Data value discovery and risk assessment system and method based on large language model

Through a data value discovery and risk assessment system based on large language models, the data silos and real-time requirements of enterprises in data management and analysis are solved, and the integrity of data sets and analysis efficiency is improved, and risks are identified in a timely manner. Scientific decision-making.

CN119047828AInactive Publication Date: 2024-11-29网靖(北京)科技有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411129354.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Enterprises have data silos in the data management and analysis process, resulting in misleading insights and decision-making. Traditional analysis methods cannot meet real-time needs, especially in terms of abnormal monitoring and risk control.

Method used

The data value discovery and risk assessment system based on large language models is adopted, including multi-source integration module, value discovery module, risk identification module and visualization module, and the topic semantic information is extracted by real-time extraction and prediction of missing data, application of LDA model and LSTM model, Build an isolated tree to calculate exception scores and generate interactive reports using Tableau.

Benefits of technology

Significantly improve the integrity and analysis efficiency of data sets, reduce analysis errors, help enterprises make scientific strategic decisions, identify potential risks in a timely manner, and improve customer satisfaction and flexibility in data analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119047828A_ABST
    Figure CN119047828A_ABST
Patent Text Reader

Abstract

The invention relates to a data value discovery and risk assessment system and method based on a large language model, in particular to the field of data element markets and data asset operation, and aims to extract business data, management data and operation data in real time, establish a prediction missing model, classify the extracted data according to prediction missing results and improve the risk assessment accuracy. The integrity of a data set is remarkably improved, analysis errors caused by missing data are reduced, topic semantic information and corresponding service data, management data and operation data are extracted by applying an LDA model, data analysis is more focused and effective, an enterprise is helped to make a more scientific strategic decision, and the enterprise development efficiency is improved. According to the method, an LSTM unit is selected to build a recurrent neural network model to output a prediction and discovery result for the value of future trends and market opportunities, an isolated tree is built to calculate abnormal scores of service data, management data and operation data, a report is automatically generated by using Tableau and data display is performed, and complex data is presented in a visual manner, so that a user can understand and analyze the data conveniently.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data factor market and data asset operation, and more specifically, to a data value discovery and risk assessment system and method based on a large language model. Background Art

[0002] With the rise of the Internet of Things, social media and online transactions, the amount of data is growing exponentially. There is a lot of potential value hidden in the massive data, and this exponential growth has brought unprecedented opportunities and challenges. Many companies have begun to realize that data is not just an accessory to business operations, but an important asset to drive strategic decision-making.

[0003] Currently, enterprises have data silos in the process of data management and analysis, and are unable to centrally manage data. Massive data contains a lot of wrong information, leading to misleading insights and decisions. In a rapidly changing market environment, traditional analysis methods cannot meet real-time requirements and are insufficient for abnormal monitoring and risk control. Summary of the invention

[0004] In view of the technical problems existing in the prior art, the present invention provides a data value discovery and risk assessment system and method based on a large language model to solve the problems raised in the above background technology.

[0005] The technical solution of the present invention to solve the above technical problems is as follows: a data value discovery and risk assessment system based on a large language model, including a multi-source integration module, a value discovery module, a risk identification module and a visualization module;

[0006] Multi-source integration module: extract business data, management data and operation data in real time, establish a prediction missing model to classify business data, management data and operation data according to the prediction missing results, use heat maps to visualize the missing status of business data, management data and operation data and fill them in;

[0007] Value discovery module: Apply the LDA model to extract topic semantic information and corresponding business data, management data, and operation data, and select LSTM units to build a recurrent neural network model to output the prediction and discovery results of the value of future trends and market opportunities;

[0008] Risk identification module: Apply large language models to convert business data, management data, and operation data into structured data and extract enterprise data features, build an isolation tree to calculate the anomaly scores of business data, management data, and operation data based on the average path length;

[0009] Visualization module: Use Tableau to automatically generate reports and display data, create interactive dashboards that integrate multiple visualization elements, and provide filtering functions to dynamically adjust visualization elements according to different enterprise dimensions;

[0010] In a preferred embodiment, the multi-source integration module is divided into a collection subunit and a processing subunit, which specifically includes the following contents:

[0011] Collection subunit: Use the SQLAlchemy tool to extract business data, management data, and operation data from the enterprise's internal database in real time and record timestamps. Use Python's requests library to interact with the RESTful API. Use ETL tools to automatically extract, convert, and load business data, management data, and operation data.

[0012] Processing subunit: Apply classification label encoding to business data, management data and operation data and convert them into numerical form, establish a prediction missing model and build a prediction tree structure, randomly and with replacement extract multiple data subsets from business data, management data and operation data, for each data subset, use the internal nodes of the tree structure to represent the features and the leaf nodes to represent the predicted missing results to build a decision tree, randomly select representative features to find the best split, repeat iterations until 80% of the decision trees of the data subsets are generated, use RandomForestClassifier to classify according to the predicted missing results, use RandomForestRegr essor to predict according to the representative features, use heat maps to visualize the missing conditions of business data, management data and operation data and fill them in.

[0013] In a preferred embodiment, the value discovery module is divided into a topic identification subunit, a value prediction subunit and a cluster classification subunit, which specifically include the following contents:

[0014] Topic identification subunit: The large language model is used to embed business data, management data, and operation data and convert them into vector representations. The word frequency-inverse document frequency algorithm is used to calculate the keyword weights of the topic semantic information of business data, management data, and operation data. The specific formula is as follows:

[0015]

[0016] TF-IDF(n) represents the keyword weight of the subject semantic information of business data, management data, and operation data. n represents the number of times the subject semantic information appears in business data, management data, and operation data, C represents the total number of semantic information in business data, management data, and operation data, M represents the total number of business data, management data, and operation data, Sn It represents the number of business data, management data and operation data containing subject semantic information, and applies the LDA model to extract the subject semantic information and the corresponding business data, management data and operation data.

[0017] Value prediction subunit: Select LSTM unit to build a recurrent neural network model. According to the hidden state of the subject semantic information corresponding to the business data, management data and operation data input at the current timestamp and the business data, management data and operation data at the previous timestamp, the activation value of the input layer is calculated through the Sigmoid activation function. The specific formula is:

[0018] T l =S(w l *[I t-1 ,x t ]+b l )

[0019] Among them, T l represents the activation value of the input layer, S() represents the Sigmoid activation function, and w l represents the weight matrix, I t-1 represents the hidden state of the subject semantic information corresponding to the business data, management data, and operation data of the previous timestamp, x t Indicates the business data, management data, and operation data input at the current timestamp, b l Represents the bias term. The output layer selects the fully connected layer and calculates the activation value of the output layer through the Sigmoid activation function according to the business data, management data and operation data input at the current timestamp, the hidden state of the subject semantic information corresponding to the business data, management data and operation data of the previous timestamp, and the cell state of the business data, management data and operation data input at the current timestamp. The specific formula is:

[0020] T o =S(w o *[I t-1 ,x t ]+b o )

[0021] Among them, T o represents the activation value of the forgetting layer, S() represents the Sigmoid activation function, and w o represents the weight matrix, I t-1 represents the hidden state of the subject semantic information corresponding to the business data, management data, and operation data of the previous timestamp, x t Indicates the business data, management data, and operation data input at the current timestamp, b oIt represents the bias term, and outputs the subject semantic information corresponding to the business data, management data, and operation data through the fully connected layer to predict the value of future trends and market opportunities.

[0022] Cluster classification subunit: Displays the distribution and relationship of the predicted results of the value of future trends and market opportunities for the subject semantic information corresponding to business data, management data and operation data through heat maps. Uses the K-means clustering algorithm to randomly select K initial cluster centers, assigns the predicted results to the cluster center closest to it, updates the cluster center and uses the average value of each cluster as the new cluster center. Repeats the iterative assignment and update operations until the cluster center no longer changes and the maximum number of iterations is reached.

[0023] In a preferred embodiment, the risk identification module applies a large language model to convert business data, management data and operation data into structured data and extract enterprise data features, randomly select enterprise data features and corresponding segmentation values, divide the structured data into two sub-data according to the selected segmentation value, recursively repeat until each sub-data contains an enterprise data feature, repeatedly iterate to build an isolated tree, for each sub-data, count the path lengths of all isolated trees to the isolated point, the path length is the number of edges that the sub-data passes through from the root node of the isolated tree to the isolated point, and calculate the abnormal scores of business data, management data and operation data according to the average path length of the sub-data, and the specific formula is:

[0024]

[0025] Among them, Score represents the anomaly score, E(h(x)) represents the average path length of the sub-data, and n represents the size of the sub-data.

[0026] In a preferred embodiment, the visualization module uses Tableau to automatically generate reports and display data, uses a heat map to display the distribution and relationship of the predicted results of the value of future trends and market opportunities of the subject semantic information corresponding to the business data, management data and operation data, uses a line graph to display the trend that changes with timestamps, uses a pie chart to display the proportion of business data, management data and operation data in the whole, uses a scatter plot to display the impact relationship between business data, management data and operation data and risks, creates an interactive dashboard integrating multiple visualization element diagrams, and provides a filtering function to dynamically adjust the visualization element diagrams according to different enterprise dimensions.

[0027] In a preferred embodiment, the method specifically includes the following steps:

[0028] S101. Extract business data, management data and operation data in real time, establish a prediction missing model to classify business data, management data and operation data according to the prediction missing results, use a heat map to visualize the missing conditions of business data, management data and operation data and fill them in;

[0029] S102. Apply the LDA model to extract topic semantic information and the corresponding business data, management data and operation data, use the word frequency-inverse document frequency algorithm to calculate the keyword weights of the topic semantic information of the business data, management data and operation data, select LSTM units to build a recurrent neural network model to output the topic semantic information corresponding to the business data, management data and operation data to predict the value of future trends and market opportunities, and use the K-means clustering algorithm to identify different user groups about enterprise services;

[0030] S103. Apply the large language model to convert the business data, management data and operation data into structured data and extract enterprise data features, build an isolation tree to calculate the anomaly scores of the business data, management data and operation data according to the average path length;

[0031] S104. Use Tableau to automatically generate reports and display data, create interactive dashboards that integrate multiple visualization elements, and provide filtering functions to dynamically adjust visualization elements according to different enterprise dimensions;

[0032] The beneficial effects of the present invention are as follows: by extracting and predicting missing data in real time, the application of missing models continuously monitors and fills missing data, significantly improves the integrity of the data set, reduces analysis errors caused by missing data, ensures data consistency and timeliness through an automated ETL process, reduces manual operation time and improves overall work efficiency, effectively extracts subject semantic information in business data, management data and operation data through the application of the LDA model, makes data analysis more focused and effective, helps enterprises make more scientific strategic decisions, provides customized services for the needs and behaviors of different users through the K-means clustering algorithm, effectively improves customer satisfaction and loyalty, applies a large language model to convert unstructured business data, management data and operation data into structured data, improves data availability and analysis efficiency, builds an isolated tree and calculates anomaly scores, can efficiently identify potential abnormal data, timely discover risk points, help enterprises take preventive measures, automatically generates reports and data displays using Tableau, can present complex data in an intuitive way, facilitates user understanding and analysis, creates interactive dashboards to dynamically adjust visualization elements as needed, and improves the flexibility of data analysis and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of the method of the present invention;

[0034] Figure 2 This is a system structure block diagram of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0036] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0037] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details to obscure the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0038] Example 1

[0039] This embodiment provides Figure 1 A data value discovery and risk assessment system based on a large language model is shown, which specifically includes the following steps:

[0040] S101. Extract business data, management data and operation data in real time, establish a prediction missing model to classify business data, management data and operation data according to the prediction missing results, use a heat map to visualize the missing conditions of business data, management data and operation data and fill them in;

[0041] S102. Apply the LDA model to extract topic semantic information and the corresponding business data, management data and operation data, use the word frequency-inverse document frequency algorithm to calculate the keyword weights of the topic semantic information of the business data, management data and operation data, select LSTM units to build a recurrent neural network model to output the topic semantic information corresponding to the business data, management data and operation data to predict the value of future trends and market opportunities, and use the K-means clustering algorithm to identify different user groups about enterprise services;

[0042] S103. Apply the large language model to convert the business data, management data and operation data into structured data and extract enterprise data features, build an isolation tree to calculate the anomaly scores of the business data, management data and operation data according to the average path length;

[0043] S104. Use Tableau to automatically generate reports and display data, create interactive dashboards that integrate multiple visualization elements, and provide filtering functions to dynamically adjust visualization elements according to different enterprise dimensions.

[0044] Example 2

[0045] This embodiment provides Figure 2 A data value discovery and risk assessment method based on a large language model is shown, specifically comprising: a multi-source integration module, a value discovery module, a risk identification module and a visualization module;

[0046] Multi-source integration module: extract business data, management data and operation data in real time, establish a prediction missing model to classify business data, management data and operation data according to the prediction missing results, use heat maps to visualize the missing status of business data, management data and operation data and fill them in;

[0047] Value discovery module: Apply the LDA model to extract topic semantic information and corresponding business data, management data, and operation data, and select LSTM units to build a recurrent neural network model to output the prediction and discovery results of the value of future trends and market opportunities;

[0048] Risk identification module: Apply large language models to convert business data, management data, and operation data into structured data and extract enterprise data features, build an isolation tree to calculate the anomaly scores of business data, management data, and operation data based on the average path length;

[0049] Visualization module: Use Tableau to automatically generate reports and display data, create interactive dashboards that integrate multiple visualization elements, and provide filtering functions to dynamically adjust visualization elements according to different enterprise dimensions;

[0050] S101. Extract business data, management data and operation data in real time, establish a prediction missing model to classify business data, management data and operation data according to the prediction missing results, use a heat map to visualize the missing conditions of business data, management data and operation data and fill them in;

[0051] Furthermore, the SQLAlchemy tool is used to extract business data, management data, and operation data from the enterprise's internal database in real time and record timestamps. The Python requests library is used to interact with the RESTful API, and the ETL tool is used to automatically extract, transform, and load business data, management data, and operation data.

[0052] Furthermore, the classification label encoding of business data, management data and operation data is converted into numerical form, a prediction missing model is established and a prediction tree structure is built, and multiple data subsets are randomly extracted with replacement from the business data, management data and operation data. For each data subset, a decision tree is constructed using the internal nodes of the tree structure to represent the features and the leaf nodes to represent the predicted missing results. The representative features are randomly selected to find the best split, and the iterations are repeated until 80% of the decision trees of the data subsets are generated. RandomForestClassifier is used to classify according to the predicted missing results, and RandomForestRegressor is used to predict according to the representative features. Heat maps are used to visualize the missing information of business data, management data and operation data and fill them in.

[0053] S102. Apply the LDA model to extract topic semantic information and the corresponding business data, management data and operation data, use the word frequency-inverse document frequency algorithm to calculate the keyword weights of the topic semantic information of the business data, management data and operation data, select LSTM units to build a recurrent neural network model to output the topic semantic information corresponding to the business data, management data and operation data to predict the value of future trends and market opportunities, and use the K-means clustering algorithm to identify different user groups about enterprise services;

[0054] Furthermore, a large language model is applied to embed the business data, management data, and operation data and convert them into vector representations to help capture the subject semantic information of the business data, management data, and operation data. The keyword weights of the subject semantic information of the business data, management data, and operation data are calculated using the word frequency-inverse document frequency algorithm. The specific formula is:

[0055]

[0056] TF-IDF(n) represents the keyword weight of the subject semantic information of business data, management data, and operation data. nrepresents the number of times the subject semantic information appears in business data, management data, and operation data, C represents the total number of semantic information in business data, management data, and operation data, M represents the total number of business data, management data, and operation data, S n It represents the number of business data, management data and operation data containing subject semantic information, and applies the LDA model to extract the subject semantic information and the corresponding business data, management data and operation data.

[0057] Furthermore, an LSTM unit is selected to build a recurrent neural network model. According to the hidden state of the subject semantic information corresponding to the business data, management data and operation data input at the current timestamp and the business data, management data and operation data at the previous timestamp, the activation value of the input layer is calculated through the Sigmoid activation function. The specific formula is:

[0058] T l =S(w l *[I t-1 ,x t ]+b l )

[0059] Among them, T l represents the activation value of the input layer, S() represents the Sigmoid activation function, and w l represents the weight matrix, I t-1 represents the hidden state of the subject semantic information corresponding to the business data, management data, and operation data of the previous timestamp, x t Indicates the business data, management data, and operation data input at the current timestamp, b l Represents the bias term. The output layer selects the fully connected layer and calculates the activation value of the output layer through the Sigmoid activation function according to the business data, management data and operation data input at the current timestamp, the hidden state of the subject semantic information corresponding to the business data, management data and operation data of the previous timestamp, and the cell state of the business data, management data and operation data input at the current timestamp. The specific formula is:

[0060] T o =S(w o *[I t-1 ,x t ]+b o )

[0061] Among them, T o represents the activation value of the forgetting layer, S() represents the Sigmoid activation function, and w o represents the weight matrix, I t-1 represents the hidden state of the subject semantic information corresponding to the business data, management data, and operation data of the previous timestamp, x tIndicates the business data, management data, and operation data input at the current timestamp, b o It represents the bias term, and outputs the subject semantic information corresponding to the business data, management data, and operation data through the fully connected layer to predict the value of future trends and market opportunities.

[0062] Furthermore, the distribution and relationship of the predicted discovery results of the value of future trends and market opportunities based on the subject semantic information corresponding to business data, management data and operation data are displayed through heat maps, helping enterprises to make adjustments in advance and seize market opportunities. The K-means clustering algorithm is used to randomly select K initial clustering centers. The initial clustering centers are related to different user groups of enterprise services, and the predicted discovery results are assigned to the clustering center closest to them. The clustering center is updated and the average value of each cluster is used as the new clustering center. The iterative allocation and update operations are repeated until the clustering center no longer changes and the maximum number of iterations is reached.

[0063] S103. Apply the large language model to convert the business data, management data and operation data into structured data and extract enterprise data features, build an isolation tree to calculate the anomaly scores of the business data, management data and operation data according to the average path length;

[0064] Furthermore, a large language model is applied to convert business data, management data and operation data into structured data and extract enterprise data features. Enterprise data features and corresponding segmentation values ​​are randomly selected. The structured data is divided into two sub-data according to the selected segmentation value. This process is repeated recursively until each sub-data contains an enterprise data feature. An isolated tree is constructed repeatedly. For each sub-data, the path length to the isolated point in all isolated trees is counted. The path length is the number of edges that the sub-data passes through from the root node of the isolated tree to the isolated point. The anomaly scores of business data, management data and operation data are calculated based on the average path length of the sub-data. The specific formula is:

[0065]

[0066] Among them, Score represents the anomaly score, E(h(x)) represents the average path length of the sub-data, n represents the size of the sub-data, and the anomaly risk score interval is defined as 0-100%. When the anomaly risk score is between 0-30%, it means that the business data, management data and operation data corresponding to the sub-data are low-risk anomaly data. When the anomaly risk score is between 30%-70%, it means that the business data, management data and operation data corresponding to the sub-data are medium-risk anomaly data. When the anomaly risk score is between 70%-100%, it means that the business data, management data and operation data corresponding to the sub-data are high-risk anomaly data.

[0067] S104. Use Tableau to automatically generate reports and display data, create interactive dashboards that integrate multiple visualization elements, and provide filtering functions to dynamically adjust visualization elements according to different enterprise dimensions;

[0068] Furthermore, Tableau is used to automatically generate reports and display data. Heat maps are used to display the distribution and relationship of the results of predictions of the value of future trends and market opportunities based on the subject semantic information corresponding to business data, management data, and operation data. Line graphs are used to display trends that change over timestamps. Pie charts are used to display the proportion of business data, management data, and operation data in the whole. Scatter plots are used to display the impact relationship between business data, management data, and operation data and risks. Interactive dashboards are created to integrate multiple visualization element graphs, and filtering functions are provided to dynamically adjust visualization element graphs according to different enterprise dimensions.

[0069] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0070] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0074] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0075] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A data value discovery and risk assessment system based on a large language model, characterized in that: It includes multi-source integration module, value discovery module, risk identification module and visualization module; Multi-source integration module: extract business data, management data and operation data in real time, establish a prediction missing model to classify business data, management data and operation data according to the prediction missing results, use heat maps to visualize the missing status of business data, management data and operation data and fill them in; Value discovery module: Apply the LDA model to extract topic semantic information and corresponding business data, management data, and operation data, and select LSTM units to build a recurrent neural network model to output the prediction and discovery results of the value of future trends and market opportunities; Risk identification module: Apply large language models to convert business data, management data, and operation data into structured data and extract enterprise data features, build an isolation tree to calculate the anomaly scores of business data, management data, and operation data based on the average path length; Visualization module: Use Tableau to automatically generate reports and display data, create interactive dashboards that integrate multiple visualization elements, and provide filtering functions to dynamically adjust visualization elements according to different enterprise dimensions.

2. According to claim 1, a data value discovery and risk assessment system based on a large language model is characterized in that: The multi-source integration module uses the SQLAlchemy tool to extract business data, management data and operation data from the enterprise's internal database in real time, records timestamps, uses Python's requests library to interact with the RESTful API, and uses the ETL tool to automatically extract, transform and load business data, management data and operation data; The classification label encoding of business data, management data and operation data is converted into numerical form, a prediction missing model is established and a prediction tree structure is built, and multiple data subsets are randomly extracted with replacement from the business data, management data and operation data. For each data subset, a decision tree is constructed using the internal nodes of the tree structure to represent the features and the leaf nodes to represent the predicted missing results. The representative features are randomly selected to find the best split, and the iterations are repeated until 80% of the decision trees of the data subsets are generated. RandomForest Classifier is used to classify according to the predicted missing results, and RandomForestRegressor is used to predict according to the representative features. Heat map is used to visualize the missing status of business data, management data and operation data and fill them in.

3. The data value discovery and risk assessment system based on a large language model according to claim 1, characterized in that: The value discovery module uses a large language model to embed business data, management data and operation data and convert them into vector representations, uses a word frequency-inverse document frequency algorithm to calculate the keyword weights of the subject semantic information of business data, management data and operation data, and uses an LDA model to extract the subject semantic information and the corresponding business data, management data and operation data; Select LSTM unit to build recurrent neural network model, and calculate the activation value of input layer through Sigmoid activation function according to the hidden state of subject semantic information corresponding to business data, management data and operation data input at current timestamp and business data, management data and operation data at previous timestamp. The specific formula is as follows: T l =S(w l *[I t-1 ,x t ]+b l ) Among them, T l represents the activation value of the input layer, S() represents the Sigmoid activation function, and w l represents the weight matrix, I t-1 represents the hidden state of the subject semantic information corresponding to the business data, management data, and operation data of the previous timestamp, x t Indicates the business data, management data, and operation data input at the current timestamp, b l Represents the bias term. The output layer selects the fully connected layer and calculates the activation value of the output layer through the Sigmoid activation function according to the business data, management data and operation data input at the current timestamp, the hidden state of the subject semantic information corresponding to the business data, management data and operation data of the previous timestamp, and the cell state of the business data, management data and operation data input at the current timestamp. The specific formula is: T o =S(w o *[I t-1 ,x t ]+b o ) Among them, T o represents the activation value of the forgetting layer, S() represents the Sigmoid activation function, and w o represents the weight matrix, I t-1 represents the hidden state of the subject semantic information corresponding to the business data, management data, and operation data of the previous timestamp, x t Indicates the business data, management data, and operation data input at the current timestamp, b o It represents the bias term, and outputs the subject semantic information corresponding to the business data, management data, and operation data through the fully connected layer to predict the value of future trends and market opportunities.

4. The data value discovery and risk assessment system based on a large language model according to claim 3, characterized in that: The specific formula for calculating the keyword weights of the subject semantic information of business data, management data and operation data by the word frequency-inverse document frequency algorithm is: TF-IDF(n) represents the keyword weight of the subject semantic information of business data, management data, and operation data. n represents the number of times the subject semantic information appears in business data, management data, and operation data, C represents the total number of semantic information in business data, management data, and operation data, M represents the total number of business data, management data, and operation data, S n Indicates the amount of business data, management data, and operational data that contain subject semantic information.

5. The data value discovery and risk assessment system based on a large language model according to claim 1, characterized in that: The risk identification module applies a large language model to convert business data, management data and operation data into structured data and extract enterprise data features, randomly selects enterprise data features and corresponding segmentation values, divides the structured data into two sub-data according to the selected segmentation value, and recursively repeats until each sub-data contains an enterprise data feature. Repeat the iteration to build an isolation tree. For each sub-data, the path length to the isolated point in all isolated trees is counted. The path length is the number of edges that the sub-data passes through from the root node of the isolation tree to the isolated point. The anomaly scores of the business data, management data and operation data are calculated based on the average path length of the sub-data.

6. The data value discovery and risk assessment system based on a large language model according to claim 5, characterized in that: The specific formula for calculating the anomaly scores of business data, management data, and operation data is: Among them, Score represents the anomaly score, E(h(x)) represents the average path length of the sub-data, and n represents the size of the sub-data.

7. The data value discovery and risk assessment system based on a large language model according to claim 1, characterized in that: The visualization module uses Tableau to automatically generate reports and display data, uses heat maps to display the distribution and relationship of the predicted results of the subject semantic information corresponding to business data, management data and operation data for the value of future trends and market opportunities, uses line graphs to display trends that change with timestamps, uses pie charts to display the proportion of business data, management data and operation data in the whole, uses scatter plots to display the impact relationship between business data, management data and operation data and risks, creates interactive dashboards that integrate multiple visualization element graphs, and provides filtering functions to dynamically adjust visualization element graphs according to different enterprise dimensions.

8. A data value discovery and risk assessment method based on a large language model is applied to a data value discovery and risk assessment system based on a large language model as described in any one of claims 1 to 7, characterized in that: The specific steps include: S101. Extract business data, management data and operation data in real time, establish a prediction missing model to classify business data, management data and operation data according to the prediction missing results, use a heat map to visualize the missing conditions of business data, management data and operation data and fill them in; S102. Apply the LDA model to extract topic semantic information and the corresponding business data, management data and operation data, use the word frequency-inverse document frequency algorithm to calculate the keyword weights of the topic semantic information of the business data, management data and operation data, select LSTM units to build a recurrent neural network model to output the topic semantic information corresponding to the business data, management data and operation data to predict the value of future trends and market opportunities, and use the K-means clustering algorithm to identify different user groups about enterprise services; S103. Apply the large language model to convert the business data, management data and operation data into structured data and extract enterprise data features, build an isolation tree to calculate the anomaly scores of the business data, management data and operation data according to the average path length; S104. Use Tableau to automatically generate reports and display data, create interactive dashboards that integrate multiple visualization elements, and provide filtering functions to dynamically adjust visualization elements according to different enterprise dimensions.