Enterprise business process intelligent management and control system and method

Abstract

The application belongs to the technical field of enterprise information management, and discloses an intelligent management and control system and method for enterprise business processes, which comprises five core modules, namely, a multi-source data fusion perception engine, a process dynamic self-evolution modeling module, a process risk prediction and intelligent decision module, a cross-process collaborative management hub and a process value quantification and closed-loop review module. Firstly, a business process knowledge graph is constructed and updated through multi-source data fusion, and then a process model suitable for a business scenario is dynamically generated in combination with a scene portrait library, while a process risk is probabilistically predicted and an optimal intervention scheme is generated, so as to realize cross-process global collaborative management. Finally, process review is completed based on a multi-dimensional value index system, and an optimization instruction is output to realize closed-loop iteration of the model and the process rules. The application improves the dynamic adaptability and global collaboration of the business process, reduces the process operation risk, realizes the continuous optimization of the process management and control system, and is suitable for intelligent management of business processes of various enterprises.

Description

Technical Field

[0001] This invention relates to the field of enterprise information management technology, and in particular to an intelligent control system and method for enterprise business processes. Background Technology

[0002] Business Process Management (BPM) is a core tool for enterprises to achieve standardized operations, improved efficiency, and controllable risks. Traditional BPM systems are typically designed based on predefined process models, driving the execution of process instances through a workflow engine and providing basic functions such as node approval, path navigation, and status monitoring. As enterprises become increasingly digitalized, the data environment on which business processes rely is becoming more complex, involving not only multiple heterogeneous internal systems (such as ERP, CRM, and OA) but also external information such as market dynamics, supply chain status, and industry policies.

[0003] The fundamental limitation of existing enterprise business process control systems lies in their "static, passive, and open-loop" control paradigm. Business process models need to be predefined statically by humans and cannot evolve dynamically with business scenarios, resulting in delayed change response. Furthermore, the control logic remains in a linear mode of "execution-monitoring-threshold alarm," which can only passively respond to anomalies that have occurred and lacks the ability to proactively predict risks and intervene based on multi-source data. The system lacks a value quantification and closed-loop review mechanism throughout the entire process lifecycle, and the process execution effect cannot be automatically fed back to the modeling and decision-making stages, causing optimization to rely on human experience and control capabilities to be unable to iterate autonomously. Summary of the Invention

[0004] The present invention aims to provide an intelligent management and control system and method for enterprise business processes to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: An intelligent management and control system for enterprise business processes includes: a multi-source data fusion perception engine, a process dynamic self-evolution modeling module, a process risk prediction and intelligent decision-making module, a cross-process collaborative management and control center, and a process value quantification and closed-loop review module. The multi-source data fusion perception engine is configured to collect multi-source heterogeneous data from within and outside the enterprise, and to fuse and process the multi-source heterogeneous data in order to build and continuously update the enterprise's business process knowledge graph. The process dynamic self-evolution modeling module is configured to: in the initial stage, construct and maintain a business scenario profile library through natural language processing technology, where each scenario profile is associated with basic process rules; in the running stage, access the knowledge graph and, based on the knowledge graph and the business scenario profile library, respond to changes in the enterprise's business scenario by dynamically combining and optimizing the basic process rules through feature matching and fuzzy reasoning algorithms to generate a business process model adapted to the current business scenario. The process risk prediction and intelligent decision-making module is configured to access the knowledge graph and business process model, identify potential risks through the prediction model, generate and execute the optimal intervention plan through the intelligent decision-making model, and output decision execution feedback data. The cross-process collaborative management and control hub is configured to access the knowledge graph and business process model, perform unified monitoring and visualization of all business processes of the enterprise, realize global resource scheduling and process flow path adjustment across processes through collaborative scheduling algorithms, and output the global process running status to the process value quantification and closed-loop review module. The process value quantification and closed-loop review module is configured to receive the knowledge graph, decision execution feedback data and global process operation status based on a preset multi-dimensional value quantification index system. It analyzes and forms a review result that includes basic process rules, prediction model and intelligent decision model parameter optimization instructions. The optimization instructions are output to the process dynamic self-evolution modeling module to optimize the basic process rules, and output to the process risk prediction and intelligent decision module to optimize the model parameters.

[0006] Preferably, the process dynamic self-evolution modeling module includes: The scenario profile building unit is configured to perform the operations of the initial stage, analyze enterprise business systems and historical business data through natural language processing technology, build and maintain a standardized business scenario profile library, and associate each scenario profile with the corresponding basic process rules. The dynamic process generation unit is configured to perform the core operations of the running phase, access the knowledge graph, and respond to changes in enterprise business scenarios based on the knowledge graph and business scenario profile library. It completes scenario profile matching through feature extraction, and dynamically combines, resolves conflicts and optimizes parameters of the associated basic process rules based on fuzzy reasoning algorithm to generate an adapted current business process instance. The self-learning optimization unit is configured to receive feedback data such as node execution efficiency and exception rate during the process execution, and iteratively update the mapping relationship between scenario profiles and process rules through reinforcement learning algorithms to optimize the business scenario profile library and basic process rules.

[0007] Preferably, the multi-source data fusion sensing engine includes: The multi-source data acquisition unit is configured to collect heterogeneous data from multiple sources inside and outside the enterprise. Specifically, it collects process execution data, business value data, and real-time process operation data by connecting to the enterprise's internal business systems through interfaces, collects external market, supply chain, and policy data by web crawling or API connection, and collects human-computer interaction behavior data by behavior recognition technology. The data fusion processing unit is configured to receive multi-source heterogeneous data collected by the multi-source data acquisition unit, clean, associate and fuse the multi-source heterogeneous data, and construct and continuously update an enterprise business process knowledge graph based on the fusion results. The knowledge graph is used to represent process nodes, participating entities, resources, control indicators, full data of process operation and the relationship between various types of data.

[0008] Preferably, the process risk prediction and intelligent decision-making module includes: The process risk prediction model is configured to access the knowledge graph and business process model, and based on time series prediction algorithms and Bayesian networks, perform probabilistic prediction and graded early warning of node timeout risk, resource shortage risk, approval backlog risk, compliance risk and value loss risk, and output risk prediction results. The intelligent decision-making model is configured to receive risk prediction results, and based on deep reinforcement learning algorithms and expert rule bases, automatically generate multiple intervention plans for predicted risks or abnormal situations in process operation. It then selects and executes the optimal intervention plan through cost-benefit analysis algorithms and outputs decision execution feedback data. The model iteration unit is configured to receive model parameter optimization instructions output by the process value quantification and closed-loop review module, as well as decision execution feedback data, and to iteratively optimize the model parameters and operating strategies of the process risk prediction model and the intelligent decision model.

[0009] Preferably, the cross-process collaborative management and control center includes: The global process monitoring unit is configured to access the knowledge graph and business process model to uniformly monitor and visualize the running status, resource usage, and risk warning information of all business processes of the enterprise. The collaborative scheduling engine is configured to receive monitoring data from the global process monitoring unit, identify the relationships between various business processes based on the knowledge graph, and dynamically adjust the global resource allocation and process flow path through the collaborative scheduling algorithm when it is detected that the running status of a certain process has a negative impact on the associated processes. The status output unit is configured to collect the full-process operation status data of the global process monitoring unit, organize it into a standardized global process operation status, and output it to the process value quantification and closed-loop review module.

[0010] Preferably, the process value quantification and closed-loop review module includes: The value indicator system management unit is configured to build and maintain a preset multi-dimensional value quantification indicator system, which includes efficiency indicators, cost indicators, value indicators and compliance indicators, providing a quantitative basis for the generation of review results; The process value analysis unit is configured to receive the knowledge graph, decision execution feedback data, and the global process operation status output by the cross-process collaborative management and control center. Based on a multi-dimensional value quantification index system, it performs multi-dimensional analysis on various types of data through data mining algorithms to identify low-value nodes and resource waste points in the process operation, generate a process value management and control report, and analyze and form a review result. The optimization instruction output unit is configured to extract basic process rule optimization instructions and parameter optimization instructions for the prediction model and intelligent decision model from the review results, output the basic process rule optimization instructions to the process dynamic self-evolution modeling module, and output the parameter optimization instructions to the process risk prediction and intelligent decision module.

[0011] A method for intelligent management and control of enterprise business processes based on the above system, characterized by the following steps: S1. Initial Profile Library Construction: Using natural language processing technology, analyze the enterprise's business systems and historical business data to build and maintain a business scenario profile library. Each scenario profile is associated with the corresponding basic process rules. S2. Multi-source data fusion and graph update: Collect multi-source heterogeneous data from within and outside the enterprise, clean, associate and fuse the multi-source heterogeneous data, and build and continuously update the enterprise business process knowledge graph. S3. Dynamic self-evolution of process generation: Access the knowledge graph, and based on the knowledge graph and business scenario profile library, respond to changes in enterprise business scenarios, dynamically combine and optimize basic process rules through feature matching and fuzzy reasoning algorithms to generate a business process model that adapts to the current business scenario. S4. Process Risk Prediction and Decision Execution: Access the knowledge graph and business process model, identify potential process risks through the prediction model, generate and execute the optimal intervention plan through the intelligent decision model, and output decision execution feedback data. S5. Cross-process collaborative management and status feedback: Access the knowledge graph and business process model to uniformly monitor and visualize all business processes of the enterprise, realize global resource scheduling and process flow path adjustment across processes through collaborative scheduling algorithms, and output the global process running status to the process value quantification and closed-loop review module. S6. Process Value Review and Instruction Output: Based on the preset multi-dimensional value quantification index system, the system receives the knowledge graph, decision execution feedback data and global process operation status, analyzes and forms a review result including basic process rules, prediction model and intelligent decision model parameter optimization instructions, and outputs the optimization instructions to the process dynamic self-evolution modeling module and the process risk prediction and intelligent decision module respectively. S7. Single Closed-Loop Iteration: Update the basic process rules and business scenario profile library based on the received basic process rule optimization instructions, and iteratively optimize the parameters and operation strategies of the prediction model and intelligent decision-making model based on the received parameter optimization instructions, thus completing a single closed-loop iteration of enterprise business process control.

[0012] Preferably, in step S3, the dynamic self-evolution of the process specifically includes: Access the knowledge graph to extract feature data of the current business activity in real time, and perform feature matching with the scene profiles in the business scenario profile library to determine the target scene profile. Based on the matched target scenario profile, the applicability assessment, conflict resolution and parameter optimization of multiple basic process rules associated with it are carried out through fuzzy inference algorithm, and process instances adapted to the current business scenario are dynamically combined to generate process instances. The generated business process model will be synchronized to the process risk prediction and intelligent decision-making module and the cross-process collaborative management and control center for them to access and call.

[0013] Preferably, in step S4, the process risk prediction and decision execution specifically includes: Access the knowledge graph and business process model, extract real-time process operation data and node relationships, and calculate the probability of occurrence and impact level of various risks based on LSTM time series prediction algorithm and Bayesian network. When the risk level exceeds the preset threshold, the intelligent decision-making model is triggered, and multiple candidate intervention schemes are generated by combining the expert rule base, including process rerouting, resource reallocation, and temporary adjustment of approval authority. The expected performance and implementation cost of each candidate intervention plan are quantitatively evaluated using a cost-benefit analysis model. The plan with the highest overall benefit is automatically selected and executed. The decision-making process and results are recorded simultaneously, and decision-making feedback data is generated and output.

[0014] Preferably, it further includes: S8. Continuous optimization of the process control system: Regularly collect full feedback data on the execution of various business processes, optimization suggestions from business departments, and logs of control indicators to build a process optimization input dataset; A multi-objective optimization algorithm is adopted to optimize the existing basic process rules, expert rule base and multi-dimensional value quantification indicator system by reducing process operation risks, shortening the overall process cycle, improving the stability of control indicators and process business value. The algorithm generates corresponding optimization and adjustment schemes. The cross-departmental management organization reviews and confirms the optimization and adjustment plan based on preset review rules. After the review is approved, the basic process rules, expert rule base and indicator system are dynamically adjusted, and the enterprise business process knowledge graph and its entity association analysis logic are updated simultaneously to realize the continuous iterative upgrade of the enterprise business process control system.

[0015] The beneficial effects of this technical solution compared to existing technologies are as follows: (1) This technical solution achieves multi-dimensional and semantic real-time perception of the internal and external environment of process operation by setting up a multi-source data fusion perception engine and constructing a unified enterprise business process knowledge graph. This enables subsequent process modeling, risk prediction and value review to be carried out on the same set of panoramic and related high-quality data foundation, ensuring the consistency, accuracy and timeliness of the entire system's cognition.

[0016] (2) By setting up a process dynamic self-evolution modeling module with dual-state operation of "initial stage" and "running stage", the system realizes the automated generation and dynamic optimization of business processes. This enables the system to respond in real time to changes in enterprise business scenarios, organizational structure or market conditions, automatically combine and adjust the optimal process path and rules, improve business agility, and avoid manual repetitive configuration and operational gaps caused by process rigidity.

[0017] (3) By setting up a process risk prediction and intelligent decision-making module, the system can identify, quantify and assess potential risks in the process operation in advance and automatically intervene. This enables the system to provide early warnings and automatically execute the optimal handling plan before problems such as node timeouts, resource depletion, and compliance deviations actually occur and cause business losses, thereby enhancing the robustness and continuity of the business process.

[0018] (4) By setting up a cross-process collaborative management and control center, unified monitoring and intelligent resource scheduling of all parallel business processes across the enterprise are realized. This enables the system to automatically identify and resolve resource competition and state conflicts between different processes from the perspective of optimizing the overall operational efficiency of the enterprise, thereby achieving cross-departmental and cross-business line collaborative efficiency and improving the overall resource utilization efficiency.

[0019] (5) By setting up a process value quantification and closed-loop review module, and forming a feedback loop with the process dynamic self-evolution modeling module and the process risk prediction and intelligent decision-making module, continuous self-optimization of the process based on objective data is realized. This enables the system to perform multi-dimensional quantitative evaluation of the effect of each process execution, automatically identify value loss points, and accurately feed back specific optimization instructions to the process rules and decision-making model, promoting the continuous iteration and autonomous evolution of the entire control system without deep human intervention. Attached Figure Description

[0020] Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention; Detailed Implementation The present invention will now be described in further detail with reference to the accompanying drawings and embodiments: like Figure 1 The system described is an intelligent management and control system for enterprise business processes, comprising a multi-source data fusion perception engine, a dynamic self-evolutionary process modeling module, a process risk prediction and intelligent decision-making module, a cross-process collaborative management and control hub, and a process value quantification and closed-loop review module. This system is built on the enterprise's existing workflow execution platform. The workflow execution platform, a component of existing technology, is responsible for parsing, instantiating, and driving the execution of process models, and provides full-scale process operation data through logs, databases, and event interfaces. The various modules of this system interact with the platform through standardized data interfaces (such as APIs, message queues, and database queries) to achieve perception, decision-making, and management functions.

[0021] 1. Multi-source data fusion perception engine The multi-source data fusion sensing engine includes a multi-source data acquisition unit and a data fusion processing unit.

[0022] Multi-source data acquisition unit This unit is configured to collect heterogeneous data from multiple sources both inside and outside the enterprise, specifically through various technical means to achieve data collection: For internal business systems (such as ERP and CRM), structured process execution data (such as process instance ID, activity node, status, and timestamp) and business value data (such as associated order amount and cost item) can be collected periodically through predefined API interfaces or database connectors in a polling or event listening manner. For external data, configure dedicated web crawler scripts to target specific policy websites and industry information platforms for crawling; at the same time, for third-party data services that provide open APIs (such as Tianyancha and logistics tracking platforms), authenticate and retrieve data through API keys; For human-computer interaction behavior data, by embedding a lightweight SDK into the front-end interface or monitoring the operation logs of the RPA robot, unstructured logs such as user click streams, dwell time, and operation sequences are captured and sent to the back-end server.

[0023] Data fusion processing unit This unit is configured to receive multi-source heterogeneous data collected by the multi-source data acquisition unit and perform the following operations to complete data fusion and knowledge graph construction: Data cleaning: Receive raw data streams from multi-source data acquisition units and perform outlier removal, missing value filling, and format standardization based on a predefined rule base; Entity Association: This unit maintains a global entity recognition database. For the cleaned data, a similarity-based entity parsing technique is used: calculating the Jaccard similarity of the name fields between different records, and the cosine similarity of the semantic vectors of the key attribute fields generated by the BERT model. The data fusion processing unit combines the two similarities using a pre-defined decision function, taking name similarity and attribute semantic similarity as features, and inputting them into a binary classification model (such as logistic regression) trained on historical positive and negative sample pairs. This model directly outputs the probability that the pair of entities is the same entity. If the probability exceeds a preset threshold, it is determined to be the same entity (such as the same supplier). Knowledge Graph Construction and Update: Based on the results of entity association, this unit converts the cleaned and associated structured data into triples of "entity-relationship-entity" or "entity-attribute-value", and uses graph database operation interfaces to insert or update these triples as incremental facts into the enterprise business process knowledge graph.

[0024] 2. Process Dynamic Self-Evolution Modeling Module The process dynamic self-evolution modeling module includes a scene profile construction unit, a dynamic process generation unit, and a self-learning optimization unit.

[0025] Scene profile building unit This unit is configured to perform operations during the initial stage of the system. It uses natural language processing technology to build and maintain a standardized business scenario profile library, with each scenario profile associated with corresponding basic process rules. The specific implementation is as follows: During system initialization, this unit uses Natural Language Processing (NLP) tools (such as spaCy and HanLP) to parse the input enterprise business policy documents. It extracts core entities such as "department," "role," "amount threshold," and "approval node" using Named Entity Recognition (NER) technology, and extracts "if...then..." type conditional constraint rules using dependency parsing technology, forming structured rule fragments. Simultaneously, this unit analyzes historical business data, performing cluster analysis on completed process instances. Using the K-means clustering algorithm, process instances with similar characteristics (such as participating departments, processing time, resource consumption, and business type) are grouped into classes, each class representing an initial business scenario profile. Each scenario profile is assigned a unique set of feature labels and associated with a corresponding basic process rule template. The rule template is stored in a declarative language (such as DMN) or structured JSON format, explicitly defining the possible nodes, node order relationships, constraints, and basic parameters of the process within that scenario, facilitating subsequent dynamic invocation and optimization.

[0026] Dynamic process generation unit This unit is configured to perform core operations during system operation. It accesses the knowledge graph and, based on the knowledge graph and business scenario profile library, responds to changes in enterprise business scenarios initiated by users or monitored by the multi-source data fusion perception engine. Through feature matching and fuzzy inference algorithms, it dynamically combines and optimizes basic process rules to generate business process instances adapted to the current business scenario. This unit has a built-in standardized fuzzy rule library, specifically implemented as follows: During system operation, this unit is automatically triggered when a new business request arrives. First, it accesses the enterprise business process knowledge graph built by the multi-source data fusion perception engine to obtain all contextual information related to the business request (such as the applicant's historical performance, the risk level of associated projects, and the occupancy status of related resources). This contextual information is then converted into standardized feature vectors, forming a feature representation of the current business scenario. Next, the unit matches the feature vector of the current scenario with the feature vectors of all scenario profiles in the business scenario profile library. A weighted cosine similarity calculation method is used to quantify the degree of matching between the current scenario and each candidate scenario profile. The feature weights can be pre-set by domain experts or automatically optimized and adjusted by training a machine learning model using historical business data. The top k scenario profiles with the highest matching degree are selected as candidate profiles. For each candidate profile's associated basic process rule template, the unit calls a built-in fuzzy inference algorithm to perform an applicability evaluation. The fuzzy rule base stores fuzzy judgment logic for rule adaptation under various scenarios (e.g., "If the business urgency is 'high' and the associated amount is 'large,' then the confidence level for simplifying the approval level is 0.8"). Through fuzzy inference algorithms, applicability judgments, dynamic combinations, and parameter adjustments are made to each candidate basic process rule (e.g., changing the node time limit parameter from "regular" to "expedited"). Simultaneously, a confidence-weighted voting method resolves potential logical conflicts between different rules, ensuring that the combined rule logic is coherent and adaptable to the current scenario. Finally, based on the optimized rule combination, a fully adaptable and directly executable business process model is generated, typically described in BPMN 2.0 XML format or a system-defined DSL. After generation, it is published to the enterprise's workflow execution engine via a standard API interface. The runtime status data generated during the engine's execution is accessible and callable by other modules.

[0027] The enterprise's workflow execution engine is a well-known basic component used to parse, instantiate, and drive the execution of the generated business process model. At the same time, it collects the running status data of the process instance in real time through three methods: log output, database synchronization, and event callback interface (including process instance ID, current node, handler, execution time, node status, resource usage, etc.), and stores it in the shared data layer according to the preset data format or actively pushes it to each related module through the message queue, providing real-time and accurate data support for the collaborative work of each module.

[0028] Self-learning optimization unit This unit is configured to receive feedback data such as node execution efficiency and exception rate during process execution, and iteratively update the mapping relationship between scenario profiles and process rules through reinforcement learning algorithms to optimize the business scenario profile library and basic process rules. The specific implementation is as follows: This unit continuously monitors the execution of all business processes within the enterprise, receiving real-time feedback data such as node execution efficiency, exception rate, and processing time pushed by the enterprise workflow execution engine through a shared data layer or message queue. It also receives process value assessment scores from the process value quantification and closed-loop review module, collecting performance feedback data for each completed process instance, including actual node processing time, overall process execution cycle, cost consumption, and the number and type of exceptions. This unit employs a reinforcement learning framework for self-learning optimization, treating the "scenario profile selection and basic process rule combination strategy" as the agent's behavioral strategy. The value assessment score after process execution (provided by the subsequent process value quantification and closed-loop review module) serves as the reward signal for reinforcement learning, quantifying the merits of the current strategy. Using policy gradient algorithms (such as the REINFORCE algorithm), and with historical feedback data and reward signals as input, the reinforcement learning model is trained, iteratively updating the parameters of the policy network to optimize the granularity of scenario profile segmentation, feature weights, and the mapping relationship between scenario profiles and basic process rule templates. At the same time, the optimized scene profiles and rule mapping relationships will be updated to the business scene profile library to achieve continuous iteration of the scene profile library and basic process rules, thereby improving the accuracy and adaptability of subsequent dynamic process generation.

[0029] 3. Process Risk Prediction and Intelligent Decision-Making Module The process risk prediction and intelligent decision-making module includes a process risk prediction model, an intelligent decision-making model, and a model iteration unit.

[0030] Process risk prediction model The model is configured to access the knowledge graph and business process model, and based on time series prediction algorithms and Bayesian networks, performs probabilistic prediction and hierarchical early warning for node timeout risk, resource shortage risk, approval backlog risk, compliance risk, and value loss risk, outputting risk prediction results. The specific implementation is as follows: This model is a hybrid prediction model that employs appropriate prediction methods for different types of process risks to ensure accuracy and timeliness. For time-series risks such as node timeouts and process delays, a Long Short-Term Memory (LSTM) network is used for prediction. The model input is a sequence of node running statuses within a preset time window (such as node running time, current processing progress, and historical processing time statistics) obtained by querying the enterprise workflow execution engine's database or listening to its event callback interface. The model learns the temporal patterns of node running through training and outputs the probability of the node timeout at the next moment. For related risks such as resource shortages and compliance risks, a Bayesian network is used for prediction. The structure of the Bayesian network is based on the predefined entity relationships in a knowledge graph. Network nodes correspond to various risk events and influencing factors, and the edges between nodes correspond to the relationships between factors and risk events. A conditional probability table (CPT) is also pre-defined for each node. When new evidence of risk impact is received by monitoring the event interface of the workflow execution engine or querying its runtime database (such as "the utilization rate of a core resource reaches 90%" or "an approval node fails to start within the prescribed time limit"), the model calculates the posterior probability of the target risk event (such as "an activity is blocked due to insufficient resources" or "a process has compliance risks") through Bayesian network probability propagation inference, quantifying the likelihood of the risk occurring. The model integrates the occurrence probability of various risks with preset risk impact levels (set by domain experts based on enterprise business needs, divided into high, medium, and low levels) to calculate the final risk score for each risk event. The risk score is the product of the occurrence probability and the impact level. Based on the preset risk score threshold range, various risks are given graded warnings (red, yellow, and blue levels). Finally, standardized risk prediction results are output, including information such as risk type, occurrence probability, risk level, and impact scope, and are simultaneously pushed to the intelligent decision-making model.

[0031] Intelligent decision-making model The model is configured to receive risk prediction results and, based on deep reinforcement learning algorithms, an expert rule base, and cost-benefit analysis algorithms, automatically generate multiple intervention plans for predicted risks or abnormal situations in process operation. It then selects and executes the optimal intervention plan and outputs decision execution feedback data. The model incorporates a standardized expert rule base, specifically implemented as follows: The model is automatically activated when it receives a risk warning from the process risk prediction model or detects anomalies in the process by monitoring the abnormal event logs of the enterprise workflow execution engine. First, the model accesses the built-in expert rule base, which stores standard handling plans for various risks and anomalies. Based on the current risk type, risk level, and anomaly description, it matches the corresponding handling plan from the expert rule base to generate basic candidate intervention solutions. Simultaneously, the model calls a decision network based on deep reinforcement learning (DRL) (such as Deep Q-Network), taking the overall state of the current system (including the process running status, resource occupancy status, risk warning information, and business priority obtained through the workflow execution engine) as input, and outputs a set of supplementary or alternative intervention actions to enrich the diversity of candidate intervention solutions. All candidate intervention solutions are quantitatively evaluated by the model using a built-in cost-benefit analysis algorithm: benefits are quantified as the expected value of risk reduction, the expected benefit of improved process efficiency, etc.; costs are quantified as the resource consumption, opportunity cost, and execution time required for the solution execution, etc., and then the net benefit of each candidate solution is calculated, which is the difference between the quantified benefit and the quantified cost. Ultimately, the model selects the candidate solution with the highest net benefit as the optimal intervention solution and automatically executes the solution through the system's preset interfaces (such as calling API interfaces to reallocate core resources, sending early warning notifications to the corresponding responsible persons, temporarily adjusting approval permissions, and replanning the process flow path through the workflow execution engine's process adjustment interface). During the solution execution process, the execution process, execution steps, and execution results are recorded in real time to form standardized decision execution feedback data, which is synchronously output to the model iteration unit and the process value quantification and closed-loop review module.

[0032] Model Iteration Unit This unit is configured to receive model parameter optimization instructions and decision execution feedback data output by the process value quantification and closed-loop review module, and to iteratively optimize the model parameters and operating strategies of the process risk prediction model and intelligent decision model. The specific implementation is as follows: This unit continuously receives two core inputs: first, model parameter optimization instructions output by the process value quantification and closed-loop review module. These instructions are in a structured format, clearly specifying the direction, range, or specific adjustment values ​​for model parameters, and are generated based on the process review analysis results; second, decision execution feedback data output by the intelligent decision-making model, including information such as the execution results of each intervention plan, risk reduction effects, actual benefits, and costs. For the process risk prediction model, this unit utilizes new labeled historical data (including historical risk occurrences, deviations between prediction results and actual results obtained through the workflow execution engine) to perform supervised fine-tuning of the parameters of the LSTM and Bayesian networks, correcting model prediction biases and improving the accuracy and timeliness of risk prediction. Simultaneously, it adjusts key parameters such as the risk score calculation logic and early warning thresholds according to the optimization instructions. For the intelligent decision-making model, this unit uses the feedback data from each decision execution as new experience samples and stores them in the experience replay buffer of the deep reinforcement learning algorithm. It periodically triggers model retraining, updates the weight parameters of the DRL decision network, and optimizes the generation logic of intervention plans and cost-benefit analysis strategies (cost-benefit analysis strategies are standardized execution logic built into the intelligent decision-making model, predefined by the enterprise based on business needs, and iteratively optimized by combining historical business data and decision execution feedback data). At the same time, it adjusts the matching logic of the expert rule base in conjunction with optimization instructions to improve the accuracy of selecting the optimal intervention plan and ensure that the model performance continues to improve with business operations.

[0033] 4. Cross-process collaborative management and control center The cross-process collaborative management and control center includes a global process monitoring unit, a collaborative scheduling engine, and a status output unit.

[0034] Global process monitoring unit This unit is configured to access knowledge graphs and business process models, enabling unified monitoring and visualization of the operational status, resource usage, and risk warning information of all enterprise business processes. The specific implementation is as follows: This unit provides a standardized, visual web console as a unified monitoring entry point for all enterprise business processes, allowing administrators to view, retrieve, and filter various process information in real time. The backend service periodically queries the knowledge graph built by the multi-source data fusion perception engine and the enterprise's workflow execution engine according to preset time intervals (e.g., every minute). By calling the workflow execution engine's status query API or directly reading its shared database, it obtains real-time running status data for all active and historical process instances, including process instance ID, current node, node manager, node processing progress, resource usage, risk warning tags, and execution time. Simultaneously, this unit calls front-end chart libraries (such as ECharts and D3.js) to render the acquired monitoring data into various visual views, including process Gantt charts, node running status graphs, resource usage heatmaps, and risk warning statistics. All views support real-time updates, ensuring administrators can grasp the overall operation of all enterprise business processes in real time. In addition, this unit supports anomaly alarm function. When an abnormal process operation or risk warning is detected by listening to the abnormal event callback of the workflow execution engine, an alarm prompt will automatically pop up in the visual console and the alarm information will be pushed to the terminal device of the corresponding person in charge for timely handling.

[0035] Cooperative scheduling engine The engine is configured to receive monitoring data from the global process monitoring unit, identify the relationships between various business processes based on a knowledge graph, and dynamically adjust global resource allocation and process flow paths through a collaborative scheduling algorithm when the operating status of a certain process is detected to have a negative impact on related processes. The specific implementation is as follows: The engine continuously scans the monitoring data output by the global process monitoring unit according to a preset cycle, gaining real-time insight into the operational status, resource utilization, and inter-process relationships of all business processes within the enterprise. By accessing a knowledge graph, the engine automatically identifies the relationships between various business processes (such as upstream / downstream processes and shared resource processes) based on the relationships between process nodes, participating entities, and resources stored within the knowledge graph, thus establishing a process relationship graph. When the engine detects that the operational status of a certain business process negatively impacts related processes (e.g., a process is delayed due to insufficient resources, affecting the normal startup of downstream processes; a process has redundant resource utilization while another related process is resource-scarce), it models the problem as a constrained optimization problem. The optimization objective is to minimize overall enterprise process delays and maximize resource utilization. The decision variables are the resource allocation scheme for the next time period and the adjustment scheme for process flow paths. The built-in collaborative scheduling algorithm (such as heuristic algorithms like genetic algorithms) solves this optimization problem. The algorithm evaluates a large number of possible scheduling schemes in batches, combining constraints such as process priority, resource requirements, and business time limits to select the optimal scheduling instruction that improves the overall process efficiency and reduces negative impacts. After the scheduling instruction is generated, it is automatically sent to each process execution node and resource management module through the resource allocation interface and process adjustment interface of the enterprise workflow execution engine. It automatically performs operations such as resource reallocation and process flow path adjustment to ensure the smooth collaborative operation of all enterprise business processes.

[0036] Status output unit This unit is configured to collect the full-process operation status data from the global process monitoring unit, organize it into standardized global process operation status data, and output it to the process value quantification and closed-loop review module. The specific implementation is as follows: This unit is a data aggregation and formatting component, primarily responsible for converting the raw monitoring data from the global process monitoring unit into a standardized data format that the process value quantification and closed-loop review module can directly receive and parse. Following a preset time period (e.g., every minute) or event-triggered mechanism (e.g., after each process instance is completed), this unit pulls raw operational status data from the global process monitoring unit, including full process operation data, resource usage data, risk warning data, and anomaly handling data obtained through the enterprise workflow execution engine. Subsequently, according to the pre-agreed data interface specifications (e.g., JSON Schema) with the process value quantification and closed-loop review module, this unit filters, cleans, aggregates, and formats the raw monitoring data, removing invalid data, supplementing missing key information, and organizing the data into a standardized global process operational status report. The report includes overall enterprise process operational efficiency, details of each process instance's operation, resource usage statistics, and risk handling statistics. Finally, the standardized global process operational status report is pushed to the process value quantification and closed-loop review module via a message queue or HTTP interface, providing accurate and standardized data support for process review analysis.

[0037] 5. Process Value Quantification and Closed-Loop Review Module The process value quantification and closed-loop review module includes a value indicator system management unit, a process value analysis unit, and an optimization instruction output unit.

[0038] Value Indicator System Management Unit This unit is configured to build and maintain a pre-defined multi-dimensional value quantification indicator system. This system includes efficiency indicators, cost indicators, value indicators, and compliance indicators, providing a quantitative basis for generating review results. The specific implementation is as follows: This unit provides a visual configuration management interface, allowing enterprise administrators to define and maintain a tree-structured, multi-dimensional value quantification indicator system based on their business needs and control objectives. The indicator system is divided into primary indicators, secondary indicators, and leaf node indicators, with clear hierarchy and well-defined responsibilities. Primary indicators include four main categories: efficiency indicators, cost indicators, value indicators, and compliance indicators. Efficiency indicators quantify the smoothness and speed of process operation (e.g., average node processing time, overall process cycle, and on-time completion rate); cost indicators quantify resource consumption during process operation (e.g., process manpower costs, resource occupancy costs, and anomaly handling costs); value indicators quantify the business value brought by process operation (e.g., process output benefits, customer satisfaction, and resource utilization improvement); and compliance indicators quantify the compliance of process operation (e.g., number of compliance violations and compliance check pass rate). Each leaf node indicator has a clearly defined name, calculation formula, data source, statistical period, and weight. The data sources mainly refer to the knowledge graph built by the multi-source data fusion perception engine, the data output by each module, and the full amount of process execution data stored by the enterprise workflow execution engine. The weights can be preset by the administrator according to the importance of the indicator, or they can be automatically optimized by training a model using historical data. This multi-dimensional value quantification indicator system is persistently stored in XML or YAML format, and a standardized runtime query interface is provided for other units within the module and other modules of the system to call.

[0039] Process Value Analysis Unit This unit is configured to receive global process operation status output from knowledge graphs, decision execution feedback data, and cross-process collaborative management centers. Based on a multi-dimensional value quantification indicator system, it uses data mining algorithms to perform multi-dimensional analysis on various types of data, identify low-value nodes and resource waste points in the process operation, generate a process value management report, and analyze and form a review result. The specific implementation is as follows: This unit is the core unit for process value quantification and closed-loop review. It mainly receives input data from three sources: First, the enterprise business process knowledge graph constructed by the multi-source data fusion perception engine, from which historical and real-time full data of process operation are extracted; second, the decision execution feedback data output by the process risk prediction and intelligent decision-making module, including the execution process, execution results, benefits and costs of intervention plans; and third, the standardized global process operation status report output by the cross-process collaborative management center, including the overall enterprise and the operational efficiency data of each process instance, the core data of which are traceable to the real-time operation data collected by the workflow execution engine. This unit first calls the multi-dimensional value quantification indicator system maintained by the value indicator system management unit, and calculates the specific scores of each process instance and each process node on each indicator according to the calculation formula and statistical period of each indicator, combined with the input data, to form a quantitative performance evaluation result. Subsequently, this unit utilizes standardized data mining algorithms (such as Apriori association rule mining and K-means clustering analysis) to conduct in-depth analysis of the quantitative evaluation results and input data. This analysis uncovers potential problems and optimization opportunities in the process. For example, association rule mining reveals the underlying rule that "when two nodes are handled by the same person, the total process time increases significantly." Clustering analysis identifies groups of "high-cost, low-value" process patterns, clearly identifying low-value nodes, resource waste points, process bottlenecks, and the root causes of anomalies. Based on the in-depth analysis results, this unit generates a standardized process value management report and simultaneously analyzes and forms a structured review result. This review result not only includes process performance diagnosis, problem analysis, and optimization directions, but more importantly, it contains targeted optimization suggestions, providing a basis for generating subsequent optimization instructions.

[0040] Optimize instruction output unit This unit is configured to extract basic process rule optimization instructions and parameter optimization instructions for the prediction model and intelligent decision-making model from the review results. It then outputs the basic process rule optimization instructions to the process dynamic self-evolution modeling module and the parameter optimization instructions to the process risk prediction and intelligent decision-making module. The specific implementation is as follows: This unit receives the debriefing results output by the process value analysis unit. It first analyzes the debriefing results, extracting two types of standardized optimization instructions: one type is basic process rule optimization instructions, primarily targeting the process dynamic self-evolution modeling module. These include specific instructions such as scenario profiling optimization, basic process rule template parameter adjustment, and rule combination logic optimization (e.g., "simplify approval node rules in a certain scenario" or "adjust the regular time limit parameter for a certain node to 1 working day"). The other type is model parameter optimization instructions, primarily targeting the process risk prediction and intelligent decision-making module. These include specific instructions such as adjusting the warning threshold of the risk prediction model, parameter fine-tuning direction, optimizing the cost-benefit analysis strategy of the intelligent decision-making model, and adjusting the expert rule base matching logic (e.g., "adjust the warning threshold for timeout risk from 70% to 65%" or "optimize the benefit quantification logic of the DRL decision network"). All optimization instructions are encapsulated in a structured format, clearly defining the instruction type, adjustment object, adjustment content, adjustment scope, and execution requirements to ensure the receiving module can accurately parse and execute them. Subsequently, this unit uses the system's preset message interface to precisely send basic process rule optimization instructions to the self-learning optimization unit of the process dynamic self-evolution modeling module, and precisely sends model parameter optimization instructions to the model iteration unit of the process risk prediction and intelligent decision-making module. Simultaneously, this unit records the delivery status, receipt status, and execution progress of each optimization instruction in real time, forming a complete optimization closed-loop log. This facilitates subsequent traceability and verification, ensuring that optimization instructions are effectively implemented and achieving closed-loop iteration of process control.

[0041] like Figure 2 The specific execution steps of the intelligent management and control method for enterprise business processes based on the above system are as follows: S1. Initial Profile Library Construction: This is executed by the scenario profile construction unit of the process dynamic self-evolution modeling module. It uses natural language processing technology to analyze enterprise business systems and historical business data, builds and maintains a business scenario profile library, and associates each scenario profile with the corresponding basic process rules, thus completing the profile library construction in the system initialization phase.

[0042] S2. Multi-source data fusion and graph update: The multi-source data acquisition unit and data fusion processing unit of the multi-source data fusion perception engine work together to collect heterogeneous data from multiple sources inside and outside the enterprise, clean, associate and fuse the heterogeneous data, build and continuously update the enterprise business process knowledge graph, and provide data support for the operation of each module.

[0043] S3. Dynamic self-evolution of process generation: Executed by the dynamic process generation unit of the dynamic self-evolution modeling module, it accesses the knowledge graph, and based on the knowledge graph and business scenario profile library, responds to changes in enterprise business scenarios. It dynamically combines and optimizes basic process rules through feature matching and fuzzy inference algorithms to generate a business process model that adapts to the current business scenario. It is then published to the enterprise's workflow execution engine through a standard API interface, and its execution status data is available for access and calls by relevant modules.

[0044] S4. Process Risk Prediction and Decision Execution: This module is executed collaboratively by the process risk prediction model and the intelligent decision model. It accesses the knowledge graph and business process model, identifies potential process risks through the prediction model, generates and executes the optimal intervention plan through the intelligent decision model, and outputs decision execution feedback data.

[0045] S5. Cross-process collaborative management and status feedback: The global process monitoring unit, collaborative scheduling engine and status output unit of the cross-process collaborative management hub work together to access the knowledge graph and business process model, and perform unified monitoring and visualization of all business processes of the enterprise. Through the collaborative scheduling algorithm, it realizes global resource scheduling and process flow path adjustment across processes, and outputs the global process running status to the process value quantification and closed-loop review module.

[0046] S6. Process Value Review and Instruction Output: The value indicator system management unit, process value analysis unit, and optimization instruction output unit of the process value quantification and closed-loop review module work together to execute the review results based on the preset multi-dimensional value quantification indicator system. It receives knowledge graph, decision execution feedback data, and global process operation status, analyzes and forms a review result that includes basic process rules, prediction models, and intelligent decision model parameter optimization instructions. The optimization instructions are then output to the process dynamic self-evolution modeling module and the process risk prediction and intelligent decision module, respectively.

[0047] S7. Single Closed-Loop Iteration: This is executed collaboratively by the self-learning optimization unit of the process dynamic self-evolution modeling module and the model iteration unit of the process risk prediction and intelligent decision-making module. Based on the received basic process rule optimization instructions, the basic process rules and business scenario profile library are updated. Based on the received parameter optimization instructions, the parameters and operating strategies of the prediction model and intelligent decision-making model are iteratively optimized to complete the single closed-loop iteration of enterprise business process control.

[0048] S8. Continuous Optimization of the Process Control System: This is executed periodically, initiated by the process value quantification and closed-loop review module. First, it collects full feedback data from all business processes, optimization suggestions from business departments, and logs of control indicator usage to construct a process optimization input dataset. Then, using a multi-objective optimization algorithm (such as NSGA-II), with the optimization goals of reducing process operation risks, shortening the overall process cycle, and improving the stability of control indicators and the business value of the process, it globally optimizes existing basic process rules, expert rule bases, and multi-dimensional value quantification indicator systems, generating corresponding optimization adjustment plans. Finally, the cross-departmental management organization system reviews and confirms the optimization adjustment plans based on preset review rules. Upon approval, the basic process rules, expert rule base, and indicator system are dynamically adjusted, and the multi-source data fusion perception engine is simultaneously triggered to update the enterprise business process knowledge graph and its entity association analysis logic, achieving continuous iterative upgrades of the enterprise business process control system.

[0049] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. An intelligent management and control system for enterprise business processes, characterized in that, include: Multi-source data fusion perception engine, process dynamic self-evolution modeling module, process risk prediction and intelligent decision-making module, cross-process collaborative management and control center, and process value quantification and closed-loop review module; The multi-source data fusion perception engine is configured to collect multi-source heterogeneous data from within and outside the enterprise, and to fuse and process the multi-source heterogeneous data in order to build and continuously update the enterprise's business process knowledge graph. The process dynamic self-evolution modeling module is configured to: in the initial stage, construct and maintain a business scenario profile library through natural language processing technology, where each scenario profile is associated with basic process rules; in the running stage, access the knowledge graph and, based on the knowledge graph and the business scenario profile library, respond to changes in the enterprise's business scenario by dynamically combining and optimizing the basic process rules through feature matching and fuzzy reasoning algorithms to generate a business process model adapted to the current business scenario. The process risk prediction and intelligent decision-making module is configured to access the knowledge graph and business process model, identify potential risks through the prediction model, generate and execute the optimal intervention plan through the intelligent decision-making model, and output decision execution feedback data. The cross-process collaborative management and control hub is configured to access the knowledge graph and business process model, perform unified monitoring and visualization of all business processes of the enterprise, realize global resource scheduling and process flow path adjustment across processes through collaborative scheduling algorithms, and output the global process running status to the process value quantification and closed-loop review module. The process value quantification and closed-loop review module is configured to receive the knowledge graph, decision execution feedback data and global process operation status based on a preset multi-dimensional value quantification index system. It analyzes and forms a review result that includes basic process rules, prediction model and intelligent decision model parameter optimization instructions. The optimization instructions are output to the process dynamic self-evolution modeling module to optimize the basic process rules, and output to the process risk prediction and intelligent decision module to optimize the model parameters.

2. The enterprise business process intelligent management and control system as described in claim 1, characterized in that, The process dynamic self-evolution modeling module includes: The scenario profile building unit is configured to perform the operations of the initial stage, analyze enterprise business systems and historical business data through natural language processing technology, build and maintain a standardized business scenario profile library, and associate each scenario profile with the corresponding basic process rules. The dynamic process generation unit is configured to perform the core operations of the running phase, access the knowledge graph, and respond to changes in enterprise business scenarios based on the knowledge graph and business scenario profile library. It completes scenario profile matching through feature extraction, and dynamically combines, resolves conflicts and optimizes parameters of the associated basic process rules based on fuzzy reasoning algorithm to generate an adapted current business process instance. The self-learning optimization unit is configured to receive feedback data such as node execution efficiency and exception rate during the process execution, and iteratively update the mapping relationship between scenario profiles and process rules through reinforcement learning algorithms to optimize the business scenario profile library and basic process rules.

3. The enterprise business process intelligent management and control system as described in claim 1, characterized in that, The multi-source data fusion sensing engine includes: The multi-source data acquisition unit is configured to collect heterogeneous data from multiple sources inside and outside the enterprise. Specifically, it collects process execution data, business value data, and real-time process operation data by connecting to the enterprise's internal business systems through interfaces, collects external market, supply chain, and policy data by web crawling or API connection, and collects human-computer interaction behavior data by behavior recognition technology. The data fusion processing unit is configured to receive multi-source heterogeneous data collected by the multi-source data acquisition unit, clean, associate and fuse the multi-source heterogeneous data, and construct and continuously update an enterprise business process knowledge graph based on the fusion results. The knowledge graph is used to represent process nodes, participating entities, resources, control indicators, full data of process operation and the relationship between various types of data.

4. The enterprise business process intelligent management and control system as described in claim 1, characterized in that, The process risk prediction and intelligent decision-making module includes: The process risk prediction model is configured to access the knowledge graph and business process model, and based on time series prediction algorithms and Bayesian networks, perform probabilistic prediction and graded early warning of node timeout risk, resource shortage risk, approval backlog risk, compliance risk and value loss risk, and output risk prediction results. The intelligent decision-making model is configured to receive risk prediction results, and based on deep reinforcement learning algorithms and expert rule bases, automatically generate multiple intervention plans for predicted risks or abnormal situations in process operation. It then selects and executes the optimal intervention plan through cost-benefit analysis algorithms and outputs decision execution feedback data. The model iteration unit is configured to receive model parameter optimization instructions output by the process value quantification and closed-loop review module, as well as decision execution feedback data, and to iteratively optimize the model parameters and operating strategies of the process risk prediction model and the intelligent decision model.

5. The enterprise business process intelligent management and control system as described in claim 1, characterized in that, The cross-process collaborative management and control center includes: The global process monitoring unit is configured to access the knowledge graph and business process model to uniformly monitor and visualize the running status, resource usage, and risk warning information of all business processes of the enterprise. The collaborative scheduling engine is configured to receive monitoring data from the global process monitoring unit, identify the relationships between various business processes based on the knowledge graph, and dynamically adjust the global resource allocation and process flow path through the collaborative scheduling algorithm when it is detected that the running status of a certain process has a negative impact on the associated processes. The status output unit is configured to collect the full-process operation status data of the global process monitoring unit, organize it into a standardized global process operation status, and output it to the process value quantification and closed-loop review module.

6. The enterprise business process intelligent management and control system as described in claim 1, characterized in that, The process value quantification and closed-loop review module includes: The value indicator system management unit is configured to build and maintain a preset multi-dimensional value quantification indicator system, which includes efficiency indicators, cost indicators, value indicators and compliance indicators, providing a quantitative basis for the generation of review results; The process value analysis unit is configured to receive the knowledge graph, decision execution feedback data, and the global process operation status output by the cross-process collaborative management and control center. Based on a multi-dimensional value quantification index system, it performs multi-dimensional analysis on various types of data through data mining algorithms to identify low-value nodes and resource waste points in the process operation, generate a process value management and control report, and analyze and form a review result. The optimization instruction output unit is configured to extract basic process rule optimization instructions and parameter optimization instructions for the prediction model and intelligent decision model from the review results, output the basic process rule optimization instructions to the process dynamic self-evolution modeling module, and output the parameter optimization instructions to the process risk prediction and intelligent decision module.

7. A method for intelligent management and control of enterprise business processes based on the system described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Initial Profile Library Construction: Using natural language processing technology, analyze the enterprise's business systems and historical business data to build and maintain a business scenario profile library. Each scenario profile is associated with the corresponding basic process rules. S2. Multi-source data fusion and graph update: Collect multi-source heterogeneous data from within and outside the enterprise, clean, associate and fuse the multi-source heterogeneous data, and build and continuously update the enterprise business process knowledge graph. S3. Dynamic self-evolution of process generation: Access the knowledge graph, and based on the knowledge graph and business scenario profile library, respond to changes in enterprise business scenarios, dynamically combine and optimize basic process rules through feature matching and fuzzy reasoning algorithms to generate a business process model that adapts to the current business scenario. S4. Process Risk Prediction and Decision Execution: Access the knowledge graph and business process model, identify potential process risks through the prediction model, generate and execute the optimal intervention plan through the intelligent decision model, and output decision execution feedback data. S5. Cross-process collaborative management and status feedback: Access the knowledge graph and business process model to uniformly monitor and visualize all business processes of the enterprise, realize global resource scheduling and process flow path adjustment across processes through collaborative scheduling algorithms, and output the global process running status to the process value quantification and closed-loop review module. S6. Process Value Review and Instruction Output: Based on the preset multi-dimensional value quantification index system, the system receives the knowledge graph, decision execution feedback data and global process operation status, analyzes and forms a review result including basic process rules, prediction model and intelligent decision model parameter optimization instructions, and outputs the optimization instructions to the process dynamic self-evolution modeling module and the process risk prediction and intelligent decision module respectively. S7. Single Closed-Loop Iteration: Update the basic process rules and business scenario profile library based on the received basic process rule optimization instructions, and iteratively optimize the parameters and operation strategies of the prediction model and intelligent decision-making model based on the received parameter optimization instructions, thus completing a single closed-loop iteration of enterprise business process control.

8. The intelligent management and control method for enterprise business processes as described in claim 7, characterized in that, In S3, the dynamic self-evolution of the process specifically includes: Access the knowledge graph to extract feature data of the current business activity in real time, and perform feature matching with the scene profiles in the business scenario profile library to determine the target scene profile. Based on the matched target scenario profile, the applicability assessment, conflict resolution and parameter optimization of multiple basic process rules associated with it are carried out through fuzzy inference algorithm, and process instances adapted to the current business scenario are dynamically combined to generate process instances. The generated business process model will be synchronized to the process risk prediction and intelligent decision-making module and the cross-process collaborative management and control center for them to access and call.

9. The intelligent management and control method for enterprise business processes as described in claim 7, characterized in that, In S4, process risk prediction and decision execution specifically include: Access the knowledge graph and business process model, extract real-time process operation data and node relationships, and calculate the probability of occurrence and impact level of various risks based on LSTM time series prediction algorithm and Bayesian network. When the risk level exceeds the preset threshold, the intelligent decision-making model is triggered, and multiple candidate intervention schemes are generated by combining the expert rule base, including process rerouting, resource reallocation, and temporary adjustment of approval authority. The expected performance and implementation cost of each candidate intervention plan are quantitatively evaluated using a cost-benefit analysis model. The plan with the highest overall benefit is automatically selected and executed. The decision-making process and results are recorded simultaneously, and decision-making feedback data is generated and output.

10. The intelligent management and control method for enterprise business processes as described in claim 7, characterized in that, Also includes: S8. Continuous optimization of the process control system: Regularly collect full feedback data on the execution of various business processes, optimization suggestions from business departments, and logs of control indicators to build a process optimization input dataset; A multi-objective optimization algorithm is adopted to optimize the existing basic process rules, expert rule base and multi-dimensional value quantification indicator system by reducing process operation risks, shortening the overall process cycle, improving the stability of control indicators and process business value. The algorithm generates corresponding optimization and adjustment schemes. The cross-departmental management organization reviews and confirms the optimization and adjustment plan based on preset review rules. After the review is approved, the basic process rules, expert rule base and indicator system are dynamically adjusted, and the enterprise business process knowledge graph and its entity association analysis logic are updated simultaneously to realize the continuous iterative upgrade of the enterprise business process control system.

Images