Data asset intelligent evaluation method and device based on large model

Abstract

The invention relates to the technical field of data asset assessment, and discloses a data asset intelligent assessment method and device based on a large model, and the method comprises the steps: obtaining target asset data; performing preprocessing operation on the target asset data to obtain feature asset data, and constructing a corresponding target knowledge graph; integrating the evaluation logic coding parameters to a target training model, and performing training operation on the target training model through a federated learning method and a target knowledge graph to obtain a target intelligent model; according to the target knowledge graph and the target intelligent model, generating a data value evaluation result corresponding to the target asset data; and obtaining real-time asset data, and performing adjustment operation on the data value evaluation result according to the real-time asset data to obtain a target asset evaluation result. Therefore, the accuracy and reliability of data asset assessment can be improved, and the intelligence and efficiency of data asset assessment can be improved.

Description

Technical Field

[0001] This invention relates to the field of data asset valuation technology, and in particular to a method and apparatus for intelligent valuation of data assets based on large models. Background Technology

[0002] With the booming development of the digital economy, the value of data assets is becoming increasingly prominent. Accurately assessing the value of data assets is crucial for enterprise decision-making, resource allocation, and market transactions. Traditional data asset valuation methods mainly include the cost approach, the income approach, and the market approach. However, these methods lack objective and unified standards for the selection of evaluation parameters and weight allocation, making them susceptible to human interference and affecting the accuracy and reliability of the evaluation results. Secondly, in today's era of data explosion, the amount of data is growing exponentially, while traditional methods are cumbersome and time-consuming in the data collection, cleaning, and analysis processes, making it difficult to cope with the dynamic changes of massive amounts of data, resulting in low efficiency in data asset valuation.

[0003] Therefore, it is particularly important to provide a new method for evaluating data assets to improve the accuracy and reliability of data asset evaluation, as well as to enhance the intelligence and efficiency of data asset evaluation. Summary of the Invention

[0004] This invention provides a data asset intelligent assessment method and apparatus based on a large model, which can help improve the accuracy and reliability of data asset assessment, as well as enhance the intelligence and efficiency of data asset assessment.

[0005] The first aspect of this invention discloses a data asset intelligent evaluation method based on a large model, the method comprising: Acquire target asset data, wherein the target asset data includes asset characteristic data and historical asset data; Preprocessing operations are performed on the target asset data to obtain feature asset data, and a target knowledge graph corresponding to the target asset data is constructed based on all the feature asset data. The predetermined evaluation logic encoding parameters are integrated into the predetermined target training model, and the predetermined target training model is trained using a preset federated learning method and the target knowledge graph to obtain the target intelligent model. Based on the target knowledge graph and the target intelligent model, a data value assessment result corresponding to the target asset data is generated; Acquire real-time asset data, and perform adjustment operations on the data value assessment results based on the real-time asset data to obtain the target asset assessment result.

[0006] As an optional implementation, in a first aspect of the present invention, the method further includes: Based on the target asset valuation results, determine the asset risk information corresponding to the target asset data; Based on the target asset valuation results and the asset risk information, visualized valuation information is generated, wherein the visualized valuation information includes one or more of the following: asset valuation report information corresponding to the target asset valuation results, asset valuation value trend information, and asset operation suggestion information; The visualized evaluation information is fed back to the target user terminal, and the evaluation feedback information corresponding to the target user terminal is obtained. The target intelligent model is then updated based on the evaluation feedback information.

[0007] As an optional implementation, in a first aspect of the present invention, the step of integrating predetermined evaluation logic encoding parameters into a predetermined target training model, and performing training operations on the predetermined target training model using a preset federated learning method and the target knowledge graph to obtain a target intelligent model includes: The predetermined evaluation logic encoding parameters are processed to obtain model encoding parameters, and all the model encoding parameters are integrated into the predetermined target training model. Perform data annotation operations on the target knowledge graph to obtain target annotation data, wherein the target annotation data includes asset value tag data; Using a pre-defined federated learning method and the target labeled data, a training operation is performed on the pre-determined target training model to obtain the parameter weights of each model parameter contained in the target training model. Based on each model parameter and the parameter weights corresponding to each model parameter, a target intelligent model is generated.

[0008] As an optional implementation, in a first aspect of the present invention, determining the asset risk information corresponding to the target asset data based on the target asset valuation result includes: Target indicator data is extracted from the target asset valuation results. The target indicator data includes data asset risk data that matches the target asset valuation results. The data asset risk data includes one or more of the following: asset integrity data, asset timeliness data, and asset scarcity data. By analyzing the data asset risk data using the target intelligent model, asset risk information corresponding to the target asset data is obtained; The asset risk information includes one or more of the following: asset compliance risk information, asset depreciation risk information, asset risk impact information, and asset risk probability information.

[0009] As an optional implementation, in the first aspect of the present invention, before acquiring real-time asset data, the method further includes: Invoke the asset interface call request, and generate the data call parameters corresponding to the asset interface call request based on the asset interface call request; The acquisition of real-time asset data includes: Based on the data call parameters, obtain the real-time asset data corresponding to the asset interface call request.

[0010] As an optional implementation, in a first aspect of the present invention, constructing a target knowledge graph corresponding to the target asset data based on all the said feature asset data includes: Perform data preprocessing operations on all the aforementioned feature asset data to obtain target feature data, wherein the data preprocessing operations include one or more of data cleaning operations, data denoising operations, and data standardization operations; Based on the target feature data, determine the graph nodes and the node connection relationships between each graph node, wherein the graph node includes one or more of the data type, data source, and data size of the target feature data, and the connection relationship includes one or more of the association strength and association type between each graph node; Based on each of the graph nodes and the node connection relationships between each of the graph nodes, a target knowledge graph corresponding to the target asset data is constructed.

[0011] As an optional implementation, in a first aspect of the present invention, the step of adjusting the data value assessment result based on the real-time asset data to obtain the target asset assessment result includes: The target intelligent model performs data processing operations on the real-time asset data to obtain the evaluation output result corresponding to the target intelligent model. The data processing operations include one or more of global feature extraction operations and complex data recognition operations. Based on the evaluation output, a local characteristic evaluation operation is performed on the real-time asset data to generate at least one local evaluation output, and a model evaluation output is generated based on all the local evaluation outputs. Based on a predetermined fusion strategy, a fusion processing operation is performed on the model evaluation output and the data value evaluation result to obtain an evaluation fusion result. Then, an adjustment operation is performed on the data value evaluation result based on the evaluation fusion result to obtain the target asset evaluation result.

[0012] A second aspect of this invention discloses a data asset intelligent evaluation device based on a large model, the device comprising: The acquisition module is used to acquire target asset data, wherein the target asset data includes asset characteristic data and historical asset data; The processing module is used to perform preprocessing operations on the target asset data to obtain feature asset data; A construction module is used to construct a target knowledge graph corresponding to the target asset data based on all the aforementioned feature asset data; The training module is used to integrate the pre-determined evaluation logic encoding parameters into the pre-determined target training model, and to perform training operations on the pre-determined target training model through a preset federated learning method and the target knowledge graph to obtain the target intelligent model; The generation module is used to generate a data value assessment result corresponding to the target asset data based on the target knowledge graph and the target intelligent model. The acquisition module is also used to acquire real-time asset data; The adjustment module is used to perform adjustment operations on the data value assessment results based on the real-time asset data to obtain the target asset assessment results.

[0013] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The determination module is used to determine the asset risk information corresponding to the target asset data based on the target asset valuation results. The generation module is further configured to generate visualized assessment information based on the target asset assessment results and the asset risk information, wherein the visualized assessment information includes one or more of the following: asset assessment report information corresponding to the target asset assessment results, asset assessment value trend information, and asset operation suggestion information; The feedback module is used to send the visualized evaluation information back to the target user terminal; The acquisition module is also used to acquire the evaluation feedback information corresponding to the target user terminal; The update module is used to perform a model update operation on the target intelligent model based on the evaluation feedback information.

[0014] As an optional implementation, in a second aspect of the present invention, the training module integrates pre-determined evaluation logic encoding parameters into a pre-determined target training model, and performs training operations on the pre-determined target training model using a preset federated learning method and the target knowledge graph to obtain the target intelligent model. Specific methods for obtaining the target intelligent model include: The predetermined evaluation logic encoding parameters are processed to obtain model encoding parameters, and all the model encoding parameters are integrated into the predetermined target training model. Perform data annotation operations on the target knowledge graph to obtain target annotation data, wherein the target annotation data includes asset value tag data; Using a pre-defined federated learning method and the target labeled data, a training operation is performed on the pre-determined target training model to obtain the parameter weights of each model parameter contained in the target training model. Based on each model parameter and the parameter weights corresponding to each model parameter, a target intelligent model is generated.

[0015] As an optional implementation, in a second aspect of the present invention, the specific method by which the determining module determines the asset risk information corresponding to the target asset data based on the target asset valuation result includes: Target indicator data is extracted from the target asset valuation results. The target indicator data includes data asset risk data that matches the target asset valuation results. The data asset risk data includes one or more of the following: asset integrity data, asset timeliness data, and asset scarcity data. By analyzing the data asset risk data using the target intelligent model, asset risk information corresponding to the target asset data is obtained; The asset risk information includes one or more of the following: asset compliance risk information, asset depreciation risk information, asset risk impact information, and asset risk probability information.

[0016] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The calling module is used to call the asset interface call request before the acquisition module obtains the real-time asset data; The generation module is also used to generate data call parameters corresponding to the asset interface call request based on the asset interface call request; The specific methods by which the acquisition module acquires real-time asset data include: Based on the data call parameters, obtain the real-time asset data corresponding to the asset interface call request.

[0017] As an optional implementation, in a second aspect of the present invention, the specific method by which the construction module constructs the target knowledge graph corresponding to the target asset data based on all the feature asset data includes: Perform data preprocessing operations on all the aforementioned feature asset data to obtain target feature data, wherein the data preprocessing operations include one or more of data cleaning operations, data denoising operations, and data standardization operations; Based on the target feature data, determine the graph nodes and the node connection relationships between each graph node, wherein the graph node includes one or more of the data type, data source, and data size of the target feature data, and the connection relationship includes one or more of the association strength and association type between each graph node; Based on each of the graph nodes and the node connection relationships between each of the graph nodes, a target knowledge graph corresponding to the target asset data is constructed.

[0018] As an optional implementation, in a second aspect of the present invention, the adjustment module performs an adjustment operation on the data value assessment result based on the real-time asset data to obtain the target asset assessment result in the following specific ways: The target intelligent model performs data processing operations on the real-time asset data to obtain the evaluation output result corresponding to the target intelligent model. The data processing operations include one or more of global feature extraction operations and complex data recognition operations. Based on the evaluation output, a local characteristic evaluation operation is performed on the real-time asset data to generate at least one local evaluation output, and a model evaluation output is generated based on all the local evaluation outputs. Based on a predetermined fusion strategy, a fusion processing operation is performed on the model evaluation output and the data value evaluation result to obtain an evaluation fusion result. Then, an adjustment operation is performed on the data value evaluation result based on the evaluation fusion result to obtain the target asset evaluation result.

[0019] A third aspect of this invention discloses another intelligent assessment device for data assets based on a large model, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the intelligent data asset evaluation method based on a large model according to any of the first aspects of the present invention.

[0020] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the data asset intelligent evaluation method based on a large model as described in any of the first aspects of the present invention.

[0021] Compared with the prior art, the present invention has the following beneficial effects: In this embodiment of the invention, target asset data is acquired; preprocessing operations are performed on the target asset data to obtain feature asset data, and a corresponding target knowledge graph is constructed; evaluation logic encoding parameters are integrated into the target training model, and training operations are performed on the target training model using a federated learning method and the target knowledge graph to obtain a target intelligent model; based on the target knowledge graph and the target intelligent model, a data value assessment result corresponding to the target asset data is generated; real-time asset data is acquired, and adjustment operations are performed on the data value assessment result based on the real-time asset data to obtain the target asset assessment result. Therefore, implementing this invention can improve the accuracy and reliability of data asset assessment, as well as enhance the intelligence and efficiency of data asset assessment. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a data asset intelligent evaluation method based on a large model disclosed in an embodiment of the present invention; Figure 2 This is a flowchart illustrating another intelligent data asset evaluation method based on a large model disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a data asset intelligent evaluation device based on a large model disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of another intelligent data asset evaluation device based on a large model disclosed in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of another intelligent data asset evaluation device based on a large model disclosed in an embodiment of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0027] This invention discloses an intelligent data asset assessment based on a large model, which can improve the accuracy and reliability of data asset assessment, as well as enhance its intelligence and efficiency. These will be described in detail below.

[0028] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a data asset intelligent evaluation method based on a large model, as disclosed in an embodiment of the present invention. Figure 1 The described intelligent data asset assessment method based on large models can be applied to intelligent data asset assessment devices based on large models. These devices can be integrated into cloud servers or local servers; this embodiment of the invention does not impose limitations. Figure 1 As shown, the intelligent assessment method for data assets based on large models can include the following operations.

[0029] 101. Obtain target asset data.

[0030] In this embodiment of the invention, the target asset data may optionally include asset characteristic data and historical asset data.

[0031] In this embodiment of the invention, optionally, the asset characteristic data includes one or more of the following: asset size characteristic data, asset quality characteristic data, and asset application scenario characteristic data; historical asset data includes one or more of the following: historical transaction data and historical evaluation data.

[0032] In this embodiment of the invention, optionally, the target asset data can be obtained by collecting data asset-related data from multiple sources such as enterprise databases, transaction records, and industry reports, including data scale, quality, application scenarios, market transaction data, etc., as the target asset data.

[0033] 102. Perform preprocessing operations on the target asset data to obtain feature asset data, and construct the target knowledge graph corresponding to the target asset data based on all feature asset data.

[0034] In this embodiment of the invention, optionally, the preprocessing operation performed on the target asset data to obtain the feature asset data may include: Preprocessing operations are performed on the target asset data to obtain preprocessed asset data, and feature data is extracted from the preprocessed asset data to obtain feature asset data. The preprocessing operations may include one or more of the following: cleaning operations, denoising operations, and standardization operations. The cleaning operations include removing invalid data and noise, the denoising operations include filtering outlier data, and the standardization operations include converting the data into a uniform format.

[0035] In this embodiment of the invention, optionally, the target knowledge graph corresponding to the target asset data may include entity identification information and relationship building information of the target asset data. The entity identification information includes entity information such as identifying data assets, enterprises, industries, policies and regulations, etc., and the relationship building information includes information on establishing association relationships between data assets and external entities.

[0036] 103. Integrate the pre-determined evaluation logic encoding parameters into the pre-determined target training model, and perform training operations on the pre-determined target training model through a preset federated learning method and target knowledge graph to obtain the target intelligent model.

[0037] In this embodiment of the invention, optionally, the pre-determined evaluation logic coding parameters may include one or more of the coding parameters corresponding to the cost approach, the income approach, and the market approach; wherein, the pre-determined evaluation logic coding parameters may also include logic coding parameters that encode the mathematical formula and logic of the cost approach into a differentiable form, logic coding parameters that encode the mathematical formula and logic of the income approach into a differentiable form, and logic coding parameters that encode the mathematical formula and logic of the market approach into a differentiable form.

[0038] In this embodiment of the invention, the target intelligent model can optionally be a DeepSeek large model.

[0039] 104. Based on the target knowledge graph and the target intelligent model, generate the data value assessment results corresponding to the target asset data.

[0040] In this embodiment of the invention, optionally, the above-mentioned generation of data value assessment results corresponding to target asset data based on the target knowledge graph and the target intelligent model may include: The target knowledge graph is input into the target intelligent model, which extracts the data asset features from the target knowledge graph and integrates them with the relationships in the target knowledge graph to obtain a comprehensive feature representation. Based on the comprehensive feature representation and according to the preset evaluation logic, the multi-dimensional value score of the data asset is calculated. The multi-dimensional value score includes, but is not limited to, the scarcity score, integrity score, timeliness score, and application scenario coverage score of the data asset. Based on multi-dimensional value scoring, a data value assessment result corresponding to the target asset data is generated. The data value assessment result includes one or more of the following: data asset value assessment result, value trend analysis, and risk warning.

[0041] 105. Obtain real-time asset data, and adjust the data value assessment results based on the real-time asset data to obtain the target asset assessment results.

[0042] In this embodiment of the invention, the real-time asset data may optionally include the latest asset data obtained from the enterprise's real-time data interface.

[0043] It is evident that implementation Figure 1 The described intelligent data asset assessment method based on a large model can acquire target asset data, perform preprocessing operations to obtain feature asset data, and then construct a target knowledge graph. The assessment logic encoding parameters are integrated into the target training model, and the target training model is trained using federated learning and the target knowledge graph to obtain a target intelligent model. Based on the target knowledge graph and the target intelligent model, corresponding data value assessment results are generated. Real-time asset data is acquired, and adjustments are made to the data value assessment results to obtain the target asset assessment results. Through the strong generalization and multimodal understanding capabilities of the large model, automated assessment of data asset value is achieved, improving the objectivity and consistency of the assessment. Federated learning technology supports rapid processing and real-time response of massive amounts of data. It can also combine real-time data streams and historical data to dynamically output assessment results, reflecting the dynamic value of data assets in real time. Furthermore, federated learning technology ensures data privacy and compliance during the assessment process. Combining real-time data and historical knowledge graphs, the large model generates multi-dimensional value scores and identifies potential risks of data assets based on the assessment results. It can dynamically adapt to changes in the value of data assets, thereby improving the accuracy, reliability, intelligence, and efficiency of data asset assessment.

[0044] Example 2 Please see Figure 2, Figure 2 This is a flowchart illustrating another intelligent data asset evaluation method based on a large model disclosed in an embodiment of the present invention. Figure 2 The described intelligent data asset assessment method based on large models can be applied to intelligent data asset assessment devices based on large models. These devices can be integrated into cloud servers or local servers; this embodiment of the invention does not impose limitations. Figure 2 As shown, this intelligent data asset assessment method based on a large model may include the following operations: 201. Obtain target asset data.

[0045] 202. Perform preprocessing operations on the target asset data to obtain feature asset data, and construct the target knowledge graph corresponding to the target asset data based on all feature asset data.

[0046] 203. Integrate the pre-determined evaluation logic encoding parameters into the pre-determined target training model, and perform training operations on the pre-determined target training model through a preset federated learning method and target knowledge graph to obtain the target intelligent model.

[0047] 204. Based on the target knowledge graph and the target intelligent model, generate the data value assessment results corresponding to the target asset data.

[0048] 205. Obtain real-time asset data, and perform adjustment operations on the data value assessment results based on the real-time asset data to obtain the target asset assessment results.

[0049] In this embodiment of the invention, for a detailed description of steps 201-205, please refer to the other descriptions of steps 101-105 in Embodiment 1. This embodiment of the invention will not repeat them.

[0050] 206. Based on the target asset valuation results, determine the asset risk information corresponding to the target asset data.

[0051] In this embodiment of the invention, optionally, the asset risk information corresponding to the target asset data may include one or more of compliance risk information, depreciation risk information, and operational risk information; wherein, compliance risk information may include the number of violations / data items: the number of violations in each stage of the data asset (such as collection, storage, and sharing); regulatory compliance degree: the score of the data asset's compliance with relevant regulatory requirements; and rectification cost: the human and technical investment cost required to fix compliance loopholes; depreciation risk information may include the data update cycle: the update frequency of the data asset (such as daily / Updated weekly; the longer the cycle, the lower the timeliness and the higher the risk of devaluation. Market supply growth rate of similar data: changes in the new supply of similar data assets in the market. A surge in supply may lead to a decrease in the scarcity of the target data asset. Probability of technological substitution: the probability that the use value of the target data asset will decrease due to the iteration of data processing technology (such as new storage technology and analysis algorithms). Operational risk information may include the similarity of historical violations: the degree of matching between the operation process of the target data asset and historical violations. The higher the similarity, the greater the risk probability. Number of process vulnerabilities: the number of security vulnerabilities in the data asset transaction process (such as authorization approval and access management). Emergency response timeliness: the emergency handling time for abnormal data asset operations (such as leakage and tampering). The longer the timeliness, the greater the potential risk loss.

[0052] 207. Generate visualized assessment information based on the target asset assessment results and asset risk information.

[0053] In this embodiment of the invention, the visualized assessment information includes one or more of the following: asset assessment report information corresponding to the target asset assessment result, asset assessment value trend information, and asset operation suggestion information.

[0054] In this embodiment of the invention, optionally, the asset valuation report information may include detailed asset value information, risk list information, and compliance recommendation information. Further, it may also include a report chapter structure (overview, valuation results, risk analysis, recommendations), data visualization components (bar charts, radar charts), and a compliance clause reference library. Asset valuation value trend information may include information on the changing trend of data asset value over time and scenario, based on historical valuation data and real-time market dynamics, displayed through line charts / heatmaps. Further, it may include time windows (last 3 months / 1 year), value fluctuation thresholds (exceeding 20% ​​triggers an alert), and scenario dimensions (business, region, industry). Asset operation recommendation information may include information on optimization strategies (such as data compliance rectification plans and asset disposal timing suggestions) provided for risk items based on the results of large-scale model inference. Further, it may include an operation strategy library (rectification, increase, and decrease strategy templates), strategy weights (dynamically adjusted according to risk level), and time constraints (such as rectification deadlines).

[0055] In this embodiment of the invention, optionally, the generation of visualized assessment information based on the target asset assessment results and asset risk information may include: Based on the target asset valuation results and asset risk information, determine the information to be interacted with, including information that needs to be fed back to the user based on the target asset valuation results and asset risk information. Based on all the information to be interacted with, generate visual evaluation information, which includes at least all the information to be interacted with.

[0056] 208. Feed back the visual evaluation information to the target user terminal, obtain the evaluation feedback information corresponding to the target user terminal, and perform model update operation on the target intelligent model based on the evaluation feedback information.

[0057] In this embodiment of the invention, optionally, the above-mentioned feeding back the visual evaluation information to the target user terminal and obtaining the evaluation feedback information corresponding to the target user terminal may include: The visual evaluation information is fed back to the target user terminal so that the target user corresponding to the target user terminal can view the visual evaluation information; Obtain evaluation feedback information generated by the target user on the target user's terminal in response to the visualized evaluation information. The evaluation feedback information includes one or more of the following: evaluation result satisfaction rating information, evaluation deviation information, and suggested correction information.

[0058] In this embodiment of the invention, optionally, the above-mentioned model update operation on the target intelligent model based on the evaluation feedback information may include: The evaluation feedback information is input into the target intelligent model, and the model is fine-tuned by adjusting parameters such as the learning rate decay factor and the weight of new samples. The optimization targets are the mean square error between the evaluation results and the actual transaction price, the accuracy of risk identification, and the improvement rate of model performance after the adoption of user feedback. The model parameters are iteratively updated to update the target intelligent model.

[0059] It is evident that implementation Figure 2The described intelligent data asset assessment method based on a large model can determine the asset risk information corresponding to the target asset data based on the target asset assessment results. It can identify potential risks of data assets in advance and generate visualized assessment information based on the target asset assessment results and asset risk information. Through visualization, users can intuitively understand the assessment results, enhancing their trust in the assessment process and results. The visualized assessment information is fed back to the target user terminal, and corresponding assessment feedback information is obtained from the target user terminal. Through user feedback, the assessment model can be adjusted and optimized in a timely manner, improving the model's accuracy and adaptability, enhancing the user experience. Based on the assessment feedback information, the target intelligent model is updated to achieve continuous model optimization. Through continuous data updates and the combination of user feedback, the model can maintain high performance and high accuracy. The feedback mechanism can also influence the optimization of the assessment model, thereby increasing trust in the system. This is conducive to improving the accuracy and reliability of data asset assessment, as well as enhancing the intelligence and efficiency of data asset assessment.

[0060] In an optional embodiment, predetermined evaluation logic encoding parameters are integrated into a predetermined target training model, and a training operation is performed on the predetermined target training model using a preset federated learning method and a target knowledge graph to obtain a target intelligent model, including: Perform parameter processing operations on the pre-determined evaluation logic encoding parameters to obtain model encoding parameters, and integrate all model encoding parameters into the pre-determined target training model; Perform data annotation operations on the target knowledge graph to obtain target labeled data, which includes asset value tag data; Using a pre-defined federated learning method and target labeled data, a training operation is performed on a pre-determined target training model to obtain the parameter weights of each model parameter contained in the target training model. Based on each model parameter and its corresponding parameter weights, a target intelligent model is generated.

[0061] In this optional embodiment, the parameter processing operation performed on the pre-determined evaluation logic encoding parameters to obtain model encoding parameters may include: The predetermined evaluation logic encoding parameters are decomposed into differentiable functions and encoded into independent modules through a plug-in architecture. The encoded modules are then standardized to unify the input and output formats, forming the model encoding parameters.

[0062] In this optional embodiment, the above-mentioned integration of all model encoding parameters into a pre-determined target training model may include: embedding the model encoding parameters into the pre-trained model; further optionally, the model encoding parameters may be embedded between the feature fusion layer and the output layer of the pre-trained model.

[0063] In this optional embodiment, the above-mentioned training operation on a pre-determined target training model using a preset federated learning method and target labeled data to obtain the parameter weights of each model parameter included in the target training model may include: The federated learning framework is deployed using a pre-defined federated learning method and connected to local data nodes of various enterprises. The target labeled data is input into the pre-determined target training model to perform training operations on the target training model, calculate the gradient of the model parameters, and continue until the loss function converges to generate the weight coefficients corresponding to each model parameter.

[0064] In this optional embodiment, the asset value tag data may optionally include one or more of the following: basic value tag data, application value tag data, market value tag data, and associated attribute tag data; wherein, the basic value tag data may include data scale value: a value score (0-100) based on parameters such as data volume and storage capacity. The data asset's basic scale effect and data quality value are reflected in the quality score generated based on indicators such as data completeness, accuracy, and consistency, which is used to measure the basic usability of the data asset. Application value label data can include scenario coverage value: marking the data asset's coverage rate in business scenarios (such as the number of covered business lines and usage frequency) and its contribution score to the business; revenue contribution label: estimating the expected contribution value of the data asset in the revenue prediction model based on historical data (such as expected revenue amount and rate of return); market value label data can include comparable transaction value: the market fair value range marked with reference to the historical transaction prices of similar data assets and market supply and demand; scarcity value label: a scarcity score generated based on the data asset's industry uniqueness and substitutability, reflecting its market competitive advantage; and related attribute label data can include time dimension labels: timestamps marking value labels (such as evaluation period and data update time) to reflect the timeliness of the value; and scenario association labels: specific attributes of the data asset's application scenario (such as industry and business scenario type) to enhance the contextual understanding of value assessment.

[0065] As can be seen, implementing this optional embodiment enables the execution of parameter processing operations on pre-determined evaluation logic encoding parameters to obtain model encoding parameters, which are then integrated into the target training model. Data annotation operations are performed on the target knowledge graph to obtain target labeled data. The target training model is then trained using a pre-defined federated learning method and the target labeled data to obtain the parameter weights for each parameter, thereby generating the target intelligent model. By performing parameter processing operations on pre-determined evaluation logic encoding parameters to obtain model encoding parameters and integrating them into the target training model, the precise integration of multiple evaluation logics is achieved, improving the accuracy and comprehensiveness of the evaluation results. Furthermore, performing data annotation operations on the target knowledge graph to obtain target labeled data not only improves the efficiency and accuracy of data annotation but also enhances the understanding of the target intelligent model through the knowledge graph. The model's understanding of the data asset context enhances the depth and breadth of the assessment. Training the model using a pre-defined federated learning method ensures data privacy and compliance during training. By using federated learning and labeled target data, training operations are performed on the target training model, dynamically optimizing the parameter weights of each model parameter. This improves the model's generalization and adaptability. Furthermore, based on each model parameter and its corresponding weight, a target intelligent model is generated. This not only improves training efficiency but also generates a high-quality target intelligent model through dynamic parameter weight optimization. This model is better adapted to different assessment scenarios and data variations, thereby improving the accuracy and reliability of data asset assessment, as well as its intelligence and efficiency.

[0066] In another optional embodiment, based on the target asset valuation results, the asset risk information corresponding to the target asset data is determined, including: Extract target indicator data from the target asset valuation results. The target indicator data includes data asset risk data that matches the target asset valuation results. The data asset risk data includes one or more of the following: asset integrity data, asset timeliness data, and asset scarcity data. By analyzing the data asset risk data through the target intelligent model, the asset risk information corresponding to the target asset data can be obtained; Among them, asset risk information includes one or more of the following: asset compliance risk information, asset depreciation risk information, asset risk impact information, and asset risk probability information.

[0067] In this optional embodiment, the extraction of target indicator data from the target asset valuation results may include: The basic characteristics of the data assets are analyzed from the target asset valuation results, and the target indicator data is extracted through a pre-set rule matching algorithm. Among them, asset integrity data includes the proportion of missing data, data consistency verification results, and the proportion of redundant data; asset timeliness data includes data update frequency, the time interval between the latest data and the assessment, and the stability of historical update cycles; asset scarcity data includes the supply of similar data in the industry, substitutability score, and data barrier indicators.

[0068] In this optional embodiment, the above-mentioned analysis of data asset risk data through the target intelligent model to obtain asset risk information corresponding to the target asset data may include: Data asset risk data is input into a target intelligent model to perform multimodal analysis. The model then calls upon the risk assessment plugin built into the target intelligent model to perform calculations on the multimodal analysis results using a weighted aggregation algorithm, yielding the asset risk information corresponding to the target asset data. The multimodal analysis results can include basic data asset feature vectors: completeness indicators (proportion of missing values, consistency verification results), timeliness indicators (update frequency, time interval), scarcity indicators (market exclusivity rate, substitutability score), and external entity association vectors: matching degree with regulatory entities, association score with market trend maps (such as price fluctuations of similar data), and similarity with historical violations. The risk assessment plugin can include cost-based risk modules, income-based risk modules, and market-based risk modules. The weighted aggregation algorithm can be calculated using a weighted aggregation calculation model trained to convergence. Asset risk information can include compliance risk information, depreciation risk information, impact and probability information, and comprehensive risk level information.

[0069] In this optional embodiment, the asset compliance risk information may include regulatory compliance information and violation rectification cost information; the asset depreciation risk information may include timeliness decay rate information and technology substitution risk information; the asset risk impact information may include value impact magnitude information, impact duration information, and risk transmission scope information; and the asset risk probability information may include one or more of the following: compliance violation probability information, market volatility probability information, and risk chain transmission probability information.

[0070] As can be seen, implementing this optional embodiment can extract target indicator data containing data asset risk data from the target asset assessment results. By analyzing the data asset risk data through a target intelligent model, the asset risk information corresponding to the target asset data can be obtained. This can accurately identify various risk types related to data assets, provide a more comprehensive understanding of the potential problems and risk probabilities of data assets, and improve the accuracy of risk assessment through intelligent risk analysis. It can also provide deeper risk insights and combine risk assessment with value assessment, providing a comprehensive assessment framework. By assessing multi-dimensional indicators of data assets, detailed asset risk information is generated, helping enterprises to fully understand the potential risks of data assets. By combining real-time data and historical knowledge graphs, the target intelligent model can dynamically analyze the risk data of data assets, promptly detect risk trends, and achieve dynamic risk monitoring. Through intelligent risk analysis and refined risk identification, manual intervention is reduced, improving the efficiency and accuracy of risk management. This, in turn, helps improve the accuracy and reliability of data asset assessment, as well as its intelligence and efficiency.

[0071] In yet another optional embodiment, before acquiring real-time asset data, the method further includes: Invoke the asset interface call request, and generate the corresponding data call parameters based on the asset interface call request; This includes acquiring real-time asset data, including: Based on the data call parameters, obtain the real-time asset data corresponding to the asset interface call request.

[0072] In this optional embodiment, the above-mentioned asset interface call request, based on the asset interface call request, generates data call parameters corresponding to the asset interface call request, which may include: By deploying an API gateway on the regulatory authority's inference server, asset interface call requests are sent to the enterprise data platform or external data sources, and corresponding structured data call parameters are generated according to the interface type. The interface type can include one of the following: real-time business data interface, market information API interface, or third-party data service interface; the data call parameters include: enterprise unique identifier (such as unified social credit code), data asset type label (basic data / transaction data / industry data), time range parameter (real-time update frequency ≤ 15 minutes / historical window period ≥ 3 months), data format parameter (JSON / CSV / Parquet), and access control parameter (data anonymization level / access frequency limit).

[0073] In this optional embodiment, the above-mentioned method of obtaining real-time asset data corresponding to the asset interface call request based on data call parameters may include: Based on data call parameters, real-time asset data is obtained from multiple source interfaces through a load balancing mechanism. The real-time asset data includes real-time data streams from enterprise business systems (such as data usage frequency and real-time logs of new data volume), dynamic price data from market APIs (the price fluctuation frequency of similar data assets is ≤5 minutes), and policy dynamic data from third-party data services (the delay in pushing regulatory updates is ≤1 hour).

[0074] As can be seen, implementing this optional embodiment can invoke asset interface call requests and generate corresponding data call parameters. Based on these data call parameters, real-time asset data corresponding to the asset interface call request can be obtained. This method of invoking asset interface call requests and generating data call parameters facilitates efficient acquisition of real-time asset data and reduces human interference, thus improving the efficiency, accuracy, and reliability of data acquisition. Furthermore, by generating data call parameters, the scope and content of data acquisition can be flexibly adjusted according to different asset interface call requests, allowing the system to better adapt to different application scenarios and user needs. By acquiring real-time asset data based on data call parameters, the system can obtain the latest data asset information in real time and promptly reflect the dynamic changes of data assets, which is beneficial... To improve the timeliness of assessment results, the system can ensure that the acquired data is up-to-date and accurate by calling asset interface requests and generating data call parameters. Parameterized calls ensure the integrity and consistency of the acquired data. By calling asset interface requests and generating data call parameters, the system can flexibly access new data sources and interfaces, enabling it to adapt to constantly changing market environments and user needs, maintaining high performance and adaptability. By acquiring the latest asset data in real time, the system can dynamically adjust assessment results, supporting real-time assessment and dynamic adjustment. This better adapts to the dynamic changes of data assets, providing more accurate assessment results, and further improving the accuracy and reliability of data asset assessment, as well as enhancing the intelligence and efficiency of data asset assessment.

[0075] In another optional embodiment, based on all feature asset data, a target knowledge graph corresponding to the target asset data is constructed, including: Perform data preprocessing operations on all feature asset data to obtain target feature data. The data preprocessing operations include one or more of the following: data cleaning, data denoising, and data standardization. Based on the target feature data, determine the graph nodes and the node connection relationships between each graph node. The graph nodes include one or more of the following: data type, data source, and data size of the target feature data. The connection relationships include one or more of the following: association strength and association type between each graph node. Based on each graph node and the node connections between each graph node, a target knowledge graph corresponding to the target asset data is constructed.

[0076] In this optional embodiment, the process of determining the map nodes and the node connection relationships between each map node based on the target feature data may include: Extract data type, data source, data size, and business scenario from the target feature data as graph nodes, and determine the correlation strength between data source and business scenario through text similarity algorithm. Define connection relationship types based on business logic between entities to generate node connection relationships. The data types can include one of the following: structured type, semi-structured type, and unstructured type; the data sources can include one or more of the following: enterprise database, industry report, and trading platform; the business scenarios can include one or more of the following: marketing scenario, risk control scenario, and operation scenario; and the connection relationship types can include one or more of the following: belonging relationship type, influence relationship type, and association relationship type.

[0077] In this optional embodiment, a target knowledge graph containing more than three layers of entity relationships can be generated through a graph visualization component. The graph update frequency is synchronized with the feature asset data change frequency (≤1 hour), and the consistency of the graph is ensured by entity relationship verification rules (such as the requirement that the association between the data source and the legal entity must have a compliance clause mapping).

[0078] In this optional embodiment, the graph nodes may optionally include data asset entities, enterprise entities, industry entities, etc.; the node connection relationships may include the relationship between data assets and enterprise entities, the relationship between data assets and industry entities, the relationship between data assets and policy and regulatory entities, etc.; the target knowledge graph is used to represent the complex relationship between data assets and external entities, and serves as contextual information for data asset value assessment.

[0079] As can be seen, implementing this optional embodiment can perform data preprocessing operations on all feature asset data to obtain target feature data, determine graph nodes and the node connections between each graph node, and then construct the target knowledge graph corresponding to the target asset data. By performing preprocessing operations such as data cleaning, denoising, and standardization, the quality of the feature asset data is ensured. By preprocessing the feature asset data to obtain target feature data, and then determining graph nodes and node connections based on this high-quality data, not only is the accuracy of the graph nodes improved, but the reliability of the node connections is also enhanced, enabling the target knowledge graph to more accurately reflect the actual situation of the data assets. Furthermore, by constructing a target knowledge graph containing multi-dimensional features and complex connections, the value of data assets can be analyzed from multiple perspectives, enhancing the comprehensiveness and depth of the assessment. This allows for a more accurate evaluation of the scarcity and completeness of data assets. The rich definition of graph nodes and connections also enables the system to have good scalability, allowing for the flexible addition of new features and relationships to adapt to different assessment scenarios. It can automatically identify and process data features, generating accurate knowledge graphs, providing a solid foundation for subsequent intelligent assessments. This, in turn, helps improve the accuracy and reliability of data asset assessments, as well as their intelligence and efficiency.

[0080] In another optional embodiment, an adjustment operation is performed on the data value assessment result based on real-time asset data to obtain the target asset assessment result, including: The target intelligent model performs data processing operations on real-time asset data to obtain the evaluation output results corresponding to the target intelligent model. The data processing operations include one or more of global feature extraction operations and complex data recognition operations. Based on the evaluation output, perform local characteristic evaluation operations on the real-time asset data to generate at least one local evaluation output, and generate the model evaluation output based on all local evaluation outputs; Based on the predetermined fusion strategy, the model evaluation output and the data value evaluation results are fused to obtain the evaluation fusion result. Then, the data value evaluation result is adjusted according to the evaluation fusion result to obtain the target asset evaluation result.

[0081] In this optional embodiment, the global feature extraction operation can optionally utilize the Transformer architecture of the model to perform global feature extraction, and capture the correlation features between data assets and market dynamics and regulatory changes through a multi-head attention mechanism; the complex data identification operation can use a convolutional neural network module to identify complex patterns in unstructured real-time data (such as policy texts and market sentiment) and generate a time-series feature vector.

[0082] In this optional embodiment, the above-mentioned process of performing a local characteristic evaluation operation on the real-time asset data based on the evaluation output results to generate at least one local evaluation output result, and generating a model evaluation output result based on all local evaluation output results, may include: Based on the evaluation output results, determine the local feature parameters corresponding to the evaluation output results, and perform local characteristic evaluation operations on the real-time asset data according to the local feature parameters to generate at least one local evaluation output result. The local evaluation output result may include one or more of the following: data scarcity evaluation result, data timeliness evaluation result, and data application scenario evaluation result. Based on all local evaluation outputs, generate model evaluation outputs, which include at least all local evaluation outputs.

[0083] In this optional embodiment, the above-mentioned fusion processing operation on the model evaluation output and the data value evaluation result based on a predetermined fusion strategy to obtain the evaluation fusion result may include: Based on a predetermined fusion strategy, a fusion processing operation is performed on the model evaluation output and the data value evaluation results. This involves performing a fusion operation on the first feature result corresponding to the model evaluation output and the second feature result corresponding to the data value evaluation results to obtain the evaluation fusion result. The first feature result may include data scale characteristics and data quality characteristics; the second feature result may include scarcity evaluation results and timeliness evaluation results. Furthermore, a fusion operation is performed on the first feature result corresponding to the model evaluation output and the second feature result corresponding to the data value evaluation result to obtain the evaluation fusion result, which may include: Align the first feature result and the second feature result to ensure that they are fused on the same feature dimension. According to the predetermined fusion strategy, assign weights to each feature and perform weighted averaging or other fusion algorithms (such as attention mechanism) to calculate the fusion result to obtain the evaluation fusion result. Furthermore, the pre-determined fusion strategy may include weighted averaging, attention mechanisms, etc., to combine model evaluation output results and data value evaluation results.

[0084] In this optional embodiment, the above-mentioned adjustment operation on the data value assessment result based on the assessment fusion result to obtain the target asset assessment result may include: The evaluation fusion results are analyzed to determine the evaluation dimensions and directions that need to be adjusted. Based on the evaluation fusion results, the adjustment parameters for each evaluation dimension are determined. The adjustment parameters include one or more of the following: adjustment weight and adjustment magnitude. Based on the determined adjustment parameters, the corresponding assessment dimensions in the data value assessment results are adjusted to obtain the adjusted data value assessment results, and the adjusted data value assessment results are integrated into the target asset assessment results.

[0085] As can be seen, implementing this optional embodiment can perform data processing operations on real-time asset data through a target intelligent model to obtain corresponding evaluation output results. Based on the evaluation output results, local feature evaluation operations are performed on the real-time asset data to generate at least one local evaluation output result, which in turn generates a model evaluation output result. Based on a pre-determined fusion strategy, a fusion processing operation is performed on the model evaluation output result and the data value evaluation result to obtain an evaluation fusion result. Finally, an adjustment operation is performed on the data value evaluation result to obtain the target asset evaluation result. Through the processing and evaluation of real-time asset data, it can dynamically adapt to real-time changes in data assets, adjust evaluation results in a timely manner, and ensure the timeliness and accuracy of the evaluation results. Furthermore, through global feature extraction and complex data recognition operations, the system can evaluate the value of data assets from multiple dimensions. This multi-dimensional evaluation improves the comprehensiveness and accuracy of the evaluation. By employing a target-based intelligent model and fusion strategies, the system enhances the intelligence and objectivity of the assessment. Through local characteristic assessment operations, it generates multiple local assessment outputs, which, combined with the model assessment outputs, enable a more accurate evaluation of various features of data assets. Furthermore, refined assessment improves the precision and reliability of the results. Using a pre-defined fusion strategy, the system can flexibly combine model assessment outputs and data value assessment results to generate a fused assessment result, adapting to different assessment scenarios and needs, thus improving the adaptability and scalability of the assessment. Through global feature extraction and complex data recognition operations using the target-based intelligent model, the system can handle complex data relationships and multi-dimensional features, supporting assessments in complex scenarios and providing more accurate results. This further contributes to improving the accuracy and reliability of data asset assessment, as well as enhancing its intelligence and efficiency.

[0086] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a data asset intelligent evaluation device based on a large model, as disclosed in an embodiment of the present invention. Figure 3 As shown, the intelligent data asset assessment device based on a large model may include: The acquisition module 301 is used to acquire target asset data, which includes asset characteristic data and historical asset data. Processing module 302 is used to perform preprocessing operations on target asset data to obtain feature asset data; Module 303 is used to construct the target knowledge graph corresponding to the target asset data based on all feature asset data; The training module 304 is used to integrate the pre-determined evaluation logic encoding parameters into the pre-determined target training model, and to perform training operations on the pre-determined target training model through a preset federated learning method and target knowledge graph to obtain the target intelligent model. The generation module 305 is used to generate data value assessment results corresponding to the target asset data based on the target knowledge graph and the target intelligent model. The acquisition module 301 is also used to acquire real-time asset data; The adjustment module 306 is used to perform adjustment operations on the data value assessment results based on real-time asset data to obtain the target asset assessment results.

[0087] It is evident that implementation Figure 3 The described apparatus can acquire target asset data and perform preprocessing operations to obtain feature asset data, thereby constructing a target knowledge graph. It integrates evaluation logic encoding parameters into a target training model and trains the model using federated learning and the target knowledge graph to obtain a target intelligent model. Based on the target knowledge graph and the target intelligent model, it generates corresponding data value evaluation results. It acquires real-time asset data and performs adjustment operations on the data value evaluation results to obtain the target asset evaluation results. Through the strong generalization and multimodal understanding capabilities of the large model, it achieves automated evaluation of data asset value, improving the objectivity and consistency of the evaluation. Through federated learning technology, it supports rapid processing and real-time response of massive amounts of data. It can also combine real-time data streams and historical data to dynamically output evaluation results, reflecting the dynamic value of data assets in real time. Furthermore, federated learning technology ensures data privacy and compliance during the evaluation process. By combining real-time data and historical knowledge graphs, it generates multi-dimensional value scores through a large model and identifies potential risks of data assets based on the evaluation results. It can dynamically adapt to changes in the value of data assets, thereby improving the accuracy and reliability of data asset evaluation, as well as its intelligence and efficiency.

[0088] In an optional embodiment, such as Figure 4 As shown, the device also includes: The determination module 307 is used to determine the asset risk information corresponding to the target asset data based on the target asset valuation results. The generation module 305 is also used to generate visualized assessment information based on the target asset assessment results and asset risk information. The visualized assessment information includes one or more of the following: asset assessment report information corresponding to the target asset assessment results, asset assessment value trend information, and asset operation suggestion information. Feedback module 308 is used to feed back the visual evaluation information to the target user terminal; The acquisition module 301 is also used to acquire the evaluation feedback information corresponding to the target user terminal; The update module 309 is used to perform model update operations on the target intelligent model based on the evaluation feedback information.

[0089] It is evident that implementation Figure 4 The described device can determine the asset risk information corresponding to the target asset data based on the target asset assessment results, identify potential risks of data assets in advance, and generate visualized assessment information based on the target asset assessment results and asset risk information. Through visualization, users can intuitively understand the assessment results, enhancing their trust in the assessment process and results. The visualized assessment information is fed back to the target user terminal, and the device obtains corresponding assessment feedback information from the target user terminal. Through user feedback, the assessment model can be adjusted and optimized in a timely manner, improving the model's accuracy and adaptability, enhancing the user experience. Based on the assessment feedback information, the device performs model update operations on the target intelligent model, achieving continuous model optimization. Through continuous data updates and the combination of user feedback, the model can maintain high performance and high accuracy. Furthermore, the feedback mechanism can influence the optimization of the assessment model, thereby increasing trust in the system. This, in turn, helps improve the accuracy and reliability of data asset assessment, as well as the intelligence and efficiency of data asset assessment.

[0090] In another alternative embodiment, such as Figure 4 As shown, the training module 304 integrates the pre-determined evaluation logic encoding parameters into the pre-determined target training model, and performs training operations on the pre-determined target training model through a preset federated learning method and a target knowledge graph. The specific methods for obtaining the target intelligent model include: Perform parameter processing operations on the pre-determined evaluation logic encoding parameters to obtain model encoding parameters, and integrate all model encoding parameters into the pre-determined target training model; Perform data annotation operations on the target knowledge graph to obtain target labeled data, which includes asset value tag data; Using a pre-defined federated learning method and target labeled data, a training operation is performed on a pre-determined target training model to obtain the parameter weights of each model parameter contained in the target training model. Based on each model parameter and its corresponding parameter weights, a target intelligent model is generated.

[0091] It is evident that implementation Figure 4The described apparatus can perform parameter processing operations on pre-determined evaluation logic encoding parameters to obtain model encoding parameters and integrate them into the target training model; perform data annotation operations on the target knowledge graph to obtain target labeled data; and train the target training model using a pre-defined federated learning method and the target labeled data to obtain the parameter weights of each parameter, thereby generating a target intelligent model. By performing parameter processing operations on pre-determined evaluation logic encoding parameters to obtain model encoding parameters and integrating them into the target training model, it achieves precise integration of multiple evaluation logics, improving the accuracy and comprehensiveness of evaluation results. Furthermore, performing data annotation operations on the target knowledge graph to obtain target labeled data not only improves the efficiency and accuracy of data annotation but also enhances the model's understanding of the target knowledge graph. Understanding the context of data assets enhances the depth and breadth of assessment. Model training using pre-defined federated learning methods ensures data privacy and compliance during training. By employing federated learning and labeled target data, training operations are performed on the target training model, dynamically optimizing the parameter weights of each model parameter. This improves the model's generalization and adaptability. Furthermore, based on each model parameter and its corresponding weight, a target intelligent model can be generated. This not only improves training efficiency but also generates a high-quality target intelligent model through dynamic parameter weight optimization. This model is better adapted to different assessment scenarios and data variations, thereby improving the accuracy and reliability of data asset assessment, as well as its intelligence and efficiency.

[0092] In yet another alternative embodiment, such as Figure 4 As shown, the specific methods by which module 307 determines the asset risk information corresponding to the target asset data based on the target asset valuation results include: Extract target indicator data from the target asset valuation results. The target indicator data includes data asset risk data that matches the target asset valuation results. The data asset risk data includes one or more of the following: asset integrity data, asset timeliness data, and asset scarcity data. By analyzing the data asset risk data through the target intelligent model, the asset risk information corresponding to the target asset data can be obtained; Among them, asset risk information includes one or more of the following: asset compliance risk information, asset depreciation risk information, asset risk impact information, and asset risk probability information.

[0093] It is evident that implementation Figure 4The described device can extract target indicator data containing data asset risk data from the target asset assessment results. Through target intelligent model analysis of the data asset risk data, it obtains the corresponding asset risk information, accurately identifying various risk types related to data assets. This allows for a more comprehensive understanding of the potential problems and risk probabilities of data assets. Furthermore, by using the target intelligent model to analyze the data asset risk data, it not only improves the accuracy of risk assessment but also provides deeper risk insights. It combines risk assessment with value assessment, providing a comprehensive evaluation framework. By evaluating multi-dimensional indicators of data assets, it generates detailed asset risk information, helping enterprises fully understand the potential risks of data assets. Combined with real-time data and historical knowledge graphs, the target intelligent model can dynamically analyze data asset risk data, promptly detect risk trends, and achieve dynamic risk monitoring. Through intelligent risk analysis and refined risk identification, it reduces manual intervention, improves the efficiency and accuracy of risk management, and further enhances the accuracy and reliability of data asset assessment, as well as its intelligence and efficiency.

[0094] In yet another alternative embodiment, such as Figure 4 As shown, the device also includes: Module 310 is used to call the asset interface call request before the acquisition module 301 acquires the real-time asset data; The generation module 305 is also used to generate data call parameters corresponding to the asset interface call request based on the asset interface call request; The specific methods by which the acquisition module 301 acquires real-time asset data include: Based on the data call parameters, obtain the real-time asset data corresponding to the asset interface call request.

[0095] It is evident that implementation Figure 4The described device can invoke asset interface call requests and generate corresponding data call parameters. Based on these parameters, it acquires real-time asset data corresponding to the asset interface call requests. This ability to efficiently acquire real-time asset data, while reducing human interference, improves the efficiency, accuracy, and reliability of data acquisition. Furthermore, by generating data call parameters, the scope and content of data acquisition can be flexibly adjusted according to different asset interface call requests, allowing the system to better adapt to different application scenarios and user needs. By acquiring real-time asset data based on these parameters, the system can obtain the latest asset information in real time, promptly reflecting dynamic changes in asset data, thus improving efficiency. The timeliness of the evaluation results can be ensured by calling asset interfaces and generating data call parameters. The system can ensure that the acquired data is up-to-date and accurate, and the parameterized calls can ensure the integrity and consistency of the acquired data. By calling asset interfaces and generating data call parameters, the system can flexibly access new data sources and interfaces, and adapt to the ever-changing market environment and user needs, maintaining high performance and high adaptability. By acquiring the latest asset data in real time, the system can dynamically adjust the evaluation results, supporting real-time evaluation and dynamic adjustment. It can better adapt to the dynamic changes of data assets, provide more accurate evaluation results, and thus improve the accuracy and reliability of data asset evaluation, as well as the intelligence and efficiency of data asset evaluation.

[0096] In yet another alternative embodiment, such as Figure 4 As shown, the specific methods by which construction module 303 constructs the target knowledge graph corresponding to the target asset data based on all feature asset data include: Perform data preprocessing operations on all feature asset data to obtain target feature data. The data preprocessing operations include one or more of the following: data cleaning, data denoising, and data standardization. Based on the target feature data, determine the graph nodes and the node connection relationships between each graph node. The graph nodes include one or more of the following: data type, data source, and data size of the target feature data. The connection relationships include one or more of the following: association strength and association type between each graph node. Based on each graph node and the node connections between each graph node, a target knowledge graph corresponding to the target asset data is constructed.

[0097] It is evident that implementation Figure 4The described apparatus can perform data preprocessing operations on all feature asset data to obtain target feature data, determine graph nodes and the node connections between each graph node, and then construct a target knowledge graph corresponding to the target asset data. By performing preprocessing operations such as data cleaning, denoising, and standardization, the quality of the feature asset data is ensured. By preprocessing the feature asset data to obtain target feature data, and then determining graph nodes and node connections based on this high-quality data, not only is the accuracy of the graph nodes improved, but the reliability of the node connections is also enhanced. This allows the target knowledge graph to more accurately reflect the actual situation of the data assets. By constructing a target knowledge graph containing multi-dimensional features and complex connections, the value of data assets can be analyzed from multiple perspectives, enhancing the comprehensiveness and depth of the assessment. This allows for a more accurate evaluation of the scarcity and completeness of data assets. Furthermore, the rich definition of graph nodes and connections enables the system to have good scalability, allowing for the flexible addition of new features and relationships to adapt to different assessment scenarios. It can automatically identify and process data features to generate accurate knowledge graphs, providing a solid foundation for subsequent intelligent assessments. This, in turn, helps improve the accuracy and reliability of data asset assessments, as well as their intelligence and efficiency.

[0098] In yet another alternative embodiment, such as Figure 4 As shown, the adjustment module 306 performs adjustment operations on the data value assessment results based on real-time asset data, and the specific methods for obtaining the target asset assessment results include: The target intelligent model performs data processing operations on real-time asset data to obtain the evaluation output results corresponding to the target intelligent model. The data processing operations include one or more of global feature extraction operations and complex data recognition operations. Based on the evaluation output, perform local characteristic evaluation operations on the real-time asset data to generate at least one local evaluation output, and generate the model evaluation output based on all local evaluation outputs; Based on the predetermined fusion strategy, the model evaluation output and the data value evaluation results are fused to obtain the evaluation fusion result. Then, the data value evaluation result is adjusted according to the evaluation fusion result to obtain the target asset evaluation result.

[0099] It is evident that implementation Figure 4The described device can perform data processing operations on real-time asset data through a target intelligent model to obtain corresponding evaluation output results. Based on the evaluation output results, it performs local feature evaluation operations on the real-time asset data to generate at least one local evaluation output result, which in turn generates a model evaluation output result. Based on a pre-determined fusion strategy, it performs fusion processing operations on the model evaluation output result and the data value evaluation result to obtain an evaluation fusion result. Finally, it performs adjustment operations on the data value evaluation result to obtain the target asset evaluation result. Through the processing and evaluation of real-time asset data, it can dynamically adapt to real-time changes in data assets, promptly adjust the evaluation results, and ensure the timeliness and accuracy of the evaluation results. Furthermore, through global feature extraction and complex data recognition operations, the system can evaluate the value of data assets from multiple dimensions. This multi-dimensional evaluation improves the comprehensiveness and accuracy of the evaluation. The target intelligent model and fusion strategy enhance the intelligence and objectivity of the assessment. By generating multiple local assessment outputs through local feature assessment operations and combining them with the model assessment outputs, the various characteristics of data assets can be assessed more accurately. Furthermore, the refined assessment improves the accuracy and reliability of the assessment results. Through a pre-determined fusion strategy, the system can flexibly combine the model assessment outputs and data value assessment results to generate a fused assessment result, adapting to different assessment scenarios and needs, and improving the adaptability and scalability of the assessment. Through global feature extraction and complex data recognition operations of the target intelligent model, the system can handle complex data relationships and multi-dimensional features, supporting assessments in complex scenarios and providing more accurate assessment results. This further contributes to improving the accuracy and reliability of data asset assessment, as well as enhancing the intelligence and efficiency of data asset assessment.

[0100] Example 4 Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of another intelligent data asset evaluation device based on a large model disclosed in an embodiment of the present invention. (See diagram below.) Figure 5 As shown, the intelligent data asset assessment device based on a large model may include: Memory 401 storing executable program code; Processor 402 coupled to memory 401; The processor 402 calls the executable program code stored in the memory 401 to execute some or all of the steps in any of the intelligent data asset assessments based on a large model in Embodiment 1 of the present invention.

[0101] Example 5 This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the data asset intelligent assessment methods based on large models disclosed in Embodiment 1 of this invention.

[0102] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0103] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0104] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data asset intelligent evaluation method based on a large model, characterized in that, The method includes: Acquire target asset data, wherein the target asset data includes asset characteristic data and historical asset data; Preprocessing operations are performed on the target asset data to obtain feature asset data, and a target knowledge graph corresponding to the target asset data is constructed based on all the feature asset data. The predetermined evaluation logic encoding parameters are integrated into the predetermined target training model, and the predetermined target training model is trained using a preset federated learning method and the target knowledge graph to obtain the target intelligent model. Based on the target knowledge graph and the target intelligent model, a data value assessment result corresponding to the target asset data is generated; Acquire real-time asset data, and perform adjustment operations on the data value assessment results based on the real-time asset data to obtain the target asset assessment results.

2. The intelligent data asset evaluation method based on a large model according to claim 1, characterized in that, The method further includes: Based on the target asset valuation results, determine the asset risk information corresponding to the target asset data; Based on the target asset valuation results and the asset risk information, visualized valuation information is generated, wherein the visualized valuation information includes one or more of the following: asset valuation report information corresponding to the target asset valuation results, asset valuation value trend information, and asset operation suggestion information; The visualized evaluation information is fed back to the target user terminal, and the evaluation feedback information corresponding to the target user terminal is obtained. The target intelligent model is then updated based on the evaluation feedback information.

3. The intelligent data asset evaluation method based on a large model according to claim 1, characterized in that, The process of integrating pre-determined evaluation logic encoding parameters into a pre-determined target training model, and then training the pre-determined target training model using a preset federated learning method and the target knowledge graph to obtain a target intelligent model includes: The predetermined evaluation logic encoding parameters are processed to obtain model encoding parameters, and all the model encoding parameters are integrated into the predetermined target training model. Perform data annotation operations on the target knowledge graph to obtain target annotation data, wherein the target annotation data includes asset value tag data; Using a pre-defined federated learning method and the target labeled data, a training operation is performed on the pre-determined target training model to obtain the parameter weights of each model parameter contained in the target training model. Based on each model parameter and the parameter weights corresponding to each model parameter, a target intelligent model is generated.

4. The intelligent data asset evaluation method based on a large model according to claim 2, characterized in that, The step of determining the asset risk information corresponding to the target asset data based on the target asset valuation results includes: Target indicator data is extracted from the target asset valuation results. The target indicator data includes data asset risk data that matches the target asset valuation results. The data asset risk data includes one or more of the following: asset integrity data, asset timeliness data, and asset scarcity data. By analyzing the data asset risk data using the target intelligent model, asset risk information corresponding to the target asset data is obtained; The asset risk information includes one or more of the following: asset compliance risk information, asset depreciation risk information, asset risk impact information, and asset risk probability information.

5. The intelligent data asset evaluation method based on a large model according to claim 1, characterized in that, Before acquiring real-time asset data, the method further includes: Invoke the asset interface call request, and generate the data call parameters corresponding to the asset interface call request based on the asset interface call request; The acquisition of real-time asset data includes: Based on the data call parameters, obtain the real-time asset data corresponding to the asset interface call request.

6. The intelligent data asset evaluation method based on a large model according to claim 1, characterized in that, The step of constructing a target knowledge graph corresponding to the target asset data based on all the aforementioned feature asset data includes: Perform data preprocessing operations on all the aforementioned feature asset data to obtain target feature data, wherein the data preprocessing operations include one or more of data cleaning operations, data denoising operations, and data standardization operations; Based on the target feature data, determine the graph nodes and the node connection relationships between each graph node, wherein the graph node includes one or more of the data type, data source, and data size of the target feature data, and the connection relationship includes one or more of the association strength and association type between each graph node; Based on each of the graph nodes and the node connection relationships between each of the graph nodes, a target knowledge graph corresponding to the target asset data is constructed.

7. The intelligent data asset evaluation method based on a large model according to claim 1, characterized in that, The step of adjusting the data value assessment result based on the real-time asset data to obtain the target asset assessment result includes: The target intelligent model performs data processing operations on the real-time asset data to obtain the evaluation output result corresponding to the target intelligent model. The data processing operations include one or more of global feature extraction operations and complex data recognition operations. Based on the evaluation output, a local characteristic evaluation operation is performed on the real-time asset data to generate at least one local evaluation output, and a model evaluation output is generated based on all the local evaluation outputs. Based on a predetermined fusion strategy, a fusion processing operation is performed on the model evaluation output and the data value evaluation result to obtain an evaluation fusion result. Then, an adjustment operation is performed on the data value evaluation result based on the evaluation fusion result to obtain the target asset evaluation result.

8. A data asset intelligent evaluation device based on a large model, characterized in that, The device includes: The acquisition module is used to acquire target asset data, wherein the target asset data includes asset characteristic data and historical asset data; The processing module is used to perform preprocessing operations on the target asset data to obtain feature asset data; A construction module is used to construct a target knowledge graph corresponding to the target asset data based on all the aforementioned feature asset data; The training module is used to integrate the pre-determined evaluation logic encoding parameters into the pre-determined target training model, and to perform training operations on the pre-determined target training model through a preset federated learning method and the target knowledge graph to obtain the target intelligent model; The generation module is used to generate a data value assessment result corresponding to the target asset data based on the target knowledge graph and the target intelligent model. The acquisition module is also used to acquire real-time asset data; The adjustment module is used to perform adjustment operations on the data value assessment results based on the real-time asset data to obtain the target asset assessment results.

9. A data asset intelligent evaluation device based on a large model, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the data asset intelligent assessment method based on a large model as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the data asset intelligent evaluation method based on a large model as described in any one of claims 1-7.

Images