Public opinion early warning and intelligent analysis system based on data platform

Abstract

The invention relates to a public opinion early warning and intelligent analysis system based on a data center, and the system comprises a public opinion data collection module which collects public opinion data, weather forecast data and enterprise operation data from a plurality of information sources inside or outside an enterprise; the data arrangement and cleaning platform module is used for cleaning the collected public opinion data, unifying heterogeneous data, classifying and marking various data and integrating the data according to a time sequence; the public opinion theme classification module is used for creating public opinion classification standards according to enterprise target themes and constructing public opinion sub-classifications; the intelligent analysis module is used for calculating a public opinion risk score based on the propagation volume and the public opinion propagation trend and generating an intelligent early warning signal; the public opinion early warning module is used for presetting an early warning trigger condition according to weather forecast and enterprise operation data and monitoring public opinion risks in combination with an intelligent early warning signal; and the intelligent coping strategy and measure module automatically recommends or generates coping strategies and measures according to the intelligent analysis result.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a public opinion early warning and intelligent analysis system based on a data platform. Background Technology

[0002] With the rapid development of the internet and social media, the speed of information dissemination has significantly increased, and the impact of online public opinion on enterprises, governments, and public institutions is intensifying. Currently, most public opinion early warnings occur during the event itself, resulting in low efficiency in tracing the root causes of problems and slow response mechanisms. Therefore, an intelligent system capable of utilizing big data technology for public opinion early warning and analysis is needed to address the information silos existing in public opinion management for institutions and enterprises. This system would enable pre-event event-related early warnings, in-event monitoring of public opinion development, and post-event early warning mechanisms for optimization and improvement, thereby enhancing the real-time monitoring and response capabilities for public opinion.

[0003] In existing technologies, public opinion data is not integrated with internal enterprise data, and information silos exist between various business systems, making it impossible to effectively predict and monitor public opinion risks in advance. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention proposes a public opinion early warning and intelligent analysis system based on a data platform.

[0005] The technical solution of the present invention is as follows: On the one hand, this invention proposes a public opinion early warning and intelligent analysis system based on a data middle platform, including: The public opinion data collection module collects public opinion data, weather forecast data, and enterprise operation data from various information sources, both internal and external to the enterprise. The data processing and cleaning platform module cleans the collected public opinion data, unifies heterogeneous data, classifies and labels various data, and integrates them in chronological order. The public opinion topic classification module creates public opinion classification standards based on the company's target themes and constructs public opinion subcategories; The intelligent analysis module classifies public opinion data by labels based on the BERT model, and combines the random forest regression model and the Lotka-Volterra model to predict the volume and trend of public opinion dissemination; then, based on the volume and trend of public opinion dissemination, it calculates the public opinion risk score and generates intelligent early warning signals. The public opinion early warning module, based on weather forecasts and enterprise operation data, pre-sets early warning trigger conditions and monitors public opinion risks in conjunction with intelligent early warning signals; The intelligent response strategy and measures module automatically recommends or generates response strategies and measures based on intelligent analysis results.

[0006] As a preferred embodiment, the method for collecting public opinion data, weather forecast data, and enterprise operation data from multiple information sources inside or outside the enterprise specifically includes: Data is obtained from external sources through web scraping, API interfaces, and RSS subscriptions; data is also synchronized with internal enterprise systems on a scheduled or real-time basis through API integration.

[0007] As a preferred embodiment, the method for unifying heterogeneous data, classifying and labeling various data, and integrating them in chronological order after cleaning the collected public opinion data specifically includes: External unstructured data is converted into a unified format through data mapping and unified coding standards, integrating data from different sources into data of a unified format; Add corresponding labels to each type of data according to its data type; Data from different sources is integrated in chronological order, linking external public opinion dynamics with internal events.

[0008] As a preferred embodiment, the pre-set warning triggering conditions based on weather forecasts and enterprise operation data are as follows: Warning thresholds for service outages triggered by extreme weather and fault warning thresholds for abnormal enterprise operational data.

[0009] As a preferred implementation, the step of classifying public opinion data by labels based on the BERT model specifically includes: Based on the BERT model, unstructured data in public opinion data is automatically stored in layers according to classification labels; among them, video files in public opinion data are converted into text by an automatic speech recognition tool, and then the text is classified.

[0010] As a preferred implementation, the data that also needs to be collected in the step of classifying public opinion data by labels based on the BERT model and predicting the volume of public opinion dissemination using a random forest regression model includes: The BERT model is used to analyze the sentiment of the target event content, including the event's reputation, credibility, accountability, and public opinion. Check the relevance of the target event's content to current popular or trending topics; Collect user attention data for the target event; Identify key influencers or opinion leaders involved in and disseminating the target event; Collect other external data that may influence the development of the target event.

[0011] As a preferred embodiment, the step of calculating the public opinion risk score based on the volume of dissemination and the trend of public opinion dissemination specifically includes: A public opinion risk scoring model is constructed based on the grey clustering method. The scores of each risk dimension are standardized to construct an event-risk assessment matrix. The risk level is then divided into several grey class levels. The membership degree of each public opinion event's score to each grey class is calculated. The membership degrees of each risk dimension are weighted to obtain the overall score of the target event in different grey classes, thus determining the final risk level of the target event.

[0012] As a preferred implementation method, corresponding strategies are recommended based on historical data and a countermeasures database, and the countermeasures database is continuously optimized through feedback information.

[0013] On the other hand, the present invention proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a public opinion early warning and intelligent analysis system based on a data middle platform as described in any embodiment of the present invention.

[0014] On the other hand, the present invention proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a public opinion early warning and intelligent analysis system based on a data middle platform as described in any embodiment of the present invention.

[0015] The present invention has the following beneficial effects: 1. This invention enables the classification and hierarchical management of public opinion data based on different topic tags. This allows the public opinion data asset layer to accurately support rapid retrieval and access across multiple topics.

[0016] 2. This invention combines weather forecasts with various internal management systems of an enterprise. By configuring rules for associating public opinion events, it automatically issues early warning signals for rules that trigger early warning conditions, thereby monitoring public opinion risks.

[0017] 3. This invention uses an intelligent analysis module to predict the volume of target events and analyze the trend of public opinion diffusion, monitor the dynamic changes of public opinion, and make further response strategies. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0021] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0023] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0024] Example 1: See Figure 1 A public opinion early warning and intelligent analysis system based on a data middle platform includes: The public opinion data collection module collects public opinion data, weather forecast data, and enterprise operation data from various information sources, both internal and external to the enterprise. In this embodiment, the system first acquires public opinion data from multiple information sources through a data acquisition module. Data sources include, but are not limited to, the following channels: External data: Information sources include social media platforms, news websites, industry forums, government announcements, and weather forecasts. External data can be acquired in real-time and automatically through web scraping, API interfaces, and RSS subscriptions.

[0025] Data types: Structured and unstructured data, such as text, images, videos, and audio.

[0026] Data collection frequency: Real-time or timed capture, with the collection frequency adjusted according to the needs of public opinion monitoring.

[0027] Internal company data: Information sources include the company's internal production billing system, safety production system, human resources system, fault reporting system, customer service system, and other business systems.

[0028] Data type: Structured data, such as production data, expense data, and operation logs.

[0029] Data acquisition method: Data is synchronized with the enterprise's internal systems through interfaces, either periodically or in real time.

[0030] This data acquisition module can provide enterprises with comprehensive public opinion monitoring and analysis. By integrating external and internal data, it can provide early warnings of public opinion risks, help enterprises respond to changes in the external environment in a timely manner, and improve internal management.

[0031] The data processing and cleaning platform module cleans the collected public opinion data, unifies heterogeneous data, classifies and labels various data, and integrates them in chronological order. The public opinion topic classification module creates public opinion classification standards based on the company's target themes and constructs public opinion subcategories; In this embodiment, firstly, based on the topics of interest to the enterprise, clear public opinion classification criteria are created, such as the following common topics: Work safety: This includes topics such as production operations, workplace accidents, and safety training.

[0032] Labor relations management: This involves topics such as employee rights, wages and benefits, and labor contract disputes.

[0033] System Failure: Monitor public opinion topics related to system operation problems, crashes, and service interruptions.

[0034] Fee irregularities: Focus on public opinion topics related to user fees, bills, and billing disputes.

[0035] Customer complaints: This mainly includes feedback and complaints from consumers, especially regarding service quality and product performance.

[0036] Marketing activities: These include brand advertising, promotional discounts, event publicity, and the selection of brand ambassadors.

[0037] The intelligent analysis module classifies public opinion data by labels based on the BERT model, and combines the random forest regression model and the Lotka-Volterra model to predict the volume and trend of public opinion dissemination; then, based on the volume and trend of public opinion dissemination, it calculates the public opinion risk score and generates intelligent early warning signals. In this embodiment, a model capable of classifying text into different topics can be trained using BERT text classification data. Utilizing the BERT model to classify and manage news, forum, or microblog content enables rapid classification of unstructured text data.

[0038] In brand event communication, the Lotka-Volterra model is called the Lotka-Volterra model, which may be simply referred to as the predator-prey model or the competition model.

[0039] This model is used to vividly describe the dynamic interaction between information (predator) and the audience (prey), providing a framework for understanding and predicting the dissemination effects of events among the public.

[0040] The Lotka-Volterra model is based on a set of differential equations that describe the dynamic interactions between two populations (predator and prey in the original ecological model). These equations reflect how population size changes over time and mathematically simulate competition and dependence between populations in the real world. In the context of public opinion or event dissemination, these two populations can be reinterpreted as information and audience. The basic form of this model is as follows: 1. Prey (x) – In event analysis, this represents the generator of information or the disseminator of public opinion. Its growth rate is affected by its own numbers and the influence of prey (i.e., the process by which information is received by the recipient), as shown in the following formula:

[0041] In the formula, a is the natural growth rate of event spreaders, bxy represents the process of event reception, and b is the reception rate.

[0042] 2. Predators (y) – In converged media event analysis, this represents the recipients or reactants of the event. Changes in this group depend not only on the number of event disseminators but also on internal factors (such as saturation, feedback, etc.), as shown in the following formula:

[0043] In the formula, cxy represents the increase in the number of receivers through exposure to information, c is the conversion rate, and d represents the natural decrease rate of receivers or the speed at which they lose interest.

[0044] 3. Event Diffusion Weight (m) – This parameter is generally used to describe the dynamic effect of external influences on a system population, especially in the context of population diffusion or event propagation. The weight m reflects the intensity of the effect of a specific event, resource, or propagation medium on different populations, affecting their growth or decline rate.

[0045] Based on the known population change rate Solve for c using the following formula:

[0046]

[0047] Will Substitute into the above formula to derive m.

[0048] 4. The percentage increase in recipients through information exposure (c) – This parameter reflects the efficiency of information transfer from disseminators to recipients. If information is frequently reshared or commented on, it means each disseminator can more effectively influence more recipients, thus this parameter should be increased. This represents an exponential increase over time in the early stages of public opinion dissemination; additionally, an event dissemination weight m is added, with different events having different weights in terms of public opinion appeal.

[0049] # Define the time dependency function for c def time_dependent_c(t,m,day): if t <day: return 0.01 + m*(t*t / 5) # Gradually increase c before day d. else: return 0.02*np.exp(day-t) # Decrease c rapidly after day d.

[0050] 5. Natural attrition rate of receivers (d) -- External interventions can suppress the spread of certain types of information, such as combating online rumors, thereby increasing the natural attrition rate of receivers of such information, because once a rumor is discovered, receivers may quickly abandon spreading or trusting such information.

[0051] # Define the time dependency function for d def time_dependent_d(t,n): if t <n: return 0.1 # Keep d at a low level else: return 0.5 # Due to external factors, intervention occurred after day n.

[0052] The public opinion early warning module, based on weather forecasts and enterprise operation data, pre-sets early warning trigger conditions and monitors public opinion risks in conjunction with intelligent early warning signals; In this embodiment, the public opinion early warning module achieves real-time monitoring and early warning management of multi-dimensional information by constructing a "public opinion event association rule engine." This solution aims to combine key information sources such as weather forecast data and enterprise operational data, and through systematic association rule configuration, to achieve comprehensive and automated identification and early warning of potential risks. The system proactively monitors and analyzes different types of early warning signals before an event occurs, and through the dynamic configuration of the rule engine, triggers corresponding response measures and notifications in advance to reduce the possibility of the event escalating into a public opinion crisis.

[0053] Association Rule Engine Configuration: By configuring the rule engine, the system dynamically associates different types of public opinion events with specific data sources, early warning conditions, and correlations. The public opinion early warning module enables a shift from "passive response" to "proactive early warning," helping companies identify and handle potentially escalating public opinion risks early, thereby improving customer satisfaction and brand influence.

[0054] Weather forecast data: Analyze the impact of weather on business operations and public services, especially during extreme weather events (typhoons, heavy rain, etc.), which may lead to service disruptions. By using weather forecast trigger thresholds (such as warning levels, frequency of weather changes, etc.), the system can provide early warnings of potential public opinion risks and take countermeasures in advance.

[0055] Enterprise operational data: Monitoring data such as system failures, production anomalies, tariff changes, and customer complaints. Once these data exceed set thresholds (such as failure rate, number of complaints, tariff anomalies, etc.), an early warning mechanism is triggered, and the data is automatically added to the early warning database.

[0056] Taking the recent Shanghai electricity bill incident as an example, the surge in electricity consumption due to tiered pricing combined with weather changes led to public outcry over rising electricity prices in Shanghai. This incident can be addressed by configuring rules based on indicators such as the percentage of users exceeding the tiered pricing system and changes in monthly electricity consumption to issue early warnings about abnormal pricing and provide relevant response strategies.

[0057] The intelligent response strategy and measures module automatically recommends or generates response strategies and measures based on intelligent analysis results.

[0058] As a preferred embodiment of this example, the method for collecting public opinion data, weather forecast data, and enterprise operation data from multiple information sources inside or outside the enterprise is specifically as follows: Data is obtained from external sources through web scraping, API interfaces, and RSS subscriptions; data is also synchronized with internal enterprise systems on a scheduled or real-time basis through API integration.

[0059] In this embodiment, external and internal data enter the system through different acquisition methods. As the core platform for data storage and management, the data platform needs to provide multi-source data access capabilities.

[0060] External data is accessed in real time or in batches via web crawlers or API interfaces, including unstructured data such as social media comments, news articles, and online posts.

[0061] Internal data is integrated with enterprise systems, such as production billing systems, safety production systems, human resources systems, fault reporting systems, and customer service systems, and structured data is synchronized to the data platform as needed.

[0062] As a preferred embodiment of this practice, the method for unifying heterogeneous data, classifying and labeling various data, and integrating them in chronological order after cleaning the collected public opinion data specifically includes: External unstructured data is converted into a unified format through data mapping and unified coding standards, integrating data from different sources into data of a unified format; Add corresponding labels to each type of data according to its data type; Data from different sources is integrated in chronological order, linking external public opinion dynamics with internal events.

[0063] In this embodiment, the above steps are more specifically as follows: Unifying heterogeneous data formats: External data is mostly unstructured text, while internal data is usually structured (such as production logs, billing records, etc.). Format conversion is needed through data mapping, unified encoding standards, and other methods to integrate data from different sources onto the same platform.

[0064] Data tagging: Tagging external and internal data, such as adding sentiment tags and topic category tags to external public opinion data, and adding production process tags and system event tags to internal data to facilitate subsequent analysis.

[0065] Timeline integration: Integrate data from different sources in chronological order to ensure that data analysis can link external public opinion dynamics with the progress of internal events.

[0066] Thematic integration: Integrating data from different sources according to different themes enables more accurate monitoring of various public opinion topics, thereby improving the company's relevant departments' ability to respond to public events involving sudden online public opinion crises.

[0067] In a preferred embodiment of this example, the pre-set warning triggering condition based on weather forecasts and enterprise operation data is as follows: Warning thresholds for service outages triggered by extreme weather and fault warning thresholds for abnormal enterprise operational data.

[0068] In this embodiment, weather forecast data is used to analyze the impact of weather on business operations and public services, especially the potential for service disruptions during extreme weather events (typhoons, heavy rain, etc.). By using weather forecast trigger thresholds (such as warning levels, weather change frequency, etc.), the system can provide early warnings of possible public opinion risks and take proactive countermeasures.

[0069] Enterprise operational data: Monitoring data such as system failures, production anomalies, tariff changes, and customer complaints. Once these data exceed set thresholds (such as failure rate, number of complaints, tariff anomalies, etc.), an early warning mechanism is triggered, and the data is automatically added to the early warning database.

[0070] As a preferred embodiment of this practice, the step of classifying public opinion data by labels based on the BERT model specifically includes: Based on the BERT model, unstructured data in public opinion data is automatically stored in layers according to classification labels; among them, video files in public opinion data are converted into text by an automatic speech recognition tool, and then the text is classified.

[0071] In a preferred embodiment of this practice, the data that needs to be collected in the step of classifying public opinion data by labels based on the BERT model and predicting the volume of public opinion dissemination using the random forest regression model includes: The BERT model is used to analyze the sentiment of the target event content, including the event's reputation, credibility, accountability, and public opinion. Check the relevance of the target event's content to current popular or trending topics; Collect user attention data for the target event; Identify key influencers or opinion leaders involved in and disseminating the target event; Collect other external data that may influence the development of the target event.

[0072] In this embodiment, the above data specifically refers to: 1. Content Analysis: The BERT model is used to analyze the sentiment of the event content, including the event's reputation, credibility, responsibility, and public opinion. Ten of the most-viewed articles are selected, and sentiment analysis is performed on these articles to obtain the corresponding metrics.

[0073] 2. Trend Relevance: Check the relevance of the content to current popular or trending topics. Content related to trending events or trends is more likely to go viral.

[0074] 3. User engagement prediction: During the event incubation period, observe and analyze user interactions, such as the speed and quantity of likes, comments, and shares.

[0075] 4. Social Network Analysis: Identify and analyze key influencers or opinion leaders who may be involved in dissemination.

[0076] 5. External influences: Whether there is external intervention, holidays, policy changes, natural disasters, and economic conditions, etc.

[0077] The volume of event dissemination predicted by the random forest regression model will serve as an important indicator for assessing the impact of the target event on the enterprise, and the public opinion risk level of the target event will be determined based on the decision-making results of the public opinion situation.

[0078] In a preferred embodiment of this invention, the step of calculating the public opinion risk score based on the volume of dissemination and the trend of public opinion dissemination specifically includes: A public opinion risk scoring model is constructed based on the grey clustering method. The scores of each risk dimension are standardized to construct an event-risk assessment matrix. The risk level is then divided into several grey class levels. The membership degree of each public opinion event's score to each grey class is calculated. The membership degrees of each risk dimension are weighted to obtain the overall score of the target event in different grey classes, thus determining the final risk level of the target event.

[0079] In this embodiment, the data composition of the public opinion risk scoring model Public opinion heat level: The public opinion heat level is quantified based on the predicted volume of event dissemination.

[0080] Event types: such as new product launches, marketing events, or crisis events, all of which greatly affect a brand's short-term and long-term influence.

[0081] Sentiment and Emotion: Analyze the sentiment (positive, neutral, negative) of public opinion content using Natural Language Processing (NLP) technology.

[0082] Dissemination scope: Assess the breadth of the event's dissemination across different channels such as social media, news websites, and forums, and whether it has received widespread attention.

[0083] Severity of the event: Analyze the potential impact of the public opinion event on the brand or enterprise using natural language processing (NLP) technology.

[0084] The construction of a public opinion risk scoring model using the grey clustering method can be achieved through the following steps: Data standardization: Each risk dimension (such as public opinion heat, sentiment tendency, etc.) is represented by a standardized score from 0 to 1, so that the scores of each dimension are comparable.

[0085] Construct an evaluation matrix: For each public opinion event, collect standardized scores from various risk dimensions to form an event-risk dimension matrix.

[0086] Divide the risk level into three gray categories, such as low, medium, and high risk, and determine an appropriate range for each gray category. The following ranges can be used: low risk (0~0.3), medium risk (0.3~0.7), and high risk (0.7~1).

[0087] Calculate the membership degree r: assess the probability (membership degree) that the score of each public opinion event belongs to each gray class, that is, measure the matching degree of the event in a certain gray class.

[0088] Comprehensive weighted scoring: The membership degree of each risk dimension is weighted and calculated to obtain the overall score of the event in different gray categories, thereby determining the final risk level.

[0089] [1] Output a comprehensive score: Based on the comprehensive score, the overall risk level of the event is determined, with higher values ​​indicating greater risk. This model can provide the risk level of public opinion events, supporting subsequent decision-making.

[0090] As a preferred embodiment of this practice, the specific method for automatically recommending or generating response strategies and measures based on intelligent analysis results is as follows: Based on historical data and a response strategy database, corresponding strategies are recommended, and the database is continuously optimized through feedback information.

[0091] In this embodiment, the various coping strategies are as follows: Pre-emptive response strategies: This module utilizes an intelligent early warning system to identify potential public opinion risks in advance and adjust the risk level promptly based on the latest public opinion dynamics. For public opinion issues arising from business operations, companies can proactively conduct public relations campaigns in conjunction with their production and operational status.

[0092] Taking Shanghai's electricity bill as an example, given the cold wave that hit Shanghai in December 2023 and the resulting surge in electricity consumption as residents sought warmth, a press release could be prepared in advance. This release could highlight the company's proactive efforts to ensure power supply during the peak winter season, particularly its commitment to guaranteeing power supply and fulfilling its corporate social responsibility. By disseminating information through authoritative channels, the press release conveys the company's determination and actions to ensure energy stability, while also effectively guiding public opinion towards energy conservation and advocating for rational electricity use to avoid waste.

[0093] In-process response strategies: Based on different levels of public opinion risk, corresponding tiered response measures can be adopted to efficiently manage public opinion while ensuring optimal resource allocation. The following are recommendations for measures at low, medium, and high risk levels: Low risk: Monitoring: Maintain daily monitoring, track changes in public opinion, and prevent the situation from escalating.

[0094] Preventative communication: Release positive information in a timely manner, actively guide public opinion, and prevent the spread of negative information.

[0095] Regular reporting: Low-risk events will be included in regular public opinion analysis reports to help identify trends.

[0096] Medium risk: Increase monitoring frequency: Increase the frequency of monitoring, track the development of public opinion in real time, and identify potential risks of dissemination.

[0097] Communication intervention: Respond promptly through official platforms to clarify facts or correct misunderstandings, thereby reducing the potential spread of negative emotions.

[0098] Crisis preparedness: Initial preparations for activating emergency plans, including developing multiple response options to prevent further escalation of the incident.

[0099] Public opinion guidance: Appropriately strengthen positive information guidance, and cooperate with the media when necessary to ensure that the incident does not escalate into a high-risk situation.

[0100] High risk: Immediately activate the emergency response: quickly establish a public opinion management team or crisis management team to unify command and decision-making.

[0101] Authoritative Releases: We release accurate and transparent information promptly through authoritative media or official platforms, proactively addressing doubts and clarifying misunderstandings.

[0102] Multi-channel communication: Release official statements and facts through multiple channels to respond to public concerns in a timely manner in order to prevent the spread of rumors.

[0103] Strengthen public interaction: Engage with the public through social media platforms, follow up on feedback in real time, and quell negative emotions.

[0104] Review and Improvement: Conduct a comprehensive review after the incident has subsided, analyze the shortcomings in public opinion management, formulate improvement measures, and enhance future emergency management capabilities.

[0105] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A public opinion early warning and intelligent analysis system based on a data middle platform, characterized in that, include: The public opinion data collection module collects public opinion data, weather forecast data, and enterprise operation data from various information sources, both internal and external to the enterprise. The data processing and cleaning platform module cleans the collected public opinion data, unifies heterogeneous data, classifies and labels various data, and integrates them in chronological order. The public opinion topic classification module creates public opinion classification standards based on the company's target themes and constructs public opinion subcategories; The intelligent analysis module classifies public opinion data by labels based on the BERT model, and combines the random forest regression model and the Lotka-Volterra model to predict the volume and trend of public opinion dissemination. Then, based on the volume of dissemination and the trend of public opinion dissemination, a public opinion risk score is calculated, and an intelligent early warning signal is generated; The public opinion early warning module, based on weather forecasts and enterprise operation data, pre-sets early warning trigger conditions and monitors public opinion risks in conjunction with intelligent early warning signals; The intelligent response strategy and measures module automatically recommends or generates response strategies and measures based on intelligent analysis results.

2. The public opinion early warning and intelligent analysis system based on a data middle platform according to claim 1, characterized in that, The method for collecting public opinion data, weather forecast data, and enterprise operation data from multiple information sources, whether internal or external, is as follows: Data is obtained from external sources through web scraping, API interfaces, and RSS subscriptions; data is also synchronized with internal enterprise systems on a scheduled or real-time basis through API integration.

3. The public opinion early warning and intelligent analysis system based on a data middle platform according to claim 1, characterized in that, The method for unifying heterogeneous data, classifying and labeling various data, and integrating them in chronological order after cleaning and collecting public opinion data is as follows: External unstructured data is converted into a unified format through data mapping and unified coding standards, integrating data from different sources into data of a unified format; Add corresponding labels to each type of data according to its data type; Data from different sources is integrated in chronological order, linking external public opinion dynamics with internal events.

4. The public opinion early warning and intelligent analysis system based on a data middle platform according to claim 1, characterized in that, The pre-set warning trigger conditions, based on weather forecasts and enterprise operational data, are as follows: Warning thresholds for service outages triggered by extreme weather and fault warning thresholds for abnormal enterprise operational data.

5. The public opinion early warning and intelligent analysis system based on a data middle platform according to claim 1, characterized in that, The specific steps for classifying public opinion data by labels based on the BERT model are as follows: Based on the BERT model, unstructured data in public opinion data is automatically stored in layers according to classification labels; among them, video files in public opinion data are converted into text by an automatic speech recognition tool, and then the text is classified.

6. The public opinion early warning and intelligent analysis system based on a data middle platform according to claim 1, characterized in that, The data that still needs to be collected in the step of classifying public opinion data by labels based on the BERT model and predicting the volume of public opinion dissemination using the random forest regression model includes: The BERT model is used to analyze the sentiment of the target event content, including the event's reputation, credibility, accountability, and public opinion. Check the relevance of the target event's content to current popular or trending topics; Collect user attention data for the target event; Identify key influencers or opinion leaders involved in and disseminating the target event; Collect other external data that may influence the development of the target event.

7. The public opinion early warning and intelligent analysis system based on a data middle platform according to claim 1, characterized in that, The specific steps for calculating the public opinion risk score based on the volume of dissemination and the trend of public opinion dissemination are as follows: A public opinion risk scoring model is constructed based on the grey clustering method. The scores of each risk dimension are standardized to construct an event-risk assessment matrix. The risk level is then divided into several grey class levels. The membership degree of each public opinion event's score to each grey class is calculated. The membership degrees of each risk dimension are weighted to obtain the overall score of the target event in different grey classes, thus determining the final risk level of the target event.

8. The public opinion early warning and intelligent analysis system based on a data middle platform according to claim 1, characterized in that, The specific method for automatically recommending or generating response strategies and measures based on intelligent analysis results is as follows: Based on historical data and a response strategy database, corresponding strategies are recommended, and the database is continuously optimized through feedback information.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a public opinion early warning and intelligent analysis system based on a data platform as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements a public opinion early warning and intelligent analysis system based on a data platform as described in any one of claims 1 to 8.

Images