An AI large model-based management method for enterprise digital transformation data
Patent Information
- Application Number
- CN202611064774.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-17
- Publication Date
- 2026-09-18
AI Technical Summary
[0003]现有的企业数字化转型数据管理方法通过人工进行管理，当私密数据内容过多时，就会降低人工管理速度，进而会导致数据管理周期增长
本方案首先通过前置数据预处理步骤，提前移除企业数字化转型数据中的噪音数据、冗余标点、无效特殊字符等大量干扰内容，过滤掉无意义的无效数据片段，直接降低后续环节的运算负载；同时依托Transformer模型的多头编码器自注意力层的并行处理能力，可同时对多维度的私密数据特征同步进行提取分析，替代传统串行运算的模式，完全摆脱传统人工管理速度慢的限制。
Smart Images
Figure CN122777526A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data management technology, specifically to a management method for enterprise digital transformation data based on AI big data models. Background Technology
[0002] Enterprise digital transformation data refers to the end-to-end data resources generated, collected, governed, applied, and accumulated by enterprises during the process of promoting digital transformation, as well as the technologies, standards, and management systems that support these data. It is not only a "record" of enterprise business operations, but also a core asset driving digital decision-making, process reshaping, and business innovation.
[0003] Existing methods for managing enterprise digital transformation data rely on manual processes. When there is too much private data, this slows down manual management and leads to longer data management cycles. Summary of the Invention
[0004] To address the aforementioned technical issues, this paper provides a management method for enterprise digital transformation data based on an AI-powered large-scale model. This technical solution resolves the problems mentioned in the background section.
[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A data management method for enterprise digital transformation based on AI big data models includes: Acquire enterprise digital transformation data, perform data preprocessing on the enterprise digital transformation data based on the data management terminal, and obtain the preprocessed enterprise digital transformation data; Based on the data management terminal, data analysis and processing are performed on the pre-processed enterprise digital transformation data to identify private data in the enterprise digital transformation data. Based on the data management terminal, the pre-processed enterprise digital transformation data is vectorized to obtain vectorized enterprise digital transformation data. Based on the data management terminal, the vectorized enterprise digital transformation data is analyzed and processed in the Transformer model to obtain private data to be verified. Based on the data management terminal, the system performs data verification processing on the private data to be verified in the enterprise's digital transformation data, and selects whether the Transformer model needs to be retrained.
[0006] Preferably, the acquisition of enterprise digital transformation data, based on a data management terminal, involves data preprocessing of the enterprise digital transformation data to obtain the preprocessed enterprise digital transformation data, specifically including the following steps: Based on the data management terminal, data is retrieved and processed from the enterprise's database system to obtain data on the enterprise's digital transformation. Based on the data management terminal, data cleaning processing is performed on enterprise digital transformation data to obtain cleaned enterprise digital transformation data. The data cleaning processing specifically involves removing noisy data, punctuation marks, and special characters. Based on the data management terminal, the cleaned enterprise digital transformation data is segmented using a word segmentation tool to obtain pre-processed enterprise digital transformation data. In the data cleaning step, full-width characters in the text information are converted to half-width characters, and duplicate and redundant characters are deduplicated. After cleaning, additional data validity verification is performed to filter out invalid text segments with a character ratio lower than a preset threshold.
[0007] Preferably, the step of performing data analysis and processing on the preprocessed enterprise digital transformation data based on the data management terminal to determine the private data in the enterprise digital transformation data specifically includes the following steps: Based on the data management terminal, data is retrieved and processed from the enterprise's database system to obtain the criteria for judging private data. Based on the data management terminal, the pre-processed enterprise digital transformation data is filtered and processed according to the private data evaluation criteria to obtain private data in the enterprise digital transformation data. The privacy data evaluation criteria adopt a hierarchical definition mechanism, covering privacy data judgment rules in at least four dimensions: personal identity information, financial information, medical information, and trade secrets. Different dimensions of privacy data are configured with independent feature keyword libraries and privacy level tags.
[0008] Preferably, the step of vectorizing the preprocessed enterprise digital transformation data based on the data management terminal to obtain vectorized enterprise digital transformation data specifically includes the following steps: Based on the data management terminal, feature extraction processing is performed on the pre-processed enterprise digital transformation data to obtain context features, phrase part-of-speech features, and phrase energy parameters. Based on the data management terminal, the context features, phrase part-of-speech features and phrase energy parameters are spliced together to obtain several sets of feature vectors with different parts of speech. Based on the data management terminal, several sets of feature vectors with different parts of speech are integrated and processed to obtain vectorized enterprise digital transformation data. The phrase energy parameter is a quantified value calculated based on the frequency of phrase occurrence and semantic weight, and the parameter value range maps the degree of suspicion of the corresponding sensitive data.
[0009] Preferably, the step of performing data analysis and processing on vectorized enterprise digital transformation data in the Transformer model based on the data management terminal to obtain the private data to be verified specifically includes the following steps: Based on the data management terminal, feature extraction processing is performed on vectorized enterprise digital transformation data to obtain context vectors in Transformer model format; Based on the data management terminal, context vectors in Transformer model format are input into the self-attention layer of the multi-head encoder of the Transformer model; Based on the data management terminal, the multi-head encoder self-attention layer of the Transformer model is controlled to reorganize the context vector with Transformer model format to obtain the private data to be verified.
[0010] Preferably, the step of performing feature extraction processing on vectorized enterprise digital transformation data based on the data management terminal to obtain a context vector with a Transformer model format specifically includes the following steps: Based on the data management terminal, several sets of different part-of-speech feature vectors are analyzed and processed to determine the contextual relationship between different part-of-speech feature vectors. Based on the data management terminal and the contextual relationship between different part-of-speech feature vectors, the vectorized enterprise digital transformation data is converted into a format using the Transformer model format converter to obtain context vectors with Transformer model format.
[0011] Preferably, the step of performing data verification processing on the private data to be verified based on the private data in the enterprise's digital transformation data, and selecting whether the Transformer model needs to be retrained, specifically includes the following steps: Based on the data management terminal, data judgment and processing are performed on private data and unverified private data in the enterprise's digital transformation data; If the private data in the enterprise's digital transformation data is exactly the same as the private data to be verified, there is no need to retrain the Transformer model; If the private data in the enterprise's digital transformation data is not exactly the same as the private data to be verified, the Transformer model should be retrained based on the data management terminal.
[0012] Furthermore, a data management system for enterprise digital transformation based on an AI big data model is proposed to implement the aforementioned data management method for enterprise digital transformation based on an AI big data model, including: A data management terminal, which is used to control data transmission and information interaction between various modules; A database system is used to store enterprise digital transformation data and private data evaluation criteria; The data preprocessing module is used to preprocess enterprise digital transformation data to obtain preprocessed enterprise digital transformation data. The private data generation module performs data filtering on the preprocessed enterprise digital transformation data according to the private data evaluation criteria to obtain private data in the enterprise digital transformation data. The vector quantization encoding module is used to perform vectorization processing on the preprocessed enterprise digital transformation data to obtain vectorized enterprise digital transformation data. The Transformer intelligent audit module is used to perform data analysis and processing on vectorized enterprise digital transformation data to obtain private data to be verified. The model accuracy verification module performs data verification processing on the private data to be verified based on the private data in the enterprise's digital transformation data, and selects whether the Transformer model needs to be retrained.
[0013] Compared with existing technologies, this invention provides a method for managing enterprise digital transformation data based on AI large-scale models, which has the following beneficial effects: This solution first removes a large amount of interference, such as noisy data, redundant punctuation, and invalid special characters, from the enterprise's digital transformation data through a pre-processing step, filtering out meaningless invalid data fragments and directly reducing the computational load of subsequent steps. At the same time, relying on the parallel processing capability of the Transformer model's multi-head encoder self-attention layer, it can simultaneously extract and analyze multi-dimensional private data features, replacing the traditional serial computing mode and completely eliminating the limitations of slow traditional manual management. Attached Figure Description
[0014] Figure 1 This is a flowchart illustrating steps S100-S500 of a data management method for enterprise digital transformation based on an AI large model proposed in this invention. Figure 2 This is a structural block diagram of a management system for enterprise digital transformation data based on an AI large model, as proposed in this invention. Detailed Implementation
[0015] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.
[0016] Reference Figure 1 As shown, a data management method for enterprise digital transformation based on AI big data models includes: S100. Obtain enterprise digital transformation data. Based on the data management terminal, perform data preprocessing on the enterprise digital transformation data and obtain the preprocessed enterprise digital transformation data. S200: Based on the data management terminal, perform data analysis and processing on the pre-processed enterprise digital transformation data to identify private data in the enterprise digital transformation data; S300, based on the data management terminal, performs vectorization processing on the pre-processed enterprise digital transformation data to obtain vectorized enterprise digital transformation data; S400, based on the data management terminal, performs data analysis and processing on vectorized enterprise digital transformation data in the Transformer model to obtain private data to be verified; S500, based on the data management terminal, performs data verification processing on the private data to be verified in the enterprise's digital transformation data, and selects whether the Transformer model needs to be retrained.
[0017] Example 1 Step S100: Obtain enterprise digital transformation data. Based on the data management terminal, perform data preprocessing on the enterprise digital transformation data to obtain the preprocessed enterprise digital transformation data. This specifically includes the following steps: S101. Based on the data management terminal, retrieve and process data from the enterprise's database system to obtain data on the enterprise's digital transformation. S102. Based on the data management terminal, perform data cleaning processing on the enterprise digital transformation data to obtain the cleaned enterprise digital transformation data. The data cleaning processing specifically involves removing noisy data, punctuation marks, and special characters. In the data cleaning process, the full-width characters in the enterprise's digital transformation data are converted to half-width characters, and duplicate and redundant characters are deduplicated. After cleaning, additional data validity verification is performed to filter out invalid text fragments with a character ratio lower than a preset threshold. S103. Based on the data management terminal, the cleaned enterprise digital transformation data is segmented using a word segmentation tool to obtain pre-processed enterprise digital transformation data. In the data cleaning step, full-width characters in the text information are converted to half-width characters, and duplicate and redundant characters are deduplicated. After cleaning, additional data validity verification is performed to filter out invalid text segments with a character ratio lower than a preset threshold.
[0018] Example 2 Step S200: Based on the data management terminal, perform data analysis and processing on the preprocessed enterprise digital transformation data to determine the private data in the enterprise digital transformation data, specifically including the following steps: S201. Based on the data management terminal, retrieve and process data from the enterprise's database system to obtain private data evaluation criteria. S202. Based on the data management terminal, perform data filtering and processing on the pre-processed enterprise digital transformation data according to the private data evaluation criteria to obtain private data in the enterprise digital transformation data. The privacy data evaluation criteria adopt a hierarchical definition mechanism, covering at least four dimensions of privacy data judgment rules: personal identity information, financial information, medical information, and trade secrets. Different dimensions of privacy data are configured with independent feature keyword libraries and privacy level tags. In addition, during the data retrieval and processing steps of the enterprise's database system, the data source path, text location, and privacy level identifier of each extracted private data content are recorded synchronously to generate a private data traceability mapping table.
[0019] Example 3 Step S300: Based on the data management terminal, the preprocessed enterprise digital transformation data is vectorized to obtain vectorized enterprise digital transformation data. This specifically includes the following steps: S301. Based on the data management terminal, perform feature extraction processing on the preprocessed enterprise digital transformation data to obtain context features, phrase part-of-speech features, and phrase energy parameters. S302. Based on the data management terminal, perform data splicing processing on context features, phrase part-of-speech features and phrase energy parameters to obtain several sets of feature vectors with different parts of speech. S303. Based on the data management terminal, integrate and process the feature vectors of several different parts of speech to obtain vectorized enterprise digital transformation data. The phrase energy parameter is a quantified value calculated based on the frequency of phrase occurrence and semantic weight, and the parameter value range maps the degree of suspicion of the corresponding sensitive data. In addition, the dimensions of different part-of-speech feature vectors are uniformly aligned and set to a preset fixed dimension. During the splicing process, independent weight coefficients are configured for different types of features. The value of the weight coefficients can be dynamically adjusted according to the scenario of private data management.
[0020] Example 4 Step S400: Based on the data management terminal, perform data analysis and processing on the vectorized enterprise digital transformation data in the Transformer model to obtain the private data to be verified. This specifically includes the following steps: S401. Based on the data management terminal, perform feature extraction processing on vectorized enterprise digital transformation data to obtain context vectors with Transformer model format; S402. Based on the data management terminal, the context vector with the Transformer model format is input into the self-attention layer of the multi-head encoder of the Transformer model. The number of heads in the self-attention layer of the multi-head encoder is set to 8 to 16. Each self-attention head is independently assigned a corresponding private feature extraction dimension, and the feature recognition task of different types of private data is completed in parallel. S403. Based on the data management terminal, control the multi-head encoder self-attention layer of the Transformer model to perform data reorganization processing on the context vector with Transformer model format to obtain the private data to be verified. After the data is reorganized, a confidence score for the private data is also generated simultaneously. Only results with a confidence score higher than a preset threshold are included in the final generated private data to be verified. Specifically, the process of extracting features from vectorized enterprise digital transformation data using a data management terminal to obtain context vectors in Transformer model format includes the following steps: S4011. Based on the data management terminal, analyze and process several sets of different part-of-speech feature vectors to determine the contextual relationship between different part-of-speech feature vectors; S4012. Based on the data management terminal and the contextual relationship between different part-of-speech feature vectors, use the Transformer model format converter to convert the vectorized enterprise digital transformation data to obtain context vectors with Transformer model format.
[0021] Example 5 Step S500: Based on the data management terminal, perform data verification processing on the private data to be verified in the enterprise's digital transformation data, and select whether the Transformer model needs to be retrained. This includes the following steps: S501. Based on the data management terminal, perform data judgment and processing on private data and private data to be verified in the enterprise's digital transformation data; S502. If the private data in the enterprise's digital transformation data is exactly the same as the private data to be verified, there is no need to retrain the Transformer model. S503. If the private data in the enterprise's digital transformation data is not completely the same as the private data to be verified, the Transformer model should be retrained based on the data management terminal. In this embodiment, during the judgment and processing of the private data to be verified, two core accuracy indicators, the false negative rate and the false positive rate, are simultaneously counted. Only when the false negative rate is 0 and the false positive rate is less than 0.1%, it is determined that the private data in the enterprise's digital transformation data is completely identical to the private data to be verified.
[0022] The specific embodiments of the present invention are as follows: Step 1: Deploy and connect to the distributed database system of the target enterprise. Pull the full amount of digital transformation data of the enterprise to be managed through the database's built-in read-only data interface. The path identifier of the original data is retained throughout the retrieval process, and no source data in the database is modified.
[0023] The second step involves preprocessing the retrieved raw enterprise digital transformation data. First, a data cleaning process is performed to remove meaningless noise fields, full-width punctuation marks, and garbled special characters. Simultaneously, full-width characters are converted to half-width characters, and duplicate characters are removed. After cleaning, the percentage of valid characters in each text is verified, and invalid short segments with a percentage of less than 30% are removed, resulting in cleaned enterprise digital transformation data. Then, an open-source Chinese word segmentation framework is used to segment the cleaned data. During segmentation, the semantic associations of the original context are preserved, and the logical connections between segmentation units are not disrupted. Finally, the preprocessed enterprise digital transformation data is output.
[0024] Step 3: Retrieve the pre-stored hierarchical and graded privacy data evaluation criteria from the database system. The criteria cover four dimensions: personal identity information, financial information, medical information, and trade secrets. Each dimension is configured with an independent feature keyword library and privacy level labels of 1-5. Based on the criteria, perform a full traversal extraction on the pre-processed enterprise digital transformation data, mark the content that meets the judgment rules as labeled privacy data content, and simultaneously record the data source path, text location and corresponding privacy level of each extraction result to generate a privacy data traceability mapping table.
[0025] Step 4: Connect to the vector quantization encoder to perform vectorization transformation on the preprocessed enterprise digital transformation data. First, extract contextual features and part-of-speech features from the preprocessed data. Simultaneously, calculate phrase energy parameters based on word frequency and semantic weights. The value of the phrase energy parameter directly maps to the privacy suspicion level of the corresponding segment. Align and map the three types of features into a fixed-length 768-dimensional vector. Configure dynamically adjustable weight coefficients for different feature types and perform a concatenation operation to generate multiple sets of different part-of-speech feature vectors. Merge all feature vectors to obtain the complete vectorized enterprise digital transformation data. This step completely converts the original plaintext data into vector form for storage, avoiding leakage of original private data during training and improving data security throughout the entire process.
[0026] Step 5: Training and inference of vectorized data based on the Transformer model: First, analyze the relationship between all different part-of-speech feature vectors to determine the context dependency logic between vectors. Then, use the Transformer's built-in format converter to convert the ordinary vectorized data into context vectors that adapt to the Transformer input rules. In this embodiment, the Transformer's multi-head encoder self-attention layer is set to 12 self-attention heads. Each head is independently assigned a private feature extraction task of one dimension. Multiple attention heads perform feature extraction and recombination operations on the input context vectors in parallel, and infer the generated private data output by the model. At the same time, a corresponding confidence score is generated for each output result. Only results with a confidence score higher than 95% are included in the final private data to be verified.
[0027] Step 6: Execute the Transformer model accuracy verification process. Perform a full comparison and statistical analysis of the manually annotated private data content and the private data to be verified output by the model. Simultaneously calculate the two core indicators: false negative rate and false positive rate. If the false negative rate is 0 and the false positive rate is less than 0.1%, the two types of results are considered to be completely matched, the model meets the admission requirements, and can be directly put into the formal online private data audit process. If the comparison results do not meet the accuracy requirements, a hierarchical fine-tuning mechanism is used to prioritize adjusting the weight parameters of the self-attention head of the corresponding false negative private features, and the iterative training process is restarted. After each round of training, the comparison verification is re-executed until the model accuracy meets the preset standard.
[0028] After the model is officially launched, the system automatically generates desensitization suggestions and risk level labels for newly identified private data, and automatically assembles and outputs standardized management reports. The entire process does not require a lot of manual intervention, avoiding oversights that may occur during manual management.
[0029] The entire process is equipped with an independent data encryption module. All vectorized enterprise digital transformation data is encrypted using national cryptographic algorithms throughout the entire storage and transmission chain, ensuring no plaintext data is leaked and further protecting the security of private data.
[0030] Reference Figure 2 As shown, a management system for enterprise digital transformation data based on an AI big data model is used to implement the aforementioned method for managing enterprise digital transformation data based on an AI big data model, including: A data management terminal, which is used to control data transmission and information interaction between various modules; A database system is used to store enterprise digital transformation data and private data evaluation criteria; The data preprocessing module is used to preprocess enterprise digital transformation data to obtain preprocessed enterprise digital transformation data. The private data generation module performs data filtering on the preprocessed enterprise digital transformation data according to the private data evaluation criteria to obtain private data in the enterprise digital transformation data. The vector quantization encoding module is used to perform vectorization processing on the preprocessed enterprise digital transformation data to obtain vectorized enterprise digital transformation data. The Transformer intelligent audit module is used to perform data analysis and processing on vectorized enterprise digital transformation data to obtain private data to be verified. The model accuracy verification module performs data verification processing on the private data to be verified based on the private data in the enterprise's digital transformation data, and selects whether the Transformer model needs to be retrained.
[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.
Claims
1. A method for managing enterprise digital transformation data based on AI large-scale models, characterized in that, include: Acquire enterprise digital transformation data, perform data preprocessing on the enterprise digital transformation data based on the data management terminal, and obtain the preprocessed enterprise digital transformation data; Based on the data management terminal, data analysis and processing are performed on the pre-processed enterprise digital transformation data to identify private data in the enterprise digital transformation data. Based on the data management terminal, the pre-processed enterprise digital transformation data is vectorized to obtain vectorized enterprise digital transformation data. Based on the data management terminal, the vectorized enterprise digital transformation data is analyzed and processed in the Transformer model to obtain private data to be verified. Based on the data management terminal, the system performs data verification processing on the private data to be verified in the enterprise's digital transformation data, and selects whether the Transformer model needs to be retrained.
2. The method for managing enterprise digital transformation data based on an AI large-scale model according to claim 1, characterized in that, The process of acquiring enterprise digital transformation data, based on a data management terminal, involves preprocessing the enterprise digital transformation data to obtain the preprocessed enterprise digital transformation data, specifically including the following steps: Based on the data management terminal, data is retrieved and processed from the enterprise's database system to obtain data on the enterprise's digital transformation. Based on the data management terminal, data cleaning processing is performed on enterprise digital transformation data to obtain cleaned enterprise digital transformation data. The data cleaning processing specifically involves removing noisy data, punctuation marks, and special characters. Based on the data management terminal, the cleaned enterprise digital transformation data is segmented using a word segmentation tool to obtain pre-processed enterprise digital transformation data. In the data cleaning step, full-width characters in the text information are converted to half-width characters, and duplicate and redundant characters are deduplicated. After cleaning, additional data validity verification is performed to filter out invalid text segments with a character ratio lower than a preset threshold.
3. The method for managing enterprise digital transformation data based on an AI large-scale model according to claim 2, characterized in that, The process of analyzing and processing pre-processed enterprise digital transformation data based on a data management terminal to identify private data within the enterprise digital transformation data includes the following steps: Based on the data management terminal, data is retrieved and processed from the enterprise's database system to obtain the criteria for judging private data. Based on the data management terminal, the pre-processed enterprise digital transformation data is filtered and processed according to the private data evaluation criteria to obtain private data in the enterprise digital transformation data. The privacy data evaluation criteria adopt a hierarchical definition mechanism, covering privacy data judgment rules in at least four dimensions: personal identity information, financial information, medical information, and trade secrets. Different dimensions of privacy data are configured with independent feature keyword libraries and privacy level tags.
4. The method for managing enterprise digital transformation data based on an AI large-scale model according to claim 3, characterized in that, The process of vectorizing preprocessed enterprise digital transformation data based on a data management terminal to obtain vectorized enterprise digital transformation data includes the following steps: Based on the data management terminal, feature extraction processing is performed on the pre-processed enterprise digital transformation data to obtain context features, phrase part-of-speech features, and phrase energy parameters. Based on the data management terminal, the context features, phrase part-of-speech features and phrase energy parameters are spliced together to obtain several sets of feature vectors with different parts of speech. Based on the data management terminal, several sets of feature vectors with different parts of speech are integrated and processed to obtain vectorized enterprise digital transformation data. The phrase energy parameter is a quantified value calculated based on the frequency of phrase occurrence and semantic weight, and the parameter value range maps the degree of suspicion of the corresponding sensitive data.
5. The method for managing enterprise digital transformation data based on an AI large-scale model according to claim 4, characterized in that, The process of analyzing and processing vectorized enterprise digital transformation data in the Transformer model based on the data management terminal to obtain private data to be verified specifically includes the following steps: Based on the data management terminal, feature extraction processing is performed on vectorized enterprise digital transformation data to obtain context vectors in Transformer model format; Based on the data management terminal, context vectors in Transformer model format are input into the self-attention layer of the multi-head encoder of the Transformer model; Based on the data management terminal, the multi-head encoder self-attention layer of the Transformer model is controlled to reorganize the context vector with Transformer model format to obtain the private data to be verified.
6. The method for managing enterprise digital transformation data based on an AI large-scale model according to claim 5, characterized in that, The process of extracting features from vectorized enterprise digital transformation data using a data management terminal to obtain context vectors in Transformer model format includes the following steps: Based on the data management terminal, several sets of different part-of-speech feature vectors are analyzed and processed to determine the contextual relationship between different part-of-speech feature vectors. Based on the data management terminal and the contextual relationship between different part-of-speech feature vectors, the vectorized enterprise digital transformation data is converted into a format using the Transformer model format converter to obtain context vectors with Transformer model format.
7. The method for managing enterprise digital transformation data based on an AI large-scale model according to claim 5, characterized in that, The process of verifying private data to be verified based on private data in enterprise digital transformation data using a data management terminal, and determining whether the Transformer model needs to be retrained, specifically includes the following steps: Based on the data management terminal, data judgment and processing are performed on private data and unverified private data in the enterprise's digital transformation data; If the private data in the enterprise's digital transformation data is exactly the same as the private data to be verified, there is no need to retrain the Transformer model; If the private data in the enterprise's digital transformation data is not exactly the same as the private data to be verified, the Transformer model should be retrained based on the data management terminal.
8. A management system for enterprise digital transformation data based on an AI big data model, used to implement the management method for enterprise digital transformation data based on an AI big data model as described in any one of claims 1-7, characterized in that, include: A data management terminal, which is used to control data transmission and information interaction between various modules; A database system is used to store enterprise digital transformation data and private data evaluation criteria; The data preprocessing module is used to preprocess enterprise digital transformation data to obtain preprocessed enterprise digital transformation data. The private data generation module performs data filtering on the preprocessed enterprise digital transformation data according to the private data evaluation criteria to obtain private data in the enterprise digital transformation data. The vector quantization encoding module is used to perform vectorization processing on the preprocessed enterprise digital transformation data to obtain vectorized enterprise digital transformation data. The Transformer intelligent audit module is used to perform data analysis and processing on vectorized enterprise digital transformation data to obtain private data to be verified. The model accuracy verification module performs data verification processing on the private data to be verified based on the private data in the enterprise's digital transformation data, and selects whether the Transformer model needs to be retrained.