Online Opinion Risk Control and Ideology Construction of College Students in New Media Environment

The main basic statistical characterizations of complex networks are as follows: 1)

Degree and degree distribution

Degree is the most basic statistical feature quantity of complex networks. For an undirected network, the degree is defined as the number of connected edges of a node, but for a directed network, the connected edges of a node are directed, and the degree of a node can be further divided into in-degree and out-degree, in-degree is the number of connected edges pointing to the node, and out-degree is the number of connected edges pointing from the node to other nodes. The degree of a node is one of the measures of the importance of a node in a network. The topology of a complex network can be represented by the adjacency matrix A = (A_ij) and the degree of a node i in the network is represented by k_i. The formula is as follows: (1)ki=∑j=1nAij$${k_i} = \sum\limits_{j = 1}^n {{A_{ij}}}$$

Where n is the number of nodes of the network. The degree of a single node is a study of the complex network from a local point of view, then the average degree of the network is a study of the complex network from a global point of view, and its value is the average of the degrees of all nodes in the network, denoted by 〈k〉, and computed by the formula: (2)〈k〉=1n∑i=1nki=1n∑i=1n∑j=1nAij$$\langle k\rangle = \frac{1}{n}\sum\limits_{i = 1}^n {{k_i}} = \frac{1}{n}\sum\limits_{i = 1}^n {\sum\limits_{j = 1}^n {{A_{ij}}} }$$

In an undirected network, each edge is connected to two nodes, and if the number of edges in the network is denoted as m, then there are nodes with 2m edges in the network. The number of nodes attached to all edges and the sum of the degrees of all nodes must be equal, then: (3)2m=∑i=1nki$$2m = \sum\limits_{i = 1}^n {{k_i}}$$

To wit: (4)m=12∑i=1nki=12∑i=1n∑j=1nAij$$m = \frac{1}{2}\sum\limits_{i = 1}^n {{k_i}} = \frac{1}{2}\sum\limits_{i = 1}^n {\sum\limits_{j = 1}^n {{A_{ij}}} }$$

The degree of all nodes in the network is ranked and the proportion of nodes with degree k among all nodes is counted and the result obtained is the degree distribution p(k). 2)

Clustering Coefficient

The clustering coefficient is used to measure the size of the probability of the occurrence of this phenomenon in a network. Its more common definition in complex networks is to measure the probability that two neighbors of a node are neighbors of each other. The clustering coefficient C_i for node i with degree k_i is calculated as: (5)Ci=Ei(ki(ki−1))/2$${C_i} = \frac{{{E_i}}}{{({k_i}({k_i} - 1))/2}}$$

where E_i is the number of pairs of nodes where the neighbor nodes of node i are neighbors of each other, and (k_i(k_i − 1))/2 is the sum of pairs of neighbor nodes of node i. From equation (5), it can be seen that as the degree k increases, C_i keeps decreasing, which means that nodes with large degrees tend to have smaller clustering coefficients, and this phenomenon can be explained by the fact that a network can be partitioned into several tight associations, with many connecting edges between nodes within the same association and fewer connecting edges between nodes belonging to different associations.

For a network with a given degree distribution, the clustering coefficient C of the network is calculated as follows: (6)C=1n[〈k2〉−〈k〉]2〈k〉3$$C = \frac{1}{n}\frac{{{{[\langle {k^2}\rangle - \langle k\rangle ]}^2}}}{{{{\langle k\rangle }^3}}}$$

where 〈k²〉 is the second order moments and 〈k〉 is the average degree. For a given network, both 〈k²〉 and 〈k〉 have fixed finite values. From Eq. (6), it can be seen that the value of C becomes smaller when n → ∞, so the clustering coefficients are usually smaller for large-scale networks. 3)

Average shortest path

In complex networks, paths refer to the pathways between nodes, starting from a node, along the edges in the network can reach another node, it is said that there is a pathway between the two nodes. The pathway with the least number of edges is called the shortest path. From a macroscopic point of view, the network has an average shortest path, which is denoted as L. Then, the formula for L is: (7)L=1n(n−1)/2∑i≥jdij$$L = \frac{1}{{n(n - 1)/2}}\sum\limits_{i \ge j} {{d_{ij}}}$$

where n is the number of nodes in the network and d_ij is the shortest path between any two nodes i and j in the network. The size of the average shortest path of a complex network represents the tightness of the network topology.

There are some complex networks where most of the nodes in the network are not connected to each other by edges, but most of the nodes have short distances between them, and complex networks with this characteristic are called small-world networks. Small world networks are intricate networks that have high clustering coefficients and low average paths. Offline interpersonal network is a typical small-world network, and the famous “six degrees of separation” theory proves that interpersonal connections can be established in a social network through only six people.

The WS small-world network model is a new network formed by changing the connecting edges of a regular network with a certain probability, which is different from a regular network and is not a completely randomized network, and the steps of constructing a WS small-world network are briefly described as follows: 1)

First, given n node with a ring distribution, then, let each node create connected edges with its left and right neighboring k/2 nodes.

It is important to note in the construction process that in step (2), no white connected and duplicate edges can be created.

The degree distribution of some complex networks shows a serious uneven distribution, i.e., a small number of nodes in the network have very large degree values, and most of the nodes have very small degree values, and such a network is defined as a scale-free network.

Scale-free networks are generally constructed by the BA model, which is a model based on the growth pattern, i.e., new nodes are constantly joining the network and will preferentially form connection edges with those nodes that have large degrees.The specific principles of constructing a BA scale-free network are briefly described as follows: 1)

Start with a very small connected network with m₀ node, then, based on this network structure, introduce new nodes in turn, and establish connection edges with the existing m(m ≤ m₀) nodes.

where n is the number of nodes in the network.

Let v be an opinion of each node (each individual is considered a node in a complex network, so in this paper we will collectively refer to them as nodes), combining the opinions to be polar opposites at either end of its boundary, and v varying in a closed subset of the boundary. Let l be the number of connections of the opinions, which is a discrete variable with a value that varies between 0 and a maximum number. The maximum number here usually refers to some fixed value, e.g., a value that is n orders of magnitude smaller than the network size.

Let f be the evolution equation for the density function: (9)f=f(v,l,t);f:V×L×R+→R+$$f = f(v,l,t)\:;\:f:V \times L \times {R^ + } \to {R^ + }$$

Where: v ∈ V(V ∈ [ − 1, 1]) is an opinion variable. l ∈ L and satisfies L = {0, 1, ⋯, l_max}, is a discrete variable describing the number of connections. t ∈ R⁺ is a common time variable. For each time t ≥ 0 the following marginal densities can be computed: (10)ρ(l,t)=∫tf(v,l,t)dv$$\rho (l,t) = \int_t f (v,l,t)\:{\text{d}}v$$

The evolution of the number of connections is equivalent to the degree distribution of the network, assuming that the number of nodes is conserved, i.e: (11)∑l=0lmaxρ(l,t)=1$$\sum\limits_{l = 0}^{{l_{\max }}} \rho (l,t)\: = \:1$$

The overall opinion opinion distribution is defined as a marginal density function of: (12)g(v,t)=∑l=0lmaxf(v,l,t)$$g(v,t) = \sum\limits_{l = 0}^{{l_{\max }}} f (v,l,t)$$

Describing the nodes from micro point of view nodes are modifying the opinion with binary interaction and number of connections, if two nodes opinion and number of connections, i.e., (v, l) and (v_*, l_*) meet then the opinion opinion after interaction becomes v′ and v*′$${v_*}^\prime$$ and the opinion after interaction takes the range of values V ∈ [ − 1, 1]. The time evolution equation of the function is represented by the Boltzmann equation [27]: (13)ddtf(v,l,t)+τ[f(v,l,t)]=Q(f,f)(v,l,t)$$\frac{d}{{dt}}f(v,l,t)\: + \:\tau \left[ {f(v,l,t)\:} \right]\: = \:Q\left( {f,f} \right)\:\left( {v,l,t} \right)$$

where: τ[·] denotes the operator associated with the evolution of network connections and Q(·, ·) denotes the binary interaction operator. The weak form of the computationally obtained formula is as follows: (14)ddt∫Vf(v,l)ψ(v)dv+∫Vτ[f(v,l)]ψ(v)dv =λ∑l*=0lmax∫V2(ψ(v′)−ψ(v))f(v*,l*)f(v,l)dvdv*$$\begin{array}{l} \frac{{\text{d}}}{{{\text{d}}t}}\int_V f (v,l)\:\psi (v)\:{\text{d}}v\: + \:\int_V \tau \:[f(v,l)\:]\psi (v)\:{\text{d}}v \\ \: = \lambda \sum\limits_{{l_*}\: = \:0}^{{l_{{\text{max}}}}} {\int_{{V^2}} {(\psi ({v^\prime })\: - \psi (v))} } f({v_*}\:,{l_*}\:)f(v,l)\:{\text{d}}v{\text{d}}{v_*} \\ \end{array}$$

Where “〈⋅〉$$\left\langle \cdot \right\rangle$$” denotes the expected value of the random variable.

Regarding the opinion evolution process, the process of adding and deleting edges in complex networks is strictly related to the static scale-free scale. In this paper, we set the operator τ[·] in the dynamic description of the network to be a detail in the model of connection evolution of nodes in discrete space, and τ[·] define expressions to describe the process of connection evolution in terms of deletion and addition of edges in the network through preferred connection combinations.

For each l = 1, 2, ⋯, l_max − 1 defined as follows: (15)τ[f(v,l,t)]=−2Sd(f;v)ω+β[(l+1+β)f(v,l+1,t)−(l+β)f(v,l,t)]− 2Sa(f;v)ω+α[(l−1+α)f(v,l−1,t)−(l+α)f(v,l,t)]$$\begin{array}{l} \tau [f(v,l,t)\:]\: = - \:\frac{{2{S_d}(f;v)}}{{\omega + \beta }}[(l\: + \:1\: + \beta )f(v,l\: + \:1,t)\: - (l + \beta )f(v,l,t)\:]\: - \\ \frac{{2{S_a}\:(f;v)}}{{\omega + \alpha }}\:[(l - 1\: + \alpha )f(v,l - 1,t)\: - (l + \alpha )f(v,l,t)] \\ \end{array}$$

Where: α, β > 0 is the attraction factor. S_d(f; v) ≥ 0 is the link removal probability. S_a(f; v) ≥ 0 is the rate of adding links, and many studies have shown that removals and additions generally occur in pairs. ω = ω(t) denotes the average connectivity density. ω(t)$$\omega \left( t \right)$$ is defined as follows: .(16)ω(t)=∑l=0lmaxlρ(l,t)$$\omega (t)\: = \:\sum\limits_{l = 0}^{{l_{\max }}} l \:\rho (l,t)$$

Eq. (15) describes the f(v, l, t) net increment due to changes in connectivity between nodes with boundaries defined as follows: (17){ τ[f(v,0,t)]=−2Sd(f;v)ω+β(1+β)f(v,1,t)+2Sa(f;v)ω+ααf(v,0,t) τ[f(v,lmax,t)]=2Sd(f;v)ω+β(lmax+β)f(v,lmax,t)− 2Sa(f;v)ω+α(lmax−1+α)f(v,lmax−1,t)$$\left\{ {\begin{array}{*{20}{l}} {\tau \left[ {f(v,0,t)} \right] = - \frac{{2{S_d}(f;v)}}{{\omega + \beta }}(1 + \beta )f(v,1,t) + \frac{{2{S_a}(f;v)}}{{\omega + \alpha }}\alpha f(v,0,t)} \\\quad\quad\quad \begin{array}{l} \tau [f(v,{l_{max}},t)] = \frac{{2{S_d}(f;v)}}{{\omega + \beta }}({l_{max}} + \beta )f(v,{l_{max}},t) - \\ \frac{{2{S_a}(f;v)}}{{\omega + \alpha }}({l_{max}} - 1 + \alpha )f(v,{l_{max}} - 1,t) \\ \end{array} \end{array}} \right.$$

Two factual existence cases should be taken into account in Eq. (17), i.e., nodes cannot be removed when there are 0 connections between them, and nodes cannot be added when they have reached a maximum value of connections between them.

If the characteristic rate is defined as: (18){ Sd(f;v)=Udω+βωf+βg(v,t) Sa(f;v)=Uaω+αωf+αg(v,t)$$\left\{ {\begin{array}{*{20}{c}} {{S_d}(f;v) = \:{U_d}\:\frac{{\omega + \beta }}{{\omega f + \beta g(v,t)}}} \\ {{S_a}(f;v) = \:{U_a}\:\frac{{\omega + \alpha }}{{\omega f + \alpha g(v,t)}}} \end{array}} \right.$$

Among them: (19)ωf(v,t)=∑l=0lmaxlf(v,l,t)$$\omega f(v,t)\: = \:\sum\limits_{l = 0}^{{l_{\max }}} l f(v,l,t)$$

density associated with the preferred combination of attachment process (α, β ≈ 0) and unification process (α, β > 0) for each opinion opinion v in Eq. (15) is f(v,l,t)g(v,t)$$\frac{{f(v,l,t)}}{{g(v,t)}}$$.

Assuming that the total number of nodes in τ[:] is conserved, considering the evolution of the average connectivity density ω over square kilometers in Eq. (16), the evolution of the average connectivity density yields that for each t ≥ 0 there is: (20)ddtω(t)=−2∫VSd(f;v)ωf+βg(v,t)ω+βdv +2∫VSa(f;v)ωf+αg(v,t)v+αdv +2βω+β∫VSa(f;v)ωf+αg(v,t)v+αdv −2(lmax+α)ω+α∫VSa(f;v)f(v,lmax,t)dv$$\begin{array}{l} \frac{d}{{dt}}\:\omega (t) = - \:2\int_V {{S_d}} (f;v)\:\frac{{\omega f + \beta g(v,t)}}{{\omega + \beta }}\:dv\: \\\quad\quad\quad + 2\int\limits_V {{S_a}} (f;v)\frac{{\omega f + \alpha g(v,t)}}{{v + \alpha }}dv \\\quad\quad\quad + \frac{{2\beta }}{{\omega + \beta }}\int\limits_V {{S_a}} (f;v)\frac{{\omega f + \alpha g(v,t)}}{{v + \alpha }}dv \\\quad\quad\quad - \frac{{2({l_{\max }} + \alpha )}}{{\omega + \alpha }}\int\limits_V {{S_a}} (f;v)f(v,{l_{\max }},t)dv \\ \end{array}$$

Typically ω is not conserved and gradually reaches a conserved state when the attraction coefficient is β = 0. The density function f(v, l, t) decays at a sufficiently fast rate if the eigenrate under the conditions of S_d = S_a is given by Eq. (19) and is set U_d = U_a.

For node conservation number setting: (21)∑l=0lmaxτ[f(v,l,t)]dv=0$$\sum\limits_{l = 0}^{{l_{\max }}} \tau \left[ {f(v,l,t)\:} \right]dv\: = \:0$$

The conservation of the total number of nodes can be obtained when the conditions required by Eq. (18) are satisfied, and the formula is as follows: (22)ddt∑l=0lmaxρ(l,t)=−∫V∑l=0lmaxτ[f(v,l,t)]dv=0$$\frac{d}{{dt}}\sum\limits_{l = 0}^{{l_{\max }}} \rho (l,t) = - \int_V {\sum\limits_{l = 0}^{{l_{\max }}} \tau } \left[ {f(v,l,t)\:} \right]dv\: = \:0$$

There is a special case where S_d and S_a are constants, when operator τ[: ] is linear, denoted by T[·], and the evolution of network connections is not affected by changes in public opinion, yielding Eq: (23)ddtρ(l,t)+T[ρ(l,t)]=0$$\frac{d}{{dt}}\rho (l,t)\: + \:T\left[ {\rho (l,t)\:} \right]\: = \:0$$

Among them: (24)T[ρ(l,t)]=−2Sdω+β[ (l+1+β)ρ(l+1,t)−(l+β)ρ(l,t)] −2Saω+α[(l−1+α)ρ(l−1,t) −(l+α)ρ(l,t)]$$T\left[ {\rho (l,t)} \right] = - \frac{{2{S_{\text{d}}}}}{{\omega + \beta }}\left[ \begin{array}{l} (l + 1 + \beta )\rho (l + 1,t) - (l + \beta )\rho (l,t)] \\ - \frac{{2{S_a}}}{{\omega + \alpha }}[(l - 1 + \alpha )\rho (l - 1,t) \\ - (l + \alpha )\rho (l,t) \\ \end{array} \right]$$

The dynamical equations of Eq. (24) correspond to the connectivity density functions of the optimal connectivity process (α, β ≈ 0) and the unification process (α, β > 0) as ρ(l, t). Due to the nature of complex network connectivity, the asymptotics of the connectivity density, denoted by ω, can be proved in general in the linear case of S_d = S_a, β = 0.

Let the static solution be equivalent for each l ∈ L: (25)(l+1)ρ∞(l+1)=1ω+α[(r(2l+α)+ωα)ρ∞(l) −ω(l−1+α)ρ∞(l−1)]$$\begin{array}{rcl} (l + 1){\rho _\infty }(l + 1) = \:\frac{1}{{\omega + \alpha }}\:[(r(2l\: + \:\alpha )\: + \:\omega \alpha ){\rho _\infty }\:(l) \\ \: - \omega (l\: - \:1\: + \:\alpha ){\rho _\infty }(l\: - \:1)\:] \\ \end{array}$$

And then get: (26)ρ∞(l)=(ωω+α)l1l!α(α+1)⋯(α+l−1)ρ∞(0)$${\rho _\infty }(l) = {(\frac{\omega }{{\omega + \alpha }})^l}\frac{1}{{l!}}\alpha (\alpha + 1) \cdots (\alpha + l - 1){\rho _\infty }(0)$$

Among them: (27)ρ∞(0)=(αα+ω)α$${\rho _\infty }(0) = {(\frac{\alpha }{{\alpha \: + \:\omega }})^\alpha }$$

Further approximate solutions are obtained in the case of α ≥ 1 and α = 0, for α taking a positive exponent of 10, where the preferential connectivity process described by Eq. (24) is corrupted and the network is approximated as a stochastic network with a degree distribution obeying a Poisson distribution, which is obtained in the practical case when the limiting value of α tends to positive infinity: (28)ρ∞(l)=limα→+∞(1+ωα)−αωl=e−ll!ωl$${\rho _\infty }\:(l) = \:\mathop {\lim }\limits_{\alpha \to + \infty } {(\:1\: + \:\frac{\omega }{\alpha })^{ - \alpha }}{\omega ^l}\: = \:\frac{{{e^{ - l}}}}{{l!}}\:{\omega ^l}$$

where e is a natural constant, and for the negative exponential case where ω ≥ 1 and α take the value 10, the density distribution can be accurately approximated as a single exponential truncated power law, i.e., denoted as: (29)ρ∞(l)=(αω)ααl$${\rho _\infty }(l) = {(\frac{\alpha }{\omega })^\alpha }\frac{\alpha }{l}$$

The factors related to the spread of online public opinion can be roughly divided into three categories: the first category is the measurable quantity category, including the number of subjects, the number of subjects that can disseminate information, the number of subjects that are not sufficiently sensitive to the same information, and the number of correlations with links to the outside world. The second category is the control category, which includes the prescribed level of public opinion dissemination, the reality of the start of the control work, its duration, and the percentage of uncredible information. The third category is the environment category, which encompasses factors like the scope of dissemination, the speed of dissemination, and the level of dissemination. In practice, not all categories of factors can be realized for their control of dynamics. The primary explanation is that the kinetic features are not significant and the kinetic constants cannot be identified. Therefore, in the practical work and modeling, the design of some kinetic constants is discarded, and the propagation variables of the public opinion propagation model are traded off under the premise of ensuring that the propagation of a certain public opinion can be demonstrated completely.

In addition to the above factors, the following concepts are also often used when expressing public opinion:

Time: In the process of public opinion dissemination, time is the key node for measuring the effect of opinion control and controlling the resolution of negative public opinion. Therefore, in this study, it is instructive to take time as a constant scalar, and it is also possible to select a specific control time node through simulation, and then derive an analysis of the effectiveness of the public opinion guidance policy through the displacement of the node.

Strength of Public Opinion: That is, the degree of strength or weakness of the ability to spread public opinion. Netizens have different sensitivities to a point of view and different concerns as a group, so public opinion itself will have different degrees of influence when it breaks out. In the simulation variables, the concept of probability is introduced, assuming that a certain propagation subject is O, and its ability to influence the surrounding is denoted by P. P∈[0,1]$$P \in \left[ {0,1} \right]$$, the closer it is to 1, the stronger the influence effect is, and after a number of successive multiplications, the greater the number of factors A that reach the far end of the influence will be.

Receiving population: from the perspective of the receiving population, each person is both a receiver and an emitter of online opinion information, with both the desire to disseminate and the possibility of terminating the dissemination at any time with the marginalization, i.e., the weakening of interest, of the BA scale-free network. Based on this, the receiving population is also considered as one of the data to be considered separately, and the percentage of each attitude in the population is used to indicate the specific stage and current status of opinion dissemination.

In the development of traditional dynamics, there is a class of VM viral models that have been widely used in a variety of different fields, ranging from computer virus propagation, social network construction, to information dissemination, which have been used as classical models in various fields of dynamics, and the viral models have been developed through continuous evolution and formed three kinds of models, i.e., the SI/SIS, and the SIR models.

In the classical transmission model, scholars classify the population into transmission state S, susceptible state I, and immune state R, where state S has a certain probability of transmitting the virus or information from one infected subject O to another infected subject A, and state A has a chance to become a transmitter like O. I represents a susceptible population that has not received any effect in the transmission environment, and R indicates complete immunity, which is no longer infected after repeated transmission.

The SI model is based on the wireless transmission scenario derived from the traditional scale-free network model, i.e., when an individual is affected, then it does not recover and has a certain amount of influence to continue affecting other subjects. Assuming that the probability that an individual is exposed to infection is p, and the total number of individuals is designed to be N, when N is distributed in the scale-free network, the infection time can be analyzed, and the propagation model is as follows: (30){ dSdt=−pSIN dIt=pSIN$$\left\{ {\begin{array}{*{20}{l}} {\frac{{dS}}{{dt}} = - p\frac{{SI}}{N}} \\ {d\frac{I}{t} = p\frac{{SI}}{N}} \end{array}} \right.$$

The individual rate of change over time is: (31)didt=pi(1−i)$$\frac{{di}}{{dt}} = pi(1 - i)$$

So: (32){ i(t)=i0ept1−i0+i0ept i0=i(0)$$\left\{ {\begin{array}{*{20}{l}} {i(t) = \frac{{{i_0}{e^{pt}}}}{{1 - {i_0} + {i_0}{e^{pt}}}}} \\ {{i_0} = i(0)} \end{array}} \right.$$

As in equation (32), the t setting is brought in to derive the number of infected individuals from the 0th moment to the tnd moment Therefore, in the initial time period, when most individuals are in the I state, any S state meets the I state individuals will be free to propagate, and therefore exponentially grow and gradually converge to the saturation state.

However, in the actual propagation process, each individual can not always be the propagation state, that is, due to the insensitivity of the information and the propagation of the willingness of the change occurs not the same change, such as the probability of r differentiation into the state of R will not spread, or after the forgetting of the state from the state of S back to the state of I. In turn, SIS and SIR models that can reflect self-healing behavior emerge.

The SIR model is the classical model of scale-free network propagation and has a very important place in most research works on propagation types [28].

The SIR model is schematically shown in Fig. 1. Based on the SI model, the SIR model considers the probability of a I state transitioning to a S state as q and a S state differentiating into a state that no longer propagates as R.

Figure 1.

SIR Model

The enhancement of legal construction can provide strong support and guarantee for the construction of network ideological security. By strengthening the construction of the network legal system and regulating the words and deeds of opinion leaders on the network, the legal system can provide support for network ideological security. By improving the real-name authentication of network information, we can also establish a binding network monitoring mechanism to regulate the words and actions of network opinion leaders. Online rumors continue to seriously mislead the public judgment, resulting in a bad social impact, the need to ferment public opinion at the same time to increase disciplinary efforts to reduce the irresponsible words and deeds of bad media, to play the mainstream opinion leaders in the supervision of public opinion on the positive leading role. At the same time, it is necessary to strengthen legal education, enhance the ability of netizens to identify and reduce the blind worship of bad online opinion leaders, and follow the trend.

As the front line of college students’ ideological education, schools should effectively shoulder the political responsibility, stand firm and guard the ideological position, and be responsible for guarding the land, guarding the land and being responsible for guarding the land. Colleges and universities can actively build campus media culture exchange platforms, by opening college ideological education positions in major mainstream media, creating official public numbers, official microblogging and other new media communication matrices, using them as effective platforms to guide the cultivation of network opinion leaders, and cultivating opinion leaders who are easily accepted by college students and who can lead the transmission of positive energy. Embedding media literacy education courses in the talent cultivation process of universities is popularized, promoting theoretical knowledge of online media literacy, especially in terms of online media and laws and regulations. Colleges and universities can carry out multi-party exchanges and communicate with young students face-to-face by going into student dormitories, laboratories and libraries to collect and understand the real situation of young students’ thought dynamics, life needs, and aspirations for growth and development, so as to promptly understand what young people are thinking about and guiding them well.

Agenda setting is the process by which the news media directs what the audience sees and thinks by setting the topics. For different audiences, the world they are exposed to through media reports is different. With the advent of the Internet era, the passive state of the audience to receive information no longer exists, and all Internet users can become the active release of information and expression of opinions “media”. The theory of “spiral of silence” in communication science refers to the fact that in the process of communication, a minority of people will intentionally cater to the dominant opinions of the group in order to avoid being isolated. The existence of more and more such a minority of silent individuals will lead to the dominant voice to occupy all the public opinion field very quickly, and the comments forwarded or emphasized by the media or opinion leaders are often more likely to evolve into the dominant opinion recognized by the majority of the people, and this opinion can be shaped. Therefore, although the Internet is full of different voices, online opinion leaders can utilize the number of fans and influence they have accumulated to selectively release specific topics and express their opinions to the audience, so the ideological direction of young college students can be driven by the use of agenda-setting by opinion leaders to create public opinion hotspots and attract the audience’s attention.

Creating a cooperative community of “party and government, public opinion leaders, and university students and netizens” will help multiple subjects to consult and communicate, and collaborate in common governance, laying a solid foundation for online ideological security. Parties and government departments need to take the main responsibility of monitoring and supervising the development of the Internet, and forming positive online public opinion. Promote rational dialogue between the public and the government. At the same time, they should fully play their positive role in encouraging online opinion leaders to analyze government decisions and hot events positively and guide positive social opinions. Correctly view the subjective initiative of college students and give full play to their strengths, so as to build a positive and healthy online public opinion environment among young college students and promote the establishment of an online ideological security team. By relying on the party and government departments to strengthen network ideological security, and mobilizing the enthusiasm of network opinion leaders and college students, we can form a situation in which the party and government, opinion leaders and college students work together to modernize the network ideological security governance system and governance capacity, and firmly grasp the network ideological security governance system and governance capacity. Only by mobilizing the active participation of online opinion leaders and college students can we form a situation where “party and government, opinion leaders and college students” work together to build a modernized network ideological security governance system and governance capacity, and firmly grasp the dominant power of network ideological work.