A data governance method and system based on a workflow engine

Abstract

The present application relates to the technical field of data processing, in particular to a data management method and system based on a workflow engine. The method embeds data management logic in each link of a business process, builds a ternary fusion management model relying on the driving capability of the workflow engine, realizes the synchronization of data management and business execution through real-time linkage of data flow collection and management rule triggering, and realizes the automation and precision of management actions through intelligent driving algorithm model and rule engine. The present application aims to realize the synchronous design, automatic execution and dynamic optimization of data management through the deep fusion of process and management, significantly improve the precision, timeliness and intelligent level of data management, provide a systematic, embedded and sustainable data management solution for enterprises in complex business environment, and ensure that data assets continuously empower business innovation and decision-making on the basis of safety, compliance and high quality.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically a data governance method and system based on a workflow engine. Background Technology

[0002] With the deepening of digital transformation, data has become a core strategic asset for organizations, and its quality, security, and compliance directly affect operational efficiency, risk management, and the effectiveness of intelligent decision-making. Enterprises and institutions have accumulated massive amounts of data in their business systems and are attempting to improve its usability and credibility through data governance to drive key scenarios such as precision marketing, intelligent risk control, and process optimization. However, traditional governance models face severe challenges: governance activities are often independent of business processes, creating a disconnect between the two, leading to rigid governance rules, delayed execution, and vague responsibilities, making it difficult to adapt to rapidly changing business needs. On the one hand, data quality issues and unclear responsibilities are often only passively traced after an incident occurs, incurring significant costs for root cause analysis; on the other hand, the lack of integrated control over data status, version conflicts, and lifecycle in cross-departmental and cross-system data flow and collaborative operations can easily lead to data inconsistencies, compliance risks, and asset loss.

[0003] Existing data governance methods largely rely on periodic batch audits, static policy documents, or isolated management tools, which have significant limitations: First, accountability relies on manual mapping, making it difficult to automatically clarify the specific roles and stages of data creation, modification, and use from the dynamic business execution chain; second, state awareness is weak, failing to automatically trigger corresponding quality checks, compliance reviews, or retention policies based on the actual progress stage of business processes, such as drafting, approval, and archiving; third, collaborative conflict control is lacking, failing to effectively warn and resolve data read / write conflicts and version discrepancies caused by concurrent operations in parallel or converged business tasks. These limitations make data governance costly, slow to respond, and difficult to measure and continuously optimize.

[0004] Therefore, there is an urgent need for a new paradigm of data governance that can be deeply integrated with business processes, linked in real time, and driven by intelligence. This method requires that governance logic be placed in advance and encoded within the business process model, and that the inherent execution capabilities of the process engine be used to achieve automated, refined, and closed-loop management of the entire data lifecycle. This fundamentally solves the core problem of the disconnect between governance and business, ensuring that data assets continue to create value on a secure, compliant, and high-quality track. Summary of the Invention

[0005] The purpose of this invention is to provide a data governance method based on a workflow engine to solve the core problems that are common in the current data governance process, such as the disconnect between governance activities and business processes, difficulty in tracing responsibility, lagging status awareness, and insufficient control of collaborative conflicts. It improves the shortcomings of traditional methods that rely on ex-post audits and static strategies, achieves comprehensive support for dynamic and complex business environments, and solves the industry pain points of lagging governance execution and difficulty in assigning responsibility.

[0006] The present invention solves its technical problem by adopting the following technical solution: A data governance method based on a workflow engine includes the following steps: S1. Dynamic data acquisition with predefined tags By using bypass monitoring technology, dynamic data in the business process is collected in real time in an event-driven manner. The dynamic data monitored includes five dimensions: user operation behavior, data change records, node attribute information, process status transitions, and task execution logs. Real-time data acquisition is achieved through non-intrusive monitoring. Dynamic data is tagged using a labeled data model, with the tags layered into a basic attribute layer, a business semantic layer, and a governance rule layer. The basic attribute layer is divided into data ID, operation time, and operation terminal; the business semantic layer is divided into process node, business status, and participating role; and the governance rule layer is divided into verification rules, compliance requirements, and responsible entity. S2. Construction of the Responsibility Origin Model Tag-based data is dynamically expanded and integrated through a fusion governance model that includes personnel, status, and tasks. The fusion governance model consists of a participant association graph, a business status graph, and a node task collision graph, forming an integrated tag data network. S3. Accountability and Conflict Detection A dynamic data-related personnel matrix is ​​constructed through an integrated tag data network and participant association graph to achieve automatic tracking of data lineage and accurate tracing of responsible parties; a multi-dimensional data collision analysis model is run through a node task collision graph to detect data conflict risks in real time and trigger control actions. S4. Intelligent governance cycle triggering Automatically trigger data quality verification, compliance checks, and tiered archiving governance by calling the tagged data model through the business status diagram; S5. Human-machine collaborative compensation processing For boundary scenarios that S4 cannot automatically manage, a compensation interaction mechanism is initiated through the governance rule layer to generate exception handling tasks and assign them to the responsible entities, thereby achieving closed-loop feedback of the processing results. S6. Visualized Closed-Loop Presentation By constructing a data link relationship map by integrating the governance records and processing results in S4 and S5 with the integrated tag data network, the linkage relationship and governance effect of the data are presented, and governance is continuously optimized based on quantitative evaluation indicators.

[0007] A data governance system based on a workflow engine, used to implement the aforementioned data governance method based on a workflow engine, includes: The data acquisition module is used to collect dynamic data in the business process in real time through bypass monitoring technology, and to perform tagging processing based on a predefined tagged data model; The intelligent governance engine module, connected to the data acquisition module, is used to construct a three-element integrated governance model and perform governance actions such as responsibility tracing, conflict detection, quality verification, compliance inspection, and hierarchical archiving. The human-machine collaborative processing module is connected to the intelligent governance engine module and is used to provide a standardized compensation and interaction mechanism for the boundary scenarios of automated governance. The data visualization module, connected to the intelligent governance engine module and the human-machine collaborative processing module, is used to construct a data relationship graph and present the governance effect.

[0008] The intelligent governance engine module includes a model management submodule, a rule engine submodule, a data processing submodule, and a tag update submodule; the data visualization module includes a data aggregation submodule, a graph construction submodule, a visualization rendering submodule, a quantitative evaluation submodule, and an optimization analysis submodule.

[0009] This method deeply embeds data governance logic into each stage of the business process, constructs a three-element integrated governance model driven by a workflow engine, synchronizes data governance with business execution through real-time linked data flow collection and governance rule triggering, and automates and refines governance actions through intelligently driven algorithm models and rule engines. Specifically, this invention aims to achieve the following objectives: 1. Achieve traceability of responsibility and lineage: Based on the many-to-many mapping rules of the roles and tasks of process executors, automatically construct a data-related person matrix, clarify the responsible parties for data creation, modification and use in each link, and strengthen data ownership management and traceability capabilities; 2. Supports conflict risk warning and data consistency management: Through a self-developed multi-dimensional data collision analysis model, data conflict detection rules are established to achieve version control and consistency maintenance under concurrent operations, thereby reducing the risk of data conflicts; 3. Promote lifecycle-driven automated governance: Based on the status changes of business processes, automatically trigger corresponding governance actions such as data quality verification, compliance review, and archiving and retention through a standardized verification and compliance system to achieve synchronization between data governance strategies and business stages; 4. Construct a closed-loop governance system for human-machine collaboration: Design a standardized compensation and interaction mechanism to solve boundary scenario problems in automated governance and ensure the integrity and accuracy of the governance process; 5. Achieve visualization and continuous optimization of governance effectiveness: Concretize the data linkage relationships and flow paths, establish a quantitative evaluation and rule optimization mechanism for governance effectiveness, and form a sustainable governance system.

[0010] Ultimately, this invention aims to achieve synchronous design, automatic execution, and dynamic optimization of data governance through the deep integration of processes and governance. This significantly improves the accuracy, timeliness, and intelligence of data governance, providing enterprises with a systematic, embedded, and sustainable data governance solution in complex business environments. It ensures that data assets continuously empower business innovation and decision-making on the basis of security, compliance, and high quality. Attached Figure Description

[0011] To more clearly illustrate the embodiments of the present invention or the existing technical solutions, the drawings used in the description of the embodiments or the prior art are briefly introduced below. The following drawings are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on the drawings.

[0012] Figure 1 is a schematic diagram of the governance model construction according to an embodiment of the present invention; Figure 2 is a flowchart of the data governance method according to an embodiment of the present invention; Figure 3 is a data link relationship diagram of an embodiment of the present invention; Figure 4 is a system block diagram of an embodiment of the present invention. Detailed Implementation

[0013] A data governance method based on a workflow engine includes the following steps: S1. Dynamic data acquisition with predefined tags By using bypass monitoring technology, dynamic data in the business process is collected in real time in an event-driven manner. The dynamic data monitored includes five dimensions: user operation behavior, data change records, node attribute information, process state transitions, and task execution logs. The real-time acquisition of data is achieved through non-intrusive monitoring without affecting the normal execution of the business process. Dynamic data is tagged using a labeled data model, with the tags layered into a basic attribute layer, a business semantic layer, and a governance rule layer. The basic attribute layer is divided into data ID, operation time, and operation terminal; the business semantic layer is divided into process node, business status, and participating role; and the governance rule layer is divided into verification rules, compliance requirements, and responsible entity. Tag adaptation: Tag adaptation rules are designed for different business process types (such as approval process, business processing process, data flow process) to achieve accurate matching between tags and business scenarios; Structured storage: After the collected dynamic data is tagged and assigned values ​​according to the predefined model, it is integrated into structured raw tag data in the form of key-value pair combination relationship, providing a real-time, complete and computable data foundation for subsequent governance.

[0014] S2. Construction of the Responsibility Origin Model Tag-based data is dynamically expanded and integrated through a fusion governance model that includes personnel, status, and tasks. The fusion governance model consists of a participant association graph, a business status graph, and a node task collision graph, forming an integrated tag data network. S3. Accountability and Conflict Detection A dynamic data-related personnel matrix is ​​constructed through an integrated tag data network and participant association graph to achieve automatic tracking of data lineage and accurate tracing of responsible parties; a multi-dimensional data collision analysis model is run through a node task collision graph to detect data conflict risks in real time and trigger control actions. S4. Intelligent governance cycle triggering Automatically trigger data quality verification, compliance checks, and tiered archiving governance by calling the tagged data model through the business status diagram; S5. Human-machine collaborative compensation processing For boundary scenarios that S4 cannot automatically manage, a compensation interaction mechanism is initiated through the governance rule layer to generate exception handling tasks and assign them to the responsible entities, thereby achieving closed-loop feedback of the processing results. S6. Visualized Closed-Loop Presentation By constructing a data link relationship map by integrating the governance records and processing results in S4 and S5 with the integrated tag data network, the linkage relationship and governance effect of the data are presented, and governance is continuously optimized based on quantitative evaluation indicators.

[0015] In step S1, the tagged data model can be predefined, wherein the governance rule layer is bound to at least one of the following: data quality verification rules, compliance requirement rules, responsible entity rules, and conflict detection rules.

[0016] The detection rules of the multi-dimensional data collision analysis model in step S3 include at least one of the following: multiple operation collision rules for the same data topic, multiple version data collision rules for the same task, and parallel task data read / write collision rules; the control actions include at least one of the following: generating version merging, locking conflicting data for read / write, and pausing the execution of related task nodes.

[0017] The data quality verification in step S4 includes integrity, accuracy, consistency, and timeliness verification. The data quality verification sets basic thresholds, early warning thresholds, and anomaly thresholds for each verification dimension of integrity, accuracy, consistency, and timeliness. The hierarchical archiving governance is triggered based on the hierarchical nature of the business type.

[0018] The compensation interaction mechanism in step S5 includes: abnormal scene identification, task assignment, multi-terminal notification, standardized processing flow, secondary verification of processing results, and timeout handling mechanism.

[0019] The nodes of the data link relationship graph in step S6 include participants, data entities, business tasks, and process status, and the edges include lineage relationships, trigger relationships, conflict relationships, and responsibility relationships; the quantitative evaluation indicators include at least the accuracy rate of responsibility tracing, conflict detection rate, quality verification pass rate, compliance inspection pass rate, and archiving completion rate.

[0020] A data governance system based on a workflow engine, used to implement data governance methods, includes: The data acquisition module is used to collect dynamic data in the business process in real time through bypass monitoring technology, and to perform tagging processing based on a predefined tagged data model; The intelligent governance engine module, connected to the data acquisition module, is used to construct a three-element integrated governance model and perform governance actions such as responsibility tracing, conflict detection, quality verification, compliance inspection, and hierarchical archiving. The human-machine collaborative processing module is connected to the intelligent governance engine module and is used to provide a standardized compensation and interaction mechanism for the boundary scenarios of automated governance. The data visualization module, connected to the intelligent governance engine module and the human-machine collaborative processing module, is used to construct a data relationship graph and present the governance effect.

[0021] The intelligent governance engine module includes a model management submodule, a rule engine submodule, a data processing submodule, and a tag update submodule; the data visualization module includes a data aggregation submodule, a graph construction submodule, a visualization rendering submodule, a quantitative evaluation submodule, and an optimization analysis submodule.

[0022] The technical solution of the present invention is further illustrated with reference to the embodiments: To deeply explore the value of dynamic data and facilitate its integration with business systems, this invention deeply embeds data governance logic into each stage of the business process, constructing a governance model that integrates "people, status, and tasks." Through a real-time linkage collection and triggering mechanism, it achieves automatic accountability, real-time conflict detection, and intelligent lifecycle management. Through an intelligently driven model and engine, it forms a closed-loop governance system covering responsibility, quality, and conflict, thoroughly solving the "two-tiered" problem from a technical perspective and significantly improving the accuracy, timeliness, and intelligence of data governance.

[0023] This involves deeply integrating bypass monitoring technology with predefined tagged data models to achieve accurate dynamic data collection, constructing a governance model that integrates "human-state-task" triad, developing a self-developed multi-dimensional data collision analysis model for conflict detection, establishing a standardized verification, compliance, and archiving system, designing a full-process human-machine collaborative compensation interaction mechanism, and visualizing and quantitatively optimizing concrete data linkage relationships. Specifically, the steps include: 1. Data Acquisition: Dynamic data acquisition using a bypass monitoring system combined with a predefined tagged data model. By employing bypass monitoring technology (a conventional technology, source: "Computer Network Traffic Monitoring Technology and Application", Tsinghua University Press, 2018), dynamic data in the business process is collected in real time in an event-driven manner. The monitoring scope accurately covers five core dimensions: user operation behavior, data change records, node attribute information, process status transitions, and task execution logs. Data acquisition is achieved in real time through non-intrusive monitoring without affecting the normal execution of the business process.

[0024] Based on the above five core dimensions, a predefined tagged data model (independent innovation, with innovations in multi-dimensional hierarchical design of tags and deep integration of business governance semantics) is defined. The predefined requirements for this model are as follows: Tag layering: Divided into a basic attribute layer, a business semantic layer, and a governance rule layer. The basic attribute layer records basic data information (such as data ID, operation time, and operation terminal). The business semantic layer associates core information of the business process (such as process nodes, business status, and participating roles). The governance rule layer binds corresponding governance strategies (such as verification rules, compliance requirements, and responsible entities). Tag adaptation: Tag adaptation rules are designed for different business process types (such as approval processes, business processing processes, and data flow processes) to achieve accurate matching between tags and business scenarios. Structured storage: After the collected dynamic data is tagged and assigned values ​​according to a predefined model, it is integrated into structured raw tag data in the form of key-value pair combinations, providing a real-time, complete, and computable data foundation for subsequent governance.

[0025] 2. Construction of a responsibility attribution model: Construction of a three-element integrated governance model of people, state, and task, and responsibility attribution. The collected tag data is injected into the ternary fusion governance model (participant relationship graph, business status graph, node task collision graph) for dynamic expansion and ternary fusion. The fusion relationship among the three is a many-to-many network relationship. The specific fusion rules and tag data generation logic are as follows: 1) Participant Relationship Graph: Depicts the many-to-many mapping relationship between process steps and execution roles (one role can participate in multiple process steps, and one process step can be executed by multiple roles), assigns responsibility governance tags to roles (such as data creator, data modifier, data reviewer, and data user), and generates role-responsibility-step tag data; 2) Business Status Diagram: Describes the status transition path of business objects in the process (e.g., drafting → submitting → reviewing → approving → archiving), attaches lifecycle governance tags to the status (e.g., pending verification, pending compliance, pending archiving, completed), and generates status-lifecycle-stage tag data; 3) Node-Task Collision Graph: Characterizes parallel or converged task nodes and the data topics they operate on, attaches version and conflict management tags (such as data version number, operation type, data topic ID) to task nodes and data topics, and generates task-data-version-conflict tag data; 4) Mesh Fusion: Using process link ID and data topic ID as the core association keys, the label data of the three diagrams are fused into a mesh to form an integrated label data network of "role-link-status-task-data", so that data in any dimension can be traced to all other related dimensions.

[0026] Based on the integrated data network, a data-related person matrix is ​​constructed according to the role-task many-to-many mapping relationship. The row dimension of the matrix is ​​the data topic ID, and the column dimension is the governance responsibility type (creation, modification, review, use, archiving). The matrix value is the role / personnel ID of the corresponding responsible entity + operation time + operation step, forming a responsibility chain mapping relationship covering the entire chain of data creation, flow, and use. Through this matrix, the responsible entity can be accurately traced based on the data ID, and the entire chain of data lineage can be traced based on the role / personnel ID, realizing automatic data lineage tracking and accurate tracing of the responsible entity.

[0027] 3. Data Linkage and Conflict Risk Detection: Conflict detection and control using a self-developed multi-dimensional data collision analysis model. The multi-dimensional data collision analysis model used in this step is independently developed. Its innovations include: multi-dimensional collision rule design based on label data, time-slice polling detection mechanism for parallel tasks, and multi-dimensional verification logic for version consistency. The core design and implementation details of the model are as follows: 1) Model input: Label data of the node task collision graph (task ID, data topic ID, operation type, data version number, operation time, execution role), and the process topology relationship of parallel / converging tasks; 2) Collision Detection Rules: Define three core rules: collisions involving multiple operations on the same data topic, collisions involving multiple versions of the same task, and collisions involving parallel task data read / write operations. Specifically: Collisions caused by multiple operations on the same data topic: Collisions are triggered when multiple roles perform mutually exclusive operations (such as simultaneous modification or simultaneous deletion) on the same data topic in the same process stage; Collisions caused by multiple versions of the same data topic in the same task: Collisions are triggered when multiple unmerged versions of the same data topic are called by the same task node; Collisions caused by parallel task data read / write: Collisions are triggered when multiple task nodes executing in parallel perform read-multiple write or multiple write-multiple read operations on the same data topic. 3) Detection mechanism: A time-slice polling detection mechanism is adopted, with 100ms as the minimum detection time slice. Data read and write operations in parallel or converged tasks are scanned in real time, and operations that meet the collision detection rules are marked in real time. 4) Risk Warning and Control: Upon detecting a collision, immediately generate conflict risk warning information (including conflict data ID, conflict type, conflict task, executing role, and operation time), and trigger corresponding control actions based on the conflict level: Low-level conflicts: Automatically generate version merge suggestions and push them to relevant execution roles; Medium-level conflicts: Lock conflicting data for reading and writing, prohibiting subsequent operations until the conflict is resolved; High-level conflicts: Suspend the execution of relevant task nodes and trigger a human-machine collaborative processing mechanism.

[0028] This model enables real-time monitoring and risk management of data linkage relationships, ensuring version consistency of data in parallel / convergence tasks.

[0029] 4. Intelligent governance cycle triggering: Full lifecycle governance of standardized verification, compliance, and archiving systems. Based on changes in business process status, this invention establishes a standardized data quality verification, compliance check, and archiving system by intelligently triggering corresponding data governance actions through a rules engine (a conventional technology, source: Drools Rules Engine official technical documentation, version 7.0 and above). The specific standards and execution rules are as follows: 1) Data quality verification standards Quantitative verification standards are established based on four dimensions: data integrity, data accuracy, data consistency, and data timeliness. Each dimension has a basic threshold, a warning threshold, and an anomaly threshold, specifically: Data Integrity: Non-empty field percentage ≥98% (basic), ≥95% (warning), <95% (anomaly); Data Accuracy: Field values ​​conforming to preset rules ≥99% (basic), ≥97% (warning), <97% (anomaly); Data Consistency: Field values ​​for the same data topic consistent across nodes ≥99% (basic), ≥98% (warning), <98% (anomaly); Data Timeliness: Time from data operation completion to receipt by the next stage ≤24h (basic), ≤48h (warning), >48h (anomaly).

[0030] Verification trigger rules: When the process status changes to key nodes such as submission or review, quality verification is automatically triggered. A quality report is generated based on the verification results, and abnormal data triggers early warnings or human-machine collaborative processing.

[0031] 2) Compliance Inspection Standards Compliance inspection standards are formulated based on industry regulatory requirements, internal corporate policies, and data security laws, and are divided into three main categories: data ownership compliance, data operation compliance, and data storage compliance. Data ownership compliance: Data operation roles match data ownership scope, with no unauthorized operations; Data operation compliance: Data creation, modification, and deletion operations comply with process permission requirements, and operation logs are complete; Data storage compliance: Sensitive data is encrypted and stored, while non-sensitive data is stored according to preset periods, with no unauthorized storage behavior. Inspection trigger rules: When the process status changes to the approval or archiving node, a compliance check is automatically triggered. Violations immediately generate compliance warnings and are traced back to the responsible party.

[0032] 3) Definition of archiving strategy and archiving effect Develop a tiered archiving strategy based on business type and data importance. Archiving effectiveness is quantified by archiving integrity, archiving traceability, and archiving storage efficiency, specifically: Data classification: Data is divided into three categories based on importance: core data, important data, and general data; Archiving strategy: Core data: Permanently archived, using multiple copies and off-site storage, with a storage format of structured + original files; Important data: Archived every ≥10 years, using dual copies and a storage format of structured; General data: Archived every ≥3 years, using single copies and a storage format of structured. Archive triggering rules: When the process status changes to the archive node, the corresponding archive action is automatically triggered according to the data level; Archiving effectiveness quantification standards: Archiving completeness: Archiving rate of data to be archived = 100%; Archiving traceability: Archived data can be traced across the entire chain through data ID, process ID, and responsible entity ID; Archiving storage efficiency: Storage and retrieval response time of archived data ≤ 3s.

[0033] Through the aforementioned standardization system, governance strategies and business progress can be synchronized and dynamically optimized.

[0034] 5. Supplementary data confirmation: Human-computer collaborative processing of standardized compensation interaction mechanism For boundary scenarios in automated governance where deterministic handling is impossible (such as high-level data conflicts, severe quality verification anomalies, compliance check violations, and special business scenarios not covered by rules), this invention designs a standardized compensation and interaction mechanism, intelligent assignment of pending tasks, standardization of processing flows, and closed-loop feedback of processing results. Specific implementation details are as follows: 1) Abnormal Scene Recognition: The system automatically identifies boundary scenes that cannot be handled by automated management, generates abnormal handling tasks, and the tasks include abnormal data information, abnormal type, abnormal cause, and preliminary handling suggestions. 2) Intelligent task assignment: Based on the data-related person matrix, tasks to be done are accurately assigned to the corresponding responsible parties (such as data creators, reviewers, and process owners), with the assignment rule being "whoever is responsible, handles it"; 3) Multi-terminal notification mechanism: Send to-do task notifications to the responsible parties through three methods: system pop-ups, WeChat / DingTalk messages, and emails to ensure timely notification; 4) Standardized processing flow: The responsible party operates through the system's human-computer collaborative processing interface. The standardized processing flow is as follows: view abnormal information → analyze the cause of the abnormality → perform processing operations (supplement / correction / confirmation) → submit the processing result. The system records the entire process processing log. ① Processing result verification and feedback: After the responsible party submits the processing result, the system automatically performs a second verification on the processing result. If the verification passes, the result is fed back to the multi-dimensional governance model to update the tag data and governance status; if the verification fails, the task is returned to the responsible party with the reason for the return. ② Timeout handling mechanism: Set the timeout period for handling pending tasks (default 24h). If the task is not handled within the timeout period, it will be automatically escalated to the superior person in charge to ensure the closed loop of the governance process.

[0035] This mechanism standardizes and normalizes the manual confirmation, supplementation, or correction processes, ensuring the integrity and accuracy of the governance closed loop.

[0036] 6. Visualized Closed-Loop Presentation: Concrete representation and quantitative optimization of data linkage relationships By integrating the chain of responsibility, conflict detection results, and governance trigger records, a data relationship graph is constructed (an independent innovation, with innovations including multi-dimensional visualization mapping, interactive graph operations, and quantitative evaluation and optimization of governance effects). This graph concretely presents the data linkages, flow paths, and governance mechanisms. Specifically, the implementation is as follows: 1) Visualizing data linkage relationships: Nodes represent participants, data entities, business tasks, and process states, while edges represent the linkage relationships between nodes, specifically: Lineage: The creation, modification, and usage relationships between data entities and participants / business tasks / process states; Triggering relationships: The triggering relationships between process states and governance actions (e.g., archive state triggers archive governance actions); Conflict relationships: Collision and conflict relationships between business tasks and data entities; Responsibility relationships: The attribution of rights and responsibilities between participants and data entities; 2) Visualization of the workflow path: Presenting the entire workflow path of data from creation to archiving in the form of timeline + process topology, marking the operation role, operation time, status change and governance actions of each link; 3) Concretization of governance mechanisms: The governance actions are presented in the form of trigger nodes and governance effect indicators to show the improvement effect of governance actions on data quality, compliance and consistency, and the execution time, execution results and responsible parties of the governance actions are marked. 4) Visualization and Interaction: Visualization is achieved through a 3D governance dashboard, supporting interactive operations such as node expansion / collapse, relationship path tracing, dimension filtering (responsibility / quality / conflict), timeline sliding, and drill-down analysis; 5) Quantitative evaluation and optimization of governance effectiveness: Establish a quantitative evaluation index system for governance effectiveness (accuracy rate of accountability traceability, conflict detection rate, quality verification pass rate, compliance inspection pass rate, and archiving completion rate). The system automatically calculates each index, identifies governance bottlenecks and risk patterns through graph analysis, generates optimization suggestions, and feeds them back to governance rules and models to achieve continuous optimization of governance strategies.

[0037] Through the above implementation, a closed-loop governance system covering the three dimensions of responsibility, quality, and conflict is formed, supporting the visualized monitoring and continuous optimization of governance effectiveness.

[0038] This system combines hardware and software to implement the aforementioned data governance methods. Its core innovations lie in the deep collaborative linkage mechanism among modules, the multi-model fusion design of the intelligent governance engine, and the quantitative optimization closed loop of the visualization module. The overall system architecture is a four-layer structure: data acquisition layer, intelligent governance layer, human-machine collaboration layer, and visualization layer. The modules are hierarchically collaborative. The data acquisition module serves as the underlying data source, outputting structured labeled data to the intelligent governance engine module. The intelligent governance engine module is the system core, outputting abnormal pending tasks to the human-machine collaboration processing module and providing full governance data to the data visualization module. The human-machine collaboration processing module provides feedback on manual processing results to the intelligent governance engine module. The data visualization module outputs governance optimization suggestions to the intelligent governance engine module, forming a fully closed-loop collaborative relationship of data-governance-processing-visualization-optimization.

[0039] The system mainly consists of the following four modules, and the specific design, technical details, and innovative points of each module are as follows: 1. Data Acquisition Module The system's underlying data source acquisition module employs a bypass monitoring technology combined with a predefined tagged data model to achieve accurate dynamic data collection. While bypass monitoring is a conventional technique, the predefined tagged data model is an independently developed innovation. It captures dynamic data from business processes in real time using an event-driven approach, focusing on five dimensions: user operations, data changes and node attributes, process state transitions, and task execution logs. Data is collected and adapted according to the predefined tagged data model and integrated into a structured raw tagged data stream, providing a real-time, complete, and computable data source for subsequent governance.

[0040] Monitoring component: Enables non-intrusive data collection based on network traffic monitoring technology, supporting adaptation to mainstream business systems (such as OA, ERP, CRM); Tagging component: Implements predefined tagged data models, completing data tag assignment and structured integration; Data storage components: A hybrid storage approach using Redis and MySQL is adopted, with Redis storing the raw data collected in real time and MySQL storing the structured tag data.

[0041] The multi-dimensional layered design of tags and the deep integration of business governance semantics enable precise matching between collected data and governance needs.

[0042] 2. Intelligent Governance Engine Module The core governance execution module of the system features innovative multi-model fusion design (responsibility tracing model + self-developed collision analysis model + standardized verification and compliance model), multi-dimensional rule engine linkage triggering, and dynamic updating of full-link data tags. Its core functions include executing core governance logic to achieve full-process governance actions such as data lineage tracing, responsibility tracing, conflict detection, quality verification, compliance checks, and archiving.

[0043] The model management submodule loads, runs, and updates the accountability traceability model, multi-dimensional data collision analysis model, and standardized verification and compliance model. The rules engine submodule is a secondary development based on the Drools rules engine (an independent innovation, the innovation being the dynamic binding of governance rules to process status), enabling flexible configuration and triggering of governance rules. The data processing submodule constructs a data-related person matrix, detects and controls conflict risks, executes quality verification and compliance checks, and implements a tiered archiving strategy. The tag update submodule updates the tag data of the three-element fusion governance model in real time based on the execution results of governance actions, ensuring the real-time nature and accuracy of the data.

[0044] The module collaboratively receives structured tagged data from the data acquisition module, outputs abnormal pending tasks to the human-machine collaborative processing module, and outputs full governance data to the data visualization module.

[0045] 3. Human-Machine Collaborative Processing Module The core innovations of the system's boundary scenario processing module lie in its standardized compensation interaction mechanism, intelligent task assignment, and closed-loop feedback of processing results. Its core function is to provide a compensation-based interaction mechanism for automated governance boundary scenarios, standardizing and normalizing manual confirmation, supplementation, or correction processes to ensure the integrity of the governance closed loop.

[0046] The module includes the following sub-modules: Anomaly Identification Sub-module: Automatically identifies boundary scenarios that the intelligent governance engine module cannot handle and generates standardized to-do tasks; Task Assignment Sub-module: Achieves intelligent and accurate assignment of to-do tasks based on a data-related person matrix; Multi-terminal Notification Sub-module: Enables multi-terminal notifications via system pop-ups, WeChat / DingTalk messages, and emails; Human-Computer Interaction Sub-module: Provides a standardized processing interface to standardize manual processing operations; Result Verification and Feedback Sub-module: Enables secondary verification and closed-loop feedback of manual processing results.

[0047] Receive abnormal pending tasks from the intelligent governance engine module and provide feedback on the manual processing results to the intelligent governance engine module.

[0048] 4. Data Visualization Module The system's governance outcome presentation and optimization module's core innovations lie in the concrete presentation of data relationship graphs, the quantitative evaluation index system for governance effectiveness, and the closed-loop drive of analysis and optimization. Its core functions include: comprehensively presenting governance outcomes, constructing data relationship graphs, and enabling visualized monitoring, quantitative evaluation, and continuous optimization of governance effectiveness.

[0049] The data aggregation submodule aggregates the full governance data from the intelligent governance engine module and the manual processing logs from the human-machine collaborative processing module to form a complete data source for graph construction. The graph construction submodule constructs a data link relationship graph, defining nodes and edges and performing multi-dimensional visualization mapping. The visualization rendering submodule renders interactive graphs based on ECharts and Three.js, supporting 3D visualization and multi-dimensional interactive operations. The quantitative evaluation submodule establishes a quantitative evaluation index system for governance effectiveness, automatically calculates various indicators, and generates a governance effectiveness evaluation report. The optimization analysis submodule identifies governance bottlenecks and risk patterns through graph analysis, formulates governance optimization suggestions, and feeds them back to the intelligent governance engine module.

[0050] It receives all data from the intelligent governance engine module and the human-machine collaborative processing module, outputs governance optimization suggestions to the intelligent governance engine module, and forms a closed loop for continuous optimization of governance strategies.

[0051] The workflow engine-based data governance system of this invention has a four-layer architecture (data acquisition module, intelligent governance engine module, human-machine collaborative processing module, and data visualization module, as shown in Figure 4). It is specifically customized for the implementation of the approval and document processing modules of an OA system. Each module consists of two parts: conventional technology reference / adaptation and independent innovative design. The conventional technologies are mature commercial / open-source solutions in the fields of OA system development and data governance, with clear technical origins, providing basic support for system implementation. The independent innovative design is a core customized design specifically for the workflow characteristics of the approval and document processing modules in OA systems. It is the core competitiveness for solving the pain points of data governance in OA systems. The following clearly distinguishes each module and explains the specific adaptation method of each module in the approval and document processing modules.

[0052] Data acquisition module ( Figure 4 S410) — The underlying data source of the dual-module OA system 1) Use / reuse of conventional technologies Bypass monitoring technology: Referencing the network traffic monitoring technology in "Computer Network Traffic Monitoring Technology and Application" (Tsinghua University Press, 2018 edition), this technology uses an open-source monitoring component (Wireshark) + a dedicated monitoring plugin for enterprise-level OA systems to achieve non-intrusive real-time monitoring without affecting the normal operation of the OA system; Hybrid storage technology: Adopting the industry-mature Redis+MySQL hybrid storage solution (generally used in OA systems), Redis stores the raw data collected in real time (operation logs, status change information), and MySQL stores the structured tagged data (role responsibility tags, data version tags); OA system data adaptation technology: It adopts the common data interface specifications of mainstream OA systems (Fanwei, Zhiyuan) to achieve seamless collection of native data from approval and document modules without modifying the underlying code of the OA system.

[0053] 2) Independently innovated design (the core of this invention, customized for OA dual modules) The predefined tagged data model (OA approval + official document customized version) features core innovations including a three-layer tag hierarchy, dual-module adaptation rules, and cross-module association rules. The three-layer tagging system consists of: a basic attribute layer that records general information such as data ID, operation time, and module; a business semantic layer that associates core business information such as process nodes, business status, and data subject type; and a governance rule layer that binds governance strategies such as verification rules, compliance requirements, and archiving standards. Dual-module adaptation rules: Design exclusive label dimensions for the characteristics of the approval and document modules. For example, the approval module adds "approval level and document type", and the document module adds "document security level and circulation scope". Cross-module association rules: Core data topics (such as basic enterprise information) adopt unified data IDs and tags to achieve data tagging linkage between the two modules.

[0054] 3) OA Approval + Document Module Adaptation Instructions In the case of the OA system of a manufacturing enterprise group, conventional bypass monitoring technology and OA general data interface are used to collect data from five core dimensions of the two modules (user operation behavior, data change records, node attribute information, process status transitions, and task execution logs). Through an independently developed tag-based data model, standardized tags are assigned to data such as "contract approval form" and "contract document", and unified tags are assigned to related data of the two modules to achieve cross-module data linkage and provide a computable data source for subsequent governance.

[0055] Intelligent Governance Engine Module (Figure 4 S420) – The core of the system, the central hub for governance execution in the OA dual-module system. 1) Use / reuse of conventional technologies Drools Rule Engine: Based on the fundamental engine capabilities of the official Drools Rule Engine technical documentation (version 7.0 and above), it serves as the underlying framework for triggering governance rules and provides common open-source technologies for developing process rules in OA systems; Basic data processing algorithms: The system adopts matrix construction, data verification, and version control algorithms commonly used in OA systems to achieve basic data calculation and processing; Cross-module data linkage technology: Using conventional data association and matching algorithms, basic associations of the same data topics between two modules are achieved.

[0056] 2) Independent innovation design (the core of this invention, OA dual-module customized multi-model fusion) Customized multi-model fusion design for OA dual modules: Integrating a responsibility traceability model, a multi-dimensional data collision analysis model, and a standardized verification and compliance model, optimized for OA dual modules, such as adding cross-module collision detection rules and customizing quality / compliance thresholds for each module; Drools rule engine secondary development (OA dual-module status linkage version): realizes dynamic binding of governance rules with dual-module status / cross-module status linkage, single-module status triggers exclusive governance rules, and cross-module status linkage triggers collaborative governance rules; OA Dual-Module Full-Link Dynamic Tag Update Mechanism: Designed real-time tag update rules, tag data is automatically updated after governance actions are completed. For example, after the approval data version is updated, the version tag of the related data in the official document module is updated synchronously to ensure data consistency between the two modules.

[0057] 3) OA Approval + Document Module Adaptation Instructions In the case of the OA system for a manufacturing enterprise group, secondary development based on the Drools rule engine was used to achieve dynamic binding of governance rules with the business status of the two modules; through customized multi-model fusion design, core governance actions such as responsibility tracing, conflict detection, and quality verification of the two modules were completed; through the dynamic update mechanism of tags, the status of the approval and document modules and data tags were updated synchronously to ensure that governance actions and business progress were linked in real time.

[0058] 3. Human-Machine Collaborative Processing Module (Figure 4 S430) – High-Level Anomaly Handling Module in the Dual-Module Structure of the OA System 1) Use / reuse of conventional technologies Multi-device notification technology: It adopts the system pop-up technology commonly used in OA systems, the open interface of WeChat Work, and the SMTP email sending protocol to realize multi-device notification of to-do tasks; Task logging technology: Utilize the OA system to develop a standard Log4j logging component to record logs of the entire manual processing process; OA to-do task integration technology: By referencing the common interface of the to-do task module of mainstream OA systems, exception handling tasks are embedded into the enterprise's original OA to-do center to improve operational convenience.

[0059] 2) Independent innovation design (the core of this invention: standardized compensation and interaction mechanism for OA dual modules) The core innovation of the customized and standardized human-machine collaborative compensation interaction mechanism of the OA system is: OA dual-module abnormal scenario intelligent identification rules: accurately distinguish between single-module / cross-module abnormalities, classify them into low / medium / high levels, and only trigger manual handling for high-level abnormalities; Dual-module precise task assignment rules: Single-module exceptions are assigned to the module manager, and cross-module exceptions are assigned to the overall manager, strictly following the principle of "whoever is responsible, handles it"; Closed-loop feedback rules for cross-module processing results: After cross-module exception handling is completed, the system automatically synchronizes the results to both modules and updates the tag data and governance status; OA office timeout handling mechanism: In line with the company's work rhythm, a 24-hour timeout escalation mechanism is set up to meet the timeliness requirements of daily office work.

[0060] 3) OA Approval + Document Module Adaptation Instructions In the case of the OA system of a manufacturing enterprise group, abnormal tasks are embedded into the original OA task center of employees by integrating the regular OA to-do tasks and multi-terminal notification technology. Through the self-innovated compensation and interaction mechanism, high-level conflicts across modules (such as inconsistencies between contract approval and document amounts) are handled accurately, and the processing results are fed back synchronously to both modules to ensure the integrity of the governance closed loop.

[0061] 4. Data Visualization Module (Figure 4 S440) – OA System Dual-Module Governance Results Presentation and Optimization Module 1) Use / reuse of conventional technologies Visualization rendering technology: Utilizing the basic rendering capabilities of ECharts (open source from Baidu) + Three.js (open source 3D visualization engine), this technology is a general-purpose technology for developing data dashboards in OA systems. Force-directed layout algorithm: This algorithm uses a conventional force-directed layout algorithm from graph theory to achieve automatic layout of graph nodes; Multi-source data aggregation technology: Using the ETL tool (DataX) commonly used in OA systems, governance data and manually processed logs are aggregated; OA system visualization integration technology: By referencing the common integration interface of mainstream OA system data dashboards, the governance map is embedded into the enterprise's original OA system data dashboard.

[0062] 2) Independent innovation design (the core of this invention: OA dual-module graph construction + quantitative optimization closed loop) OA Dual-Module Data Link Relationship Graph: The graph is designed with a three-layer architecture (basic data layer + entity-relationship network layer + management representation layer) + a five-step construction process (corresponding to Figure 3). The core innovation is the addition of nodes and edges for cross-module linkage relationships, realizing the integrated presentation of dual-module governance relationships. OA Dual-Module Governance Effectiveness Quantitative Evaluation Index System: Establish five quantitative indicators (accuracy rate of responsibility traceability, conflict detection rate, quality verification pass rate, compliance inspection pass rate, and archiving completion rate), calculated separately for single module + dual module as a whole, to achieve refined evaluation; OA Dual-Module Analysis - Optimization Closed-Loop Driven Mechanism: Design a graph-based governance bottleneck identification rule, and the optimization suggestions derived from the analysis are automatically fed back to the intelligent governance engine module to update the governance rules, realizing a complete closed loop of "analysis-execution-visualization-optimization".

[0063] 3) OA Approval + Document Module Adaptation Instructions In the case of the OA system of a manufacturing enterprise group, a dual-module 3D governance map is generated in the enterprise's original OA screen by using conventional visualization rendering and OA data screen integration technology. It supports interactive operations such as node expansion and cross-module path tracing. Through an independently developed quantitative indicator system, the governance effect of the dual modules can be monitored in a refined manner. By analyzing the map, governance bottlenecks can be identified and optimization suggestions can be generated to achieve continuous optimization of governance strategies.

[0064] The governance system of this invention features a four-layer architecture that is highly customized for implementation in the OA system's approval and document processing modules. Conventional technologies ensure seamless integration with the enterprise's existing OA system without affecting daily office workflows. The independently developed design enables deep integration of governance and business workflows, unified governance of data from both modules, and real-time control and management across modules.

[0065] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. The embodiments described are part of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0066] Furthermore, the described features, structures, or characteristics can be combined in a suitable manner or form more exemplary implementations. Some of the block diagrams shown in the accompanying drawings are functional entities, which can be implemented in one or more hardware modules or integrated circuits.

[0067] Figure 1. Schematic diagram of governance model construction The label-driven ternary fusion data governance model construction process of this invention is S110-S150, and the specific implementation details of each step are as follows: S110. Define and collect basic workflow elements: Obtain three basic element diagrams from the business process to be managed, and clarify the core content and data dimensions of each diagram: l Participant Relationship Diagram: Defines the many-to-many mapping relationship between process steps and execution roles, with data collection dimensions including process step ID, role ID, and responsibility type; l Business State Diagram: Describes the state transition path of a business object throughout the entire process. The data collection dimensions are process ID, business state ID, state transition conditions, and transition time. Node-task collision graph: This depicts parallel or converging task nodes and the data topics they operate on. The data collection dimensions include task ID, node ID, data topic ID, operation type, and process topology. This step aims to establish the initial inputs and structural baselines for governance.

[0068] S120. Attach dynamic governance tags to target elements: Attach structured three-layer dynamic governance tags (basic attribute layer, business semantic layer, and governance rule layer) to the key elements in the above three diagrams, specifically: In the participant relationship diagram, the "roles" are tagged with responsibility governance, including responsibility type, scope of authority, and operational permissions; in the business status diagram, the "status" is tagged with lifecycle governance, including lifecycle stage, triggered governance actions, and verification compliance standards; and in the node task collision diagram, the "data topic" is tagged with version and conflict dimension, including data version number, version update rules, and conflict detection rules. The tag content is dynamically generated and updated based on business rules and governance strategies, encoding governance semantics into process elements.

[0069] S130. Multi-dimensional Standardized Integration Based on Tags: Using process step IDs and data topic IDs as unified indexes and association keys, information from three independent dimensions is aligned and integrated to form a many-to-many mesh network. For example, a certain "data topic tag" can be linked to all "task nodes" that operate on it, the "participant roles" involved in these tasks, and the current "business status" of the data, thereby constructing a three-element integrated association network with tags as the link, spanning people, events, and things (data), thus forming a tag-driven three-element integrated data governance model.

[0070] S140. Governance Rule Mapping Based on Integrated Model: Based on the integrated model constructed in the previous step, specific data governance rules are precisely mapped to specific dimensions or label combinations of the model. The mapping rules are as follows: Data quality verification rules → business status tags (e.g., "Submitted for Review" or "Review Passed"); compliance check rules → business status tags + responsibility governance tags (e.g., "Archived" status + "Data Archiver" tag); conflict detection rules → node task collision tag combinations (e.g., same data subject ID + multiple operation type tags); archiving strategy rules → business status tags + version and conflict tags (e.g., "Archived" status + "Data Classification" tag). This model enables governance rules to be precisely anchored to specific scenarios within business processes.

[0071] S150. Output a computable and visualized governance relationship graph: The system transforms the ternary fusion model and its mapped governance rules into a computable and traversable data link relationship graph. This graph is presented in a visual form, where nodes represent participants, tasks, data, or states, and edges represent the relationships between them (such as lineage, triggers, and conflicts). It can also highlight the responsibility links, risk points, and governance action trigger points identified by the governance rules, thereby providing an intuitive operational and decision-making view for closed-loop governance.

[0072] As shown in Figure 2, the specific execution flow of the data governance method based on the workflow engine of the present invention is S210-S260. The implementation details and technical implementation of each step have been described in detail in the invention content: S210: Construct a tag-driven ternary fusion data governance model as a unified data foundation and analysis framework for all subsequent governance activities; S220: Construct a data-related person matrix based on the participant's association dimensions and tags to achieve automatic tracking of data lineage and accurate tracing of responsible parties, with a responsibility tracing accuracy rate of ≥99%; S230: Based on the node task collision dimension and its label, it runs a self-developed multi-dimensional data collision analysis model with a collision detection rate of ≥98%, realizing real-time detection of data collisions and version consistency management. S240: Based on business status dimensions and their tags, trigger full lifecycle governance actions through a standardized verification and compliance system to achieve intelligent synchronization of governance strategies and business progress, with a quality verification pass rate and compliance inspection pass rate of ≥95%; S250: Activate a standardized human-machine collaborative compensation mechanism to handle boundary scenarios of automated governance, with an anomaly handling completion rate of 100%, ensuring the integrity of the governance closed loop; S260: Construct a data link relationship graph to achieve visualized presentation and quantitative optimization of governance effects, forming a three-dimensional closed-loop governance system covering responsibility, quality, and conflict, with an archiving completion rate of 100%.

[0073] Figure 3 shows the data link relationship diagram of the present invention, which is a three-layer architecture plus a five-step construction process. The three-layer architecture consists of a basic data layer, an entity-relationship network layer, and a management representation layer. The five-step construction process is S310-S350, and the specific implementation details are as follows: Basic data layer: Defined with dynamic tags as the core, it gathers the full tag data of the three-element fusion governance model, providing a data foundation for map construction; Entity-Relationship Network Layer: Abstracts participants, data entities, business tasks, and process states into nodes, and abstracts logical relationships such as lineage, operation, trigger, and conflict into edges to construct a mesh-like entity-relationship network; Management Presentation Layer: Enables intelligent mapping and graph construction of data. Through multi-dimensional visualization mapping and labeling, it forms an interactive visualization graph and builds an intelligent governance and optimization closed loop. S310. Aggregating Multi-Source Governance Data and Relationships: The system aggregates multi-source data from the ternary fusion governance model and various governance activities in real time, including data lineage links, responsibility entity mapping, data conflict events, version associations, quality verification records, compliance inspection results, lifecycle operation logs, etc., providing a complete structured data foundation for graph construction; S320. Merge and construct an entity-relationship network: Merge the aggregated data, abstract nodes and edges, and use dynamic tags as unified semantic identifiers for nodes and edges to construct an entity-relationship network that can express complex business contexts and governance logic. S330. Visual mapping and marking by governance dimensions: Visual mapping is performed based on the three major governance dimensions of responsibility, quality and conflict, and visual styles such as color, icon, highlight and alarm are used to achieve rapid identification of different governance concerns. S340, Rendering and Generating Interactive Relationship Graphs: Based on the force-directed layout algorithm, the labeled network is automatically laid out and rendered to generate a visual data link relationship graph, supporting interactive operations such as node expansion / collapse, relationship path tracing, dimension filtering, time axis sliding, and drill-down analysis; S350, Governance Optimization Closed Loop Based on Graph Analysis: By analyzing the global relationships, bottlenecks and risk patterns revealed by the graph, governance insights are formed and fed back to the system to optimize governance rules, responsibility matrices or process design, realizing a complete closed loop of "analysis-execution-visualization-optimization".

[0074] Figure 4 shows that the workflow-based data governance system of the present invention has a four-layer architecture, including a data acquisition module (S410), an intelligent governance engine module (S420), a human-machine collaborative processing module (S430), and a data visualization module (S440). The functions, technical details, innovations, and collaborative relationships of each module have been described in detail in the invention description. The core implementation points are: 1. The modules are hierarchically collaborative with a closed-loop data system, ensuring efficient data flow within the system and precise execution of governance actions; 2. All modules of the system are deployed on cloud-native networks, supporting highly available and high-concurrency business scenarios and adapting to enterprise-level large-scale data governance needs; 3. The system's software products are implemented using a microservice architecture, with each module capable of independent deployment, upgrades, and expansion, supporting seamless integration with existing enterprise business systems.

[0075] This patent proposes a data governance method and system based on a workflow engine. Through three core technologies—deep integration, real-time linkage, and intelligent driving—it deeply embeds governance activities into all aspects of the business process, achieving automated and refined control over the entire data lifecycle. This method achieves accurate dynamic data collection through bypass monitoring and a predefined tagged data model. It comprehensively utilizes multi-dimensional analysis, including responsibility association, state transitions, and task collisions, to construct a "human-state-task" three-element intelligent governance model. Using a self-developed multi-dimensional data collision analysis model and a secondary-developed rule engine, it automatically identifies data quality risks, conflicts of authority and responsibility, and compliance issues. A standardized verification and compliance system is used to implement layered and categorized governance strategies. Relying on the driving capabilities of the workflow engine, based on process status and the rule model, it dynamically performs governance actions such as quality verification, conflict resolution, and compliance review on identified data issues. A standardized compensation interaction mechanism addresses boundary scenario issues in automated governance, and a data relationship graph enables visualized monitoring and strategy optimization of governance effects.

[0076] The method and system of this invention completely solve the core industry problem of the disconnect between data governance and business processes from a technical perspective. It integrates a human-machine collaborative compensation processing mechanism and the dynamic optimization capability of governance strategies, further improving the system's adaptability, timeliness and intelligence level, and providing a systematic and embedded integrated solution for data governance in complex business environments.

[0077] Furthermore, the workflow engine-based data governance method of this invention has been implemented in the approval and document modules of the OA system of a large manufacturing enterprise group. This OA system encompasses three core approval modules (financial approval, personnel approval, and contract approval) and three document modules (internal document creation, external document circulation, and document archiving), covering the group headquarters and 12 branch offices. It suffers from typical problems such as fragmented cross-module data flow, difficulty in tracing the responsibility for documents and approvals, data version conflicts caused by parallel approvals, and lagging document lifecycle governance. This method perfectly addresses the industry pain point of traditional data governance being disconnected from business processes. The following section, using the governance method flow of this invention (Figure 2), briefly describes the specific implementation steps and effects of this case study on the approval and document modules of the OA system, intuitively demonstrating the method's implementability, practicality, and scenario adaptability.

[0078] (I) Case Background The manufacturing enterprise group's OA system's approval module processes over 500 transactions daily, including financial reimbursements, personnel appointments and dismissals, and contract approvals. The document module generates / circulates over 2000 internal documents and over 800 external documents annually, involving 10 core roles such as initiators, department reviewers, group approvers, document drafters, proofreaders, and archive administrators. Data is stored in the OA system's local database and the business sub-databases of each branch office. Before applying this method, the system had three major problems: The linkage between approval data and official document data is poor. For example, after a contract is approved, the information of the company issuing the corresponding official document is inconsistent, and there is no real-time conflict detection mechanism. The responsible party is vague; after errors in drafting official documents or mistakes in modifying approval data, it takes more than 1.5 hours to manually trace the responsible party on average. Governance actions were lagging behind, official documents were only subject to simple verification during the archiving stage, there was no full-process quality control for approval data, compliance checks on sensitive financial / personnel data were lacking, and the data archiving completion rate was only 82%.

[0079] This invention addresses the commonalities and interconnectedness of the approval workflow (drafting → submission → multi-level review → approval / rejection) and the official document workflow (drafting → review → signing → circulation → archiving) in OA systems, and customizes a governance model to achieve integrated, automated, and intelligent data governance for both modules.

[0080] (II) Specific implementation steps (corresponding to Figure 2 S210-S260 process) S210: Building a Tag-Driven Tripartite Data Governance Model For the dual-module workflow of approval and document processing in the OA system, the first step is to collect participant relationship diagrams, business status diagrams, and node task collision data. Figure 3The basic elements are then labeled with three layers of dynamic governance tags, ultimately integrating them into a dual-module, three-element fusion governance model: Participant Relationship Diagram: Depicting the many-to-many mapping relationships between 10 core roles and 8 core stages: "approval drafting / review / final review, document drafting / proofreading / issuance / archiving"; Business Status Diagram: Unified definition of 6 statuses in the approval module: "Draft, Submit, Department Review, Group Review, Final Approval, Final Rejection"; and 7 statuses in the document module: "Drafting, Review, Issuance, Circulation, Pending Archiving, Archiving, and Cancellation". Clarify the cross-module status linkage relationship (e.g., "Contract Approval Final Approval" triggers "Contract Document Drafting"). Node task collision graph: Identifies parallel nodes in the approval module, parallel nodes in the document module, and cross-module collision nodes, and sorts out 7 core data themes such as basic enterprise information, approval document information, and document text information; Tag mounting: Mount responsibility governance tags for roles (such as "Data Creator - Approval" "Data Archiver - Dual Module"), mount lifecycle governance tags for statuses (such as "Pending Quality Verification - Approval" "Pending Data Linkage Verification"), and mount version and conflict governance tags for data topics (such as "Version Number V1.0 / Operation Module / Conflict Detection Rules"). Model integration: Using the process ID of the OA system and the data topic ID as the core association key, the integrated association of data, roles, and status of the two modules is realized, and a tag-driven three-element fusion data governance model is constructed.

[0081] S220: Execution of Data Lineage Tracing and Automatic Accountability Tracking Based on the tag data of the participant association graph, a data association matrix adapted to the OA dual modules is constructed: the matrix row dimension is 7 types of core data theme IDs (distinguishing between approval / document modules), the column dimension is 5 types of governance responsibility "creation / modification / approval / use / archiving", and the matrix value is "role / person ID + operation time + operation step + module".

[0082] This matrix enables single-module traceability and cross-module traceability. For example, if there is an error in the financial approval form data, it can be traced back to the auditor within 5 seconds. If the information in the contract document is inconsistent with the approval form information, it can be traced back to the operators of both modules at the same time. In this case, the accuracy rate of responsibility traceability is 99.5%, which is more than 95% more efficient than traditional manual traceability.

[0083] S230: Implement real-time data conflict detection and consistency management The self-developed multi-dimensional data collision analysis model was adapted to the OA dual-module, with a 100ms time slice for polling detection. Real-time early warning and control were implemented for three major collision scenarios. In this case, the collision detection rate reached 98.8%, and the collision handling time was reduced from 24 hours to 40 minutes. Internal collision within the approval module: When multiple departments simultaneously modify the same financial expense report, a medium-level conflict is triggered. The system immediately locks read and write operations and sends an alert. Internal collision within the document module: When multiple branches circulate the same document and multiple versions appear, triggering a low-level conflict, the system automatically generates a version merging suggestion. Cross-module collision: After the contract amount is modified in the approval module, the official document is not synchronized, triggering a medium-level conflict. The system locks the document drafting stage and pushes a linkage modification warning to the operators of both modules.

[0084] S240: Trigger intelligent governance actions throughout the entire lifecycle Based on the dual-module state transitions and cross-module linkage of OA, governance actions are automatically triggered through a standardized verification and compliance system, strictly adhering to the quality verification, compliance inspection, and hierarchical archiving standards of this invention, thereby achieving intelligent synchronization between governance and business progress.

[0085] Data quality verification: Triggered at the "Submit for Review" node of the approval, "Review" node of the official document, and "Approval Passed + Draft" node of the cross-module, the quantitative threshold is customized according to the two modules. The pass rate of quality verification is 97.2% for the approval module and 96.5% for the official document module. Compliance checks are triggered at the "final approval" and "issue" stages of the approval process. Sensitive data related to finance, personnel, contracts, and document security classifications are verified from three aspects: ownership, operation, and storage. The compliance check pass rate is 96.3%. Tiered archiving: Triggered at the "Final Approval" and "Archived" nodes of the approval, the data is classified according to "data importance + module". Core data (contract approval, official documents) are permanently archived with multiple copies, important data (financial reimbursement) is archived for ≥10 years, and general data (approval drafts) is archived for ≥3 years. The archiving completion rate is 100%, and the storage retrieval response time is ≤3s.

[0086] S250: Initiate the human-machine collaborative compensation mechanism. For high-level conflict / anomaly scenarios that automated governance cannot handle (such as severe inconsistencies in core data across modules, or compliance violations of sensitive data), a standardized compensation and interaction mechanism is activated, strictly adhering to the principle of "whoever is responsible, handles it." In this case, the anomaly handling completion rate was 100%, and all high-level issues were resolved within 24 hours. Anomaly Detection: The system automatically identifies high-level scenarios and generates standardized to-do tasks that include anomaly information, causes, and handling suggestions; Intelligent assignment: Single-module exceptions are assigned to the module manager, and cross-module exceptions are assigned to the overall process manager; Multi-terminal notifications: Notifications are pushed via pop-ups in the OA system, WeChat Work, and email. Standardized processing: The responsible person operates in the OA system according to the process of "view → analyze → correct → submit", and the system records the entire process log; Closed-loop feedback: The system performs a second verification of the processing results. If the verification passes, the label data is updated; if it fails, the result is returned with a reason. A 24-hour timeout escalation mechanism is set up. If the processing is not completed within the timeout period, the result is automatically pushed to the superior person in charge.

[0087] S260: Generate a visual governance map and form an optimization closed loop. Based on the dual-module responsibility chain, conflict detection results, governance records, and manual processing logs, a data link relationship diagram of the OA system's dual modules is constructed (corresponding to Figure 3). This diagram is visualized through a three-dimensional governance dashboard, and a quantitative evaluation index system is established to form a closed loop for governance optimization. Visualization of the graph: Nodes represent participants, data entities, tasks, and states of the two modules, while edges represent lineage, triggers, conflicts, responsibilities, and cross-module linkages, realizing an integrated presentation of the governance relationships of the two modules; Visualization of workflow: Presents the entire path of data from creation to archiving using a "timeline + process topology", supporting cross-module path tracing; Quantitative assessment: Five core indicators were established. The actual test results in this case were: 99.5% accuracy rate in tracing responsibility, 98.8% conflict detection rate, 96.8% pass rate in quality verification, 96.3% pass rate in compliance inspection, and 100% completion rate in archiving. Continuous optimization: By identifying governance bottlenecks through graph analysis (such as a high error rate initiator of financial approval in a branch), the system automatically generates optimization suggestions (adding pre-verification rules) and feeds them back to the intelligent governance engine module to update governance rules, realizing a complete closed loop of "analysis-execution-visualization-optimization".

[0088] (III) Overall Effect of Case Implementation After applying this invention to the approval and document processing module of the OA system of this manufacturing enterprise group, the problem of the disconnect between data governance and business workflow was completely solved. The core effects are as follows: Significantly improved governance efficiency: The average processing time for data issues has been reduced from 48 hours to 1 hour, and cross-module data collaboration efficiency has increased by 85%; Significantly improved data quality: Data integrity, accuracy, and consistency improved by more than 30% across both modules; comprehensive compliance management of sensitive data. Accurate identification of responsible parties: Enables traceability and location of responsibility across the entire approval and document processing process, improving traceability efficiency by 95%; Standardized archiving management: The completion rate of dual-module data archiving increased from 82% to 100%, and storage retrieval efficiency improved by 60%; Continuous optimization of the governance system: The governance effect is quantitatively monitored through a visual graph, forming a closed-loop optimization mechanism that adapts to the dynamic adjustment of business processes in the OA system.

[0089] This case fully demonstrates the high adaptability, feasibility, and practical application value of the present invention in the approval and document modules of OA systems.

[0090] Definition: A labeled data model is an unstructured or semi-structured data model that organizes and describes data using "label-value" pairs. Unlike strict relational databases, it offers high flexibility and scalability, making it ideal for handling scenarios involving dynamic attributes, user profiles, and log records.

[0091] The convergence governance model is a system that is business value-oriented and achieves unified management and compliant use of multi-source heterogeneous data through the synergy of technology, processes, and organization.

[0092] A participant relationship diagram is an analytical tool used to visualize all key roles and their interactions within a business or data ecosystem. Through a network diagram composed of "nodes (participants)" and "edges (relationships)," it visually reveals who is participating, how they collaborate, and the flow of data and permissions.

[0093] A node task collision graph is a visual analysis tool used to analyze and optimize multi-task scheduling and resource conflicts. It is commonly used in areas such as data pipeline governance, distributed computing scheduling, and data asset dependency management.

[0094] Multidimensional data collision analysis model is an advanced analytical technique that cross-compares data from different sources and with different attributes to uncover hidden correlations, identify anomalies, or predict trends.

Claims

1. A data governance method based on a workflow engine, characterized in that, Includes the following steps: S1. Dynamic data acquisition with predefined tags By using bypass monitoring technology, dynamic data in the business process is collected in real time in an event-driven manner. The dynamic data monitored includes five dimensions: user operation behavior, data change records, node attribute information, process status transitions, and task execution logs. Real-time data acquisition is achieved through non-intrusive monitoring. Dynamic data is tagged using a labeled data model, with the tags layered into a basic attribute layer, a business semantic layer, and a governance rule layer. The basic attribute layer is divided into data ID, operation time, and operation terminal; the business semantic layer is divided into process node, business status, and participating role; and the governance rule layer is divided into verification rules, compliance requirements, and responsible entity. S2. Construction of the Responsibility Origin Model Tag-based data is dynamically expanded and integrated through a fusion governance model that includes personnel, status, and tasks. The fusion governance model consists of a participant association graph, a business status graph, and a node task collision graph, forming an integrated tag data network. S3. Accountability and Conflict Detection A dynamic data-related personnel matrix is ​​constructed through an integrated tag data network and participant association graph to achieve automatic tracking of data lineage and accurate tracing of responsible parties; a multi-dimensional data collision analysis model is run through a node task collision graph to detect data conflict risks in real time and trigger control actions. S4. Intelligent governance cycle triggering Automatically trigger data quality verification, compliance checks, and tiered archiving governance by calling the tagged data model through the business status diagram; S5. Human-machine collaborative compensation processing For boundary scenarios that S4 cannot automatically manage, a compensation interaction mechanism is initiated through the governance rule layer to generate exception handling tasks and assign them to the responsible entities, thereby achieving closed-loop feedback of the processing results. S6. Visualized Closed-Loop Presentation By constructing a data link relationship map by integrating the governance records and processing results in S4 and S5 with the integrated tag data network, the linkage relationship and governance effect of the data are presented, and governance is continuously optimized based on quantitative evaluation indicators.

2. The data governance method based on a workflow engine according to claim 1, characterized in that... In step S1, the tagged data model can be predefined, wherein the governance rule layer is bound to at least one of the following: data quality verification rules, compliance requirement rules, responsible entity rules, and conflict detection rules.

3. The data governance method based on a workflow engine according to claim 1, characterized in that... The detection rules of the multi-dimensional data collision analysis model in step S3 include at least one of the following: multiple operation collision rules for the same data topic, multiple version data collision rules for the same task, and parallel task data read / write collision rules; the control actions include at least one of the following: generating version merging, locking conflicting data for read / write, and pausing the execution of related task nodes.

4. The data governance method based on a workflow engine according to claim 1, characterized in that... The data quality verification in step S4 includes integrity, accuracy, consistency, and timeliness verification. The data quality verification sets basic thresholds, early warning thresholds, and anomaly thresholds for each verification dimension of integrity, accuracy, consistency, and timeliness. The hierarchical archiving governance is triggered based on the hierarchical nature of the business type.

5. A data governance method based on a workflow engine according to claim 1, characterized in that... The compensation interaction mechanism in step S5 includes: abnormal scene identification, task assignment, multi-terminal notification, standardized processing flow, secondary verification of processing results, and timeout handling mechanism.

6. The data governance method based on a workflow engine according to claim 1, characterized in that, The nodes of the data link relationship graph in step S6 include participants, data entities, business tasks, and process status, and the edges include lineage relationships, trigger relationships, conflict relationships, and responsibility relationships; the quantitative evaluation indicators include at least the accuracy rate of responsibility tracing, conflict detection rate, quality verification pass rate, compliance inspection pass rate, and archiving completion rate.

7. A data governance system based on a workflow engine, used to implement the data governance method based on a workflow engine as described in any one of claims 1 to 6, characterized in that, include: The data acquisition module is used to collect dynamic data in the business process in real time through bypass monitoring technology, and to perform tagging processing based on a predefined tagged data model; The intelligent governance engine module, connected to the data acquisition module, is used to construct a three-element integrated governance model and perform governance actions such as responsibility tracing, conflict detection, quality verification, compliance inspection, and hierarchical archiving. The human-machine collaborative processing module is connected to the intelligent governance engine module and is used to provide a standardized compensation and interaction mechanism for the boundary scenarios of automated governance. The data visualization module, connected to the intelligent governance engine module and the human-machine collaborative processing module, is used to construct a data relationship graph and present the governance effect.

8. The data governance system based on a workflow engine according to claim 7, characterized in that, The intelligent governance engine module includes a model management submodule, a rule engine submodule, a data processing submodule, and a tag update submodule; the data visualization module includes a data aggregation submodule, a graph construction submodule, a visualization rendering submodule, a quantitative evaluation submodule, and an optimization analysis submodule.

Images