Enterprise financial risk determination method and related device

Abstract

The invention discloses an enterprise financial risk determination method and a related device, and relates to the technical field of artificial intelligence. Six dimensions of mobility, credit, market, operation, industry macroscopic and public opinion reputation are respectively processed through six special and precise agents including a mobility risk analysis agent, a credit risk analysis agent, a market risk analysis agent, an operation risk agent, an industry macroscopic risk analysis agent and a public opinion and reputation risk analysis agent; and full-quadrant scanning of enterprise financial risks is realized. The cross validation coordinator and the dynamic cooperation engine form a double-wheel drive, and the cross validation coordinator can obtain a cooperation agent identifier set, a cooperation task type and cooperation configuration information; the dynamic collaboration engine dispatches multi-agent parallel collaboration in real time based on the collaboration agent identifier set, the collaboration task type and the collaboration configuration information, information islands are eliminated, and misjudgment and missing report are remarkably reduced. Therefore, accurate enterprise financial risks can be obtained.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method for determining corporate financial risks and related devices. Background Art

[0002] Financial risk refers to the risk that, due to the uncertainty of capital flows (such as financing, investment, capital recovery, and profit distribution) during the operation of an enterprise, it may lose its solvency, disrupt its capital chain, or reduce its profits, potentially leading to economic losses or even bankruptcy. As the business environment becomes increasingly complex, financial risk is increasingly characterized by multi-source heterogeneity, dynamic linkages, and cross-entity transmission. Determining a company's financial risk is a technical challenge that urgently needs to be addressed in this field. Summary of the Invention

[0003] In view of the above problems, this application provides a method and related device for determining the financial risk of an enterprise to achieve the purpose of determining the financial risk of the enterprise. The specific solution is as follows:

[0004] The first aspect of the present application provides a method for determining enterprise financial risk, comprising:

[0005] Input the enterprise's cash flow gap value within a first preset future time period, the proportion of highly liquid assets, the proportion of liabilities due within a second preset future time period, the bid-ask spread of key assets, the available credit line, and the industry negative public opinion index into a pre-built liquidity risk analysis agent, and output liquidity risk information through the liquidity risk analysis agent. The liquidity risk information includes a liquidity risk score, an explainability analysis, and first cross-agent collaboration information; the first cross-agent collaboration information includes an agent identifier that affects the liquidity risk score;

[0006] Input the debt-to-asset ratio, interest coverage ratio, operating cash flow, total liabilities, industry prosperity index, and customer and supplier concentration of the enterprise into a pre-built credit risk analysis agent, and output credit risk information through the credit risk analysis agent. The credit risk information includes a credit risk score, a first key risk propagation link, and second cross-agent collaboration information. The second cross-agent collaboration information includes the identity of the agent that affects the credit risk score. The first key risk propagation link includes a link consisting of risk events that lead to the credit risk score.

[0007] Input the enterprise's volatility index, profit volatility, key interest rate curve, commodity main contract price, stock-home correlation index, and foreign exchange risk exposure ratio into a pre-built market risk analysis agent, and output market risk information through the market risk analysis agent. The market risk information includes the market risk score at the current time, a second key risk propagation link, and third cross-agent collaboration information; the second key risk propagation link includes a link consisting of risk events that lead to the market risk score; the third cross-agent collaboration information includes the identity of the agent that affects the market risk score;

[0008] Input the company's compliance check abnormality frequency, system failure time ratio, abnormal access request volume, and supplier environmental and social governance score into a pre-built operational risk agent, and output operational risk information through the operational risk agent. The operational risk information includes the risk source that causes the operational risk, a third key risk propagation link, and fourth cross-agent collaboration information; the third key risk propagation link includes a link consisting of risk events that cause the risk source, and the fourth cross-agent collaboration information includes the identity of the agent that affects the risk source;

[0009] Input the enterprise's capacity utilization fluctuations, enterprise market share changes, public opinion sentiment index, core supplier concentration, and upstream raw material price fluctuation transmission coefficient into a pre-built industry macro risk analysis agent, and output industry macro risk information through the industry macro risk analysis agent. The industry macro risk information includes the industry macro risk rating and the fourth key risk propagation link. The fourth key risk propagation link includes the link consisting of risk events that lead to the industry macro risk rating.

[0010] The negative KOL dissemination heat of the enterprise, the network-wide sentiment polarity index, the brand search volume volatility, the industry negative public opinion index, and the frequency of regulatory penalties are input into a pre-built public opinion and reputation risk analysis agent, and the public opinion and reputation risk analysis agent outputs public opinion and reputation risk information, which includes the reputation risk score at the current time, the public opinion propagation path, and the fifth cross-agent collaboration information; the public opinion propagation path includes a link composed of key nodes that lead to the reputation risk score, and the fifth cross-agent collaboration information includes the identity of the agent that affects the reputation risk score;

[0011] The liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, and the first preset prompt text are input into a pre-built cross-validation coordinator, and a collaborative agent identification set, a collaborative task type, and collaborative configuration information are obtained through the cross-validation coordinator; the collaborative agent identification set includes multiple collaborative target agent identifications, and the collaborative configuration information includes the collaborative processes corresponding to the multiple target agent identifications;

[0012] Inputting the set of collaborative agent identifiers, the collaborative task type, the collaborative configuration information, and a second preset prompt text into a pre-built dynamic collaboration engine, and communicating with the agent having the target agent identifier through the dynamic collaboration engine according to the collaborative configuration information to obtain collaboration risk information;

[0013] The liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, the collaborative risk information, the collaborative intelligent entity identification set, the collaborative task type and the collaborative configuration information are input into a pre-built risk assessment and early warning generation model, and the enterprise financial risk is output through the risk assessment and early warning generation model.

[0014] In a possible implementation, the step of obtaining, by the cross-validation coordinator, a set of collaborative agent identifiers, a collaborative task type, and collaborative configuration information includes:

[0015] Obtaining, through the cross-validation coordinator, conflict detection results of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0016] Obtaining, through the cross-validation coordinator, risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0017] Obtaining the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information through the cross-validation coordinator;

[0018] The accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro-risk information, and the accuracy of the public opinion and reputation risk information are obtained through the conflict detection results, the risk association information and the cross-validation coordinator, and the collaborative intelligent agent identification set, the collaborative task type and the collaborative configuration information are obtained.

[0019] In one possible implementation, the step of obtaining, through the cross-validation coordinator, conflict detection results of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information includes:

[0020] Obtaining a first field value of the same first field of a numerical type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0021] Obtaining a first comparison result between the absolute value of the difference between any two of the first field values ​​included in the liquidity risk information, the first field value included in the credit risk information, the first field value included in the market risk information, the first field value included in the operational risk information, the first field value included in the industry macro risk information, and the first field value included in the public opinion and reputation risk information and a first preset threshold;

[0022] Obtaining a second field value of the same second field of the enumeration type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0023] detecting whether the second field value included in the liquidity risk information, the second field value included in the credit risk information, the second field value included in the market risk information, the second field value included in the operational risk information, the second field value included in the industry macro risk information, and the second field value included in the public opinion and reputation risk information are the same, to obtain a second comparison result;

[0024] Obtaining a third field value of the same third field of text type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0025] Obtaining semantic similarities among the third field value included in the liquidity risk information, the third field value included in the credit risk information, the third field value included in the market risk information, the third field value included in the operational risk information, the third field value included in the industry macro risk information, and the third field value included in the public opinion and reputation risk information, to obtain a third comparison result;

[0026] The conflict detection result includes the first comparison result, the second comparison result, and the third comparison result.

[0027] In one possible implementation, the step of obtaining, by the cross-validation coordinator, risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information includes:

[0028] The same risk events in the first key risk propagation link, the second key risk propagation link, the third key risk propagation link, and the fourth key risk propagation link are merged to obtain the risk association information.

[0029] In one possible implementation, the step of obtaining, by the cross-validation coordinator, the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information includes:

[0030] Determining the accuracy of the liquidity risk information based on the uncertainty metadata output by the liquidity risk analysis agent;

[0031] Determining the accuracy of the credit risk information based on the uncertainty metadata output by the credit risk analysis agent;

[0032] Determining the accuracy of the market risk information based on the uncertainty metadata output by the market risk analysis agent;

[0033] Determining the accuracy of the operational risk information based on the uncertainty metadata output by the operational risk agent;

[0034] Determining the accuracy of the industry macro risk information based on the uncertainty metadata output by the industry macro risk analysis agent;

[0035] Based on the uncertainty metadata output by the public opinion and reputation risk analysis agent, the accuracy of the public opinion and reputation risk information is determined.

[0036] A second aspect of the present application provides an enterprise financial risk determination device, comprising:

[0037] A first acquisition module is configured to input the enterprise's cash flow gap value within a first preset future time period, the proportion of highly liquid assets, the proportion of liabilities due within a second preset future time period, the bid-ask spread of key assets, the available credit limit, and the industry negative public opinion index into a pre-built liquidity risk analysis agent, and output liquidity risk information through the liquidity risk analysis agent, the liquidity risk information including a liquidity risk score, an explainability analysis, and first cross-agent collaboration information; the first cross-agent collaboration information includes an agent identifier that affects the liquidity risk score;

[0038] A second acquisition module is configured to input the debt-to-asset ratio, interest coverage ratio, operating cash flow, total liabilities, industry prosperity index, and customer and supplier concentration of the enterprise into a pre-built credit risk analysis agent, and output credit risk information through the credit risk analysis agent, wherein the credit risk information includes a credit risk score, a first key risk propagation link, and second cross-agent collaboration information, wherein the second cross-agent collaboration information includes the identity of the agent that affects the credit risk score; the first key risk propagation link includes a link consisting of risk events that lead to the occurrence of the credit risk score;

[0039] A third acquisition module is configured to input the enterprise's volatility index, profit volatility, key interest rate curve, commodity main contract price, stock-home correlation index, and foreign exchange risk exposure ratio into a pre-built market risk analysis agent, and output market risk information through the market risk analysis agent. The market risk information includes a market risk score at the current time, a second key risk propagation link, and third cross-agent collaboration information; the second key risk propagation link includes a link consisting of risk events that lead to the market risk score; and the third cross-agent collaboration information includes the identity of the agent that affects the market risk score.

[0040] A fourth acquisition module is configured to input the enterprise's compliance check exception frequency, system failure time ratio, abnormal access request volume, and supplier environmental and social governance score into a pre-built operational risk agent, and output operational risk information through the operational risk agent. The operational risk information includes the risk source that causes the operational risk, a third key risk propagation link, and fourth cross-agent collaboration information; the third key risk propagation link includes a link consisting of risk events that cause the risk source, and the fourth cross-agent collaboration information includes the identity of the agent that affects the risk source.

[0041] A fifth acquisition module is configured to input the enterprise's capacity utilization fluctuations, changes in enterprise market share, public opinion sentiment index, core supplier concentration, and upstream raw material price fluctuation transmission coefficient into a pre-built industry macro-risk analysis agent, and output industry macro-risk information through the industry macro-risk analysis agent. The industry macro-risk information includes an industry macro-risk rating and a fourth key risk propagation link. The fourth key risk propagation link includes a link consisting of risk events that lead to the industry macro-risk rating.

[0042] The sixth acquisition module is used to input the enterprise's KOL negative communication heat, the network-wide sentiment polarity index, the brand search volume volatility, the industry negative public opinion index, and the frequency of regulatory penalties into a pre-built public opinion and reputation risk analysis agent, and output public opinion and reputation risk information through the public opinion and reputation risk analysis agent. The public opinion and reputation risk information includes the reputation risk score at the current time, the public opinion propagation path, and the fifth cross-agent collaboration information; the public opinion propagation path includes a link composed of key nodes that lead to the reputation risk score, and the fifth cross-agent collaboration information includes the identity of the agent that affects the reputation risk score;

[0043] a seventh acquisition module, configured to input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, and a first preset prompt text into a pre-built cross-validation coordinator, and obtain, through the cross-validation coordinator, a collaborative agent identification set, a collaborative task type, and collaborative configuration information; the collaborative agent identification set includes multiple collaborative target agent identifications, and the collaborative configuration information includes collaborative processes corresponding to the multiple target agent identifications;

[0044] an eighth acquisition module, configured to input the set of collaborative agent identifiers, the collaborative task type, the collaborative configuration information, and a second preset prompt text into a pre-built dynamic collaboration engine, and communicate with the agent having the target agent identifier through the dynamic collaboration engine according to the collaborative configuration information to obtain collaboration risk information;

[0045] The ninth acquisition module is used to input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, the collaborative risk information, the collaborative intelligent body identification set, the collaborative task type and the collaborative configuration information into a pre-built risk assessment and early warning generation model, and output the enterprise financial risk through the risk assessment and early warning generation model.

[0046] In a possible implementation, the seventh acquisition module includes:

[0047] A first acquisition unit is configured to acquire, through the cross-validation coordinator, a conflict detection result of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0048] A second acquisition unit is configured to acquire, through the cross-validation coordinator, risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0049] A third acquisition unit is configured to acquire, through the cross-validation coordinator, the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information;

[0050] The fourth acquisition unit is used to obtain the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information through the conflict detection results, the risk association information and the cross-validation coordinator, and obtain the collaborative intelligent agent identification set, the collaborative task type and the collaborative configuration information.

[0051] A third aspect of the present application provides a computer program product comprising computer-readable instructions, which, when executed on an electronic device, enables the electronic device to implement the enterprise financial risk determination method of the first aspect or any implementation of the first aspect.

[0052] A fourth aspect of the present application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0053] The memory is used to store computer programs;

[0054] The processor is used to execute the computer program so that the electronic device can implement the enterprise financial risk determination method of the first aspect or any implementation of the first aspect.

[0055] In a fifth aspect, the present application provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the enterprise financial risk determination method of the above-mentioned first aspect or any implementation of the first aspect.

[0056] Through the above technical solution, this application provides a method for determining corporate financial risks. Through six specialized intelligent agents, including liquidity risk analysis agent, credit risk analysis agent, market risk analysis agent, operational risk agent, industry macro risk analysis agent, and public opinion and reputation risk analysis agent, the six dimensions of liquidity, credit, market, operation, industry macro, and public opinion and reputation are processed respectively to achieve a "full quadrant" scan of corporate financial risks. The cross-validation coordinator and the dynamic collaboration engine constitute a "two-wheel drive". The cross-validation coordinator can obtain the collaborative agent identification set, collaborative task type, and collaborative configuration information; the dynamic collaboration engine is based on the collaborative agent identification set, collaborative task type, and collaborative configuration information, and schedules multiple agents to collaborate in parallel in real time, eliminating information islands and significantly reducing misjudgments and omissions. In this way, a more accurate corporate financial risk can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that the originals and elements are not necessarily drawn to scale.

[0058] Figure 1 A schematic diagram of the system architecture provided for this application;

[0059] Figure 2 A flowchart of a method for determining corporate financial risk provided in an embodiment of the present application;

[0060] Figure 3 A schematic diagram of the structure of a device for determining enterprise financial risk provided in an embodiment of the present application;

[0061] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] The following describes the embodiments of the present application in conjunction with the accompanying drawings. The terms used in the implementation methods of the present application are only used to explain the specific embodiments of the present application and are not intended to limit the present application.

[0063] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0064] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is merely a way of distinguishing the objects of the same attributes when describing them in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0065] As the business environment becomes increasingly complex, financial risks are increasingly characterized by multi-source heterogeneity, dynamic correlations, and cross-subject transmission. Businesses are shifting their demand for risk early warning from single-indicator analysis to multi-dimensional collaborative perception and closed-loop management. Consequently, the limitations of traditional rule-based financial risk early warning systems are becoming increasingly prominent.

[0066] In recent years, artificial intelligence (AI) technology has continuously penetrated the financial risk management field. Intelligent agents, as large-scale model systems with autonomous perception, decision-making, and action capabilities, are driving the financial sector's transition from automation to intelligence. Each intelligent agent, built on a large model, possesses specialized analytical capabilities and the ability to input and output structured and unstructured data.

[0067] Despite technological advancements, intelligent agents focus on completing a single specialized task, and financial risk analysis still lacks multi-role collaboration and structured communication mechanisms. This leads to information silos, a static single perspective, and a lack of coordination.

[0068] In the embodiment of this application, six major intelligent agents, namely liquidity risk analysis intelligent agent, credit risk analysis intelligent agent, market risk analysis intelligent agent, operational risk intelligent agent, industry macro risk analysis intelligent agent, and public opinion and reputation risk analysis intelligent agent, are used to integrate heterogeneous information such as financial statements, market data, public opinion texts, regulatory penalties, and ESG scores into the evaluation system, which significantly improves the integrity and accuracy of risk identification and avoids the blind spot of the traditional model of "only looking at the three financial statements".

[0069] The detailed description is given below.

[0070] See also Figure 1 , Figure 1 A schematic diagram of a system architecture is shown. The system may include a terminal 100 and a server 200. The server 200 may provide the method provided in the embodiment of the present application to one or more terminals.

[0071] Among them, an application can be installed on the terminal 100, and the above application can provide an interface. The terminal 100 can receive relevant parameters entered by the user on the interface, for example, the input of the six major intelligent bodies, and send the above parameters to the server 200. The server 200 can obtain processing results based on the received parameters and return the processing results to the terminal 100.

[0072] It should be understood that in some optional implementations, the terminal 100 can also complete the action of obtaining the processing result based on the received parameters by itself without the need for the cooperation of the server, and the embodiments of the present application are not limited to this.

[0073] Next describe Figure 1 The product form of the mid-terminal 100;

[0074] The terminal 100 in the embodiment of the present application can be a mobile phone, a tablet computer, a wearable device, an in-vehicle device, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiment of the present application does not impose any restrictions on this.

[0075] The terminal 100 may include components such as a radio frequency unit, a memory, an input unit, a display unit, a camera (optional), an audio circuit (optional), a speaker (optional), a microphone (optional), a headphone jack (optional), a processor, an external interface, and a power supply. Those skilled in the art will appreciate that the aforementioned components are merely examples and do not limit the terminal or multi-function device. The terminal or multi-function device may include more or fewer components, or may combine certain components or use different components.

[0076] The input unit can be used to receive input digital or character information and generate key signal input related to user settings and function control of the portable multi-function device. Specifically, the input unit may include a touch screen (optional) and / or other input devices. Specifically, other input devices may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, power keys, etc.), a trackball, a mouse, a joystick, etc.

[0077] Among them, the input device can receive input data and so on.

[0078] The display unit can be used to display information input by the user or information provided to the user, various menus of the terminal, interactive interface, file display and / or playback of any multimedia file. In an embodiment of the present application, the display unit can be used to display an interface, processing results, etc.

[0079] Among them, the memory can be used to store software codes related to the enterprise financial risk determination method, the processor can execute the steps of the enterprise financial risk determination method, and can also schedule other units (such as the above-mentioned input unit and display unit) to achieve corresponding functions.

[0080] The radio frequency unit (optional) can be used to send and receive information or receive and send signals during a call.

[0081] In this embodiment of the present application, the radio frequency unit can send data to the server 200 and receive the processing results sent by the server 200.

[0082] It should be understood that the radio frequency unit is optional and can be replaced by other communication interfaces, such as a network port.

[0083] The terminal 100 further includes a power source (such as a battery) for supplying power to various components.

[0084] The terminal 100 also includes an external interface, which may be a standard Micro USB interface or a multi-pin connector, and may be used to connect the terminal 100 to other devices for communication, or to connect a charger to charge the terminal 100.

[0085] The server 200 includes a bus, a processor, a communication interface, and a memory. The processor, the memory, and the communication interface communicate with each other via the bus.

[0086] Among them, the memory can be used to store software codes related to the enterprise financial risk determination method, the processor can execute the steps of the enterprise financial risk determination method of the chip, and can also schedule other units to implement corresponding functions.

[0087] Reference Figure 2 , Figure 2 A flow chart of a method for determining corporate financial risk provided in an embodiment of the present application is shown as follows: Figure 2 As shown, a method for determining enterprise financial risk provided by an embodiment of the present application may include steps S201 to S209, and these steps are described in detail below.

[0088] Step S201: Input the cash flow gap value of the enterprise in the first preset future time period, the proportion of high-liquidity assets, the proportion of maturing liabilities in the second preset future time period, the bid-ask spread of key assets, the available credit limit, and the industry negative public opinion index into a pre-built liquidity risk analysis agent, and output liquidity risk information through the liquidity risk analysis agent. The liquidity risk information includes a liquidity risk score, an explainability analysis, and first cross-agent collaboration information; the first cross-agent collaboration information includes an agent identifier that affects the liquidity risk score.

[0089] Exemplarily, the first preset future time period may be determined based on actual conditions and is not limited here, for example, one week, one month, one year, etc.

[0090] Exemplarily, the second preset future time period may be determined based on actual conditions and is not limited here, for example, one week, one month, one year, etc.

[0091] The following describes each parameter.

[0092] The cash flow gap value within the first preset future time period may be the expected cash flow gap value within the first preset future time period for each business line in the future.

[0093] The business line can be a retail business line, a wholesale business line, a manufacturing business line, etc.

[0094] Assuming the first preset future time period is the next 7 or 30 days, the cash flow gap value for the first preset future time period is the difference between "cash to be paid" and "cash expected to be received" within the next 7 or 30 days. The cash flow gap value for the first preset future time period can be used to measure the degree of short-term capital chain tension. The larger the gap, the higher the liquidity risk.

[0095] The proportion of highly liquid assets refers to the percentage of cash and near-cash (such as demand deposits, money market funds, and reverse repos of government bonds) in total assets. Understandably, a higher proportion of highly liquid assets indicates a company's ability to quickly liquidate and repay debts in an emergency.

[0096] The proportion of liabilities due within the second predetermined future time period refers to the ratio of liabilities that must be repaid within the second predetermined future time period (such as short-term loans, notes payable, and long-term loans due within one year) to total liabilities. It is understood that the higher the proportion of liabilities due within the second predetermined future time period, the greater the company's near-term debt repayment pressure.

[0097] The bid-ask spread of key assets refers to the difference between the current bid and ask prices of large, tradable assets held by a company (such as stocks, bonds, and commodities). It's understandable that the wider the bid-ask spread, the greater the asset's discount when liquidated, and the lower its liquidity.

[0098] Available credit refers to the amount of credit that a bank has approved but not yet used. It's understandable that the larger the available credit, the more likely a business will be able to "draw down" funds in an emergency.

[0099] The industry negative public opinion index is a quantitative indicator of negative public opinion (typically ranging from 0 to 100, with higher scores indicating greater negativity) derived through natural language processing, using sentiment scores from news, social media, and announcements about a company. Understandably, negative public opinion can trigger banks to tighten credit, customers to delay payments, and exacerbate liquidity risks.

[0100] Exemplarily, the liquidity risk information also includes: one or more of: gap warning signals, stress levels, and explainability analysis.

[0101] Liquidity risk score refers to a quantitative score ranging from 0 to 100 or AAA to D grade generated by comprehensively considering dimensions such as cash flow gap, asset liquidity, liability maturity, and public opinion.

[0102] A shortfall warning signal automatically triggers a "red light" or "orange light" alert if the cash flow shortfall exceeds a pre-set threshold within a pre-set time period. Understandably, this shortfall warning signal can prompt management to take immediate emergency measures, such as delaying expenditures or accelerating payment collections.

[0103] Stress levels are a grading of the probability of a company's capital chain being disrupted under extreme scenarios (e.g., a sudden drop in sales, bank loan withdrawals), with the probability being classified as mild, moderate, or severe. Understandably, stress levels can help decision-makers determine whether to activate crisis plans.

[0104] A fund optimization plan is a list of actionable suggestions, such as redeeming a financial product early to replenish cash, negotiating with a bank for additional credit, or postponing a large purchase. Understandably, a fund optimization plan provides operational guidance on how to move money out or save money.

[0105] Explainable analysis uses natural language or charts to explain why a certain score was assigned. For example, "A red flag has been triggered due to a 200 million yuan cash flow shortfall over the next seven days and only 50 million yuan in available credit." Understanding this, explainable analysis allows even non-financial decision-makers to instantly understand the source of risk.

[0106] The first cross-agent collaboration information includes the agent identification that affects the liquidity risk score. For example, if it is found that the liquidity risk score may be affected by credit risk or public opinion risk, the agent identification that affects the liquidity risk score includes the agent identification of the credit risk analysis agent and the agent identification of the public opinion and reputation risk analysis agent.

[0107] For example, the liquidity risk analysis agent can analyze cash flow, quick ratio, financing capacity, etc. to predict short-term debt repayment ability.

[0108] Step S202: Input the debt-to-asset ratio, interest coverage ratio, operating cash flow, total liabilities, industry prosperity index, and customer and supplier concentration of the enterprise into a pre-built credit risk analysis agent, and output credit risk information through the credit risk analysis agent. The credit risk information includes a credit risk score, a first key risk propagation link, and a second cross-agent collaboration information. The second cross-agent collaboration information includes an agent identifier that affects the credit risk score; the first key risk propagation link includes a link consisting of risk events that lead to the credit risk score.

[0109] The following describes each parameter.

[0110] The debt-to-asset ratio is defined as: total liabilities divided by total assets. As you can see, the higher the debt-to-asset ratio, the greater the debt repayment pressure and the higher the credit risk.

[0111] The interest coverage ratio is calculated as (earnings before interest and taxes, EBIT) divided by interest expense. As you can see, the higher the interest coverage ratio, the stronger the company's ability to pay interest and the lower its credit risk.

[0112] Operating cash flow refers to the net cash received from a company's core activities (after deducting cash costs). Understandably, persistently negative operating cash flow indicates a reliance on borrowed funds to maintain operations, raising credit risk.

[0113] Total liabilities refer to the total amount of all debts that an enterprise needs to repay, including short-term loans, long-term loans, bonds payable, etc.

[0114] The industry prosperity index is a comprehensive score (0–100 or year-over-year or month-over-month) that reflects indicators such as overall industry supply and demand, profitability, and capacity utilization. Understandably, a decline in the industry prosperity index signals industry decline and an increased probability of customer defaults.

[0115] Customer and supplier concentration refers to the ratio of sales (or purchases) from the top N customers (or suppliers) to total sales (or purchases). As you can see, the higher the customer and supplier concentration, the greater the impact on a company's creditworthiness if a major customer or supplier becomes risky. N is a positive integer greater than or equal to 1.

[0116] Exemplarily, credit risk information also includes, but is not limited to, an explainability report.

[0117] A credit risk score is a scale of 0 to 100 or AAA to D, which quantifies the probability of a company defaulting within the next 12 months.

[0118] The first key risk transmission chain is the chain of events that progressively evolve into a risk. This chain is clearly laid out using a chain or path diagram, helping managers clearly understand "where it starts, where it progresses, and where it ultimately ends." For example: industry downturn → decreased orders from major customers → deteriorating cash flow → inability to repay bonds.

[0119] An explainable report means: using natural language to explain the basis for the score, for example: "Because the debt-to-asset ratio is >80% and the interest coverage ratio is <2, the BBB score is downgraded."

[0120] The second cross-agent collaboration information includes the agent identification that affects the credit risk score. For example, if it is found that the credit risk may be linked to the market risk (rising interest rates) or liquidity risk (cash depletion), the agent identification that affects the credit risk score includes: the agent identification of the market risk analysis agent and the agent identification of the liquidity risk analysis agent.

[0121] For example, the credit risk analysis agent is used to analyze debt repayment ability, customer credit changes, financing costs, etc.

[0122] Step S203: Input the enterprise's volatility index, profit volatility, key interest rate curve, commodity main contract price, stock-home correlation index, and foreign exchange risk exposure ratio into a pre-built market risk analysis agent, and output market risk information through the market risk analysis agent. The market risk information includes the market risk score at the current time, the second key risk propagation link, and the third cross-agent collaboration information; the second key risk propagation link includes a link composed of risk events that lead to the market risk score; the third cross-agent collaboration information includes the agent identifier that affects the market risk score.

[0123] The following describes each parameter.

[0124] The higher the VIX (Volatility Index) value, the more panic the global risk sentiment is, which will raise the volatility expectations of all risky assets.

[0125] Profit volatility is calculated as the standard deviation of a company's net profit over the last 12 months divided by its mean. Interest rate volatility measures earnings stability; greater volatility makes hedging more difficult.

[0126] The key interest rate curve (LIBOR / SHIBOR) refers to the 3M / 6M / 1Y term structure of LIBOR (offshore) or SHIBOR (onshore), which directly determines the floating rate financing cost and bond valuation.

[0127] The price of a major commodity futures contract refers to the real-time price of the most active futures contract for a company's primary commodity (such as crude oil, copper, iron ore, etc.). Understandably, the price of a major commodity futures contract can be a source of revenue or cost fluctuations.

[0128] The Stock-Real Estate Correlation Index (SHRI) is the 60-day rolling correlation coefficient between stock and real estate assets. Understandably, a sudden increase in the SHRI indicates a loss of diversification and amplified portfolio drawdowns.

[0129] Foreign exchange exposure percentage is calculated as: net foreign currency-denominated assets or liabilities divided by total assets. Understandably, the higher the foreign exchange exposure percentage, the greater the impact of exchange rate fluctuations on the balance sheet.

[0130] Exemplarily, market risk information may also include, but is not limited to: one or more of: key asset drawdown range, hedge ratio adjustment signal, and explainability analysis.

[0131] For example, the market risk score at the current time refers to: "The maximum loss amount that may occur within 1 day / 10 days with a 95% confidence level under the current position.

[0132] The key asset retracement refers to the percentage drop from the historical high to the current price of a preset number of assets (for example, 5) in the company's holdings if the market continues to deteriorate.

[0133] The second key risk propagation chain is the chain of events that gradually evolve into a risk. For example, the second key risk propagation chain could be "rising interest rates → falling bond valuations → shrinking net assets → triggering collateral liquidation."

[0134] For example, the hedge ratio adjustment signal refers to an executable instruction such as "increase the interest rate swap position by 20%" or "adjust the foreign exchange forward hedging ratio from 60% to 80%."

[0135] Explainability analysis means using natural language to explain the scoring basis. For example, "The 50 basis point increase in SHIBOR 3M has led to an increase of 20 million in floating interest expenses, which is the main reason for the increase in VaR today."

[0136] The third cross-agent collaboration information includes the agent identification that affects the market risk score. For example, if "exchange rate fluctuation → import cost ↑ → cash flow gap ↑" is found, the agent identification that affects the market risk score includes: the agent identification of the liquidity risk analysis agent.

[0137] For example, the market risk analysis agent is used to analyze the impact of interest rate, exchange rate, and commodity price fluctuations on corporate assets, liabilities, and revenue.

[0138] Step S204: Input the company's compliance inspection abnormality frequency, system failure time ratio, abnormal access request volume, and supplier environmental and social governance score into a pre-built operational risk agent, and output operational risk information through the operational risk agent. The operational risk information includes the risk source that causes the operational risk, the third key risk propagation link, and the fourth cross-agent collaboration information; the third key risk propagation link includes the link composed of risk events that cause the occurrence of the risk source, and the fourth cross-agent collaboration information includes the agent identifier that affects the risk source.

[0139] The following describes each parameter.

[0140] The frequency of compliance inspection anomalies refers to the number of violations discovered during audits, regulatory oversight, or internal control inspections within a specific timeframe. As you can imagine, the higher the frequency of compliance inspection anomalies, the greater the risk of a violation.

[0141] System downtime percentage is calculated as: System unavailable minutes per month divided by total monthly minutes. As you can see, the higher the system downtime percentage, the greater the risk of business interruption.

[0142] Abnormal access requests refer to the number of logins / operations identified as suspicious or unauthorized by the firewall / log.

[0143] A supplier's ESG score is a comprehensive assessment of a supplier's environmental (E), social (S), and governance (G) performance by a third-party organization. Understandably, the lower a supplier's ESG score, the higher the compliance and reputational risks they pose.

[0144] Exemplarily, operational risk information may also include, but is not limited to: immediate warning level and event type.

[0145] Risk source refers to the use of "people / system / external events" labels to point out the real root cause of the operational risk, such as "internal employee unauthorized transfer" or "UPS failure in the computer room."

[0146] Immediate warning levels refer to lights that are automatically divided by severity, for example, Green / Amber / Red, or levels 1 to 5; red lights require immediate attention.

[0147] Event type refers to the classification of risk events into standard dictionaries, such as data leakage, system interruption, compliance violation, supplier default, etc., to facilitate subsequent statistics and governance.

[0148] The third key risk propagation chain is how multiple risk events gradually evolve into a risk. For example, UPS failure → server downtime → payment system downtime → regulatory fines.

[0149] The third cross-agent collaboration information includes the identity of the agent that affects the risk source.

[0150] For example, the operational risk agent is used to analyze internal control deficiencies, compliance risks, and major events using text analysis and knowledge graphs.

[0151] Step S205: Input the fluctuations in the enterprise's capacity utilization rate, changes in the enterprise's market share, public opinion sentiment index, core supplier concentration, and upstream raw material price fluctuation transmission coefficient into a pre-built industry macro-risk analysis intelligent body, and output industry macro-risk information through the industry macro-risk analysis intelligent body. The industry macro-risk information includes the industry macro-risk rating and the fourth key risk propagation link. The fourth key risk propagation link includes a link consisting of risk events that lead to the occurrence of the industry macro-risk rating.

[0152] The following describes each parameter.

[0153] Fluctuations in capacity utilization refer to whether the machine operating rate is rising or falling. Large fluctuations indicate an imbalance between supply and demand, and the industry's business climate is fluctuating.

[0154] Changes in the market share of leading companies refer to increases or decreases in the market share of the top N companies by ≥ X%, reflecting any sudden changes in industry concentration and competitive landscape. N and X can be determined based on actual circumstances and are not specified here.

[0155] The public sentiment index refers to the positive or negative ratings of the industry by online news and social media, such as 0 to 100. It is understandable that the lower the public sentiment index, the more pessimistic the sentiment and the higher the demand or policy risks.

[0156] Core supplier concentration refers to the proportion of supply from the top N key suppliers to the industry's total procurement volume. N can be determined based on actual circumstances and is not a fixed number here. It's understandable that a high concentration of core suppliers can lead to a shortage across the entire industry if supply is disrupted.

[0157] The upstream raw material price fluctuation transmission coefficient measures the percentage increase or decrease in the industry's average cost / sell price for every 1% increase or decrease in raw material prices. Understandably, a transmission coefficient of upstream raw material price fluctuations greater than 0.7 indicates that "a shock upstream can severely impact downstream markets."

[0158] Exemplarily, industry macro-risk information may also include, but is not limited to: one or more of: differentiated positioning solutions, key information penetration rate, and communication ROI.

[0159] For example, the industry macro risk rating is AAA to D or 1 to 5. Based on the industry macro risk rating, the overall risk level of the industry can be understood at a glance.

[0160] A differentiated positioning plan means: instructing the company on "which segment to enter and which model to use" to avoid industry resonance and reduce risks.

[0161] Key information penetration refers to the percentage of companies and channels within an industry that have fully communicated important risk news (such as policy tightening). As you can imagine, the higher the key information penetration rate, the faster the collective action.

[0162] Communication ROI refers to how much negative impact can be reduced by spending 1 yuan on risk communication (such as announcements and public opinion guidance), thereby quantifying whether it is worth refuting rumors.

[0163] The fourth key risk propagation chain is how multiple risk events gradually evolve into a risk. For example, upstream price increases → losses for small and medium-sized manufacturers → price cuts by leading manufacturers to grab market share → industry price wars → overall profit margin decline. This can help identify the first link to explode.

[0164] For example, the industry macro-risk analysis agent is used to analyze policy changes, industrial cycles, upstream and downstream impacts, and competitive landscape.

[0165] Step S206: Input the KOL negative communication heat of the enterprise, the network-wide sentiment polarity index, the brand search volume volatility, the industry negative public opinion index, and the frequency of regulatory penalties into the pre-built public opinion and reputation risk analysis intelligent entity, and output the public opinion and reputation risk information through the public opinion and reputation risk analysis intelligent entity. The public opinion and reputation risk information includes the reputation risk score at the current time, the public opinion propagation path, and the fifth cross-agent collaboration information; the public opinion propagation path includes a link composed of key nodes that lead to the reputation risk score, and the fifth cross-agent collaboration information includes the identity of the intelligent entity that affects the reputation risk score.

[0166] The following describes each parameter.

[0167] The negative KOL (Key Opinion Leader) spread heat refers to the total number of reads, reposts, and comments on negative content published by key opinion leaders (such as influencers, bloggers, and experts) within 24 hours. As you can see, a higher KOL negative spread heat value indicates a stronger spread.

[0168] The online sentiment polarity index uses NLP (Natural Language Processing) to categorize news, Weibo posts, short videos, and other content as positive, neutral, or negative, then calculates the percentage of negative content on a scale of 0 to 100. As you can imagine, the higher the online sentiment polarity index, the greater the negative impact.

[0169] Brand search volume volatility refers to the standard deviation of the average daily search volume for a brand keyword on search engines over a specified period of time. Understandably, a sudden spike or plunge in brand search volume volatility often signals a public opinion storm.

[0170] The industry negative public opinion index refers to the normalized total amount of negative information in the entire industry, with a score between 0 and 100.

[0171] The frequency of regulatory penalties refers to the number of times a company and its subsidiaries have been publicly penalized by regulators in the past 12 months. It is understandable that the more frequent regulatory penalties are, the greater the reputational and legal risks.

[0172] For example, public opinion and reputation risk information may also include, but is not limited to: a KOL relationship database activation list and an expected value recovery curve.

[0173] The current reputation risk score refers to the risk level of the brand reputation.

[0174] The public opinion propagation path uses a timeline and network diagram to record the source of negative information, the key nodes it passes through (media / KOLs / platforms), and how it is amplified step by step, ultimately forming the complete chain of transmission of the public opinion storm. For example, a time-series network diagram of "who first posts → who amplifies → who triggers the outbreak" can be traced back to the original post and key nodes.

[0175] The KOL relationship database activation list refers to: a list of KOLs with the top N forwarding volumes and negative sentiment in negative public opinion. For example, the list can also include the number of fans and interaction rate for precise public relations.

[0176] The expected value recovery curve refers to: based on the historical pattern of the decline of similar public opinions, predicting the number of days and path required for brand reputation to return to pre-crisis levels.

[0177] The fifth cross-agent collaboration information includes the agent identifier that affects the reputation risk score. If the reputation risk may trigger liquidity or credit risk, the agent identifier that affects the reputation risk score includes: the agent identifier of the liquidity risk analysis agent and the agent identifier of the credit risk analysis agent.

[0178] For example, the public opinion and reputation risk analysis agent is used to monitor corporate risk signals in news, social media, and analytical reports.

[0179] It is understood that the six agents mentioned in steps S201 through S206 are specialized agents. Each specialized agent can process data including, but not limited to: structured financial data, such as financial statements; unstructured data, such as news, social media, and announcements; real-time / quasi-real-time data, such as public opinion hotspots and market trends; and industry / macro data, such as GDP (Gross Domestic Product), PPI (Producer Price Index), and CPI (Consumer Price Index). Each specialized agent can output analytical information, including, but not limited to, risk type, key indicators, risk level, risk confidence, risk impact estimate, and risk trigger points.

[0180] The risk information output by the six agents mentioned in steps S201 through S206 is accompanied by interpretable analysis, including feature contributions, trigger thresholds, and event chains. This interpretable analysis allows regulators, investors, and management to quickly identify the root causes of risk, mitigate "black box" concerns, and meet regulatory compliance and information disclosure requirements.

[0181] In an optional implementation, the training process of the six agents mentioned in steps S201 to S206 may include the following steps A1 to A2.

[0182] Step A1: Input sample training data into the agent and obtain prediction results through the agent.

[0183] Step A2: Train the agent based on the comparison between the prediction results and the labeled results of the sample training data.

[0184] Different agents use different sample training data. For example, if the agent is a liquidity risk analysis agent, the sample training data includes the cash flow gap value in the first preset future time period, the proportion of highly liquid assets, the proportion of liabilities due in the second preset future time period, the bid-ask spread of key assets, available credit limits, and the industry negative public opinion index. The labeled results of the sample training data also include liquidity risk information. If the agent is a credit risk analysis agent, the sample training data includes the debt-to-asset ratio, interest coverage ratio, operating cash flow, total liabilities, industry prosperity index, customer and supplier concentration, and the labeled results of the sample training data also include credit risk information.

[0185] Step S207: Input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, and the first preset prompt text into a pre-built cross-validation coordinator, and obtain the collaborative agent identification set, collaborative task type, and collaborative configuration information through the cross-validation coordinator; the collaborative agent identification set includes multiple collaborative target agent identifications, and the collaborative configuration information includes the collaborative processes corresponding to the multiple target agent identifications.

[0186] Exemplarily, the first preset prompt text includes but is not limited to the following: core role definition, input processing specifications (for standardizing data parsing methods), conflict detection method prompts, risk association and deduction, and collaborative scheduling.

[0187] For example, a core role definition might be: As the Cross-Validation Coordinator (CVC), you are the core decision-making hub of the risk analysis system. Required skills: Role: A senior risk governance expert, proficient in risk transmission logic in finance, supply chain, public opinion, and other fields; Core capabilities: Real-time analysis of risk information (JSON format) output by six specialized agents (Liquidity Risk Analysis Agent, Credit Risk Analysis Agent, Market Risk Analysis Agent, Operational Risk Agent, Industry Macro Risk Analysis Agent, and Public Opinion and Reputation Risk Analysis Agent); Identify conflicts (data bias, semantic contradictions, and type conflicts) between risk information; Build a cross-domain risk transmission map, quantify uncertainty, and schedule multi-agent collaborative deduction; Decision-making principles; Mark high-risk uncertainty events with confidence interval thresholds (>0.2 span); All outputs must comply with field standardization and JSON Schema validation.

[0188] For example, the input processing specification may be: the input you receive is risk information of 6 types of specialized agents, and the data structure is as follows:

[0189] The analysis requirements for the risk information structure of specialized agents are as follows:

[0190] Conflict detection is performed from the following perspectives: data deviation verification: comparing numerical differences between the same entity and the same indicator; semantic consistency verification: identifying descriptive contradictions (e.g., an operational risk report stating "system vulnerabilities have been fixed" versus a public opinion report stating "ongoing customer data leakage"); type verification: verifying the consistency of enumeration values ​​(e.g., a market risk recommendation of "buy" versus an industry risk recommendation of "sell" for the same security).

[0191] Exemplarily, the conflict detection method prompt may include:

[0192] Step 1: Data deviation verification method, traverse all risk information fields with the same entity and indicator, and calculate the value difference rate; if the difference rate is greater than the threshold (default 0.4), mark it as a conflict. For example, the first comparison result is a JSON string containing the conflict type, conflict entity, and conflict value.

[0193] Step 2: Semantic consistency check. Descriptive fields are compared using LLM (Large Language Model) embedding vectors. If the semantic similarity is less than 0.6, a conflict is flagged. The third comparison result is typically output as a JSON string.

[0194] Step 3: Type Verification. Check whether the action instructions for the same entity are mutually exclusive. The second comparison result is a JSON string containing the conflicting subject, conflicting fields, and field values.

[0195] Exemplarily, risk association and deduction include:

[0196] If the task is to generate a risk transmission map, it is necessary to merge overlapping nodes to construct a map; identify nodes that connect >= 3 paths as key hub nodes; output the map and deduction instructions including the starting node, deduction steps, and related specialized intelligent agents.

[0197] Exemplarily, collaborative scheduling includes collaborative task generation rules. Collaborative tasks are triggered by the detection of conflicts, high uncertainty (confidence interval span > 0.2), and the presence of risky hub nodes. Collaborative tasks contain a task ID, collaborative task type, task objective, target agent identifiers for participating agents, and collaborative operation steps, i.e., collaborative configuration information.

[0198] The Cross-Validation Coordinator (CVC) is connected to each intelligent agent, receives the risk information output by each intelligent agent, performs integrated analysis, and coordinates the intelligent agents to conduct collaborative analysis.

[0199] Exemplarily, the collaborative task types include but are not limited to the following four collaborative task types.

[0200] For example, for the four types of protocol tasks, the trigger conditions and execution process of the protocol task type can be specified, and the response method requirements, time limits, and execution round limits can be defined. The following describes the four types of collaborative tasks respectively.

[0201] The first type of collaborative task is Information Sharing Request (ISR).

[0202] An information sharing request is a one-way data request initiated by the Cross-Validation Coordinator (CVC) to a designated specialized agent to supplement or verify key information required for risk analysis. This request is submitted in the form of a structured message, requiring the agent identified by the target agent to return highly relevant, traceable data or evidence within a specified timeframe to eliminate uncertainties, resolve conflicts, or fill information gaps.

[0203] In summary, information sharing requests require clear request and response messages.

[0204] For example, if the confidence level of a risk point is lower than a threshold (e.g., <70%) and there is a data discrepancy across agents (e.g., the cash flow gap predicted by the liquidity risk agent does not match the debt maturity assessment of the credit risk agent), the collaboration task type is determined to be an information sharing request, and the collaboration configuration information is as follows:

[0205] {"taskType": "information_sharing_request",

[0206] "triggerReason": "confidence<0.7_and_cross_agent_conflict",

[0207] "agentIds": ["liquidity_risk_agent", "credit_risk_agent"],

[0208] "config": {

[0209] "liquidity_risk_agent": {

[0210] "role": "data_provider",

[0211] "deliverables": [

[0212] {

[0213] "type": "cash_flow_forecast",

[0214] "granularity": "daily",

[0215] "horizon": "90_days",

[0216] "format": "CSV+JSON",

[0217] "required_fields": ["operating_cf", "investing_cf", "financing_cf", "net_cf"],

[0218] "confidence_sources": ["model_version", "key_assumptions"]

[0219] }

[0220] ],

[0221] "deadline": "2025-07-28T10:00:00Z",

[0222] "response_schema": "ISR_Response_v1.3",

[0223] "fallback_action": "mark_as_data_deficient"

[0224] },

[0225] "credit_risk_agent": {

[0226] "role": "data_provider",

[0227] "deliverables": [

[0228] {

[0229] "type": "debt_maturity_schedule",

[0230] "granularity": "per_instrument",

[0231] "horizon": "90_days",

[0232] "format": "CSV+JSON",

[0233] "required_fields": ["due_date", "principal", "interest", "covenant_flags"],

[0234] "confidence_sources": ["contract_source", "borrower_rating"]

[0235] }

[0236] ],

[0237] "deadline": "2025-07-28T10:00:00Z",

[0238] "response_schema": "ISR_Response_v1.3",

[0239] "fallback_action": "mark_as_data_deficient"}}}

[0240] The second type of collaborative task is the Joint Analysis Task (JAT).

[0241] When the Cross-Validation Coordinator (CVC) discovers that the risk information output by a single agent is insufficient or that the risk information across agents is contradictory, and that a reliable conclusion requires multi-agent modeling, the CVC defines the collaborative task type as a joint analysis task. An example of this collaborative configuration involves having multiple agents share all relevant input data in a controlled temporary storage area, perform joint modeling based on a unified protocol (such as Bayesian model averaging, weighted voting, or joint simulation), and ultimately output a joint risk report with a consensus score.

[0242] For example, in the process of executing tasks belonging to the joint analysis task, the agent with the target agent identifier shares the relevant data to a temporary storage area managed by DCE that is bound to the task for shared use, votes on key conclusions through voting, and calculates the final consensus according to the voting result calculation method defined in the protocol (such as majority voting, weighted voting and other rules). The output report (i.e., collaborative risk information) is integrated into a result message by DCE, the core content of which includes: consensus core conclusions, referenced key data, and the final consensus degree.

[0243] The third type of collaborative task is controversial debate.

[0244] When the Cross-Validation Coordinator (CVC) detects irreconcilable conflict between two or more agents regarding the same risk point, and this conflict cannot be resolved through supplementary data (information sharing) or joint modeling, the CVC defines the collaborative task as a dispute debate. Exemplarily, the collaborative configuration includes: The conflicting parties (or multiple parties) submit statements and rebuttals in a round-by-round format (e.g., ≤10 rounds per party), publicly disclosing their core arguments, data support, and reasoning chain. The DCE ultimately compiles a list of core disagreements and a summary of each party's arguments, providing a transparent basis for subsequent manual arbitration or model weight adjustment. Specifically, agents with conflicting conclusions communicate within a prescribed mechanism, outputting their disagreements and supporting evidence. The CVC then determines the conflicting conclusions. The prescribed mechanism refers to the generation rules included in the second preset prompt text provided to the DCE (e.g., requiring evidence to support the argument, requiring a three-paragraph narrative format, etc.). The communication between the agents is a multi-round dialogue between the large language model.

[0245] For example, during a controversial debate, the protocol defines the pro- and con-siders, as well as the moderator. Each agent sends a statement or rebuttal message in a round-based system, with a threshold of 10 rounds. The DCE or a designated agent aggregates all statements and rebuttal messages, along with related citation data, extracts a list of core disagreements, and outputs a result message (i.e., collaborative risk information) that includes the core disagreements, each party's core arguments, and related citation data.

[0246] The fourth collaborative task type is Risk Cascade Simulation (RCS).

[0247] When the Cross-Validation Coordinator (CVC) identifies that a single initial risk event could trigger a multi-stage chain reaction along the business chain, it determines the collaborative task type as a risk propagation simulation. Exemplarily, the collaborative configuration information includes: Under the unified scheduling of the Dynamic Collaboration Engine (DCE), agents identified by the target agent quantify the impact of upstream risks on downstream nodes (0%–100%) step by step along a pre-defined transmission path (e.g., "raw material price fluctuations → gross profit margin → cash flow → debt repayment ability"). Ultimately, they output a complete risk propagation chain diagram, key amplification nodes, and systemic risk threshold alerts, providing decision-making basis for early intervention and scenario-based stress testing. Specifically, agents identified by the target agent simulate an initial risk event and simulate the chain reaction it triggers. For example, the collaborative configuration information includes a message to initiate the simulation, which contains: a description of the initial risk event (e.g., raw material price +30%); a recommended simulation domain path (e.g., market risk, profitability, liquidity risk); and the number of simulation steps.

[0248] For example, when executing a risk propagation task, the risk event description is initialized, followed by a definition of the risk propagation path (e.g., purchase price, profitability, and liquidity risk). The DCE sends a prediction request to the agent corresponding to each node along the path, requesting it to calculate the impact on the node under the upstream impact level (0%-100%). The process repeats until the prediction path is completed.

[0249] For example, the DCE sends a request message to Agent A, requesting an estimate of the direct impact I of the initial shock on its domain. Agent A returns a response message containing the expected impact on its domain, along with potentially impacted downstream domain B (the next business domain in the risk transmission chain), and possible impact J. Based on the potential downstream domain B and the deduction message, the DCE sends a message to Agent B (a specialized agent responsible for downstream domain B), informing it of the potential impact J and requesting an estimate of the direct impact J' and the potentially impacted downstream domain C. This process repeats until the predetermined number of deduction steps is reached or no potentially impacted downstream domains are found.

[0250] It is understandable that the triggering conditions of the four types of collaborative tasks are different, that is, under different triggering conditions, different collaborative task types are determined. Exemplarily, the triggering conditions include but are not limited to: conflict, low confidence, etc.

[0251] Exemplarily, the target agent identifier may be included in the agent role list.

[0252] Step S208: Input the collaborative agent identification set, the collaborative task type, the collaborative configuration information and the second preset prompt text into a pre-built dynamic collaboration engine, and communicate with the agent with the target agent identification according to the collaborative configuration information through the dynamic collaboration engine to obtain collaborative risk information.

[0253] Exemplarily, the basic communication framework of the Dynamic Collaboration Engine (DCE) includes, but is not limited to: structured messages in JSON Schema format; structured messages include natural language summaries, which include key points and information routing functionality. Exemplarily, DCE can be used as a message bus to record all communication history.

[0254] Exemplarily, the second preset prompt text includes, but is not limited to: execution instructions, task lifecycle, communication protocol, result requirements, and feedback loop. Exemplarily, the result requirements include format requirements for collaboration risk information, such as compliance with DCE collaboration rules. For example, the second preset prompt text is as follows:

[0255] As a Dynamic Collaboration Engine (DCE), you must strictly execute the following collaboration instructions upon receipt: 1. Task Parsing: Read the `taskType`, `agentIds`, and `config` fields, complete parsing within 500 ms, and create a task queue item. 2. Resource Scheduling: If the current CPU usage is > 80% or the memory usage is > 85%, reduce the priority of this task by 10% according to the rules and log the reason. 3. Communication Protocol: All messages must use JSON Schema v2.4; each structured message must include a natural language summary of ≤ 200 characters to facilitate human auditing. 4. Result Delivery: - Information Sharing Request: Returns a data package with a link to the original source, with missing fields explicitly marked as null. - Joint Analysis: Returns a consensus report with a consensus degree (0-1) and a table of key assumptions. - Dispute Debate: Returns a list of disagreements within 10 rounds and the core arguments of each party. - Risk Propagation Simulation: Returns a node influence matrix with the ID of the amplified node. + Whether the system threshold has been triggered (true / false). 5. Timeouts and exceptions: If any agent does not respond within 60 seconds of the deadline, the DCE automatically marks the task as timed out and updates the task status to the CVC. 6. Feedback: Within 1 second of task completion, the optimization module will receive collaborative efficiency metadata (e.g., time consumption, network bytes, peak memory usage, and protocol compliance score).

[0256] For example, DCE explicitly defines collaborative task types focused on financial risk early warning scenarios (such as information sharing, joint analysis tasks, dispute resolution, and risk propagation simulation). This differs from the common simple messaging and task allocation mechanisms. This is based on the dynamic, interconnected, and multi-source nature of financial risk. It requires collaborative task types to effectively locate risk points, support in-depth interaction from multiple perspectives, and simulate risk propagation pathways.

[0257] Exemplarily, the DCE maintains a queue of collaborative tasks to be executed, and upon receiving a new task request, calculates the real-time priority of each task. The real-time priority calculation includes the following steps C1 to C4.

[0258] Step C1: Determine the initial priority of the collaborative task type to which the task to be processed belongs.

[0259] Exemplarily, the initial priorities of tasks of different collaborative task types may be determined based on actual conditions, which is not limited here.

[0260] Step C2: Obtain the time factor of the task to be processed.

[0261] It is understandable that the priority level of time-sensitive tasks dynamically decays over time, that is, the time factor decreases, and the priority level of periodic tasks gradually increases with the waiting time, that is, the time factor increases.

[0262] For example, a time-sensitive task is one that must be completed within a specific time window. If the time window is exceeded, the task becomes ineffective or its value decreases dramatically. A periodic task is one that is repeated at a fixed interval. A certain delay is allowed for each task, but the longer the delay, the greater the impact on the overall cadence.

[0263] Step C3: Determine a priority adjustment factor based on resource utilization.

[0264] Exemplarily, the DCE monitors resource utilization and the load of each agent. For example, if the resource utilization reaches 80%, the DCE proactively reduces the priority of non-critical tasks, for example, by reducing the priority adjustment factor.

[0265] Step C4: Determine the real-time priority of the task to be processed based on the initial priority, time factor, and priority adjustment factor of the task to be processed.

[0266] For example, the collaboration risk information can also be input into the optimization module, and self-learning optimization is performed based on the early warning effect of the collaboration risk information and expert feedback to adjust the dynamic collaboration engine (DCE).

[0267] Step S209: Input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, the collaborative risk information, the collaborative intelligent entity identification set, the collaborative task type and the collaborative configuration information into a pre-built risk assessment and early warning generation model, and output the enterprise financial risk through the risk assessment and early warning generation model.

[0268] For example, the weight of the collaborative risk information output by the DCE is 0.8, the weight of the information output by the CVC is 0.1, and the weight of the risk information output by the agent is 0.1.

[0269] For example, the risk assessment and early warning generation model will call the interface of the early warning module and transmit the enterprise financial risk in the form of a message.

[0270] Exemplarily, DCE is an AI (Artificial Intelligence) agent that can implement a communication hub, understand natural language, possess logical reasoning capabilities, and call external tools and algorithmic functions. The specific functions of DCE include, but are not limited to: DCE first calculates the priority of each collaborative task, then manages a communication queue for each collaborative task. This communication queue represents the order in which agents are activated during the collaborative process. It connects to the current agent in the collaborative task sequentially through communication interface calls and informs it of the collaborative context. It then obtains the output of the current agent and activates the next agent according to the collaborative rules and communication queue. This generates collaborative risk information, which includes metadata and conclusions.

[0271] The metadata includes: task ID, indicating the task type and priority; target agent ID of the participating agents, indicating the agents participating in the task; timestamp; consensus indicator, generated by the agents during the collaboration process.

[0272] Among them, the conclusions include: for joint analysis tasks, such as consensus conclusions; for controversial debate tasks, such as dispute lists; summaries of pros and cons arguments (summarized by DCE); for risk deduction tasks, such as risk propagation paths.

[0273] Exemplarily, the process of training the risk assessment and early warning generation model includes the following steps B1.

[0274] Step B1: The sample collaborative agent identification set, sample collaborative task type and sample collaborative configuration information are used as inputs of the risk assessment and early warning generation model, and the labeled enterprise financial risk is used as the training target.

[0275] For example, corporate financial risks include but are not limited to: comprehensive risk rating, early warning information, core risk sources, risk trigger points, risk chain and other final conclusions.

[0276] For example, the risk assessment and early warning generation model can also output user-friendly explainable reports, providing users with intuitive, actionable, and deeply explorable early warning results and analysis conclusions.

[0277] The embodiment of the present application provides a method for determining the financial risk of an enterprise. Through six specialized intelligent agents, namely, liquidity risk analysis agent, credit risk analysis agent, market risk analysis agent, operational risk agent, industry macro risk analysis agent, and public opinion and reputation risk analysis agent, the six dimensions of liquidity, credit, market, operation, industry macro, and public opinion and reputation are processed respectively to achieve a "full quadrant" scan of the enterprise's financial risk. The cross-validation coordinator and the dynamic collaboration engine constitute a "two-wheel drive". The cross-validation coordinator can obtain the collaborative agent identification set, collaborative task type, and collaborative configuration information; the dynamic collaboration engine is based on the collaborative agent identification set, collaborative task type, and collaborative configuration information, and schedules multiple agents to collaborate in parallel in real time, eliminating information islands and significantly reducing misjudgments and omissions. Thus, a more accurate picture of the financial risk of the enterprise is obtained.

[0278] It is understandable that there are multiple ways to implement step S207. This application provides but is not limited to the following method, which includes the following steps D1 to D4.

[0279] Step D1: Obtain conflict detection results of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information through the cross-validation coordinator.

[0280] Before performing conflict detection, the Cross Validation Coordinator (CVC) needs to convert the risk information output by the six agents into a common machine-readable format, including but not limited to:

[0281] Unify field names: such as risk_score, hedge_ratio, and confidence_interval.

[0282] Unified enumeration values: for example, action_type can only be ["hold", "increase", "reduce", "freeze"].

[0283] Unified timestamp precision: All timestamps are in UTC (Coordinated Universal Time) milliseconds to prevent time zone ambiguity.

[0284] Conflict detection is used to discover contradictions in the risk information output by different agents. For example, the risk information output by each agent is used to construct a global operation timeline according to time; and one or more of the action type, warning information, and risk level in the same field of the risk information output by each agent at the same time are compared.

[0285] Step D2: Obtain risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information and the public opinion and reputation risk information through the cross-validation coordinator.

[0286] Step D3: Obtain the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information through the cross-validation coordinator.

[0287] The process of obtaining the accuracy of risk information output by the intelligent agent is to screen out risk events that are inaccurately calculated due to insufficient information, too wide confidence intervals or too large model divergence, and quantify their degree of uncertainty to provide information for subsequent "data supplementation and joint analysis".

[0288] Step D4: Obtain the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro-risk information, and the accuracy of the public opinion and reputation risk information through the conflict detection results, the risk association information, and the cross-validation coordinator, and obtain the collaborative intelligent agent identification set, the collaborative task type, and the collaborative configuration information.

[0289] It can be understood that the cross-validation coordinator (CVC) actively initiates collaborative tasks and provides them to the dynamic collaboration engine based on the conflict detection results, risk association information and the accuracy of each risk information.

[0290] Collaboration tasks require the identification of the target agent and the specific purpose of the collaboration, i.e., the collaboration configuration information. Collaboration tasks orchestrate the interactions of multiple agents to produce analytical results. Each collaborative task type is based on a specific collaboration protocol, which specifies the interaction logic between agents and the mechanism for reaching conclusions.

[0291] Exemplarily, interactions between agents are communicated in the form of JSON Schema messages, which include task id, sender, message type (request, response, conclusion, status update), collaborative task type, and timestamp.

[0292] In an optional implementation, step D1 includes the following steps D11 to D16.

[0293] Step D11: Obtain the first field value of the same first field of the numerical type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information.

[0294] For example, the number of the first field can be one or more. The following description takes the first field cash_gap_million (cash flow gap (million yuan)) as an example. The details are shown in Table 1.

[0295] Table 1

[0296] Agent Name First field First field value Liquidity Risk Agent "cash_gap_million" 120 Credit Risk Agent "cash_gap_million" 135 Market Risk Agent "cash_gap_million" 118 Operational Risk Agent "cash_gap_million" 120 Industry Macro Intelligence "cash_gap_million" 130 Public Opinion Reputation Intelligent Agent "cash_gap_million" 125

[0297] Step D12: Obtain a first comparison result between the absolute value of the difference between any two of the first field values ​​contained in the liquidity risk information, the first field value contained in the credit risk information, the first field value contained in the market risk information, the first field value contained in the operational risk information, the first field value contained in the industry macro risk information, and the first field value contained in the public opinion and reputation risk information and a first preset threshold.

[0298] Steps D11 to D12 are for performing data deviation verification.

[0299] Still taking Table 1 as an example, the difference between the six first field values ​​is calculated to see whether it exceeds the first preset threshold. Assume that the first preset threshold is 10 (million).

[0300] The following example uses the absolute value of any pairwise difference to illustrate.

[0301] |120-135| = 15. Since 15 > 10, the first comparison result includes: the first field value output by the operational risk agent conflicts with the first field value output by the credit risk agent.

[0302] |120-118| = 2. Since 2 ≤ 10, the first comparison result includes: there is no conflict between the first field value output by the operational risk agent and the first field value output by the market risk agent.

[0303] For example, if the first field values ​​output by different agents conflict, the “hedge ratio adjustment signal” can be determined as a conflict point.

[0304] Step D13: Obtain the second field value of the same second field of the enumeration type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information.

[0305] Exemplarily, the number of the second field may be one or more.

[0306] The following description uses the second field action_type as an example. The legal enumeration values ​​of the second field action_type include: hold, increase, reduce, and freeze. See Table 2 for details.

[0307] Table 2

[0308] Name of the agent Second field Second field value Liquidity Risk Agent "action_type" "hold" Credit Risk Agent "action_type" "increase" Market Risk Agent "action_type" "hold" Operational Risk Agent "action_type" "hold" Industry Macro Intelligence "action_type" "reduce" Public Opinion Reputation Intelligent Agent "action_type" "hold"

[0309] Step D14: Detect whether the second field value contained in the liquidity risk information, the second field value contained in the credit risk information, the second field value contained in the market risk information, the second field value contained in the operational risk information, the second field value contained in the industry macro risk information, and the second field value contained in the public opinion and reputation risk information are the same to obtain a second comparison result.

[0310] Steps D13 and D14 belong to type verification.

[0311] Still taking Table 2 as an example, the second comparison result includes: the second field value output by the credit risk agent, the second field value output by the industry macro agent and the second field value output by other agents conflict.

[0312] Step D15: Obtain the third field value of the same third field of text type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information.

[0313] Exemplarily, the number of the third field may be one or more.

[0314] The following uses the third field, risk_explanation (interpretability analysis), as an example for explanation, as shown in Table 3.

[0315] Table 3

[0316] Name of the agent The third field The third field value Liquidity Risk Agent "risk_explanation" Concentrated payments from suppliers have led to short-term liquidity constraints. It is recommended to maintain the existing financing limit. Credit Risk Agent "risk_explanation" The payment period of major customers has been extended. It is recommended to immediately increase the financing limit to hedge against potential defaults. …… …… ……

[0317] Step D16: Obtain the semantic similarity of the third field value contained in the liquidity risk information, the third field value contained in the credit risk information, the third field value contained in the market risk information, the third field value contained in the operational risk information, the third field value contained in the industry macro risk information, and the third field value contained in the public opinion and reputation risk information to obtain a third comparison result.

[0318] The conflict detection result includes the first comparison result, the second comparison result, and the third comparison result.

[0319] Steps D15 to D16 belong to semantic consistency verification.

[0320] Exemplarily, the text similarity can be calculated using an embedded large language model. Exemplarily, the text can be vectorized and cosine similarity can be calculated.

[0321] Assuming that the preset semantic threshold is 0.5, if the similarity between "Concentrated payments from suppliers lead to short-term liquidity tightness, it is recommended to maintain the existing financing limit" and "The payment period of major customers is extended, it is recommended to immediately increase the financing limit to hedge against potential defaults" is 0.31<0.5, then the third comparison result includes: the third field value output by the liquidity risk agent conflicts with the third field value output by the credit risk agent.

[0322] Exemplarily, the conflict detection result is a combination of the first comparison result, the second comparison result, and the third comparison result. Still taking Table 1, Table 2, and Table 3 as an example, the conflict detection result is:

[0323] {

[0324] "numeric_conflicts": [

[0325] ["cash_gap_million", "liquidity_vs_credit", 15]

[0326] ],

[0327] "enum_conflicts": [

[0328] ["action_type", "liquidity_vs_credit", "hold≠increase"]

[0329] ],

[0330] "semantic_conflicts": [

[0331] ["risk_explanation", "liquidity_vs_credit", 0.31] ]

[0333] }

[0334] In an optional implementation, step D2 includes the following step D21.

[0335] Step D21: Merge the same risk events in the first key risk propagation link, the second key risk propagation link, the third key risk propagation link, and the fourth key risk propagation link to obtain the risk association information.

[0336] It is understandable that the first key risk propagation link, the second key risk propagation link, the third key risk propagation link, and the fourth key risk propagation link are all links composed of multiple risk events. The first key risk propagation link, the second key risk propagation link, the third key risk propagation link, and the fourth key risk propagation link are all directed links.

[0337] Each risk event in the first, second, third, and fourth key risk propagation links can be used as a node. Identical nodes in the first, second, third, and fourth key risk propagation links can be merged into a single node. This avoids treating the same risk event as multiple different risk events.

[0338] Step D21 weaves the diverse risk events across various agents into a comprehensive causal network. By merging risk events with the same name across multiple key risk propagation links into a single node, and then connecting these nodes in their original causal order to form a larger graph, the system can identify which risks are actually the same, eliminating ambiguity and clarifying the true chain reaction path.

[0339] The following is an example, see Table 4 for details.

[0340] Table 4

[0341] Name of the agent Key risk link name Specific key risk links Credit Risk Analysis Agent The first key risk communication link [Large customer payment extension → Decreased accounts receivable turnover → Decreased operating cash flow → Interest coverage ratio <1.5 → Credit default] Market Risk Agent The second key risk communication link [Rising raw material prices → rising costs → declining gross profit margins → declining operating cash flow → worsening debt coverage ratios → hindered refinancing] Operational Risk Agent The third key risk transmission link [Financial system downtime → payment order delays → supplier collection harassment → operating cash flow decline → bank credit restrictions triggered → cross-default] Industry Macro Risk Analysis Agent The fourth key risk communication link [The company's credit line has been cut → the liquidity gap has widened → debt rollover has become more difficult → the industry's systemic default risk has increased]

[0342] As shown in Table 4, since the first key risk propagation link, the second key risk propagation link, and the third key risk propagation link all have the "decline in operating cash flow" node, these three nodes can be merged into one node, thereby achieving the role of risk event disambiguation.

[0343] In an optional implementation, step D3 includes the following steps D31 to D36.

[0344] Exemplarily, the uncertainty metadata includes, but is not limited to, risk level and confidence interval.

[0345] For example, the accuracy of the risk information output by each agent can be determined based on preset rules. For example, the preset rules include: if the confidence interval width (i.e., |upper bound - lower bound|) > a preset length, the risk level is recorded as an uncertain event, i.e., the risk information is inaccurate; otherwise, the risk level is considered a credible event, i.e., the risk information is accurate.

[0346] Illustratively, the confidence interval expresses a possible range of confidence levels for the risk level.

[0347] For example, the risk level of a risk event is 78 (out of 100) and the confidence interval is [0.72, 0.84]. Assuming the preset length is 0.2, since the span of the confidence interval = 0.12 < 0.2, the risk level of this risk event is marked as a credible event.

[0348] It can be understood that uncertainty metadata is the structured information that is additionally output by the intelligent agent when generating risk information. This structured information is used to quantify the credibility, missing information and sample limitations of the risk information, so that the cross-validation coordinator (CVC) can quickly determine the accuracy of the risk information.

[0349] Step D31: Determine the accuracy of the liquidity risk information based on the uncertainty metadata output by the liquidity risk analysis agent.

[0350] Assume that the uncertainty metadata corresponding to the liquidity risk analysis agent includes: {"risk_point": "Cash flow gap","risk_level": 82,"confidence": [0.70, 0.95]}, then the span of the confidence interval = 0.25 > 0.20, so the risk level 82 is an uncertain event, that is, the accuracy of the liquidity risk information is inaccurate.

[0351] Step D32: Determine the accuracy of the credit risk information based on the uncertainty metadata output by the credit risk analysis agent.

[0352] Assume that the uncertainty metadata corresponding to the credit risk analysis agent includes: {"risk_point": "Large customer default probability","risk_level": 65, "confidence": [0.60, 0.70]}, and the span of the confidence interval = 0.10 ≤ 0.20. Therefore, risk level 65 is a credible event, that is, the accuracy of the credit risk information is accurate.

[0353] Step D33: Determine the accuracy of the market risk information based on the uncertainty metadata output by the market risk analysis agent.

[0354] Assume that the uncertainty metadata corresponding to the market risk analysis agent includes: {"risk_point": "The impact of rising interest rates on bond valuations", "risk_level": 74, "confidence": [0.65, 0.90]}, and the span of the confidence interval = 0.25 > 0.20. Therefore, risk level 74 is an uncertain event, that is, the accuracy of market risk information is inaccurate.

[0355] Step D34: Determine the accuracy of the operational risk information based on the uncertainty metadata output by the operational risk agent.

[0356] Assume that the uncertainty metadata corresponding to the operational risk agent includes: {"risk_point": "Overall industry prosperity declines", "risk_level": 81, "confidence": [0.68, 0.91]}, and the span of the confidence interval = 0.23 > 0.20. Therefore, risk level 81 is an uncertain event, that is, the accuracy of the operational risk information is inaccurate.

[0357] Step D35: Determine the accuracy of the industry macro-risk information based on the uncertainty metadata output by the industry macro-risk analysis agent.

[0358] Assume that the uncertainty metadata corresponding to the industry macro-risk analysis agent includes: {"risk_point": "Negative hot search diffusion speed", "risk_level": 88, "confidence": [0.85, 0.89]}, and the span of the confidence interval = 0.04 ≤ 0.20. Therefore, risk level 88 is a credible event, that is, the accuracy of the industry macro-risk information is accurate.

[0359] Step D36: Determine the accuracy of the public opinion and reputation risk information based on the uncertainty metadata output by the public opinion and reputation risk analysis agent.

[0360] Assume that the uncertainty metadata corresponding to the public opinion and reputation risk analysis agent includes: { "risk_point":"Negative hot search diffusion speed", "risk_level": 88, "confidence": [0.85, 0.89]}, and the span of the confidence interval = 0.04 ≤ 0.20, so the risk level 88 is a credible event, that is, the accuracy of the public opinion and reputation risk information is accurate.

[0361] The intelligent agents mentioned in the embodiments of this application refer to large models pre-trained with massive amounts of data from the financial and accounting fields. This application is a collaborative financial risk early warning technology based on dynamic collaboration among multiple intelligent agents, including a group of specialized financial risk early warning intelligent agents. In addition to financial risk analysis capabilities, data processing capabilities, natural language dialogue capabilities, data collection capabilities, data query capabilities, file reading capabilities, etc., each intelligent agent also focuses on specific financial risk identification, assessment, and quantification, and has independent risk perception, computational reasoning, and result output capabilities.

[0362] For example, the risk information output by each agent can be a risk analysis report. This analysis report is structured data standardized by a unified semantic framework. The unified semantic framework establishes a standard terminology library. For example, action types include payment freeze and collateral add, and entity types include supplier and debtor. This information is expressed using a standardized JSON format and uniformly named. When designing various CVC features, comparative analysis of the analysis report data based on standardized specialized agents is performed on a field-by-field basis.

[0363] The above describes a method for determining enterprise financial risk provided by an embodiment of the present application. The following describes an apparatus for executing the above method for determining enterprise financial risk.

[0364] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of a device for determining corporate financial risk provided by an embodiment of the present application. Figure 3 As shown, the enterprise financial risk determination device includes:

[0365] The first acquisition module 301 is used to input the enterprise's cash flow gap value within a first preset future time period, the proportion of highly liquid assets, the proportion of liabilities due within a second preset future time period, the bid-ask spread of key assets, the available credit limit, and the industry negative public opinion index into a pre-built liquidity risk analysis agent, and output liquidity risk information through the liquidity risk analysis agent, the liquidity risk information including a liquidity risk score, an explainability analysis, and first cross-agent collaboration information; the first cross-agent collaboration information includes an agent identifier that affects the liquidity risk score;

[0366] The second acquisition module 302 is configured to input the debt-to-asset ratio, interest coverage ratio, operating cash flow, total liabilities, industry prosperity index, and customer and supplier concentration of the enterprise into a pre-built credit risk analysis agent, and output credit risk information through the credit risk analysis agent. The credit risk information includes a credit risk score, a first key risk propagation link, and second cross-agent collaboration information. The second cross-agent collaboration information includes the identifiers of agents that affect the credit risk score. The first key risk propagation link includes a link consisting of risk events that lead to the credit risk score.

[0367] The third acquisition module 303 is configured to input the enterprise's volatility index, profit volatility, key interest rate curve, commodity main contract price, stock-home correlation index, and foreign exchange risk exposure ratio into a pre-built market risk analysis agent, and output market risk information through the market risk analysis agent. The market risk information includes the current market risk score, a second key risk propagation link, and third cross-agent collaboration information; the second key risk propagation link includes a link consisting of risk events that lead to the market risk score; and the third cross-agent collaboration information includes the identity of the agent that affects the market risk score.

[0368] A fourth acquisition module 304 is configured to input the enterprise's compliance check exception frequency, system failure time percentage, abnormal access request volume, and supplier environmental and social governance score into a pre-built operational risk agent, and output operational risk information through the operational risk agent. The operational risk information includes the risk source that causes the operational risk, a third key risk propagation link, and fourth cross-agent collaboration information; the third key risk propagation link includes a link consisting of risk events that cause the risk source, and the fourth cross-agent collaboration information includes the identity of the agent that affects the risk source.

[0369] A fifth acquisition module 305 is configured to input the enterprise's capacity utilization fluctuations, changes in enterprise market share, public opinion sentiment index, core supplier concentration, and upstream raw material price fluctuation transmission coefficient into a pre-built industry macro-risk analysis agent, and output industry macro-risk information through the industry macro-risk analysis agent. The industry macro-risk information includes an industry macro-risk rating and a fourth key risk propagation link. The fourth key risk propagation link includes a link consisting of risk events that lead to the industry macro-risk rating.

[0370] The sixth acquisition module 306 is configured to input the enterprise's KOL negative communication heat, the network-wide sentiment polarity index, the brand search volume volatility, the industry negative public opinion index, and the frequency of regulatory penalties into a pre-built public opinion and reputation risk analysis agent, and output public opinion and reputation risk information through the public opinion and reputation risk analysis agent. The public opinion and reputation risk information includes the reputation risk score at the current time, the public opinion propagation path, and fifth cross-agent collaboration information; the public opinion propagation path includes a link consisting of key nodes that lead to the reputation risk score, and the fifth cross-agent collaboration information includes the identity of the agent that affects the reputation risk score;

[0371] A seventh acquisition module 307 is configured to input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, and the first preset prompt text into a pre-built cross-validation coordinator, and obtain, through the cross-validation coordinator, a collaborative agent identification set, a collaborative task type, and collaborative configuration information; the collaborative agent identification set includes multiple collaborative target agent identifications, and the collaborative configuration information includes collaborative processes corresponding to the multiple target agent identifications;

[0372] An eighth acquisition module 308 is configured to input the set of collaborative agent identifiers, the collaborative task type, the collaborative configuration information, and the second preset prompt text into a pre-built dynamic collaboration engine, and communicate with the agent having the target agent identifier through the dynamic collaboration engine according to the collaborative configuration information to obtain collaboration risk information;

[0373] The ninth acquisition module 309 is used to input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, the collaborative risk information, the collaborative intelligent entity identification set, the collaborative task type and the collaborative configuration information into a pre-built risk assessment and early warning generation model, and output the enterprise financial risk through the risk assessment and early warning generation model.

[0374] In a possible implementation, the seventh acquisition module includes:

[0375] A first acquisition unit is configured to acquire, through the cross-validation coordinator, a conflict detection result of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0376] A second acquisition unit is configured to acquire, through the cross-validation coordinator, risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0377] A third acquisition unit is configured to acquire, through the cross-validation coordinator, the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information;

[0378] The fourth acquisition unit is used to obtain the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information through the conflict detection results, the risk association information and the cross-validation coordinator, and obtain the collaborative intelligent agent identification set, the collaborative task type and the collaborative configuration information.

[0379] In an optional implementation, the first acquiring unit includes:

[0380] A first acquisition subunit is configured to acquire a first field value of a same first field of a numerical type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0381] a second acquisition subunit, configured to obtain a first comparison result between the absolute value of the difference between any two of the first field values ​​included in the liquidity risk information, the first field value included in the credit risk information, the first field value included in the market risk information, the first field value included in the operational risk information, the first field value included in the industry macro risk information, and the first field value included in the public opinion and reputation risk information and a first preset threshold;

[0382] A third acquisition subunit is configured to acquire a second field value of the same second field of the enumeration type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0383] a first detection subunit, configured to detect whether the second field value included in the liquidity risk information, the second field value included in the credit risk information, the second field value included in the market risk information, the second field value included in the operational risk information, the second field value included in the industry macro risk information, and the second field value included in the public opinion and reputation risk information are the same, to obtain a second comparison result;

[0384] a fourth acquisition subunit, configured to acquire a third field value of the same third field of text type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information;

[0385] a fifth acquiring subunit, configured to acquire semantic similarities among the third field value included in the liquidity risk information, the third field value included in the credit risk information, the third field value included in the market risk information, the third field value included in the operational risk information, the third field value included in the industry macro risk information, and the third field value included in the public opinion and reputation risk information, to obtain a third comparison result;

[0386] The conflict detection result includes the first comparison result, the second comparison result, and the third comparison result.

[0387] In an optional implementation, the second obtaining unit includes:

[0388] The sixth acquisition subunit is used to merge the same risk events in the first key risk propagation link, the second key risk propagation link, the third key risk propagation link and the fourth key risk propagation link to obtain the risk association information.

[0389] In an optional implementation, the third obtaining unit includes:

[0390] A first determining subunit, configured to determine the accuracy of the liquidity risk information based on the uncertainty metadata output by the liquidity risk analysis agent;

[0391] A second determining subunit is configured to determine the accuracy of the credit risk information based on the uncertainty metadata output by the credit risk analysis agent;

[0392] A third determining subunit is configured to determine the accuracy of the market risk information based on the uncertainty metadata output by the market risk analysis agent;

[0393] a fourth determining subunit, configured to determine the accuracy of the operational risk information based on the uncertainty metadata output by the operational risk agent;

[0394] a fifth determining subunit, configured to determine the accuracy of the industry macro risk information based on the uncertainty metadata output by the industry macro risk analysis agent;

[0395] The sixth determination subunit is used to determine the accuracy of the public opinion and reputation risk information based on the uncertainty metadata output by the public opinion and reputation risk analysis agent.

[0396] An electronic device is also provided in an embodiment of the present application. Figure 4 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 4 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0397] like Figure 4 As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 401, which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 402 or programs loaded from a storage device 408 into a random access memory (RAM) 403. When the electronic device is powered on, the RAM 403 also stores various programs and data required for the operation of the electronic device. The processing device 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0398] Typically, the following devices may be connected to the I / O interface 405: an input device 406 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 407 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 408 including, for example, a memory card, a hard disk, etc.; and a communication device 409. The communication device 409 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Figure 4 The electronic device is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0399] An embodiment of the present application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the enterprise financial risk determination methods provided in the embodiments of the present application.

[0400] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any one of the enterprise financial risk determination methods provided in the embodiment of the present application.

[0401] It should also be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines.

[0402] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memories, special components, etc. In general, all functions performed by computer programs can be easily implemented with corresponding hardware, and the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits or special circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, training equipment, or network equipment, etc.) to execute the methods described in each embodiment of the present application.

[0403] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0404] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, a computer, a training device or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website, a computer, a training device or a data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims

1. A method for determining enterprise financial risk, characterized in that: include: Input the enterprise's cash flow gap value within a first preset future time period, the proportion of highly liquid assets, the proportion of liabilities due within a second preset future time period, the bid-ask spread of key assets, the available credit line, and the industry negative public opinion index into a pre-built liquidity risk analysis agent, and output liquidity risk information through the liquidity risk analysis agent. The liquidity risk information includes a liquidity risk score, an explainability analysis, and first cross-agent collaboration information; the first cross-agent collaboration information includes an agent identifier that affects the liquidity risk score; Input the debt-to-asset ratio, interest coverage ratio, operating cash flow, total liabilities, industry prosperity index, and customer and supplier concentration of the enterprise into a pre-built credit risk analysis agent, and output credit risk information through the credit risk analysis agent. The credit risk information includes a credit risk score, a first key risk propagation link, and second cross-agent collaboration information, wherein the second cross-agent collaboration information includes the identity of the agent that affects the credit risk score; The first key risk propagation link includes a link consisting of risk events that lead to the occurrence of the credit risk score; Input the enterprise's volatility index, profit volatility, key interest rate curve, commodity main contract price, stock-home correlation index, and foreign exchange risk exposure ratio into a pre-built market risk analysis agent, and output market risk information through the market risk analysis agent. The market risk information includes the market risk score at the current time, the second key risk propagation link, and the third cross-agent collaboration information; The second key risk propagation link includes a link consisting of risk events that lead to the occurrence of the market risk score; the third cross-agent collaboration information includes the identity of the agent that affects the market risk score; Input the company's compliance check exception frequency, system failure time ratio, abnormal access request volume, and supplier environmental and social governance score into a pre-built operational risk agent, and output operational risk information through the operational risk agent. The operational risk information includes the risk source that causes the operational risk, the third key risk propagation link, and the fourth cross-agent collaboration information; The third key risk propagation link includes a link consisting of risk events that lead to the occurrence of the risk source, and the fourth cross-agent collaboration information includes the identity of the agent that affects the risk source; Input the enterprise's capacity utilization fluctuations, enterprise market share changes, public opinion sentiment index, core supplier concentration, and upstream raw material price fluctuation transmission coefficient into a pre-built industry macro risk analysis agent, and output industry macro risk information through the industry macro risk analysis agent. The industry macro risk information includes the industry macro risk rating and the fourth key risk propagation link. The fourth key risk propagation link includes the link consisting of risk events that lead to the industry macro risk rating. The negative KOL dissemination heat of the enterprise, the network-wide sentiment polarity index, the brand search volume volatility, the industry negative public opinion index, and the frequency of regulatory penalties are input into a pre-built public opinion and reputation risk analysis agent, and the public opinion and reputation risk analysis agent outputs public opinion and reputation risk information, which includes the reputation risk score at the current time, the public opinion propagation path, and the fifth cross-agent collaboration information; the public opinion propagation path includes a link composed of key nodes that lead to the reputation risk score, and the fifth cross-agent collaboration information includes the identity of the agent that affects the reputation risk score; The liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, and the first preset prompt text are input into a pre-built cross-validation coordinator, and a collaborative agent identification set, a collaborative task type, and collaborative configuration information are obtained through the cross-validation coordinator; the collaborative agent identification set includes multiple collaborative target agent identifications, and the collaborative configuration information includes the collaborative processes corresponding to the multiple target agent identifications; Inputting the set of collaborative agent identifiers, the collaborative task type, the collaborative configuration information, and a second preset prompt text into a pre-built dynamic collaboration engine, and communicating with the agent having the target agent identifier through the dynamic collaboration engine according to the collaborative configuration information to obtain collaboration risk information; The liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, the collaborative risk information, the collaborative intelligent entity identification set, the collaborative task type and the collaborative configuration information are input into a pre-built risk assessment and early warning generation model, and the enterprise financial risk is output through the risk assessment and early warning generation model.

2. The method for determining enterprise financial risk according to claim 1, characterized in that: The step of obtaining the collaborative agent identification set, collaborative task type, and collaborative configuration information through the cross-validation coordinator includes: Obtaining, through the cross-validation coordinator, conflict detection results of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information; Obtaining, through the cross-validation coordinator, risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information; Obtaining the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information through the cross-validation coordinator; The accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro-risk information, and the accuracy of the public opinion and reputation risk information are obtained through the conflict detection results, the risk association information and the cross-validation coordinator, and the collaborative intelligent agent identification set, the collaborative task type and the collaborative configuration information are obtained.

3. The method for determining enterprise financial risk according to claim 2, characterized in that: The step of obtaining, through the cross-validation coordinator, conflict detection results of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information includes: Obtaining a first field value of the same first field of a numerical type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information; Obtaining a first comparison result between the absolute value of the difference between any two of the first field values ​​included in the liquidity risk information, the first field value included in the credit risk information, the first field value included in the market risk information, the first field value included in the operational risk information, the first field value included in the industry macro risk information, and the first field value included in the public opinion and reputation risk information and a first preset threshold; Obtaining a second field value of the same second field of the enumeration type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information; detecting whether the second field value included in the liquidity risk information, the second field value included in the credit risk information, the second field value included in the market risk information, the second field value included in the operational risk information, the second field value included in the industry macro risk information, and the second field value included in the public opinion and reputation risk information are the same, to obtain a second comparison result; Obtaining a third field value of the same third field of text type in the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information; Obtaining semantic similarities among the third field value included in the liquidity risk information, the third field value included in the credit risk information, the third field value included in the market risk information, the third field value included in the operational risk information, the third field value included in the industry macro risk information, and the third field value included in the public opinion and reputation risk information, to obtain a third comparison result; The conflict detection result includes the first comparison result, the second comparison result, and the third comparison result.

4. The method for determining enterprise financial risk according to claim 2, characterized in that: The step of obtaining risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information through the cross-validation coordinator includes: The same risk events in the first key risk propagation link, the second key risk propagation link, the third key risk propagation link, and the fourth key risk propagation link are merged to obtain the risk association information.

5. The method for determining enterprise financial risk according to claim 2, characterized in that: The step of obtaining, through the cross-validation coordinator, the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information includes: Determining the accuracy of the liquidity risk information based on the uncertainty metadata output by the liquidity risk analysis agent; Determining the accuracy of the credit risk information based on the uncertainty metadata output by the credit risk analysis agent; Determining the accuracy of the market risk information based on the uncertainty metadata output by the market risk analysis agent; Determining the accuracy of the operational risk information based on the uncertainty metadata output by the operational risk agent; Determining the accuracy of the industry macro risk information based on the uncertainty metadata output by the industry macro risk analysis agent; Based on the uncertainty metadata output by the public opinion and reputation risk analysis agent, the accuracy of the public opinion and reputation risk information is determined.

6. An enterprise financial risk determination device, characterized in that: include: A first acquisition module is configured to input the enterprise's cash flow gap value within a first preset future time period, the proportion of highly liquid assets, the proportion of liabilities due within a second preset future time period, the bid-ask spread of key assets, the available credit limit, and the industry negative public opinion index into a pre-built liquidity risk analysis agent, and output liquidity risk information through the liquidity risk analysis agent, the liquidity risk information including a liquidity risk score, an explainability analysis, and first cross-agent collaboration information; the first cross-agent collaboration information includes an agent identifier that affects the liquidity risk score; A second acquisition module is configured to input the debt-to-asset ratio, interest coverage ratio, operating cash flow, total liabilities, industry prosperity index, and customer and supplier concentration of the enterprise into a pre-built credit risk analysis agent, and output credit risk information through the credit risk analysis agent, the credit risk information including a credit risk score, a first key risk propagation link, and second cross-agent collaboration information, the second cross-agent collaboration information including agent identifiers that influence the credit risk score; The first key risk propagation link includes a link consisting of risk events that lead to the occurrence of the credit risk score; The third acquisition module is used to input the enterprise's volatility index, profit volatility, key interest rate curve, commodity main contract price, stock-home correlation index, and foreign exchange risk exposure ratio into a pre-built market risk analysis agent, and output market risk information through the market risk analysis agent. The market risk information includes the market risk score at the current time, the second key risk propagation link, and the third cross-agent collaboration information; The second key risk propagation link includes a link consisting of risk events that lead to the occurrence of the market risk score; the third cross-agent collaboration information includes the identity of the agent that affects the market risk score; A fourth acquisition module is configured to input the enterprise's compliance check exception frequency, system failure time ratio, abnormal access request volume, and supplier environmental and social governance score into a pre-built operational risk agent, and output operational risk information through the operational risk agent. The operational risk information includes the risk source causing the operational risk, the third key risk propagation link, and the fourth cross-agent collaboration information. The third key risk propagation link includes a link consisting of risk events that lead to the occurrence of the risk source, and the fourth cross-agent collaboration information includes the identity of the agent that affects the risk source; A fifth acquisition module is configured to input the enterprise's capacity utilization fluctuations, changes in enterprise market share, public opinion sentiment index, core supplier concentration, and upstream raw material price fluctuation transmission coefficient into a pre-built industry macro-risk analysis agent, and output industry macro-risk information through the industry macro-risk analysis agent. The industry macro-risk information includes an industry macro-risk rating and a fourth key risk propagation link. The fourth key risk propagation link includes a link consisting of risk events that lead to the industry macro-risk rating. The sixth acquisition module is used to input the enterprise's KOL negative communication heat, the network-wide sentiment polarity index, the brand search volume volatility, the industry negative public opinion index, and the frequency of regulatory penalties into a pre-built public opinion and reputation risk analysis agent, and output public opinion and reputation risk information through the public opinion and reputation risk analysis agent. The public opinion and reputation risk information includes the reputation risk score at the current time, the public opinion propagation path, and the fifth cross-agent collaboration information; the public opinion propagation path includes a link composed of key nodes that lead to the reputation risk score, and the fifth cross-agent collaboration information includes the identity of the agent that affects the reputation risk score; a seventh acquisition module, configured to input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, and a first preset prompt text into a pre-built cross-validation coordinator, and obtain, through the cross-validation coordinator, a collaborative agent identification set, a collaborative task type, and collaborative configuration information; the collaborative agent identification set includes multiple collaborative target agent identifications, and the collaborative configuration information includes collaborative processes corresponding to the multiple target agent identifications; an eighth acquisition module, configured to input the set of collaborative agent identifiers, the collaborative task type, the collaborative configuration information, and a second preset prompt text into a pre-built dynamic collaboration engine, and communicate with the agent having the target agent identifier through the dynamic collaboration engine according to the collaborative configuration information to obtain collaboration risk information; The ninth acquisition module is used to input the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, the public opinion and reputation risk information, the collaborative risk information, the collaborative intelligent body identification set, the collaborative task type and the collaborative configuration information into a pre-built risk assessment and early warning generation model, and output the enterprise financial risk through the risk assessment and early warning generation model.

7. The enterprise financial risk determination device according to claim 6, characterized in that: The seventh acquisition module includes: A first acquisition unit is configured to acquire, through the cross-validation coordinator, a conflict detection result of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information; A second acquisition unit is configured to acquire, through the cross-validation coordinator, risk correlation information of the liquidity risk information, the credit risk information, the market risk information, the operational risk information, the industry macro risk information, and the public opinion and reputation risk information; A third acquisition unit is configured to acquire, through the cross-validation coordinator, the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information; The fourth acquisition unit is used to obtain the accuracy of the liquidity risk information, the accuracy of the credit risk information, the accuracy of the market risk information, the accuracy of the operational risk information, the accuracy of the industry macro risk information, and the accuracy of the public opinion and reputation risk information through the conflict detection results, the risk association information and the cross-validation coordinator, and obtain the collaborative intelligent agent identification set, the collaborative task type and the collaborative configuration information.

8. A computer program product, characterized in that The method comprises computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the enterprise financial risk determination method according to any one of claims 1 to 5.

9. An electronic device, characterized in that: comprising at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is configured to execute the computer program so that the electronic device can implement the enterprise financial risk determination method according to any one of claims 1 to 5.

10. A computer storage medium, characterized in that The storage medium carries one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the enterprise financial risk determination method as described in any one of claims 1 to 5.

Images