A consultation processing method and device, a storage medium and an electronic device

Abstract

The specification discloses a consultation processing method and device, a storage medium and an electronic equipment, wherein the method comprises: obtaining risk control consultation information input by a user to a risk control explanation intelligent agent service for a transaction risk control system; controlling a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consultation information; obtaining risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information; and outputting the risk control decision explanation information to the user based on the risk control explanation intelligent agent service.

Description

TECHNICAL FIELD

[0001] The present specification relates to the technical field of computer technology, and particularly relates to a consultation processing method and device, a storage medium and an electronic equipment. BACKGROUND

[0002] Risk control (referred to as risk control) on platform transaction services such as consumer financial transaction services, network shopping transaction services, express logistics transaction services, etc. is mainly realized by pre-developed transaction risk control systems. Data collection, three-party service integration, algorithm execution and other functions are integrated in the transaction risk control system. After the development of the transaction risk control system, the transaction risk control system is equivalent to a black box for non-system developers. Non-system developers have risk control consultation needs for the transaction risk control system in actual application. SUMMARY

[0003] The present specification provides a consultation processing method, device, storage medium and electronic equipment, and the technical solution is as follows:

[0004] In a first aspect, the present specification provides a consultation processing method, which comprises:

[0005] Obtaining risk control consultation information for a transaction risk control system input by a user to a risk control explanation intelligent agent service, wherein the risk control consultation information comprises consultation information for a system risk control transaction state;

[0006] Controlling a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consultation information, and obtaining risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information;

[0007] Outputting risk control decision explanation information to the user based on the risk control explanation intelligent agent service.

[0008] In a second aspect, the present specification provides a risk control explanation large language model training method, which comprises

[0009] Creating an initial risk control explanation large language model under a risk control explanation scenario by using a basic large language model;

[0010] Obtaining sample risk control consultation information for a transaction risk control system under the risk control explanation scenario;

[0011] The sample risk control consultation information is used for model training of an initial risk control explanation large language model. In the model training process, sample risk control attribution tool information is determined by the initial risk control explanation large language model, and prediction risk control decision explanation information for the transaction risk control system is obtained based on the sample risk control attribution tool information. The initial risk control explanation large language model is adjusted in model parameters based on the prediction risk control decision explanation information, until the initial risk control explanation large language model ends the model training, and a risk control explanation large language model is obtained.

[0012] In a third aspect, the present specification provides a consultation processing device, the device comprising:

[0013] An acquisition module is configured to acquire risk control consultation information for a transaction risk control system input by a user to a risk control explanation intelligent agent service, wherein the risk control consultation information comprises consultation information for a system risk control transaction state;

[0014] A determination module is configured to control a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consultation information, and obtain risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information;

[0015] An output module is configured to output risk control decision explanation information to the user based on the risk control explanation intelligent agent service.

[0016] In a fourth aspect, the present specification provides a risk control explanation large language model training device, the device comprising

[0017] A model creation module is configured to create an initial risk control explanation large language model in a risk control explanation scenario by using a basic large language model;

[0018] An information acquisition module is configured to acquire sample risk control consultation information for a transaction risk control system in the risk control explanation scenario;

[0019] A model training module is configured to use the sample risk control consultation information to train an initial risk control explanation large language model. In the model training process, sample risk control attribution tool information is determined by the initial risk control explanation large language model, and prediction risk control decision explanation information for the transaction risk control system is obtained based on the sample risk control attribution tool information. The initial risk control explanation large language model is adjusted in model parameters based on the prediction risk control decision explanation information, until the initial risk control explanation large language model ends the model training, and a risk control explanation large language model is obtained.

[0020] In a fifth aspect, the present specification provides a computer storage medium, which stores at least one instruction adapted to be loaded by a processor and execute the method steps of one or more embodiments of the present specification.

[0021] In a sixth aspect, the present specification provides a computer program product, which stores at least one instruction suitable for being loaded by a processor and executing the method steps of one or more embodiments of the present specification.

[0022] In a seventh aspect, the present specification provides an electronic device, which can include a processor and a memory; wherein the memory stores a computer program suitable for being loaded by the processor and executing the method steps of one or more embodiments of the present specification.

[0023] The technical solutions provided by some embodiments of the present specification have at least the following beneficial effects:

[0024] In one or more embodiments of the present specification, the electronic device obtains the risk control consulting information for the transaction risk control system input by the user to the risk control explanation intelligent agent service, controls the target risk control attribution tool information based on the risk control consulting information determined by the risk control explanation large language model, and obtains the risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information. Based on the risk control explanation intelligent agent service, the user can output the risk control decision explanation information to the user. By using the method of combining the risk control explanation intelligent agent service and the risk control explanation large language model, the automation and accuracy of risk control management are significantly improved, so that the user can obtain intuitive and personalized risk control explanation through simple natural language input. In addition to reducing the dependence on professional risk control attribution tool knowledge and improving user experience, the response speed of risk control decision is also accelerated, effectively supporting fast and accurate risk control evaluation and management. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the present specification or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present specification, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0026] Figure 1 is a scene schematic diagram of a consulting processing system provided by the present specification;

[0027] Figure 2 is a flow schematic diagram of a consulting processing method provided by the present specification;

[0028] Figure 3 is a schematic diagram of a consulting processing interface provided by the present specification;

[0029] Figure 4 is a schematic diagram of another consulting processing interface provided by the present specification;

[0030] Figure 5 is a flowchart of a consultation processing method provided by the present specification;

[0031] Figure 6 is a flowchart of information generation provided by the present specification;

[0032] Figure 7 is a flowchart of a risk control explanation large language model training method provided by the present specification;

[0033] Figure 8 is a device structure diagram of a consultation processing device provided by the present specification;

[0034] Figure 9 is a structure diagram of a risk control explanation large language model training device provided by the present specification;

[0035] Figure 10 is a structure diagram of an electronic device provided by the present specification. DETAILED DESCRIPTION

[0036] The technical solutions in the present specification will be described in detail below with reference to the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present specification, not all embodiments. Based on the embodiments in the present specification, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present specification.

[0037] In the description of the present specification, it should be understood that the terms "first", "second" and the like are used only for the purpose of description and should not be understood as indicating or implying relative importance. In the description of the present specification, it should be noted that, unless otherwise explicitly specified and limited, "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units not listed, or optionally includes other steps or units inherent to the process, method, product or device. The specific meaning of the above terms in the present specification can be understood by the person of ordinary skill in the art. In addition, in the description of the present specification, "multiple" means two or more, unless otherwise specified. "And / or", which describes the relationship between the associated objects, means that there can be three relationships, for example, A and / or B can mean that there are three cases of A alone, A and B together, and B alone. The character " / " generally represents an "or" relationship between the associated objects before and after it.

[0038] In the related art, non-system developers such as customer side, customer service side, user side, and regulatory side have risk control consulting needs in actual application of the transaction risk control system, and need to perform risk control abnormal attribution on platform service transactions maintained based on the transaction risk control system. The risk control abnormal attribution is a relatively complex proposition. In daily processing work, non-system developers such as customer side, customer service side, user side, and regulatory side need to consult system developers (such as risk control experts). System developers occupy a large amount of platform human resources and time cost because they need to frequently answer various risk abnormal questions.

[0039] The present specification will be described in detail below with reference to specific embodiments.

[0040] Please refer to Figure 1 , a scene schematic diagram of a consultation processing system provided by the present specification. As Figure 1 indicated, the consultation processing system can at least include a client cluster and a service platform 100.

[0041] The client cluster can include at least one client, such as Figure 1 indicated, specifically including a client 1 corresponding to a user 1, a client 2 corresponding to a user 2, …, and a client n corresponding to a user n, n is an integer greater than 0.

[0042] Each client in the client cluster can be an electronic device with communication function, including but not limited to: wearable devices, handheld devices, personal computers, tablet computers, vehicle-mounted devices, smart phones, computing devices, or other processing devices connected to wireless modems, etc. In different networks, the electronic device can be called by different names, such as: user equipment, access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device, cellular phone, cordless phone, personal digital assistant (PDA), 5G network or future evolution network electronic device, etc.

[0043] The service platform 100 can be a separate server device, such as: rack-mounted, blade, tower, or cabinet server device, or using workstations, mainframe computers, and other hardware devices with strong computing power; It can also be a server cluster composed of multiple servers. Each server in the service cluster can be composed in a symmetrical manner, where each server is functionally equivalent and positionally equivalent in the transaction link. Each server can independently provide services to the outside. The independent service can be understood as not needing the assistance of another server.

[0044] In one or more embodiments of the present specification, the service platform 100 can establish a communication connection with at least one client in the client cluster, and based on the communication connection, complete the interaction of data in the consultation processing process.

[0045] It should be noted that the service platform 100 and at least one client in the client cluster establish a communication connection through a network for interactive communication, wherein the network can be a wireless network or a wired network. The wireless network includes but is not limited to a cellular network, a wireless local area network, an infrared network or a Bluetooth network. The wired network includes but is not limited to an Ethernet, a universal serial bus (USB) or a controller area network. In one or more embodiments of the present specification, technologies and / or formats including Hyper Text Mark-up Language (HTML), Extensible Markup Language (XML) and the like are used to represent data (such as target compression package) exchanged through the network. In addition, all or some links can be encrypted using conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec) and the like. In other embodiments, custom and / or dedicated data communication technologies can be used instead of or in addition to the above data communication technologies.

[0046] The consultation processing system embodiments provided in the present specification and the consultation processing method in one or more embodiments belong to the same concept. The execution subject corresponding to the consultation processing method involved in one or more embodiments of the present specification can be an electronic device, which can be the service platform 100 described above. The execution subject corresponding to the consultation processing method involved in one or more embodiments of the present specification can also be an electronic device corresponding to a client, which is determined based on the actual application environment. The consultation processing system embodiment embodies the implementation process, which can be seen from the method embodiments described below, and will not be described here.

[0047] Based on Figure 1 The scene schematic diagram shown below will be described in detail.

[0048] Please refer to Figure 2A flowchart of an advisory processing method is provided for one or more embodiments of the present specification, which can rely on a computer program implementation and can be run on an advisory processing device based on the Von Neumann system. The computer program can be integrated in an application or run as an independent tool application. The advisory processing device can be an electronic device.

[0049] Specifically, the advisory processing method includes:

[0050] S102: Obtain the risk control consulting information input by the user to the risk control explanation intelligent agent service for the transaction risk control system, wherein the risk control consulting information includes consulting information for the system risk control transaction state;

[0051] Risk control explanation, i.e., risk control decision attribution process: the transaction risk control decision result is affected by many aspects of information, including traffic distribution, feature data, strategy model, engineering link, etc. Generally, when the transaction risk control system makes a risk decision transaction for a certain transaction matter and presents a system risk control transaction state such as risk control abnormal state, the user may need to quickly locate the root cause of the system risk control transaction state and find the root cause of the transaction decision. The process of finding the root cause of the transaction decision is also called risk control decision attribution.

[0052] The system risk control transaction state is set based on the actual transaction application scenario for the transaction risk control system;

[0053] In the present specification, the risk control explanation intelligent agent service (Agent service) is a risk control explanation computer program with perception, reasoning, and decision relying on an electronic device. The risk control explanation intelligent agent service is a service carrying object of the intelligent agent Agent. Based on the risk control explanation large language model in the present specification, the risk control explanation intelligent agent service provides an autonomous explanation processing function for the risk control decision of the transaction risk control system for the user's risk control consulting information for the transaction risk control system;

[0054] Optionally, the risk control explanation intelligent agent service can be a service control with a risk control explanation function that carries the intelligent agent service. The user can trigger the service control corresponding to the risk control explanation intelligent agent service, and input the risk control consulting information for the transaction risk control system in the intelligent agent service interface corresponding to the service control;

[0055] Non-system developers such as customer side, customer service side, user side, regulatory side, etc. can be users involved in one or more embodiments of the present specification. In practical applications, there is a risk control consulting demand for the transaction risk control system. Users may need to attribute risk control exceptions to the system risk control transaction state of the transaction risk control system for a certain transaction matter. Risk control exception attribution: the transaction risk control decision result is affected by many aspects of information, including traffic distribution, feature data, strategy model, engineering link, etc. Generally, when the risk decision made by the transaction risk control system for a certain transaction matter is in a risk control exception state, the user may need to quickly locate the root cause of the risk control exception problem to find the root cause of the decision exception. The process of finding the root cause of the decision exception is also called exception attribution.

[0056] At this time, the user may need to attribute risk control exceptions to platform service transactions maintained based on the transaction risk control system. Based on this, the consulting processing method involved in the present specification can be used. The user inputs risk control consulting information for the transaction risk control system to the risk control explanation intelligent agent service through the device held by the user. At this time, the electronic device can obtain the risk control consulting information.

[0057] For example, taking a consumer finance transaction as an example, a consumer credit transaction applied by a certain user is in an abnormal risk control state, and the transaction risk control system rejects the credit processing for the consumer credit transaction. At this time, the risk control consulting information for the abnormal risk control state can be output to the risk control explanation intelligent agent service as “help me query why the consumer credit transaction of a certain user is rejected”;

[0058] For example, taking a consumer finance transaction as an example, the credit rate of the platform's large-cap consumer credit transaction is lower than the usual travel credit rate in a certain time period. At this time, the risk control consulting information for the credit rate reduction state of the large-cap consumer credit transaction can be output to the risk control explanation intelligent agent service as “help me query why the credit rate of the large-cap consumer credit transaction in a certain time period is reduced”;

[0059] Optionally, the risk control consulting information can refer to the user's inquiry or question about the risk control state of a certain transaction in the transaction risk control system. The risk control consulting information includes but is not limited to the following aspects:

[0060] 1. Transaction identification information

[0061] The user may specify the identification of a specific transaction in the risk control consulting information, such as transaction number, user ID or other related unique identifiers, so that the intelligent agent service can accurately locate the specific transaction of interest to the user through the risk control explanation large language model.

[0062] 2. Inquiry information about transaction state

[0063] The user may have specific questions about the current risk control status of the transaction, such as asking why a certain transaction is marked as high risk or why a certain account is restricted. Such inquiries require the transaction risk control system to provide an explanation of the current risk control status, which may include the reason for the risk label, the measures taken, and their consequences.

[0064] 3. Decision details of the risk control decision

[0065] The user may be interested in the basis for the risk control decision made by the system, such as the risk assessment based on what kind of data analysis or historical information. At this time, the user usually expects the system to provide a detailed explanation of the various decision factors considered in the decision-making process.

[0066] 4. Motivation and background of the consultation

[0067] The user's motivation for consultation is also important information, such as they may want to understand how to optimize operations to avoid future risk control problems, or have objections to the risk control results and want to appeal. Understanding the user's motivation and background can help the agent better customize the response strategy through the risk control explanation large language model.

[0068] S104: Control the risk control explanation large language model based on the risk control consultation information to determine the target risk attribution tool information, and obtain the risk control decision explanation information for the transaction risk control system based on the target risk attribution tool information;

[0069] For example, the risk control consultation information of the user is received through the risk control explanation large language model, and the most suitable target risk attribution tool for the transaction risk control system is determined through semantic analysis and intent recognition of the risk control consultation information. The target risk attribution tool information corresponding to the target risk attribution tool is generated, so as to subsequently obtain the risk control decision explanation information for the transaction risk control system based on the target risk attribution tool information by using the target risk attribution tool. Specific tools include case attribution tools, fluctuation attribution tools, etc.

[0070] In practical applications, a number of types of risk attribution tools are pre-configured for the transaction risk control system. The risk attribution tool plays a crucial role in the risk control system, and its main function is to analyze and explain specific risk control events or decisions after the transaction risk control system goes online, helping users understand the root cause of the risk and the impact and effect of the relevant decisions. The application of risk attribution tools not only improves the transparency and efficiency of risk control, but also helps management develop more scientific and reasonable risk control strategies.

[0071] The risk attribution tool can include one or more combinations of case attribution tools, fluctuation attribution tools, risk interpretation tools, and user profile tools based on the type of actual system risk control transaction status;

[0072] The target risk control attribution tool information at least includes the currently required risk control attribution tool, and in some embodiments, also includes the tool required parameter query information (i.e., the target risk control transaction query parameter) determined from the risk control consulting information;

[0073] For example, the case attribution tool, the tool function determines the risk control rejection reason of a single user in a specific scenario according to the user ID or trace ID, and the tool application scenario: can be used for individual user risk control analysis and management.

[0074] For example, the fluctuation attribution tool, the tool function specifies a time range, analyzes the root cause of risk fluctuation in the period, and the tool application scenario: helps to understand the change of risk in a specific time, and performs time series analysis.

[0075] For example, the risk interpretation tool, the tool function provides a textual interpretation of the user's risk situation through the user ID, and the tool application scenario: suitable for manual review by the risk control department, and quickly obtains detailed instructions of the user's risk situation.

[0076] For example, the user profile tool, the tool function queries the historical risk event record of the user through the user ID, and the tool application scenario: used to collect and analyze the long-term risk performance of the user, and facilitate the establishment of a risk management profile.

[0077] Optionally, the risk control attribution tools configured by the transaction risk control system are usually designed to be complex, and usually require risk control analysts or managers with professional knowledge. Due to the high complexity of the tool technology and the strong professional nature of the tool programming, ordinary users cannot directly handle it. In this specification, the large language model and the intelligent agent service are creatively combined as intermediaries to simplify the operation process of the user. The user only needs to input the regular consulting information, and the intelligent agent service can provide a simple and intuitive user interface. The user only needs to input the risk control consulting information in the user interface corresponding to the risk control interpretation intelligent agent service. The intelligent agent service first receives the risk control consulting information input by the user, then performs semantic analysis and intent recognition on the risk control consulting information through the large language model to obtain an analysis result, and determines the most suitable risk control attribution tool according to the analysis result. For example, if the user queries the risk control reason of a transaction, the case attribution tool can be called.

[0078] Subsequently, the risk control interpretation large language model or the risk control interpretation intelligent agent service directly operates the target risk control attribution tool based on the target risk control attribution tool information, and uses the target risk control attribution tool to process complex data analysis and model calling. The user does not need to directly contact these complex operations, and can obtain the risk control decision interpretation information of the transaction risk control system. This step is fully automated, ensuring the accuracy and efficiency of the analysis. It can make the risk control management more friendly and accessible to non-professional users.

[0079] Exemplarily, the following approach can be adopted:

[0080] A2: input the risk control consulting information into the risk control interpretation large language model;

[0081] A4: control the risk control interpretation large language model to determine the consulting semantic intent of the risk control consulting information, and based on the consulting semantic intent, function call tool identification is performed in the large model function call tool set to obtain target risk control attribution tool information.

[0082] Exemplarily, the risk control interpretation large language model performs semantic intent recognition based on the risk control consulting information to obtain the consulting semantic intent, for example, by integrating a large language model to process the consulting information input by the user. The model analyzes the text, identifies keywords and phrases such as "high risk", "transaction", etc., to determine the core content and intent of the user's consultation, and then according to the analysis result, the model understands the user's query intent, judges the user's consulting semantic intent, for example, the consulting semantic intent can be a specific transaction interpretation, risk control policy interpretation, or a detailed explanation of the risk control mark, etc. Then, based on the consulting semantic intent, function call tool identification is performed in the large model function call tool set to obtain the target risk control attribution tool to generate target risk control attribution tool information.

[0083] Exemplarily, the risk control interpretation large language model is associated with a function call tool set, which is also called function_call. In the model training stage of the risk control interpretation large language model, the corpus and parameters of a plurality of risk control attribution tools are registered in the function_call of the large model in advance. In the actual application stage, once the target risk control attribution tool information is determined, the tool is automatically called, and relevant data and parameters in the risk control attribution tool information are input, such as user ID, transaction ID, time range, etc., to obtain the data analysis result of the transaction risk control system, and based on the data analysis result, risk control decision explanation information is generated.

[0084] Risk control decision explanation information: the risk control attribution tool usually queries professional data analysis results, and based on the data analysis results, result analysis is performed to generate detailed explanation about the risk control decision to obtain risk control decision explanation information, which can include the basis of the decision, related risk factors and recommended countermeasures.

[0085] S106: based on the risk control interpretation intelligent agent service, output the risk control decision explanation information to the user.

[0086] For example, the risk control explanation big language model can transmit risk control decision explanation information to the risk control explanation intelligent agent service, which then outputs the risk control decision explanation information to the user on the user interface. The risk control decision explanation information is displayed in a clear and easy-to-understand format in the user interface, ensuring that the user can understand the reasons and background of the risk control decision.

[0087] For example, please refer to Figure 3 , Figure 3 This is a schematic diagram of a consultation processing interface. Because a user applied for a consumption credit service under a specific scenario, the transaction risk control system rejected the credit application. At this point, the user explains the abnormal risk control status to the risk control system regarding the intelligent agent service (e.g., ...). Figure 3 The risk control attribution robot shown can be input with, for example, consultation information, such as... Figure 3 The message "Help me find out why my Huabei credit (a type of consumer credit matter) for user ID 208830228686252981 was rejected" is explained by executing the consultation and explanation method described in this manual. Based on risk control consultation information, the risk control explanation language model determines the target risk control attribution tool information as case-specific attribution tool information. Based on this case-specific attribution tool information, the case-specific attribution tool is used to obtain risk control decision explanation information for the transaction risk control system. Then, the risk control explanation intelligent agent service outputs the following to the user on the user interface: Figure 3 The dialog box shown contains explanations of risk control decisions.

[0088] For example, please refer to Figure 4 , Figure 4 This is a diagram illustrating another consultation processing interface. It shows that the credit approval rate for overall consumer credit matters on the platform has decreased compared to the usual travel credit approval rate over a certain period. In this case, the user explains the decreased credit approval rate for this overall consumer credit matter to the risk control team using the intelligent agent service (e.g., ...). Figure 4 The risk control attribution robot shown can be input with, for example, consultation information, such as... Figure 4 The message "Please help me check why the approval rate for Huabei credit approval (a type of consumer credit matter) has decreased in the last 10 minutes" is generated by executing the consultation and explanation method described in this manual. Based on the risk control consultation information, the risk control explanation language model determines the target risk control attribution tool information as fluctuation attribution tool information. Based on the fluctuation attribution tool information, the fluctuation attribution tool is used to obtain risk control decision explanation information for the transaction risk control system. Then, the risk control explanation intelligent agent service outputs this information to the user on the user interface, such as... Figure 4 The dialog box shown contains explanations of risk control decisions.

[0089] In one or more embodiments of the present specification, the electronic device obtains risk control consulting information for a transaction risk control system input by a user to a risk control explanation intelligent agent service, controls a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consulting information, and obtains risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information. The risk control decision explanation information can be output to the user based on the risk control explanation intelligent agent service. By using the method of combining the risk control explanation intelligent agent service and the risk control explanation large language model, the automation and accuracy of risk control management are significantly improved, so that the user can obtain intuitive and personalized risk control explanation through simple natural language input. In addition, the dependence on professional risk control attribution tool knowledge is reduced, the user experience is improved, the response speed of risk control decision is accelerated, and fast and accurate risk control evaluation and management are effectively supported.

[0090] Please refer to Figure 5 , Figure 5 is a flowchart of another embodiment of a consulting processing method proposed in one or more embodiments of the present specification. Specifically:

[0091] S202: input the risk control consulting information into the risk control explanation large language model;

[0092] S204: control the risk control explanation large language model to determine the consulting semantic intent of the risk control consulting information;

[0093] For example, after controlling the risk control explanation large language model to determine the consulting semantic intent of the risk control consulting information, the subsequent function calling tool recognition in the large model function calling tool set based on the consulting semantic intent obtains target risk control attribution tool information, which includes target risk control attribution tool and target risk control transaction query parameters.

[0094] S206: determine the target risk control attribution tool matching the consulting semantic intent from the large model function calling tool set, and obtain the target tool query parameter type corresponding to the target risk control attribution tool;

[0095] For example, the risk control explanation large language model identifies the semantic intent of the consultation based on the risk control consultation information, such as processing the consultation information input by the user through the integrated large language model. The model analyzes the text, identifies keywords and phrases such as "high risk", "transaction", etc., to determine the core content and intent of the user's consultation, and then according to the analysis result, the model understands the user's query intent, judges the semantic intent of the user's consultation, such as whether the semantic intent of the consultation is a specific transaction explanation, a risk control policy interpretation, or a detailed explanation of the risk control label, etc. Then, based on the semantic intent of the consultation, the target risk control attribution tool is identified from the function call tool set, and the tool query parameters required by the target risk control attribution tool are also obtained, that is, the target tool query parameter type corresponding to the target risk control attribution tool (pre-configured during tool development) is determined.

[0096] S208: Extract the target tool query parameter type corresponding to the target tool query parameter type based on the risk control consultation information.

[0097] For example, the target tool query parameter type is a user ID type, a transaction ID type, a time range type, etc. Then, based on the risk control consultation information, the user ID, transaction ID, and time range are extracted to generate the target risk control transaction query parameter.

[0098] The risk control explanation large language model generates target risk control attribution tool information based on the target risk control transaction query parameter and the target risk control attribution tool. Subsequently, based on the target risk control attribution tool information, the target risk control attribution tool and the target risk control transaction query parameter can be determined.

[0099] Then, based on the target risk control attribution tool information, the risk control decision explanation information for the transaction risk control system is obtained.

[0100] S210: Based on the target risk control transaction query parameter, control the risk control explanation intelligent agent to call the target risk control attribution tool to perform risk control decision information query processing, and obtain the target system risk control decision information.

[0101] Illustratively, the risk control explanation large language model is associated with function_call. During the model training stage of the risk control explanation large language model, the corpus and parameters of a plurality of risk control attribution tools are registered in the function_call of the large model. In the actual application stage, once the target risk control attribution tool information is determined, the target risk control attribution tool is automatically called, and the target risk control transaction query parameter in the risk control attribution tool information, such as user ID, transaction ID, time range, etc., is input into the target risk control attribution tool to obtain the data analysis result of the transaction risk control system. Subsequently, based on the data analysis result, the risk control decision explanation information is generated.

[0102] S212: generating risk control decision explanation information for the transaction risk control system based on the target system risk control decision information.

[0103] Exemplarily, the target system risk control decision information can be directly used as the risk control decision explanation information for the transaction risk control system.

[0104] Exemplarily, considering that the risk control attribution tool usually queries professional data analysis results, the data analysis results are analyzed to generate detailed explanations about the risk control decision, and the risk control decision explanation information is obtained, which can include the basis of the decision, relevant risk factors and recommended countermeasures.

[0105] In one or more embodiments of the present specification, the electronic device obtains risk control consultation information for the transaction risk control system input by the user to the risk control explanation intelligent agent service, controls the risk control explanation large language model based on the risk control consultation information to determine target risk control attribution tool information, and obtains risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information. The risk control decision explanation information can be output to the user based on the risk control explanation intelligent agent service. By using the method of combining the risk control explanation intelligent agent service and the risk control explanation large language model, the automation and accuracy of risk control management are significantly improved, so that the user can obtain intuitive and personalized risk control explanations through simple natural language input. In addition, the dependence on professional risk control attribution tool knowledge is reduced, the user experience is improved, the response speed of risk control decision is accelerated, and fast and accurate risk control evaluation and management are effectively supported.

[0106] Please refer to Figure 6 , Figure 6 is a flowchart of information generation. In a feasible implementation manner, the risk control decision explanation information for the transaction risk control system based on the target system risk control decision information can be generated in the following manner:

[0107] S3002: determining a target risk control rule item based on the target system risk control decision information.

[0108] By analyzing the target system risk control decision information, the target risk control rule item that specifically affects the decision is identified, such as a credit score threshold, a transaction behavior pattern matching, etc.

[0109] Exemplarily, the key rules in the transaction risk control system that currently affect the specific risk control decision are identified and determined from the target system risk control decision information. These key rules, i.e., the target risk control rule item, are the direct cause of the specific system risk control transaction state (such as transaction rejection, user risk rating adjustment, etc.).

[0110] S3004: Perform system rule running evaluation on the transaction risk control system based on the target risk control rule items to obtain system rule running evaluation information, and determine transaction optimization suggestion information for the target risk control rule items;

[0111] For example, after determining the key target risk control rule items, system rule running evaluation is performed on the effect of these target risk control rule items in actual operation to determine whether the risk control decision system is false positive. System rule running evaluation includes analyzing performance indicators such as hit rate, false positive rate, and false negative rate of the rules to evaluate their effectiveness and accuracy, thereby obtaining system rule running evaluation information.

[0112] For example, based on the system rule running evaluation information of rule running, optimization measures are determined for the target risk control rule items, and all optimization measures generate transaction optimization suggestion information. For example, if the false positive rate of a rule is high, it may be suggested to adjust its parameters; or update the rule logic according to the latest risk data to improve the overall performance and responsiveness of the system.

[0113] S3006: Generate risk control decision explanation information for the transaction risk control system based on the target system risk control decision information, the system rule running evaluation information, and the transaction optimization suggestion information.

[0114] Based on the target system risk control decision information, the system rule running evaluation information, and the transaction optimization suggestion information, comprehensive information processing is performed: the risk control explanation large language model combines the risk control decision information, the system rule running evaluation result, and the optimization suggestion to generate comprehensive risk control decision explanation information, thereby providing a clear and transparent risk control decision background view for users or decision makers.

[0115] Output content: The generated risk control decision explanation information includes detailed explanations of the role, evaluation results, and suggested optimization measures of each risk control rule. In addition, it also explains how these rules work together to make specific risk control decisions, and how optimization suggestions can improve system performance and decision quality.

[0116] In one or more embodiments of the present specification, the risk control decision explanation information generated in the above manner can significantly improve the transparency and reliability of the risk control system, making the risk control decision process more reasonable and scientific. At the same time, through continuous rule evaluation and optimization, the risk control system can keep pace with market environment and enterprise strategy, effectively responding to various risk challenges. This method not only helps internal management, but also enhances user trust.

[0117] See Figure 7A flowchart of a risk control explanation large language model training method is provided for one or more embodiments of the present specification. The method can rely on a computer program and can run on a Von Neumann architecture-based consultation processing device. The computer program can be integrated into an application or run as a standalone tool application. The consultation processing device can be an electronic device.

[0118] Specifically, the consultation processing method includes:

[0119] S4002: Create an initial risk control explanation large language model under the risk control explanation scenario using a basic large language model;

[0120] In a feasible implementation, the risk control explanation large language model can be obtained by training a basic large language model for the risk control explanation scenario. The basic large language model (Large Language Model, LLM) is an artificial intelligence content generation model designed to understand and generate human language. The basic large language model is trained on a large amount of data and can perform a wide range of tasks, including text summarization, translation, sentiment analysis, etc.

[0121] Optionally, the basic large language model can be a general-purpose large language model, a hundred-ling large language model, a GPT system large language model, etc.

[0122] In some embodiments, a trained basic large language model can be obtained, and the basic large language model can be adapted to the risk control explanation scenario to obtain a risk control explanation large language model. Generally, the basic large language model is directly applied to the risk control explanation scenario for dialogue response, which is difficult to adapt to new risk control explanation scenarios. Therefore, an initial risk control explanation large language model is first obtained based on the basic large language model, and sample data under the new risk control explanation scenario is obtained. The sample data is sample risk control consultation information. Since the basic large language model is usually a trained AIGC model with content generation capability, only the adaptation of the risk control explanation scenario is required in the present specification. Specifically, the sample data can be used to fine-tune the initial risk control explanation large language model. After the fine-tuning training is completed, the risk control explanation large language model adapted to the risk control explanation scenario is obtained.

[0123] In a feasible implementation, since the risk control decisions of the transaction risk control system need to be explained, the function call tool set, i.e., function_call, is configured in the model training phase of the risk control explanation large language model. The corpus and parameters of several risk control attribution tools are registered in the function_call of the large model. Specifically, the following methods can be used:

[0124] B2: Obtain a risk control attribution query tool for a transaction risk control system in a risk control explanation scenario and a tool query parameter type;

[0125] The risk control attribution tool can include one or more combinations of a case attribution tool, a fluctuation attribution tool, a risk interpretation tool, and a user profile tool, according to different types of actual system risk control transaction states.

[0126] The tool query parameter type is a parameter type required for using the risk control attribution tool, which is determined and configured in the tool development stage.

[0127] B4: Create an initial risk control explanation large language model in a risk control explanation scenario using a basic large language model, and configure a large model function call tool set of the initial risk control explanation large language model based on the risk control attribution query tool and the tool query parameter type.

[0128] For example, model creation: obtain a basic large language model, create a risk control explanation scenario adaptation module in a risk control explanation scenario, and create a basic large language model network using a basic large language model, combine the initial risk control explanation large language model based on the risk control explanation scenario adaptation module and the basic large language model network; at the same time, configure the large model function call tool set of the initial risk control explanation large language model in function_call based on the risk control attribution query tool and the tool query parameter type.

[0129] Optionally, the risk control explanation scenario adaptation module can be created based on a machine learning model.

[0130] It should be noted that the machine learning model involved in one or more embodiments of the present specification includes but is not limited to one or more of the following machine learning models: convolutional neural network (CNN) model, deep neural network (DNN) model, recurrent neural network (RNN), embedding model, gradient boosting decision tree (GBDT) model, logistic regression (LR) model, etc.

[0131] S4004: Obtain sample risk control consultation information for a transaction risk control system in the risk control explanation scenario;

[0132] Sample data acquisition: Obtain sample risk control consultation information in a new risk control explanation scenario. The sample risk control consultation information can be historical risk control consultation information of the platform.

[0133] Optionally, sample data labeling: based on the risk control explanation scene risk control explanation demand calling expert end service labeling corresponding risk control decision explanation information label.

[0134] S4006: training the initial risk control explanation large language model using the sample risk control consultation information, determining sample risk control attribution tool information through the initial risk control explanation large language model during model training, obtaining predicted risk control decision explanation information for the transaction risk control system based on the sample risk control attribution tool information, adjusting the model parameters of the initial risk control explanation large language model based on the predicted risk control decision explanation information, until the initial risk control explanation large language model ends model training, obtaining a risk control explanation large language model.

[0135] In a feasible implementation, the determination of sample risk control attribution tool information through the initial risk control explanation large language model during model training, and the obtaining of predicted risk control decision explanation information for the transaction risk control system based on the sample risk control attribution tool information can be implemented in the following manner:

[0136] C2: during model training, controlling the initial risk control explanation large language model to determine the sample consultation semantic intent of the sample risk control consultation information, and performing function call tool identification based on the sample consultation semantic intent in the large model function call tool set to obtain predicted risk control attribution tool information;

[0137] C4: determining predicted risk control attribution tools and predicted risk control transaction query parameters based on the predicted risk control attribution tool information

[0138] C6: based on the predicted risk control transaction query parameters, controlling the risk control explanation intelligent agent service to call the predicted risk control attribution tools to perform risk control decision information query processing, and obtaining predicted system risk control decision information;

[0139] C8: generating predicted risk control decision explanation information for the transaction risk control system based on the predicted system risk control decision information.

[0140] In S4006, the model parameter adjustment of the initial risk control explanation large language model based on the predicted risk control decision explanation information can be implemented in the following manner:

[0141] determining a target model loss based on the predicted risk control decision explanation information, adjusting the model parameters of the risk control explanation scene adaptation module using the target model loss, and controlling the model parameters of the basic large language model network to remain unchanged.

[0142] An exemplary model training process: input sample risk control consulting information into the initial risk control explanation large language model for at least one round of model training to obtain predicted risk control decision explanation information, determine a target model loss based on the predicted risk control decision explanation information and the risk control decision explanation information label using a model loss function, adjust the model parameters of the risk control explanation scene adaptation module using the target model loss, and control the model parameters of the basic large language model network to remain unchanged until a model training end condition is met to obtain the risk control explanation scene adaptation module and the basic large language model network, complete model fusion of the basic large language model network and the risk control explanation scene adaptation module, and obtain the trained risk control explanation large language model.

[0143] Optionally, the model end training condition of the model can include, for example, that the value of the loss function is less than or equal to a preset loss function threshold, the number of iterations reaches a preset number threshold, etc. The specific model end training condition can be determined based on actual conditions, which is not limited here.

[0144] Model fusion of the basic large language model network and the risk control explanation scene adaptation module: the weight of each model structure layer of the risk control explanation scene adaptation module is fused with the basic large language model network, the corresponding target model structure layer of the model structure layer weight in the basic large language model network is determined, the model structure layer parameters of the target model structure layer are fused with the model structure layer weight of the risk control explanation scene adaptation module, and the model structure layer weight of the risk control explanation scene adaptation module can only exist in part of the corresponding model structure layer weight among all the model structure layers of the basic large language model network. The parameter update of the model structure layer based on the model structure layer weight is completed for these target model structure layers, and the parameter update process of all model structure layer weights is completed in this way, thereby obtaining the risk control explanation large language model.

[0145] In one or more embodiments of the present specification, the risk control explanation large language model is trained in the above-mentioned manner, and based on the risk control explanation large language model, the transparency and reliability of the transaction risk control system can be significantly improved, and the risk control decision process is more reasonable and scientific. At the same time, through continuous rule evaluation and optimization, it can be ensured that the risk control system keeps pace with the market environment and enterprise strategy, effectively coping with various risk challenges. This method not only helps internal management, but also enhances user trust.

[0146] The following will be combined Figure 8 The consulting processing device provided in the present specification will be described in detail. It should be noted that Figure 8 The consulting processing device shown in the present specification is used to execute the method of the present specification Figures 1 to 7 The method of the embodiment shown in the present specification is only shown with parts related to the present specification for the convenience of description, and the specific technical details not disclosed are referred to the present specification Figures 1 to 7The illustrated embodiment.

[0147] See Figure 8 which shows a structural schematic diagram of the consultation processing device of the present specification. The consultation processing device 1 can be realized by software, hardware or a combination of the two to become all or part of the device. According to some embodiments, the consultation processing device 1 comprises an acquisition module 11, a determination module 12 and an output module 13, which are specifically used for:

[0148] The acquisition module 11 is configured to acquire the risk control consultation information for the transaction risk control system input by the user to the risk control explanation intelligent agent service, wherein the risk control consultation information comprises consultation information for the system risk control transaction state;

[0149] The determination module 12 is configured to control the risk control explanation large language model to determine target risk control attribution tool information based on the risk control consultation information, and acquire risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information;

[0150] The output module 13 is configured to output the risk control decision explanation information to the user based on the risk control explanation intelligent agent service.

[0151] Optionally, the determination module 12 is configured to:

[0152] input the risk control consultation information into the risk control explanation large language model;

[0153] control the risk control explanation large language model to determine the consultation semantic intent of the risk control consultation information, and perform function call tool identification based on the consultation semantic intent in the large model function call tool set to obtain target risk control attribution tool information.

[0154] Optionally, the target risk control attribution tool information comprises a target risk control attribution tool and a target risk control transaction query parameter, and the determination module 12 is configured to:

[0155] determine the target risk control attribution tool matching the consultation semantic intent from the large model function call tool set, and acquire the target tool query parameter type corresponding to the target risk control attribution tool;

[0156] extract the target risk control transaction query parameter corresponding to the target tool query parameter type based on the risk control consultation information.

[0157] Optionally, the determination module 12 is configured to:

[0158] determine the target risk control attribution tool and the target risk control transaction query parameter based on the target risk control attribution tool information;

[0159] The target risk control transaction query parameter based risk control explanation intelligent agent service invokes the target risk control attribution tool to perform risk control decision information query processing, and obtains target system risk control decision information;

[0160] The target system risk control decision information is used to generate risk control decision explanation information for the transaction risk control system.

[0161] Optionally, the determination module 12 is configured to:

[0162] The target system risk control decision information is used to determine a target risk control rule item;

[0163] The target risk control rule item is used to perform system rule running evaluation on the transaction risk control system to obtain system rule running evaluation information, and determine transaction optimization suggestion information for the target risk control rule item;

[0164] The target system risk control decision information, the system rule running evaluation information, and the transaction optimization suggestion information are used to generate risk control decision explanation information for the transaction risk control system.

[0165] It should be noted that the consulting processing device provided in the above embodiments is only used as an example to illustrate the division of the above functional modules in the execution of the consulting processing method. In actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the consulting processing device and the consulting processing method provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments. Here, it is not repeated.

[0166] The serial number in the above description is only for description, not representing the pros and cons of the embodiments.

[0167] In one or more embodiments of the present specification, the electronic device obtains risk control consulting information for a transaction risk control system input by a user to a risk control explanation intelligent agent service, controls a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consulting information, and obtains risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information. The risk control decision explanation information can be output to the user based on the risk control explanation intelligent agent service. By using the method of combining the risk control explanation intelligent agent service and the risk control explanation large language model, the automation and accuracy of risk control management are significantly improved, so that the user can obtain intuitive and personalized risk control explanation through simple natural language input. In addition to reducing the dependence on professional risk control attribution tool knowledge and improving user experience, the response speed of risk control decision is also accelerated, effectively supporting fast and accurate risk control evaluation and management.

[0168] Please refer toFigure 9 FIG. 1 shows a structural schematic diagram of a risk control explanation large language model training apparatus according to the present specification. The risk control explanation large language model training apparatus 2 can be realized by software, hardware, or a combination of both to become all or part of the apparatus. According to some embodiments, the risk control explanation large language model training apparatus 2 includes a model creation module 11, an information acquisition module 12, and a model training module 13, and is specifically used for:

[0169] The model creation module 11 is configured to create an initial risk control explanation large language model in a risk control explanation scenario by using a basic large language model.

[0170] The information acquisition module 12 is configured to acquire sample risk control consulting information for a transaction risk control system in the risk control explanation scenario.

[0171] The model training module 13 is configured to perform model training on the initial risk control explanation large language model by using the sample risk control consulting information, determine sample risk control attribution tool information by the initial risk control explanation large language model during the model training, acquire predicted risk control decision explanation information for the transaction risk control system based on the sample risk control attribution tool information, perform model parameter adjustment on the initial risk control explanation large language model based on the predicted risk control decision explanation information, and end the model training of the initial risk control explanation large language model until a risk control explanation large language model is obtained.

[0172] Optionally, the model creation module 11 is configured to:

[0173] acquire a risk control attribution query tool and a tool query parameter type for a transaction risk control system in the risk control explanation scenario;

[0174] create an initial risk control explanation large language model in the risk control explanation scenario by using a basic large language model, and configure a large model function call tool set of the initial risk control explanation large language model based on the risk control attribution query tool and the tool query parameter type.

[0175] Optionally, the model training module 13 is configured to:

[0176] during the model training, control the initial risk control explanation large language model to determine a sample consulting semantic intent of the sample risk control consulting information, and perform function call tool identification in a large model function call tool set based on the sample consulting semantic intent to obtain predicted risk control attribution tool information;

[0177] determine a predicted risk control attribution tool and a predicted risk control transaction query parameter based on the predicted risk control attribution tool information;

[0178] Based on the predicted risk control transaction query parameter, the risk control explanation intelligent agent service calls the predicted risk control attribution tool for risk control decision information query processing, and obtains predicted system risk control decision information.

[0179] Based on the predicted system risk control decision information, predicted risk control decision explanation information for the transaction risk control system is generated.

[0180] Optionally, the model training module 13 is configured to:

[0181] A risk control explanation scene adaptation module in a risk control explanation scene is created, and a basic large language model network is created using a basic large language model, and an initial risk control explanation large language model is obtained based on the risk control explanation scene adaptation module and the basic large language model network.

[0182] The model parameter adjustment of the initial risk control explanation large language model based on the predicted risk control decision explanation information includes:

[0183] Based on the predicted risk control decision explanation information, a target model loss is determined, the model parameter adjustment of the risk control explanation scene adaptation module is performed using the target model loss, and the model parameters of the basic large language model network remain unchanged.

[0184] It should be noted that the risk control explanation large language model training device provided in the above embodiments is used to execute the risk control explanation large language model training method, and only the division of the above functional modules is used as an example for illustration. In actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the risk control explanation large language model training device and the risk control explanation large language model training method provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments. Therefore, it is not repeated here.

[0185] The serial numbers in the above description are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0186] In one or more embodiments of the present specification, the trained risk control explanation large language model can be obtained in the above manner. The electronic device obtains the risk control consultation information for the transaction risk control system input by the user to the risk control explanation intelligent agent service, controls the risk control explanation large language model to determine the target risk control attribution tool information based on the risk control consultation information, and obtains the risk control decision explanation information for the transaction risk control system based on the target risk control attribution tool information. The risk control decision explanation information can be output to the user based on the risk control explanation intelligent agent service. By using the method of combining the risk control explanation intelligent agent service and the risk control explanation large language model, the automation and accuracy of risk control management are significantly improved, so that the user can obtain intuitive and personalized risk control explanation through simple natural language input. In addition, the dependence on professional risk control attribution tool knowledge is reduced, the user experience is improved, the response speed of risk control decision is accelerated, and fast and accurate risk control evaluation and management are effectively supported.

[0187] The present specification also provides a computer storage medium, which can store a plurality of instructions. The instructions are suitable for being loaded and executed by a processor to perform the consultation processing method of the above-mentioned Figures 1 to 7 embodiments. For specific implementation process, please refer to the specific description of the above-mentioned Figures 1 to 7 embodiments, which will not be repeated here.

[0188] The present specification also provides a computer program product, which stores at least one instruction. The at least one instruction is loaded and executed by the processor to perform the consultation processing method of the above-mentioned Figures 1 to 7 embodiments. For specific implementation process, please refer to the specific description of the above-mentioned Figures 1 to 7 embodiments, which will not be repeated here.

[0189] Please refer to Figure 10 for a structural block diagram of an electronic device provided by the embodiments of the present specification. The electronic device in the present specification can include one or more of the following components: a processor 1010, a memory 1020, an input device 1030, an output device 1040 and a bus 1050. The processor 1010, the memory 1020, the input device 1030 and the output device 1040 can be connected through the bus 1050.

[0190] The processor 1010 can include one or more processing cores. The processor 1010 connects various parts within the terminal through various interfaces and lines, performs various functions of the terminal 100 and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 1020, and calling data stored in the memory 1020. Alternatively, the processor 1010 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), a programmable logic array (PLA). The processor 1010 can integrate a combination of one or several of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes an operating system, a user interface, and an application program; the GPU is responsible for rendering and drawing display content; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 1010, but be implemented by a separate communication chip.

[0191] The memory 1020 can include a random access memory (RAM) and can also include a read-only memory (ROM). Alternatively, the memory 1020 includes a non-transitory computer-readable storage medium. The memory 1020 can be used to store instructions, programs, codes, code sets or instruction sets.

[0192] Among them, the input device 1030 is used to receive input instructions or data, and the input device 1030 includes but is not limited to a keyboard, a mouse, a camera, a microphone or a touch device. The output device 1040 is used to output instructions or data, and the output device 1040 includes but is not limited to a display device and a speaker. In the embodiment of the present specification, the input device 1030 can be a temperature sensor for obtaining the operating temperature of the terminal. The output device 1040 can be a speaker for outputting an audio signal.

[0193] In addition, those skilled in the art will understand that the structure of the terminal shown in the above figures does not constitute a limitation on the terminal. The terminal may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, the terminal may also include radio frequency circuits, input units, sensors, audio circuits, wireless fidelity (WIFI) modules, power supplies, Bluetooth modules, etc., which will not be described in detail here.

[0194] In the embodiments of this specification, the executing entity for each step can be the terminal described above. Optionally, the executing entity for each step is the terminal's operating system. The operating system can be Android, iOS, or other operating systems; this specification does not limit this.

[0195] exist Figure 10 In the electronic device, the processor 1010 can be used to call a program stored in the memory 1020 and execute it to implement the consultation processing method and / or the risk control interpretation large language model training method as described in the various method embodiments of this specification.

[0196] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.

[0197] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in the embodiments of this specification are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the risk control consulting information involved in this specification was obtained under full authorization.

[0198] The above-disclosed embodiments are merely preferred embodiments of this specification and should not be construed as limiting the scope of this specification. Therefore, any equivalent variations made in accordance with the claims of this specification shall still fall within the scope of this specification.

Claims

1. A consulting processing method, the method comprising: obtaining risk control consulting information input by a user to a risk control explanation intelligent agent service for a transaction risk control system, the risk control consulting information comprising consulting information for a system risk control transaction state; controlling a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consulting information, calling a target risk control attribution tool based on the target risk control attribution tool information to obtain target system risk control decision information, determining a target risk control rule item based on the target system risk control decision information, performing system rule running evaluation on the transaction risk control system based on the target risk control rule item to obtain system rule running evaluation information, and determining transaction optimization suggestion information for the target risk control rule item, and generating risk control decision explanation information for the transaction risk control system based on the target system risk control decision information, the system rule running evaluation information, and the transaction optimization suggestion information; outputting the risk control decision explanation information to the user based on the risk control explanation intelligent agent service.

2. The method of claim 1, wherein controlling a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consulting information comprises: inputting the risk control consulting information into a risk control explanation large language model; controlling the risk control explanation large language model to determine a consulting semantic intent of the risk control consulting information, and identifying a target risk control attribution tool information based on the consulting semantic intent in a large model function calling tool set.

3. The method of claim 2, wherein the target risk control attribution tool information comprises a target risk control attribution tool and a target risk control transaction query parameter, and wherein identifying a target risk control attribution tool information based on the consulting semantic intent in a large model function calling tool set comprises: determining a target risk control attribution tool matching the consulting semantic intent from a large model function calling tool set, and obtaining a target tool query parameter type corresponding to the target risk control attribution tool; extracting a target risk control transaction query parameter corresponding to the target tool query parameter type based on the risk control consulting information.

4. The method of claim 1, wherein calling a target risk control attribution tool based on the target risk control attribution tool information to obtain target system risk control decision information comprises: determining a target risk control attribution tool and a target risk control transaction query parameter based on the target risk control attribution tool information; controlling the risk control explanation intelligent agent service to call the target risk control attribution tool for risk control decision information query processing based on the target risk control transaction query parameter, and obtaining target system risk control decision information.

5. The method of claim 1, comprising: creating an initial risk control explanation large language model in a risk control explanation scenario using a basic large language model; obtaining sample risk control consulting information for a transaction risk control system in the risk control explanation scenario; The sample risk control consulting information is used for model training of the initial risk control explanation large language model. In the model training process, sample risk control attribution tool information is determined by the initial risk control explanation large language model, and prediction risk control decision explanation information for the transaction risk control system is obtained based on the sample risk control attribution tool information. The initial risk control explanation large language model is adjusted in model parameters based on the prediction risk control decision explanation information, until the initial risk control explanation large language model ends the model training, and a risk control explanation large language model is obtained.

6. The method of claim 5, wherein the initial risk control explanation large language model in the risk control explanation scene is created using the base large language model, comprising: obtaining a risk control attribution query tool and a tool query parameter type for the transaction risk control system in the risk control explanation scene; creating an initial risk control explanation large language model in the risk control explanation scene using the base large language model, and configuring a large model function call tool set of the initial risk control explanation large language model based on the risk control attribution query tool and the tool query parameter type.

7. The method of claim 5, wherein the sample risk control attribution tool information is determined by the initial risk control explanation large language model in the model training process, and the prediction risk control decision explanation information for the transaction risk control system is obtained based on the sample risk control attribution tool information, comprising: in the model training process, controlling the initial risk control explanation large language model to determine a sample consulting semantic intent of the sample risk control consulting information, and identifying a prediction risk control attribution tool information in the large model function call tool set based on the sample consulting semantic intent; determining a prediction risk control attribution tool and a prediction risk control transaction query parameter based on the prediction risk control attribution tool information; controlling the risk control explanation intelligent agent service to call the prediction risk control attribution tool for risk control decision information query processing based on the prediction risk control transaction query parameter, and obtaining prediction system risk control decision information; generating prediction risk control decision explanation information for the transaction risk control system based on the prediction system risk control decision information.

8. The method of claim 5, wherein the initial risk control explanation large language model in the risk control explanation scene is created using the base large language model, comprising: creating a risk control explanation scene adaptation module in the risk control explanation scene and creating a base large language model network using the base large language model, and obtaining an initial risk control explanation large language model based on the risk control explanation scene adaptation module and the base large language model network; based on the prediction risk control decision explanation information, adjusting model parameters of the initial risk control explanation large language model, comprising: determining a target model loss based on the prediction risk control decision explanation information, adjusting model parameters of the risk control explanation scene adaptation module using the target model loss, and keeping model parameters of the base large language model network unchanged.

9. A consulting processing device, comprising: An acquisition module is configured to acquire risk control consultation information input by a user to a risk control explanation intelligent agent service, the risk control consultation information including consultation information for a system risk control transaction state; A determination module is configured to control a risk control explanation large language model to determine target risk control attribution tool information based on the risk control consultation information, to call a target risk control attribution tool based on the target risk control attribution tool information to obtain target system risk control decision information, to determine a target risk control rule item based on the target system risk control decision information, to perform system rule operation evaluation on the transaction risk control system based on the target risk control rule item to obtain system rule operation evaluation information, to determine transaction optimization suggestion information for the target risk control rule item, and to generate risk control decision explanation information for the transaction risk control system based on the target system risk control decision information, the system rule operation evaluation information, and the transaction optimization suggestion information; An output module is configured to output the risk control decision explanation information to the user based on the risk control explanation intelligent agent service.

10. A computer storage medium, the computer storage medium storing a plurality of instructions, the instructions being adapted to be loaded and executed by a processor to perform the method steps of any one of claims 1-8.

11. A computer program product, the computer program product storing at least one instruction, the at least one instruction being loaded and executed by a processor to perform the method steps of any one of claims 1-8.

12. An electronic device comprising: A processor and a memory; wherein the memory stores a computer program, the computer program being adapted to be loaded and executed by the processor to perform the method steps of any one of claims 1-8.

Images