Social public opinion hyper-network key public opinion element identification method based on structural hole

Abstract

The invention is suitable for the field of social network analysis, and provides a social public opinion hypernetwork key public opinion element identification method based on a structural hole, and the method comprises the steps: collecting public opinion event related data; constructing a social public opinion hypernetwork model of five layers of subnets; constructing a conversion hypergraph, and constructing a conversion line graph according to the conversion hypergraph; and calculating the sum of scores of all hyperedges formed by node participation, and calculating the final score of the node. According to the method, the development situation of public opinion propagation is researched by utilizing a multilayer super network model, a structural hole theory is introduced into a method for identifying public opinion elements, and a conventional hyperedge sorting method is improved to a certain extent; compared with the traditional method, the method provided by the invention can effectively and accurately identify various key public opinion elements in the social public opinion super network, and can analyze by combining with the node importance index of the super network, so that the key public opinion information can be interpreted from multiple dimensions and multiple perspectives.

Description

Technical Field

[0001] The present invention belongs to the field of social network analysis, and in particular relates to a method for identifying key public opinion elements of a social public opinion super network based on structural holes. Background Art

[0002] The management and control of public opinion is an important part of ensuring social security. As a key carrier reflecting public sentiment and social hot spots, the nodes of the social public opinion network present diversified characteristics, including users, core content, release time, emotional tendencies, and topic participation. How to identify hot topics and predict the development trend of public opinion events has become the focus of current public opinion research.

[0003] Most of the previous studies on identifying key nodes of structural holes are based on the application of complex networks, such as the application of structural hole theory in social networks. Relevant research in hypernetworks is relatively scarce. Complex networks usually use nodes and edges to represent the relationships in the system. Taking the identification of important nodes in social networks as an example, traditional research usually uses complex networks to represent them, but this often ignores the group relationships on social networks, as well as the relationships between groups. The nodes identified based on this will have errors and are not enough to prove the importance of the node. Summary of the invention

[0004] The purpose of the present invention is to provide a method for identifying key public opinion elements of a social public opinion super network based on structural holes, aiming to solve the technical problems existing in the prior art identified in the background technology.

[0005] The present invention is implemented by a method for identifying key public opinion elements of a social public opinion super network based on structural holes, the method comprising:

[0006] Collect data related to public opinion events based on social platforms;

[0007] Constructing a social public opinion super network model with five subnets, wherein the five subnets include: a social subnet, a content subnet, a time sequence subnet, an emotion subnet and a topic subnet;

[0008] Construct a transformation hypergraph, and construct a transformation line graph based on the transformation hypergraph;

[0009] Calculate the constraint coefficient of the hyperedge, identify the key hyperedges and key nodes, and combine the three indicators of local, positional, and global to calculate the total score of all hyperedges formed by the node, and use the Euclidean distance formula to calculate the final score of the node.

[0010] As a further solution of the present invention, the social public opinion super network model with five sub-networks is constructed, wherein each piece of content published by a social platform corresponds to a super edge, and the super edge SE o Indicates the subject of public opinion a i At time tz External forces and psychological j Under the internal force of n and core content m , and based on this, a super network model of social public opinion is constructed.

[0011] As a further solution of the present invention, the five-layer subnet, wherein:

[0012] The social subnet represents the reply relationship between the opinion subjects participating in the discussion in the social opinion;

[0013] The content subnet represents the dissemination process of information released by the public opinion subject, where a core content corresponds to a node in the content subnet, and the information dissemination influence is expressed as I(e i )=R(e i )×D(e i );

[0014] The temporal subnetwork represents the different stages and degrees of the development of public opinion events, and the temporal similarity is expressed as

[0015] The emotional subnet represents the psychological type of the opinion subject when publishing the speech, and there is a conversion correlation between different psychological types. The psychological conversion correlation is expressed as

[0016] The topic subnet represents the topics involved in the content posted by netizens, and the topic similarity is expressed as

[0017] As a further solution of the present invention, the transformation hypergraph is constructed, a transformation line graph is constructed according to the transformation hypergraph, and constraint coefficients of hyperedges are calculated to identify key hyperedges and key nodes;

[0018] S1 obtains the transformation hypergraph of the social public opinion hypernetwork model based on the model;

[0019] S2 constructs a transformation line graph from the transformation hypergraph, taking into account the connectivity of the strengthening nodes and the contribution of each node to information dissemination, and takes the number of common nodes between adjacent hyperedges in the transformation hypergraph as the weight of the edge in the transformation line graph;

[0020] S3 uses the structural hole constraint coefficient formula in the hypernetwork to calculate the constraint coefficient HyperC of each hyperedge. The calculation formula is:

[0021] HyperC(i)=∑ j∈Γ(i) (w ij +∑ q w iq w qj ) 2 ;

[0022] Among them, w ij is the number of common nodes between hyperedges i and j, j is the neighbor of hyperedge i, and q is the common neighbor between hyperedges i and j;

[0023] S4 identifies the key hyperedges HBurt_Hyperedge and key nodes HBurt_Node in the social public opinion hypernetwork based on the HyperC(i) value.

[0024] As a further solution of the present invention, the transformation hypergraph is composed of nodes and hyperedges obtained by superimposing subnets of various layers.

[0025] As a further solution of the present invention, each edge of the transformation hypergraph contains five nodes, wherein SE o ={a i , t z , p j , k n , c m}.

[0026] As a further solution of the present invention, the key hyperedge HBurt_Hyperedge and key node HBurt_Node in the social public opinion hypernetwork are identified, wherein the key hyperedge identification includes:

[0027] A time-series subnetwork is added to the basic model of the social public opinion hypernetwork, and an improved hyperedge sorting algorithm is generated based on the time-series similarity index. The algorithm formula is:

[0028]

[0029] Where m is the total number of hyperedges, is the influence of information dissemination, p ij Indicates psychological i With psychological p j The relationship between ij Represents super edge SE i With super edge SE j The temporal similarity between ij Represents hyperedge Sim i With Super Edge Sim j Similarity of topics included; L(SE j ) represents the super edge SE j The hyperedge connectivity of

[0030] The structural hole constraint coefficient formula is combined with the hyperedge sorting algorithm formula to calculate the hyperedge SE in the hypernetwork. i The score is recorded as:

[0031] HBurt_HyperedgeScore(SEi )=αSuperedgerank(SE i )+βHyperC(i);

[0032] Among them, α and β are balance coefficients, and α+β=1.

[0033] As a further solution of the present invention, the key hyperedge HBurt_Hyperedge and key node HBurt_Node in the social public opinion hypernetwork are identified, wherein the key node identification includes:

[0034] Calculate the sum of the scores of all hyperedges formed by node i, and use the Euclidean distance formula to calculate its value with the nuclear degree centrality to determine the final score of the node HBurt_NodeScore, specifically:

[0035]

[0036] in, represents the core centrality of node i, HBurt_HyperedgeScore(SE i ) is the score of the hyperedge formed by node i.

[0037] The beneficial effects of the present invention are:

[0038] This method uses a multi-layer hypernetwork model to study the development trend of public opinion dissemination, and introduces the structural hole theory in the method of identifying public opinion elements and makes certain improvements to the previous hyperedge sorting method. Compared with previous traditional methods, the method proposed in this application can effectively and accurately identify a variety of key public opinion elements in the social public opinion hypernetwork, and analyze them in combination with the hypernetwork node importance index, which can interpret key public opinion information from multiple dimensions and perspectives;

[0039] (1) Five-layer hypernetwork model: This model can analyze the system from different levels and capture the interactions and influences between the layers. Each layer can represent a specific relationship or attribute, making the understanding of the system more comprehensive. By using hyperedges, the model can naturally represent multivariate relationships, making it possible to better simulate complex interactions in the real world and suitable for studying the dynamic processes of complex systems.

[0040] (2) Structural Hole Theory: Structural hole theory emphasizes the identification of key nodes that connect different groups in the network. By introducing this theory, we can more effectively identify the nodes that play a bridging role in the five-layer hypernetwork, which may play an important role in information dissemination and resource acquisition;

[0041] (3) Improved hyperedge sorting algorithm: The improved algorithm not only considers the information attributes contained in different hyperedges, but also considers the impact of the topological relationship between neighbors on the importance of hyperedges, which increases the accuracy of hyperedge identification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A diagram of the social public opinion super-network model architecture provided by an embodiment of the present invention;

[0043] Figure 2 A social public opinion conversion hypergraph provided by an embodiment of the present invention;

[0044] Figure 3 A social public opinion conversion line graph provided by an embodiment of the present invention;

[0045] Figure 4 A comparison diagram of sentiment similarity and topic similarity of some hyperedges provided in an embodiment of the present invention;

[0046] Figure 5 A graph showing the calculation results of the hyperedge ranking scores provided in an embodiment of the present invention;

[0047] Figure 6 A comparison diagram of node propagation influence based on the RP strategy provided in an embodiment of the present invention;

[0048] Figure 7 This is a diagram of opinion leader identification and evaluation results provided by an embodiment of the present invention; DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0050] A method for identifying key public opinion elements of a social public opinion super network based on structural holes, the method comprising:

[0051] Collect data related to public opinion events based on social platforms;

[0052] Constructing a social public opinion super network model with five subnets, wherein the five subnets include: a social subnet, a content subnet, a time sequence subnet, an emotion subnet and a topic subnet;

[0053] Construct a transformation hypergraph, and construct a transformation line graph based on the transformation hypergraph;

[0054] Calculate the constraint coefficient of the hyperedge, identify the key hyperedges and key nodes, and combine the three indicators of local, positional, and global to calculate the total score of all hyperedges formed by the node, and use the Euclidean distance formula to calculate the final score of the node.

[0055] like Figure 1 As shown in the figure, it is the architecture diagram of the social public opinion super network model. This scheme divides the social public opinion super network into five subnets: social subnet, content subnet, time subnet, emotion subnet and topic subnet. Each piece of content published on a social platform (such as Weibo) corresponds to a hyperedge. o Indicates the subject of public opinion a i At time t z External forces and psychological j Under the internal force of n and core content m Based on this, a social public opinion super network model is constructed. The social public opinion super network model and its sub-network attributes are shown in Figure 1 As shown, hyperedges are represented by solid lines and connections between subnetworks are represented by dotted lines.

[0056] The five-layer subnet, wherein:

[0057] The social subnet represents the reply relationship between the opinion subjects participating in the discussion in the social opinion;

[0058] The content subnet represents the dissemination process of information released by the public opinion subject, where a core content corresponds to a node in the content subnet, and the information dissemination influence is expressed as I(e i )=R(e i )×D(e i );

[0059] Among them, I(e i ) represents the influence of information dissemination;

[0060] R(e i ) is the breadth of information dissemination, that is, the scope of information dissemination, through information e i The total number of occurrences and total number of hyperedges in the online opinion hypernetwork model are obtained, that is, e i The ratio of the number of hyperedges participating in the composition to the total number of hyperedges in the model;

[0061] D(e i ) indicates the depth of information dissemination, that is, the depth to which information is disseminated by the subject of public opinion. i The total number of times it appears in the hyperedge and the number of times it is affected by the information e i Obtain the number of subjects in the affected social subnetwork;

[0062] The temporal subnetwork represents the different stages and degrees of the development of public opinion events. According to the amount of public opinion information and the life cycle theory, the evolution of social public opinion can be divided into four stages: the incubation period, the outbreak period, the continuous period, and the recurrence period. The temporal similarity is expressed as

[0063] Among them, SE i , SE j denote the i-th and j-th hyperedges respectively;

[0064] For row i and row j :

[0065] The time series subnet contains four time nodes T1, T2, T3, and T4. The matrix C is used to indicate whether the time of posting opinions is at the same time node. C = C ij ,in:

[0066]

[0067] Then row i Represents super edge SE i The temporal relationship of row i =(C i，1 , C i，2 , …, C i，m )(i=1,2,…,m);

[0068] row j Represents super edge SE j The temporal relationship of row j =(C j，1 , C j，2 , …, C j，m )(j=1,2,…,m);

[0069]

[0070] The emotional subnet represents the psychological type of the opinion subject when publishing the speech, and there is a conversion correlation between different psychological types. The psychological conversion correlation is expressed as

[0071] Among them, sign(x) is the sign function,

[0072] When x≥0, sign(x)=1, when x<0, sign(x)=-1;

[0073] p i 、p j denote psychological type i and psychological type j respectively;

[0074] The topic subnet represents the topics involved in the content posted by netizens, and the topic similarity is expressed as

[0075]

[0076] Among them, w 1j 、w 2jare the weight values ​​of the j-th keyword of hyperedge SE1 and hyperedge SE2 respectively, 1≤j≤m.

[0077] 4. The method according to claim 3, characterized in that the transformation hypergraph is constructed, a transformation line graph is constructed according to the transformation hypergraph, and constraint coefficients of hyperedges are calculated to identify key hyperedges and key nodes;

[0078] S1 obtains the transformation hypergraph of the social public opinion hypernetwork model, such as Figure 2 As shown, each edge of the transformation hypergraph contains five nodes, among which SE o ={a i , t z , p j , k n , c m}, SE o represents a hyperedge, a i represents the user node in the social subnetwork, t z represents the time node in the timing subnet, p j represents the psychological node in the emotion subnet, k n represents the topic node in the topic subnet, c m Represents an information node in the content subnet;

[0079] S2, such as Figure 3 As shown, the transformation line graph is constructed from the transformation hypergraph, taking into account the connectivity of the strengthening nodes and the contribution of each node to the information propagation, and the number of common nodes between adjacent hyperedges in the transformation hypergraph is used as the weight of the edge in the transformation line graph;

[0080] S3 uses the structural hole constraint coefficient formula in the hypernetwork to calculate the constraint coefficient HyperC of each hyperedge. The calculation formula is:

[0081] HyperC(i)=∑ j∈Γ(i) (w ij +∑ q w iq w qj ) 2 ;

[0082] Among them, w ij is the number of common nodes between hyperedges i and j, j is the neighbor of hyperedge i, q is the common neighbor between hyperedges i and j, w iq is the number of common nodes between hyperedges i and q, w qj is the number of common nodes between hyperedges q and j, Γ(i) is the set of neighbor nodes of node i;

[0083] When the constraint coefficient of a hyperedge is low, it means that the number of neighbors of the hyperedge is small and the degree of closure between these neighbors is high. Such hyperedges find it difficult to obtain new relationship resources, so they are often at a disadvantage in competition. On the contrary, as the constraint coefficient of a hyperedge increases, it is more conducive to forming structural holes and obtaining new relationship resources from them.

[0084] S4 identifies the key hyperedges HBurt_Hyperedge and key nodes HBurt_Node in the social public opinion hypernetwork based on the HyperC(i) value.

[0085] The transformation hypergraph is composed of nodes and hyperedges formed by superimposing subnetworks at each layer.

[0086] Key hyperedge identification includes:

[0087] According to the "golden 4 hours" principle, timely intervention of key microblog information in social public opinion is of great significance. Considering that some users fail to follow up on events in a timely manner, or review and mention the early development of events at a later time point, the influence of time sequence is ignored. Therefore, a time sequence subnet is added to the basic model of social public opinion hypernetwork, and an improved hyperedge sorting algorithm is generated based on the time sequence similarity index. The algorithm formula is:

[0088]

[0089] Where m is the total number of hyperedges, is the influence of information dissemination, p ij Indicates psychological i With psychological p j The relationship between ij represents the temporal similarity between node i and node j, Sim ij Represents hyperedge Sim i With Super Edge Sim j Similarity of topics included; L(SE j ) represents the super edge SE j The hyperedge connectivity of

[0090] Most of the existing hyperedge sorting algorithms only consider the information attributes contained in different hyperedges, ignoring the impact of the topological relationship between neighbors on the importance of hyperedges. To address this problem, the structural hole constraint coefficient formula is combined with the hyperedge sorting algorithm formula to calculate the SE of hyperedges in the hypernetwork. i The score is recorded as:

[0091] HBurt_HyperedgeScore(SE i )=αSuperedgerank(SE i )+βHyperC(i);

[0092] Among them, α and β are balance coefficients, and α+β=1.

[0093] Key node identification includes:

[0094] Identifying opinion leaders in social public opinion can timely intervene and guide the direction of public opinion, and analyzing core emotions and focus topics ensures a multi-dimensional understanding and evaluation of events. Local indicators, location indicators, and global indicators of the network all play an important role in identifying key nodes, but each indicator has a different focus. In this regard, a key node identification method is proposed, which comprehensively considers different types of indicators:

[0095] Calculate the sum of the scores of all hyperedges formed by node i, and use the Euclidean distance formula to calculate its value with the nuclear degree centrality to determine the final score of the node HBurt_NodeScore, specifically:

[0096]

[0097] in, represents the core centrality of node i, HBurt_HyperedgeScore(SE i ) is the score of the hyperedge formed by node i.

[0098] Example:

[0099] 1. Case Introduction

[0100] Regarding the case A that has been fermenting on the Internet, considering that the public opinion information of the incident contains discussions from multiple perspectives, covering different public attitudes and evaluations, and can ensure the diversity of data, this paper chooses this incident for empirical analysis. This paper crawls the Weibo data of a certain period of time under specific keywords and conducts simulation analysis. The Weibo data is crawled with the case A as the keyword, and Jieba word segmentation is used to segment the effective comments, and high-frequency words, punctuation marks, numbers, and letters with meaningless connotations are removed, and 1,224 effective Weibo posts are obtained.

[0101] 2. Emotional subnet quantification results

[0102] The crawled microblog texts are preprocessed by removing invalid symbols and stop words to obtain candidate information texts. The Dalian University of Technology sentiment vocabulary database is used to score the sentiment of the collected microblog texts and divide them into 5 sentiment types, where the sentiment type is recorded as p i :

[0103] (1)p i <-10, indicating that the emotional tendency is extremely negative and the psychological intensity is |p i |, the sentiment type is recorded as p1.

[0104] (2) -10<p i <0, indicating that the emotional tendency is negative and the psychological intensity is |p i |, the sentiment type is recorded as p2.

[0105] (3)p i =0, indicating that the emotional tendency is neutral and the psychological intensity is |p i |, the sentiment type is recorded as p3.

[0106] (4)0<p i <10, indicating that the emotional tendency is positive and the psychological intensity is |p i |, the sentiment type is recorded as p4.

[0107] (5)p i >10, indicating that the emotional tendency is extremely positive and the psychological intensity is |p i |, the sentiment type is recorded as p5.

[0108] Some of the emotion scoring results are shown in Table 1.

[0109] Table 1. Some sentiment scoring results

[0110]

[0111] 3. Topic subnet quantification results

[0112] LDA is used to calculate the consistency score and topic word spacing of microblog texts with 2-20 topics to determine the optimal number of topics for this model. According to the consistency score and topic word distribution, this paper determines that the number of topics for this event is 9. Table 2 shows the topic subnet keyword extraction results obtained after data processing:

[0113] Table 2. Topic quantification results

[0114]

[0115]

[0116] 4. Quantification results of social public opinion super network

[0117] Users who participated in the public opinion discussion during the entire development process of the incident were selected as the research objects, and the public opinion subjects and their corresponding content, time sequence, emotion, and topic were constructed respectively. On this basis, a social public opinion super network model was established, in which the public opinion subjects published a total of 9 topics, involving 5 core contents, 5 psychological types and 4 time periods. Each hyperedge only contains one public opinion subject, one core content, one emotional type, one topic and one time node. The quantitative results of some super networks are shown in Table 3.

[0118] Table 3. Quantization results of some hypernetworks

[0119]

[0120] 5. Hyperedge sorting calculation results

[0121] According to the steps of the hyperedge sorting algorithm, the sentiment similarity and hyperedge topic similarity are calculated respectively. The specific calculation results are shown in Figure 4 .

[0122] Based on the calculation results, the experiment uses the structural holes of the hypernetwork, the classic hyperedge ranking algorithms SuperEdgeRank, HyperEdgeRank, the improved SuperEdgeRank and HBurt_Hyperedge to identify important hyperedges. The top 5 results after normalization are SE 1099 , SE 972 , SE 1038 , SE 1192 , SE 1156 Since the SuperEdgeRank method only considers the properties of the information contained between different hyperedges, SE 1099 and SE 972 , SE 1038 and SE 1192 The scores are equal; HyperEdgeRank and the improved SuperEdgeRank also have the same problem, making SE 1099 , SE 972 and SE 1038 , SE 1192 The scores are equal. 972 , SE 1038 They occupy similar positions in the network structure, resulting in similar scores for their structural holes. However, HBurt_Hyperedge comprehensively considers the topological relationship between neighbors and the information attributes between different hyperedges, and the scores of the top 5 rankings are different. Figure 5 .

[0123] 6. Analysis of network communication dynamics

[0124] The accuracy of the proposed model is further demonstrated from the perspective of network propagation dynamics. The classic Susceptible-Infected-Susceptible (SIS) and Susceptible-Infected-Recovered (SIR) information propagation models are used to analyze the network. The SIS model contains two types of nodes: S state (susceptible state) and I state (infected state). The following formula shows the rate of change of the number of individuals in different states over time:

[0125]

[0126] Among them, β and γ represent the probability that an s-state (susceptible state) individual is infected to the I-state (infected state) and the probability that an I-state individual recovers to the S-state, respectively. s(t) represents the proportion of s-state individuals at time t, and i(t) represents the proportion of I-state individuals at time t.

[0127] The SIR model contains three types of nodes: S state (susceptible state), I state (infected state) and R state (recovered state). The following formula shows the rate of change of the number of individuals in different states over time:

[0128]

[0129] Among them, s(t), i(t) and γ(t) represent the proportions of individuals in the S state, I state and R state in the network at time t, β represents the probability of an S state individual being infected by an I state individual, and γ represents the probability of an I state transforming into an R state.

[0130] Based on the above information model, this paper selects the RP (Reactive Process) strategy in the information propagation process. The RP strategy refers to propagating information to all nodes in the hypernetwork, presenting global propagation, that is, a certain proportion of nodes are randomly selected as infected nodes at the beginning. At each time step, the I-state node transmits information to all neighboring nodes in the S-state in the hyperedge with a certain probability, and recovers to the S-state with a certain probability. Repeat this step until there are no I-state nodes in the network.

[0131] The experiment uses 6 different influential node identification methods and 2 propagation models to analyze the propagation of social public opinion super network. To ensure that enough data points are collected before reaching the steady state, α is set to 0.5 and β is set to 0.5. Figure 6 The figure shows the comparison of node propagation influence under different propagation models.

[0132] Figure 6 The transformation line diagram of the social public opinion hypernetwork is based on the propagation results of different models. By comparing the propagation of different methods in various network models, it can be clearly seen that the seed nodes selected by the method in this paper cause the proportion of infected nodes to increase rapidly in the early stage of propagation. In each time step before the steady state, the proportion of infected nodes remains at a high level and reaches the steady state first. This shows that the high-influence hyperedges identified in this paper propagate the fastest. Through the hypernetwork propagation dynamics experiment, the effectiveness of this method in identifying high-influence hyperedges is verified.

[0133] 7.AUC evaluation indicator

[0134] AUC (Area Under the Curve) is a common indicator for measuring the accuracy of link prediction algorithms, which represents the area under the ROC (Receiver Operating Characteristic) curve. Different algorithms are used to rank the importance of hyperedges in the social public opinion hypernetwork. The more ranking values ​​a hyperedge is assigned, the clearer the importance of the hyperedge identified by the algorithm. This paper compares a variety of hyperedge importance evaluation algorithms, including SuperEdgeRank, structural holes in hypernetworks, improved SuperEdgeRank, HyperEdgeRank, and HBurt_Hyperedge. Table 4 shows the specific AUC values ​​of each algorithm:

[0135] Table 4. AUC metrics

[0136]

[0137]

[0138] From Table 4, we can observe that the AUC index of HBurt_Hyperedge is higher than that of other algorithms.

[0139] 8. Calculation results of hypernetwork attribute indicators

[0140] The calculation of social public opinion key node identification is carried out according to the hypernetwork node importance index. Node hyperdegree is used to identify active figures in the hypernetwork; hyperedge connectivity is the basis for identifying focal figures; hyperedge overlap plays an important role in identifying communication figures; and proximity centrality is closely related to identifying potential active figures, as shown in Table 5:

[0141] Table 5. Identification results of the top three key nodes ranked by hypernetwork attribute indicators

[0142]

[0143] Using the hypernetwork node importance index to mine various types of nodes, they all show different roles and influences. For example, more active users such as "B" can play a positive role in public opinion events, but they may also bring risks. In order to better manage and guide these users, we should consider building communication bridges and feedback mechanisms to make them beneficial members of the platform. For example, focus users such as "D" have a significant impact on the public, so it is particularly important to establish effective communication with them. After the outbreak of public opinion, the focus of users can be monitored in real time by supervising the content published by focus users. For example, communication users such as "C" should be provided with more information resources to make decisions and express opinions after formulating standardized user behavior guidelines. "A" has multiple identities. He is not only an active person, but also the focus of the public, the core of information dissemination and a potential active source. He is undoubtedly a key person. Since the content published by key people has a wide range of dissemination and a large audience, his speech should be continuously monitored and managed, and timely guidance and intervention should be made to effectively control the spread of public opinion. According to the role played by different types of key users, different strategies and methods for managing and monitoring public opinion can be set.

[0144] 9. Key node identification results and verification

[0145] Based on the "Case A" event, the HBurt_Hyperedge method was used to sort all the hyperedges in the model to identify opinion leaders. Five opinion leaders were identified from 916 participants, and the normalized results are shown in Table 6:

[0146] Table 6. Ranking of opinion leaders

[0147]

[0148] In order to quantitatively evaluate the effectiveness of the proposed method, this paper uses the opinion leader identification evaluation index, which is defined as:

[0149]

[0150] Among them, SO i Indicates the isolation of opinion subject a i The hyperedge overlap of the newly formed hypernetwork after the hyperedge; To isolate the subject of public opinion i The shortest distance between the hyperedges of the newly formed hypernetwork after the hyperedge where it is located; E' i The higher the value, the i More likely to be an opinion leader. i For opinion leaders, the network structure becomes loose after isolation, and the hyperedge overlap decreases (i.e., 1-SO i The shortest distance between the super edges increases.

[0151] The experimental results show that the HBurt_Node method is consistent with the opinion leader identification and evaluation results, thus verifying the effectiveness of the method. Figure 7 .

[0152] The HBurt_Node method is used to calculate the scores of each layer of subnets, from which the identification results of the key nodes of each layer of subnets can be obtained, as shown in Table 7:

[0153] Table 7 Key node identification results of each layer subnet

[0154]

[0155] According to the method proposed in this paper, the key nodes of each layer of subnetwork are mined. In the social subnetwork, according to the node scoring calculation formula, the identified opinion leader is user "A". This shows that the public opinion content he released has a wide range of dissemination, involving many social groups and topic cycles. At the same time, this user holds similar views to many users and has a high influence in the entire event. The information he releases promotes in-depth communication and collision of views among users, and can arouse emotional resonance among most users. In the topic subnetwork, the key node mined is the focus topic. Among them, the node score of the focus topic is 0.35, which means that under the consideration of multiple indicator dimensions, the topic node "l9" has become the focus of public discussion and has the ability to guide public opinion and deepen discussion. The extraction of focus topics can provide valuable information for decision makers, help understand the needs of the public, and improve the interaction effect with information recipients. The mined emotional nodes are core emotions, and the emotional node "neutral" has the highest score, indicating that this emotion is highly consistent with the public emotion of the entire event, effectively enhancing the scope and acceptance of information dissemination, and also promoting the cohesion and participation motivation of the community. Therefore, "neutral" is defined as the core emotion of the event. The extraction of its nodes helps decision makers accurately capture the changes in public emotions, which in turn has a positive impact on the trend of public opinion. In addition, in the process of public opinion evolution, key time nodes mark important stages in the development of public opinion. By monitoring the peak moments and sub-peak moments in the development of public opinion, it can help identify the time points of public opinion fermentation and surge, thereby providing useful suggestions for understanding the development and changes of public opinion and real-time monitoring of public opinion.

[0156] This scheme applies the hypernetwork theory to the process of social public opinion propagation, and proposes a method for identifying key public opinion elements in the social public opinion hypernetwork based on structural holes to identify key nodes and hyperedges in social public opinion. First, a social public opinion hypernetwork model is constructed, including social subnetwork, content subnetwork, temporal subnetwork, emotional subnetwork and topic subnetwork. Secondly, by quantifying the information propagation influence, temporal similarity and emotional similarity, the correlation between hyperedges in the hypernetwork model is calculated, and an improved hyperedge sorting method is designed. The structural hole theory is introduced into the model to identify key hyperedges in combination with the hyperedge sorting method. Finally, data related to "Case A" are obtained through the Weibo platform for empirical research. The experimental results show that compared with the existing methods, the proposed method performs better in the discrimination of hyperedge sorting scores; in the SIS and SIR hypernetwork propagation models based on the RP strategy, the proposed method has a faster propagation speed and a shorter steady-state reaching time; the accuracy of the identification results is verified by the AUC indicator, and the results show that the AUC values ​​of the proposed scheme are better than those of the existing hyperedge sorting methods. The results of multiple experiments show that the proposed method can effectively identify multiple key public opinion elements in the social public opinion super network, and analyze them in combination with the importance index of the super network nodes, interpreting key public opinion information from multiple dimensions and perspectives. This helps to enhance the in-depth understanding of public opinion events and provides new research perspectives and practical ideas for relevant scholars and public opinion managers.

[0157] Future research can apply this method to a wider range of online public opinion events to verify its universality, and further explore corresponding intervention and guidance strategies based on the social public opinion super-network model.

[0158] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0159] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0160] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying key public opinion elements of social public opinion super-network based on structural holes, characterized by: The method comprises: Collect data related to public opinion events based on social platforms; Constructing a social public opinion super network model with five subnets, wherein the five subnets include: a social subnet, a content subnet, a time sequence subnet, an emotion subnet and a topic subnet; Construct a transformation hypergraph, and construct a transformation line graph based on the transformation hypergraph; Calculate the constraint coefficient of the hyperedge, identify the key hyperedges and key nodes, and combine the three indicators of local, positional, and global to calculate the total score of all hyperedges formed by the node, and use the Euclidean distance formula to calculate the final score of the node.

2. The method according to claim 1, characterized in that The social public opinion super network model with five sub-networks is constructed, in which each piece of content published on a social platform corresponds to a super edge. o Indicates the subject of public opinion a i At time t z External forces and psychological j Under the internal force of n and core content m , and based on this, a super network model of social public opinion is constructed.

3. The method according to claim 1, characterized in that The five-layer subnet, wherein: The social subnet represents the reply relationship between the opinion subjects participating in the discussion in the social opinion; The content subnet represents the dissemination process of information released by the public opinion subject, where a core content corresponds to a node in the content subnet, and the information dissemination influence is expressed as I(e i )=R(e i )×D(e i ); Among them, I(e i ) represents the influence of information dissemination; R(e i ) is the breadth of information dissemination, that is, the scope of information dissemination, through information e i The total number of occurrences and total number of hyperedges in the online opinion hypernetwork model are obtained, that is, e i The ratio of the number of hyperedges participating in the composition to the total number of hyperedges in the model; D(e i ) indicates the depth of information dissemination, that is, the depth to which information is disseminated by the subject of public opinion. i The total number of times it appears in the hyperedge and the number of times it is affected by the information e i Obtain the number of subjects in the affected social subnetwork; The temporal subnetwork represents the different stages and degrees of the development of public opinion events, and the temporal similarity is expressed as Among them, SE i , SE j denote the i-th and j-th hyperedges respectively; For row i and row j : The time series subnet contains four time nodes T1, T2, T3, and T4. The matrix C is used to indicate whether the time of posting opinions is at the same time node. C = C ij ,in: Then row i Represents super edge SE i The temporal relationship of row i =(C i,1 ,C i,2 ,…,C i,m )(i=1,2,…,m); row j Represents super edge SE j The temporal relationship of row j =(C j,1 ,C j,2 ,…,C j,m )(j=1,2,…,m); The emotional subnet represents the psychological type of the opinion subject when publishing the speech, and there is a conversion correlation between different psychological types. The psychological conversion correlation is expressed as Among them, sign(x) is the sign function, When x ≥ 0, sign(x) = 1, when x < 0, sign(x) = -1; p i 、p j denote psychological type i and psychological type j respectively; The topic subnet represents the topics involved in the content posted by netizens, and the topic similarity is expressed as Among them, w 1j 、w 2j are the weight values ​​of the j-th keyword of hyperedge SE1 and hyperedge SE2 respectively, 1≤j≤m.

4. The method according to claim 3, characterized in that The transformation hypergraph is constructed, a transformation line graph is constructed according to the transformation hypergraph, and constraint coefficients of hyperedges are calculated to identify key hyperedges and key nodes; S1 obtains the transformation hypergraph of the social public opinion hypernetwork model based on the model; S2 constructs a transformation line graph from the transformation hypergraph, taking into account the connectivity of the strengthening nodes and the contribution of each node to information dissemination, and takes the number of common nodes between adjacent hyperedges in the transformation hypergraph as the weight of the edge in the transformation line graph; S3 uses the structural hole constraint coefficient formula in the hypernetwork to calculate the constraint coefficient HyperC of each hyperedge. The calculation formula is: HyperC(i)=∑ j∈Γ(i) (w ij +∑ q w iq w qj ) 2 ; Among them, w ij is the number of common nodes between hyperedges i and j, j is the neighbor of hyperedge i, q is the common neighbor between hyperedges i and j, w iq is the number of common nodes between hyperedges i and q, w qj is the number of common nodes between hyperedges q and j, Γ(i) is the set of neighbor nodes of node i; S4 identifies the key hyperedges HBurt_Hyperedge and key nodes HBurt_Node in the social public opinion hypernetwork based on the HyperC(i) value.

5. The method according to claim 4, characterized in that The transformation hypergraph is composed of nodes and hyperedges formed by superimposing subnetworks at each layer.

6. The method according to claim 3, characterized in that Each edge of the transformation hypergraph contains five nodes, among which SE o ={a i ,t z ,p j ,k n ,c m }, SE o represents a hyperedge, a i represents the user node in the social subnetwork, t z represents the time node in the timing subnet, p j represents the psychological node in the emotion subnet, k n represents the topic node in the topic subnet, c m Represents an information node in a content subnet.

7. The method according to claim 4, characterized in that The identification of key hyperedges HBurt_Hyperedge and key nodes HBurt_Node in the social public opinion hypernetwork, wherein the key hyperedge identification includes: A time-series subnetwork is added to the basic model of the social public opinion hypernetwork, and an improved hyperedge sorting algorithm is generated based on the time-series similarity index. The algorithm formula is: Where m is the total number of hyperedges, is the influence of information dissemination, p ij Indicates psychological i With psychological p j The relationship between ij represents the temporal similarity between node i and node j, Sim ij Represents hyperedge Sim i With Super Edge Sim j Similarity of topics included; L(SE j ) represents the super edge SE j The hyperedge connectivity of The structural hole constraint coefficient formula is combined with the hyperedge sorting algorithm formula to calculate the hyperedge SE in the hypernetwork. i The score is recorded as: HBurt_HyperedgeScore(SE i )=αSuperedgerank(SE i )+βHyperC(i); Among them, α and β are balance coefficients, and α+β=1.

8. The method according to claim 7, characterized in that The step of identifying key hyperedges HBurt_Hyperedge and key nodes HBurt_Node in the social public opinion hypernetwork includes: Calculate the sum of the scores of all hyperedges formed by node i, and use the Euclidean distance formula to calculate its value with the nuclear degree centrality to determine the final score of the node HBurt_NodeScore, specifically: in, represents the core centrality of node i, HBurt_HyperedgeScore(SE i ) is the score of the hyperedge formed by node i.

Images