A business risk executable program generation method, device, equipment and storage medium

Abstract

The application discloses a service risk executable program generation method, device and equipment and a storage medium. The method is applied to a main intelligent agent of a service risk program generation intelligent agent, the service risk program generation intelligent agent is realized based on a super intelligent agent operation framework, and the method comprises the following steps: in response to a service risk program generation request of a target demand side, performing request analysis on the service risk program generation request, and determining a risk service type; according to the risk service type, generating a risk reference sample data set based on pre-set sample feature dimensions; according to the risk reference sample data set and the risk service type, generating a service risk executable program of the target demand side; and feeding back the service risk executable program to the target demand side, so that the target demand side calls and executes.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, and storage medium for generating executable programs for business risks. Background Technology

[0002] In the financial sector, intelligent risk decision-making for financial businesses such as lending is a key component of a financial institution's core risk control capabilities. In recent years, with the rapid development of artificial intelligence technology, business risk control systems have evolved from traditional rule engines to large-scale models, but existing business risk decision-making methods still have many shortcomings.

[0003] Existing traditional risk control solutions primarily rely on manual review and static rule engines. These solutions are mainly based on fixed rule engines and static scoring logic, with risk control rules and model parameters preset and fixed manually. Business approval processes are executed in a linear, sequential manner. In contrast, intelligent risk control decision-making solutions based on large-scale models leverage semantic understanding, unstructured data parsing, and complex reasoning capabilities to make business risk decisions in the financial sector.

[0004] However, traditional static rule-based engines have rigid risk processes that cannot adapt to the dynamic risk changes required in financial scenarios. This results in low accuracy of business risk decisions in financial settings, and the large number of rules and long iteration cycles lead to inefficiency in risk decision-making. Furthermore, existing large-model decision-making methods suffer from the prominent "large model illusion" problem, which may directly cause risk decisions to deviate from reality, resulting in unreliable outputs. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and storage medium for generating executable programs for business risks, in order to improve the efficiency, accuracy, and reliability of business risk decision-making in financial scenarios.

[0006] According to one aspect of the present invention, a method for generating a business risk executable program is provided, applied to a main intelligent agent of a business risk program generation intelligent agent, wherein the business risk program generation intelligent agent is implemented based on a super intelligent agent runtime framework, and the method includes: In response to a business risk procedure generation request from a target demander, the business risk procedure generation request is parsed to determine the type of risky business. Based on the risk business type, a risk reference sample dataset is generated using pre-defined sample feature dimensions. Based on the risk reference sample dataset and the risk business type, generate an executable program for the business risk of the target requester; The executable program for the business risk is fed back to the target requester for execution.

[0007] According to another aspect of the present invention, an apparatus for generating a business risk executable program is provided, applied to a main intelligent agent of a business risk program generation intelligent agent, the business risk program generation intelligent agent being implemented based on a super intelligent agent runtime framework, the apparatus comprising: The request parsing module is used to respond to the business risk program generation request from the target demander, parse the business risk program generation request, and determine the risk business type. The sample data generation module is used to generate a risk reference sample dataset based on the risk business type and pre-defined sample feature dimensions. The program generation module is used to generate an executable program for the business risk of the target requester based on the risk reference sample dataset and the risk business type. The program feedback module is used to feed back the executable program for the business risk to the target requester, so that the target requester can call and execute it.

[0008] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the method for generating a business risk executable program according to any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the method for generating a business risk executable program according to any embodiment of the present invention.

[0010] The technical solution of this invention parses the business risk program generation request to determine the risk business type. Based on the risk business type and pre-defined sample feature dimensions, a risk reference sample dataset is generated. Based on the risk reference sample dataset and the risk business type, a business risk executable program for the target requester is generated and fed back to the target requester for execution. This technical solution utilizes a main intelligent agent implemented based on a super-intelligent agent framework to handle request response and parsing, and automatically constructs the risk reference sample dataset. The risk business type is considered during dataset construction, ensuring the accuracy and efficiency of the risk reference sample dataset generation for the corresponding risk business type. The automated generation of the business risk executable program based on the risk reference sample dataset and risk business type improves the accuracy and reliability of the generated program, thereby enhancing the efficiency, accuracy, and reliability of subsequent business risk decisions in financial scenarios based on this executable program.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a method for generating a business risk executable program according to Embodiment 1 of the present invention; Figure 2 This is a flowchart of a method for generating a business risk executable program according to Embodiment 2 of the present invention; Figure 3 This is a flowchart of a method for generating a business risk executable program according to Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of a device for generating a business risk executable program according to Embodiment 4 of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device that implements the method for generating a business risk executable program according to an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] Example 1 Figure 1 This is a flowchart illustrating a method for generating a business risk executable program according to Embodiment 1 of the present invention. This embodiment is applicable to situations in the financial field where risk decisions are made for businesses such as credit. The method can be executed by a device for generating a business risk executable program, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes: S110. In response to the business risk procedure generation request from the target demander, perform request parsing on the business risk procedure generation request to determine the risk business type.

[0017] S120. Based on the risk business type and pre-defined sample feature dimensions, generate a risk reference sample dataset.

[0018] S130. Based on the risk reference sample dataset and the risk business type, generate an executable procedure for the target demander's business risk.

[0019] S140. Feed back the executable procedure for business risks to the target requester for them to call and execute.

[0020] Specifically, the SuperAgentharness framework can be DeerFlow 2.0. DeerFlow 2.0 is essentially an operating system that enables agents to autonomously complete complex, long-term tasks. As an open-source SuperAgentharness framework, DeerFlow 2.0 includes technologies such as LangGraph and LangChain (an open-source framework for developing large-scale model applications). It centers on a single master agent, which has a built-in middleware chain for data processing and context management. It can also dynamically schedule sub-agents to execute in parallel, possessing core capabilities such as long-duration complex task scheduling, multi-sub-agent parallel collaboration, dynamic task decomposition, cross-session long-term memory persistence, context engineering, breakpoint resumption and state persistence, a unified interface for multiple models, a progressive Skills system (standardized capability modules built into DeerFlow 2.0), and MCP (ModelContextProtocol) server extensions.

[0021] This embodiment introduces the DeerFlow 2.0 multi-agent collaboration paradigm into the business risk decision-making domain of the financial industry, constructing a business risk program generation agent based on the DeerFlow 2.0 framework. This agent is used to automatically generate executable business risk programs that can meet the business risk decision-making needs of different stakeholders in financial scenarios. The technical solution in this embodiment can be executed by the main agent of the business risk program generation agent.

[0022] Among them, the target demander can be banking institutions or other entities with business risk decision-making needs in the financial sector; the business risk procedure generation request can be a request generated by the target demander based on its own actual needs; and the business risk procedure can be used to make business risk decisions.

[0023] Specifically, the main intelligent agent can receive business risk program generation requests through a unified API (Application Programming Interface) gateway, and identify the risk business type by parsing the key time periods in the request header. In financial scenarios, risk business types can include application review, fraud detection, risk rating, and post-loan risk warning.

[0024] In one optional embodiment, a risk reference sample dataset is generated based on the risk business type and pre-defined sample feature dimensions, including: generating feature reference data under the corresponding sample feature dimensions according to the risk business type; the sample feature dimensions include the scenario general business rule dimension, the risk assessment standard feature dimension, and the model output reliability verification dimension; and generating a risk reference sample dataset based on the feature reference data of each sample feature dimension.

[0025] The sample feature dimensions can include the general business rules dimension for the scenario, the risk assessment standard feature dimension, and the model output reliability verification dimension. Specifically, the general business rules dimension can be a set of general business rules for financial businesses, such as access rules, credit rules, and data source rules; the risk assessment standard feature dimension can be a set of risk features in the financial industry; and the model output reliability verification dimension can include mechanisms such as model illusion constraints. It should be noted that different risk business types require different feature datasets for each sample feature dimension.

[0026] Taking the scenario of "risk control review for loan applications from a small and micro manufacturing enterprise" as an example, the corresponding risk business type for this scenario is "application review". For the general business rules dimension of this scenario, the characteristic reference data for the corresponding admission rules may include: {"The applicant must be the legal representative or actual controller of the enterprise", "The enterprise must have been established for more than one year", "No records of business irregularities, administrative penalties, or being a dishonest executor"}; the characteristic reference data for the corresponding credit rules may include: {"No current overdue payments for individuals and related entities", "No frequent credit inquiries"}; the characteristic reference data for the corresponding data source rules may include: {"All review conclusions must be based on original, authentic data sources such as business registration data, credit data, transaction records, and tax data"}.

[0027] The feature reference data for the risk assessment standard feature dimensions corresponding to the above-mentioned business types of incoming document review risks may include: {supply chain risk: {"concentration of the top five suppliers", "inventory turnover days"}; financial risk: {"debt-to-equity ratio", "cash flow coverage ratio"}}, as well as the thresholds corresponding to each risk assessment standard feature, such as the number of days threshold corresponding to inventory turnover days, the debt ratio threshold corresponding to debt-to-equity ratio, etc.

[0028] The feature reference data for the reliability verification dimension of the model output corresponding to the above-mentioned business types of application review risks may include: {"a. Prohibition of unfounded inference", "b. Mandatory conclusion tracing", "c. Strong red line locking", "d. Real-time detection constraints"}. Specifically, prohibiting unfounded inference means that "the large model can only infer based on real original data such as collected business registration, tax, credit, and bank statements, and it is prohibited to arbitrarily determine that a company's operations are abnormal, fabricate overdue records, or make arbitrary assumptions about fraudulent behavior"; mandatory conclusion tracing means that "all risk ratings, risk judgments, and review conclusions must be labeled with the corresponding original data fields and data sources. Conclusions without data source support that are directly deemed invalid or whose information cannot be traced must not be used as a basis for decision-making, such as: "Based on page 3, line 5 of XXX bank statement", "Based on the snapshot of the National Enterprise Credit Information Publicity System on July 8, 2026", etc. The red line lock-in can specifically mean that "the hard regulatory rules such as credit access thresholds, overdue judgment standards, multiple borrowing thresholds, and fund use control are written into an immutable rule base. The big model has no right to arbitrarily relax the standards, weaken the risk level, or exempt violations, thus eliminating the illusion of soft judgment and lenient judgment from the root." The real-time detection constraint can specifically mean that "real-time detection constraint, all risk reasoning must meet the linear compliance link of "data facts, feature performance, and risk conclusion", automatically verify the causal rationality of reasoning, and forcibly mark and reset contradictory conclusions and conclusions that deviate from industry characteristics."

[0029] Specifically, the main intelligent agent can obtain feature reference data for the general business rule dimension of the risk business type corresponding to the above scenarios from the rule base, obtain feature reference data for the risk assessment standard feature dimension from the industry risk feature library, and dynamically generate multiple types of illusion constraints according to the risk business type to obtain feature reference data for the model output reliability verification dimension; and generate a risk reference sample dataset including feature reference data for each sample feature dimension.

[0030] The above technical solution comprehensively considers multiple sample feature dimensions in the process of generating risk reference sample datasets, and generates risk reference sample datasets based on feature reference data of risk business types under each sample feature dimension, thereby improving the comprehensiveness, diversity, reliability and accuracy of the generation of risk reference sample datasets.

[0031] Based on the risk reference sample dataset and the risk business type, an executable business procedure for the target requester is generated. Specifically, the main agent can generate target prompt words based on the risk reference sample dataset and the risk business type, and then call the LLM (Large Language Model) to generate an executable business risk procedure based on the target prompt words.

[0032] The main intelligent agent feeds back the executable business risk procedure to the target requester, who then invokes and executes it. Specifically, the target requester can invoke the executable business risk procedure according to its own business risk control needs to obtain the business risk decision result.

[0033] For example, consider a small manufacturing enterprise initiating a risk control review request for a business loan application to a target borrower. The target borrower obtains the enterprise's application request, parses it, and retrieves relevant application data, such as the loan amount requested, relevant business information, and the legal representative's information—all necessary for the review. The target borrower inputs this relevant data into a business risk executable program, which automatically makes a business risk decision and outputs the result. The decision can include approval or rejection; if the decision is rejection, the output may also include additional information such as the reason for rejection and recommendations.

[0034] For example, the business risk execution procedure, after analysis, outputs a final decision of rejection. Reasons for rejection include: "Red line rule: The company has been established for 11 months, failing to meet the minimum requirement of 12 months," and "Industry characteristics: The concentration of the top five suppliers is 85%, exceeding the industry threshold of 60%." The recommended suggestion is to have the company reapply after it has been established for at least 12 months.

[0035] The technical solution of this invention parses the business risk program generation request to determine the risk business type. Based on the risk business type and pre-defined sample feature dimensions, a risk reference sample dataset is generated. Based on the risk reference sample dataset and the risk business type, a business risk executable program for the target requester is generated and fed back to the target requester for execution. This technical solution utilizes a main intelligent agent implemented based on a super-intelligent agent framework to handle request response and parsing, and automatically constructs the risk reference sample dataset. The risk business type is considered during dataset construction, ensuring the accuracy and efficiency of the risk reference sample dataset generation for the corresponding risk business type. The automated generation of the business risk executable program based on the risk reference sample dataset and risk business type improves the accuracy and reliability of the generated program, thereby enhancing the efficiency, accuracy, and reliability of subsequent business risk decisions in financial scenarios based on this executable program.

[0036] Example 2 Figure 2 This is a flowchart of a method for generating a business risk executable program according to Embodiment 2 of the present invention. This embodiment is an optimization and improvement based on the above technical solutions.

[0037] Furthermore, the step "generating a business risk executable program for the target requester based on the risk reference sample dataset and the risk business type" is refined into "matching subtasks to a preset subtask combination template based on the risk business type to determine at least one target subtask required to execute the business risk program generation request, as well as the task execution relationship between each target subtask; generating a standard risk execution pipeline definition file based on the risk reference sample dataset, according to each target subtask and the task execution relationship between each target subtask; and generating a business risk executable program for the target requester based on the standard risk execution pipeline definition file." This improves the method for generating business risk executable programs.

[0038] It should be noted that for parts not described in detail in the embodiments of the present invention, please refer to the descriptions in other embodiments. For example... Figure 2 As shown, the method includes the following specific steps: S210. In response to the business risk procedure generation request from the target demander, perform request parsing on the business risk procedure generation request to determine the risk business type.

[0039] S220. Based on the risk business type and pre-defined sample feature dimensions, generate a risk reference sample dataset.

[0040] S230. Based on the risk business type, perform subtask matching on the preset subtask combination template to determine at least one target subtask required to generate the business risk procedure, as well as the task execution relationship between each target subtask.

[0041] S240. Based on the target sub-tasks and the task execution relationships between them, and using the risk reference sample dataset, generate a standard risk execution pipeline definition file.

[0042] S250. Generate executable procedures for business risks of the target demander based on the standard risk execution pipeline definition document.

[0043] S260. Feed back the executable procedure for business risks to the target requester for them to call and execute.

[0044] It should be noted that different risk business types correspond to different sub-task combination templates. The mapping relationship between risk business types and sub-task combination templates can be pre-generated and stored by the main intelligent agent, or it can be pre-built and stored by relevant technical personnel. For example, taking risk business types including application review, fraud behavior identification, risk rating, and post-loan risk warning as an example, the mapping relationship between risk business types and sub-task combination templates can be seen in Table 1.

[0045] Table 1 Specifically, subtask matching can be performed based on the subtask combination templates corresponding to the aforementioned risk business types to determine at least one target subtask required for the business risk procedure generation request or the corresponding risk business type, as well as the execution relationships between these target subtasks. It should be noted that the execution relationship between target subtasks is also the execution order between subtasks; the "→" in the aforementioned subtask combination template indicates the task execution order. For example, taking application review as an example, the "data collection and cleaning" subtask needs to be executed after the "identity verification" subtask is completed. However, it should be noted that some of the aforementioned subtasks can be executed in parallel. For example, "static risk screening" and "dynamic behavioral risk assessment" can be executed in parallel without affecting each other. Therefore, optionally, target subtasks that can be executed in parallel can be marked by adding parallel execution identifiers to the subtasks.

[0046] Taking the risky business type "application review" as an example, its corresponding target sub-tasks can include eight major target sub-tasks: "identity verification," "data collection and cleaning," "static risk screening," "dynamic behavioral risk assessment," "related risk mining," "illusion detection and error correction," "score level determination," and "compliance verification." For example, identity verification is the verification of the authenticity of the user and enterprise identities, and data collection and cleaning is the collection and standardized cleaning of multi-source business data.

[0047] Based on the target subtasks and the task execution relationships between them, a standard risk execution pipeline definition file is generated using the risk reference sample dataset. This standard risk execution pipeline definition file is also known as a standardized DAG (Directed Acyclic Graph) file.

[0048] In an optional embodiment, based on each target subtask and the task execution relationships between them, a standard risk execution pipeline definition file is generated using a risk reference sample dataset, including: Step a: Generate the basic topology of a directed acyclic graph (DAG) based on each target subtask and the task execution relationship between them; the DAG basic topology file includes the original node objects and their corresponding node edges.

[0049] Specifically, the main agent traverses each target subtask, generating the original node objects of the DAG, which are the target subtasks. Based on the task execution relationships between the target subtasks, directed node edges of the DAG are generated using a dependency depth computation algorithm. For target subtasks with parallel relationships, original node objects with parallel branch relationships can be generated. This results in the basic topology of the DAG containing the original node objects and the set of node edges.

[0050] Step b: Determine the dedicated intelligent agent for each target subtask and its corresponding intelligent agent constraints.

[0051] A task execution dedicated agent is used to execute target subtasks. Different target subtasks correspond to their own task execution dedicated agent. For example, the task execution dedicated agent A corresponding to target subtask a is only used to execute target subtask a.

[0052] The eight major objectives and sub-tasks are: "identity verification", "data collection and cleaning", "static risk screening", "dynamic behavioral risk assessment", "related risk mining", "hallucination detection and error correction", "score level determination" and "compliance verification".

[0053] Specifically, the dedicated intelligent agent for executing the sub-task of "Identity Verification" is the "Identity Verification Intelligent Agent"; the dedicated intelligent agent for executing the sub-task of "Data Collection and Cleaning" is the "Data Collection and Cleaning Intelligent Agent"; the dedicated intelligent agent for executing the sub-task of "Static Risk Screening" is the "Basic Rule Screening Intelligent Agent"; the dedicated intelligent agent for executing the sub-task of "Dynamic Behavioral Risk Assessment" is the "Dynamic Behavioral Risk Assessment Intelligent Agent"; the dedicated intelligent agent for executing the sub-task of "Association Risk Mining" is the "Multi-Party Association Risk Mining Intelligent Agent"; the dedicated intelligent agent for executing the sub-task of "Hallucination Detection and Correction" is the "Hallucination Verification and Correction Intelligent Agent"; the dedicated intelligent agent for executing the sub-task of "Scoring Level Determination" is the "Risk Rating and Scoring Intelligent Agent"; and the dedicated intelligent agent for executing the sub-task of "Compliance Verification" is the "Compliance Final Review and Archiving Intelligent Agent".

[0054] It should be noted that different task-specific intelligent agents have their own constraints during the execution process. For example, the constraints may include whether an independent sandbox is needed during task execution and the timeout threshold for agent execution.

[0055] Step c: Based on the task execution agent and its corresponding constraints for each target subtask, update the original node objects and their corresponding node edges of the DAG basic topology to obtain the intermediate DAG topology.

[0056] Specifically, the main agent uses the task-specific agent and its corresponding agent constraints corresponding to each target subtask to update the original node objects and their corresponding node edges of the DAG basic topology, generating an intermediate DAG topology with references to the task-specific agent and its agent constraints, which is the DAG intermediate state.

[0057] Step d: Update the intermediate topology of the DAG using the risk reference sample dataset to obtain the target topology of the DAG.

[0058] Specifically, feature data from the risk reference sample dataset is injected into the node objects and edges of the intermediate DAG topology to obtain the target DAG topology. However, it should be noted that the target subtasks have an execution order and required resources. The execution order includes sequential execution and parallel execution. Therefore, the execution order can be represented using task priority features, and features of task execution priority and computing resource allocation are introduced into the intermediate DAG structure.

[0059] Optionally, the intermediate DAG topology is updated using a risk reference sample dataset to obtain the target DAG topology. This includes: generating engineering prompts that include priority ranking and computing resource allocation mechanisms for each target subtask; generating task priority ranking results and computing resource allocation results for each target subtask based on the engineering prompts; updating the intermediate DAG topology using the task priority ranking results and computing resource allocation results for each target subtask to obtain a DAG reference topology; and updating the DAG reference topology using a risk reference sample dataset to obtain the target DAG topology.

[0060] It should be noted that each target subtask has its corresponding execution order. The subtask execution template only provides the executable sequence of subtasks. However, during the execution of one or more subtasks, there may be subtasks that can be executed concurrently, as well as subtasks that can run throughout the entire task execution chain. Therefore, determining the task priority for each target subtask can improve the execution efficiency of the target subtasks and effectively avoid task execution conflicts. Furthermore, the computing resources required for the execution of target subtasks vary. Therefore, to ensure that the execution of target subtasks is not abnormal, differentiated computing resources can be allocated to different target subtasks during execution, thereby solving the resource contention and timing disorder problems in multi-agent parallel scheduling.

[0061] The system generates project prompts, which may include the priority ranking and computing resource allocation mechanisms for each target subtask. These mechanisms may include priority allocation, dynamic priority adjustment, concurrency cap control, multi-level queue scheduling, dynamic resource allocation, and resource isolation.

[0062] The prompts corresponding to the priority allocation mechanism can be generated by referring to the following feature dimensions: For example, a multi-level priority system can be pre-set, such as P0-P5, where P0 is the highest priority and P5 is the lowest. The identity verification subtask is a prerequisite for access control; without identity verification, all subsequent tasks are invalid. Therefore, it should be allocated the highest CPU (Central Processing Unit) and memory resources, prioritized for execution, and given priority for retrying in case of exceptions. Its corresponding priority should be considered as P0. The data collection and cleaning subtask provides the basic data source for the entire process, relying on successful identity verification. It consumes batch IO (Input / Output) and container resources and should be allocated dedicated sandbox container cluster resources to ensure data throughput efficiency. Its corresponding priority should be considered as the second highest, P1.

[0063] The static risk screening subtask can begin lightweight rule calculations after data cleaning. It can run in parallel with some deep feature tasks and should be allocated a parallel priority of P2, along with standard computing resources. The dynamic behavioral risk assessment subtask relies on time-series behavioral data and can run in parallel with the static risk screening subtask. This subtask focuses more on time-series model calculations and should be allocated dedicated computing resources for time-series inference. Its corresponding priority should be allocated a parallel priority of P2. The association risk mining subtask relies on the fused feature graph and must wait for the data fusion step to complete before execution. It should be allocated dedicated computing resources for graph computation and its corresponding priority should be allocated a medium priority of P3.

[0064] The illusion detection and correction subtask is a full-process constraint task, involving pre-implementation, in-process verification, and post-review. It utilizes illusion constraint middleware throughout, without monopolizing computing power but intercepting illusion outputs in real time. Its priority should be considered across priorities P0-P3. The scoring level determination subtask depends on the convergence of all analysis and correction results, starting after all results are summarized. It should be allocated computing power for scoring modeling. Its priority should be considered a later priority P4. The compliance verification subtask is the final closing task, relying on all pre-tasks for completion. It should be allocated dedicated resources for compliance auditing. Its priority should be considered a final priority P5.

[0065] The prompt for the dynamic priority adjustment mechanism could be: When a backlog of high-priority tasks is detected, temporarily reduce the weight of low-priority tasks. The prompt for the concurrency cap control mechanism could be: The SubagentLimitMiddleware middleware component in DeerFlow 2.0, which enforces the maximum number of concurrent sub-agents, dynamically controls the number of concurrent sub-agents to prevent overload. The prompt for the multi-level queue scheduling mechanism could be: Assign subtasks of different priorities to their corresponding queues to determine the scheduling order. The prompt for the dynamic resource allocation mechanism could be: The resource scheduler can dynamically allocate computing resources based on task priority and load. The prompt for the resource isolation mechanism could be: Allocate independent sandbox environments for high-priority subtasks through the SandboxMiddleware middleware component to prevent resource contention.

[0066] Based on the prompts corresponding to the aforementioned mechanisms, engineering prompt information is generated. The main agent then uses this information to generate task priority ranking results and computing resource allocation results for each target sub-task. These results are then used as node attributes in the intermediate DAG topology to update the DAG reference topology. Finally, the risk reference sample dataset is injected into the node attributes of each node object in the DAG reference topology to obtain the target DAG topology.

[0067] Step e: Instantiate the DAG target topology and generate a standard risk execution pipeline definition file.

[0068] Specifically, the main agent can call the middleware TodoMiddleware to instantiate the DAG target topology and aggregate all metadata, such as node objects, node edges, node attributes, etc. Node attributes, such as task priority, resource quota, model constraints and rules, into JSON (JavaScriptObjectNotation) format to obtain a standard risk execution pipeline definition file that can be parsed by LangGraph.

[0069] Generate a business risk executable program for the target requester based on a standard risk execution pipeline definition file. In one optional embodiment, generating the business risk executable program for the target requester based on the standard risk execution pipeline definition file includes: generating a middleware executable environment based on the standard risk execution pipeline definition file; generating a workflow instance from the standard risk execution pipeline definition file to run on the middleware executable environment based on a pre-selected intelligent agent graph workflow framework, and encoding the instance; using the compiled workflow instance as the business risk executable program for the target requester.

[0070] Optionally, generating a middleware executable environment based on a standard risk execution pipeline definition file includes: parsing the standard risk execution pipeline definition file to determine capability requirement tags; determining the middleware components that perform the corresponding functions based on the capability requirement tags, and generating a middleware processing chain based on the execution order of each middleware component; injecting middleware configuration parameters into the middleware processing chain to generate a middleware executable environment.

[0071] The capability requirement tag is used to declare the capabilities required for the environment to operate, that is, which middleware capabilities are needed for the task environment to run. The capability requirement tag is used for capability requirement declaration, while the middleware acts as the capability provider. The standard risk execution pipeline definition file defines the node attributes and edge attributes of different nodes, represented as fields in the file, such as the "requires" field and "max_concurrency". The capability requirement tag can be determined by parsing specific fields in the standard risk execution pipeline definition file. For example, parsing the "requires" field and obtaining "requires_sandbox: true" indicates that a sandbox isolation environment is required.

[0072] Specifically, the standard risk execution pipeline definition file is parsed to obtain capability requirement tags, such as the need for a sandbox isolation environment, concurrency control, phantom verification, file upload handling, and cross-session and persistence capabilities. It should be noted that DeerFlow 2.0 has a fixed-order middleware chain mechanism, specifically comprising 14 layers of middleware, whose order cannot be changed. For example, the middleware SandboxMiddleware in layer 2 must precede the middleware ThreadDataMiddleware in layer 3, because the execution of SandboxMiddleware requires the thread_id of the aforementioned middleware.

[0073] The required middleware components are determined based on the capability requirement tags, i.e., which middlewares are to be launched. For example, if the capability requirement tag includes "requires_sandbox: true", the corresponding middleware to be launched is SandboxMiddleware (layer 2). Each middleware component is instantiated sequentially in ascending order of its layer hierarchy. Based on the execution order of each middleware component, DeerFlow 2.0 uses an Onion Model architecture to generate the middleware processing chain. The middleware configuration parameters are injected into the middleware processing chain according to the standard risk execution pipeline definition file, resulting in the middleware executable environment. For example, the middleware configuration parameter to be injected for the middleware SubagentLimitMiddleware includes the concurrency limit (max_concurrency).

[0074] The mapping relationship between the hierarchy, name and core responsibilities of the 14-layer middleware components in DeerFlow 2.0 can be found in Table 2.

[0075] Table 2 The main agent, based on a pre-selected agent graph workflow framework, generates a workflow instance from the standard risk execution pipeline definition file, which runs on the middleware executable environment. This instance is then coded and compiled into an executable program for the target client's business risks. Specifically, the agent graph workflow framework can be the LangGraph workflow framework.

[0076] This embodiment's technical solution matches pre-defined subtask combination templates based on risk business types to determine at least one target subtask required to execute the business risk program generation request, as well as the task execution relationships between each target subtask. Based on each target subtask and the task execution relationships between them, and using a risk reference sample dataset, a standard risk execution pipeline definition file is generated. Based on the standard risk execution pipeline definition file, a business risk executable program for the target requester is generated. During the execution of the business risk executable program, the target subtasks are automatically determined, and a pipeline definition file is constructed based on the execution relationships of the subtasks and the risk reference sample dataset. This further improves the efficiency and accuracy of the main intelligent agent in generating business risk executable programs.

[0077] Furthermore, leveraging the long-term task processing and cross-session long-term memory iteration capabilities of the DeerFlow 2.0 framework, the main agent can autonomously optimize large model illusion error samples, and autonomously optimize the dimensions of general business rules, risk assessment standard features, and model output reliability verification. This eliminates the need for frequent manual updates of rules and model illusion constraints. Compared to traditional manual iteration methods, this significantly improves the efficiency of risk control rule iteration and reduces manual maintenance costs. It can adapt to the problem of random illusions in dynamically iterated large models in real time, solving the problems of poor adaptability and continuous interference from large model illusions in traditional risk control. The technical solution of task decomposition by the main agent and parallel execution by multiple dedicated agents overcomes the traditional linear sequential risk decision-making process. Multi-dimensional data processing, risk assessment, and illusion verification can be performed simultaneously, further shortening the time required for risk decision-making.

[0078] Example 3 Figure 3 This is a schematic diagram of the flow structure of a method for generating a business risk executable program according to Embodiment 3 of the present invention. Based on the above embodiments, this embodiment provides a preferred example. In this embodiment, the main intelligent agent for generating the business risk program is implemented based on the super-intelligent agent runtime framework DeerFlow 2.0, and specifically includes the following steps: S30. In response to the business risk procedure generation request from the target demander, perform request parsing on the business risk procedure generation request to determine the risk business type.

[0079] S31. Based on the risk business type and pre-defined sample feature dimensions, generate a risk reference sample dataset.

[0080] S32. Based on the risk business type, perform subtask matching on the preset subtask combination template to determine at least one target subtask required to generate the business risk procedure request, as well as the task execution relationship between each target subtask.

[0081] S33. Generate the basic topology of a directed acyclic graph (DAG) based on each target subtask and the task execution relationship between each target subtask.

[0082] S34. Determine the dedicated agent for each target subtask and its corresponding agent constraints. Based on the dedicated agent for each task and its corresponding agent constraints, update the basic DAG topology to obtain the intermediate DAG topology.

[0083] S35. Generate the task priority ranking results and computing resource allocation results corresponding to each target subtask, and update the DAG intermediate topology using the task priority ranking results and computing resource allocation results to obtain the DAG reference topology.

[0084] S36. Update the DAG reference topology using the risk reference sample dataset to obtain the DAG target topology, and instantiate the DAG target topology to generate a standard risk execution pipeline definition file.

[0085] S37. Analyze the standard risk execution pipeline definition document and determine the capability requirement labels.

[0086] S38. Determine the middleware components that perform the corresponding functions based on the capability requirement tags, and generate a middleware processing chain according to the execution order of each middleware component. Inject middleware configuration parameters into the middleware processing chain to obtain the middleware executable environment.

[0087] S39. Based on the LangGraph workflow framework, the standard risk execution pipeline definition file is generated into a workflow instance that runs on the middleware executable environment and the instance is coded to serve as the business risk executable program for the target demand party.

[0088] S310. Feed back the executable procedure for business risks to the target requester for invocation and execution.

[0089] Example 4 Figure 4 This is a schematic diagram of a business risk executable program generation device provided in Embodiment 4 of the present invention. The business risk executable program generation device provided in this embodiment of the present invention is applicable to situations in the financial field where risk decisions are made for businesses such as credit. This business risk executable program generation device can be implemented in hardware and / or software, such as... Figure 4 As shown, the device includes: a request parsing module 401, a sample data generation module 402, a program generation module 403, and a program feedback module 404. Among them, The request parsing module 401 is used to respond to the business risk program generation request from the target requester, parse the business risk program generation request, and determine the risk business type. The sample data generation module 402 is used to generate a risk reference sample dataset based on the risk business type and a pre-defined sample feature dimension. The program generation module 403 is used to generate an executable program for the business risk of the target demander based on the risk reference sample dataset and the risk business type. The program feedback module 404 is used to feed back the executable program for business risks to the target demand party for the target demand party to call and execute.

[0090] The technical solution of this invention parses the business risk program generation request to determine the risk business type. Based on the risk business type and pre-defined sample feature dimensions, a risk reference sample dataset is generated. Based on the risk reference sample dataset and the risk business type, a business risk executable program for the target requester is generated and fed back to the target requester for execution. This technical solution utilizes a main intelligent agent implemented based on a super-intelligent agent framework to handle request response and parsing, and automatically constructs the risk reference sample dataset. The risk business type is considered during dataset construction, ensuring the accuracy and efficiency of the risk reference sample dataset generation for the corresponding risk business type. The automated generation of the business risk executable program based on the risk reference sample dataset and risk business type improves the accuracy and reliability of the generated program, thereby enhancing the efficiency, accuracy, and reliability of subsequent business risk decisions in financial scenarios based on this executable program.

[0091] Optionally, the program generation module 403 includes: The subtask determination unit is used to perform subtask matching on a preset subtask combination template based on the risk business type, determine at least one target subtask required to execute the business risk program generation request, and the task execution relationship between each target subtask. The standard file generation unit is used to generate a standard risk execution pipeline definition file based on the risk reference sample dataset, according to each of the target sub-tasks and the task execution relationship between each of the target sub-tasks. The program generation unit is used to generate an executable program for the business risks of the target demander based on the standard risk execution pipeline definition file.

[0092] Optional, standard document generation units include: The DAG basic structure generation sub-unit is used to generate a directed acyclic graph (DAG) basic topology structure based on each target subtask and the task execution relationship between each target subtask; the DAG basic topology structure file includes the original node objects and their corresponding node edges. A dedicated agent generation subunit is used to determine the dedicated agent for task execution and its corresponding agent constraints for each of the target sub-tasks. The DAG intermediate structure generation subunit is used to update the original node objects and their corresponding node edges of the DAG basic topology structure according to the task execution dedicated agent and its corresponding agent constraint conditions corresponding to each of the target sub-tasks, so as to obtain the DAG intermediate topology structure. DAG target structure generation sub-unit is used to update the intermediate topology of the DAG using the risk reference sample dataset to obtain the target topology of the DAG. The standard file generation subunit is used to instantiate the DAG target topology and generate a standard risk execution pipeline definition file.

[0093] Optionally, the DAG target structure generation sub-unit is specifically used for: Generate engineering prompt information including the priority ranking and computing resource allocation mechanism of each target subtask, and generate task priority ranking results and computing resource allocation results corresponding to each target subtask based on the engineering prompt information; The intermediate topology of the DAG is updated using the task priority ranking results and computing resource allocation results corresponding to each of the target sub-tasks to obtain the DAG reference topology. The DAG reference topology is updated using the risk reference sample dataset to obtain the DAG target topology.

[0094] Optionally, the program generation unit includes: An executable environment generation subunit is used to generate a middleware executable environment based on the standard risk execution pipeline definition file. The program generation subunit is used to generate the standard risk execution pipeline definition file into a workflow instance that runs on the middleware executable environment based on the pre-selected intelligent agent graph workflow framework, and to encode the instance. The compiled workflow instance is used as the business risk executable program of the target demand party.

[0095] Optional, executable environment generation subunit, specifically used for: The standard risk execution pipeline definition file is parsed to determine the capability requirement tags; The middleware components that perform the corresponding functions are determined based on the capability requirement tags, and a middleware processing chain is generated according to the execution order of each middleware component. Middleware configuration parameters are injected into the middleware processing chain to generate a middleware executable environment.

[0096] Optionally, the sample data generation module 402 is specifically used for: Based on the risk business type, generate feature reference data under the corresponding sample feature dimensions; the sample feature dimensions include the general business rule dimension for the scenario, the risk assessment standard feature dimension, and the model output reliability verification dimension. A risk reference sample dataset is generated based on the feature reference data of each of the aforementioned sample feature dimensions.

[0097] The business risk executable program generation apparatus provided in this embodiment of the invention can execute the business risk executable program generation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0098] Example 5 Figure 5 A schematic diagram of an electronic device 50 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0099] like Figure 5 As shown, the electronic device 50 includes at least one processor 51 and a memory, such as a read-only memory (ROM) 52 and a random access memory (RAM) 53, communicatively connected to the at least one processor 51. The memory stores computer programs executable by the at least one processor. The processor 51 can perform various appropriate actions and processes based on the computer program stored in the ROM 52 or loaded from storage unit 58 into the RAM 53. The RAM 53 can also store various programs and data required for the operation of the electronic device 50. The processor 51, ROM 52, and RAM 53 are interconnected via a bus 54. An input / output (I / O) interface 55 is also connected to the bus 54.

[0100] Multiple components in electronic device 50 are connected to I / O interface 55, including: input unit 56, such as keyboard, mouse, etc.; output unit 57, such as various types of monitors, speakers, etc.; storage unit 58, such as disk, optical disk, etc.; and communication unit 59, such as network card, modem, wireless transceiver, etc. Communication unit 59 allows electronic device 50 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0101] Processor 51 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 51 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 51 performs the various methods and processes described above, such as methods for generating business risk executable programs.

[0102] In some embodiments, the method for generating a business risk executable program may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 58. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 50 via ROM 52 and / or communication unit 59. When the computer program is loaded into RAM 53 and executed by processor 51, one or more steps of the method for generating a business risk executable program described above may be performed. Alternatively, in other embodiments, processor 51 may be configured to execute the method for generating a business risk executable program by any other suitable means (e.g., by means of firmware).

[0103] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0104] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0105] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0106] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0107] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0108] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0109] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0110] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for generating an executable program for business risks, characterized in that, A master agent applied to the business risk program generation agent, wherein the business risk program generation agent is implemented based on a super agent runtime framework, the method comprising: In response to a business risk procedure generation request from a target demander, the business risk procedure generation request is parsed to determine the type of risky business. Based on the risk business type, a risk reference sample dataset is generated using pre-defined sample feature dimensions. Based on the risk reference sample dataset and the risk business type, generate an executable program for the business risk of the target requester; The executable program for the business risk is fed back to the target requester for execution.

2. The method according to claim 1, characterized in that, The step of generating an executable program for the business risk of the target client based on the risk reference sample dataset and the risk business type includes: Based on the risk business type, subtask matching is performed on the preset subtask combination template to determine at least one target subtask required to execute the business risk program generation request, as well as the task execution relationship between each target subtask. Based on the target sub-tasks and the task execution relationships between them, a standard risk execution pipeline definition file is generated based on the risk reference sample dataset. Based on the standard risk execution pipeline definition file, generate the business risk executable program for the target demand party.

3. The method according to claim 2, characterized in that, The step of generating a standard risk execution pipeline definition file based on the risk reference sample dataset, according to each of the target sub-tasks and the task execution relationships between them, includes: Based on each target subtask and the task execution relationship between each target subtask, a directed acyclic graph (DAG) basic topology structure is generated; the DAG basic topology structure file includes the original node objects and their corresponding node edges. Determine the task execution agent and its corresponding agent constraints for each of the target sub-tasks; Based on the task execution agent and its corresponding agent constraints corresponding to each target subtask, the original node objects and their corresponding node edges of the DAG basic topology are updated to obtain the intermediate DAG topology. The intermediate topology of the DAG is updated using the risk reference sample dataset to obtain the target topology of the DAG; The DAG target topology is instantiated to generate a standard risk execution pipeline definition file.

4. The method according to claim 3, characterized in that, The step of updating the intermediate topology of the DAG using the risk reference sample dataset to obtain the target topology of the DAG includes: Generate engineering prompt information including the priority ranking and computing resource allocation mechanism of each target subtask, and generate task priority ranking results and computing resource allocation results corresponding to each target subtask based on the engineering prompt information; The intermediate topology of the DAG is updated using the task priority ranking results and computing resource allocation results corresponding to each of the target sub-tasks to obtain the DAG reference topology. The DAG reference topology is updated using the risk reference sample dataset to obtain the DAG target topology.

5. The method according to claim 2, characterized in that, The step of generating an executable program for the target client's business risks based on the standard risk execution pipeline definition file includes: Based on the aforementioned standard risk execution pipeline definition file, a middleware executable environment is generated; Based on the pre-selected intelligent agent graph workflow framework, the standard risk execution pipeline definition file is generated into a workflow instance that runs on the middleware executable environment and the instance is coded. The compiled workflow instance is used as the business risk executable program of the target demand party.

6. The method according to claim 5, characterized in that, The middleware executable environment generated based on the standard risk execution pipeline definition file includes: The standard risk execution pipeline definition file is parsed to determine the capability requirement tags; The middleware components that perform the corresponding functions are determined based on the capability requirement tags, and a middleware processing chain is generated according to the execution order of each middleware component. Middleware configuration parameters are injected into the middleware processing chain to generate a middleware executable environment.

7. The method according to claim 1, characterized in that, The step of generating a risk reference sample dataset based on the risk business type and pre-defined sample feature dimensions includes: Based on the risk business type, generate feature reference data under the corresponding sample feature dimensions; the sample feature dimensions include the general business rule dimension for the scenario, the risk assessment standard feature dimension, and the model output reliability verification dimension. A risk reference sample dataset is generated based on the feature reference data of each of the aforementioned sample feature dimensions.

8. An apparatus for generating a business risk executable program, characterized in that, A master agent applied to a business risk program generation agent, the business risk program generation agent being implemented based on a super agent runtime framework, the device comprising: The request parsing module is used to respond to the business risk program generation request from the target demander, parse the business risk program generation request, and determine the risk business type. The sample data generation module is used to generate a risk reference sample dataset based on the risk business type and pre-defined sample feature dimensions. The program generation module is used to generate an executable program for the business risk of the target requester based on the risk reference sample dataset and the risk business type. The program feedback module is used to feed back the executable program for the business risk to the target requester, so that the target requester can call and execute it.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the method for generating a business risk executable program according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method for generating a business risk executable program according to any one of claims 1-7.

Images