Reliable model management system based on dynamic authority control and automatic deployment
The EasyData Model Management System, with its dynamic access control and automated deployment, solves the problems of fragmented model management, inefficient deployment, and static permissions in traditional data governance systems. It achieves efficient model deployment and secure data governance, improves model reuse rate and system scalability, and is suitable for data sharing scenarios in finance, healthcare, and government.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHANGCHUN WHY E SCI & TECH
- Filing Date
- 2025-12-26
- Publication Date
- 2026-05-05
AI Technical Summary
Traditional data governance systems suffer from fragmented model management, inefficient deployment processes, static access control, insufficient multimodal data processing capabilities, and low model reuse rates. These issues result in high update and maintenance costs, numerous security vulnerabilities, difficulties in compliance auditing, slow business response times, and low model reuse rates.
The EasyData Model Management System, which adopts dynamic access control and automated deployment, integrates dynamic access control, automated deployment, and model classification and storage technologies. Through the architecture design of the front-end interaction layer, core service layer, data persistence layer, and external integration layer, it realizes model classification, version control, multimodal adaptation, and automated deployment. Combined with the hybrid control strategy of RBAC and ABAC, it supports structured and unstructured data governance.
It improved model deployment efficiency, reduced the risk of sensitive data leakage, enhanced system scalability and model reuse rate, met the real-time auditing requirements of regulations such as GDPR, shortened model deployment time, and reduced costs.
Smart Images
Figure CN121980581A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of data governance and artificial intelligence interdisciplinary technology, specifically involving an easy-to-use model management system based on dynamic access control and automated deployment. Background Technology
[0002] With the advancement of the digital economy, data has become a core asset for enterprises. Traditional data governance models, lacking intelligent means, struggle to cope with complex scenarios where unstructured data accounts for over 80%. For example, the financial industry needs to analyze customer behavior data in real time to prevent fraud, while the healthcare industry needs to achieve cross-institutional data collaboration while protecting patient privacy. Against this backdrop, AI-driven data governance technologies (such as automated data cleaning and intelligent lineage analysis) have become essential, and the standardization and dynamism of model management are key to unlocking the value of data.
[0003] Current data governance systems suffer from the following major flaws: (1) Fragmented model management: Traditional data governance systems focus on data integration and cleaning, but have not established a unified architecture for model classification and storage, dynamic access control and automated deployment. For example, data governance models such as financial risk control rules and medical diagnostic algorithms are scattered in different systems, resulting in high update and maintenance costs.
[0004] (2) The deployment process is inefficient and prone to errors. Manual deployment is prone to causing environment configuration errors. For example, an e-commerce company caused a rollback of its data governance model due to version confusion, resulting in several hours of service interruption and increased workload for permission configuration.
[0005] (3) Static Access Control: Most data governance systems use role-based access control (RBAC), which cannot dynamically adjust permissions based on data sensitivity (e.g., public / confidential) or user behavior. On the one hand, fixed permissions can easily lead to security vulnerabilities, and static RBAC policies cannot adapt to dynamic scenarios. For example, when a bank analyst exported customer data in bulk at night, the system did not trigger secondary authentication, resulting in data leakage. At the same time, when the analyst is processing highly sensitive customer data, the system may still grant him data export permissions, leading to leakage risks.
[0006] Furthermore, regulations such as GDPR require data governance to achieve full lifecycle traceability, while traditional data governance systems lack dynamic access control auditing capabilities. Compliance auditing is costly, and static access logs are insufficient to meet the real-time auditing requirements of regulations like GDPR. For example, a multinational corporation needs to hire a dedicated team to analyze data access records, incurring annual costs exceeding one million US dollars.
[0007] (4) Fragmented deployment process: In traditional data governance systems, model updates require manual deployment and cannot be integrated with CI / CD pipelines. For example, a bank's data governance team needs to spend 2 hours to complete model verification, environment configuration and deployment, which seriously affects business response speed.
[0008] (5) Insufficient multimodal data processing capabilities: Traditional data governance systems only support structured data governance and lack classification and quality inspection methods for unstructured data (such as contract texts and medical images). For example, a legal technology company needs to manually review key contract clauses, which is inefficient.
[0009] (6) Low model reuse rate: Traditional data governance systems tightly bind data governance models to business systems, resulting in low model reuse rate. For example, a government affairs platform needs to maintain a separate data quality detection model for each department, which is costly. Summary of the Invention
[0010] To address the problems of low model management and deployment efficiency, static access control, and manual reliance on deployment processes in existing data governance systems, this invention provides an EasyData model management system based on dynamic access control and automated deployment. By integrating dynamic access control, automated deployment, and model classification and storage technologies, this invention improves model deployment efficiency, reduces the risk of sensitive data leakage, enhances system scalability, and increases model reusability. It is suitable for scenarios requiring high security and high availability, such as financial risk control, government data sharing, and medical master data governance. Specifically, the data governance technologies involved in this invention cover data classification, quality inspection, access management, and model lifecycle management; model management technologies mainly include model version control, multimodal adaptation (structured / unstructured data), and automated deployment; dynamic access control technology combines a hybrid control strategy of role-based access control (RBAC) and attribute-based access control (ABAC); and automated process integration technology seamlessly connects with data governance components such as ETL tools, data warehouses, and BI platforms through API interfaces.
[0011] The technical solution adopted by this invention to solve the technical problem is as follows: This invention provides an easy-sense model management system based on dynamic access control and automated deployment, comprising: The front-end interaction layer provides the user interface for interacting with the system; The core service layer is used to implement model classification, version control and metadata management, as well as to dynamically allocate access permissions, automatically deploy models, and adapt data governance models. The data persistence layer is used to store model files and metadata, as well as to store and quickly retrieve unstructured models. The external integration layer is used to interface with the data governance platform to achieve unified management of model assets and to automatically trigger model deployment.
[0012] Furthermore, the front-end interaction layer includes a model management panel, a permission configuration workbench, and a monitoring dashboard; the model management panel is used to display a model classification tree diagram and to filter data governance models according to different dimensions; the permission configuration workbench is used to visually configure the mapping relationship between user roles and data sensitivity; the monitoring dashboard is used to display the model deployment status in real time and generate indicator reports.
[0013] Furthermore, the core service layer includes a model classification and storage engine module, a dynamic access control engine module, an automated deployment service module, and a multimodal data processing adapter module. The model classification and storage engine module is used to classify, version control, and manage metadata of data governance models. The dynamic access control engine module is used to dynamically allocate access permissions based on user roles, data sensitivity, and user behavior. The automated deployment service module is used to integrate with the data governance process to achieve automatic deployment after model updates. The multimodal data processing adapter module is used to adapt governance models for structured and unstructured data.
[0014] Furthermore, the model classification and storage engine module includes a model classification module, a model version control module, and a metadata management module; the model classification module automatically classifies models based on metadata using a decision tree algorithm; the model version control module uses a Git-like version control mechanism to implement model version rollback and branch management; the metadata management module uses Elasticsearch to store model metadata and supports full-text search and correlation analysis.
[0015] Furthermore, the dynamic access control engine module includes a hybrid control model and a real-time decision engine module; the hybrid control model combines RBAC and ABAC, with RBAC defining basic roles and assigning default access permissions; ABAC dynamically adjusts access permissions based on data sensitivity and user behavior; the real-time decision engine module adopts a Redis cached permission strategy and evaluates access permission requests in real time through a rule engine.
[0016] Furthermore, the automated deployment service module includes an API interface and an automated deployment pipeline; the API interface adopts a RESTful API for interfacing with external systems; the automated deployment pipeline integrates Jenkins or GitLab CI to automate the entire process of model verification, environment configuration, and deployment.
[0017] Furthermore, the multimodal data processing adapter module includes a structured data processing module and an unstructured data processing module; the structured data processing module connects to a relational database via JDBC / ODBC and automatically generates SQL scripts to perform data cleaning and classification; the unstructured data processing module integrates OCR and NLP models to extract key information from text and generate structured metadata.
[0018] Furthermore, the data persistence layer includes a model storage module and a permission policy storage module; the model storage module uses MongoDB to store model files and metadata, supporting the storage and fast retrieval of unstructured models; the permission policy storage module uses MySQL to store the mapping relationship between user roles and data sensitivity, as well as audit logs, to ensure ACID transaction consistency.
[0019] Furthermore, the external integration layer includes a data governance tool integration module and a CI / CD pipeline integration module; the data governance tool integration module interfaces with the data governance platform via API to achieve unified management of model assets; the CI / CD pipeline integration module integrates with Jenkins and GitLab CI to support automated triggering of the model automated deployment pipeline.
[0020] This invention provides a method for managing the Yishu model based on dynamic access control and automated deployment, comprising the following steps: Step S1: Model upload; Data engineers upload customer binning models through the front-end interaction layer, the model management panel filters data governance models, and the model classification module automatically classifies the data governance models using a decision tree algorithm and generates version numbers. Step S2: Permission configuration; Administrators configure the mapping relationship between user roles and data sensitivity through the permission configuration workbench; they define basic roles and assign default access permissions through a hybrid control model, and dynamically adjust access permissions based on data sensitivity and user behavior. Step S3: Deployment triggered; After the model passes unit testing, the automated deployment service module calls the API interface to push the model to the production environment data governance cluster through the automated deployment pipeline; Step S4: Real-time governance; After customer data enters the system, the model automatically performs binning operations. The real-time decision engine module monitors access behavior in real time and uses Redis caching permission policies to evaluate access permission requests in real time through the rule engine, blocking large-scale data export requests at unusual times.
[0021] The beneficial effects of this invention are: 1. Improve efficiency and quantify benefits; To address the problems of isolated models, rigid permissions, and inefficient deployment in traditional data governance systems, this invention, for the first time, integrates model classification and storage, dynamic permission control, and automated deployment into a unified architecture. By associating model metadata with permission policies, it automates the entire process from model update to permission adjustment to deployment effectiveness. Model deployment time is reduced from 2 hours of manual operation to 5 minutes, and the deployment success rate is increased to 99.9%. Automated permission allocation reduces the workload of permission configuration by 80%.
[0022] For example, a bank saved over 2 million yuan in annual human resource costs by dynamically adjusting analyst permissions using an ABAC strategy.
[0023] 2. Enhanced security; This invention combines RBAC and ABAC models, supporting dynamic permission allocation based on user roles, data sensitivity, and real-time behavior (such as access time and device IP). Compared to traditional static RBAC, dynamic permission control reduces the risk of sensitive data leakage by 70% and meets the real-time auditing requirements of regulations such as GDPR.
[0024] For example, a certain medical platform automatically freezes abnormal export operations by monitoring user behavior in real time, reducing annual security incidents by 90%.
[0025] 3. Improved scalability; This invention integrates OCR, NLP, and database connectors to support a unified governance model for both structured and unstructured data, solving the problem that traditional data governance systems only support structured data. For example, in the medical industry, the system can automatically extract key indicators from medical record texts to support clinical decision-making.
[0026] This invention supports unstructured data governance. For example, an autonomous driving company integrates image recognition models into its system to achieve credibility assessment of labeled datasets, improving model training efficiency by 3 times.
[0027] This invention enables cross-platform compatibility with mainstream data governance tools (such as Collibra and Alation), reducing integration costs by more than 50%.
[0028] 4. Improved model reuse rate; This invention employs a Git-like version control mechanism combined with Elasticsearch metadata storage to achieve traceability and rapid retrieval of model changes. Model reusability is improved by 50%; for example, government platforms can quickly reuse classification models from other departments, reducing redundant development costs. Attached Figure Description
[0029] Figure 1This invention provides a structural block diagram of an IF model management system based on dynamic access control and automated deployment.
[0030] Figure 2 The flowchart illustrates an IF model management method based on dynamic access control and automated deployment provided by this invention.
[0031] In the diagram, the front-end interaction layer is 1, the model management panel is 101, the permission configuration workbench is 102, and the monitoring dashboard is 103. Core Service Layer 2, Model Classification and Storage Engine Module 201, Model Classification Module 2011, Model Version Control Module 2012, Metadata Management Module 2013, Dynamic Access Control Engine Module 202, Hybrid Control Model 2021, Real-time Decision Engine Module 2022, Automated Deployment Service Module 203, API Interface 2031, Automated Deployment Pipeline 2032, Multimodal Data Processing Adapter Module 204, Structured Data Processing Module 2041, Unstructured Data Processing Module 2042; Data persistence layer 3, model storage module 301, permission policy storage module 302; External integration layer 4, data governance tool integration module 401, CI / CD pipeline integration module 402. Detailed Implementation
[0032] The present invention will be further described in detail below with reference to the accompanying drawings.
[0033] In a first aspect, the present invention provides an easy-to-use model management system based on dynamic access control and automated deployment.
[0034] like Figure 1 As shown, the present invention provides an EasyData Model Management System based on dynamic access control and automated deployment. It adopts a layered architecture design, consisting of a front-end interaction layer 1, a core service layer 2, a data persistence layer 3, and an external integration layer 4. The specific functions of each layer are as follows: Front-end interaction layer 1 primarily provides the user interface for interacting with the system. It mainly includes a model management panel 101, a permission configuration workbench 102, and a monitoring dashboard 103. The model management panel 101 displays a tree diagram of model categories, supporting filtering of data governance models by data type (structured / unstructured), version number, creation time, and other dimensions. The permission configuration workbench 102 visually configures the mapping relationship between user roles and data sensitivity, such as binding the "analyst" role to "public data" sensitivity. The monitoring dashboard 103 displays the real-time model deployment status and generates reports on metrics such as model deployment success rate and deployment time.
[0035] Core Service Layer 2 mainly includes four core modules: Model Classification and Storage Engine Module 201, Dynamic Access Control Engine Module 202, Automated Deployment Service Module 203, and Multimodal Data Processing Adapter Module 204.
[0036] Among them, the model classification and storage engine module 201 is mainly used to classify, version control and manage metadata of data governance models (such as classification rules, cleaning algorithms and quality detection scripts), specifically including model classification module 2011, model version control module 2012 and metadata management module 2013.
[0037] Specifically, the model classification module 2011 classifies data governance models by automatically classifying them using a decision tree algorithm based on metadata (such as data type, source system, and creation time). For example, a model whose source is a Customer Relationship Management System (CRM) and whose data type is "structured" is labeled as a "Customer Master Data Governance Model".
[0038] Specifically, the Model Version Control Module 2012 performs version control on the data governance model as follows: it adopts a Git-like version control mechanism, supporting model version rollback and branch management. Each model update automatically generates a version number (e.g., v1.2.3) and records a change log.
[0039] Specifically, the metadata management module 2013 manages the metadata of the data governance model by using Elasticsearch to store model metadata (such as author, creation time, and associated data assets), supporting full-text search and correlation analysis.
[0040] This invention employs a Git-like version control mechanism combined with Elasticsearch metadata storage to achieve traceability and rapid retrieval of model changes. Model reusability is improved by 50%; for example, government platforms can quickly reuse classification models from other departments, reducing redundant development costs.
[0041] The dynamic access control engine module 202 is mainly used to dynamically allocate access permissions based on user roles, data sensitivity, and user behavior. It mainly includes the hybrid control model 2021 and the real-time decision engine module 2022.
[0042] Specifically, the Hybrid Control Model 2021 combines RBAC (Role-Based Access Control) and ABAC (Attribute-Based Access Control). For example, the RBAC component defines basic roles such as "Administrator," "Analyst," and "Auditor," assigning default access permissions; for instance, administrators have full permissions. The ABAC component dynamically adjusts access permissions based on data sensitivity (e.g., public / internal / confidential) and user behavior (e.g., access time, device IP). For example, access to confidential data outside of working hours requires secondary authentication. This invention, through a dynamic permission engine based on the Hybrid Control Model 2021, combines RBAC and ABAC models. Compared to traditional static RBAC, it reduces the risk of sensitive data leakage by 70% and meets the real-time auditing requirements of regulations such as GDPR.
[0043] Specifically, the real-time decision engine module 2022 adopts a Redis cache permission strategy and evaluates access permission requests in real time through a rule engine (such as Drools), with a response time of less than 100ms.
[0044] The automated deployment service module 203 is primarily used for integration with data governance processes (such as ETL and data quality inspection) to achieve automatic deployment after model updates. The automated deployment service module 203 mainly includes an API interface 2031 and an automated deployment pipeline 2032.
[0045] Specifically, API interface 2031 can adopt a RESTful API, mainly used for interfacing with external systems. For example: POST / api / deploy interface: receives model ID and deployment environment (test / production) to trigger the deployment process; GET / api / status / {modelId} interface: queries the deployment status.
[0046] Specifically, the Automated Deployment Pipeline 2032 integrates Jenkins or GitLab CI to automate the entire process of model verification (such as unit testing and compatibility checks), environment configuration (such as database connection), and deployment.
[0047] This invention achieves seamless integration with CI / CD tools such as Jenkins and GitLab CI by defining the API interface 2031 specification. Model deployment time is reduced from 2 hours of manual operation to 5 minutes, and advanced features such as canary releases and rollbacks are supported.
[0048] The multimodal data processing adapter module 204 is mainly used to adapt the governance model for structured (such as database tables) data and unstructured (such as PDF contracts and images) data. The multimodal data processing adapter module 204 mainly includes a structured data processing module 2041 and an unstructured data processing module 2042.
[0049] Specifically, the structured data processing module 2041 connects to the relational database via JDBC / ODBC and automatically generates SQL scripts to perform data cleaning and classification.
[0050] Specifically, the unstructured data processing module 2042 integrates OCR (such as Tesseract) and NLP models (such as BERT) to extract key information from text and generate structured metadata. For example, it extracts the "contract amount" and "contracting parties" fields from a PDF contract and associates them with the "customer master data governance model".
[0051] In this invention, the multimodal data processing adapter module 204 integrates OCR, NLP and database connectors to support the adaptation of a unified governance model for structured and unstructured data, solving the problem that traditional data governance systems only support structured data. For example, the medical industry can use the system to automatically extract key indicators from medical record texts to support clinical decision-making.
[0052] The data persistence layer 3 mainly includes the model storage module 301 and the permission policy storage module 302.
[0053] Among them, the model storage module 301 mainly uses MongoDB to store model files and metadata, and supports the storage and fast retrieval of unstructured models (such as Python cleaning scripts).
[0054] Among them, the permission policy storage module 302 mainly uses MySQL to store the mapping relationship between user roles and data sensitivity, as well as audit logs, to ensure ACID transaction consistency.
[0055] The external integration layer 4 mainly includes the data governance tool integration module 401 and the CI / CD pipeline integration module 402.
[0056] Among them, the data governance tool integration module 401 connects with data governance platforms such as Collibra and Alation via API to achieve unified management of model assets.
[0057] Among them, the CI / CD pipeline integration module 402 integrates with Jenkins and GitLab CI, and supports the automated triggering of the model automated deployment pipeline 2032.
[0058] This invention, by constructing a model-permission-deployment linkage architecture, integrates model categorization and storage, dynamic permission control, and automated deployment into a unified architecture for the first time, solving the problems of isolated models, fixed permissions, and inefficient deployment in traditional systems. Through the association between model metadata and permission policies, this invention automates the entire process of "model update - permission adjustment - deployment take-off," improving deployment efficiency by 90%.
[0059] Secondly, this invention provides a method for managing Yishu models based on dynamic access control and automated deployment. This method is implemented using the Yishu model management system based on dynamic access control and automated deployment provided in the first aspect.
[0060] like Figure 2 As shown, the present invention provides a method for managing the Yishu model based on dynamic access control and automated deployment. The specific implementation process is as follows: Step S1: Model upload; Data engineers upload customer binning models through the front-end interaction layer 1. The model management panel 101 filters data governance models according to dimensions such as data type (structured / unstructured), version number, and creation time. The model classification module 2011 automatically classifies the data governance models using a decision tree algorithm and generates version numbers, such as v1.0.0. Finally, the model classification tree diagram can also be displayed through the model management panel 101. At the same time, the model version control module 2012 adopts a Git-like version control mechanism to support model version rollback and branch management, and the metadata management module 2013 uses Elasticsearch to store model metadata, supporting full-text search and correlation analysis.
[0061] Step S2: Permission configuration; Administrators can visualize the mapping relationship between user roles and data sensitivity through the permission configuration workbench 102; they can define basic roles and assign default access permissions through the hybrid control model 2021, and dynamically adjust access permissions based on data sensitivity and user behavior.
[0062] Step S3: Deployment triggered; After the model passes unit testing, the automated deployment service module 203 calls the API interface 2031 to push the model to the production environment data governance cluster through the automated deployment pipeline 2032. After automated deployment is completed, the model deployment status can be displayed in real time through the monitoring dashboard 103, and reports on indicators such as model deployment success rate and model deployment time can be generated.
[0063] Step S4: Real-time governance; After customer data enters the system, the model automatically performs binning operations. The real-time decision engine module 2022 monitors access behavior in real time and uses Redis caching permission policies to evaluate access permission requests in real time through the rule engine, blocking large-scale data export requests at unusual times.
[0064] The following detailed description of the Yishu model management method based on dynamic permission control and automated deployment provided by the present invention, with reference to specific embodiments, provides a detailed explanation.
[0065] Taking the customer data governance scenario in the financial industry as an example, the system operation process is as follows: (1) Model upload: Data engineers upload customer binning models (Python scripts) through the front-end interface. The system automatically classifies them as "structured data governance models" and generates version number v1.0.0.
[0066] (2) Permission configuration: The administrator configures the “Risk Control Analyst” role through the permission configuration workbench 102 to only access the “Public Data” sensitivity mapping model. Access at night (22:00-6:00) requires secondary authentication.
[0067] (3) Deployment trigger: After the model passes the unit test, the automated deployment service module 203 calls the POST / api / deploy interface to push the model to the production environment data governance cluster.
[0068] (4) Real-time governance: After customer data enters the system, the model automatically performs binning operations. The real-time decision engine module 2022 monitors access behavior in real time and blocks large-scale data export requests at unusual times.
[0069] This invention can be applied to financial risk control, medical data governance, and e-government digitalization. In the financial risk control industry, by updating anti-fraud models in real time, the risk identification response time can be reduced from hours to minutes. In medical data governance, dynamic anonymization technology achieves a balance between patient privacy protection and data sharing, supporting cross-institutional research collaboration. In the field of e-government digitalization, by constructing a unified data asset catalog—for example, a smart city project improved policy-making efficiency by 40% through the system integration of data models from over 200 departments. Through a model-permission-deployment linkage design, this invention promotes the transformation of data governance from a "cost center" to a "value creation center," providing a technological foundation for innovative practices such as enterprise data asset entry and cross-border data flow.
[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. However, these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A data model management system based on dynamic access control and automated deployment, characterized in that, include: The front-end interaction layer provides the user interface for interacting with the system. The core service layer is used to implement model classification, version control and metadata management, as well as to dynamically allocate access permissions, automatically deploy models, and adapt data governance models. The data persistence layer is used to store model files and metadata, as well as to store and quickly retrieve unstructured models. The external integration layer is used to interface with the data governance platform to achieve unified management of model assets and to automatically trigger model deployment.
2. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 1, characterized in that, The front-end interaction layer includes a model management panel, a permission configuration workbench, and a monitoring dashboard. The model management panel is used to display a model classification tree diagram and to filter data governance models according to different dimensions. The permission configuration workbench is used to visually configure the mapping relationship between user roles and data sensitivity. The monitoring dashboard is used to display the model deployment status in real time and generate indicator reports.
3. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 1, characterized in that, The core service layer includes a model classification and storage engine module, a dynamic access control engine module, an automated deployment service module, and a multimodal data processing adapter module. The model classification and storage engine module is used to classify, version control, and manage metadata of data governance models. The dynamic access control engine module is used to dynamically allocate access permissions based on user roles, data sensitivity, and user behavior. The automated deployment service module is used to integrate with the data governance process to achieve automatic deployment after model updates. The multimodal data processing adapter module is used to adapt governance models for structured and unstructured data.
4. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 3, characterized in that, The model classification and storage engine module includes a model classification module, a model version control module, and a metadata management module. The model classification module automatically classifies models based on metadata using a decision tree algorithm. The model version control module uses a Git-like version control mechanism to implement model version rollback and branch management. The metadata management module uses Elasticsearch to store model metadata and supports full-text search and correlation analysis.
5. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 3, characterized in that, The dynamic access control engine module includes a hybrid control model and a real-time decision engine module; the hybrid control model combines RBAC and ABAC, with RBAC defining basic roles and assigning default access permissions. ABAC dynamically adjusts access permissions based on data sensitivity and user behavior; the real-time decision engine module adopts a Redis caching permission strategy and evaluates access permission requests in real time through a rule engine.
6. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 3, characterized in that, The automated deployment service module includes an API interface and an automated deployment pipeline; the API interface adopts a RESTful API for interfacing with external systems; the automated deployment pipeline integrates Jenkins or GitLab CI to automate the entire process of model verification, environment configuration, and deployment.
7. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 3, characterized in that, The multimodal data processing adapter module includes a structured data processing module and an unstructured data processing module. The structured data processing module connects to a relational database via JDBC / ODBC and automatically generates SQL scripts to perform data cleaning and classification. The unstructured data processing module integrates OCR and NLP models to extract key information from text and generate structured metadata.
8. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 1, characterized in that, The data persistence layer includes a model storage module and a permission policy storage module. The model storage module uses MongoDB to store model files and metadata, supporting the storage and fast retrieval of unstructured models. The permission policy storage module uses MySQL to store the mapping relationship between user roles and data sensitivity, as well as audit logs, to ensure ACID transaction consistency.
9. The Easy Number Model Management System based on dynamic access control and automated deployment according to claim 1, characterized in that, The external integration layer includes a data governance tool integration module and a CI / CD pipeline integration module. The data governance tool integration module interfaces with the data governance platform via API to achieve unified management of model assets. The CI / CD pipeline integration module integrates with Jenkins and GitLab CI to support automated triggering of the model automated deployment pipeline.
10. A method for managing Yi-number models based on dynamic access control and automated deployment, implemented using the Yi-number model management system based on dynamic access control and automated deployment as described in any one of claims 1-9, characterized in that, Includes the following steps: Step S1: Model upload; Data engineers upload customer binning models through the front-end interaction layer, the model management panel filters data governance models, and the model classification module automatically classifies the data governance models using a decision tree algorithm and generates version numbers. Step S2: Permission configuration; Administrators configure the mapping relationship between user roles and data sensitivity through the permission configuration workbench; they define basic roles and assign default access permissions through a hybrid control model, and dynamically adjust access permissions based on data sensitivity and user behavior. Step S3: Deployment triggered; After the model passes unit testing, the automated deployment service module calls the API interface to push the model to the production environment data governance cluster through the automated deployment pipeline; Step S4: Real-time governance; After customer data enters the system, the model automatically performs binning operations. The real-time decision engine module monitors access behavior in real time and uses Redis caching permission policies to evaluate access permission requests in real time through the rule engine, blocking large-scale data export requests at unusual times.