Financial activity public opinion data fusion system based on knowledge graph

Abstract

The application belongs to the field of financial technology, and specifically relates to a financial activity public opinion data fusion system based on a knowledge graph, steps of which are as follows: step one: an intelligent public opinion data fusion and processing module, which extracts public opinion data from multiple data sources, performs standardized processing, adjusts data source weights for fusion; step two: a natural language processing in a risk feature extraction and intelligent analysis module extracts risk features in public opinion data; step three: a knowledge graph construction and complex network analysis module constructs a knowledge graph of local illegal financial activities; step four: a deep neural network risk prediction and situation awareness module uses a deep neural network to train public opinion data; step five: according to a deep learning model and network analysis results, risk early warning is triggered and public opinion changes are monitored; step six: a risk feedback and system optimization module collects risk feedback and adjusts model parameters; and step seven: a report generation and analysis system generates a detailed risk report according to risk early warning results.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to a financial activity public opinion data fusion system based on knowledge graphs. Background Art

[0002] The supervision of local illegal financial activities faces the huge challenge of multi-source data. With the surge in online information, traditional financial monitoring methods can no longer cope with the huge and complex unstructured data (such as social media content, news and public opinion). These illegal activities often affect the public and financial markets through opaque capital flows, false advertising and media manipulation, and are usually closely related to financial institutions, non-financial enterprises, illegal financial organizations, etc. Existing technologies focus on monitoring methods based on structured data, which have problems such as narrow monitoring scope, low accuracy and information lag.

[0003] How to effectively integrate multi-source data (especially social media and online public opinion) and extract potential risks of illegal financial activities has become a key challenge in local supervision of illegal financial activities.

[0004] Therefore, it is particularly urgent to explore a new way to integrate financial public opinion data in order to achieve efficient and accurate real-time monitoring.

[0005] In response to the above-mentioned technical defects, a financial activity public opinion data fusion system solution based on knowledge graph is proposed. Summary of the Invention

[0006] In order to solve the above problems, the present invention provides the following technical solutions:

[0007] The financial activity public opinion data fusion system based on knowledge graph includes the public opinion data intelligent fusion and processing module, risk feature extraction and intelligent analysis module, knowledge graph construction and complex network analysis module, deep neural network risk prediction and situation awareness module, real-time warning and monitoring module, risk feedback and system optimization module and report generation and analysis module.

[0008] The public opinion data intelligent fusion and processing module includes a data acquisition unit, a data processing unit, and a data fusion unit, which provides high-quality data input for the risk feature extraction module through real-time capture, cleaning, and fusion of multi-source data;

[0009] The risk feature extraction and intelligent analysis module includes a public opinion data feature extraction unit, a risk indicator scoring unit, and a sentiment analysis and public opinion heat analysis unit. It uses natural language processing technology and deep learning models to intelligently extract potential risk information from public opinion texts processed by the public opinion data intelligent fusion and processing module.

[0010] The knowledge graph construction and complex network analysis module extracts potential risk information in the public opinion text based on the risk feature extraction and intelligent analysis module, constructs a knowledge graph of local illegal financial activities in real time, dynamically updates nodes and relationships, and identifies risk transmission paths through network analysis;

[0011] The deep neural network risk prediction and situation awareness module includes a deep neural network based on the transmission path analyzed by the knowledge graph construction and complex network analysis module, simultaneously based on the deep neural network for real-time training of public opinion data, prediction of illegal financial activity risks, and dynamic adjustment of model parameters to optimize prediction results;

[0012] The real-time early warning and monitoring module includes a real-time risk warning unit and a public opinion change monitoring unit, which triggers a risk warning in real time and monitors public opinion changes according to the deep learning model and network analysis results in the deep neural network risk prediction and situation awareness module, and adjusts the warning level;

[0013] The risk feedback and system optimization module includes a feedback mechanism unit and a dynamic optimization and adaptive learning unit, which evaluates the accuracy of the prediction according to the feedback results of the actual illegal financial events, and uses the feedback data for model optimization through adaptive learning to adjust the model parameters;

[0014] The report generation and analysis module automatically generates detailed risk reports according to the risk warning results, provides decision support for regulatory agencies and financial institutions, and saves historical data and analysis results for long-term tracking and prediction of the trend of illegal financial activities.

[0015] Further, the public opinion data intelligent fusion and processing module includes a data acquisition unit, a data fusion unit and a data processing unit, the data acquisition unit acquires data from various social media, news websites, financial advertising platforms and records them, and defines the collected data as D, the data processing unit standardizes the collected data to form a standardized public opinion data set Normalize(D i (t)), the data fusion unit dynamically adjusts the weight of each data source by calculating the correlation and importance between data sources through the core algorithm of data fusion, the formula of public opinion data fusion is as follows:

[0016]

[0017] Where D fusion (t) represents the fusion data at time point t, D i (t), represents the public opinion data of the i-th data source, α i(t) represents the weight of the i-th data source, which is dynamically adjusted as time t changes. Normalize(D i (t)) indicates the standardization of the public opinion data of the i-th data source.

[0018] Weight α i (t) Adjustments are made by analyzing the real-time response speed, accuracy, content credibility, and immediate impact of each data source on public opinion. For example, social media may have a stronger influence on illegal financial activities at a certain moment, while news websites may have a smaller influence. In this case, the weight of social media will be dynamically increased to ensure that the integrated data reflects the most accurate public opinion situation.

[0019] Furthermore, the risk feature extraction and intelligent analysis module includes a public opinion data feature extraction unit, a risk index scoring unit, and a sentiment analysis and public opinion heat analysis unit. The public opinion data feature extraction unit uses natural language processing technology to analyze public opinion data, identify sentiment tendencies, risk-related keywords, and public opinion hotspots, and simultaneously applies modeling technology to extract risk features related to illegal financial activities from the data. The real-time feature extraction modeling formula is as follows:

[0020]

[0021] where R score (t) Risk score at time point t, indicating the overall risk level of the current public opinion data; ω j (t), represents the weight of the j-th risk feature; f j (D(t)) represents the extraction function of the jth feature; D(t) represents the public opinion data set at time point t; m represents the total number of risk features.

[0022] The weight ω of each feature j (t) will be dynamically adjusted according to its timeliness and impact on illegal financial activities. For example, the popularity of public opinion about certain illegal activities may rise sharply in a short period of time, resulting in an increase in the weight of risk features related to popularity. The feature extraction function f j (D(t)) will extract each feature from the current public opinion data in real time and assign corresponding scores according to its importance.

[0023] Furthermore, the knowledge graph construction and complex network analysis module builds and updates the knowledge graph of local illegal financial activities in real time, identifies the associations between financial institutions, non-financial enterprises, public opinion events, and equity relationship nodes. The system updates the relationship between graph nodes and edges according to the changes of each data point, timely reflects the complex network structure of financial activities, and uses complex network analysis algorithms to predict risk transmission paths. The real-time knowledge graph construction formula is specifically analyzed as follows:

[0024] G(t)=(V(t),E(t),W(t))

[0025] Where G(t) represents the knowledge graph constructed at time point t; V(t) represents the set of nodes in the graph at time point t; E(t) represents the set of edges between nodes at time point t, and W(t) represents the weight of the edge at time point t.

[0026] The knowledge graph is a dynamically updated process. Whenever new public opinion data or financial activity information appears, the system will reflect the changes in the local illegal financial activity network in real time by updating the nodes and edges in the graph. If a financial institution is linked to an illegal financial activity, the system will update the network structure in real time by adding corresponding edges and adjusting weights, revealing potential risk transmission paths.

[0027] Furthermore, the deep neural network risk prediction and situational awareness module includes a deep neural network, which processes public opinion data in real time and uses deep neural networks and convolutional neural networks to predict risks. The model is trained in real time based on historical public opinion data and the occurrence pattern of risk events, and can dynamically adjust the prediction results and provide risk trends of illegal financial activities in a timely manner. The real-time risk prediction formula is specifically analyzed as follows:

[0028]

[0029] Where L(θ(t)) represents the loss function at time point t; y i represents the true label of the i-th sample; x i (t) The characteristics of the i-th sample; f(x i (t),θ(t)) represents the sample x at time point t i (t) is the predicted value; θ(t) is the model parameter, which is the real-time parameter at time point t; λ is the regularization coefficient, which is used to control the complexity of the model; N is the number of samples.

[0030] In the deep learning model, the loss function L(θ(t)) is dynamically calculated over time, using historical data x i (t) and the current parameter θ(t) to predict risk. With each data input, the model parameters are optimized in real time based on the actual results.

[0031] Furthermore, the real-time warning and monitoring module includes a real-time risk warning unit and a public opinion change monitoring unit, and defines L(θ(t)) 阈值 is a safe value, and L(θ(t)) 阈值Compare the results with those of L(θ(t)) and trigger risk warnings in real time based on the deep learning model and network analysis results within the deep neural network risk prediction and situational awareness module. This also monitors changes in public opinion and adjusts the warning level.

[0032] The public opinion change monitoring unit tracks and monitors changes in relevant public opinion in real time, and monitors whether new illegal financial activities may cause greater financial risks.

[0033] Furthermore, the risk feedback and system optimization module includes a feedback mechanism unit and a dynamic optimization and adaptive learning unit. Based on the data from the real-time warning and monitoring module, the feedback mechanism unit monitors the actual occurrence of illegal financial activities in real time. The system collects feedback information and verifies the prediction results of the model. Define L(θ(t)) 标准 , represents the standard value of the difference between the predicted result and the actual label, L(θ(t)) 标准 Compared with L(θ(t)), if the prediction is wrong, the feedback data will be used to adjust the model parameters.

[0034] Furthermore, the dynamic optimization and adaptive learning unit automatically adjusts the model parameters and data processing strategy based on the feedback data from the feedback mechanism unit. The specific real-time optimization formula is:

[0035] θ new (t) = θ old (t)-η·▽ θ L(θ(t))

[0036] θ new (t) represents the updated model parameters at time point t; θ old (t) represents the model parameters before time point t; η represents the learning rate, which controls the step size of parameter update; ▽ θ L(θ(t)) represents the gradient of the loss function with respect to the model parameters, indicating the influence of the model parameters on the loss function; L(θ(t)) represents the loss function.

[0037] The system calculates the gradient of the loss function θ L(θ(t)), and adjust the model parameters θ based on real-time data feedback new (t) Ensure that the system is continuously optimized to adapt to new public sentiment and patterns of illegal financial activities.

[0038] Furthermore, the report generation and analysis module: the system automatically generates detailed risk reports based on risk warning results, provides decision support to regulators and financial institutions, and saves historical data and analysis results to facilitate long-term tracking and prediction of trends in illegal financial activities.

[0039] Furthermore, the method for fusion of financial activity public opinion data based on knowledge graph according to claim 1 is characterized in that the system for fusion of financial activity public opinion data based on knowledge graph according to any one of claims 1 to 9 is adopted, and the steps are as follows:

[0040] Step 1: Utilize the public opinion data intelligent fusion and processing module to capture public opinion data from multiple data sources (such as social media, news websites, etc.) in real time, clean and standardize it, and dynamically adjust the weights of the data sources for fusion.

[0041] Step 2: After the data is integrated, the risk features in the public opinion data are extracted through natural language processing in the risk feature extraction and intelligent analysis module, and the risk score is calculated and sentiment and popularity analysis is performed;

[0042] Step 3: The knowledge graph construction and complex network analysis module is then used to construct a knowledge graph of local illegal financial activities in real time based on the data extracted by the risk feature extraction and intelligent analysis modules. The nodes and relationships are dynamically updated, and the risk propagation paths are identified through network analysis.

[0043] Step 4: The deep neural network risk prediction and situational awareness module then uses the knowledge graph construction and network analysis of the complex network analysis module to identify risk propagation path data. It uses the deep neural network to conduct real-time training on public opinion data, predict the risk of illegal financial activities, and dynamically adjust model parameters to optimize the prediction results.

[0044] Step 5: The prediction results of the deep neural network risk prediction and situational awareness module are compared with the thresholds set by real-time warning and monitoring. Based on the deep learning model and network analysis results, risk warnings are triggered in real time, and changes in public opinion are monitored to adjust the warning level.

[0045] Step 6: Based on the data from the real-time warning and monitoring module, the risk feedback and system optimization module collects actual risk feedback, adjusts model parameters for adaptive optimization, and improves future recognition accuracy;

[0046] Step 7: Report generation and analysis: The system automatically generates a detailed risk report based on the risk warning results and provides decision support to regulators and financial institutions.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. In the financial activity public opinion data fusion system based on knowledge graph of the present invention, the present invention aggregates public opinion data from different platforms through multi-source data fusion technology, so that the system can more comprehensively and accurately identify the risks of illegal financial activities.

[0049] 2. The present invention's knowledge graph-based financial activity public opinion data fusion system, based on deep learning and knowledge graphs, achieves dynamic monitoring and situational awareness of complex financial activities, significantly improving the accuracy and real-time nature of monitoring.

[0050] 3. In the financial activity public opinion data fusion system based on the knowledge graph of the present invention, the system continuously improves its recognition ability through continuous feedback and optimization, has self-learning and adaptability, and can continuously respond to new forms of illegal financial activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0052] Figure 1 This is a schematic diagram of the overall steps of the financial activity public opinion data fusion system based on knowledge graph of the present invention. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0054] like Figure 1 As shown in the figure, the financial activity public opinion data fusion system based on knowledge graph includes public opinion data intelligent fusion and processing module, risk feature extraction and intelligent analysis module, knowledge graph construction and complex network analysis module, deep neural network risk prediction and situation awareness module, real-time warning and monitoring module, risk feedback and system optimization module and report generation and analysis module.

[0055] The public opinion data intelligent fusion and processing module includes a data acquisition unit, a data processing unit, and a data fusion unit. It provides high-quality data input for the risk feature extraction module through real-time capture, cleaning, and fusion of multi-source data.

[0056] The risk feature extraction and intelligent analysis module includes a public opinion data feature extraction unit, a risk indicator scoring unit, and a sentiment analysis and public opinion heat analysis unit. Through natural language processing technology and deep learning models, it intelligently extracts potential risk information from public opinion texts processed by the public opinion data intelligent fusion and processing module.

[0057] The knowledge graph construction and complex network analysis module builds a knowledge graph of local illegal financial activities in real time based on the potential risk information in public opinion texts intelligently extracted by the risk feature extraction and intelligent analysis module, dynamically updates nodes and relationships, and identifies risk propagation paths through network analysis.

[0058] The deep neural network risk prediction and situational awareness module includes a deep neural network, based on the knowledge graph construction and the propagation path analyzed by the complex network analysis module. At the same time, it uses the deep neural network to train public opinion data in real time to predict the risk of illegal financial activities and dynamically adjust the model parameters to optimize the prediction results.

[0059] The real-time warning and monitoring module includes a real-time risk warning unit and a public opinion change monitoring unit. Based on the deep learning model and network analysis results in the deep neural network risk prediction and situation awareness module, it triggers risk warnings in real time, monitors public opinion changes, and adjusts the warning level.

[0060] The risk feedback and system optimization module includes a feedback mechanism unit and a dynamic optimization and adaptive learning unit. The system evaluates the accuracy of predictions based on feedback from actual illegal financial incidents, uses the feedback data for model optimization, and adjusts model parameters through adaptive learning.

[0061] Report generation and analysis module: The system automatically generates detailed risk reports based on risk warning results, providing decision support to regulators and financial institutions, while saving historical data and analysis results for subsequent long-term tracking and prediction of trends in illegal financial activities.

[0062] Specifically, the public opinion data intelligent fusion and processing module includes a data collection unit, a data fusion unit, and a data processing unit. The data collection unit collects data from various channels such as major social media, news websites, and financial advertising platforms, and records the collected data, and defines the collected data as D. The data processing unit normalizes the collected data to form a standardized public opinion data set Normalize (D i (t));

[0063] Specifically, the data fusion unit dynamically adjusts the weight of each data source by calculating the correlation and importance between data sources through the core algorithm of data fusion. The formula for public opinion data fusion is as follows:

[0064]

[0065] Among them D fusion (t) represents the fusion data at time point t, D i (t), represents the public opinion data of the i-th data source, α i(t) represents the weight of the ith data source, which is dynamically adjusted as the change of time t, Normalize(D i (t)) represents the standardization processing of the public opinion data of the ith data source.

[0066] The weight α i (t) is adjusted by analyzing the real-time response speed, accuracy, content credibility and immediate impact on public opinion of each data source. For example, social media may have stronger influence on illegal financial activities at a certain moment, while the influence of news websites may be smaller. In this case, the weight of social media will be dynamically increased to ensure that the fused data reflects the most accurate public opinion situation.

[0067] The risk feature extraction and intelligent analysis module includes a public opinion data feature extraction unit, a risk index scoring unit and an emotional analysis and public opinion heat analysis unit. The public opinion data feature extraction unit analyzes the public opinion data using natural language processing technology to identify the emotional tendency, risk-related keywords and public opinion hotspots in the data. At the same time, modeling technology is applied to extract risk features related to illegal financial activities from the data. The real-time feature extraction modeling formula is as follows:

[0068]

[0069] where R score (t) is the risk score at time point t, indicating the overall risk level of the current public opinion data; ω j (t), represents the weight of the jth risk feature; f j (D(t)) represents the extraction function of the jth feature; D(t) represents the public opinion data set at time point t; and m represents the total number of risk features.

[0070] The weight ω j (t) of each feature is dynamically adjusted according to its timeliness and impact on illegal financial activities. For example, the public opinion heat of some illegal activities may rise sharply in a short period of time, resulting in an increase in the weight of the risk features related to the heat. The feature extraction function f j (D(t)) will extract each feature from the current public opinion data in real time and assign a corresponding score according to its importance.

[0071] Specifically, the knowledge graph construction and complex network analysis module establishes and updates the knowledge graph of local illegal financial activities in real time, identifies the correlation between financial institutions, non-financial enterprises, public opinion events and equity relationship nodes, and updates the graph nodes and edge relationships according to the changes of each data point, timely reflects the complex financial activity network structure, and predicts the risk propagation path using complex network analysis algorithms. The real-time knowledge graph construction formula is analyzed as follows:

[0072] G(t)=(V(t),E(t),W(t))

[0073] Where G(t) represents the knowledge graph constructed at time point t; V(t) represents the set of nodes in the graph at time point t; E(t) represents the set of edges between nodes at time point t, and W(t) represents the weight of the edge at time point t.

[0074] The knowledge graph is a dynamically updated process. Whenever new public opinion data or financial activity information appears, the system will reflect the changes in the local illegal financial activity network in real time by updating the nodes and edges in the graph. If a financial institution is linked to an illegal financial activity, the system will update the network structure in real time by adding corresponding edges and adjusting weights, revealing potential risk transmission paths.

[0075] Specifically, the deep neural network risk prediction and situational awareness module includes a deep neural network. The deep neural network processes public opinion data in real time and uses deep neural networks and convolutional neural networks to predict risks. The model is trained in real time based on historical public opinion data and the occurrence patterns of risk events. It can dynamically adjust the prediction results and provide risk trends of illegal financial activities in a timely manner. The real-time risk prediction formula is specifically analyzed as follows:

[0076]

[0077] Where L(θ(t)) represents the loss function at time point t; y i represents the true label of the i-th sample; x i (t) The characteristics of the i-th sample; f(x i (t),θ(t)) represents the sample x at time point t i (t) is the predicted value; θ(t) is the model parameter, which is the real-time parameter at time point t; λ is the regularization coefficient, which is used to control the complexity of the model; N is the number of samples.

[0078] In the deep learning model, the loss function L(θ(t)) is dynamically calculated over time, using historical data x i (t) and the current parameter θ(t) to predict risk. With each data input, the model parameters are optimized in real time based on the actual results.

[0079] Specifically, the real-time warning and monitoring module includes a real-time risk warning unit and a public opinion change monitoring unit. Define L(θ(t)) 阈值 is a safe value, and L(θ(t)) 阈值Compare the results with those of L(θ(t)) and trigger risk warnings in real time based on the deep learning model and network analysis results within the deep neural network risk prediction and situational awareness module. This also monitors changes in public opinion and adjusts the warning level.

[0080] The public opinion change monitoring unit tracks and monitors changes in relevant public opinion in real time, and monitors whether new illegal financial activities may cause greater financial risks.

[0081] Specifically, the risk feedback and system optimization module includes a feedback mechanism unit and a dynamic optimization and adaptive learning unit. Based on the data from the real-time warning and monitoring module, the feedback mechanism unit monitors the actual occurrence of illegal financial activities in real time. The system collects feedback information and verifies the prediction results of the model. Define L(θ(t)) 标准 , represents the standard value of the difference between the predicted result and the actual label, L(θ(t)) 标准 Compared with L(θ(t)), if the prediction is wrong, the feedback data will be used to adjust the model parameters.

[0082] Specifically, the dynamic optimization and adaptive learning unit automatically adjusts the model parameters and data processing strategies based on the feedback data from the feedback mechanism unit. The specific real-time optimization formula is:

[0083] θ new (t) = θ old (t)-η·▽ θ L(θ(t))

[0084] θ new (t) represents the updated model parameters at time point t; θ old (t) represents the model parameters before time point t; η represents the learning rate, which controls the step size of parameter update; ▽ θ L(θ(t)) represents the gradient of the loss function with respect to the model parameters, indicating the influence of the model parameters on the loss function; L(θ(t)) represents the loss function.

[0085] The system calculates the gradient of the loss function θ L(θ(t)), and adjust the model parameters θ based on real-time data feedback new (t) Ensure that the system is continuously optimized to adapt to new public sentiment and patterns of illegal financial activities.

[0086] Specifically, the report generation and analysis module: the system automatically generates detailed risk reports based on risk warning results, provides decision support to regulators and financial institutions, and saves historical data and analysis results for subsequent long-term tracking and prediction of trends in illegal financial activities.

[0087] Specifically, the method for fusion of financial activity public opinion data based on knowledge graph according to claim 9 is characterized in that the system for fusion of financial activity public opinion data based on knowledge graph according to any one of claims 1 to 9 is adopted, and the steps are as follows:

[0088] Step 1: Utilize the public opinion data intelligent fusion and processing module to capture public opinion data from multiple data sources (such as social media, news websites, etc.) in real time, clean and standardize it, and dynamically adjust the weights of the data sources for fusion.

[0089] Step 2: After the data is integrated, the risk features in the public opinion data are extracted through natural language processing in the risk feature extraction and intelligent analysis module, and the risk score is calculated and sentiment and popularity analysis is performed;

[0090] Step 3: The knowledge graph construction and complex network analysis module is then used to construct a knowledge graph of local illegal financial activities in real time based on the data extracted by the risk feature extraction and intelligent analysis modules. The nodes and relationships are dynamically updated, and the risk propagation paths are identified through network analysis.

[0091] Step 4: The deep neural network risk prediction and situational awareness module then uses the knowledge graph construction and network analysis of the complex network analysis module to identify risk propagation path data. It uses the deep neural network to conduct real-time training on public opinion data, predict the risk of illegal financial activities, and dynamically adjust model parameters to optimize the prediction results.

[0092] Step 5: The prediction results of the deep neural network risk prediction and situational awareness module are compared with the thresholds set by real-time warning and monitoring. Based on the deep learning model and network analysis results, risk warnings are triggered in real time, and changes in public opinion are monitored to adjust the warning level.

[0093] Step 6: Based on the data from the real-time warning and monitoring module, the risk feedback and system optimization module collects actual risk feedback, adjusts model parameters for adaptive optimization, and improves future recognition accuracy;

[0094] Step 7: Report generation and analysis: The system automatically generates a detailed risk report based on the risk warning results and provides decision support to regulators and financial institutions.

[0095] Working principle: Using the intelligent fusion and processing module of public opinion data, public opinion data is captured from multiple data sources (such as social media, news websites, etc.) in real time, and then cleaned and standardized to form a standardized public opinion data set called Normalize (D i (t)), and dynamically adjust the weight of the data source for fusion. The formula for public opinion data fusion is as follows:

[0096]

[0097] wherein D fusion (t) represents the fusion data at time point t, D i (t) represents the public opinion data of the i-th data source, a i (t) represents the weight of the i-th data source, which is dynamically adjusted as time t changes, and Normalize(D i (t)) represents the standardization processing of the public opinion data of the i-th data source; after fusion, the data is extracted through the natural language processing in the risk feature extraction and intelligent analysis module to extract the risk features in the public opinion data, and the real-time feature extraction modeling formula is as follows:

[0098]

[0099] wherein R score (t) represents the risk score at time point t, indicating the overall risk degree of the current public opinion data; ω j (t) represents the weight of the j-th risk feature; f j (D(t)) represents the extraction function of the j-th feature; D(t) represents the public opinion data set at time point t; and m represents the total number of risk features; then, the knowledge graph construction and complex network analysis module is used to construct the knowledge graph of local illegal financial activities based on the data extracted by the risk feature extraction and intelligent analysis module, dynamically update the nodes and relationships, and identify the risk propagation path through network analysis, and the real-time knowledge graph construction formula is analyzed as follows:

[0100] G(t) = (V(t), E(t), W(t))

[0101] wherein G(t) represents the knowledge graph constructed at time point t; V(t) represents the node set in the graph at time point t; E(t) represents the edge set between nodes at time point t, and W(t) represents the weight of the edge at time point t; then, the deep neural network risk prediction and situation awareness module is used to identify the risk propagation path data based on the network analysis of the knowledge graph construction and complex network analysis module, use the deep neural network to train the public opinion data in real time, predict the illegal financial activity risk, and dynamically adjust the model parameters to optimize the prediction results, and the real-time risk prediction formula is analyzed as follows:

[0102]

[0103] wherein L(θ(t)) represents the loss function at time point t; y i represents the true label of the i-th sample; x i (t) represents the feature of the i-th sample; f(x i (t), θ(t)) represents the prediction result of the i-th sample at time point t; and x i(t) is the predicted value; θ(t) is the model parameter, the real-time parameter at time point t; λ is the regularization coefficient, which is used to control the complexity of the model; N is the number of samples; then the prediction results of the deep neural network risk prediction and situation awareness module are compared with the threshold set by the real-time warning and monitoring, that is, L(θ(t)) is defined 阈值 is a safe value, and L(θ(t)) 阈值 Compare with the detection results of L(θ(t)), trigger risk warnings in real time based on the deep learning model and network analysis results, monitor changes in public opinion, and adjust the warning level; then, based on the data of the real-time warning and monitoring module, the risk feedback and system optimization module collects actual risk feedback, adjusts model parameters for adaptive optimization, and uses the real-time optimization formula:

[0104]

[0105] θ new (t) represents the updated model parameters at time point t; θ old (t) represents the model parameters before time point t; η represents the learning rate, which controls the step size of parameter update; It represents the gradient of the loss function with respect to the model parameters, and the influence of the model parameters on the loss function; L(θ(t)) represents the loss function; then the report generation and analysis module is used: the system automatically generates a detailed risk report based on the risk warning results, and provides decision support to regulators and financial institutions.

[0106] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. The financial activity public opinion data fusion system based on knowledge graph is characterized by: It includes the public opinion data intelligent fusion and processing module, risk feature extraction and intelligent analysis module, knowledge graph construction and complex network analysis module, deep neural network risk prediction and situation awareness module, real-time warning and monitoring module, risk feedback and system optimization module, and report generation and analysis module; The public opinion data intelligent fusion and processing module includes a data acquisition unit, a data processing unit, and a data fusion unit, which provides high-quality data input for the risk feature extraction module through real-time capture, cleaning, and fusion of multi-source data; The risk feature extraction and intelligent analysis module includes a public opinion data feature extraction unit, a risk indicator scoring unit, and a sentiment analysis and public opinion heat analysis unit. It uses natural language processing technology and deep learning models to intelligently extract potential risk information from public opinion texts processed by the public opinion data intelligent fusion and processing module. The knowledge graph construction and complex network analysis module builds a knowledge graph of local illegal financial activities in real time based on the potential risk information in the public opinion text intelligently extracted by the risk feature extraction and intelligent analysis module, dynamically updates nodes and relationships, and identifies risk propagation paths through network analysis; The deep neural network risk prediction and situational awareness module includes a deep neural network, which is based on the knowledge graph construction and the propagation path analyzed by the complex network analysis module. At the same time, it uses the deep neural network to train public opinion data in real time to predict the risk of illegal financial activities and dynamically adjust the model parameters to optimize the prediction results. The real-time warning and monitoring module includes a real-time risk warning unit and a public opinion change monitoring unit. Based on the deep learning model and network analysis results in the deep neural network risk prediction and situation awareness module, it triggers risk warnings in real time, monitors public opinion changes, and adjusts the warning level. The risk feedback and system optimization module includes a feedback mechanism unit and a dynamic optimization and adaptive learning unit. The system evaluates the accuracy of predictions based on the feedback results of actual illegal financial incidents, and uses the feedback data for model optimization and adjusts model parameters through adaptive learning. The report generation and analysis module: The system automatically generates detailed risk reports based on risk warning results, provides decision support to regulators and financial institutions, and saves historical data and analysis results to facilitate long-term tracking and prediction of trends in illegal financial activities.

2. The financial activity public opinion data fusion system based on knowledge graph according to claim 1 is characterized in that: The public opinion data intelligent fusion and processing module includes a data collection unit, a data fusion unit and a data processing unit. The data collection unit collects data from various channels such as major social media, news websites and financial advertising platforms for recording, and defines the collected data as The data processing unit performs standardization processing on the collected data to form a standardized public opinion data set The data fusion unit dynamically adjusts the weight of each data source by calculating the correlation and importance between data sources through the core algorithm of data fusion. The formula for public opinion data fusion is as follows: ； in represents the fused data at time point t, , represents the public opinion data of the i-th data source, Represents the weight of the i-th data source, which is dynamically adjusted as time t changes. Indicates that the public opinion data of the i-th data source is standardized.

3. The financial activity public opinion data fusion system based on knowledge graph according to claim 2 is characterized in that: The risk feature extraction and intelligent analysis module includes a public opinion data feature extraction unit, a risk index scoring unit, and a sentiment analysis and public opinion heat analysis unit. The public opinion data feature extraction unit uses natural language processing technology to analyze public opinion data, identify sentiment tendencies, risk-related keywords, and public opinion hotspots, and simultaneously applies modeling technology to extract risk features related to illegal financial activities from the data. The real-time feature extraction modeling formula is as follows: ； in The risk score at time point t represents the overall risk level of the current public opinion data; , represents the weight of the j-th risk feature; represents the extraction function of the jth feature; represents the public opinion data set at time point t; Represents the total number of risk characteristics.

4. The financial activity public opinion data fusion system based on knowledge graph according to claim 3 is characterized in that: The knowledge graph construction and complex network analysis module builds and updates the knowledge graph of local illegal financial activities in real time, identifying the associations between financial institutions, non-financial enterprises, public opinion events, and equity relationship nodes. The system updates the relationship between graph nodes and edges based on the changes in each data point, promptly reflecting the complex network structure of financial activities and using complex network analysis algorithms to predict risk transmission paths. The real-time knowledge graph construction formula is specifically analyzed as follows: ； in Represents the knowledge graph constructed at time point t; represents the set of nodes in the graph at time point t; represents the set of edges between nodes at time point t, represents the weight of the edge at time point t.

5. The financial activity public opinion data fusion system based on knowledge graph according to claim 4 is characterized in that: The deep neural network risk prediction and situational awareness module includes a deep neural network, which processes public opinion data in real time and uses deep neural networks and convolutional neural networks to predict risks. The model is trained in real time based on historical public opinion data and the occurrence patterns of risk events. It can dynamically adjust the prediction results and provide risk trends of illegal financial activities in a timely manner. The real-time risk prediction formula is specifically analyzed as follows: ； in Represents the loss function at time point t; represents the true label of the i-th sample; The characteristics of the i-th sample; Indicates the sample at time point t The predicted value of Model parameters, real-time parameters at time point t; Represents the regularization coefficient, which is used to control the complexity of the model; Indicates the sample size.

6. The financial activity public opinion data fusion system based on knowledge graph according to claim 5 is characterized in that: The real-time warning and monitoring module includes a real-time risk warning unit and a public opinion change monitoring unit. is a safe value, and and The system compares the detection results with those of the previous ones, triggers risk warnings in real time based on the deep learning model and network analysis results in the deep neural network risk prediction and situation awareness module, monitors changes in public opinion, and adjusts the warning level; The public opinion change monitoring unit tracks and monitors changes in relevant public opinion in real time, and monitors whether new illegal financial activities may cause greater financial risks.

7. The financial activity public opinion data fusion system based on knowledge graph according to claim 6 is characterized in that: The risk feedback and system optimization module includes a feedback mechanism unit and a dynamic optimization and adaptive learning unit. Based on the data of the real-time warning and monitoring module, the feedback mechanism unit monitors the actual occurrence of illegal financial activities in real time. The system collects feedback information and verifies the prediction results of the model. , represents the standard value of the difference between the predicted result and the actual label, If the prediction is wrong, the feedback data will be used to adjust the model parameters.

8. The financial activity public opinion data fusion system based on knowledge graph according to claim 7 is characterized in that: The dynamic optimization and adaptive learning unit automatically adjusts model parameters and data processing strategies based on the feedback data from the feedback mechanism unit. The specific real-time optimization formula is: ； Represents the updated model parameters at time point t; Represents the model parameters before time point t; Represents the learning rate, which controls the step size of parameter updates; Represents the gradient of the loss function with respect to the model parameters, indicating the influence of the model parameters on the loss function; represents the loss function.

9. The financial activity public opinion data fusion system based on knowledge graph according to claim 7 is characterized in that: The report generation and analysis module: The system automatically generates detailed risk reports based on risk warning results, provides decision support to regulators and financial institutions, and saves historical data and analysis results to facilitate long-term tracking and prediction of trends in illegal financial activities.

10. The method for integrating public opinion data of financial activities based on knowledge graph is characterized by: The financial activity public opinion data fusion system based on knowledge graph described in any one of claims 1 to 9 is adopted, and the steps are as follows: Step 1: Utilize the public opinion data intelligent fusion and processing module to capture public opinion data from multiple data sources in real time, perform cleaning and standardization, and dynamically adjust the weights of data sources for fusion; Step 2: After the data is integrated, the risk features in the public opinion data are extracted through natural language processing in the risk feature extraction and intelligent analysis module, and the risk score is calculated and sentiment and popularity analysis is performed; Step 3: The knowledge graph construction and complex network analysis module is then used to construct a knowledge graph of local illegal financial activities in real time based on the data extracted by the risk feature extraction and intelligent analysis modules. The nodes and relationships are dynamically updated, and the risk propagation paths are identified through network analysis. Step 4: The deep neural network risk prediction and situational awareness module then uses the knowledge graph construction and network analysis of the complex network analysis module to identify risk propagation path data. It uses the deep neural network to conduct real-time training on public opinion data, predict the risk of illegal financial activities, and dynamically adjust model parameters to optimize the prediction results. Step 5: The prediction results of the deep neural network risk prediction and situational awareness module are compared with the thresholds set by real-time warning and monitoring. Based on the deep learning model and network analysis results, risk warnings are triggered in real time, and changes in public opinion are monitored to adjust the warning level. Step 6: Based on the data from the real-time warning and monitoring module, the risk feedback and system optimization module collects actual risk feedback, adjusts model parameters for adaptive optimization, and improves future recognition accuracy; Step 7: Report generation and analysis: The system automatically generates a detailed risk report based on the risk warning results and provides decision support to regulators and financial institutions.

Images