Data processing method and device of risk control system and computer equipment

Abstract

The invention discloses a data processing method and device of a risk control system and computer equipment, and the method comprises the steps: obtaining risk identification attribute data of the risk control system, the risk identification attribute data being used for representing the attribute of the risk identification capability of the risk control system; obtaining project information of risk control projects from a risk control project knowledge base, and sending the project information and the risk identification attribute data to a large language model, so that the large language model analyzes the risk identification attribute data based on the project information to obtain an operation parameter adjustment strategy of the risk control system; the project information is used for describing the risk control project, and the project information is generated by the big language model based on demand information of the risk control project provided by a risk control expert; and adjusting operation parameters of the risk control system based on the parameter adjustment strategy.

Description

Technical Field

[0001] This specification relates to the field of risk control technology, and in particular to a data processing method and apparatus for a risk control system, as well as computer equipment. Background Technology

[0002] Risk control (or simply risk management) primarily refers to the identification, assessment, and response to various risks that may affect the achievement of corporate or individual goals through a series of methods and technologies. In related technologies, risk control is generally implemented through risk control systems. With technological advancements, intelligent risk control systems based on big data and artificial intelligence are gradually becoming mainstream, enabling more efficient prediction and prevention of risks. Faced with complex and ever-changing service environments and constantly evolving risk factors, risk control systems need continuous adjustments to maintain their accuracy and effectiveness. However, in related technologies, adjustments to risk control systems are mainly implemented manually by operations and maintenance personnel. This method is inefficient and struggles to meet the requirements of real-time performance and flexibility. Summary of the Invention

[0003] Firstly, embodiments of this specification provide a data processing method for a risk control system, the method comprising:

[0004] Obtain risk identification attribute data from the risk control system, wherein the risk identification attribute data is used to characterize the risk identification capability of the risk control system;

[0005] The project information of the risk control project is obtained from the risk control project knowledge base. The project information and the risk identification attribute data are sent to the big language model so that the big language model can analyze the risk identification attribute data based on the project information to obtain the operation parameter adjustment strategy of the risk control system. The project information is used to describe the risk control project and is generated by the big language model based on the requirement information of the risk control project provided by the risk control expert.

[0006] The operating parameters of the risk control system are adjusted based on the parameter adjustment strategy.

[0007] Secondly, embodiments of this specification provide a data processing device for a risk control system, the device comprising:

[0008] The first acquisition module is used to acquire risk identification attribute data of the risk control system, wherein the risk identification attribute data is used to characterize the risk identification capability of the risk control system;

[0009] The second acquisition module is used to acquire project information of risk control projects from the risk control project knowledge base, and send the project information and the risk identification attribute data to the big language model, so that the big language model can analyze the risk identification attribute data based on the project information to obtain the operating parameter adjustment strategy of the risk control system; the project information is used to describe the risk control project, and the project information is generated by the big language model based on the requirement information of the risk control project provided by the risk control expert;

[0010] The parameter adjustment module is used to adjust the operating parameters of the risk control system based on the parameter adjustment strategy.

[0011] Thirdly, embodiments of this specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the methods described in any embodiment of this specification.

[0012] In the embodiments of this specification, the project information of the risk control project is stored in the risk control project knowledge base. After obtaining the risk identification attribute data of the risk control system, the project information of the risk control project is first obtained from the risk control project knowledge base. Then, the project information and the risk identification attribute data are sent to the big language model. Since the risk identification attribute data can characterize the risk identification capability of the risk control system, the big language model can combine the risk identification attribute data and the project information to analyze whether the attributes of the current risk identification capability of the risk control system meet expectations, and automatically generate targeted operating parameter adjustment strategies, thereby realizing automatic adjustment of the risk control system, reducing the cost of manual intervention, and enhancing the real-time performance and flexibility of the system.

[0013] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this specification and, together with the specification, serve to explain the technical solutions described herein.

[0015] Figure 1 This is a schematic diagram illustrating the application scenarios of the embodiments in this specification.

[0016] Figure 2 This is a flowchart of the data processing method of the risk control system in an embodiment of this specification.

[0017] Figure 3 This is a schematic diagram of the system architecture of an embodiment of this specification.

[0018] Figure 4This is a block diagram of the data processing device of the risk control system in an embodiment of this specification.

[0019] Figure 5 This is a schematic diagram of a computer device according to an embodiment of this specification. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this specification as detailed in the appended claims.

[0021] The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of this specification. The singular forms “a,” “the,” and “the” as used in this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items. Additionally, the term “at least one” herein means any combination of at least two of any one or more of a plurality.

[0022] It should be understood that although the terms first, second, third, etc., may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this specification, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0023] To enable those skilled in the art to better understand the technical solutions in the embodiments of this specification, and to make the above-mentioned objectives, features and advantages of the embodiments of this specification more apparent and understandable, the technical solutions in the embodiments of this specification will be further described in detail below with reference to the accompanying drawings.

[0024] Risk control primarily refers to the identification, assessment, and response to various risks that may affect the achievement of corporate or individual goals through a series of methods and technologies. Risk control can be applied to various fields, such as finance, the internet, logistics, and healthcare. In the financial sector, risk control can be used to prevent risks such as credit defaults and credit card fraud; in the internet sector, it can prevent risks such as malicious attacks and data breaches; in the logistics sector, it can address risks such as transportation delays and lost goods; and in the healthcare sector, it can reduce the probability of medical accidents and medical insurance fraud.

[0025] In related technologies, risk control is generally achieved through risk control systems. With the rapid development of technology, intelligent risk control systems based on big data and artificial intelligence are gradually becoming mainstream. For example... Figure 1 As shown, taking the credit sector as an example, the risk control process in the credit sector mainly includes the following steps: First, collecting multi-dimensional data such as the borrower's personal information, credit history, and financial status; second, using big data analytics to clean, organize, and analyze this data to uncover risk characteristics and correlations; then, using artificial intelligence algorithms, such as decision trees and neural networks in machine learning, to accurately assess the borrower's credit risk and predict the likelihood of default; finally, based on the risk assessment results, formulating corresponding risk response strategies, such as determining the loan amount, interest rate level, and whether collateral or guarantees are required.

[0026] Faced with a complex and ever-changing service environment and constantly evolving risk factors, risk control systems require continuous adjustments to maintain their accuracy and effectiveness. However, in related technologies, these adjustments are primarily implemented manually by operations and maintenance (O&M) personnel. These personnel need to periodically monitor and evaluate the system, manually analyzing the causes and making corresponding adjustments when they detect a decline in system performance or a significant deviation between risk assessment results and actual conditions. These adjustments might include updating the parameters of the risk assessment model or adjusting the scope and frequency of data collection. This manual adjustment method has several drawbacks. Firstly, it is inefficient, making it difficult to respond quickly to real-time changes in the service environment and risk factors. Secondly, it fails to meet the system's high requirements for real-time performance and flexibility, easily leading to lag in the risk control system's response to sudden risks or rapidly changing service scenarios. This inability to identify and respond to risks in a timely and effective manner can potentially cause significant losses to businesses or individuals.

[0027] Based on this, the embodiments of this specification provide a data processing method for a risk control system, see [link to documentation]. Figure 2 The method includes:

[0028] Step S12: Obtain the risk identification attribute data of the risk control system. The risk identification attribute data is used to characterize the risk identification capability of the risk control system.

[0029] Step S14: Obtain project information of risk control projects from the risk control project knowledge base, and send the project information and risk identification attribute data to the big language model so that the big language model can analyze the risk identification attribute data based on the project information to obtain the risk control system's operating parameter adjustment strategy; the project information is used to describe the risk control project, and the project information is generated by the big language model based on the risk control project requirement information provided by risk control experts;

[0030] Step S16: Adjust the operating parameters of the risk control system based on the parameter adjustment strategy.

[0031] This specification's embodiments combine a risk control project knowledge base, a large language model, and risk control experts to achieve automatic adjustment of the risk control system. Risk control experts provide requirement information for risk control projects, thereby generating project information. The risk control project knowledge base stores this project information. The large language model automatically generates targeted operational parameter adjustment strategies based on the project information and the risk identification attribute data of the risk control system. This achieves automatic adjustment of the risk control system, reduces manual intervention costs, and enhances the system's real-time performance and flexibility. The specific implementation details of this specification's embodiments are illustrated below.

[0032] Figure 3 A schematic diagram of the risk control system according to an embodiment of this specification is shown. See also Figure 2 The risk control system includes:

[0033] The risk control intelligent agent is responsible for coordinating the various functional modules of the risk control system, including but not limited to interaction with the large language model and risk control project knowledge base, dynamic decision flow control, and scheduling of various resources.

[0034] The large language model, as the engine for semantic understanding, generation, and reasoning, is responsible for functions such as information acquisition and parsing, risk reasoning and response plan generation, feature and model code generation, rule logic and code generation, and supports natural language interaction.

[0035] Risk control experts, as human oversight nodes, are responsible for reviewing key decisions, marking complex cases, and adjusting risk preference parameters, etc.

[0036] The risk control project knowledge base, serving as a dedicated knowledge repository for specific risk control projects, covers comprehensive information across all stages of a risk control project, from planning to implementation and management. This knowledge base can be jointly built by a large language model and risk control experts and is continuously updated. Risk control experts can provide requirements information for risk control projects, and the large language model can generate project information describing these projects based on this information, thus establishing the risk control project knowledge base.

[0037] The requirements information for risk control projects is a key guide provided by risk control experts based on the actual needs and risk control objectives of the project. It lays the foundation for building a risk control project knowledge base. The requirements information defines in detail the specific service indicators that the project should achieve, such as project type, risk control objectives, project scale and scope, etc.; it clarifies the data requirements, covering internal data (borrower basic information, historical lending records, etc.), external data (credit data, industry and macroeconomic data, etc.), and data update frequency; it specifies model-related requirements, including model type, performance indicator targets, and interpretability requirements; it outlines service rules and strategies, such as risk assessment rules, approval processes, and post-loan management strategies; and it also covers many compliance and management requirements, such as applicable laws and regulations, management policies, and data security and privacy protection standards.

[0038] Project information for risk control projects, generated by a large language model based on the requirements information, provides detailed content to enrich the risk control project knowledge base. This information includes, but is not limited to: project overview, industry analysis and management policy information, service process information, core risk control system information (data, models, strategies and rules, monitoring and early warning, etc.), technical implementation information, case studies and debriefings, and / or assistance and operational information. The project overview provides a comprehensive introduction to the risk control project, covering background, objectives, scope, and plans to quickly provide a complete picture. Industry analysis and management policy information focuses on the dynamics and management requirements of the industry involved in the risk control project (e.g., the credit industry). Service process information details each stage of the industry involved in the risk control project; taking the credit industry as an example, this includes, but is not limited to, operational procedures for pre-loan application, loan approval, and post-loan management. Core risk control system information covers data management, model building, the formulation of risk control processing rules, and monitoring and early warning mechanisms, serving as the core guide for risk control technology. Technical implementation information focuses on the system's architecture design, technology selection, and data processing details to ensure the project's technical feasibility. This may include model description information (such as model type and version) used to describe the risk detection models employed in the risk control system. Case studies and debriefing information analyzes specific cases to summarize lessons learned and facilitate project optimization. Assistance and operational documents cover internal communication mechanisms, personnel training, and external collaborations to ensure smooth project operation and sustainable development. These documents collectively constitute the risk control project knowledge base, providing comprehensive support for the successful implementation of risk control projects.

[0039] The aforementioned project information provides a comprehensive and systematic description of the risk control project. Regarding service processes, it details the specifics and standards of each stage, the steps and methods of risk assessment, and the specific procedures, standards, and personnel responsibilities for approval and decision-making. For data processing and analysis, it covers data collection sources, methods, and frequency; data cleaning and preprocessing methods; data analysis techniques; and data visualization formats. For model development and application, it details each step of the model development process, model evaluation methods, and mechanisms for model monitoring and updating. At the risk identification and monitoring level, it clarifies the specific methods for risk identification, risk monitoring indicators and their threshold settings, as well as the triggering and response mechanisms for risk warnings. In the compliance and reporting section, it clearly defines the compliance inspection process, the specific content and format requirements of the reporting mechanism, and the scope, frequency, and cooperation methods for audits. This comprehensive and systematic project information provides detailed knowledge support for the smooth progress of risk control projects, ensuring that each stage of the project is based on evidence and carried out in an orderly manner.

[0040] The system comprises an enterprise knowledge base and an internet data access module. The enterprise knowledge base serves as the organizational-level knowledge foundation, integrating enterprise-specific knowledge such as service system data, service process documents, and contract texts. This internal knowledge allows for the generation of project information that better aligns with the company's actual situation and personalized needs. This provides more accurate and targeted project information for risk control projects, better addressing potential risks and ensuring their smooth progress. The internet data access module acts as an external information access layer, capturing news, public opinion, external knowledge bases and data, and management announcements in real time. This information helps to grasp the macro-environment and external risk factors of risk control projects from a macro perspective. By deploying the enterprise knowledge base and internet data access module, the large language model can use RAG (Retrieval-Augmented Generation) to obtain rich information from these modules to generate project information for risk control projects. This ensures that project information meets both the company's personalized needs and keeps pace with changes in the macro environment. It helps project management teams more accurately identify risk points, develop risk response strategies in advance, and optimize project planning and resource allocation. At the same time, this project information generation method, which combines internal and external information, can also improve the scientific nature and timeliness of project decision-making, enhance the project's adaptability and competitiveness in a complex and ever-changing environment, and provide strong support for the successful implementation of the project.

[0041] The feature engineering module can dynamically generate risk features (including time-series derived features and / or correlation network features, etc.). The features generated by the feature engineering module can be used as input features for the risk detection model, so that the risk detection model can perform risk detection based on the input features. In some embodiments, the risk control project knowledge base is also used to store descriptive information of the input data of the risk control system. The feature engineering module can obtain the descriptive information of the input data sent by the risk control agent and perform feature engineering operations based on the descriptive information of the input data to obtain the input features of the risk detection model.

[0042] The model management module manages the risk detection model throughout its entire lifecycle, supporting automatic hyperparameter tuning, drift detection, and hot model replacement. In some embodiments, the project information includes model description information that describes the risk detection model. The model management module can obtain the model description information sent by the risk control agent and generate a risk detection model based on the model description information.

[0043] The rule management module, acting as a rule evolution engine, supports the dynamic generation of rules that meet risk control objectives and requirements based on statistical analysis and machine learning rule iteration. In some embodiments, the project information includes rule description information for describing the risk control processing rules. The rule management module can obtain the rule description information sent by the risk control agent and generate risk control processing rules based on the rule description information.

[0044] The data platform, as a data governance infrastructure, provides a unified data lake, integrated stream and batch processing, and privacy computing capabilities (supporting federated learning);

[0045] The model platform, as a model service platform, supports online risk detection model configuration and management, online or offline inference, and also provides model interpretability reports;

[0046] The decision engine, serving as an online risk decision center, integrates the query services of the rule engine and the model, supporting the import, updating, deployment, testing, and monitoring of knowledge base rules.

[0047] The management platform, serving as the system's health and risk awareness center, provides visualized real-time / near real-time tracking of system indicators defined in various risk control project knowledge bases, enabling risk warning and analysis.

[0048] It is understood that the above architecture is merely illustrative and not intended to limit this specification. In other examples, the risk control system may include only some of the above modules, or it may include other modules besides those mentioned above. The specific steps of the methods in the embodiments of this specification are illustrated below with reference to the above system architecture.

[0049] The steps in the embodiments of this specification can be executed by the risk control agent in the above system architecture.

[0050] In step S12, the risk control agent can acquire risk identification attribute data from the risk control system. This risk identification attribute data characterizes the risk identification capability of the risk control system. The attributes of risk identification capability may include, but are not limited to, the strength of risk identification capability and / or the type of risk identification capability. The strength of risk identification capability reflects the accuracy and effectiveness of the risk control system in identifying risks; the stronger the risk identification capability, the more accurate and effective the risk identification. The type of risk identification capability reflects the types and sources of risks that the risk control system can identify.

[0051] In some embodiments, the risk control system includes a model platform deployed with a risk detection model, which is used to perform risk detection on the input data of the risk control system. Based on this, the risk identification attribute data may include model validity data, used to characterize the effectiveness of the risk detection model adopted by the risk control system. Model validity data may include model operating parameters, such as the accuracy of the risk control detection model, KS value (Kolmogorov-Smirnov Statistic), PSI (Population Stability Index), and IV value (Information Value). These parameters collectively constitute the key performance indicators of the risk detection model. By regularly monitoring these parameters, performance degradation or failure of the risk detection model can be detected in a timely manner, thereby enabling corresponding optimization measures to ensure the effectiveness and stability of the risk control system.

[0052] In some embodiments, the risk control system includes a decision engine deployed with risk control processing rules. The decision engine processes the risk detection results of the risk detection model based on these rules. Therefore, the risk identification attribute data may include rule validity data, used to characterize the effectiveness of the risk control processing rules adopted by the risk control system. Rule validity data may include the occurrence rate of risk events. For example, in the credit field, a risk event may be "overdue repayment of the first loan," and the rule validity data may include the overdue repayment ratio of the first loan; in the e-commerce field, a risk event may be "order fraud," and the rule validity data may include the order fraud occurrence rate; in the financial transaction field, a risk event may be "abnormal transaction behavior," and the rule validity data may include the abnormal transaction behavior occurrence rate.

[0053] In step S14, the risk control agent can obtain project information of risk control projects from the risk control project knowledge base. In some embodiments, the risk control system may include a data platform, on which project information of risk control projects (such as the model validity data and rule validity data mentioned above) can be stored. Further, the risk control system may also include a management platform, which can collect project information stored in the data platform based on preset conditions (such as according to preset time intervals) and synchronize the collected project information to the risk control project knowledge base for the risk control agent to access.

[0054] After acquiring project information for risk control projects, the risk control agent can send the project information and risk identification attribute data to the large language model. The large language model can then analyze the risk identification attribute data based on the project information to derive adjustment strategies for the operating parameters of the risk control system.

[0055] Specifically, the risk control intelligent agent can analyze risk identification attribute data based on project information to obtain the operational status information of the risk control system. This operational status information can characterize whether any anomalies have occurred in the operational status of the risk control system.

[0056] In some embodiments, the operational status information includes the status information of the risk detection model. This status information characterizes whether the risk detection model's status is abnormal. If the risk detection model's status information is abnormal, the operational parameter adjustment strategy for the risk control system can be determined to include strategies for adjusting the model parameters of the risk detection model. For example, the risk detection model's status information may include key indicators such as the model's response time, prediction accuracy, recall, and resource consumption. If the risk detection model's response time exceeds a set threshold, or the prediction accuracy and recall are lower than expected standards, it can be determined that the risk detection model's status information is abnormal. In this case, the determined operational parameter adjustment strategy for the risk control system includes strategies for optimizing the model parameters of the risk detection model, such as retraining the model, adjusting the model's hyperparameters, or updating the model's algorithm structure, thereby ensuring that the risk detection model can accurately and efficiently identify potential risks, providing a reliable guarantee for the stable operation of the risk control system.

[0057] In some embodiments, the state information of the risk detection model includes the changing trend of the operating parameters of the risk detection model over a historical time period. Anomalies in the state information of the risk detection model may include a mismatch between the changing trend of the group stability index of the risk control system's features and the changing trend of the operating parameters of the risk detection model. Based on this, strategies for adjusting the model parameters of the risk detection model include: a strategy of incrementally training the risk detection model based on target data generated within a historical time period; wherein the rate of change of the group stability index of the target data's features is higher than a preset rate of change threshold.

[0058] In practical applications, risks escalate, and new risks may emerge that bypass existing risk detection models. In such cases, the operating parameters of the risk detection model may not change significantly. However, the group stability index of the features in a risk control system is usually stable under normal circumstances. When it drifts sharply, it reflects potential risks that the risk detection model may have failed to detect. This embodiment considers both the operating parameters of the risk detection model and the group stability index of the features in the risk control system to determine the adjustment strategy for the operating parameters. In this strategy, target data with high rates of change in the group stability index of the features are marked and used for incremental training of the risk detection model, thereby effectively improving the model's ability to identify new risks.

[0059] In some embodiments, the state information of the risk detection model includes the expected trend of changes in the operating parameters of the risk detection model over a future time period. These operating parameters include the detection accuracy of the risk detection model for input data of a specified category. Anomalies in the state information of the risk detection model may include a decreasing trend in the expected change of the detection accuracy of the risk detection model for the specified category. Strategies for adjusting the model parameters of the risk detection model may include adjusting the model parameters related to the specified category.

[0060] In practical applications, project information for risk control projects may be adjusted. These adjustments could cause the risk level of input data for a specific category to change over a future period. The risk detection model may not be able to respond promptly to these changes, thus, it can be expected that the accuracy of the risk detection model for the specified category of input data will decrease in the future. Therefore, the model parameters related to the specified category in the risk detection model can be adjusted to optimize model performance in a timely manner to adapt to new risk characteristics and changes. This embodiment predicts the changing trends of the operating parameters of the risk detection model over a future time period and determines the adjustment strategy based on this prediction. This achieves predictive risk assessment that surpasses current data performance, upgrading the risk detection process from passive response to proactive management.

[0061] In some embodiments, the operational status information includes the status information of risk control processing rules. This status information characterizes whether the risk control processing rules are abnormal. If the status information of a risk control processing rule is abnormal, it can be determined that the operational parameter adjustment strategy of the risk control system includes a strategy to adjust the risk control processing rules. The status information of risk control processing rules may involve multiple dimensions such as the matching frequency of rules, the conflict between rules, and the execution effect of rules. For example, if a risk control processing rule is not triggered for a long time, frequently conflicts with other rules, or its execution effect fails to achieve the expected risk control objective, then the status information of the risk control processing rule can be considered abnormal. In such cases, the determined operational parameter adjustment strategy of the risk control system will include strategies for targeted adjustments to the risk control processing rules, such as optimizing the rule's condition settings, adjusting the rule's priority, or modifying the rule's logical structure, to ensure that the risk control processing rules can accurately and efficiently respond to various risk scenarios and guarantee the overall effectiveness and stability of the risk control system.

[0062] In some embodiments, the status information of risk control processing rules includes the hit rate and accuracy of the risk control processing rules within a historical time period. Anomalies in the status information of risk control processing rules may include a mismatch between the trend of the hit rate and the trend of the accuracy of the risk control processing rules. The hit rate of a risk control processing rule can be characterized by the ratio of the number of input data points that hit the rule to the total number of input data points. The accuracy of a risk control processing rule can be characterized by the ratio of the number of risky data points among the input data points that hit the rule to the total number of input data points that hit the rule. Strategies for adjusting risk control processing rules may include setting a secondary verification process within the risk control processing rules. This secondary verification process is used to verify the validity of risky input data before processing it through the risk control processing rules.

[0063] The hit rate reflects the coverage and trigger frequency of risk control rules in the input data, i.e., the extent to which the rule can capture relevant input data. A high hit rate may indicate overly strict rules, causing many data points without potential risk to be mistakenly identified as risky. Accuracy, on the other hand, reflects the precision with which risk control rules identify risky data, i.e., the proportion of input data that actually contains risk within the hit rules. A low accuracy rate leads to numerous false positives, wasting review resources and potentially disrupting normal service processes. Secondary validation allows for more precise labeling and filtering of risky data, reducing false positives caused by overly strict rules. This helps to effectively intercept high-risk data while minimizing disruption to normal service processes, improving the accuracy and adaptability of the risk control model.

[0064] In step S16, the operating parameters of the risk control system can be adjusted based on a parameter adjustment strategy. For example, if the parameter adjustment strategy includes optimizing the model parameters of the risk detection model, the model parameters of the risk detection model can be optimized. Specifically, a control command can be sent to the model management module to control the model management module to retrain the model, adjust the model's hyperparameters, or update the model's algorithm structure. As another example, if the parameter adjustment strategy includes adjusting the risk control processing rules, the risk control processing rules can be adjusted. Specifically, a control command can be sent to the rule management module to control the rule management module to regenerate the risk control processing rules.

[0065] In some embodiments, parameter adjustment strategies can be written into the risk control project knowledge base, which can be viewed by risk control experts. Specifically, the risk control project knowledge base can provide a user interface through which risk control experts can view parameter adjustment strategies and other information in the knowledge base, and also modify information within it. This allows risk control experts to gain a deeper understanding of past adjustment approaches and operational details, providing crucial reference for subsequent actions such as formulating or optimizing risk control processing rules, adjusting model parameters of risk control detection models, and / or modifying project information, thereby helping to improve the accuracy and effectiveness of risk control work.

[0066] Furthermore, the generated parameter adjustment strategies can first be confirmed by risk control experts. After the risk control experts confirm the parameter adjustment strategies in the risk control project knowledge base, the operating parameters of the risk control system can be adjusted based on these strategies. The risk control experts, leveraging their professional knowledge and experience, evaluate the rationality, effectiveness, and potential impact of the parameter adjustment strategies. Once the risk control experts confirm the parameter adjustment strategies in the risk control project knowledge base, these professionally reviewed parameter adjustment strategies will serve as a reliable basis for precisely adjusting the operating parameters of the risk control system to optimize its performance.

[0067] Furthermore, risk control experts can modify the parameter adjustment strategies in the risk control project knowledge base. After modifying these strategies, the operating parameters of the risk control system can be adjusted accordingly. Risk control experts can optimize or revise existing strategies based on the latest risk assessment results, market dynamics, or changes in corporate policies. The modified parameter adjustment strategies will better align with current service needs and risk conditions, providing more precise guidance for adjusting the operating parameters of the risk control system. In this way, the risk control system can reflect the latest risk control requirements in real time, improving its adaptability and effectiveness in complex and ever-changing environments.

[0068] In some embodiments, the risk control system further includes a management platform for monitoring the status of the risk control system. The management platform can alert risk control experts upon detecting an abnormal state, prompting them to update the project information in the risk control project knowledge base. Based on this, in response to the detection of an update to the project information in the risk control project knowledge base, the operating parameters of the risk control system can be adjusted according to the updated project information. In this embodiment, when the project information in the risk control project knowledge base is updated, the change can be automatically detected and the update process can be triggered, ensuring that the operating parameters of the risk control system are always optimized and adjusted according to the latest project information. This not only improves the intelligence level of the risk control system but also enhances its adaptability and effectiveness in a variable risk environment. In some embodiments, a target level or target type of abnormal state requiring alerts can be preset. If the level of the detected abnormal state reaches the target level of the abnormal state requiring alerts, or the type of the detected abnormal state matches the target type, an alert is sent to the risk control experts.

[0069] This specification's embodiments achieve automated and intelligent adjustments to the risk control system through the collaboration of "risk control experts, risk control project knowledge base, and large language models." On one hand, the large language model possesses rich industry knowledge, efficient risk control system construction capabilities, and automated and intelligent system operation and maintenance capabilities, reducing the burden on risk control experts in knowledge acquisition and system construction, allowing them to focus more on in-depth analysis and decision-making in specific situations. On the other hand, feedback from risk control experts helps the large language model correct biases, supplement non-public knowledge, and achieve continuous learning and optimization in specific projects, improving the accuracy and adaptability of the large language model. In summary, this specification's embodiments effectively combine the advantages of large language models and risk control experts while avoiding their respective shortcomings, achieving automated and intelligent construction and improvement of the risk control system for specific risk control projects, effectively lowering the threshold for risk control work and improving its efficiency and effectiveness.

[0070] In some embodiments, the current risk control strategy and the test risk control strategy of the risk control system can also be compared. The current risk control strategy refers to the risk control strategy currently adopted by the risk control system, including but not limited to the risk detection model and risk processing rules currently used by the risk control system. Specifically, in response to obtaining the test risk control strategy, a traffic allocation strategy for the current risk control strategy and the test risk control strategy can be determined. The traffic allocation strategy is used to determine the ratio of input data for risk detection based on the current risk control strategy to input data for risk detection based on the test risk control strategy. Then, the risk control effects of the current risk control strategy and the test risk control strategy on their respective acquired input data can be obtained. A comparison report of the current risk control strategy and the test risk control strategy is generated by the large language model based on the risk control effects, and the comparison report is written to the risk control project knowledge base. This allows risk control experts to determine whether to switch the current risk control strategy to the test risk control strategy based on the comparison report in the risk control project knowledge base.

[0071] In this embodiment, the large language model plays the role of an "intelligent experimental platform". It can automatically allocate traffic to the current risk control strategy and the test risk control strategy, conduct comparative tests on the two risk control strategies, and automatically generate a comparison report so that users can determine the performance of the two risk control strategies and realize the iteration of risk control strategies.

[0072] See Figure 4 This specification also provides a data processing device for a risk control system, the device comprising:

[0073] The first acquisition module 102 is used to acquire risk identification attribute data of the risk control system, wherein the risk identification attribute data is used to characterize the risk identification capability of the risk control system;

[0074] The second acquisition module 104 is used to acquire project information of risk control projects from the risk control project knowledge base, and send the project information and the risk identification attribute data to the big language model, so that the big language model can analyze the risk identification attribute data based on the project information to obtain the operating parameter adjustment strategy of the risk control system; the project information is used to describe the risk control project, and the project information is generated by the big language model based on the requirement information of the risk control project provided by the risk control expert;

[0075] The parameter adjustment module 106 is used to adjust the operating parameters of the risk control system based on the parameter adjustment strategy.

[0076] This specification also provides a computer device, which includes at least a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the methods described in any of the foregoing embodiments.

[0077] Figure 5 This diagram illustrates a more specific hardware structure of a computer device provided in an embodiment of this specification. The device may include: a processor 202, a memory 204, an input / output interface 206, a communication interface 208, and a bus 210. The processor 202, memory 204, input / output interface 206, and communication interface 208 are interconnected internally via the bus 210.

[0078] The processor 202 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification. The processor 202 may also include a graphics card, such as an Nvidia Titan X graphics card or a 1080Ti graphics card.

[0079] The memory 204 can be implemented in the form of read-only memory (ROM), random access memory (RAM), static storage device, dynamic storage device, etc. The memory 204 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 204 and is called and executed by the processor 202.

[0080] Input / output interface 206 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0081] The communication interface 208 is used to connect the communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, Wi-Fi, Bluetooth, etc.).

[0082] Bus 210 includes a pathway for transmitting information between various components of the device, such as processor 202, memory 204, input / output interface 206, and communication interface 208.

[0083] It should be noted that although the above-described device only shows the processor 202, memory 204, input / output interface 206, communication interface 208, and bus 210, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0084] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. When implementing the embodiments of this specification, the functions of each module can be implemented in one or more software and / or hardware. Alternatively, some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0085] The above description is merely a specific implementation of the embodiments of this specification. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles of the embodiments of this specification, and these improvements and modifications should also be considered within the protection scope of the embodiments of this specification.

Claims

1. A data processing method for a risk control system, the method comprising: Obtain risk identification attribute data from the risk control system, wherein the risk identification attribute data is used to characterize the risk identification capability of the risk control system; The project information of the risk control project is obtained from the risk control project knowledge base. The project information and the risk identification attribute data are sent to the big language model so that the big language model can analyze the risk identification attribute data based on the project information to obtain the operation parameter adjustment strategy of the risk control system. The project information is used to describe the risk control project and is generated by the big language model based on the requirement information of the risk control project provided by the risk control expert. The operating parameters of the risk control system are adjusted based on the parameter adjustment strategy.

2. The method according to claim 1, wherein the risk control system comprises a model platform with a risk detection model deployed and a decision engine with risk control processing rules deployed, the risk detection model being used to perform risk detection on the input data of the risk control system, and the decision engine being used to process the risk detection results of the risk detection model based on the risk control processing rules; the risk identification attribute data includes: Model validity data is used to characterize the effectiveness of the risk detection model used in the risk control system, and Rule validity data is used to characterize the validity of the risk control processing rules adopted by the risk control system.

3. The method according to claim 2, wherein analyzing the risk identification attribute data based on the project information to obtain the operating parameter adjustment strategy of the risk control system includes: Based on the project information, the risk identification attribute data is analyzed to obtain the operating status information of the risk control system. The operating status information includes the status information of the risk detection model and the status information of the risk control processing rules. If the status information of the risk detection model is abnormal, the operational parameter adjustment strategy of the risk control system is determined to include a strategy of adjusting the model parameters of the risk detection model; If the status information of the risk control processing rule is abnormal, the operating parameter adjustment strategy of the risk control system is determined to include the strategy of adjusting the risk control processing rule.

4. The method according to claim 3, wherein the status information of the risk detection model includes the changing trend of the operating parameters of the risk detection model within a historical time period; The abnormal state information of the risk detection model includes: the changing trend of the group stability index of the risk control system's features does not match the changing trend of the operating parameters of the risk detection model; The strategy for adjusting the model parameters of the risk detection model includes: an incremental training strategy for the risk detection model based on target data generated within the historical time period; wherein the rate of change of the population stability index of the target data features is higher than a preset rate of change threshold.

5. The method according to claim 3, wherein the state information of the risk detection model includes the expected change trend of the operating parameters of the risk detection model in a future time period, and the operating parameters of the risk detection model include the detection accuracy of the risk detection model for a specified category of data; The abnormal state information of the risk detection model includes: the expected trend of the risk detection model's detection accuracy for the input data of the specified category is a decreasing trend; The strategy for adjusting the model parameters of the risk detection model includes: adjusting the model parameters of the risk detection model that are related to the specified category.

6. The method according to claim 3, wherein the status information of the risk control processing rule includes the hit rate and accuracy of the risk control processing rule within a historical time period; The abnormal status information of the risk control processing rule includes: The trend of the hit rate of the risk control processing rule does not match the trend of the accuracy of the risk control processing rule. The strategy for adjusting the risk control processing rules includes: setting a secondary verification process in the risk control processing rules, wherein the secondary verification process is used to verify the validity of the risky input data before processing the risky input data through the risk control processing rules.

7. The method according to claim 1, wherein the risk control system includes a model platform deployed with a risk detection model, the risk detection model being used to perform risk detection on the input data of the risk control system; the risk control project knowledge base is further used to store descriptive information of the input data of the risk control system; the method further includes: The description information is sent to the feature engineering module so that the feature engineering module performs feature engineering operations based on the description information to obtain the input features of the risk detection model.

8. The method according to claim 1, wherein adjusting the operating parameters of the risk control system based on the parameter adjustment strategy includes: The parameter adjustment strategy is written into the risk control project knowledge base, and the parameter adjustment strategy in the risk control project knowledge base can be viewed by risk control experts. After the risk control expert confirms the parameter adjustment strategy in the risk control project knowledge base, the operating parameters of the risk control system are adjusted based on the parameter adjustment strategy.

9. A data processing device for a risk control system, the device comprising: The first acquisition module is used to acquire risk identification attribute data of the risk control system, wherein the risk identification attribute data is used to characterize the risk identification capability of the risk control system; The second acquisition module is used to acquire project information of risk control projects from the risk control project knowledge base, and send the project information and the risk identification attribute data to the big language model, so that the big language model can analyze the risk identification attribute data based on the project information to obtain the operating parameter adjustment strategy of the risk control system; the project information is used to describe the risk control project, and the project information is generated by the big language model based on the requirement information of the risk control project provided by the risk control expert; The parameter adjustment module is used to adjust the operating parameters of the risk control system based on the parameter adjustment strategy.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method of any one of claims 1 to 8.

Images