Intelligent analysis system of network public opinion based on data mining

Abstract

The application provides a network public opinion intelligent analysis system based on data mining, relates to the technical field of data mining, and comprises the following modules: a public opinion propagation relationship graph construction module, which collects multiple-source network public opinion related data sets to construct a public opinion propagation relationship graph; a public opinion topology feature vector acquisition module, which extracts topology features from the multiple-source network public opinion related data sets based on the public opinion propagation relationship graph; a public opinion feature representation space acquisition module, which performs public opinion data mining on the public opinion topology feature vector; and a public opinion risk early warning result output module, which performs risk analysis on the public opinion feature representation space and outputs a public opinion risk early warning result by using a public opinion propagation risk index. The application can solve the technical problem that the existing technology cannot accurately predict the diffusion trend and propagation risk of public opinion events, and achieves the technical effect of accurately predicting the diffusion trend, propagation risk and potential crisis of public opinion events by analyzing the topology structure and calculating the propagation node energy.

Description

Technical Field

[0001] This application relates to the field of data mining technology, and in particular to an intelligent analysis system for online public opinion based on data mining. Background Technology

[0002] With the rapid development of online information, the speed and scope of public opinion dissemination are constantly expanding. The spread of public opinion events not only influences the formation of public opinion but also has a profound impact on social stability and corporate reputation. Especially in the era of the internet and social media, the speed of information dissemination far surpasses that of traditional media, making public opinion monitoring and risk warning particularly important. Current technologies in public opinion monitoring systems generally employ methods such as keyword matching and sentiment analysis to filter and analyze online information. However, although these methods can identify public opinion hotspots and sentiment trends to some extent, they still have many shortcomings.

[0003] Currently, most existing public opinion monitoring technologies rely on static public opinion data models, lacking in-depth analysis of the propagation path, information topology, and inter-node relationships. These technologies typically focus only on changes in a single information flow or sentiment, neglecting the complex propagation relationships between various nodes in the public opinion dissemination process. Furthermore, existing technologies often fail to comprehensively consider the topological structure of the propagation path when processing public opinion networks, resulting in the inability to promptly capture the potential spread risks of public opinion events. Because these technologies fail to accurately identify the criticality of propagation nodes, the hierarchical relationships of information diffusion, and changes in propagation energy, they often result in ineffective early warning of public opinion events, thus severely impacting public opinion management and crisis response.

[0004] In summary, existing technologies suffer from a lack of comprehensive analysis of the topology and energy of public opinion propagation nodes, which makes it impossible to accurately predict the spread trend and risks of public opinion events. This further affects the timely early warning and effective intervention of public opinion events, and makes it impossible to achieve in-depth control over the overall and detailed aspects of public opinion propagation. Summary of the Invention

[0005] The purpose of this application is to provide a data mining-based intelligent analysis system for online public opinion, in order to solve the technical problems in the existing technology that, due to the lack of comprehensive analysis of the topology and energy of the propagation nodes of public opinion, it is impossible to accurately predict the spread trend and risk of public opinion events, which further affects the timely early warning and effective intervention of public opinion events, and makes it impossible to achieve in-depth control over the overall and detailed aspects of public opinion propagation.

[0006] In view of the above problems, this application provides a data mining-based intelligent analysis system for online public opinion, including: a public opinion propagation relationship graph construction module, used to collect multi-source online public opinion related datasets to construct a public opinion propagation relationship graph; a public opinion topology feature vector acquisition module, used to extract multi-source public opinion semantic features from the multi-source online public opinion related datasets, and perform topology feature extraction on the multi-source public opinion semantic features based on the public opinion propagation relationship graph to obtain a public opinion topology feature vector; a public opinion feature representation space acquisition module, used to construct a topology constraint data mining model based on a topology consistency constraint threshold, and call the topology constraint data mining model to perform public opinion data mining on the public opinion topology feature vector to obtain a public opinion feature representation space, wherein the topology consistency constraint threshold is obtained through dynamic calculation; and a public opinion risk warning result output module, used to perform risk analysis on the public opinion feature representation space to calculate a public opinion propagation risk index, and use the public opinion propagation risk index to output a public opinion risk warning result.

[0007] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a propagation association information extraction unit, used to analyze the propagation relationship of the multi-source online public opinion related dataset and extract propagation association information, the propagation association information including information release relationship, information forwarding relationship, information comment relationship and information citation relationship; a propagation node set extraction unit, used to identify the propagation subject of the multi-source online public opinion related dataset and extract a propagation node set; a propagation relationship edge set obtaining unit, used to determine the information propagation relationship between nodes in the propagation node set according to the propagation association information and obtain a propagation relationship edge set; and a propagation relationship graph construction unit, used to construct a propagation relationship graph according to the propagation node set and the propagation relationship edge set.

[0008] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a feature calculation unit, used to calculate node propagation centrality features, node clustering coefficient features, and propagation path hierarchy features based on the public opinion propagation relationship graph; a first fused topological feature vector acquisition unit, used to fuse the node propagation centrality features, node clustering coefficient features, and propagation path hierarchy features to obtain a first fused topological feature vector; and a public opinion topological feature vector output unit, used to fuse the first fused topological feature vector based on the multi-source public opinion semantic features to obtain a second fused topological feature vector, and output the second fused topological feature vector as the public opinion topological feature vector.

[0009] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a node topology similarity matrix acquisition unit, used to obtain node propagation structure features based on the public opinion propagation relationship graph, calculate the topology similarity between nodes according to the node propagation structure features, and obtain a node topology similarity matrix; a topology consistency probability distribution function generation unit, used to calculate the topology consistency distribution interval according to the node topology similarity matrix, and generate a topology consistency probability distribution function by performing probability distribution fitting on all similarity values ​​in the topology consistency distribution interval; and a topology consistency constraint threshold determination unit, used to determine the topology consistency constraint threshold according to the topology consistency probability distribution function.

[0010] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a topological adjacency constraint set filtering unit, used to filter topological adjacency constraint sets that are greater than or equal to the topological consistency constraint threshold; and a topological constraint data mining model obtaining unit, used to construct a topological constraint objective function based on the public opinion topological feature vector to obtain a topological constraint data mining model, wherein the topological constraint objective function includes a data mining loss function and a topological consistency constraint term, and the topological consistency constraint term is used to constrain the selection of node pairs with topological consistency relationships from the topological adjacency constraint set.

[0011] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a public opinion feature representation space initialization unit, used to initialize the public opinion feature representation space according to the public opinion topology feature vector; and a public opinion feature representation space acquisition unit, used to iteratively optimize and update the initialized public opinion feature representation space by introducing the topology constraint objective function of the topology constraint data mining model, and to acquire the public opinion feature representation space when the topology constraint objective function reaches a preset number of iterations; wherein each public opinion feature representation vector in the public opinion feature representation space includes public opinion semantic information and public opinion propagation structure information.

[0012] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a public opinion propagation energy calculation unit, used to calculate the public opinion propagation energy of each node in the public opinion propagation relationship graph according to the public opinion feature representation space; a public opinion propagation topology energy field construction unit, used to construct a public opinion propagation topology energy field in the public opinion propagation relationship graph according to the public opinion propagation energy; and a public opinion propagation risk index obtaining unit, used to calculate the public opinion propagation topology energy field based on a preset public opinion propagation energy threshold to obtain a public opinion propagation risk index.

[0013] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a topological similarity acquisition channel, used to recalculate the updated public opinion feature representation vector after each iteration of optimization to obtain the updated topological similarity; a topological consistency distribution interval calculation channel, used to recalculate the topological consistency distribution interval based on the updated topological similarity; and a topological consistency constraint threshold calculation channel, used to recalculate the topological consistency constraint threshold based on the updated topological consistency distribution interval.

[0014] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: a public opinion propagation sub-network identification unit, used to identify multiple public opinion propagation sub-networks according to the public opinion propagation relationship graph; an abnormal propagation structure marking unit, used to calculate the structural consistency index of the multiple public opinion propagation sub-networks and mark the public opinion propagation sub-networks with a consistency threshold less than a preset threshold as abnormal propagation structures; and a public opinion risk warning unit, used to provide public opinion risk warnings for public opinion events corresponding to the abnormal propagation structures.

[0015] Preferably, the data mining-based intelligent analysis system for online public opinion further includes: in the public opinion topology feature vector acquisition module, the multi-source public opinion semantic features include at least one of text sentiment polarity features, text sentiment intensity features, text stance tendency features, text topic distribution features, named entity features, and event trigger word features.

[0016] The technical solution provided in this application has at least the following technical effects or advantages: by realizing the comprehensive analysis of the relationship between nodes, the propagation path and the propagation energy in the public opinion propagation network, it can accurately predict the spread trend, propagation risk and potential crisis of public opinion events through topological structure analysis and propagation node energy calculation, improve the timeliness and accuracy of public opinion monitoring and risk warning, and ensure the technical effect of effective intervention and control in the early stage of public opinion events.

[0017] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the structure of the data mining-based intelligent analysis system for online public opinion in this application.

[0020] Figure 2 This is a flowchart illustrating the data mining-based intelligent analysis system for online public opinion in this application.

[0021] Figure labeling: Module 1 for constructing a public opinion dissemination relationship graph, Module 2 for obtaining public opinion topology feature vectors, Module 3 for obtaining public opinion feature representation space, and Module 4 for outputting public opinion risk warning results. Detailed Implementation

[0022] This application provides a data mining-based intelligent analysis system for online public opinion, addressing the technical problem in existing technologies where the lack of comprehensive analysis of the topology and energy of propagation nodes hinders accurate prediction of the spread and risks of public opinion events. This, in turn, affects timely early warning and effective intervention, preventing in-depth control over both the overall and detailed aspects of public opinion dissemination. The system achieves comprehensive analysis of the relationships between nodes, propagation paths, and propagation energy within the public opinion network. Through topology analysis and propagation node energy calculation, it accurately predicts the spread, risks, and potential crises of public opinion events, improving the timeliness and accuracy of public opinion monitoring and risk warning. This ensures effective intervention and control in the early stages of public opinion events, thereby effectively safeguarding social stability and corporate reputation.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0024] Please see Figure 1 and Figure 2 This application provides a data mining-based intelligent analysis system for online public opinion, specifically including:

[0025] The public opinion dissemination relationship graph construction module 1 is used to collect multi-source online public opinion related datasets to construct a public opinion dissemination relationship graph.

[0026] Furthermore, this application also includes: a propagation association information extraction unit, used to analyze the propagation relationship of the multi-source online public opinion related dataset and extract propagation association information, the propagation association information including information release relationship, information forwarding relationship, information comment relationship and information citation relationship; a propagation node set extraction unit, used to identify the propagation subject of the multi-source online public opinion related dataset and extract the propagation node set; a propagation relationship edge set obtaining unit, used to determine the information propagation relationship between nodes in the propagation node set according to the propagation association information and obtain the propagation relationship edge set; and a propagation relationship graph construction unit, used to construct a propagation relationship graph according to the propagation node set and the propagation relationship edge set.

[0027] Furthermore, this application also includes: a public opinion propagation sub-network identification unit, used to identify multiple public opinion propagation sub-networks based on the public opinion propagation relationship graph; an abnormal propagation structure marking unit, used to calculate the structural consistency index of the multiple public opinion propagation sub-networks and mark the public opinion propagation sub-networks that are less than a preset consistency threshold as abnormal propagation structures; and a public opinion risk warning unit, used to provide public opinion risk warnings for public opinion events corresponding to the abnormal propagation structures.

[0028] Specifically, the process of analyzing the propagation relationships and extracting propagation-related information from multi-source online public opinion datasets refers to the unified data processing and relationship identification analysis of data sources from different online platforms. Multi-source online public opinion datasets represent collections of public opinion data gathered from online information carriers such as social media platforms, news websites, forums, short video platforms, and blog systems. The data formats include structured or semi-structured data such as text content, user identifiers, publication times, interaction records, and link information. Propagation relationship analysis involves using text structure analysis, behavioral relationship identification, and time-series correlation analysis methods to identify relationships and extract propagation paths of public opinion information among different user entities. Propagation-related information represents structured relationship data that reflects the propagation process of public opinion information in cyberspace. This includes information publication relationships, information forwarding relationships, information commenting relationships, and information citation relationships. Information release relationships represent the original information generation relationship formed when public opinion information is first generated and publicly released by the disseminating entity. Information forwarding relationships represent the diffusion relationship formed when online users further disseminate existing information. Information commenting relationships represent the interactive relationship formed when users provide feedback on original or forwarded information. Information citation relationships represent the information association relationship formed when the disseminating entity cites or links to existing public opinion information when releasing new information. By analyzing and extracting various dissemination relationships, a set of dissemination-related information that reflects the dissemination behavior of public opinion information can be formed, thus providing basic data for subsequent dissemination structure modeling.

[0029] Furthermore, identifying the main actors in the dissemination of public opinion and extracting the set of dissemination nodes from multi-source online public opinion datasets refers to the identification and entity extraction of information publishers, disseminators, and participants in information interaction within the dataset after completing the analysis of public opinion dissemination relationships. The main actors in public opinion dissemination refer to network user entities or information publishing entities that participate in information generation, dissemination, or interaction during the dissemination of online public opinion information. These main actors can include personal user accounts, media organization accounts, organizational accounts, and automated information publishing nodes. The set of dissemination nodes represents the set of dissemination actors identified and organized from multi-source online public opinion datasets using entity recognition algorithms and user identifier extraction methods. Each dissemination node corresponds to a uniquely identified public opinion dissemination entity. The set of dissemination nodes constitutes the node foundation in the public opinion dissemination network structure, providing a set of subject objects for subsequent dissemination relationship connections and network structure construction.

[0030] Subsequently, determining the information propagation relationships between nodes in the propagation node set based on propagation association information and obtaining the propagation relationship edge set refers to establishing explicit propagation connections between nodes in the propagation node set using propagation association information. Information propagation relationships represent the information interaction paths between nodes formed by information publishing, forwarding, commenting, or quoting during the public opinion propagation process. Propagation relationship edges represent the relational units used to connect two propagation nodes in the propagation network structure. Each propagation relationship edge describes the information propagation behavior between two propagation subjects and can contain attribute information such as propagation direction, propagation time, and propagation type. The propagation relationship edge set represents the set structure of all propagation relationship edges formed by traversing the propagation node set and combining it with propagation association information. The propagation relationship edge set can comprehensively describe the propagation connection structure formed between different propagation subjects of public opinion information.

[0031] Furthermore, constructing a propagation relationship graph based on the set of propagation nodes and the set of propagation relationship edges refers to formally representing the structure of public opinion propagation using graph structure modeling methods. The propagation relationship graph represents the overall modeling and structural expression of the public opinion information propagation path, using propagation nodes as node units in the graph structure and propagation relationship edges as connections between nodes. The propagation relationship graph can describe the multi-level propagation structure and information diffusion path formed by different propagation subjects in online public opinion, while also reflecting the structural characteristics, propagation density, and propagation hierarchy relationships in the public opinion propagation network. By constructing a propagation relationship graph, discrete propagation behavior data can be transformed into a data representation with structured topological relationships, providing a network structure foundation for subsequent topological feature extraction, data mining analysis, and public opinion risk assessment.

[0032] Furthermore, the process of identifying multiple sub-networks of public opinion dissemination based on the public opinion dissemination relationship graph refers to the decomposition and structural division of the dissemination network structure based on the already constructed public opinion dissemination relationship graph. The public opinion dissemination relationship graph represents a network structure model composed of dissemination nodes and dissemination relationship edges. Dissemination nodes represent network subjects participating in the generation, dissemination, or interaction of public opinion information, while dissemination relationship edges represent the information dissemination connections formed between dissemination subjects. A public opinion dissemination sub-network represents a local dissemination structure unit within the overall dissemination network structure, composed of dissemination nodes with close dissemination links and corresponding dissemination relationship edges. Public opinion dissemination sub-networks can reflect the local dissemination scope and path structure of a specific public opinion event or topic in cyberspace. By dividing the public opinion dissemination relationship graph into network structures, multiple relatively independent dissemination sub-structures can be identified. Each public opinion dissemination sub-network corresponds to a public opinion dissemination unit with a relatively complete dissemination chain, thus providing a structured analysis object for subsequent dissemination structure feature analysis.

[0033] Furthermore, the process of calculating the structural consistency index for multiple public opinion dissemination sub-networks and marking those with a consistency threshold below a preset threshold as abnormal dissemination structures refers to the quantitative assessment of the stability of the dissemination structure and the consistency of the dissemination pattern for each public opinion dissemination sub-network. The structural consistency index is an evaluation metric used to measure the stability of the dissemination relationship structure within a dissemination sub-network. It can be calculated based on dissemination path distribution characteristics, node connection density characteristics, dissemination hierarchy distribution characteristics, and node topological similarity characteristics. By comprehensively analyzing the regularity of the connection relationships between nodes in the dissemination sub-network and the degree of structural stability, a numerical result reflecting the consistency of the dissemination structure is obtained. The preset consistency threshold is a pre-set structural judgment standard based on historical statistical patterns or empirical rules of public opinion dissemination structures. When the structural consistency index is lower than the preset consistency threshold, it indicates that the dissemination relationship within the corresponding dissemination sub-network exhibits characteristics such as structural instability, abnormal dissemination paths, or abnormal node connection relationships. In the above situation, the corresponding public opinion dissemination subnetwork is marked as an abnormal dissemination structure. An abnormal dissemination structure indicates a dissemination network structure whose dissemination pattern is significantly different from the normal public opinion dissemination pattern. Abnormal dissemination structures are usually related to abnormal information diffusion behavior, malicious information dissemination behavior, or sudden public opinion diffusion behavior.

[0034] Subsequently, the process of issuing public opinion risk warnings for public opinion events corresponding to abnormal propagation structures refers to the risk assessment and early warning handling of public opinion events associated with abnormal propagation structures after identifying them. A public opinion event refers to a set of related public opinion information generated around a specific social event, public issue, or online topic. Public opinion events are typically composed of multiple information release behaviors, dissemination behaviors, and user interaction behaviors. Public opinion risk warnings involve comprehensively analyzing the abnormal characteristics of public opinion propagation structures, trends in propagation scale, and changes in propagation speed to identify and warn of public opinion events that may trigger risks of public opinion spread, social impact, or information distortion. By associating abnormal propagation structures with public opinion events, public opinion information with potential propagation risks or abnormal spread trends can be identified, thereby providing decision support information for online public opinion supervision, public opinion guidance, and risk prevention and control.

[0035] The public opinion topology feature vector acquisition module 2 is used to extract multi-source public opinion semantic features from the multi-source network public opinion related dataset, and to extract topology features from the multi-source public opinion semantic features based on the public opinion propagation relationship graph to obtain the public opinion topology feature vector.

[0036] Furthermore, this application also includes: in the public opinion topology feature vector acquisition module 2, the multi-source public opinion semantic features include at least one of the following: text sentiment polarity features, text sentiment intensity features, text stance tendency features, text topic distribution features, named entity features, and event trigger word features.

[0037] Furthermore, this application also includes: a feature calculation unit, used to calculate node propagation centrality features, node clustering coefficient features, and propagation path hierarchy features based on the public opinion propagation relationship graph; a first fused topological feature vector obtaining unit, used to fuse the node propagation centrality features, node clustering coefficient features, and propagation path hierarchy features to obtain a first fused topological feature vector; and a public opinion topological feature vector output unit, used to fuse the first fused topological feature vector based on the multi-source public opinion semantic features to obtain a second fused topological feature vector, and output the second fused topological feature vector as the public opinion topological feature vector.

[0038] Specifically, extracting multi-source semantic features from multi-source online public opinion datasets refers to the process of extracting and calculating features representing sentiment, stance, and theme from online public opinion data from multiple sources, based on specific analytical objectives and technical means. Multi-source online public opinion datasets represent collections of public opinion data from different online platforms, including news reports, social media posts, forum discussions, blog articles, etc., involving different types of information and viewpoints. Through semantic feature extraction, valuable semantic information for public opinion analysis can be extracted from this diverse data, and these features can characterize the potential meaning and attitude of public opinion information, further providing effective support for public opinion monitoring and risk assessment. Multi-source public opinion semantic features include at least one of the following: text sentiment polarity features, text sentiment intensity features, text stance tendency features, text theme distribution features, named entity features, and event trigger word features. Text sentiment polarity features represent the emotional orientation expressed in the text, typically including positive, negative, or neutral sentiment. Sentiment polarity features can help analyze the public's emotional attitude towards specific public opinion topics. Text sentiment intensity features further quantify the intensity of emotional expression in the text, reflecting the strength of the emotions expressed and assessing the depth of emotional expression. Text stance characteristics refer to the attitude and position held by the text when discussing a particular issue, typically supporting, opposing, or remaining neutral, revealing the text's bias towards a particular event or topic. Text topic distribution features extract potential topic or theme information from the text, showcasing the thematic structure of the text's discussion and revealing the main content direction. Named entity features represent specific entity names marked in the text, such as names of people, places, and organizations, used to identify and classify key objects in the text. Event trigger word features refer to keywords or phrases appearing in the text that can trigger discussion of a specific event or public opinion topic, indicating the occurrence or change of an event. By extracting various public opinion semantic features, we can comprehensively understand and capture the emotional attitudes, stances, discussion topics, and related events in the text, thus providing multi-dimensional data support for public opinion analysis, trend prediction, and risk assessment.

[0039] Furthermore, the process of calculating node propagation centrality, node clustering coefficient, and propagation path hierarchy features based on the public opinion propagation relationship graph refers to analyzing each node in the graph and its connections to calculate features reflecting the importance of each node in the entire propagation network. Node propagation centrality describes the propagation influence of each node in the network, specifically referring to the degree of connection between the node and other nodes through the propagation path, reflecting the role the node plays in the propagation process. Node clustering coefficient indicates the density of neighboring nodes around each node; a higher clustering coefficient indicates a closer connection between the node and its neighbors, enabling more effective information propagation. Propagation path hierarchy describes the hierarchical structure of information propagation in the network, reflecting the path hierarchy and complexity of information from the source node to the target node. By calculating these features, a deeper analysis of the role and influence of nodes in the public opinion propagation network can be achieved.

[0040] The process of fusing node propagation centrality features, node clustering coefficient features, and propagation path hierarchy features to obtain the first fused topological feature vector refers to combining and integrating the calculated node features to form a new, more comprehensive feature representation. Feature fusion integrates multiple independent features into a unified feature vector, taking into account both the independence and complementarity of each feature, thereby improving the accuracy and information content of the feature representation. The first fused topological feature vector represents the fused node feature set, containing multiple information such as the node's centrality, clustering, and propagation hierarchy in the propagation network. It can comprehensively reflect the topological characteristics of the nodes, providing richer input data for subsequent data mining and analysis.

[0041] The process of fusing the first fused topological feature vector based on multi-source public opinion semantic features to obtain the second fused topological feature vector refers to further combining multi-source public opinion semantic features with topological features on the basis of feature fusion. Multi-source public opinion semantic features represent feature data obtained through analysis of text in public opinion data at the sentiment, theme, and other levels, reflecting information such as the emotional attitude, stance bias, and discussion topics of public opinion information. Fusing multi-source public opinion semantic features with the first fused topological feature vector combines the physical characteristics of the network propagation structure with the semantic content of public opinion information. This results in a feature vector that includes both the topological information of the propagation network and the semantic information in the public opinion data, thus providing comprehensive feature support for more in-depth public opinion analysis and risk assessment. The second fused topological feature vector, i.e., the fused comprehensive feature vector, can be output as the public opinion topological feature vector, providing accurate data representation for subsequent public opinion monitoring, early warning analysis, and other tasks.

[0042] The public opinion feature representation space acquisition module 3 is used to construct a topology constraint data mining model based on the topology consistency constraint threshold, call the topology constraint data mining model to perform public opinion data mining on the public opinion topology feature vector, and obtain the public opinion feature representation space. The topology consistency constraint threshold is obtained through dynamic calculation.

[0043] Furthermore, this application also includes: a node topology similarity matrix obtaining unit, used to obtain node propagation structure features based on the public opinion propagation relationship graph, calculate the topology similarity between nodes according to the node propagation structure features, and obtain a node topology similarity matrix; a topology consistency probability distribution function generating unit, used to calculate the topology consistency distribution interval according to the node topology similarity matrix, and generate a topology consistency probability distribution function by performing probability distribution fitting on all similarity values ​​in the topology consistency distribution interval; and a topology consistency constraint threshold determining unit, used to determine the topology consistency constraint threshold according to the topology consistency probability distribution function.

[0044] Furthermore, this application also includes: a topological adjacency constraint set filtering unit, used to filter topological adjacency constraint sets that are greater than or equal to the topological consistency constraint threshold; and a topological constraint data mining model obtaining unit, used to construct a topological constraint objective function based on the public opinion topological feature vector to obtain a topological constraint data mining model, wherein the topological constraint objective function includes a data mining loss function and a topological consistency constraint term, and the topological consistency constraint term is used to constrain the selection of node pairs with topological consistency relationships from the topological adjacency constraint set.

[0045] Furthermore, this application also includes: a public opinion feature representation space initialization unit, used to initialize the public opinion feature representation space according to the public opinion topological feature vector; and a public opinion feature representation space acquisition unit, used to iteratively optimize and update the initialized public opinion feature representation space by introducing the topological constraint objective function of the topological constraint data mining model, and to acquire the public opinion feature representation space when the topological constraint objective function reaches a preset number of iterations; wherein, each public opinion feature representation vector in the public opinion feature representation space includes public opinion semantic information and public opinion propagation structure information.

[0046] Furthermore, this application also includes: a topological similarity acquisition channel, used to recalculate the updated public opinion feature representation vector after each iteration of optimization to obtain the updated topological similarity; a topological consistency distribution interval calculation channel, used to recalculate the topological consistency distribution interval based on the updated topological similarity; and a topological consistency constraint threshold calculation channel, used to recalculate the topological consistency constraint threshold based on the updated topological consistency distribution interval.

[0047] Specifically, the process of obtaining node propagation structure characteristics based on public opinion propagation relationship graphs refers to extracting node structure information related to the propagation process from an established propagation relationship graph. A public opinion propagation relationship graph represents a network structure formed by connecting propagation nodes and their propagation relationships. Node propagation structure characteristics describe the relative position and importance of each node in the propagation network. Based on these structural characteristics, the properties of nodes in terms of influence, centrality, and propagation path during the propagation process can be revealed.

[0048] Calculating topological similarity between nodes based on their propagation structure characteristics refers to evaluating the similarity reflected in the propagation structure by calculating the similarity between each pair of nodes. Topological similarity assesses the degree of similarity between nodes throughout the entire propagation network by comparing factors such as propagation paths, adjacent nodes, and propagation levels. The topological similarity matrix between nodes is a symmetric matrix, where each element represents the similarity value between two nodes in the graph. This matrix allows for the quantification and comparison of the similarity between different nodes in the network, providing a foundation for subsequent data analysis and model building.

[0049] The process of calculating the topology consistency distribution interval based on the node topology similarity matrix refers to analyzing the topology similarity values ​​between nodes to calculate the range and interval of the topology consistency distribution. The topology consistency distribution interval represents the distribution range of the topology similarity values ​​among all nodes, reflecting the changes in the degree of similarity between nodes. Through this distribution interval, it is possible to identify which nodes in the network have high consistency and which nodes exhibit anomalies in their topology.

[0050] The process of generating a topology consistency probability distribution function by fitting a probability distribution to all similarity values ​​within a topology consistency distribution interval refers to statistically analyzing the similarity values ​​within the interval and fitting them to a mathematical function to obtain a probability function describing the distribution pattern of similarity. Probability distribution fitting can employ common statistical methods, such as maximum likelihood estimation or least squares, to fit the similarity values ​​within the distribution interval into a mathematical function, thereby obtaining a probabilistic model that reflects the similarity characteristics of network nodes.

[0051] The process of determining the topology consistency constraint threshold based on the topology consistency probability distribution function refers to using this function to determine a numerical threshold that distinguishes whether nodes in a network have a consistency relationship. Specifically, the topology consistency constraint threshold is calculated by analyzing the node propagation structure characteristics in the public opinion propagation relationship graph. This threshold reflects the similarity and structural relationship between nodes. In data mining, when the similarity between nodes is higher than this threshold, it indicates that the two nodes have strong topology consistency and can maintain a proximity relationship in the feature space.

[0052] Furthermore, the process of filtering the set of topological adjacency constraints that is greater than or equal to the topological consistency constraint threshold refers to selecting node pairs that meet certain consistency requirements from the inter-node relationships in the public opinion propagation network based on the dynamically calculated topological consistency constraint threshold. The topological adjacency constraint set represents a group of node pairs in which the connections between nodes conform to specific topological structure constraints. This filtering process helps ensure that the node pairs used in data mining have sufficient structural consistency, ensuring that the mining results reflect the true characteristics of the propagation network.

[0053] The process of constructing a topology-constrained objective function based on public opinion topology feature vectors refers to using the topology feature vectors extracted from the public opinion propagation network as input to construct an objective function for data mining. The public opinion topology feature vectors represent the comprehensive feature information obtained by analyzing nodes in the public opinion propagation relationship graph. The constructed topology-constrained objective function constrains node feature learning and model optimization during the data mining process by comprehensively considering the topological relationships and node features of each node in the network. The role of the topology-constrained objective function is to ensure that the influence of the network topology is fully considered during the data mining process, enabling the model to effectively learn node features consistent with the network structure in the data feature space. The topology-constrained objective function includes a data mining loss function and a topology consistency constraint term. The data mining loss function measures the error between the model's predictions and the actual results during the data mining process; by minimizing the loss function, the accuracy of the model is optimized.

[0054] The topology-constrained data mining model is a data mining model built upon a topology-constrained objective function. It is used for node feature extraction and data analysis in public opinion propagation networks. Specifically, by constraining the relationships between similar nodes, it optimizes the data mining process, making the representation of the node feature space more consistent with the actual structure and propagation patterns of the network. The topology consistency constraint term is used to constrain the selection of node pairs with topology consistency relationships from the set of topology adjacency constraints. This ensures that the relationships between node pairs in the feature space conform to the topology of the actual propagation network, guaranteeing that the model can automatically identify and retain topology-consistent node pairs during data mining, thereby improving the stability of feature learning and the reliability of the results.

[0055] Furthermore, the process of initializing the public opinion feature representation space based on the topological feature vectors of public opinion refers to constructing an initial feature representation space based on the topological feature vectors extracted from the public opinion propagation network. The public opinion feature representation space represents a multi-dimensional feature space, where each dimension corresponds to a certain feature or attribute of the public opinion data. By initializing the public opinion feature representation space, a preliminary spatial framework can be provided for subsequent data analysis and mining, so as to further optimize and adjust the data representation in the feature space.

[0056] The process of iteratively optimizing and updating the initialized public opinion feature representation space by introducing a topology constraint objective function into the topology constraint data mining model refers to the process of introducing a topology constraint data mining model based on the initial feature representation space, and continuously adjusting the node features in the public opinion feature representation space through iterative optimization of the objective function until the preset optimization conditions are met. The topology constraint objective function is used to restrict the relationships between nodes in the feature representation space, ensuring that the nodes in the feature space maintain topological consistency in the network structure. By iteratively updating the objective function, the public opinion feature representation space can be gradually optimized, making the node features in the feature space more consistent with the topology of the actual propagation network, thereby improving the effectiveness and accuracy of data mining.

[0057] Furthermore, the process of recalculating the updated public opinion feature representation vectors in each iteration refers to recalculating each public opinion feature vector in the public opinion feature representation space in each round of optimization to ensure that it can more accurately reflect the characteristics of public opinion data. The public opinion feature representation vectors contain the semantic features and propagation structure features of public opinion information. By recalculating each feature vector, its position in the feature space can be adjusted, allowing the optimized vector to better conform to the topological rules of the network structure. The recalculation after each iteration causes the feature space to gradually converge and reflects the latest node propagation relationships and public opinion content features in each round of optimization.

[0058] The updated topological similarity refers to calculating the topological similarity between nodes in the network based on the updated feature vectors after recalculating each sentiment feature representation vector. Topological similarity measures the similarity between nodes by comparing the strength and relationships of their interconnections within the propagation network. Higher topological similarity indicates a stronger structural association within the propagation network. The updated topological similarity reflects the changes in the similarity of the optimized sentiment feature representation vectors within the propagation structure.

[0059] The process of recalculating the topology consistency distribution interval based on updated topology similarity refers to reassessing the distribution range of node similarity in the network based on new topology similarity data. The topology consistency distribution interval represents the statistical range of topology similarity between nodes, reflecting the range of changes in node similarity throughout the network. By recalculating the consistency distribution interval, the criteria for judging node similarity can be dynamically adjusted, ensuring the accuracy of topology consistency analysis. The updated distribution interval can better capture changes in relationships between nodes and reflect fluctuations in factors such as propagation paths and propagation density within the network.

[0060] The process of recalculating the topology consistency constraint threshold based on the updated topology consistency distribution interval refers to recalculating the topology consistency constraint threshold after the topology consistency distribution interval is updated. The topology consistency constraint threshold is a dynamically determined value based on the distribution of node topology similarity, used to constrain the topological relationships between nodes during data mining. When the topology similarity of nodes exceeds this threshold, the two nodes are considered to have sufficient similarity to maintain consistency in the feature space. The recalculated topology consistency constraint threshold ensures that, during data mining, constraints can be reasonably adjusted according to the latest changes in network topology, thereby improving the accuracy and adaptability of the mining model.

[0061] When the topological constraint objective function reaches the preset number of iterations, the optimization process of the public opinion feature representation space will be completed, and the final public opinion feature representation space will be obtained. The preset number of iterations represents the optimization rounds in the data mining process. That is, in each round of optimization, the model will adjust the feature space according to the objective function until the set iteration limit is reached, ensuring that the model can complete the optimization under limited computing resources. The final public opinion feature representation space contains comprehensive features optimized from the initial data, which can reflect the important node features and propagation network structure features in the public opinion propagation process.

[0062] Each public opinion feature representation vector in the public opinion feature representation space includes semantic information and structural information about public opinion dissemination. Semantic information includes features such as sentiment polarity, topic distribution, and sentiment intensity extracted from public opinion text analysis; these reflect the sentiment tendency and topic information in the public opinion data. Structural information about public opinion dissemination originates from features such as node propagation centrality, clustering coefficient, and propagation path hierarchy in the propagation relationship graph, reflecting the propagation patterns, node relationships, and network structure within the public opinion dissemination network. By combining these two types of information in the feature representation vector, public opinion information can be more comprehensively represented, providing stronger data support for subsequent tasks such as public opinion risk assessment and public opinion trend analysis.

[0063] The public opinion risk warning result output module 4 is used to perform risk analysis on the public opinion feature representation space, calculate the public opinion dissemination risk index, and output the public opinion risk warning result using the public opinion dissemination risk index.

[0064] Furthermore, this application also includes: a public opinion dissemination energy calculation unit, used to calculate the public opinion dissemination energy of each node in the public opinion dissemination relationship graph based on the public opinion feature representation space; a public opinion dissemination topology energy field construction unit, used to construct a public opinion dissemination topology energy field in the public opinion dissemination relationship graph based on the public opinion dissemination energy; and a public opinion dissemination risk index obtaining unit, used to calculate the public opinion dissemination topology energy field based on a preset public opinion dissemination energy threshold to obtain a public opinion dissemination risk index.

[0065] Specifically, the process of calculating the public opinion propagation energy of each node in the public opinion propagation relationship graph based on the public opinion feature representation space refers to calculating the energy value of each node in the public opinion propagation network by using the node feature information in the public opinion feature representation space. Public opinion propagation energy represents the influence or propagation capability of a node during the propagation process, and is usually related to the node's propagation centrality, propagation intensity, and the degree of influence on the propagation of other nodes. By calculating based on the node characteristics in the public opinion feature representation space, the role and importance of each node in the entire propagation network can be effectively assessed, providing a basis for subsequent risk assessment and propagation path analysis.

[0066] The process of constructing a topological energy field for public opinion dissemination within a public opinion dissemination relationship graph, based on the dissemination energy of each node, refers to building a global topological energy field for public opinion dissemination. This topological energy field represents the distribution of dissemination energy among all nodes in the entire public opinion dissemination network, revealing the strength of interactions between nodes and the density of dissemination paths. By constructing this topological energy field, the energy distribution within the dissemination network can be visually displayed, thus aiding in the analysis of which nodes play key roles in public opinion dissemination, as well as potential risks and dissemination directions during the process.

[0067] The process of calculating the public opinion propagation risk index based on a preset public opinion propagation energy threshold involves further analyzing the propagation nodes and their paths within the energy field by setting a threshold for public opinion propagation energy. The public opinion propagation energy threshold is a critical energy value set based on experience or historical data, used to determine which nodes and propagation paths have a high propagation risk. By calculating the nodes and paths in the public opinion propagation topology energy field, a public opinion propagation risk index can be generated. This risk index reflects the risk level of the entire public opinion propagation network and can provide effective data support for public opinion monitoring and early warning.

[0068] The process of using a public opinion dissemination risk index to output public opinion risk early warning results refers to generating corresponding public opinion risk early warning information based on the calculated public opinion dissemination risk index, by quantifying the dissemination energy of dissemination nodes, the density of dissemination paths, and the complexity of dissemination structures. The public opinion risk early warning results, through the analysis of the public opinion dissemination risk index, output warning information on potential public opinion crises or social events. These results include the identification of high-risk dissemination nodes, hot topics, and key dissemination paths, and provide information such as risk level, risk source, and possible dissemination trends. This can help relevant decision-making departments or public opinion management personnel identify potential public opinion crises in advance and conduct targeted intervention and management.

[0069] In summary, the data mining-based intelligent analysis system for online public opinion provided in this application has the following technical effects: by comprehensively analyzing the relationships, propagation paths, and propagation energy among nodes in the public opinion propagation network, it can accurately predict the spread trend, propagation risks, and potential crises of public opinion events through topological structure analysis and propagation node energy calculation, thereby improving the timeliness and accuracy of public opinion monitoring and risk warning, and ensuring effective intervention and control in the early stages of public opinion events.

[0070] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0071] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A network public opinion intelligent analysis system based on data mining, characterized in that, The system includes: The public opinion dissemination relationship graph construction module is used to collect multi-source online public opinion-related datasets to construct a public opinion dissemination relationship graph; The public opinion topology feature vector acquisition module is used to extract multi-source public opinion semantic features from the multi-source network public opinion related dataset, and to extract topology features from the multi-source public opinion semantic features based on the public opinion propagation relationship graph to obtain the public opinion topology feature vector. The public opinion feature representation space acquisition module is used to construct a topology constraint data mining model based on the topology consistency constraint threshold, call the topology constraint data mining model to perform public opinion data mining on the public opinion topology feature vector, and obtain the public opinion feature representation space, wherein the topology consistency constraint threshold is obtained through dynamic calculation; The public opinion feature representation space acquisition module includes: The node topology similarity matrix is ​​obtained by using a unit to acquire node propagation structure features based on the public opinion propagation relationship graph, and to calculate the topology similarity between nodes based on the node propagation structure features to obtain the node topology similarity matrix. The topology consistency probability distribution function generation unit is used to calculate the topology consistency distribution interval based on the node topology similarity matrix, and generate the topology consistency probability distribution function by performing probability distribution fitting on all similarity values ​​in the topology consistency distribution interval; A topology consistency constraint threshold determination unit is used to determine the topology consistency constraint threshold according to the topology consistency probability distribution function; The public opinion risk warning result output module is used to perform risk analysis on the public opinion feature representation space, calculate the public opinion dissemination risk index, and output the public opinion risk warning result using the public opinion dissemination risk index.

2. The intelligent network public opinion analysis system based on data mining as described in claim 1, characterized in that, The public opinion dissemination relationship graph construction module includes: The propagation association information extraction unit is used to analyze the propagation relationship of the multi-source network public opinion related dataset and extract propagation association information, which includes information release relationship, information forwarding relationship, information comment relationship and information citation relationship; The propagation node set extraction unit is used to identify the propagation subject of the multi-source network public opinion related dataset and extract the propagation node set. The propagation relationship edge set is obtained by a unit, which is used to determine the information propagation relationship between nodes in the propagation node set based on the propagation association information, thereby obtaining the propagation relationship edge set; The propagation relationship graph construction unit is used to construct a propagation relationship graph according to the propagation node set and the propagation relationship edge set.

3. The intelligent network public opinion analysis system based on data mining as described in claim 1, characterized in that, The public opinion topology feature vector acquisition module includes: The feature calculation unit is used to calculate node propagation centrality features, node clustering coefficient features, and propagation path hierarchy features based on the public opinion propagation relationship graph. The first fusion topological feature vector obtaining unit is used to fuse the node propagation centrality feature, node clustering coefficient feature and propagation path hierarchy feature to obtain the first fusion topological feature vector; The public opinion topology feature vector output unit is used to fuse the first fused topology feature vector according to the multi-source public opinion semantic features to obtain the second fused topology feature vector, and output the second fused topology feature vector as the public opinion topology feature vector.

4. The intelligent network public opinion analysis system based on data mining as described in claim 1, characterized in that, The public opinion feature representation space acquisition module includes: A topological adjacency constraint set filtering unit is used to filter topological adjacency constraint sets that are greater than or equal to the topological consistency constraint threshold. The topology constraint data mining model is obtained by a unit used to construct a topology constraint objective function based on the public opinion topology feature vector, thereby obtaining a topology constraint data mining model. The topology constraint objective function includes a data mining loss function and a topology consistency constraint term. The topology consistency constraint term is used to constrain the selection of node pairs with topology consistency relationships from the topology adjacency constraint set.

5. The intelligent network public opinion analysis system based on data mining as described in claim 4, characterized in that, The public opinion feature representation space acquisition module further includes: The public opinion feature representation space initialization unit is used to initialize the public opinion feature representation space according to the public opinion topology feature vector. The public opinion feature representation space acquisition unit is used to iteratively optimize and update the initial public opinion feature representation space by introducing the topological constraint objective function of the topological constraint data mining model. When the topological constraint objective function reaches a preset number of iterations, the public opinion feature representation space is acquired. Each public opinion feature representation vector in the public opinion feature representation space includes public opinion semantic information and public opinion propagation structure information.

6. The intelligent network public opinion analysis system based on data mining as described in claim 1, characterized in that, The public opinion risk warning result output module includes: The public opinion propagation energy calculation unit is used to calculate the public opinion propagation energy of each node in the public opinion propagation relationship graph based on the public opinion feature representation space. The public opinion dissemination topology energy field construction unit is used to construct a public opinion dissemination topology energy field in the public opinion dissemination relationship graph based on the public opinion dissemination energy. The public opinion dissemination risk index acquisition unit is used to calculate the public opinion dissemination topological energy field based on a preset public opinion dissemination energy threshold to obtain the public opinion dissemination risk index.

7. The intelligent network public opinion analysis system based on data mining as described in claim 5, characterized in that, The public opinion feature representation space acquisition unit includes: The topological similarity is used to obtain channels, which are then used to recalculate the updated public opinion feature representation vector after each iteration of optimization to obtain the updated topological similarity. The topology consistency distribution interval calculation channel is used to recalculate the topology consistency distribution interval based on the updated topology similarity. The topology consistency constraint threshold calculation channel is used to recalculate the topology consistency constraint threshold based on the updated topology consistency distribution interval.

8. The intelligent network public opinion analysis system based on data mining as described in claim 1, characterized in that, The public opinion dissemination relationship graph construction module also includes: The public opinion dissemination sub-network identification unit is used to identify multiple public opinion dissemination sub-networks based on the public opinion dissemination relationship graph; An abnormal propagation structure marking unit is used to calculate the structural consistency index of the multiple public opinion propagation sub-networks and mark the public opinion propagation sub-networks that are less than the preset consistency threshold as abnormal propagation structures. The public opinion risk early warning unit is used to provide public opinion risk warnings for public opinion events corresponding to the abnormal propagation structure.

9. The intelligent network public opinion analysis system based on data mining as described in claim 1, characterized in that, In the public opinion topology feature vector acquisition module, the multi-source public opinion semantic features include at least one of the following: text sentiment polarity features, text sentiment intensity features, text stance tendency features, text topic distribution features, named entity features, and event trigger word features.

Images