Process engine automatic configuration method and device and medium

Abstract

The application provides a process engine automatic configuration method and device and medium, and relates to the technical field of computers. The method comprises the following steps: acquiring a value function of a process engine processing a business, wherein the value function represents a first value of a single optional node processing a single link of the business according to an optional action of the single optional node, and a second value of multiple optional nodes jointly and completely processing the business according to respective optional actions of the multiple optional nodes; and solving the value function to obtain an optimal strategy of the process engine processing the business, wherein the optimal strategy represents a single optimal node and an optimal action of the single optimal node that can obtain an optimal first value in processing the single link of the business, and multiple optimal nodes and optimal actions of the multiple optimal nodes that can obtain an optimal second value in jointly and completely processing the business. The application constructs a value function of the process engine processing the business, and solves the value function to obtain an optimal strategy of the process engine processing the business.

Description

Technical Field

[0001] This application relates at least to the field of computer technology, and in particular to a method, apparatus and medium for automatic configuration of a process engine. Background Technology

[0002] Currently, the configuration of the business process engine is mainly done manually. As business processes become more complex, manual configuration becomes inefficient and fails to adequately assess the value of individual process nodes and the configured chain of nodes, potentially leading to unsatisfactory process execution. Summary of the Invention

[0003] To address the aforementioned shortcomings, this application provides a method, apparatus, and medium for automatic configuration of a process engine, thereby solving the following technical problem: how to evaluate the value of node and action processing business in a process engine in order to select the optimal node and optimal action configuration for the process engine to process the business.

[0004] Firstly, this application provides a method for automatic configuration of a process engine, the method comprising:

[0005] Obtain the value function of the process engine for processing business. The value function represents the first value of a single optional node processing a single step of the business according to its own optional actions, and the second value of multiple optional nodes jointly processing the business completely according to their respective optional actions.

[0006] The optimal solution is obtained by solving the value function to obtain the optimal strategy for the configuration process engine to process business. The optimal strategy represents the single optimal node and its optimal action that can obtain the optimal first value in a single link of processing business in the process engine, and the multiple optimal nodes and their respective optimal actions that can obtain the optimal second value in the complete processing of business.

[0007] Furthermore, the value function of the process engine in handling business is obtained, specifically including:

[0008] Obtain the various stages of a business process to be configured, the input / output conditions of each stage, and the several optional algorithms for processing each stage;

[0009] For each stage, an optional node is matched in the process engine, and each optional node has at least one optional algorithm to process the corresponding stage;

[0010] Based on several input / output conditions of each stage, the optional connection relationships of each optional node are obtained. The optional algorithm and optional connection relationships constitute the optional actions of each optional node.

[0011] Using the node status and optional actions of each optional node as input, and the single-node reward and cumulative reward of each optional node in handling a certain business as output, the value function of the process engine for handling the business is constructed respectively.

[0012] Furthermore, taking the node status and optional actions of each optional node as input, and the single-node reward and cumulative reward of each optional node in handling a certain business as output, the value function for the process engine to handle the business is constructed, specifically including:

[0013] The experiment obtains the single-node return of each optional node in various node states, based on different connection relationships and processing algorithms, for processing a certain business process. The single-node return is obtained based on the processing cost and processing time of the corresponding process of a certain business process.

[0014] Based on the reports from each individual node, construct the single-node value function for each optional node in handling a certain business based on each state and action. Based on the connection relationships of each optional node and the single-node value function, construct the cumulative value function for each optional node in handling a certain business.

[0015] Furthermore, based on the returns from each individual node, a single-node value function is constructed for each optional node to handle a specific business process according to each state and action. Based on the connection relationships of each optional node and the single-node value function, a cumulative value function for each optional node to handle a specific business process is constructed, specifically including:

[0016] The function to obtain the single-node state-action value of each optional node is as follows: In the formula, S k A represents the node state of the optional node k. k This represents a node action consisting of the upper connectivity of an optional node k and an optional algorithm, where i represents the number of trials, 0 ≤ γ ≤ 1 represents the reward discount factor, and R... i Let represent the reward obtained by the optional node k in the i-th trial, and s and a represent the specific values ​​of the node state and node action, respectively. This represents the average of the rewards obtained from multiple trials;

[0017] The function to obtain the single-node state value of each optional node: This represents the average reward for an optional node k performing j node actions in state s;

[0018] The function to obtain the accumulated state-action value of each optional node: Q π (S k A k )=[Q(S k A k )+Q π (Sk-1 A k-1 )|S k =s,A k =a], where π represents the optional strategy by which the process engine fully processes a certain business through optional nodes, several connection relationships and several optional algorithms, and k-1 and k represent the two adjacent optional nodes in the optional strategy respectively;

[0019] Obtain the cumulative state value function for each optional node: V π (S k )=[V(S k )+V π (S k-1 )|S k =s).

[0020] Furthermore, the value function is optimized to obtain the optimal strategy for the configuration process engine to handle business, specifically including:

[0021] The gene fragment of each optional node in the coding process engine, each gene fragment includes the node state and optional actions of the corresponding optional node;

[0022] Construct a set of optional strategies for the process engine to handle a certain business. The set of optional strategies includes all possible combinations of optional nodes and optional actions in the process engine that can collectively and completely handle a certain business.

[0023] With the set of optional strategies as constraints, a process engine is constructed to fully process a certain business's chromosome individual, and a genetic algorithm is used to solve for the chromosome individual with the optimal value function.

[0024] The optimal strategy for the chromosome individual with the best value function is obtained and used as the optimal strategy for the configuration process engine to handle business.

[0025] Furthermore, using the set of optional strategies as constraints, a process engine is constructed to fully process a certain business process (chromosome-like individual). A genetic algorithm is then used to solve for the chromosome-like individual with the optimal value function, specifically including:

[0026] Obtain the population used for this round of iteration of the genetic algorithm. The population is either a randomly generated initial population or a population obtained through iterative optimization. The population includes multiple chromosome individuals, and each chromosome individual corresponds to an optional strategy.

[0027] Calculate the first value of the single-node value function for each optional node of each chromosome individual in the population of this iteration. If the first value is less than the first threshold, perform crossover and mutation operations on the corresponding optional node based on the optional strategy set to obtain the population of the next iteration.

[0028] When the termination condition of the genetic algorithm iteration is met, calculate the second value of the cumulative value function of the last optional node of each chromosome individual in the population obtained in the last iteration, and obtain the chromosome individual with the largest second value.

[0029] Furthermore, the first value of the single-node value function for each optional node of each chromosome individual in the population during this iteration is calculated. If the first value is less than a first threshold, crossover and mutation operations are performed on the corresponding optional node based on the optional strategy set, specifically including:

[0030] Calculate the third value of the single-node state-action value function for each optional node of each chromosome individual in the population during this iteration. If the third value is less than the third threshold, perform optional action mutation on the gene fragment of the corresponding optional node based on the optional strategy set.

[0031] Calculate the fourth value of the single-node state value function for each optional node of each chromosome individual in the population during this iteration. If the fourth value is less than the fourth threshold, replace the corresponding optional node with a new optional node based on the optional policy set.

[0032] Furthermore, the second value of the cumulative value function of the last optional node of each chromosome individual in the population obtained in the last iteration is calculated, and the chromosome individual with the largest second value is obtained, specifically including:

[0033] Calculate the fifth value of the cumulative state value function of the last optional node of each chromosome individual in the population obtained in the last iteration, and obtain several last optional nodes whose fifth value is greater than the fifth threshold;

[0034] Calculate the sixth value of the cumulative state-action value function for several last optional nodes whose fifth value is greater than the fifth threshold, and obtain the chromosome individual corresponding to the last optional node with the largest sixth value.

[0035] Secondly, this application provides an automatic configuration device for a process engine, the device comprising:

[0036] The value function module is used to obtain the value function of the process engine in processing business. The value function represents the first value of a single optional node processing a single link of the business according to its own optional actions, and the second value of multiple optional nodes jointly processing the business completely according to their respective optional actions in the process engine.

[0037] The optimization solution module, connected to the value function module, is used to optimize the solution of the value function in order to obtain the optimal strategy for the configuration process engine to process business. The optimal strategy represents the single optimal node and its optimal action in the process engine that can obtain the optimal first value in a single link of processing business, and the multiple optimal nodes and their respective optimal actions that can obtain the optimal second value in the complete processing of business.

[0038] Thirdly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the process engine automatic configuration method as described above.

[0039] This application provides a method, apparatus, and medium for automatic configuration of a process engine. Based on the individual steps of the optional nodes in the process engine that process business according to optional actions, and the complete steps of jointly processing business, a value function for the process engine to process business is constructed. The value function is then optimized to obtain the optimal strategy for configuring the process engine to process business. The optimal strategy includes the single node with the highest value and the entire node chain. Finally, the optimal automatic configuration of the process engine to process business can be achieved based on the optimal strategy. Attached Figure Description

[0040] Figure 1 This is a flowchart of an automatic configuration method for a process engine according to an embodiment of this application;

[0041] Figure 2 This is a schematic diagram of the structure of an automatic configuration device for a process engine according to an embodiment of this application;

[0042] Figure 3 This is a flowchart of another automatic configuration method for a process engine according to an embodiment of this application. Detailed Implementation

[0043] To enable those skilled in the art to better understand the technical solution of this application, the embodiments of this application will be further described in detail below with reference to the accompanying drawings.

[0044] It is understood that the specific embodiments and accompanying drawings described herein are merely for explaining this application and are not intended to limit this application.

[0045] It is understood that, without conflict, the various embodiments and features in the embodiments of this application can be combined with each other.

[0046] It is understood that, for ease of description, only the parts relevant to this application are shown in the accompanying drawings, while parts unrelated to this application are not shown in the drawings.

[0047] It is understood that each module or unit involved in the embodiments of this application may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple modules or units may be integrated into one entity structure.

[0048] It is understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of this application may occur in a different order than that marked in the accompanying drawings.

[0049] It is understood that the flowcharts and block diagrams of this application illustrate the possible architecture, functions, and operations of systems, apparatuses, devices, and methods according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, unit, program segment, or code, containing executable instructions for implementing the specified function. Furthermore, each block or combination of blocks in the block diagrams and flowcharts may be implemented using a hardware-based device to implement the specified function, or using a combination of hardware and computer instructions.

[0050] It is understood that the modules and units involved in the embodiments of this application can be implemented by software or by hardware. For example, the modules and units can be located in the processor.

[0051] Example 1:

[0052] like Figure 1 As shown, this application provides a method for automatic configuration of a process engine, the method comprising:

[0053] S1. Obtain the value function of the process engine for processing business. The value function represents the first value of a single optional node processing a single link of the business according to its own optional actions, and the second value of multiple optional nodes jointly processing the business completely according to their respective optional actions in the process engine.

[0054] S2. Optimize the value function to obtain the optimal strategy for the configuration process engine to process business. The optimal strategy represents the single optimal node and its optimal action in the process engine that can obtain the optimal first value in a single link of processing business, and the multiple optimal nodes and their respective optimal actions that can obtain the optimal second value in the complete processing of business.

[0055] In this embodiment, the method constructs a value function for the process engine's business processing based on the individual steps of the optional nodes in the process engine and the complete steps of the business processing together. It then optimizes and solves this value function to obtain the optimal strategy for configuring the process engine's business processing. The optimal strategy includes the single node with the highest value and the entire node chain. Finally, based on the optimal strategy, the process engine's business processing can be automatically and optimally configured. Figure 1 The method shown is applicable to, for example, Figure 2 The apparatus shown.

[0056] Specifically, this embodiment provides a method for automating the configuration of a process engine based on maximizing process value. This method can be applied to relevant business systems in mobile communications, and mainly uses the process engine to automate complex business processes and improve system collaboration efficiency.

[0057] Common process engines in the communications field include: Activiti (a business process management (BPM) and workflow system) / Flowable (a lightweight business process engine written in Java), suitable for scenarios such as work order processing and network resource scheduling in the communications industry; Camunda (an open-source workflow engine and business process management platform), often used in scenarios requiring dynamic resource adjustment, such as 5G (5th Generation Mobile Communication Technology) network slicing management and service orchestration; and self-developed engines, which are often developed by large communications companies based on the BPMN (Business Process Modeling Notation) specification to adapt to proprietary protocols and business logic.

[0058] Typical application process examples in the communications field include: Network service activation process, which includes user broadband application → resource verification → port configuration → device activation → billing system synchronization. The engine's role is to orchestrate cross-system operation nodes through a visual process designer and automatically trigger OSS (Operation Support Systems) interface calls; Automated fault handling process, which includes network alarm triggering → root cause analysis → automatic repair (such as switching to backup links) → fault report generation. The engine's role is to integrate AI (Artificial Intelligence) analysis modules to automate fault location and handling strategies; 5G network slice management, which includes slice creation request → resource allocation → QoS (Quality of Service) policy configuration → end-to-end testing → service launch. The engine's role is to coordinate NFV (Network Functions Virtualization) components to ensure low-latency, high-reliability dynamic resource allocation; and user complaint handling process, which includes complaint acceptance → work order dispatch → cross-departmental collaborative processing → result feedback → satisfaction follow-up. The engine's role is to integrate CRM (Customer Relationship Management). The Management system enables automatic determination of work order priorities and dynamic adjustment of approval paths.

[0059] The workflow engine is the core of a workflow management system, responsible for defining, creating, executing, and managing workflows. The workflow engine interfaces with the data required for specific application scenarios. During implementation, requirements and development personnel may focus only on the design of workflow configurations, workflow nodes, and workflow forms. However, with the continuous emergence and iteration of various systems within operators, traditional workflow engines in these systems face new challenges regarding workflow status and value. For example, failure to evaluate workflow value during configuration and failure to automate workflow design based on the evaluation results may result in a large number of low-value workflow engine nodes, while also posing a risk of missing high-value workflow engine nodes.

[0060] To address the aforementioned issues, this embodiment proposes a method for automating the configuration of a process engine based on maximizing process value. By using an algorithm to solve for the maximum value of the process engine in process states and actions, the method ensures the maximization of process value in the process system, guarantees optimal process design in complex processes, and automates the configuration of the process engine based on the process with maximized value.

[0061] More specifically, an engine is a highly abstracted and encapsulated business of a certain type, detached from specific business scenarios. The value function represents the reward that can be obtained by selecting optional nodes in the process engine to process the business according to optional actions. The optimal strategy represents the optimal node and optimal action in the process engine that can obtain the best reward for processing the business. Actions include the node's own processing actions in the business process and the connection relationships between nodes. The value function is solved based on an improved genetic algorithm to configure an engine device that maximizes the value of the process. By solving the maximum value of the process engine at the levels of process node state, process node action, and process node connection through the genetic algorithm, it is possible to ensure that the process engine selects the node with the optimal state, adopts the optimal action, and forms the optimal node connection, thereby ensuring the maximization of the value of the process in the process system and ensuring the optimal process design in complex processes.

[0062] like Figure 3As shown, the method provided in this embodiment includes: constructing a process state and action model in the process engine by combining reinforcement learning and decision theory; obtaining the highest value action based on an improved ant colony algorithm to select the optimal action; and finally converging to the optimal strategy through reinforcement learning value iteration to complete the automated configuration of the process engine. In the process engine, reinforcement learning and decision theory are used to construct value functions for relevant states and actions. The potential future benefits of the current state are evaluated to guide the decision-making process and select the optimal action path. An improved genetic algorithm is used to solve for the maximum value of the process engine in process states and actions, ensuring the maximization of the process engine's value at process nodes and in process chains, guaranteeing optimal process design in complex processes. Through reinforcement learning value iteration, the optimal strategy is finally converged, and the automated configuration of the process engine is completed.

[0063] In one implementation, S1, obtaining the value function for the business processed by the process engine specifically includes:

[0064] Obtain the various stages of a business process to be configured, the input / output conditions of each stage, and the several optional algorithms for processing each stage;

[0065] For each stage, an optional node is matched in the process engine, and each optional node has at least one optional algorithm to process the corresponding stage;

[0066] Based on several input / output conditions of each stage, the optional connection relationships of each optional node are obtained. The optional algorithm and optional connection relationships constitute the optional actions of each optional node.

[0067] Using the node status and optional actions of each optional node as input, and the single-node reward and cumulative reward of each optional node in handling a certain business as output, the value function of the process engine for handling the business is constructed respectively.

[0068] In this embodiment, constructing the value function requires certain preparatory steps: extracting specific process steps, execution conditions, tasks, decision points, and other information from existing business process definitions. The process of extracting business processes and actions typically requires understanding the following aspects: the structure and storage method of the process engine (e.g., BPMN or XML format). The choice of different storage structures needs to balance standardization with business specificity, achieving the optimal solution between process compliance auditing (strong standardization) and rapid iteration (high flexibility); obtaining each step (task, gateway, event, etc.) in the process definition. The gateway's role in process steps is to control process flow and manage execution status; the specific actions and execution conditions of task nodes; and the APIs (Application Programming Interfaces) or management interfaces provided by the process engine. By understanding these elements and combining them with appropriate tools or APIs, the specific steps and actions of the business process can be extracted, thereby enabling mathematical modeling of the process engine nodes.

[0069] In a process engine, state and action models guide decision-making. Within the framework of the process engine, each state of the process can be viewed as an "environmental state." In a given state, an appropriate action is chosen to propel the process towards a better state. Each state receives a reward based on certain rules (e.g., task completion time, resource consumption). The goal is to select the optimal action based on the current state and historical data, ensuring optimal overall performance (i.e., maximizing long-term returns). State: Each state in the process engine can include the current process node's status, execution progress, resource status, etc. Action: An action is the operation the engine can perform in a given state, such as advancing to the next process node, pausing the task, or executing an optimization algorithm. Reward: The reward measures the benefit of performing an action. For example, a reward function can be defined based on factors such as process efficiency, task completion time, and cost. Based on the state and action models in the process engine, the engine can calculate the expected reward for each possible action based on the current state, thereby selecting the optimal action.

[0070] In one implementation, the value function for processing business by the process engine is constructed by taking the node state and optional actions of each optional node as input, and the single-node reward and cumulative reward of each optional node in processing a certain business as output. Specifically, this includes:

[0071] The experiment obtains the single-node return of each optional node in various node states, based on different connection relationships and processing algorithms, for processing a certain business process. The single-node return is obtained based on the processing cost and processing time of the corresponding process of a certain business process.

[0072] Based on the reports from each individual node, construct the single-node value function for each optional node in handling a certain business based on each state and action. Based on the connection relationships of each optional node and the single-node value function, construct the cumulative value function for each optional node in handling a certain business.

[0073] In this embodiment, for example, the existing system processes of the operator are designed by considering steps such as process organizational structure design, form configuration, data binding, and process interface design to design the system's process engine. This design can basically meet the operator's needs for the process engine in specific daily work. However, with the increase in the number of internal systems and the iteration of system processes, most processes have certain deficiencies in the rationality of value, and have not fully selected high-value nodes. Therefore, this embodiment uses reinforcement learning methods and decision theory in the process engine to construct value functions for the relevant states and actions of the process engine. The current state is used to evaluate the potential future benefits, thereby guiding the decision-making process and selecting the optimal action path. This includes: comprehensively analyzing the process engine in the operator's existing system, considering its performance, scalability, ease of use, and maintainability from multiple dimensions; mathematically modeling the state of the process engine based on the comprehensive consideration results of the process engine, according to reinforcement learning and decision-making processes; mathematically modeling the actions of the process engine based on the comprehensive consideration results of the process engine, according to reinforcement learning and decision-making processes; and determining the optimal action (the action with the highest value) based on the current state and historical data, so that the execution of the process engine can achieve optimal overall performance.

[0074] Specific considerations may include: performance metrics, including response time and throughput. Key indicators such as process instance startup latency must be controlled within milliseconds (<200ms), single cluster throughput must support tens of thousands of TPS (process instances / second), and core network scenarios require over 5,000 TPS. This includes in-memory computing optimization (e.g., pre-compiled process definitions), asynchronous task queues, and leveraging the Netty event-driven model to improve concurrent processing capabilities. Resource consumption control is also considered, such as memory / CPU (Central Processing Unit) usage: single-node memory usage must be <3GB, CPU utilization <50%, suitable for edge computing scenarios (e.g., lightweight deployment of 5G base stations), and hardware compatibility: support for GPUs (Graphics Processing Units). ProcessingUnit (PSU) acceleration or FPGA (Field-Programmable Gate Array) hardware offloading reduces the load pressure on the main processing unit from high-density computing tasks; scalability considerations include architectural expansion capabilities, such as horizontal expansion capabilities: decoupling the process orchestration layer and execution nodes through a microservice architecture, supporting the dynamic addition of computing nodes to cope with sudden traffic surges; vertical expansion capabilities: providing pluggable interfaces for rapid integration of proprietary protocols (such as NETCONF, TR-069) and third-party rule engines (such as Drools); multi-scenario adaptation, such as network slice management: supporting dynamic allocation of 5G slice resources and QoS policy adjustment to meet the needs of URLLC (ultra-reliable and low-latency communication) scenarios with end-to-end latency <10ms; cross-domain collaboration: encapsulating the differences between network management systems in different regions through standardized APIs, reducing the time for cross-provincial fault handling from hours to minutes; usability considerations include improved development efficiency, such as visual design: integration with BPMN 2.A zero-code modeler supports drag-and-drop workflow design, reducing the development time for complex work order workflows from 8 hours to 2 hours. It also features a low-code interface providing RESTful APIs. The API and SDK toolkit allows third-party systems to create process instances and query status with less than 10 lines of code. It also includes intelligent assistance features such as AI-generated logic: automatically generating sub-process templates based on natural language descriptions (e.g., "base station fault self-healing"), reducing manual coding by 70%; real-time debugging: a built-in process simulator allows verification of the correctness of cross-system interface calls before deployment; maintainability considerations include an operation and maintenance monitoring system, such as full-link tracing: monitoring 20+ key indicators (e.g., process execution error rate, node time distribution) through Prometheus + Grafana to quickly locate hardware problems such as optical module failures; hot-fix capabilities: supporting non-stop process definition updates and reducing version upgrade risks through canary releases; and modular maintenance, such as component decoupling: loosely coupling the core engine with modules like logs and permission management, supporting independent upgrades (e.g., separate deployment of security patches); and automated operation and maintenance: built-in intelligent diagnostic tools reduce the average fault repair time from 6 hours (open-source solutions) to 1 hour.

[0075] In one implementation, a single-node value function is constructed for each optional node based on its state and actions in handling a specific service, according to the reports from each single node. Based on the connection relationships and single-node value functions of each optional node, a cumulative value function for each optional node handling a specific service is constructed, specifically including:

[0076] The function to obtain the single-node state-action value of each optional node is as follows: In the formula, S k A represents the node state of the optional node k. k This represents a node action consisting of the upper connectivity of an optional node k and an optional algorithm, where i represents the number of trials, 0 ≤ γ ≤ 1 represents the reward discount factor, and R... i Let represent the reward obtained by the optional node k in the i-th trial, and s and a represent the specific values ​​of the node state and node action, respectively. This represents the average of the rewards obtained from multiple trials;

[0077] The function to obtain the single-node state value of each optional node: This represents the average reward for an optional node k performing j node actions in state s;

[0078] The function to obtain the accumulated state-action value of each optional node: Q π (S k A k )=[Q(S k A k )+Qπ (S k-1 A k-1 )|S k =s,A k =a], where π represents the optional strategy by which the process engine fully processes a certain business through optional nodes, several connection relationships and several optional algorithms, and k-1 and k represent the two adjacent optional nodes in the optional strategy respectively;

[0079] Obtain the cumulative state value function for each optional node: V π (S k )=[V(S k )+V π (S k-1 )|S k =s).

[0080] In this embodiment, the value function specifically includes: a single-node state-action value function, representing the reward that each selectable node can obtain by performing a certain action processing step in its current node state, which is its value; the reward is obtained by averaging multiple trials; a single-node state value function, representing the average reward that each selectable node can obtain by performing all possible action processing steps in its current node state; a cumulative state-action value function, representing the reward that each selectable node can obtain by performing a certain action processing step in its current node state, summed with the rewards already obtained by preceding nodes in processing business based on strategy π; and a cumulative state value function, representing the average reward that each selectable node can obtain by performing all possible action processing steps in its current node state, summed with the average rewards already obtained by preceding nodes in processing business based on strategy π. The single-node function is used to select the optimal node, and the cumulative function is used to select the optimal node chain.

[0081] Another way to build a value model for the state and actions related to the process engine is as follows:

[0082] By introducing a value function, mathematical modeling is performed on optimal decision-making and state selection problems based on reinforcement learning and decision theory. With the goal of selecting the optimal action in the process engine to obtain more rewards, the definition of the value function can help the process engine evaluate the potential future benefits through the current state, thereby guiding the decision-making process and selecting the path of the process engine.

[0083] The state modeling of the process engine uses the state value function V(s) to describe the expected reward (i.e., the expected value of the reward) that can be obtained by starting from state s and executing according to a certain strategy π. The specific modeling results are as follows: Where: V π (s) represents the expected reward obtained from state s according to policy π; Rt γ is the reward obtained at time t, where t represents the time of each of the repeated trials; γ is the discount factor, representing the importance of future rewards, with a value range of 0 ≤ γ ≤ 1. The expected value represents the average reward across multiple trials. Nodes typically contain state attributes such as "inactive," "in progress," and "completed." In probability theory, the expected reward is the long-term average of the reward obtained when an event occurs across multiple trials. A policy can be expressed as a mapping table of actions.

[0084] Action value modeling in a process engine considers the value of an action 'a' in state 's'. The action value function Q(s,a) represents the expected reward after performing action 'a'. The specific modeling result is as follows: Among them: Q π (s,a) represents the expected return obtained according to strategy π after taking action a starting from state s; A0=a represents the action a taken at time 0; return refers to the long-term comprehensive benefit obtained by an individual or organization after taking an action; reward specifically refers to the external incentive given immediately for a specific behavior, which usually has a clear purpose and short-term nature.

[0085] Probability and weighted modeling: In a certain state of a process engine, after executing a process action and receiving an immediate reward, this immediate reward can be added to the value function of future subsequent states to obtain the total reward of the process state, i.e., γR in the formula. t+1 +V π (S t+1 ), where t represents the moment of each node's processing flow. Q π (s,a) is the process action value function, and π is the strategy function. Q represents the expected reward, achievable by choosing process action a in the current process state s. * (s,a) is the optimal process action value function, and there are multiple policy functions π. * indicates that the policy function π that maximizes the process action value function is selected. V π (s) is the process state value function, V π It is Q π The expectation.

[0086] In one implementation, in S2, the value function is optimized to obtain the optimal strategy for the configuration process engine to handle business, specifically including:

[0087] The gene fragment of each optional node in the coding process engine, each gene fragment includes the node state and optional actions of the corresponding optional node;

[0088] Construct a set of optional strategies for the process engine to handle a certain business. The set of optional strategies includes all possible combinations of optional nodes and optional actions in the process engine that can collectively and completely handle a certain business.

[0089] With the set of optional strategies as constraints, a process engine is constructed to fully process a certain business's chromosome individual, and a genetic algorithm is used to solve for the chromosome individual with the optimal value function.

[0090] The optimal strategy for the chromosome individual with the best value function is obtained and used as the optimal strategy for the configuration process engine to handle business.

[0091] In this embodiment, an improved genetic algorithm is used to solve for the maximum value of the process engine in the process state and process value, thereby ensuring the maximization of the process value within the process system and the optimal process design in complex processes. Specifically, this includes: the genes in the value-maximizing process engine selection algorithm are process nodes, and the process order of these nodes represents all paths required by the process, thus using ordered real number encoding; the fitness function of the improved genetic algorithm (such as an ant colony improved genetic algorithm) is determined based on the process nodes in the process engine. The fitness function guides the search by evaluating the quality of the solution. The search process influences the distribution of pheromones, thereby determining the ants' choices and behaviors (pheromone-genetic coordination mechanism, dynamic mutation strategy). Ultimately, through continuous iteration and optimization of the search path, the optimal or near-optimal solution to the problem is reached. Chromosome crossover is used to generate new individuals by combining the genetic information of parent chromosomes. Chromosome mutation is used to increase population diversity, escape local optima, restore lost superior genes, and explore unknown spaces. Mutation operations are combined with crossover operations and selection strategies, and mutation probabilities and methods are rationally selected to balance the needs of exploration and utilization, thereby achieving the best search results. The design of a population needs to balance maintaining population diversity and achieving rapid convergence. From a mathematical perspective, allowing superior individuals from the parent generation to enter the next round of competition ensures the iterative stability of the optimal solution, while eliminating inferior individuals from the offspring early accelerates the optimization process. A population update method is adopted where both superior offspring and superior parent individuals simultaneously enter the next generation. Specifically, parent individuals undergo crossover and mutation operations to obtain temporary offspring. The parent individuals and temporary offspring are compared, and individuals with higher fitness are selected as offspring to enter the next generation of competition. This population update method ensures that each crossover and mutation operation produces two better offspring. Individuals are included to ensure the convergence of the algorithm; based on the improved genetic algorithm, the process state and behavioral value are embedded into the workflow of the process engine. It is a type of business process that can be fully executed automatically. According to a series of process node rules, documents, information or tasks are transferred and executed between different executors; through reinforcement learning value iteration, it finally converges to the optimal policy and completes the automated configuration of the outflow engine. In reinforcement learning, the value function is usually estimated by a method called value iteration. This method updates the value of the state repeatedly and finally converges to the optimal policy. The automated configuration of the process engine is completed according to the finally converged optimal policy of the process engine.

[0092] In one implementation, a chromosome individual is constructed to fully process a certain business using a set of optional strategies as constraints. A genetic algorithm is then used to solve for the chromosome individual with the optimal value function, specifically including:

[0093] Obtain the population used for this round of iteration of the genetic algorithm. The population is either a randomly generated initial population or a population obtained through iterative optimization. The population includes multiple chromosome individuals, and each chromosome individual corresponds to an optional strategy.

[0094] Calculate the first value of the single-node value function for each optional node of each chromosome individual in the population of this iteration. If the first value is less than the first threshold, perform crossover and mutation operations on the corresponding optional node based on the optional strategy set to obtain the population of the next iteration.

[0095] When the termination condition of the genetic algorithm iteration is met, calculate the second value of the cumulative value function of the last optional node of each chromosome individual in the population obtained in the last iteration, and obtain the chromosome individual with the largest second value.

[0096] In this embodiment, the process engine can calculate the expected reward of each possible action based on the current state, and thus select the optimal action. The process engine can achieve this by using an improved genetic algorithm to solve for the maximum value of the process state and process actions, and selecting the action with the highest value. In some cases, the process engine needs to explore new possible actions to prevent getting trapped in local optima. Therefore, a certain degree of randomness or exploration strategy is often introduced into the strategy.

[0097] Taking the operator's complaint ticket processing flow as an example, this illustrates the application of coding in process optimization: The core process nodes of the complaint ticket include: A-Complaint acceptance (multi-channel access), B-Information integrity verification, C-Complaint classification coding (such as signal quality, tariff disputes), D-Preliminary responsibility determination, E-Technical investigation and data collection, F-Solution formulation, G-Customer feedback closed loop, H-Case archiving and analysis; Basic process constraints include: Information verification (B) must be completed before classification coding (C), responsibility determination (D) must be performed after technical investigation (E), and archiving analysis (H) is the final node, etc. An example of the correspondence between nodes and genes is nodes=["A","B","C","D","E","F","G","H"], which obtains the real number coding sequence chromosome of the node.

[0098] In the context of applying genetic algorithms to process optimization, the definitions of individuals, superior individuals, and degraded nodes need to be analyzed in conjunction with the characteristics of the process engine. An individual represents a complete process path solution, usually composed of real numbers / binary codes. For example, in a complaint handling process, an individual may be encoded as a node sequence A→B→F→C→D→E→G→H. A single gene corresponds to the execution order parameter of a process node. The criteria for judging superior and degraded individuals are shown in Table 1 below. The value of a single node is quantified by its frequency of occurrence among superior individuals. For example, a high-value node is a node with an occurrence rate ≥85% in the Top-10% of individuals, and a low-value node is a redundant node with an occurrence rate ≥60% in eliminated individuals. The node value weight Wk = number of times a superior individual passes / total number of executions × 1 / average processing time. High-weight nodes are preferentially retained during crossover mutation. Through the collaborative optimization of the node layer and the parameter layer, a balance between global search and local tuning of the process structure is achieved.

[0099] Table 1. Examples of criteria for judging superior and inferior individuals in genetic algorithms.

[0100]

[0101] In one implementation, the first value of the single-node value function for each optional node of each chromosome individual in the population during this iteration is calculated. If the first value is less than a first threshold, crossover and mutation operations are performed on the corresponding optional node based on the optional strategy set, specifically including:

[0102] Calculate the third value of the single-node state-action value function for each optional node of each chromosome individual in the population during this iteration. If the third value is less than the third threshold, perform optional action mutation on the gene fragment of the corresponding optional node based on the optional strategy set.

[0103] Calculate the fourth value of the single-node state value function for each optional node of each chromosome individual in the population during this iteration. If the fourth value is less than the fourth threshold, replace the corresponding optional node with a new optional node based on the optional policy set.

[0104] In this embodiment, the optimal nodes and node combinations are searched through crossover and mutation using a genetic algorithm. For each node, the search includes finding the highest-value alternative node and the node with the highest overall value considering multiple alternative actions. An improved genetic algorithm is used to solve for the maximum value of the process engine in both process states and actions, specifically including:

[0105] In the chromosome coding and value maximization process engine selection algorithm, genes are process nodes. The process order of these nodes represents all the paths required by the process. Therefore, an ordered real number coding method is adopted. For each node, an executor / equipment resource pool is bound (e.g., approval node → department head, quality inspection node → AI quality inspection instrument). Business rules are formulated, such as the preconditions that the node must meet, and value weights are obtained, such as calculating the node value coefficient through historical data. Valid paths in the business process are defined, such as a legal node sequence from the starting point O to the ending point D. The coding content includes the node order and unique identifier in the path.

[0106] Construct a fitness function model, defining the fitness function based on the node p that is connected to the process nodes in the process engine: Where p(m) represents the number of effective gene segments on the m-th chromosome, P max This represents the maximum number of effective gene segments among all individuals in this generation. Adding 1 is to avoid a fitness of 1 (if it is 1, it means 100% fitness, and the chromosome will no longer mutate). When the calculation reaches a certain generation, the fitness function is defined based on the path resistance (process value) of the chromosome. The effective gene segment is the sum of the path resistances traversed by a certain chromosome from process node O to process node D, which meets the requirement. That is, the chromosome's p(m) = ∑d(m1,m2), where d(m1,m2) is the node between m1 and m2 whose process value is greater than the threshold.

[0107] Chromosome crossover involves randomly selecting a crossover point Q from the parent chromosomes A and B. The crossover point Q should ensure that the effective gene segments of the parent chromosomes are not destroyed, and to remove duplicate and redundant process nodes from the two new chromosomes obtained by the crossover.

[0108] Chromosomal mutation involves modifying invalid gene segments. If the mutated chromosome has a longer valid gene segment than the original chromosome, it replaces the parent chromosome. The length of the valid gene segment is determined based on whether it conforms to business process rules and the efficiency / cost / quality of path execution; otherwise, it is not replaced.

[0109] If the first digit of an invalid gene fragment is D, and the chromosome is invalid, then a node is inserted at the position before D. After insertion, the resulting chromosome is guaranteed to be valid, i.e., it conforms to the encoding rules, and the length of the chromosome cannot be changed.

[0110] Chromosome selection first involves using roulette wheel selection, crossover, and mutation to obtain the next generation of chromosomes. Since a large number of chromosomes are considered unreasonable in the genetic calculations of previous generations, reasonable chromosomes from the current generation are used to replace unreasonable chromosomes with small gene segments in the next generation.

[0111] If there is no reasonable chromosome in the current generation, no replacement is performed. When the genetic calculation has been performed for a certain number of generations, most chromosomes are reasonable. At this time, the path resistance of the current generation is compared with the chromosome with the largest path resistance. If the best chromosome in the current generation has a smaller path resistance than the worst chromosome in the next generation, it is replaced; otherwise, it is not replaced. If the path resistance of the current generation and the smallest individual are not the most reasonable process path, no replacement operation is performed.

[0112] In chromosome update methods, the population design needs to balance maintaining population diversity and achieving fast convergence. Mathematically, allowing superior individuals from the parent generation to enter the next round of competition ensures the iterative stability of the optimal solution, while eliminating inferior individuals from the offspring early accelerates the optimization process. This is achieved by simultaneously allowing both superior offspring and superior parent individuals to enter the next generation of the population update. Specifically, parent individuals undergo crossover and mutation operations to obtain temporary offspring. The parent individuals and temporary offspring are compared, and the individual with higher fitness is selected as the next generation's offspring. This population update method ensures that each crossover and mutation operation produces two better offspring, thus guaranteeing the algorithm's convergence.

[0113] Iterative optimization involves ants gradually optimizing task order and resource allocation through multiple iterations, ultimately finding a configuration that maximizes the value function.

[0114] In one implementation, the second value of the cumulative value function of the last optional node of each chromosome individual in the population obtained in the last iteration is calculated, and the chromosome individual with the largest second value is obtained, specifically including:

[0115] Calculate the fifth value of the cumulative state value function of the last optional node of each chromosome individual in the population obtained in the last iteration, and obtain several last optional nodes whose fifth value is greater than the fifth threshold;

[0116] Calculate the sixth value of the cumulative state-action value function for several last optional nodes whose fifth value is greater than the fifth threshold, and obtain the chromosome individual corresponding to the last optional node with the largest sixth value.

[0117] In this embodiment, the overall value of the node chain performing the complete business process is evaluated based on the accumulated value of the last node. This includes evaluating the accumulated state value and accumulated state-action value of the last optional node in the node chain, ensuring that the node chain can remain in an optimized state by adjusting actions even after some node states have changed. In reinforcement learning, the value function is typically estimated using a value iteration method. This method, by repeatedly updating the value of the states, eventually converges to the optimal policy, and can be expressed as: Where R(s,a) is the immediate reward obtained by taking action a in state s; P(s'|s,a) is the state transition probability, that is, the probability of transitioning to state s' after taking action a in state s. The system can continuously update the value of the state and finally select the best strategy to optimize the execution of the process.

[0118] Example 2:

[0119] like Figure 2 As shown, this application provides an automatic configuration device for a process engine, the device comprising:

[0120] Value function module 1 is used to obtain the value function of the process engine in processing business. The value function represents the first value of a single optional node processing a single link of the business according to its own optional actions, and the second value of multiple optional nodes jointly processing the business completely according to their respective optional actions in the process engine.

[0121] The optimization solution module 2, connected to the value function module 1, is used to optimize the solution of the value function in order to obtain the optimal strategy for the configuration process engine to process business. The optimal strategy represents the single optimal node and its optimal action in the process engine that can obtain the optimal first value in a single link of processing business, and the multiple optimal nodes and their respective optimal actions that can obtain the optimal second value in the complete processing of business.

[0122] In one embodiment, the value function module 1 specifically includes:

[0123] The business information unit is used to obtain the various stages of a business process to be configured in the process engine, the input / output conditions of each stage, and the optional algorithms for processing each stage.

[0124] The node matching unit, connected to the business information unit, is used to match optional nodes in the process engine for each stage. Each optional node has at least one optional algorithm for processing the corresponding stage.

[0125] The optional action unit, connected to the node matching unit, is used to obtain the optional connection relationship of each optional node based on several input / output conditions of each stage. The optional algorithm and the optional connection relationship constitute the optional action of each optional node.

[0126] The model building unit, connected to the optional action unit, is used to construct the value function for the process engine to handle business, taking the node state and optional action of each optional node as input and the single-node reward and cumulative reward of each optional node in handling a certain business as output.

[0127] In one embodiment, the model building unit specifically includes:

[0128] The expected return unit is used to obtain the single-node return of each optional node in various node states, based on different connection relationships and processing algorithms, for processing a certain business in a corresponding stage. The single-node return is obtained based on the processing cost and processing time of the corresponding stage of processing a certain business.

[0129] The single-node and cumulative model units are connected to the expected return unit. They are used to construct the single-node value function of each optional node for processing a certain business based on each state and action, according to the single-node return of each node. Based on the connection relationship of each optional node and the single-node value function, the cumulative value function of each optional node for processing a certain business is constructed.

[0130] In one embodiment, the single-node and cumulative model unit specifically includes:

[0131] The single-node state-action value function unit is used to obtain the single-node state-action value function for each selectable node. In the formula, S k A represents the node state of the optional node k. k This represents a node action consisting of the upper connectivity of an optional node k and an optional algorithm, where i represents the number of trials, 0 ≤ γ ≤ 1 represents the reward discount factor, and R... i Let represent the reward obtained by the optional node k in the i-th trial, and s and a represent the specific values ​​of the node state and node action, respectively. This represents the average of the rewards obtained from multiple trials;

[0132] The single-node state value function unit, connected to the single-node state-action value function unit, is used to obtain the single-node state value function for each selectable node: This represents the average reward for an optional node k performing j node actions in state s;

[0133] The accumulated state-action value function unit, connected to the single-node state-action value function unit, is used to obtain the accumulated state-action value function for each selectable node: Q π (S k A k )=[Q(S k A k )+Q π (S k-1 A k-1 )|S k =s,A k =a], where π represents the optional strategy by which the process engine fully processes a certain business through optional nodes, several connection relationships and several optional algorithms, and k-1 and k represent the two adjacent optional nodes in the optional strategy respectively;

[0134] The cumulative state value function unit, connected to the single-node state value function unit, is used to obtain the cumulative state value function for each selectable node: V π (S k )=[V(S k )+V π (S k-1 )|S k =s).

[0135] In one embodiment, the optimization solution module 2 specifically includes:

[0136] Encoding unit, used to encode the gene fragment of each optional node in the encoding process engine, each gene fragment includes the node state and optional action of the corresponding optional node;

[0137] The strategy set unit is used to build a set of optional strategies for the process engine to handle a certain business. The optional strategy set includes all optional combinations of optional nodes and optional actions in the process engine that can jointly and completely handle a certain business.

[0138] The solution unit, connected to the encoding unit and the strategy set unit, is used to construct a chromosome individual that can fully process a certain business by the process engine with the optional strategy set as a constraint, and to use a genetic algorithm to solve for the chromosome individual with the optimal value function.

[0139] The strategy selection unit, connected to the solution unit, is used to obtain the optional strategies corresponding to the chromosome individuals with the optimal value function, and to serve as the optimal strategy for the configuration process engine to handle business.

[0140] In one embodiment, the solving unit specifically includes:

[0141] Population unit, used to obtain the population for the current iteration of the genetic algorithm. The population is either a randomly generated initial population or a population obtained through iterative optimization. The population includes multiple chromosome individuals, and each chromosome individual corresponds to an optional strategy.

[0142] The iteration unit, connected to the population unit, is used to calculate the first value of the single-node value function of each optional node of each chromosome individual in the population in the current iteration. If the first value is less than the first threshold, crossover and mutation operations are performed on the corresponding optional node based on the optional strategy set to obtain the population in the next iteration.

[0143] The termination unit, connected to the iteration unit, is used to calculate the second value of the cumulative value function of the last optional node of each chromosome individual in the population obtained in the last iteration when the termination condition of the genetic algorithm iteration is met, and to obtain the chromosome individual with the largest second value.

[0144] In one embodiment, the iteration unit specifically includes:

[0145] The mutation unit is used to calculate the third value of the single-node state-action value function of each optional node of each chromosome individual in the population in this round of iteration. If the third value is less than the third threshold, the gene fragment of the corresponding optional node is mutated based on the optional strategy set.

[0146] The crossover unit is used to calculate the fourth value of the single-node state value function for each optional node of each chromosome individual in the population during this iteration. If the fourth value is less than the fourth threshold, the corresponding optional node is replaced with a new optional node based on the optional policy set.

[0147] In one embodiment, the termination unit specifically includes:

[0148] The first selection unit is used to calculate the fifth value of the cumulative state value function of the last selectable node of each chromosome individual in the population obtained in the last iteration, and to obtain several last selectable nodes whose fifth value is greater than the fifth threshold.

[0149] The second selection unit, connected to the first selection unit, is used to calculate the sixth value of the cumulative state-action value function of several last selectable nodes whose fifth value is greater than the fifth threshold, and to obtain the chromosome individual corresponding to the last selectable node with the largest sixth value.

[0150] Example 3:

[0151] Embodiment 3 of this application provides a computer-readable storage medium storing a computer program. When the computer program is run by a processor, it implements the automatic configuration method of the process engine as described in Embodiment 1, or the automatic configuration device of the process engine as described in Embodiment 2.

[0152] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program units, or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other memory technologies, CD-ROM (Compact Disc Read-Only Memory), DVD or other optical disc storage, cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer.

[0153] Additionally, this application may provide a computer device including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the process engine automatic configuration method as described in Embodiment 1. This computer device may be the process engine automatic configuration apparatus as described in Embodiment 2.

[0154] The memory is connected to the processor. The memory can be flash memory, read-only memory or other types of memory. The processor can be a central processing unit or a microcontroller.

[0155] Embodiments 1-3 of this application provide an automatic configuration method, apparatus, and medium for a process engine. Based on the individual steps of the optional nodes in the process engine that process the business according to optional actions, and the complete steps of the business that process the business together, a value function for the process engine to process the business is constructed. The value function is optimized and solved to obtain the optimal strategy for configuring the process engine to process the business. The optimal strategy includes the single node with the highest value and the entire node chain. Finally, the optimal automatic configuration of the process engine to process the business can be achieved based on the optimal strategy.

[0156] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of this application, and this application is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of this application, and these modifications and improvements are also considered to be within the scope of protection of this application.

Claims

1. A method for automatically configuring a process engine, characterized in that, The method includes: Obtain the value function of the process engine for processing business. The value function represents the first value of a single optional node processing a single step of the business according to its own optional actions, and the second value of multiple optional nodes jointly processing the business completely according to their respective optional actions. The optimal solution is obtained by solving the value function to obtain the optimal strategy for the configuration process engine to process business. The optimal strategy represents the single optimal node and its optimal action that can obtain the optimal first value in a single link of processing business in the process engine, and the multiple optimal nodes and their respective optimal actions that can obtain the optimal second value in the complete processing of business.

2. The method according to claim 1, characterized in that, The value functions for the business processing by the workflow engine are obtained, specifically including: Obtain the various stages of a business process to be configured, the input / output conditions of each stage, and the several optional algorithms for processing each stage; For each stage, an optional node is matched in the process engine, and each optional node has at least one optional algorithm to process the corresponding stage; Based on several input / output conditions of each stage, the optional connection relationships of each optional node are obtained. The optional algorithm and optional connection relationships constitute the optional actions of each optional node. Using the node status and optional actions of each optional node as input, and the single-node reward and cumulative reward of each optional node in handling a certain business as output, the value function of the process engine for handling the business is constructed respectively.

3. The method according to claim 2, characterized in that, Using the node status and optional actions of each optional node as input, and the single-node reward and cumulative reward of each optional node in handling a certain business as output, the value function for the business processing of the process engine is constructed, specifically including: The experiment obtains the single-node return of each optional node in various node states, based on different connection relationships and processing algorithms, for processing a certain business process. The single-node return is obtained based on the processing cost and processing time of the corresponding process of a certain business process. Based on the reports from each individual node, construct the single-node value function for each optional node in handling a certain business based on each state and action. Based on the connection relationships of each optional node and the single-node value function, construct the cumulative value function for each optional node in handling a certain business.

4. The method according to claim 3, characterized in that, Based on the reports from each individual node, construct the single-node value function for each optional node in handling a specific business process according to each state and action. Based on the connection relationships of each optional node and the single-node value function, construct the cumulative value function for each optional node in handling a specific business process, specifically including: The function to obtain the single-node state-action value of each optional node is as follows: In the formula, S k A represents the node state of the optional node k. k This represents a node action consisting of the upper connectivity of an optional node k and an optional algorithm, where i represents the number of trials, 0 ≤ γ ≤ 1 represents the reward discount factor, and R... i Let represent the reward obtained by the optional node k in the i-th trial, and s and a represent the specific values ​​of the node state and node action, respectively. This represents the average of the rewards obtained from multiple trials; The function to obtain the single-node state value of each optional node: This represents the average reward for an optional node k performing j node actions in state s; The function to obtain the accumulated state-action value of each optional node: Q π (S k A k )=[Q(S k A k )+Q π (S k-1 A k-1 )|S k =s,A k =a], where π represents the optional strategy by which the process engine fully processes a certain business through optional nodes, several connection relationships and several optional algorithms, and k-1 and k represent the two adjacent optional nodes in the optional strategy respectively; Obtain the cumulative state value function for each optional node: V π (S k )=[V(S k )+V π (S k-1 )|S k =s).

5. The method according to any one of claims 1-4, characterized in that, The optimal solution to the value function is obtained to determine the optimal strategy for the configuration process engine to handle business operations. This includes: The gene fragment of each optional node in the coding process engine, each gene fragment includes the node state and optional actions of the corresponding optional node; Construct a set of optional strategies for the process engine to handle a certain business. The set of optional strategies includes all possible combinations of optional nodes and optional actions in the process engine that can collectively and completely handle a certain business. With the set of optional strategies as constraints, a process engine is constructed to fully process a certain business's chromosome individual, and a genetic algorithm is used to solve for the chromosome individual with the optimal value function. The optimal strategy for the chromosome individual with the best value function is obtained and used as the optimal strategy for the configuration process engine to handle business.

6. The method according to claim 5, characterized in that, Given a set of optional strategies as constraints, a process engine is constructed to fully process a specific business process using chromosome individuals. A genetic algorithm is then used to find the chromosome individual with the optimal value function, specifically including: Obtain the population used for this round of iteration of the genetic algorithm. The population is either a randomly generated initial population or a population obtained through iterative optimization. The population includes multiple chromosome individuals, and each chromosome individual corresponds to an optional strategy. Calculate the first value of the single-node value function for each optional node of each chromosome individual in the population of this iteration. If the first value is less than the first threshold, perform crossover and mutation operations on the corresponding optional node based on the optional strategy set to obtain the population of the next iteration. When the termination condition of the genetic algorithm iteration is met, calculate the second value of the cumulative value function of the last optional node of each chromosome individual in the population obtained in the last iteration, and obtain the chromosome individual with the largest second value.

7. The method according to claim 6, characterized in that, Calculate the first value of the single-node value function for each optional node of each chromosome individual in the population during this iteration. If the first value is less than a first threshold, perform crossover and mutation operations on the corresponding optional node based on the optional strategy set, specifically including: Calculate the third value of the single-node state-action value function for each optional node of each chromosome individual in the population during this iteration. If the third value is less than the third threshold, perform optional action mutation on the gene fragment of the corresponding optional node based on the optional strategy set. Calculate the fourth value of the single-node state value function for each optional node of each chromosome individual in the population during this iteration. If the fourth value is less than the fourth threshold, replace the corresponding optional node with a new optional node based on the optional policy set.

8. The method according to claim 6, characterized in that, Calculate the second value of the cumulative value function of the last optional node of each chromosome individual in the population obtained in the last iteration, and obtain the chromosome individual with the largest second value. Specifically, this includes: Calculate the fifth value of the cumulative state value function of the last optional node of each chromosome individual in the population obtained in the last iteration, and obtain several last optional nodes whose fifth value is greater than the fifth threshold; Calculate the sixth value of the cumulative state-action value function for several last optional nodes whose fifth value is greater than the fifth threshold, and obtain the chromosome individual corresponding to the last optional node with the largest sixth value.

9. An automatic configuration device for a process engine, characterized in that, The device includes: The value function module is used to obtain the value function of the process engine in processing business. The value function represents the first value of a single optional node processing a single link of the business according to its own optional actions, and the second value of multiple optional nodes jointly processing the business completely according to their respective optional actions in the process engine. The optimization solution module, connected to the value function module, is used to optimize the solution of the value function in order to obtain the optimal strategy for the configuration process engine to process business. The optimal strategy represents the single optimal node and its optimal action in the process engine that can obtain the optimal first value in a single link of processing business, and the multiple optimal nodes and their respective optimal actions that can obtain the optimal second value in the complete processing of business.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the automatic configuration method for the process engine as described in any one of claims 1-8.

Images