Data asset analysis method, equipment and medium

Abstract

The invention discloses a data asset analysis method and device and a medium, and relates to the technical field of data assets. The method comprises the following steps: inputting to-be-analyzed asset data into a preset industry classification model to obtain an industry classification label of the to-be-analyzed asset data; wherein the industry classification model comprises classification algorithms corresponding to different data modes; based on the industry classification label, matching with a label in a preset industry knowledge base to obtain an industry knowledge fragment with the highest association degree with the to-be-analyzed asset data; wherein the industry knowledge base comprises industry knowledge for data resource quality analysis and value analysis; and inputting the industry knowledge fragment into a preset large language model to obtain an analysis result of the to-be-analyzed asset data. According to the method, the to-be-analyzed data assets are labeled with the industry labels through the industry classification model, the industry knowledge base is matched based on the classification result, and the industry knowledge with the highest association degree is injected into the large language model, so that the analysis result better meeting the specific requirements of the industry is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data assets, and in particular to a data asset analysis method, device and medium. BACKGROUND

[0002] In the field of data assetization, existing technologies usually rely on traditional data analysis methods, such as data exploration and value assessment based on rule engines or general machine learning models (e.g., decision trees, clustering algorithms). These methods often lack deep integration of industry-specific knowledge and semantic understanding capabilities when dealing with large-scale, multi-source heterogeneous data. At the same time, the models in existing technologies are mostly based on static rules or general training data, making it difficult to adapt to different industry business backgrounds and semantic characteristics, resulting in a disconnect between data analysis results and real application scenarios. In addition, traditional methods have obvious shortcomings in cross-industry data generalization, dynamic knowledge retrieval, and context enhancement, and cannot effectively utilize industry knowledge bases to improve the accuracy and professionalism of analysis. There are widespread problems such as low data analysis precision, poor industry adaptability, low efficiency, and difficulty in breaking through industry knowledge silos.

[0003] Therefore, how to build an industry classification model and combine the semantic understanding and logical reasoning capabilities of large language models to dynamically match industry knowledge bases and achieve intelligent and specialized analysis and value mining of data assets has become a technical problem that needs to be solved. SUMMARY

[0004] The embodiments of the present application provide a data asset analysis method, device and medium, to solve the technical problem of how to build an industry classification model and combine the semantic understanding and logical reasoning capabilities of large language models to dynamically match industry knowledge bases and achieve intelligent and specialized analysis and value mining of data assets.

[0005] In a first aspect, the embodiments of the present application provide a data asset analysis method, which includes: inputting asset data to be analyzed into a preset industry classification model to obtain an industry classification label of the asset data to be analyzed; based on the industry classification label, matching with labels in a preset industry knowledge base to obtain an industry knowledge segment with the highest correlation degree associated with the asset data to be analyzed; and inputting the industry knowledge segment into a preset large language model to obtain an analysis result of the asset data to be analyzed.

[0006] In one implementation of this application, the method further includes: constructing an industry classification model, specifically including: obtaining a training sample set; wherein the training sample set includes structured data samples, text data samples, and image data samples labeled with industry classification tags; training the structured data samples using a decision tree algorithm to generate a first classification sub-model; training the text data samples using a BERT model to generate a second classification sub-model; training the image data samples using a CLIP model to generate a third classification sub-model; and integrating the first classification sub-model, the second classification sub-model, and the third classification sub-model to obtain an industry classification model.

[0007] In one implementation of this application, the asset data to be analyzed is input into a preset industry classification model to obtain the industry classification label of the asset data to be analyzed. Specifically, this includes: parsing the data modality of the asset data to be analyzed to obtain the data modality type; selecting a target classification sub-model corresponding to the data modality type from the industry classification model; and using the target classification sub-model to process the asset data to be analyzed to obtain the industry classification label of the asset data to be analyzed.

[0008] In one implementation of this application, the method further includes: constructing an industry knowledge base, specifically including: collecting standard documents and data standards from various industries to generate raw knowledge materials; preprocessing the raw knowledge materials to obtain standard knowledge materials, and extracting key features from the standard knowledge materials to generate knowledge tags corresponding to the standard knowledge materials; and constructing an industry knowledge base based on the key features and knowledge tags.

[0009] In one implementation of this application, the industry knowledge fragments are matched with tags in a preset industry knowledge base based on industry classification tags to obtain the industry knowledge fragments with the highest relevance to the asset data to be analyzed. Specifically, this includes: calculating the semantic similarity between the industry classification tags and each tag in the industry knowledge base; sorting the knowledge fragments in the industry knowledge base based on the semantic similarity to obtain a sorting result; and selecting the knowledge fragment with the highest semantic similarity from the sorting result to obtain the industry knowledge fragment.

[0010] In one implementation of this application, industry knowledge fragments are input into a preset large language model to obtain analysis results of the asset data to be analyzed. Specifically, this includes: associating industry knowledge fragments with the asset data to be analyzed to generate input prompts; inputting the input prompts into the large language model to generate quality assessment content and data value assessment content of the asset data to be analyzed; and integrating and verifying the quality assessment content and data value assessment content to obtain analysis results of the asset data to be analyzed.

[0011] In one implementation of this application, after obtaining the analysis results of the asset data to be analyzed, the method further includes: obtaining application feedback information of the asset data to be analyzed; wherein, the application feedback information includes the accuracy evaluation and applicability evaluation of the analysis results in the data usage scenario; based on the application feedback information, calculating the degree of deviation between the analysis results and the actual application requirements; if the degree of deviation exceeds a preset threshold, then based on the application feedback information, iteratively optimizing the industry knowledge base and industry classification model until an analysis result that meets the actual application requirements is obtained.

[0012] In one implementation of this application, the method further includes: extracting quality assessment content and data value assessment content, and adding corresponding industry classification tags, timestamps, and knowledge base matching source identifiers to generate quality assessment sub-reports and value assessment sub-reports; and integrating the quality assessment sub-reports and value assessment sub-reports into a standardized analysis document for data traceability.

[0013] Secondly, embodiments of this application also provide a data asset analysis device, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: input asset data to be analyzed into a preset industry classification model to obtain industry classification labels for the asset data to be analyzed; match the industry classification labels with labels in a preset industry knowledge base to obtain industry knowledge fragments with the highest relevance to the asset data to be analyzed; and input the industry knowledge fragments into a preset large language model to obtain analysis results for the asset data to be analyzed.

[0014] Thirdly, embodiments of this application also provide a non-volatile computer storage medium for data asset analysis, storing computer-executable instructions, which are configured to: input the asset data to be analyzed into a preset industry classification model to obtain industry classification labels for the asset data to be analyzed; match the industry classification labels with labels in a preset industry knowledge base to obtain industry knowledge fragments with the highest relevance to the asset data to be analyzed; and input the industry knowledge fragments into a preset large language model to obtain analysis results for the asset data to be analyzed.

[0015] The data asset analysis method, device, and medium provided in this application have the following beneficial effects: They accurately determine the industry affiliation of source data through an industry classification model, dynamically match the industry knowledge base based on the classification results, and inject the most relevant industry knowledge into the model, making quality exploration and value analysis more aligned with specific industry needs; combining the semantic understanding of the large model with the professional content of the industry knowledge base enhances the model's ability to interpret the semantics of heterogeneous data, reduces interference from invalid information, and improves the industry adaptability and result authenticity of data asset analysis. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of a data asset analysis method provided in this application embodiment; Figure 2 This is a schematic diagram of the internal structure of a data asset analysis device provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] This application provides a data asset analysis method, device, and medium to solve the following technical problem: how to dynamically match industry knowledge bases by constructing industry classification models and combining the semantic understanding and logical reasoning capabilities of large language models, thereby achieving intelligent, professional analysis and value mining of data assets.

[0019] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating a data asset analysis method provided in an embodiment of this application. Figure 1 As shown in the figure, the data asset analysis method provided in this application embodiment specifically includes the following steps: Step 10: Input the asset data to be analyzed into the preset industry classification model to obtain the industry classification labels of the asset data to be analyzed.

[0021] As an optional embodiment, the asset data to be analyzed is input into a preset industry classification model to obtain the industry classification label of the asset data to be analyzed. Specifically, it may include: Step 101: Constructing an industry classification model.

[0022] As an optional embodiment, constructing an industry classification model may specifically include: Step 1011: Obtaining a training sample set; wherein the training sample set includes structured data samples, text data samples, and image data samples labeled with industry classification tags.

[0023] In this step, the first step in building the industry classification model is to obtain a high-quality training sample set. This training sample set is the foundation for the model's learning and optimization. Its construction process is systematic and targeted, and the sample set needs to comprehensively cover multiple data modalities to ensure that the subsequently constructed classification model has the ability to handle heterogeneous data. In this application, the training sample set includes three main types of data samples: structured data samples, text data samples, and image data samples. Each type of data sample is precisely labeled with its corresponding industry affiliation tag based on its source and content. By introducing these labeled samples, the model can learn the characteristic patterns and inherent laws exhibited by different industries under different data modalities, thereby ensuring that the final integrated industry classification model can accurately and reliably classify and identify the industry of various asset data to be analyzed.

[0024] Step 1012: Use the decision tree algorithm to train the structured data samples to generate the first sub-model.

[0025] In this step, the structured data samples in the training sample set are input into the decision tree algorithm. The algorithm automatically learns and establishes a set of classification rules for industry discrimination by analyzing the inherent relationship between different feature fields and corresponding industry labels in these samples. This set of rules is presented in a tree structure, where each internal node represents a judgment on a data feature, each branch represents a judgment result, and each leaf node corresponds to a final industry classification conclusion. The training process is essentially about finding the features and thresholds that best distinguish different industry categories, so that the generated decision tree model can automatically and efficiently complete industry classification based on the input structured data. Finally, the first classification sub-model obtained through this training process has the ability to accurately classify new, unlabeled structured asset data into the corresponding industry category.

[0026] Step 1013: Use the BERT model to train the text data samples to generate a second classification sub-model.

[0027] In this step, for training the text data samples in the training sample set, firstly, the text data samples labeled with industry classification tags are input into the BERT model. Through the deep language understanding capabilities obtained from pre-training on massive amounts of text, the BERT model can extract semantic features highly related to industry attributes from the input text samples. These features go beyond shallow word matching and can capture subtle differences in language style, terminology usage, and contextual semantics of specific industries. During training, the model continuously adjusts its internal parameters through supervised learning, aiming to enable it to map the deep semantic features extracted from the text data to the correct industry classification tags. This process essentially teaches the model how to accurately distinguish based on the industry characteristics contained in the text content. Finally, the trained second classification sub-model becomes an industry classifier specifically designed to process text modal data, possessing the ability to automatically identify and output the industry classification of new textual asset data.

[0028] Step 1014: Use the CLIP model to train the image data samples to generate a third-class sub-model.

[0029] In this step, the CLIP model is used as the basic framework for training image data samples in the training sample set. The CLIP model possesses powerful cross-modal understanding capabilities due to its association training of images and text in a unified semantic space, and is specifically used here to identify industry features from image data. During training, image data samples with precisely labeled industry classifications are input into the model, guiding it to learn to deeply associate visual information in the images, such as typical scene compositions, specific object types, symbols, or chart formats, with corresponding industry semantic concepts. By learning from these labeled samples, the model continuously optimizes its internal parameters, gradually mastering the ability to map the visual features of input images to specific industry classifications. The resulting third-classification sub-model becomes a dedicated industry classifier for processing image modal data. It can independently perform reasoning analysis on new image-based asset data, outputting its industry category, effectively expanding the comprehensive analysis scope of the entire system for heterogeneous data assets.

[0030] Step 1015: Integrate the first sub-model, the second sub-model, and the third sub-model to obtain the industry classification model.

[0031] In this step, after training three independent sub-models for structured data, text data, and image data respectively, the three independent sub-models are integrated into a unified and fully functional industry classification model. This integration process is not a simple model stacking, but a design to build a multimodal classification system. The integrated industry classification model, as a whole, encapsulates professional processing capabilities for different data modalities. The model has an intelligent routing judgment mechanism that can automatically schedule and allocate the input data to the corresponding optimal sub-model for processing based on the specific modal characteristics of the input data.

[0032] Step 102: Parse the data modality of the asset data to be analyzed to obtain the data modality type.

[0033] In this step, the system first receives the input asset data to be analyzed and detects and analyzes its data format, internal structure, and content characteristics. Through pre-set parsing logic and feature recognition mechanisms, it can effectively distinguish whether the data is structured data with regular fields and table forms, text data composed of natural language, or image data containing visual information. This automated parsing process does not rely on manual intervention and can accurately output the judgment result of the data modality type. This type judgment is the decision basis of the entire workflow. It directly determines which of the three professional sub-models will be called to perform the specific industry classification task, thereby ensuring the efficiency and professionalism of the entire analysis system when processing heterogeneous data assets.

[0034] Step 103: Select the target classification sub-model corresponding to the data modality type from the industry classification model, and use the target classification sub-model to process the asset data to be analyzed in order to obtain the industry classification label of the asset data to be analyzed.

[0035] In this step, the data modality type obtained through parsing is used as input for the selection logic. A target classification sub-model that precisely matches the modality is invoked. For example, if the data is determined to be structured data, a first classification sub-model based on decision trees is used; if it is determined to be text data, a second classification sub-model based on BERT is used; and if it is determined to be image data, a third classification sub-model based on CLIP is used. After selecting the target classification sub-model, the asset data to be analyzed is input for processing. This target sub-model utilizes its internally learned industry-specific discrimination knowledge and feature extraction capabilities for this modality to perform in-depth analysis and inference calculations on the input data, ultimately outputting a clear classification result that characterizes the data's industry affiliation—the industry classification label. Obtaining this label marks the completion of the industry classification stage, laying a crucial foundation for subsequent in-depth value analysis based on industry knowledge.

[0036] Step 20: Based on industry classification tags, match them with tags in the preset industry knowledge base to obtain industry knowledge fragments that are most relevant to the asset data to be analyzed.

[0037] As an optional implementation, based on industry classification tags, the tags are matched with those in a preset industry knowledge base to obtain the industry knowledge fragments that are most relevant to the asset data to be analyzed. Specifically, this includes: Step 201: Constructing an industry knowledge base.

[0038] As an optional implementation, building an industry knowledge base may specifically include: Step 2011: collecting standard documents and data standards from various industries to generate original knowledge materials.

[0039] In this step, to build an industry knowledge base, we first target a wide range of industry sectors and collect publicly released or internally circulated normative documents. These materials typically include, but are not limited to, industry-wide data governance frameworks, business terminology standards, data quality evaluation criteria, data security and compliance requirements, and data value assessment guidelines for specific fields. By massively aggregating these reliable and authoritative documents, we form a collection of original knowledge materials covering multiple industries and dimensions of professional knowledge. This collection constitutes the underlying data foundation of the industry knowledge base. The quality and breadth of its content directly determine the depth and professionalism of the industry background knowledge that can be accessed in future data asset analysis, providing a fundamental information source for subsequent knowledge structuring, tagging, and ultimately intelligent retrieval and application.

[0040] Step 2012: Preprocess the original knowledge materials to obtain standard knowledge materials, and extract key features from the standard knowledge materials to generate knowledge tags corresponding to the standard knowledge materials.

[0041] In this step, after collecting the original knowledge materials, they need to be systematically preprocessed to form standard knowledge materials. The preprocessing stage includes a series of operations such as text cleaning, format unification, and structural standardization of the original materials, aiming to eliminate noise and non-standard content in the source data and ensure that the knowledge materials have a consistent expression form and a regular data structure. Subsequently, key features are extracted from the preprocessed standard knowledge materials. This process uses natural language processing and feature engineering techniques to identify and extract semantic features that can represent core industry concepts, key technical indicators, typical business scenarios, and important evaluation dimensions, forming a machine-readable expression of industry knowledge. Based on the extracted key features, knowledge tags that precisely correspond to each piece of standard knowledge material are further generated. These tags are highly condensed and semantically summarized knowledge content, representing the core industry attributes and professional scope covered by the knowledge fragment, thereby effectively supporting knowledge enhancement and contextual understanding in the data asset analysis process.

[0042] Step 2013: Construct an industry knowledge base based on key features and knowledge tags.

[0043] In this step, the three elements of previously obtained standard knowledge materials, key features extracted from their content, and corresponding knowledge tags are organically integrated to establish internal connections for each knowledge item. This creates a close correspondence between the original content, the feature vector representing its core information, and the tags that serve as semantic identifiers. Based on this connection structure, all knowledge items are organized and stored in an orderly manner according to their inherent industry attributes and logical categories. This process constructs a structured professional knowledge container rich in semantic information. The resulting industry knowledge base is not only a static knowledge repository but also a dynamic system with efficient indexing and retrieval capabilities. This allows for the rapid and accurate location and retrieval of the most relevant professional knowledge fragments based on the industry classification tags of the data in subsequent data asset analysis processes. This provides deep domain context support for big data models, enabling more precise and professional data asset analysis.

[0044] Step 202: Calculate the semantic similarity between the industry category tags and the tags in the industry knowledge base.

[0045] In this step, the industry classification tags and all predefined knowledge tags in the knowledge base are first mapped to vector representations in a high-dimensional semantic space using natural language processing technology. The semantic connotation of each tag is thus transformed into a feature vector with mathematical meaning. Based on this, a specific similarity calculation algorithm is used to measure the proximity of the vector of the industry classification tag of the data to be analyzed to the vector of each knowledge tag in the knowledge base in the semantic space. Through this semantic similarity, a quantitative relevance score is generated for each tag in the knowledge base. This score provides the core basis for the accurate screening and sorting of knowledge fragments, ensuring that the professional knowledge content that best fits the current industry background of the data assets can be dynamically and intelligently acquired in the future.

[0046] Step 203: Based on semantic similarity, sort the knowledge fragments in the industry knowledge base to obtain the sorting results.

[0047] In this step, after calculating the semantic similarity between industry classification tags and various knowledge tags in the industry knowledge base, the corresponding knowledge fragments in the knowledge base are systematically sorted based on these quantified similarity values. This process is achieved by calling a specific sorting algorithm, which uses semantic similarity as the core comparison indicator to reorganize all knowledge fragments involved in the matching. The sorting operation arranges all candidate knowledge fragments in descending order according to the semantic relevance strength, generating an ordered list of knowledge fragments. This sorting result intuitively shows the degree of relevance between different knowledge fragments and the data assets to be analyzed in the industry context, providing a direct basis for achieving accurate knowledge retrieval and effectively ensuring that the knowledge content finally selected and injected into the large language model has the highest industry relevance and contextual relevance.

[0048] Step 204: Select the knowledge fragment with the highest semantic similarity from the sorting results to obtain industry knowledge fragments.

[0049] In this step, after systematically ranking the knowledge fragments in the industry knowledge base based on semantic similarity, the knowledge fragment ranked first in the ranking results is located. This fragment is identified as the most relevant professional knowledge content because it has the highest proximity in semantic space to the industry classification label of the asset data to be analyzed. This fragment is then extracted from the knowledge base and identified as the industry knowledge fragment for subsequent analysis. This selection mechanism ensures that the obtained knowledge fragment has the strongest semantic consistency and background relevance with the current data assets in terms of industry context, professional terminology, and evaluation dimensions. This provides the most accurate and effective domain knowledge support for subsequently constructing high-quality prompt words and inputting them into a large language model for in-depth analysis.

[0050] Step 30: Input industry knowledge fragments into the preset large language model to obtain the analysis results of the asset data to be analyzed.

[0051] As an optional embodiment, industry knowledge fragments are input into a preset large language model to obtain the analysis results of the asset data to be analyzed. Specifically, this may include: Step 301: Associating industry knowledge fragments with the asset data to be analyzed to generate input prompts.

[0052] In this step, after obtaining precisely selected industry knowledge fragments, they are intelligently associated with the asset data to be analyzed. This association process is not a simple information splicing, but a structured prompt word construction process. By organically combining industry knowledge fragments as key background context and professional analytical basis with the specific content of the asset data to be analyzed, the generated input prompts typically contain three core parts: the first part is the injected industry knowledge fragment, which sets the professional industry background and evaluation criteria for the analysis task; the second part is the asset data to be analyzed itself, as the core object to be analyzed; and the third part is the specific analysis instructions based on this context, guiding the large language model to use the aforementioned professional knowledge to conduct in-depth analysis of the given data. Through this structured association method, the generated input prompts can effectively integrate industry professional knowledge, specific data content, and analysis task objectives into a coherent whole, providing the large language model with sufficient and accurate contextual information, thereby significantly enhancing the professionalism, accuracy, and reliability of its subsequent data quality assessment and value analysis.

[0053] Step 302: Input the input prompts into the large language model to generate quality assessment content and data value assessment content for the asset data to be analyzed.

[0054] In this step, the large language model receives a complete prompt containing industry knowledge fragments, the asset data to be analyzed, and analysis instructions. It then performs deep semantic understanding and logical reasoning. First, the model fully understands the professional background, evaluation dimensions, and industry standards defined by the injected industry knowledge fragments, using them as the basic framework for analysis. Subsequently, within this professional framework, the model performs multi-dimensional analysis of the simultaneously provided asset data. Based on its powerful language understanding and knowledge reasoning capabilities, the model compares, correlates, and deeply assesses the characteristics of the data with the requirements of industry knowledge. During processing, the model generates two aspects of professional analysis content in parallel: first, a quality assessment of the data assets, systematically analyzing their performance in terms of accuracy, completeness, consistency, and compliance; second, a value assessment of the data assets, deeply exploring their potential value and utility in business applications, decision support, process optimization, or risk control. Finally, the model outputs a structured, comprehensive analysis that deeply integrates industry professional insights, thereby achieving intelligent and professional evaluation of data assets from quality to value.

[0055] Step 303: Integrate and verify the quality assessment content and data value assessment content to obtain the analysis results of the asset data to be analyzed.

[0056] In this step, after obtaining the data quality assessment content and data value assessment content generated separately by the large language model, the two assessment contents are first merged and organized into a unified, structured output framework. The logical consistency of the two contents is then verified to ensure that the quality judgments and value inferences derived from the same data characteristics and industry background are mutually supportive and do not contain semantic or logical contradictions. Based on this, the core findings are extracted and summarized to generate a comprehensive analysis report that is complete, well-organized, and has clear conclusions. This final analysis result not only comprehensively reflects the performance of the asset data under analysis in key quality dimensions, but also deeply explains its potential value and applicable scenarios in a specific industry context, thus providing a direct and reliable decision-making basis for the management and value realization of data assets.

[0057] Step 304: Obtain application feedback information for the asset data to be analyzed; wherein, the application feedback information includes the accuracy evaluation and applicability evaluation of the analysis results in the data usage scenarios.

[0058] In this step, after completing the analysis of data assets and outputting preliminary results, feedback information on the application of the analyzed asset data in actual business scenarios is obtained. This feedback information comes from the actual application and verification process of the data asset analysis results in subsequent real business activities, and mainly includes two dimensions of evaluation: the first is the evaluation of the accuracy of the analysis results, that is, in actual use, verifying whether the data quality assessment conclusions are consistent with the true state of the data, and whether the prediction of data value assessment is consistent with the actual business utility generated by the data; the second is the evaluation of the applicability of the analysis results, that is, assessing the relevance, operability and practical effect of the analysis conclusions in actual scenarios such as guiding specific business decisions, supporting data transactions or driving data product development.

[0059] Step 305: Based on the application feedback information, calculate the degree of deviation between the analysis results and the actual application requirements.

[0060] In this step, after obtaining feedback information on the application of data asset analysis results in actual business scenarios, the collected accuracy and applicability evaluations are first structurally analyzed and transformed into quantifiable evaluation indicators. The actual data status and business utility reflected in the feedback information are systematically compared and analyzed with the quality judgments and value inferences in the previous analysis report. Through specific evaluation algorithms, the differences between the analysis results and the requirements of real application scenarios are quantitatively measured from multiple dimensions such as semantic consistency, conclusion consistency, and business fit. The final deviation results provide clear optimization directions and calibration basis for subsequent knowledge base updates and model optimization, thereby promoting the continuous improvement of the accuracy and adaptive evolution of the entire analysis system.

[0061] Step 306: If the deviation exceeds the preset threshold, the industry knowledge base and industry classification model are iteratively optimized based on the application feedback information until an analysis result that meets the actual application requirements is obtained.

[0062] In this step, when the calculated deviation exceeds a preset allowable threshold, the core components will undergo targeted iterative optimization based on collected application feedback. This optimization process is carried out simultaneously on the industry knowledge base and the industry classification model. For the industry knowledge base, based on the knowledge gaps or cognitive biases revealed in the feedback information, the corresponding knowledge entries are corrected, supplemented, or updated to ensure that the industry standards, evaluation dimensions, and professional knowledge it contains can more accurately reflect the needs of actual business scenarios. At the same time, for the industry classification model, application feedback data containing correct industry labels is used as new training samples and re-input into the model training process. Through incremental learning or model fine-tuning techniques, the model parameters are adjusted to make its industry classification judgment more accurate. This iterative optimization process will continue until the deviation calculated based on the new feedback falls back to within the preset threshold. At this point, the analysis results output by the system are determined to meet the actual application needs. This mechanism ensures that the entire analysis system has the ability to continuously learn and improve itself, and can continuously adapt to changes and developments in business scenarios.

[0063] Step 307: Extract the quality assessment content and data value assessment content, and add corresponding industry classification tags, timestamps, and knowledge base matching source identifiers to generate quality assessment sub-reports and value assessment sub-reports.

[0064] In this step, the core arguments specifically for data quality assessment and data value assessment are extracted from the integrated content. Then, important metadata information is added to these extracted contents, including the industry classification label determined by the industry classification model to clarify its professional field background; a timestamp indicating the time context of the report generation to record the analysis operation; and a matching source identifier from the associated industry knowledge base to trace the authoritative knowledge supporting the analysis. Based on the above processing, the quality assessment content and value assessment content, each with complete metadata, are independently packaged into a structured quality assessment sub-report. These two sub-reports together constitute the total output of the data asset analysis. They possess independent reference value and are interconnected through shared metadata, ensuring the traceability of the analysis process, the domain relevance of the analysis conclusions, and the completeness and professionalism of the reporting system.

[0065] Step 308: Integrate the quality assessment sub-report and the value assessment sub-report into a standardized analysis document for data traceability.

[0066] In this step, after generating separate quality assessment and value assessment sub-reports, they are integrated into a unified, standardized analysis document. This integration process is not a simple content merging, but rather follows a predefined document structure and data specifications. The two sub-reports are used as core modules, and their associated metadata is systematically organized. The document establishes a complete data traceability chain by embedding key metadata such as industry classification tags for data assets, timestamps of analysis operations, and source identifiers of referenced knowledge bases. This structured design ensures that the document not only fully presents the comprehensive assessment conclusions of the data assets but also possesses powerful traceability capabilities. Any analysis conclusion can be traced back to the industry knowledge fragment upon which it is based using the source identifier in the document, its timeliness can be confirmed by the timestamp, and its professional context can be understood based on the industry classification tags. Ultimately, this standardized analysis document serves as the authoritative output of intelligent data asset analysis, providing a reliable and traceable complete basis for the auditing, trading, management, and value re-mining of data assets.

[0067] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a data asset analysis device, the structure of which is as follows: Figure 2 As shown.

[0068] Figure 2 This is a schematic diagram of the internal structure of a data asset analysis device provided in an embodiment of this application. Figure 2 As shown, the device includes: At least one processor 201; And a memory 202 that is communicatively connected to at least one processor; The memory 202 stores instructions that can be executed by at least one processor. The instructions are executed by at least one processor 201 to enable at least one processor 201 to: input the asset data to be analyzed into a preset industry classification model to obtain industry classification labels for the asset data to be analyzed; match the industry classification labels with labels in a preset industry knowledge base to obtain industry knowledge fragments with the highest relevance to the asset data to be analyzed; and input the industry knowledge fragments into a preset large language model to obtain the analysis results of the asset data to be analyzed.

[0069] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium for data asset analysis stores computer-executable instructions, which are configured to: input the asset data to be analyzed into a preset industry classification model to obtain industry classification labels for the asset data to be analyzed; match the industry classification labels with labels in a preset industry knowledge base to obtain industry knowledge fragments with the highest relevance to the asset data to be analyzed; and input the industry knowledge fragments into a preset large language model to obtain the analysis results of the asset data to be analyzed.

[0070] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0071] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0072] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0073] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0074] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0075] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0076] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0077] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0078] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0079] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0080] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A data asset analysis method characterized by, The method comprises: inputting asset data to be analyzed into a preset industry classification model to obtain an industry classification label of the asset data to be analyzed; wherein the industry classification model comprises classification algorithms corresponding to different data modalities; based on the industry classification label, matching with labels in a preset industry knowledge base to obtain an industry knowledge segment with the highest correlation degree with the asset data to be analyzed; wherein the industry knowledge base contains industry knowledge for data resource quality analysis and value analysis; inputting the industry knowledge segment into a preset large language model to obtain an analysis result of the asset data to be analyzed.

2. The data asset analysis method of claim 1, wherein, The method further comprises: constructing an industry classification model, specifically comprising: obtaining a training sample set; wherein the training sample set comprises structured data samples, text data samples and image data samples with labeled industry classification labels; training the structured data samples using a decision tree algorithm to generate a first classification sub-model; training the text data samples using a BERT model to generate a second classification sub-model; training the image data samples using a CLIP model to generate a third classification sub-model; integrating the first, second and third classification sub-models to obtain the industry classification model.

3. The data asset analysis method of claim 1, wherein, Inputting asset data to be analyzed into a preset industry classification model to obtain an industry classification label of the asset data to be analyzed, specifically comprising: analyzing the data modality of the asset data to be analyzed to obtain a data modality type; selecting a target classification sub-model corresponding to the data modality type from the industry classification model, and processing the asset data to be analyzed using the target classification sub-model to obtain an industry classification label of the asset data to be analyzed.

4. The data asset analysis method of claim 1, wherein, The method further comprises: constructing an industry knowledge base, specifically comprising: collecting specification documents and data standards of various industries to generate original knowledge materials; preprocessing the original knowledge materials to obtain standard knowledge materials, and extracting key features from the standard knowledge materials to generate knowledge labels corresponding to the standard knowledge materials; based on the key features and the knowledge labels, constructing the industry knowledge base.

5. The data asset analysis method of claim 1, wherein, Based on the industry classification label, matching with labels in a preset industry knowledge base to obtain an industry knowledge segment with the highest correlation degree with the asset data to be analyzed, specifically comprising: calculating the semantic similarity between the industry classification label and each label in the industry knowledge base; based on the semantic similarity, sorting the knowledge segments in the industry knowledge base to obtain a sorting result; selecting a knowledge segment with the highest semantic similarity from the sorting result to obtain the industry knowledge segment.

6. The data asset analysis method of claim 1, wherein, Inputting the industry knowledge segment into a preset large language model to obtain an analysis result of the asset data to be analyzed, specifically comprising: associating the industry knowledge segment with the asset data to be analyzed to generate an input prompt; inputting the input prompt into the large language model to generate quality evaluation content and data value evaluation content of the asset data to be analyzed; The quality evaluation content and the data value evaluation content are integrated and verified to obtain an analysis result of the asset data to be analyzed.

7. The data asset analysis method of claim 1, wherein, After obtaining the analysis result of the asset data to be analyzed, the method further comprises: obtaining application feedback information of the asset data to be analyzed; wherein the application feedback information comprises accuracy evaluation and applicability evaluation of the analysis result in a data use scenario; based on the application feedback information, calculating a deviation degree of the analysis result from actual application requirements; if the deviation degree exceeds a preset threshold, based on the application feedback information, iteratively optimizing the industry knowledge base and the industry classification model until an analysis result meeting the actual application requirements is obtained.

8. The data asset analysis method of claim 6, wherein, The method further comprises: extracting the quality evaluation content and the data value evaluation content, and adding corresponding industry classification labels, time stamps and knowledge base matching source identifiers to generate a quality evaluation sub-report and a value evaluation sub-report; integrating the quality evaluation sub-report and the value evaluation sub-report into a standardized analysis document for data tracing.

9. A data asset analysis device, characterized by, The device comprises: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a method according to any one of claims 1-8.

10. A non-transitory computer storage medium storing computer-executable instructions for data asset analysis, the computer-executable instructions comprising: The computer executable instructions, when executed, implement a method according to any one of claims 1-8.

Images