλϕ2+adeϕ1+bceϕ3+adγ1ϕ2+bcγ2ϕ2)μecd(λcd+adγ1+bcγ2)(a+b+λ)

    where a,b,c,d,e satisfies Eq (3.7).

    Theorem 1. If R0<1,Λμ<ecdλcd+γ1ad+γ2bc,c=d and γ1a<λc holds true, then the rumor-free equilibrium point E0=(Λμ,0,0,0,0) is locally asymptotically stable; If Λ[α1(1+β1)+α2(1β2)+λ]μ2 holds true, then the rumor-free equilibrium point E0=(Λμ,0,0,0,0) is globally asymptotically stable. Where a,b,c,d,e satisfies the above Eq (3.7).

    Proof. The Jacobian matrix of system (2.1) at E0=(Λμ,0,0,0,0) is

    J(E0)=|μ00aΛμλΛμbΛμ00c0aΛμ000dbΛμ00γ1γ2λΛμe00ϕ1ϕ3ϕ2μ|. (3.8)

    The characteristic equation of matrix J(E0) is

    |θEJ(E0)|=|θ+μ00aΛμ+λΛμ+bΛμ00θ+c0aΛμ000θ+dbΛμ00γ1γ2θλΛμ+e00ϕ1ϕ3ϕ2θ+μ|=(θ+μ)2(θ3+Aθ2+Bθ+C)=0 (3.9)

    where,

    {A=e+c+dλΛμB=ec+ed+cd(λc+γ1a+λd+γ2b)ΛμC=ecd(λcd+γ1ad+γ2bc)Λμ. (3.10)

    This gives us the eigenvalues θ1=θ2=μ and the simplified characteristic equation

    θ3+Aθ2+Bθ+C=0. (3.11)

    We then construct a cubic polynomial using a0,a1,a2,a3 instead of the coefficients to find the other eigenvalues of Eq (3.11). We can then rewrite Eq (3.11) as

    a0θ3+a1θ2+a2θ+a3=0 (3.12)

    where,

    a0=1
    a1=A=e+c+dλΛμ
    a2=B=ec+ed+cd(λc+γ1a+λd+γ2b)Λμ
    a3=C=ecd(λcd+γ1ad+γ2bc)Λμ

    and

    a2a1a3a0=(λ2c+λγ1a+λγ2b+λ2d)(Λμ)2(2λcd+2λec+2λed+λc2+λd2+γ1ae+γ1ac+γ2be+γ2bd)Λμ+(2ecd+e2c+ec2+e2d+ed2+c2d+cd2). (3.13)

    The conditions for local asymptotic stability of the rumor-free equilibrium point E0=(Λμ,0,0,0,0) based on the Routh-Hurwitz criterion [20] are as follows: (1) an>0,n=0,1,2,3 (2) a2a1a3a0>0. Through calculation, it can be concluded that when the condition of Λμ<ecdλcd+γ1ad+γ2bc,c=d and γ1a<λc is met, the rumor-free equilibrium point E0=(Λμ,0,0,0,0) is locally asymptotically stable.

    The above is the proof of the local asymptotic stability of the rumor-free equilibrium point E0=(Λμ,0,0,0,0), followed by the proof of the global asymptotic stability.

    We construct a Lyapunov function [21] L(t)=D(t)+E(t)+I(t)+R(t). Then, we can get L(t),

    L(t)=D(t)+E(t)+I(t)+R(t)=α1(1+β1)SIγ1Dϕ1DμD+α2(1β2)SIγ2Eϕ3EμE+λSI+γ1D+γ2Eϕ2IμI+ϕ1D+ϕ2I+ϕ3EμR=α1(1+β1)SI+α2(1β2)SI+λSIμIμDμEμR={[α1(1+β1)+α2(1β2)+λ]Sμ}Iμ(D+E+R). (3.14)

    We know SΛμ from Lemma 1. Obviously, if Λ[α1(1+β1)+α2(1β2)+λ]μ2 and SΛμ, then L(t)0 is always true.

    Furthermore, L(t)=0 is true if and only if S(t)=S0,D(t)=E(t)=I(t)=R(t)=0. From system (2.1) we know that when L(t)=0, E0 is the only solution in Γ. Therefore, based on Lyapunov function and LaSalle invariance principle [22], it can be shown that every solution of the system (2.1) tends to E0 infinitely for t. Therefore, the rumor-free existence equilibrium point E0=(Λμ,0,0,0,0) of system (2.1) is globally asymptotically stable. □

    Theorem 2. If R0>1, P3P2P4P1>0 and P3P2P1P4P21P23>0, then the rumor existence equilibrium point E=(S,D,E,I,R) is locally asymptotically stable; If R0>1, then the rumor existence equilibrium point E=(S,D,E,I,R) is globally asymptotically stable. Where a,b,c,d,e satisfies the above Eq (3.7).

    Proof. The Jacobian matrix of system (2.1) at E=(S,D,E,I,R) is

    J(E)=[aIλIbIμ00aSλSbS0aIc0aS0bI0dbS0λIγ1γ2λSe00ϕ1ϕ3ϕ2μ]. (3.15)

    The characteristic equation of matrix J(E) is

    |θEJ(E)|=|θ+aI+λI+bI+μ00aS+λS+bS0aIθ+c0aS0bI0θ+dbS0λIγ1γ2θλS+e00ϕ1ϕ3ϕ2θ+μ|=(θ+μ)(θ4+P1θ3+P2θ2+P3θ+P4)=0, (3.16)

    where,

    {P1=(e+c+d)λecdλcd+γ1ad+γ2bc+Λ(λcd+γ1ad+γ2bc)ecdP2=(ec+ed+cd)(λc+γ1a+γ2b+λμ+λd)ecdλcd+γ1ad+γ2bc+Λ(e+c+d)(λcd+γ1ad+γ2bc)ecdP3=Λ(ec+cd+ed)(λcd+γ1ad+γ2bc)ecd(λc+γ1a+γ2b+λd)μecdλcd+γ1ad+γ2bcP4=Λ(λcd+γ1ad+γ2bc)μecd.

    From this, the eigenvalues θ1=μ<0 and the simplified characteristic equation θ4+P1θ3+P2θ2+P3θ+P4=0 can be obtained.

    According to the simplified characteristic equation above, with the help of Routh-Hurwitz criterion, it can be seen that all the coefficients of the characteristic equation are positive numbers, and the values of each coefficient in the first column of table Routh-Hurwitz are also greater than 0, that is, the conditions of local asymptotic stability are satisfied [23]. Therefore, we can judge that the condition of local stability of the model is R0>1, P3P2P4P1>0 and P3P2P1P4P21P23>0 when the equilibrium point E=(S,D,E,I,R) exists.

    The above is the proof of the local asymptotic stability of the rumor equilibrium point E=(S,D,E,I,R), followed by the proof of the global asymptotic stability.

    First, we construct a Lyapunov function W(t)=[(S(t)S)+(D(t)D)+(E(t)E)+(I(t)I)+(R(t)R)]2. Then, you get W(t),

    W(t)=2[(S(t)S)+(D(t)D)+(E(t)E)+(I(t)I)+(R(t)R)][S(t)+D(t)+E(t)+I(t)+R(t)]=2[(S(t)S)+(D(t)D)+(E(t)E)+(I(t)I)+(R(t)R)](ΛμSμDμEμIμR). (3.17)

    From the above discussion, it can be seen that the equilibrium point E=(S,D,E,I,R) of rumors exists. Therefore, ΛμSμDμEμIμR=0. In other words, Λ=μS+μD+μE+μI+μR. So, Eq (3.17) can be calculated as

    W(t)=2[(S(t)S)+(D(t)D)+(E(t)E)+(I(t)I)+(R(t)R)][μS+μD+μE+μI+μRμSμDμEμIμR]=2μ[(SS)+(DD)+(EE)+(II)+(RR)]20. (3.18)

    Furthermore, W(t)=0 is true if and only if S(t)=S,D(t)=D,E(t)=E,I(t)=I,R(t)=R. Therefore, based on Lyapunov function and LaSalle invariance principle, it can be obtained that the rumor existence equilibrium point E=(S,D,E,I,R) of system (2.1) is globally asymptotically stable. □

    The model developed in this study is predicated on the social environment with the aim of preventing the propagation of misinformation. Optimal control is a well-known concept involving the purposeful exertion of control over systems and processes in a variety of contexts, such as production and life operations, to achieve specific performance targets [24]. Given the alignment between the objectives of this study and the optimal control theory, the model is analyzed in this section using established methods from existing literature [25].

    First, it is assumed that the model discussed in this paper is operating in a social communication environment that is closed.

    Second, our purpose of this paper is to reduce the number of rumor hesitators and rumor spreaders, while increasing the number of rumor immunizers. To achieve this goal, we controlled for the following five parameters: The natural contact rate α1 between rumor unknowers with higher education and rumor spreaders, and the natural contact rate α2 between rumor unknowers with less education and rumor spreaders. The proportion β1 of rumor unknowers with less education level and the proportion β2 of rumor unknowers with higher education level, as well as the probability λ of transforming uneducated rumor unknowers into rumor spreaders.

    Next, we carry out the following analysis:

    (1) We convert the five parameters α1,α2,β1,β2,λ in the model into control variables α1(t),α2(t),β1(t),β2(t) and λ(t).

    (2) In this model, control variables satisfy 0α1(t),α2(t),β1(t),β2(t),λ(t)1. When α1(t)=0,α2(t)=0,β1(t)=0,β2(t)=1,λ(t)=0, we can conclude that rumor unknowers did not transform into rumor spreaders and rumor hesitators, but directly transformed into rumor immunizers people. At this point, the desired result of the control has been achieved, which means that the control measure has been a hundred percent effective; When α1(t)=1,α2(t)=1,β1(t)=1,β2(t)=0,λ(t)=1, we can conclude that the rumor unknowers all transform into rumor hesitators, and then all transform into rumor spreaders, which means that the control measures are completely ineffective.

    From the above statement, we set up an objective function, defined as follows:

    J(α1,α2,β1,β2,λ)=tf0eαt[D(t)+E(t)+I(t)+c12α21(t)+c22α22(t)+c32β21(t)+c42β22(t)+c52λ2(t)] (4.1)

    and satisfy the following state system:

    {dS(t)dt=Λα1(t)(1+β1(t))SIλ(t)SIα2(t)(1β2(t))SIμSdD(t)dt=α1(t)(1+β1(t))SIγ1Dϕ1DμDdE(t)dt=α2(t)(1β2(t))SIγ2Eϕ3EμEdI(t)dt=λ(t)SI+γ1D+γ2Eϕ2IμIdR(t)dt=ϕ1D+ϕ2I+ϕ3EμR. (4.2)

    The initial conditions for satisfying formula (4.2) are as follows:

    S(0)=S00,D(0)=D00,E(0)=E00,I(0)=I00,R(0)=R00 (4.3)

    where,

    α1(t),α2(t),β1(t),β2(t),λ(t)U[α1,α2,β1,β2,λ|(α1(t),α2(t),β1(t),β2(t),λ(t)),0α1(t),α2(t),β1(t),β2(t),λ(t)1,t[0,tf]], (4.4)

    and U is the admissible control set. The control time interval is between 0 and tf. c1,c2,c3,c4,c5 is a positive weight coefficient indicating the prevention intensity and importance of the five control measures.

    In this paper, we demonstrate the existence of an optimal control policy in the above optimal control problem by applying the following theorem.

    Theorem 4.1. There exists an optimal control for (α1,α2,β1,β2,λ)U such that the function is established as follows [26]:

    J(α1,α2,β1,β2,λ)=min{J(α1,α2,β1,β2,λ),α1(t),α2(t),β1(t),β2(t),λ(t)U} (4.5)

    Proof. In order to prove the existence of optimal control, it is only necessary to verify the following conditions [27,28].

    (1) The control admissible set and the set of state variables are non-empty.

    (2) The control allows set U to be convex and closed.

    (3) The integrand in the objective functional is convex in the control admissible set U.

    (4) The right end of the state system is linear with respect to state variables and control variables.

    (5) The existence of constants d1,d2>0 and ε>1 makes the integrand expression in the target functional.

    Let X=(S(t),D(t),E(t),I(t),R(t))T and

    L(t,X(t),α1(t),α2(t),β1(t),β2(t),λ(t))eαt[D(t)+E(t)+I(t)+c12α21(t)+c22α22(t)+c32β21(t)+c42β22(t)+c52λ2(t))] (4.6)

    satisfy

    L(t,α1,α2,β1,β2,λ)d1(|α1|2+|α2|2+|β1|2+|β2|2+|λ|2)ε2d2. (4.7)

    Conditions (1)–(3) are obviously true, only the last two need to be verified. From Lemma 1

    max{S(t),D(t),E(t),I(t),R(t)}Λμ. (4.8)

    The following inequality can be obtained:

    SΛ,Dα1(1+β1)SI,Eα2(1β2)SI,Iγ1D+λSI+γ2E,Rϕ1D+ϕ2E+ϕ3R. (4.9)

    Thus, the fourth condition holds. If we take d1=eαt02min{c1,c2,c3,c4,c5} and d2R+, and ε=2, then the fifth condition holds. □

    According to the optimal control expression given in 4.1, the optimal control system is obtained. To do this, define the following enlarge Hamiltonian operator L with a penalty term [29]:

    L=eαt[D(t)+E(t)+I(t)+c12α21(t)+c22α22(t)+c32β21(t)+c42β22(t)+c52λ2(t))]+η1(t)[Λα1(t)(1+β1(t))SIλ(t)SIα2(t)(1β2(t))SIμS]+η2(t)[α1(t)(1+β1(t))SIγ1Dϕ1DμD]+η3(t)[α2(t)(1β2(t))SIγ2Eϕ3EμE]+η4(t)[λ(t)SI+γ1D+γ2Eϕ2IμI]+η5(t)[ϕ1D+ϕ2I+ϕ3EμR]ω11(t)α1(t)ω12(t)(1α1(t))ω21(t)α2(t)ω22(t)(1α2(t))ω31(t)β1(t)ω32(t)(1β1(t))ω41(t)β2(t)ω42(t)(1β2(t))ω51(t)λ(t)ω52(t)(1λ(t)), (4.10)

    where ωij0(i,j=1,2,3,4,5) is the penalty operator, it satisfies ω11(t)α1(t)=ω12(1α1(t))=0, at optimal control α1;

    ω21(t)α2(t)=ω22(1α2(t))=0, at optimal control α2;

    ω31(t)β1(t)=ω32(1β1(t))=0, at optimal control β1;

    ω41(t)β2(t)=ω42(1β2(t))=0, at optimal control β2;

    ω51(t)λ(t)=ω52(1λ(t))=0, at optimal control λ.

    Theorem 4.2. Given the optimal control pair (α1(t),α2(t),β1(t),β2(t),λ(t)) and solution S(t),D(t),E(t),I(t),R(t) of Eq (2.1) of state system, there is a costate variable ηi(t),i=1,2,3,4,5, which satisfies the equation:

    {η1=η1(t)[α1(t)(1+β1(t))I+λ(t)I+α2(t)(1β2(t))I+μ]η2(t)α1(t)(1+β1(t))Iη3(t)α2(t)(1β2(t))Iη4(t)λ(t)Iη2=eαt+η2(t)(ϕ1+γ1+μ)η4(t)γ1η5(t)ϕ1η3=eαt+η3(t)(ϕ3+γ2+μ)η4(t)γ2η5(t)ϕ3η4=eαtη1(t)[α1(t)(1+β1(t))S+λ(t)S+α2(t)(1β2(t))S]η2(t)α1(t)(1+β1(t))Sη3(t)α2(t)(1β2(t))Sη4(t)(λ(t)Sμϕ2)η5(t)ϕ2η5=η5(t)μ (4.11)

    and terminal conditions

    ηi(tf)=0,i=1,2,3,4,5. (4.12)

    Moreover, the expression for optimal control (α1(t),α2(t),β1(t),β2(t),λ(t)) is expressed as follows:

    {α1(t)=min(1,max(0,η1(t)(1+β1(t))S(t)I(t)η2(t)(1+β1(t))S(t)I(t))eαtc1))α2(t)=min(1,max(0,η1(t)(1β2(t))S(t)I(t)η3(t)(1β2(t))S(t)I(t))eαtc2))β1(t)=min(1,max(0,η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))eαtc3))β2(t)=min(1,max(0,η1(t)α2(t)S(t)I(t)+η3(t)α2(t)S(t)I(t))eαtc4))λ(t)=min(1,max(0,η1(t)S(t)I(t)η4(t)S(t)I(t)eαtc5)) (4.13)

    Proof. Based on Pontriagin's maximum principle [30], partial derivatives of the Hamiltonian operator L with respect to each state variable give the following costate system η1=LS,η2=LD,η3=LE,η4=LI,η5=LR, and the final value condition ηi(tf)=0,i=1,2,3,4,5.

    In order to obtain the necessary conditions for optimality, the partial derivative of the operator L with respect to the control variable α1(t),α2(t),β1(t),β2(t),λ(t) is obtained separately. Subsequently, the value should be set to zero and get

    {Lα1=eαtc1α1(t)η1(t)(1+β1(t))SI+η2(t)(1+β1(t))SIω11(t)+ω12(t)=0Lα2=eαtc2α2(t)η1(t)(1β2(t))SI+η3(t)(1β2(t))SIω21(t)+ω22(t)=0Lβ1=eαtc3β1(t)η1(t)α1(t)SI+η2(t)α1(t)SIω31(t)+ω32(t)=0Lβ2=eαtc4β2(t)+η1(t)α2(t)SIη3(t)α2(t)SIω41(t)+ω42(t)=0Lλ=eαtc5λ1(t)η1(t)SI+η4(t)SIω51(t)+ω52(t)=0. (4.14)

    To solve it, the expression of optimal control can be obtained in the following form:

    {α1(t)=η1(t)(1+β1(t))S(t)I(t)η2(t)(1+β1(t))S(t)I(t))+ω11(t)ω12(t)eαtc1α2(t)=η1(t)(1β2(t))S(t)I(t)η3(t)(1β2(t))S(t)I(t))+ω21(t)ω22(t)eαtc2β1(t)=η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))+ω31(t)ω32(t)eαtc3β2(t)=η1(t)α2(t)S(t)I(t)+η3(t)α2(t)S(t)I(t))+ω41(t)ω42(t)eαtc4λ(t)=η1(t)S(t)I(t)η4(t)S(t)I(t)+ω51(t)ω52(t)eαtc5. (4.15)

    In order to obtain the concrete expression of optimal control ηi(t)(i=1,2,3,4,5) without penalty term ωij(i=1,2,3,4,5;j=1,2), the standard practice of reference [31] is adopted. Taking β1(t) as an example, the specific process is divided into three situations for implementation:

    (1) In case 1, when {t|0<β1(t)<1} is satisfied, let ω31(t)=ω32(t)=0.

    Therefore, the optimal control is β1(t)=η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))eαtc3.

    (2) In case 2, when {t|β1(t)=1} is satisfied, let ω31(t)=0.

    Therefore, 1=β1(t)=η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))ω32(t)eαtc3.

    This means that when ω32(t)0, η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))eαtc31.

    (3) In case 3, when {t|β1(t)=0} is satisfied, let ω32(t)=0.

    Therefore, 0=β1(t)=η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))+ω31(t)eαtc3.

    This means that when ω32(t)0, η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))eαtc30.

    Combining the above three cases, the expression of optimal control β1(t) can be obtained: β1(t)=min(1,max(0,η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))eαtc3)).

    Similar to the discussion, the corresponding expressions for optimal control α1(t),α2(t),β1(t),β2(t) and λ(t) can be obtained:

    {α1(t)=min(1,max(0,η1(t)(1+β1(t))S(t)I(t)η2(t)(1+β1(t))S(t)I(t))eαtc1))α2(t)=min(1,max(0,η1(t)(1β2(t))S(t)I(t)η3(t)(1β2(t))S(t)I(t))eαtc2))β1(t)=min(1,max(0,η1(t)α1(t)S(t)I(t)η2(t)α1(t)S(t)I(t))eαtc3))β2(t)=min(1,max(0,η1(t)α2(t)S(t)I(t)+η3(t)α2(t)S(t)I(t))eαtc4))λ(t)=min(1,max(0,η1(t)S(t)I(t)η4(t)S(t)I(t)eαtc5)).

    It is worth noting that the optimal control system comprises the state system (4.1) and its initial condition (4.3), the costate system (4.11) and its terminal condition (4.12), and the optimal control form expression (4.13). Every optimal control must fulfill this complex optimal control system structure. The following is the optimal control system for future simulations:

    {dS(t)dt=Λα1(t)(1+β1(t))SIλ(t)SIα2(t)(1β2(t))SIμS,dD(t)dt=α1(t)(1+β1(t))SIγ1Dϕ1DμD,dE(t)dt=α2(t)(1β2(t))SIγ2Eϕ3EμE,dI(t)dt=λ(t)SI+γ1D+γ2Eϕ2IμI,dR(t)dt=ϕ1D+ϕ2I+ϕ3EμR,η1=η1(t)[α1(t)(1+β1(t))I+λ(t)I+α2(t)(1β2(t))I+μ]η2(t)α1(t)(1+β1(t))Iη3(t)α2(t)(1β2(t))Iη4(t)λ(t)I,η2=eαt+η2(t)(ϕ1+γ1+μ)η4(t)γ1η5(t)ϕ1,η3=eαt+η3(t)(ϕ3+γ2+μ)η4(t)γ2η5(t)ϕ3,η4=eαtη1(t)[α1(t)(1+β1(t))S+λ(t)S+α2(t)(1β2(t))S]η2(t)α1(t)(1+β1(t))Sη3(t)α2(t)(1β2(t))Sη4(t)(λ(t)Sμϕ2)η5(t)ϕ2,η5=η5(t)μ,S(0)=S00,D(0)=D00,E(0)=E00,I(0)=I00,R(0)=R00,ηi(tf)=0,i=1,2,3,4,5.

    To ensure the accuracy of the aforementioned conclusions, we adopted established methodologies from existing literature [23] and employed Matlab as a calculation tool for numerical simulation of said conclusions. We initially introduced hesitators and their corresponding parameters to the fundamental SIR model. Numerical simulation graphs were drawn to illustrate changes in population density over time in SIR and SDEIR models. Results indicated that including hesitators and their corresponding parameters in the SIR model could moderately reduce the speed of rumor propagation. The impact of each parameter on chamber population density was charted by varying their values. Finally, the model includes optimal control conditions to numerically simulate and confirm the accuracy of the aforementioned conclusions. Additionally, understanding how to manage the spread of rumors in real-life holds considerable practical significance.

    To facilitate a more precise comparison between the outcomes of SIR and SDEIR models, we initially constructed an SIR model, adopting the parameters specified in this paper. The model's flow chart is presented below (see Figure 2):

    Figure 2.  Flow chart of SIR model.

    The model dynamics equation is as follows:

    {dSdt=ΛλSIμSdIdt=λSIϕ2IμIdRdt=ϕ2IμR.

    In the year 2020, there have been a number of rumors circulating in the city of Anshan regarding the potential prevention of COVID-19 infection. These rumors include the suggestion that alcohol consumption may offer protection against the virus, that drinking garlic water may prevent infection, and that the virus may only spread at certain times. In order to simulate these rumors, we initially conducted a demographic analysis of the city. A survey of the Anshan City population revealed that the total number of educated individuals is 3170657*, of which 439634*have completed university studies, 497423* have completed high school, 170689894* have completed junior high school, and 526706*have completed primary school. The total number of uneducated individuals, or those who are illiterate, is 19911*. The average duration of the formation period for medical and health rumors is 2–3 days, the average duration of the outbreak period is 5–6 days, the average duration of the decline period is 2–3 days, and the duration of the rumor is mostly 8 days. Therefore, 1λ=5 is selected, namely a transmission rate of λ=0.2, assuming a number of people entering the system Λ=50, an immunisation rate ϕ2=0.65 and a removal rate μ=0.3.

    *http://tjj.anshan.gov.cn/html/TJJ/202106/0162398536049853.html

    Ren Qun. A Study on the Law of Rumors in Domeestic Network-An Empirical Study Based on 100 Hot Stops of Rumors Since 2016[D], Shandong University, China, 2017

    Figure 3 shows the change of population density over time in the SIR model. Among them, Λ=50,ϕ2=0.65,μ=0.3,λ=0.3.

    Figure 3.  Changes of crowd density over time in the SIR model.

    Figure 4 illustrates the population density variation in the SDEIR model with the inclusion of two chambers of rumor hesitators possessing diverse educational backgrounds on the basis of the SIR model. The results depict the changes over time, among them, Λ=50,α1=0.1,α2=0.13, β1=0.15,β2=0.33,γ1=0.23, γ2=0.27,ϕ1=0.25,ϕ2=0.65,ϕ3=0.21,μ=0.3,λ=0.3.

    Figure 4.  Changes of crowd density over time in the SDEIR model.

    By comparing Figures 3 and 4, it can be seen that the scale of rumor outbreaks (the peak of the spreader curve) decreases significantly after adding the level of education under the condition of ensuring that the original parameter remains unchanged, (the peak of the number of spreaders in Figure 3 reaches 0.58524, and then decreases, and finally it is stabilized at 0.30682; the peak of the number of spreaders in Figure 4 reaches 0.34041, and then decreases, and finally it is stable at 0.20293). This means that, when people are educated, people will think when facing rumors, which reduces the scale of rumor outbreaks to a certain extent. By comparing the results of these two models, the practical significance of the model constructed in this paper can be visualized.

    Let Λ=1,α1=0.1,α2=0.13, β1=0.15,β2=0.33,γ1=0.23,γ2=0.27, ϕ1=0.25,ϕ2=0.85,ϕ3=0.21,μ=0.3,λ=0.23, calculate R0=0.8781<1, and the setting of parameter values satisfies the conditions of Theorem 1, then rumor-free equilibrium point E0 is stable under different initial conditions.

    As illustrated in Figure 5, over time, the number of D(t),E(t),I(t),R(t) will approach zero, while the number of S(t) will approach one. In other words, as time passes, the number of S(t) will eventually equal the total number of people in the rumor system. Consequently, as time continues to elapse, rumors will eventually cease to exist.

    Figure 5.  Stability of equilibrium E0.

    Let Λ=1,α1=0.5,α2=0.6, β1=0.15,β2=0.33,γ1=0.23, γ2=0.27,ϕ1=0.25,ϕ2=0.85, ϕ3=0.21,μ=0.3,λ=0.23. R0=1.3364>1 is calculated, and the setting of parameter values satisfies the conditions of Theorem 2. At this time, Figure 6 clearly shows that the equilibrium point E of rumor propagation is locally asymptotically stable under this set of parameters.

    Figure 6.  Stability of equilibrium point E.

    As can be seen in Figure 6, in the early stage of rumor spreading, rumor unknowers begin to receive different levels of education and then start to change their views on rumors. As a result, the number of D(t),E(t),R(t) will drop sharply in a short period of time and then level off, and the number of R(t) will rise sharply and reach a peak in a short period of time, and then begin to drop and eventually level off. The above trend shows that over time, rumors can be controlled, but they do not disappear and will continue to exist.

    In order to discuss the effect of different values of parameters in the system (2.1) on rumor propagation, the following Table 2 is plotted using the uniform distribution assumption. Based on the data in Table 2, the changes in the number of cabin corresponding to the parameter changes are plotted separately and conclusions are drawn.

    Table 2.  The individual compartments and values of the individual parameters.
    Compartments and parameters value
    S(t) 100
    D(t) 50
    E(t) 50
    I(t) 50
    R(t) 50
    Λ 100
    μ 0.3
    α1 0.1
    β1 0.15
    α2 0.13
    β2 0.33
    λ 0.23
    γ1 0.23
    γ2 0.27
    ϕ1 0.25
    ϕ2 0.85
    ϕ3 0.21

     | Show Table
    DownLoad: CSV

    Figure 7 illustrates the variation in the number of S(t) as a function of different values of parameters α1,β1,λ,α2 and β2, respectively. It can be observed that parameters α1,β1,λ,α2 and S(t) exhibit a negative correlation, whereas parameters β2 and S(t) exhibit a positive correlation. Consequently, in order to increase the number of S(t), it is necessary to decrease the value of parameter α1,β1,λ,α2 and increase the value of parameter β2.

    Figure 7.  Variation in the number of S(t) under different parameter conditions ((a) is the number of S(t) under different values of parameter α1; (b) is the number of S(t) under different values of parameter β1; (c) is the number of S(t) under different values of parameter λ; (d) is the number of S(t) under different values of parameter α2; (e) is the number of S(t) under different values of parameter β2).

    Figure 8 illustrates the variation in the number of D(t) as a function of different values of parameters α1,β1,γ1 and ϕ1. It can be observed that parameters α1,β1 and D(t) exhibit a positive correlation, while parameters γ1,ϕ1 and D(t) exhibit a negative correlation. Consequently, in order to reduce the number of D(t), it is necessary to decrease the value of parameter α1,β1 and increase the value of parameter γ1,ϕ1.

    Figure 8.  Variation in the number of D(t) under different parameter conditions ((a) is the number of D(t) under different values of parameter α1; (b) is the number of D(t) under different values of parameter β1; (c) is the number of D(t) under different values of parameter γ1; (d) is the number of D(t) under different values of parameter ϕ1.

    Figure 9 illustrates the variation in the number of E(t) as a function of different values of parameters α2,β2,γ2 and ϕ3. It can be observed that parameters α2 and E(t) exhibit a positive correlation, while parameters β2,γ2,ϕ3 and E(t) exhibit a negative correlation. Consequently, in order to reduce the number of E(t), it is necessary to decrease the value of parameter α2 and increase the value of parameter β2,γ2,ϕ3.

    Figure 9.  Variation in the number of E(t) under different parameter conditions ((a) is the number of E(t) under different values of parameter α2; (b) is the number of E(t) under different values of parameter β2; (c) is the number of E(t) under different values of parameter γ2; (d) is the number of E(t) under different values of parameter ϕ3.

    Figure 10 illustrates the variation in the number of I(t) as a function of different values of parameters λ,γ1,γ2 and ϕ2. It can be observed that parameters λ,γ1,γ2 and I(t) exhibit a positive correlation, while parameters ϕ2 and I(t) exhibit a negative correlation. Consequently, in order to reduce the number of I(t), it is necessary to decrease the value of parameter λ,γ1,γ2 and increase the value of parameter ϕ2.

    Figure 10.  Variation in the number of I(t) under different parameter conditions ((a) is the number of I(t) under different values of parameter λ; (b) is the number of I(t) under different values of parameter γ1; (c) is the number of I(t) under different values of parameter γ2; (d) is the number of I(t) under different values of parameter ϕ2.

    Figure 11 illustrates the variation in the number of R(t) as a function of different values of parameters ϕ1,ϕ2 and ϕ3. It can be observed that parameters ϕ1,ϕ2,ϕ3 and R(t) exhibit a positive correlation. Consequently, in order to increase the number of R(t), it is necessary to increase the value of parameter ϕ1,ϕ2,ϕ3.

    Figure 11.  Variation in the number of R(t) under different parameter conditions ((a) is the number of R(t) under different values of parameter ϕ1; (b) is the number of R(t) under different values of parameter ϕ2; (c) is the number of R(t) under different values of parameter ϕ3.

    In conclusion, in order to reduce the spread of rumors, it is necessary to decrease the natural contact rate between rumor unknowers and rumor spreaders, increase the percentage of rumor unknowers with a higher level of education, and decrease the transmission rate and increase the immunity rate. Consequently, in order to mitigate the spread of rumors, the following section will continue to detail the most effective method of controlling the spread of rumors by manipulating parameter α1,α2,β1,β2,λ values.

    In this section, by controlling the above five parameters, that is, α1,α2,β1,β2,λ, Figure 12 is drawn.

    Figure 12.  The change of the density of D(t),E(t),I(t) with time under different strategies.

    Because optimal control mainly controls the population density of three chambers: Rumor hesitators with less education, rumor hesitators with higher education, and rumor spreaders. Therefore, during the drawing process, we only display D(t), E(t) and I(t).

    In Figure 12,

    (a) optimal control: α1=0,α2=0,β1=0,β2=1,λ=0

    (b) middle control: α1=0.5,α2=0.5,β1=0.5,β2=0.5,λ=0.5

    (c) constant control: α1=1,α2=1,β1=1,β2=0,λ=1.

    After observing Figure 12, it becomes clear that optimal control is more effective than middle control or constant control at all. With optimal control, the number of less-educated rumor hesitators, higher-educated rumor hesitators, and rumor spreaders can all be reduced to zero, resulting in successful rumor control.

    Combined with real-life scenarios, specific measures can be employed to regulate the five key parameters outlined in the article, thus controlling the spread of rumors. The government and official media are recommended to reinforce their supervision of online information propagation, which could effectively decrease the natural contact rate between rumor spreaders and rumor unknowers. Furthermore, the government and schools could enhance the education of those rumors unknowers and uplift people's education level to limit the propagation of rumors.

    We studied the influence of education level on rumor propagation, established the SDEIR model, analyzed the basic reproduction number, rumor equilibrium point and stability, optimized the control strategy, and verified the theoretical results through numerical simulation.

    The main conclusions of this paper are as follows:

    (1) When R0<1, the rumor will disappear from the system over time; when R0>1, the rumor will not disappear in the future, but will reach a stable state.

    (2) Based on the classical SIR rumor propagation model, the scale of rumor outbreak is significantly reduced after the education level is introduced.

    (3) Reducing the contact rate of I(t) and S(t) with different levels of education, increasing the proportion of S(t) with higher education level, reducing the transmission rate and improving the immunization rate can inhibit the spread of rumors to varying degrees.

    (4) When controlling parameter α1(t)=0,α2(t)=0,β1(t)=0,β2(t)=1,λ(t)=0, the optimal control effect is achieved and the purpose of controlling rumor propagation is realized.

    (5) Although improving people's education level can reduce the spread of rumors to a certain extent, it cannot completely stop the spread of rumors.

    The spread of rumors is a complex social phenomenon, and further research and comprehensive measures are needed to effectively deal with the spread of rumors. In fact, in addition to the changes between different chambers considered in this paper, S(t) may be directly transformed into R(t), which also has an important impact on the process of rumor propagation. In addition to government and state media, other media have the potential to change the way rumors spread. However, these cases are not detailed in this article. It is hoped that future studies can comprehensively consider the influence of various factors and provide more effective control measures and prevention strategies to meet the challenges brought by rumor propagation.

    Hongshuang Wang: Writing-original draft and writing-review & editing; Sida Kang: Writing-original draft; Yuhan Hu: Supervision and writing-review & editing. All authors have read and approved the final version of the manuscript for publication.

    The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

    We thank the anonymous Editors and Reviewers warm work earnestly.

    This work was supported by the Social Science Planning Fund of Liaoning Province China (No. L22AGL015).

    The authors declare no conflict of interest.



    [1] X. Zhao, J. Wang, Dynamical model about rumor spreading with medium, Discrete Dyn. Nat. Soc., 2013 (2013). http://dx.doi.org/10.1155/2013/586867 doi: 10.1155/2013/586867
    [2] D. J. Daley, D. G. Kendall, Stochastic rumors, IMA J. Appl. Math., 1 (1965), 42–55. https://doi.org/10.1093/imamat/1.1.42 doi: 10.1093/imamat/1.1.42
    [3] D. P. Maki, Mathematical models and applications, 1973.
    [4] J. Gu, W. Li, X. Cai, The effect of the forget-remember mechanism on spreading, Eur. Phys. J. B, 62 (2008), 247–255. https://doi.org/10.1140/epjb/e2008-00139-4 doi: 10.1140/epjb/e2008-00139-4
    [5] L. Zhao, J. Wang, Y. Chen, Q. Wang, J. Cheng, H. Cui, Sihr rumor spreading model in social networks, Physica A, 391 (2012), 2444–2453. https://doi.org/10.1016/j.physa.2011.12.008 doi: 10.1016/j.physa.2011.12.008
    [6] W. Jing, L. Min, W. Y. Qi, Z. Z. Chen, Z. L. Qiong, The influence of oblivion-recall mechanism and loss-interest mechanism on the spread of rumors in complex networks, Int. J. Mod. Phys. C, 30 (2019), 1950075. https://doi.org/10.1142/S012918311950075X doi: 10.1142/S012918311950075X
    [7] L. L. Xia, G. P. Jiang, B. Song, Y. R. Song, Rumor spreading model considering hesitating mechanism in complex social networks, Physica A, 437 (2015), 295–303. https://doi.org/10.1016/j.physa.2015.05.113 doi: 10.1016/j.physa.2015.05.113
    [8] Y. H. Xing, G. X. Yang, SE2IR invest market rumor spreading model considering hesitating mechanism, J. Syst. Sci. Inform., 7 (2018), 54–69. https://doi.org/10.21078/JSSI-2019-054-16 doi: 10.21078/JSSI-2019-054-16
    [9] Y. Zan, J. Wu, P. Li, Q. Yu, Sicr rumor spreading model in complex networks: Counterattack and self-resistance, Physica A, 405 (2014), 159–170. https://doi.org/10.1016/j.physa.2014.03.021 doi: 10.1016/j.physa.2014.03.021
    [10] A. Jain, J. Dhar, V. K. Gupta, Optimal control of rumor spreading model on homogeneous social network with consideration of influence delay of thinkers, Differ. Equat. Dyn. Syst., 31 (2023), 113–134. https://doi.org/10.1007/s12591-019-00484-w doi: 10.1007/s12591-019-00484-w
    [11] H. Yao, Y. Zou, Research on rumor spreading model with time delay and control effect, J. Syst. Sci. Inform., 7 (2019), 373–389. https://doi.org/10.21078/JSSI-2019-373-17 doi: 10.21078/JSSI-2019-373-17
    [12] Y. Zhang, Y. Su, L. Weigang, H. Liu, Interacting model of rumor propagation and behavior spreading in multiplex networks, Chaos Soliton. Fract., 121 (2019), 168–177. https://doi.org/10.1016/j.chaos.2019.01.035 doi: 10.1016/j.chaos.2019.01.035
    [13] L. Yang, Z. Li, A. Giua, Containment of rumor spread in complex social networks, Inform. Sci., 506 (2020), 113–130. https://doi.org/10.1016/j.ins.2019.07.055 doi: 10.1016/j.ins.2019.07.055
    [14] K. Afassinou, Analysis of the impact of education rate on the rumor spreading mechanism, Physica A, 414 (2014), 43–52. https://doi.org/10.1016/j.physa.2014.07.041 doi: 10.1016/j.physa.2014.07.041
    [15] Y. Hu, Q. Pan, W. Hou, M. He, Rumor spreading model considering the proportion of wisemen in the crowd, Physica A, 505 (2018), 1084–1094. https://doi.org/10.1016/j.physa.2018.04.056 doi: 10.1016/j.physa.2018.04.056
    [16] D. Li, Y. Zhao, Y. Deng, Rumor spreading model with a focus on educational impact and optimal control, Nonlinear Dynam., 112 (2024), 1575–1597. https://doi.org/10.1007/s11071-023-09102-5 doi: 10.1007/s11071-023-09102-5
    [17] W. Pan, W. Yan, Y. Hu, R. He, L. Wu, G. Rasool, Dynamic analysis and optimal control of rumor propagation model with reporting effect, Adv. Math. Phys., 2022 (2022), 1–14. https://doi.org/10.1155/2022/5503137 doi: 10.1155/2022/5503137
    [18] O. Diekmann, J. Heesterbeek, M. G. Roberts, The construction of next generation matrices for compartmental epidemic models, J. R. Soc. Interface, 7 (2010), 873–885. https://doi.org/10.1098/rsif.2009.0386 doi: 10.1098/rsif.2009.0386
    [19] P. Van den Driessche, J. Watmough, Reproduction numbers and subthreshold endemic equilibria for compartmental models of disease transmission, Math. Biosci., 180 (2002), 29–48. https://doi.org/10.1016/s0025-5564(02)00108-6 doi: 10.1016/s0025-5564(02)00108-6
    [20] J. J. Anagnost, C. A. Desoer, An elementary proof of the routh-hurwitz stability criterion, Circ. Syst. Signal Pr., 10 (1991), 101–114. https://doi.org/10.1007/BF01183243 doi: 10.1007/BF01183243
    [21] S. F. Marinósson, Lyapunov function construction for ordinary differential equations with linear programming, Dynam. Syst., 17 (2002), 137–150. https://doi.org/10.1080/0268111011011847 doi: 10.1080/0268111011011847
    [22] J. P. Lasalle, The stability of dynamical systems, Soc. Ind. Appl. Math., 1976. https://doi.org/10.1137/1.9781611970432
    [23] S. Kang, X. Hou, Y. Hu, H. Liu, Dynamical analysis and optimal control of the developed information transmission model, Plos One, 17 (2022), e0268326. https://doi.org/10.1371/journal.pone.0268326 doi: 10.1371/journal.pone.0268326
    [24] F. L. Lewis, D. Vrabie, V. L. Syrmos, Optimal control, John Wiley Sons, 2012. https://doi.org/10.1002/9781118122631
    [25] S. Kang, X. Hou, Y. Hu, H. Liu, Dynamic analysis and optimal control considering cross transmission and variation of information, Sci. Rep., 12 (2022), 18104. https://doi.org/10.1038/s41598-022-21774-4 doi: 10.1038/s41598-022-21774-4
    [26] O. Sharomi, T. Malik, Optimal control in epidemiology, Ann. Oper. Res., 251 (2017), 55–71. https://doi.org/10.1007/s10479-015-1834-4 doi: 10.1007/s10479-015-1834-4
    [27] W. H. Fleming, R. W. Rishel, Deterministic and stochastic optimal control, Springer Science & Business Media, 1 (2012).
    [28] S. Kang, X. Hou, Y. Hu, H. Liu, Dynamic analysis and optimal control of a stochastic information spreading model considering super-spreader and implicit exposer with random parametric perturbations, Front. Phys., 11 (2023), 377. https://doi.org/10.3389/fphy.2023.1194804 doi: 10.3389/fphy.2023.1194804
    [29] M. Hale, Y. Wardi, H. Jaleel, M. Egerstedt, Hamiltonian-based algo rithm for optimal control, arXiv: 1603.02747, 2016. https://doi.org/10.48550/arXiv.1603.02747
    [30] R. E. Kopp, Pontryagin maximum principle, Math. Sci. Eng., Elsevier, 5 (1962), 255–279. https://doi.org/10.1016/S0076-5392(08)62095-0
    [31] L. Bittner, L. S. Pontryagin, V. G. Boltyanskii, R. V. Gamkrelidze, E. F. Mishchenko, The mathematical theory of optimal processes, New York, ZAMM-J. Appl. Math. Mech. Z. Angew. Math. Mech., 43 (1962), 514–515. https://doi.org/10.1002/zamm.19630431023 doi: 10.1002/zamm.19630431023
  • Reader Comments
  • © 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1279) PDF downloads(77) Cited by(0)

Figures and Tables

Figures(12)  /  Tables(2)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog