Service auditing method and device based on supply chain, equipment, medium and program product

Abstract

The invention provides a service auditing method and device based on a supply chain, equipment, a medium and a program product, can be applied to the field of artificial intelligence and the field of financial science and technology, and relates to application of a large model in a service auditing scene. The method comprises the steps of submitting a service application in response to an application subject, and obtaining application data corresponding to the service application; constructing a federated learning framework according to the participant and the application subject, and collecting multi-source data based on the federated learning framework; wherein a supply chain exists between the application subject and the participating subject; based on the multi-source data, performing primary review processing on the application data by adopting a multi-agent collaborative strategy to obtain a primary review result; under the condition that the primary review result is passed, the application data is rechecked through a risk conduction simulation engine and an anomaly detection engine based on the multi-source data, and a recheck result is obtained; and generating an auditing result based on the primary auditing result or the rechecking result, and sending the auditing result to the application subject.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and fintech, and to the application of large models in business auditing scenarios. More specifically, it relates to a supply chain-based business auditing method, apparatus, equipment, media, and program products. Background Technology

[0002] Against the backdrop of accelerated digital transformation in industries, supply chain finance, as a key path to solving the financing difficulties of SMEs, is seeing increasingly diverse participants and a continuously extending transaction chain. However, the technological shortcomings of traditional models are becoming increasingly apparent. The core issue in supply chain management lies in the deep-rooted information asymmetry. SMEs have low levels of IT literacy and a weak willingness to disclose information, making it difficult for financial institutions to fully grasp their true operating and transaction conditions. Coupled with the highly concealed nature of abnormal behaviors such as fraudulent trade and double pledging, this directly leads to a sharp increase in the difficulty of risk control.

[0003] In the existing business audit services of the banking supply chain, it is difficult for SMEs to collect and standardize data, information asymmetry leads to difficulties in risk control, the audit process is inefficient, and the problem of data silos is prominent. Existing applications mostly focus on optimizing a single link in the audit and lack the ability to coordinate the entire process. Summary of the Invention

[0004] In view of the above problems, this application provides supply chain-based business auditing methods, apparatus, equipment, media and program products to improve the intelligence of auditing.

[0005] According to a first aspect of this application, a supply chain-based business audit method is provided, comprising: in response to an applicant submitting a business application, obtaining application data corresponding to the business application; constructing a federated learning framework based on participating entities and the applicant, and collecting multi-source data based on the federated learning framework; wherein a supply chain exists between the applicant and the participating entities; performing preliminary audit processing on the application data based on the multi-source data using a multi-agent collaborative strategy to obtain a preliminary audit result; if the preliminary audit result is passed, reviewing the application data based on the multi-source data using a risk transmission simulation engine and an anomaly detection engine to obtain a review result; and generating an audit result based on the preliminary audit result or the review result, and sending the audit result to the applicant.

[0006] According to an embodiment of this application, the multi-agent collaborative strategy includes: transmitting task review instructions between target agents through a message queue based on a distributed agent architecture; wherein the target agents include one or more of a document parsing agent, a logistics sensing agent, a product evaluation agent, a pricing engine agent, and a decision coordination agent; and synchronizing the initial review processing status of the target agents using a central cluster.

[0007] According to an embodiment of this application, the preliminary review of the application data based on the multi-source data and employing a multi-agent collaborative strategy to obtain a preliminary review result includes: parsing the application data through the document parsing agent and verifying the information of the multi-source data and the parsed application data to obtain an information verification result; extracting logistics information of the application data through the logistics sensing agent based on the multi-source data and reviewing the logistics information to obtain a logistics review result; conducting a product risk assessment of the application data through the product evaluation agent based on the multi-source data to obtain a product evaluation result; calculating a pricing reference result by calling the multi-source data through the pricing engine agent based on the pricing information of the application data; and analyzing the multi-dimensional review results through the decision coordination agent to generate the preliminary review result; wherein the multi-dimensional review result includes one or more of the information verification result, the logistics review result, the product evaluation result, and the pricing reference result.

[0008] According to an embodiment of this application, the risk transmission simulation engine includes: constructing a dynamic supply chain relationship graph with the participating entities and the applicant entities as graph nodes and business relationships as graph edges; and simulating risk propagation paths based on the dynamic supply chain relationship graph and the multi-source data through a risk assessment agent, and assessing the specific business risks of the application data based on the risk propagation results.

[0009] According to an embodiment of this application, the anomaly detection engine includes: scanning the supply chain dynamic correlation graph to identify abnormal correlation subgraphs; and detecting the abnormal correlation subgraphs based on the multi-source data using a pre-trained decision tree model to obtain anomaly risk results.

[0010] According to an embodiment of this application, the step of constructing a federated learning framework based on the participating entities and the applicant entities includes: constructing federated nodes based on the participating entities and the applicant entities; setting security rules for transmission parameters between the federated nodes and the central coordination node according to a data encryption protocol; and connecting the federated nodes to the corresponding node database based on a preset data interface to construct the federated learning framework.

[0011] According to an embodiment of this application, the step of collecting multi-source data based on the federated learning framework includes: extracting business data and standardizing the business data based on the node database accessed by the federated node in the federated node; performing homomorphic encryption on the standardized business data in the federated node to obtain encrypted feature data, and uploading the encrypted feature data to the central coordination node; and aggregating the encrypted feature data through the central coordination node to obtain the multi-source data.

[0012] A second aspect of this application provides a supply chain-based business auditing device, comprising: a business application module, configured to obtain application data corresponding to the business application in response to a business application submitted by an applicant; a data acquisition module, configured to construct a federated learning framework based on participating entities and the applicant, and to collect multi-source data based on the federated learning framework; wherein a supply chain exists between the applicant and the participating entities; a preliminary review processing module, configured to perform preliminary review processing on the application data based on the multi-source data using a multi-agent collaborative strategy to obtain a preliminary review result; a review processing module, configured to review the application data based on the multi-source data using a risk transmission simulation engine and an anomaly detection engine if the preliminary review result is passed, to obtain a review result; and a result feedback module, configured to generate an audit result based on the preliminary review result or the review result, and send the audit result to the applicant.

[0013] A third aspect of this application provides an electronic device comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0014] A fourth aspect of this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0015] The fifth aspect of this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0016] In this embodiment, a federated learning framework is constructed to collect multi-source data from the supply chain. Combined with multi-agent collaborative preliminary review, risk transmission simulation and anomaly detection engine review, the entire process review is completed. This effectively overcomes the data silo problem, solves the information asymmetry problem, improves risk control accuracy and risk identification coverage, significantly reduces review and approval time, optimizes business efficiency, improves the intelligence level of business review, and reduces operation and data collaboration costs. Attached Figure Description

[0017] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0018] Figure 1 The illustrations depict application scenarios of supply chain-based business auditing methods, apparatuses, devices, media, and program products according to embodiments of this application.

[0019] Figure 2A flowchart illustrating a supply chain-based business audit method according to an embodiment of this application is shown schematically.

[0020] Figure 3 The diagram illustrates the construction flowchart of a federated learning framework for a supply chain-based business auditing method according to an embodiment of this application.

[0021] Figure 4 This illustration schematically shows a multi-source data acquisition flowchart of a supply chain-based business audit method according to an embodiment of this application;

[0022] Figure 5 This illustration schematically shows a preliminary review process flowchart of a supply chain-based business review method according to an embodiment of this application;

[0023] Figure 6 This schematically illustrates a structural block diagram of a supply chain-based business auditing apparatus according to an embodiment of this application; and

[0024] Figure 7 A block diagram schematically illustrates an electronic device suitable for implementing a supply chain-based business auditing method according to an embodiment of this application. Detailed Implementation

[0025] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0026] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0027] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0028] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0029] Supply chain finance, as a core tool for solving the financing difficulties of SMEs, has long faced three major pain points: First, information asymmetry leads to difficulties in risk control, with traditional manual due diligence only covering 30% of supply chain relationships and static credit models having an error rate as high as 25%; second, low process efficiency, with procedures such as document verification taking up to 72 hours and financing approval cycles generally exceeding 14 days; and third, the problem of data silos is prominent, with core enterprises, logistics providers, and banks finding it difficult to securely share data, which restricts the comprehensiveness of risk assessment.

[0030] Information and data are scattered across multiple entities, including core enterprises, financial institutions, and logistics providers, forming heavily isolated data silos. This results in high costs for cross-entity verification, severely hindering process efficiency. Existing Artificial Intelligence (AI) applications are limited to optimizing single aspects such as invoice recognition and individual credit scoring, lacking the ability to integrate data across the entire supply chain and failing to achieve collaborative risk control across capital flows, logistics, and information flows. While open-source financial AI platforms have achieved data integration, and multi-agent transaction frameworks demonstrate the value of role division, none have been technically adapted for banking supply chain finance scenarios, exhibiting significant shortcomings, particularly in cross-entity risk transmission analysis and privacy-preserving data utilization. Therefore, there is an urgent need to build a full-chain intelligent collaborative system to overcome the technical bottlenecks in supply chain audit services.

[0031] This application provides a supply chain-based business review method, comprising: in response to a business application submitted by an applicant, obtaining application data corresponding to the business application; constructing a federated learning framework based on participating entities and the applicant, and collecting multi-source data based on the federated learning framework; wherein a supply chain exists between the applicant and participating entities; performing preliminary review of the application data based on the multi-source data using a multi-agent collaborative strategy to obtain a preliminary review result; if the preliminary review result is satisfactory, reviewing the application data based on the multi-source data using a risk transmission simulation engine and an anomaly detection engine to obtain a review result; and generating an review result based on the preliminary review result or the review result, and sending the review result to the applicant. In this application, the federated learning framework enables the collection of multi-source supply chain data, and the combined multi-agent collaborative preliminary review, risk transmission simulation, and anomaly detection engine review completes the entire review process. This effectively overcomes the data silo problem, solves the information asymmetry problem, improves risk control accuracy and risk identification coverage, significantly reduces review and approval time, optimizes business efficiency, improves the intelligence level of business review, and reduces operational and data collaboration costs.

[0032] It should be noted that the supply chain-based business audit method and apparatus of this application can be used in the fields of artificial intelligence and fintech, as well as in any field other than artificial intelligence and fintech, involving the application of large models in business audit scenarios. The application fields of the supply chain-based business audit method and apparatus of this application are not limited.

[0033] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0034] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this application all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0035] Figure 1 The illustration shows an application scenario of a supply chain-based business audit method, apparatus, device, medium, and program product according to embodiments of this application.

[0036] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0037] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0038] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0039] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0040] It should be noted that the supply chain-based business auditing method provided in this application embodiment can generally be executed by server 105. Correspondingly, the supply chain-based business auditing device provided in this application embodiment can generally be located in server 105. The supply chain-based business auditing method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the supply chain-based business auditing device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0041] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0042] The following will be based on Figure 1 The described scene, through Figures 2-5 A detailed description is provided of the supply chain-based business audit method according to embodiments of this application.

[0043] Figure 2 A flowchart illustrating a supply chain-based business audit method according to an embodiment of this application is shown.

[0044] like Figure 2 As shown, the supply chain-based business auditing method of this embodiment includes operations S210 to S250. This supply chain-based business auditing method is not limited to a specific executing entity. The executing entity can be any electronic device, such as a terminal device or a server device, etc. The executing entity can also be any software application or client.

[0045] In operation S210, in response to the applicant submitting a business application, the application data corresponding to the business application is obtained.

[0046] The applicant is the core responsible entity that initiates a specific business application, such as a manufacturing enterprise submitting a financing application in the supply chain or an e-commerce platform initiating a logistics cooperation. It is the triggerer of the business process and the proposer of the demand.

[0047] Applicants can submit business applications through the unified system / platform portal. They can choose from various application types, such as financing, credit, discounting, account opening, changes, data integration, business changes, access, and qualification filing, and upload corresponding materials. For example, access applications require the submission of the company's business license and industry qualifications, discounting applications require the submission of the original bills and transaction contracts, and data integration applications require the submission of the integration requirements specification and interface specifications.

[0048] After receiving a business application, the system / platform uses natural language processing technology to identify the application type and extracts basic application data such as the applicant's name, the type of business applied for, explanatory materials, and supporting materials in order to trigger the review process.

[0049] In operating S220, a federated learning framework is constructed based on the participating entities and the applicant entities, and multi-source data is collected based on the federated learning framework; among them, there is a supply chain between the applicant entities and the participating entities.

[0050] Participating entities are related entities that are driven by the applicant's needs and participate in the business process collaboration. In the supply chain scenario, they typically include suppliers, logistics service providers, warehousing companies, financial institutions, etc., and need to cooperate with the applicant to complete order review, order fulfillment, cargo transportation, fund settlement and other links.

[0051] The participating entities and the applicant entities are bound together through supply chain relationships. The applicant entity initiates an application based on production or sales needs, and the participating entities provide corresponding services based on the division of labor in the supply chain. Together, they complete the entire process from the proposal of needs to the implementation of business. The applicant entity bears the primary responsibility for the business outcome, and the participating entities fulfill their collaborative obligations according to their division of labor.

[0052] Within the federated learning framework, a cross-entity data access federated node is constructed, encompassing multiple data sources from participating / applicant entities in the supply chain. These include core Enterprise Resource Planning (ERP) transaction data sources, IoT sensing data sources from logistics companies (such as transportation trajectories and warehouse videos), and internal bank credit data sources. Based on the federated learning framework, data from the corresponding data sources of participating / applicant entities is collected, acquiring historical transaction data, logistics fulfillment data, credit records, and other multi-source data from each entity. Homomorphic encryption within the federated learning framework ensures data is usable but not visible, providing multi-dimensional data support for subsequent review. Only model parameters, not raw data, are transmitted between nodes, resolving the conflict between data silos and privacy protection. Referencing open-source platform data source integration solutions, over a hundred free and compliant data source interfaces can be integrated, supporting Python secondary development and integration with internal enterprise databases to obtain multi-source data.

[0053] In operation S230, based on multi-source data, a multi-agent collaborative strategy is used to perform preliminary review of the application data and obtain the preliminary review results.

[0054] A distributed intelligent agent architecture is adopted, setting up multiple types of professional intelligent agents (such as document parsing intelligent agents, logistics perception intelligent agents, product evaluation intelligent agents, pricing engine intelligent agents, and decision coordination intelligent agents, etc.), and using multi-agent collaboration strategies, collaborative analysis is carried out for the core review requirements of different business applications to generate preliminary review results.

[0055] The process of building a distributed intelligent agent architecture is as follows: First, the intelligent agents are split into multiple types based on functional layers (perception-decision-execution), such as the perception layer corresponding to logistics data. Second, nodes are deployed using containerized deployment of each intelligent agent, and cross-node communication is achieved through general remote calls. Finally, a fault tolerance mechanism is configured to monitor the status of the intelligent agents and avoid single points of failure through master-slave backup.

[0056] It is worth noting that intelligent agents are modular software entities with a large language model as their core brain. They can perceive the environment, make autonomous plans and decisions, call tools to execute tasks, and can cooperate with other intelligent agents of the same kind to process tasks and achieve specific goals. Intelligent agents are key carriers for the implementation of large models.

[0057] Each intelligent agent is trained using fine-tuning methods. For example, an intelligent agent generator automatically generates training tasks based on multi-source historical data (such as simulating different business applications and logistics scenarios). Then, based on AI multimodal technology, the intelligent agents are fine-tuned using real business data (such as financing data and logistics trajectories). The collaborative capabilities of each intelligent agent are gradually improved through task evolution (such as from simple orders to complex supply chains).

[0058] When operating S240, if the initial review result is passed, the application data is reviewed based on multi-source data, and the risk transmission simulation engine and anomaly detection engine are used to obtain the review result.

[0059] If the initial review result is unsuccessful, a rejection reason will be automatically generated for the unsuccessful application (such as incomplete materials or excessive risk).

[0060] If the initial review result is passed, the risk transmission simulation engine and the anomaly detection engine will be activated for the approved application for review. The two engines will carry out in-depth review in parallel to realize dynamic risk control and provide final security guarantee for the review conclusion.

[0061] During the review process, hidden abnormal subgraphs with implicit connections in the graph are identified to capture hidden risks such as guarantee chains and nominee shareholding. A self-evolving decision tree is established, and the model parameters are iterated based on tens of thousands of transaction data iterations to improve the accuracy of risk warnings. It can access public opinion and industry policy data, and update risk weights in real time through natural language processing. For example, risk events can be extracted through natural language processing (such as "a certain industry has introduced restrictive policies"), and the risk weights of corresponding enterprises / industries can be automatically adjusted (e.g., the weight of the policy-restricted industry increases by 0.2).

[0062] In operation S250, an audit result is generated based on the preliminary review result or the review result, and the audit result is sent to the applicant.

[0063] If the review results generated by the two engines are complementary, the in-depth review will find no anomalies, the review will be confirmed as passed, and the review result will be "passed." If either engine detects a risk, the decision-making and coordination agent will regenerate the review opinion (such as rejection or adjustment of application conditions), and the review result will be "disapproved." The review results will be automatically fed back to the applicant. If the review is passed, the applicant will be informed of the subsequent procedures simultaneously. If the review is failed, the reasons for rejection and suggestions for supplementary materials will be provided simultaneously.

[0064] By connecting to the bank's core system, it achieves zero human intervention throughout the entire process, from business application actions (such as financing application, document review, and risk assessment) to the generation of loan disbursement instructions. Addressing the splittable nature of supply chain invoices, it has developed intelligent splitting and circulation tracking functions, supporting the splitting of million-yuan invoices into multiple small-amount payment vouchers, adapting to the high-frequency, small-amount financing needs of SMEs.

[0065] For example, for financing applications, a loan disbursement instruction (including corporate account, amount, and interest rate) is generated and sent to the bank's core system to trigger fund transfer. For bill financing, intelligent splitting (breaking large bills into smaller payment vouchers) and circulation tracking (recording the circulation path through blockchain) are carried out simultaneously. For credit applications, a credit approval document is generated and synchronized to the bank's credit system to complete the credit limit entry. For discount applications, a discount approval document is generated and sent to the bank's core system to complete the discount fund transfer.

[0066] We continuously monitor the transaction data, public opinion, and policy changes of applicant entities, iterate the decision tree model parameters weekly, update risk weights, and monitor the use of funds and the circulation of bills in real time for financing, credit, discounting and other businesses. If an abnormal warning is triggered, we automatically trigger collection or risk control adjustment processes.

[0067] In the embodiments of this application, the federated learning framework is used to collect multi-source data from the supply chain. Combined with multi-agent collaborative preliminary review, risk transmission simulation and anomaly detection engine review, the entire process review is completed. This effectively breaks through the data silo problem, solves the problem of information asymmetry, improves the accuracy of risk control and the coverage of risk identification, significantly reduces the review and approval time, optimizes business efficiency, improves the intelligence level of business review, and reduces the cost of operation and data collaboration.

[0068] Figure 3 The diagram illustrates the construction flowchart of a federated learning framework for a supply chain-based business auditing method according to an embodiment of this application.

[0069] like Figure 3 As shown, the federal learning framework is constructed based on the participating entities and the applicant entities, including operations S310 to S330.

[0070] When operating S310, a federated node is constructed based on the participating entities and the applicant entities.

[0071] Deploy federated nodes among participating / applicant entities such as core enterprises, logistics companies, and banks. Each federated node includes modules such as data access and encrypted computing, and is configured with a central coordination node.

[0072] When operating the S320, security rules are set for the transmission parameters between the federated nodes and the central coordination node according to the data encryption protocol.

[0073] Configure the privacy protocol, select the homomorphic encryption algorithm as the data encryption protocol, and set the security rules for model parameters / data transmission.

[0074] When operating the S330, federated nodes are connected to the corresponding node database based on the preset data interface in order to build a federated learning framework.

[0075] By using open-source data interfaces, we can adapt data interfaces and connect to multi-source node databases (such as database tables) from ERP, IoT, and credit sectors.

[0076] In the embodiments of this application, a federated node is built based on the supply chain participants and the applicant, data security transmission rules are formulated in combination with data encryption protocols, and a federated learning framework is constructed by accessing the node database through a preset interface. Under the premise of strictly protecting data privacy and security, data silos are broken down, which facilitates the compliant collection of multi-source data across entities, provides a data infrastructure framework for full-process review, reduces the cost of cross-institutional data collaboration, and improves the efficiency of data application.

[0077] like Figure 4 As shown, multi-source data is collected based on the federated learning framework, including operations S410 to S430.

[0078] When operating the S410, in the federated nodes, business data is extracted and standardized based on the node database accessed by the federated nodes.

[0079] Each participating / applicant entity accesses its business data (core enterprise ERP, logistics IoT data, bank credit data) at the local federated node and standardizes the format.

[0080] When operating S420, in the federated nodes, the standardized business data is homomorphically encrypted to obtain encrypted feature data, and the encrypted feature data is uploaded to the central coordination node.

[0081] Local federated nodes perform homomorphic encryption on standardized business data and only upload the encrypted data features to the central coordination node of the federated learning framework.

[0082] Homomorphic encryption is the security foundation of federated data collection, enabling joint collection of data from multiple sources without exposing the original content. This achieves data sharing while resolving the conflict between data silos and privacy protection.

[0083] When operating the S430, encrypted feature data is aggregated through the central coordination node to obtain multi-source data.

[0084] The central coordinating node aggregates the encryption features of each participating / applying entity to generate a usable but invisible joint dataset, i.e., multi-source data.

[0085] Historical multi-source data can be used for agent and model training. After participating / applicant entities train the model locally, they homomorphically encrypt the model parameters (such as weights and gradients) using the public key of the federated framework. Only the encrypted parameters are transmitted to the central node to avoid leakage of the original parameters. The central coordination node performs ciphertext operations (such as weighted averaging) on ​​the encrypted parameters to generate global model parameters, and then decrypts them with its private key and distributes them to each participating / applicant entity.

[0086] In the embodiments of this application, business data extraction and standardization are completed at the federated node, encrypted feature data is generated through homomorphic encryption, and uploaded to the central coordination node, which aggregates the data to form multi-source data. This not only achieves efficient aggregation and unified organization of cross-entity data in the supply chain, but also relies on encryption technology to ensure data privacy and security, making the data usable but invisible, resolving the contradiction between data silos and privacy protection, and significantly reducing the cost of cross-organizational collaboration.

[0087] According to embodiments of this application, a multi-agent collaborative strategy includes: transmitting task review instructions between target agents through a message queue based on a distributed agent architecture; wherein the target agents include one or more of a document parsing agent, a logistics sensing agent, a product evaluation agent, a pricing engine agent, and a decision coordination agent; and synchronizing the initial review processing status of the target agents using a central cluster.

[0088] Based on a distributed general-purpose toolkit, intelligent agents such as document parsing, logistics sensing, and product evaluation are broken down into atomic components and deployed on different distributed nodes. These agents communicate with each other via message queues (e.g., the document parsing agent transmits order information to the logistics sensing agent). A central cluster is built as a unified state-sharing center, enabling agents such as document parsing, risk assessment, and pricing engines to read, write, and synchronize the initial review and processing status of each other's tasks in real time, ensuring consistent progress / result perception across all agents for the same task. The central cluster synchronizes the initial review and processing status of each agent (e.g., order progress material review passed, logistics review completed, risk assessment completed), ensuring unified collaborative logic.

[0089] The preliminary review results, such as the material analysis results (analyzed application data), information verification results, logistics review results, and product evaluation results, are synchronized to the distributed cluster to ensure that each agent obtains a unified review status.

[0090] It should be noted that the target intelligent agent is selected from several types of intelligent agents, such as document parsing, logistics perception, product evaluation, pricing engine, and decision coordination, based on the specific content and requirements of the business application. For example, for a business application that includes product information and pricing, the product evaluation intelligent agent and the pricing engine intelligent agent will be selected as the target intelligent agents to complete the product qualification review and price reasonableness verification, respectively.

[0091] In the embodiments of this application, based on a distributed intelligent agent architecture, review instructions are transmitted through a message queue, multiple types of intelligent agents cooperate to carry out preliminary review, and the status is processed synchronously by a central cluster, so as to realize the efficient distribution and collaborative advancement of review tasks, greatly improve the efficiency of preliminary review, and ensure the uniformity and real-time nature of review process control.

[0092] Figure 5 The illustration shows a flowchart of the preliminary review process of a supply chain-based business review method according to an embodiment of this application.

[0093] like Figure 5 As shown, based on multi-source data, a multi-agent collaborative strategy is adopted to perform preliminary review of the application data and obtain the preliminary review results, including operations S510 to S550.

[0094] When operating S510, the application data is parsed by the document parsing intelligent agent, and the multi-source data and the parsed application data are verified to obtain the information verification result.

[0095] The document parsing agent is a smart agent specifically designed to process and verify various documents. Based on the multimodal processing technology of supply chain finance AI, it integrates computer vision and optical character recognition capabilities to complete the identification of the authenticity of bills and contracts and the extraction of key information within seconds, with an error rate of less than 0.1%.

[0096] The document parsing intelligent agent enables intelligent review of application materials across all categories, performing multi-type material parsing and cross-source information verification based on the material characteristics of different business applications.

[0097] The analysis of various document types includes qualification documents, transaction documents, technical documents, and account documents. Qualification documents (such as access applications and qualification filing applications) are analyzed to verify the authenticity of business licenses, industry permits, credit rating reports, etc., and to extract key information such as the company's registered address, registered capital, and qualification validity period. Transaction documents (such as financing applications and discounting applications) are analyzed to extract transaction amounts, payment terms, goods information, and logistics tracking information from invoices, contracts, and logistics documents. Technical documents (such as data integration applications) are analyzed to extract integration methods, data transmission scope, and security measures from interface specification documents and data security commitment letters. Account documents (such as account opening / change applications) are analyzed to extract account information and reasons for change from account opening permits, legal representative ID cards, and change application forms.

[0098] Cross-source information verification includes basic information verification and business-specific information verification. Basic information verification compares the parsed applicant information with core enterprise resource planning (ERP) data and bank credit data collected federally to confirm the authenticity of the enterprise's identity and its continued existence. Business-specific information verification conducts specific verifications for different business types. For example, discount applications require verification of the authenticity and validity of invoices and the authenticity of the transaction background; data integration applications require verification of the matching degree between the integration requirements and the platform's interface capabilities; and account change applications require verification of the reasonableness of the change reasons and the completeness of relevant supporting documents.

[0099] When operating the S520, based on multi-source data, the logistics information of the application data is extracted by the logistics sensing agent, and the logistics information is reviewed to obtain the logistics review result.

[0100] The logistics sensing intelligent agent extracts core information related to application data, such as logistics timeliness, transportation route, and carrier qualifications, based on multi-source data. It then calls preset verification rules to compare logistics information with industry standards and historical compliance data, and finally outputs the compliant / abnormal logistics audit results.

[0101] When operating the S530, based on multi-source data, the product evaluation agent performs product risk assessment on the application data to obtain product evaluation results.

[0102] The product evaluation AI, based on multi-source data, extracts core information from application data, including product category, specifications, market circulation, and past transaction records. It can connect to the bank's internal credit product database to retrieve reference indicators such as credit risk benchmarks and collateral valuation coefficients for similar products. Next, it conducts quantitative analysis from dimensions such as market volatility, compliance, and liquidity, comparing each indicator with benchmarks. Finally, it comprehensively calculates the product's risk level and generates a product evaluation result that includes risk point annotations and assessment criteria.

[0103] When operating the S540, based on the pricing information in the application data, the pricing engine agent calls up multi-source data to calculate the pricing reference result.

[0104] The pricing engine intelligent agent performs targeted pricing and cost calculation (applicable to financing, credit, discounting and other business applications). It calls multi-source data in real time to calculate pricing reference values, such as market interest rates, applicant risk values, logistics timeliness (for supply chain financing), bill term (for discounting), etc., and then dynamically calculates and generates pricing reference results such as financing interest rates, credit lines, and discount rates based on the pricing reference values.

[0105] For other services without pricing requirements (such as access applications and account opening applications), the pricing engine agent will not intervene, and other agents will complete the review.

[0106] When operating the S550, the decision-coordinating intelligent agent analyzes the multi-dimensional audit results to generate preliminary audit results; among them, the multi-dimensional audit results include one or more of the following: information verification results, logistics audit results, product evaluation results, and pricing reference results.

[0107] The decision-making and coordination intelligent agent integrates the results and generates preliminary review opinions. Based on a large-scale model cognitive engine, it integrates multi-dimensional review results, such as the information verification results of the document parsing intelligent agent, the logistics review results of the logistics perception intelligent agent, the product evaluation results of the product evaluation intelligent agent, and the pricing reference results of the pricing engine intelligent agent. It comprehensively analyzes these multi-dimensional review results to generate preliminary review results. For example, for applications for access and qualification registration, it determines whether the enterprise meets the access conditions and whether to approve the registration; for applications for financing, credit, and discounting, it determines whether to approve the application and clarifies the amount, interest rate, and term; for data integration applications, it determines whether to approve the integration and clarifies the integration plan and security requirements; and for applications for account opening / change and business change, it determines whether to approve the processing and clarifies the processing procedures and precautions.

[0108] In the embodiments of this application, each intelligent agent performs its own function, completing preliminary review actions such as document parsing and verification, logistics information review, product risk assessment, and pricing reference calculation. Then, the decision-coordination intelligent agent integrates the multi-dimensional results to generate a preliminary review conclusion, realizing full-dimensional and refined verification in the preliminary review process. This not only greatly improves the efficiency and accuracy of the preliminary review, but also improves the efficiency of approval and the accuracy of risk control.

[0109] According to an embodiment of this application, the risk transmission simulation engine includes: constructing a dynamic supply chain relationship graph with participating entities and applicant entities as graph nodes and business relationships as graph edges; and simulating risk propagation paths based on the dynamic supply chain relationship graph and multi-source data through a risk assessment agent, and assessing the specific business risks of the application data based on the risk propagation results.

[0110] By mapping each entity in the supply chain to a graph node and business relationships such as transactions, guarantees, and equity to graph edges, a dynamic supply chain relationship graph is constructed. The risk assessment agent, equipped with a self-supervised graph neural network, simulates the risk propagation path and scope of impact for approved applications based on this dynamic supply chain relationship graph (e.g., simulating the path and scope of risk spreading to other companies along the guarantee chain and transaction chain after a company defaults). This determines the scope of impact of defaults by related companies of the applicant and the impact of defaults by financing companies on financial institutions.

[0111] Risk assessment intelligence is used to evaluate specific business risks. For example, for financing, credit, and discounting applications, the main assessments are repayment ability, authenticity of transaction background, and bill risks. For access applications and qualification filing applications, the focus is on assessing the company's industry qualifications, credit status, and compliance. For data connection applications, the main assessments are data security risks and interface connection risks. For account opening / change and business change applications, the main assessments are the rationality of the change and whether there are any abnormal transaction risks.

[0112] In the embodiments of this application, a dynamic correlation graph is constructed based on the participants in the supply chain and the applicant. A risk assessment agent simulates the risk propagation path and assesses the risks of specific businesses, accurately capturing the risk transmission links between upstream and downstream of the supply chain, greatly improving the risk identification coverage and assessment accuracy, effectively solving the problem of information asymmetry, and consolidating the risk control defense line in the review stage.

[0113] According to an embodiment of this application, the anomaly detection engine includes: scanning the dynamic correlation graph of the supply chain to identify abnormal correlation subgraphs; and detecting abnormal correlation subgraphs based on multi-source data using a pre-trained decision tree model to obtain anomaly risk results.

[0114] Scanning the dynamic relationship map of the supply chain, linking it with dynamic risk detection, identifies hidden risks (such as circular guarantees and nominee shareholding) within the map, captures hidden risk points, and identifies hidden abnormal relationship subgraphs to ensure no hidden risks are missed. These subgraphs are marked with risk characteristics (such as nominee shareholding and hidden relationships) to achieve the capture of hidden risks.

[0115] At regular intervals (e.g., weekly), tens of thousands of transaction data points are used to incrementally train and update the decision tree model. During training, the branching rules of the decision tree (such as "transaction amount fluctuation threshold" and "related enterprise number threshold") are automatically adjusted to continuously optimize the accuracy of anomaly detection.

[0116] The trained decision tree model is invoked to detect abnormal features of different business types (such as fluctuations in transaction amounts of financing applications and abnormal legal person information in account applications), detect abnormal correlation subgraphs, and identify hidden abnormal risks.

[0117] In the embodiments of this application, abnormal subgraphs are locked by scanning the dynamic correlation graph of the supply chain, and abnormal risks are accurately detected by using a decision tree model in combination with multi-source data. This can efficiently identify hidden abnormal behaviors such as false trade and double pledging, and reduce the assessment error rate and bad debt rate.

[0118] Based on the aforementioned supply chain-based business auditing method, this application also provides a supply chain-based business auditing device. The following will combine... Figure 6 The device is described in detail.

[0119] Figure 6 The schematic diagram illustrates a structural block diagram of a supply chain-based business audit apparatus according to an embodiment of this application.

[0120] like Figure 6 As shown, the supply chain-based business review device 600 of this embodiment includes a business application module 610, a data acquisition module 620, a preliminary review processing module 630, a review processing module 640, and a result feedback module 650.

[0121] The business application module 610 is used to obtain the application data corresponding to the business application in response to the application subject submitting the business application. In one embodiment, the business application module 610 can be used to perform the operation S210 described above, which will not be repeated here.

[0122] The data acquisition module 620 is used to construct a federated learning framework based on the participating entities and the applicant entity, and to collect multi-source data based on the federated learning framework; wherein, a supply chain exists between the applicant entity and the participating entities. In one embodiment, the data acquisition module 620 can be used to perform the operation S220 described above, which will not be repeated here.

[0123] The preliminary review processing module 630 is used to perform preliminary review processing on the application data based on the multi-source data and employing a multi-agent collaborative strategy to obtain a preliminary review result. In one embodiment, the preliminary review processing module 630 can be used to execute the operation S230 described above, which will not be repeated here.

[0124] The review processing module 640 is used to review the application data based on the multi-source data, using a risk transmission simulation engine and an anomaly detection engine, to obtain a review result if the initial review result is satisfactory. In one embodiment, the review processing module 640 can be used to perform the operation S240 described above, which will not be repeated here.

[0125] The result feedback module 650 is used to generate an audit result based on the preliminary review result or the review result, and send the audit result to the applicant. In one embodiment, the result feedback module 650 can be used to perform the operation S250 described above, which will not be repeated here.

[0126] According to an embodiment of this application, the preliminary review processing module 630 includes: an intelligent agent coordination unit, used to transmit task review instructions between target intelligent agents through a message queue based on a distributed intelligent agent architecture; wherein the target intelligent agents include one or more of a document parsing intelligent agent, a logistics perception intelligent agent, a product evaluation intelligent agent, a pricing engine intelligent agent, and a decision coordination intelligent agent; and to synchronize the preliminary review processing status of the target intelligent agents using a central cluster.

[0127] According to an embodiment of this application, the preliminary review processing module 630 further includes: a document parsing unit, used to parse the application data through the document parsing intelligent agent, and to perform information verification on the multi-source data and the parsed application data to obtain an information verification result; a logistics sensing unit, used to extract logistics information of the application data based on the multi-source data through the logistics sensing intelligent agent, and to review the logistics information to obtain a logistics review result; a product evaluation unit, used to perform product risk assessment on the application data based on the multi-source data through the product evaluation intelligent agent to obtain a product evaluation result; a pricing unit, used to calculate a pricing reference result based on the pricing information of the application data by calling the multi-source data through the pricing engine intelligent agent; and a decision coordination unit, used to analyze the multi-dimensional review results through the decision coordination intelligent agent to generate the preliminary review result; wherein the multi-dimensional review result includes one or more of the information verification result, the logistics review result, the product evaluation result, and the pricing reference result.

[0128] According to an embodiment of this application, the review processing module 640 includes: a risk transmission simulation unit, used to construct a dynamic supply chain association graph with the participating entities and the applicant entity as graph nodes and business relationships as graph edges; and to simulate risk propagation paths based on the dynamic supply chain association graph and the multi-source data through a risk assessment agent, and to assess the specific business risks of the application data based on the risk propagation results.

[0129] According to an embodiment of this application, the review processing module 640 includes: an anomaly detection unit, used to scan the supply chain dynamic correlation graph to identify abnormal correlation subgraphs; and based on the multi-source data, to detect the abnormal correlation subgraphs through a pre-trained decision tree model to obtain an anomaly risk result.

[0130] According to an embodiment of this application, the data acquisition module 620 includes: a node construction unit, used to construct a federated node based on the participating entity and the applicant entity; a rule setting unit, used to set security rules for the transmission parameters between the federated node and the central coordination node according to a data encryption protocol; and a database access unit, used to connect the federated node to the corresponding node database based on a preset data interface to construct the federated learning framework.

[0131] According to an embodiment of this application, the data acquisition module 620 further includes: a data extraction unit, used to extract business data and standardize the business data based on the node database accessed by the federated node; a homomorphic encryption unit, used to perform homomorphic encryption on the standardized business data in the federated node to obtain encrypted feature data, and upload the encrypted feature data to the central coordination node; and a data aggregation unit, used to aggregate the encrypted feature data through the central coordination node to obtain the multi-source data.

[0132] According to embodiments of this application, any multiple modules among the business application module 610, data acquisition module 620, preliminary review processing module 630, review processing module 640, and result feedback module 650 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the business application module 610, data acquisition module 620, preliminary review processing module 630, review processing module 640, and result feedback module 650 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the business application module 610, data collection module 620, preliminary review processing module 630, review processing module 640, and result feedback module 650 may be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0133] Figure 7 A block diagram schematically illustrates an electronic device suitable for implementing a supply chain-based business auditing method according to an embodiment of this application.

[0134] like Figure 7As shown, an electronic device 900 according to an embodiment of this application includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage portion 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0135] RAM 903 stores various programs and data required for the operation of electronic device 900. Processor 901, ROM 902, and RAM 903 are interconnected via bus 904. Processor 901 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 902 and / or RAM 903. It should be noted that the programs may also be stored in one or more memories other than ROM 902 and RAM 903. Processor 901 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0136] According to embodiments of this application, the electronic device 900 may further include an input / output (I / O) interface 905, which is also connected to a bus 904. The electronic device 900 may also include one or more of the following components connected to the input / output (I / O) interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the input / output (I / O) interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 910 as needed so that computer programs read from it can be installed into the storage section 908 as needed.

[0137] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0138] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 902 and / or RAM 903 and / or one or more memories other than ROM 902 and RAM 903 described above.

[0139] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.

[0140] When the computer program is executed by the processor 901, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0141] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via the communication section 909, and / or installed from a removable medium 911. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0142] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, it performs the functions defined in the system of this application embodiment. According to the embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0143] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0144] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0145] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A business auditing method based on the supply chain, characterized in that, The method includes: In response to the applicant submitting a business application, the application data corresponding to the business application is obtained; A federated learning framework is constructed based on the participating entities and the applicant entities, and multi-source data is collected based on the federated learning framework; wherein, there is a supply chain between the applicant entities and the participating entities; Based on the multi-source data, a multi-agent collaborative strategy is used to perform preliminary review of the application data to obtain the preliminary review result; If the initial review result is positive, the application data is reviewed based on the multi-source data using a risk transmission simulation engine and an anomaly detection engine to obtain a review result; and An audit result is generated based on the preliminary review result or the review result, and the audit result is sent to the applicant.

2. The method according to claim 1, characterized in that, The multi-agent cooperative strategy includes: Based on a distributed intelligent agent architecture, task review instructions are transmitted between target intelligent agents via message queues; wherein, the target intelligent agents include one or more of the following: document parsing intelligent agent, logistics sensing intelligent agent, product evaluation intelligent agent, pricing engine intelligent agent, and decision coordination intelligent agent; and The initial review status of the target intelligent agent is synchronized using a central cluster.

3. The method according to claim 2, characterized in that, The preliminary review of the application data based on the multi-source data, using a multi-agent collaborative strategy, yields the following results: The document parsing agent parses the application data and performs information verification on the multi-source data and the parsed application data to obtain the information verification result. Based on the multi-source data, the logistics information of the application data is extracted by the logistics sensing agent, and the logistics information is reviewed to obtain the logistics review result; Based on the multi-source data, the product evaluation agent performs a product risk assessment on the application data to obtain the product evaluation result. Based on the pricing information in the application data, the pricing engine agent invokes the multi-source data to calculate a pricing reference result; and The decision-coordination intelligent agent analyzes the multi-dimensional review results to generate the preliminary review results; wherein, the multi-dimensional review results include one or more of the information verification results, the logistics review results, the product evaluation results, and the pricing reference results.

4. The method according to claim 1, characterized in that, The risk transmission simulation engine includes: Using the participating entities and the applicant entities as graph nodes and business relationships as graph edges, a dynamic supply chain relationship graph is constructed; and The risk assessment agent simulates risk propagation paths based on the supply chain dynamic correlation graph and the multi-source data, and assesses the specific business risks of the application data based on the risk propagation results.

5. The method according to claim 4, characterized in that, The anomaly detection engine includes: Scan the dynamic relationship graph of the supply chain to identify abnormal relationship subgraphs; and Based on the multi-source data, the abnormal correlation subgraph is detected by a pre-trained decision tree model to obtain the abnormal risk result.

6. The method according to claim 1, characterized in that, The construction of the federated learning framework based on the participating entities and the applicant entities includes: Based on the participating entities and the applying entities, a federated node is constructed; According to the data encryption protocol, security rules are set for the transmission parameters between the federated nodes and the central coordination node; and Based on a preset data interface, the federated nodes are connected to the corresponding node database to construct the federated learning framework.

7. The method according to claim 6, characterized in that, The collection of multi-source data based on the federated learning framework includes: In the federated node, business data is extracted based on the node database accessed by the federated node, and the business data is standardized. In the federated node, the standardized business data is homomorphically encrypted to obtain encrypted feature data, and the encrypted feature data is uploaded to the central coordination node; and The encrypted feature data is aggregated through the central coordination node to obtain the multi-source data.

8. A business auditing device based on the supply chain, characterized in that, The device includes: The business application module is used to respond to the application subject submitting a business application and obtain the application data corresponding to the business application; The data acquisition module is used to construct a federated learning framework based on the participating entities and the applicant entity, and to collect multi-source data based on the federated learning framework; wherein, there is a supply chain between the applicant entity and the participating entities; The preliminary review processing module is used to perform preliminary review processing on the application data based on the multi-source data and adopt a multi-agent collaborative strategy to obtain the preliminary review result; The review processing module is used to review the application data based on the multi-source data, using a risk transmission simulation engine and an anomaly detection engine, and obtain a review result if the initial review result is satisfactory; and The result feedback module is used to generate an audit result based on the preliminary review result or the review result, and send the audit result to the applicant.

9. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

Images