Risk control board automatic generation method and device, computer device and storage medium

Abstract

The embodiment of the application belongs to the technical field of artificial intelligence, and relates to a risk control board automatic generation method and device, a computer device and a storage medium. The method comprises the following steps: receiving a risk control board demand sent by a user terminal; obtaining a three-level prompt chain; performing structural analysis on the risk control board demand according to the three-level prompt chain to obtain a standardized generation instruction; inputting the standardized generation instruction into a high-performance computing engine to build a board to obtain a target risk control board. The application can be used for automatic generation of related risk control boards in a financial technology business system. By receiving a risk control board demand sent by a user terminal, using an open source programming framework to build a three-level prompt chain to perform structural analysis on the demand, obtaining a standardized generation instruction, and then automatically building a target risk control board with the aid of a high-performance computing engine, the application realizes quick, accurate and standardized generation of a risk control board, and improves the efficiency and flexibility of risk control work.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment and storage medium for automatically generating risk control dashboards. Background Technology

[0002] In the financial sector, risk control is of paramount importance. Risk control dashboards serve as crucial tools for risk control personnel to intuitively understand risk conditions, conduct risk assessments, and make decisions; their quality and generation efficiency directly impact the effectiveness of risk control efforts.

[0003] Traditionally, the generation of risk control dashboards relies primarily on manual operation. Risk control personnel need to manually collect, organize, and analyze large amounts of data based on their own experience and business needs, and then use specific software tools to draw and build the dashboards.

[0004] However, the applicant found that manually collecting and organizing data consumed a lot of time and energy, especially when dealing with massive amounts of data, the processing speed was slow and it was difficult to meet the needs of real-time risk control. In addition, manual operation is prone to problems such as data entry errors and analysis biases, which leads to inaccurate information presented on the risk control dashboard, thereby affecting the scientific nature of risk control decisions.

[0005] This shows that traditional methods for generating risk control dashboards suffer from low efficiency and limited accuracy. Summary of the Invention

[0006] The purpose of this application is to propose a method, apparatus, computer equipment, and storage medium for automatically generating risk control dashboards, so as to solve the problems of low efficiency and limited accuracy in traditional risk control dashboard generation methods.

[0007] To address the aforementioned technical problems, this application provides a method for automatically generating risk control dashboards, employing the following technical solution: Receive risk control dashboard requests sent from user terminals; Obtain the three-level hint chain; The risk control dashboard requirements are structured and parsed based on the three-level prompt chain to obtain standardized generation instructions; The standardized generation instructions are input into the high-performance computing engine to build the dashboard, resulting in the target risk control dashboard.

[0008] To address the aforementioned technical problems, this application also provides an automatic risk control dashboard generation device, which employs the following technical solution: The request receiving module is used to receive risk control dashboard requests sent by user terminals; The hint chain building module is used to obtain a three-level hint chain; The structured parsing module is used to perform structured parsing of the risk control dashboard requirements based on the three-level prompt chain to obtain standardized generation instructions; The Kanban building module is used to input the standardized generation instructions into the high-performance computing engine to build the Kanban and obtain the target risk control Kanban.

[0009] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution: It includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the risk control dashboard automatic generation method as described above.

[0010] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the risk control dashboard automatic generation method described above.

[0011] This application provides a method for automatically generating risk control dashboards, comprising: receiving risk control dashboard requests sent by a user terminal; obtaining a three-level prompt chain; performing structured parsing of the risk control dashboard requests based on the three-level prompt chain to obtain standardized generation instructions; and inputting the standardized generation instructions into a high-performance computing engine to build the dashboard, thereby obtaining the target risk control dashboard. Compared with existing technologies, this application, by receiving risk control dashboard requests sent by a user terminal, utilizing an open-source programming framework to construct a three-level prompt chain to perform structured parsing of the requests to obtain standardized generation instructions, and then automatically building the target risk control dashboard with the help of a high-performance computing engine, achieves rapid, accurate, and standardized generation of risk control dashboards, improving the efficiency and flexibility of risk control work. Attached Figure Description

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

[0013] Figure 1 This is an exemplary system architecture diagram to which this application can be applied; Figure 2 This is a flowchart illustrating the implementation of the risk control dashboard automatic generation method provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of the automatic risk control dashboard generation device provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0014] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0015] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0016] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0017] like Figure 1 As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0018] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0019] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0020] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0021] It should be noted that the automatic risk control dashboard generation method provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the automatic risk control dashboard generation device is generally set in the server / terminal device.

[0022] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0023] Continue to refer to Figure 2 The diagram shows a flowchart of an embodiment of the automatic risk control dashboard generation method according to this application. The above-described automatic risk control dashboard generation method includes steps S201, S202, S203, and S204.

[0024] In step S201, the risk control dashboard request sent by the user terminal is received.

[0025] In the embodiments of this application, a user inputs a request through their terminal device, which is received by the system. The terminal device may be a mobile terminal such as a mobile phone, smartphone, laptop, digital broadcast receiver, PDA (personal digital assistant), PAD (tablet computer), PMP (portable multimedia player), navigation device, etc., or a fixed terminal such as a digital TV, desktop computer, etc. It should be understood that the examples of terminal devices given herein are for convenience of understanding only and are not intended to limit this application.

[0026] In this application embodiment, the above-mentioned input data is the specific content that the user wants the system to answer. Specifically, the input data can be "transaction data, payment data, business data, or purchase data" related to financial institutions (such as banks). The input data can also be medical data related to medical scenarios. For example, personal health records, prescriptions, examination reports, etc. It should be understood that the examples of input data here are only for ease of understanding and are not intended to limit this application.

[0027] In step S202, a three-level hint chain is obtained.

[0028] In this embodiment, an open-source programming framework with strong extensibility and flexibility is selected, such as Python's Django, LangChain, or Flask frameworks. These frameworks provide rich libraries and tools, facilitating data processing, model building, and system development. Specifically: (1) First-level prompt chain: Based on general knowledge and rules in the risk control field, user needs are initially classified and identified. For example, needs are divided into major categories such as data indicators, display methods, and time ranges. Through preset keyword matching and semantic analysis algorithms, the approximate category of user needs is quickly located; (2) Second-level suggestion chain: Based on the first-level suggestion chain, the specific content under each major category is further refined. Taking the data indicator category as an example, the second-level suggestion chain can include specific indicators such as default rate, delinquency rate, and non-performing loan rate. Through deeper semantic analysis and contextual understanding, the specific indicators involved in user needs are determined; (3) Three-level prompt chain: For the specific content determined by the two-level prompt chain, detailed parameter settings and rule definitions are made. For example, for the default rate indicator, the three-level prompt chain can define parameters such as the data source, calculation method, and statistical period for calculating the default rate. In this way, the three-level prompt chain can transform the user's vague needs into specific and operable technical parameters.

[0029] In some optional implementations of this application's embodiments, this application can combine the engineering design paradigm of PromptChaining with the business characteristics of financial risk control dashboards to construct a three-level prompt chain of "strategy decoding → indicator construction → dashboard generation," achieving automatic generation of the entire chain from business intent to visual dashboards. Specifically, the overall architecture of the three-level prompt chain is shown in the table below:

[0030] Table 1: Overall Architecture of the Three-Level Hint Chain The core design logic of the three-level hint chain architecture is that the output of the preceding layer serves as the input of the following layer, and each layer enforces structured output (JSON format) to ensure the reliability of chained transmission. This is the key difference between the hint chain and a single hint—modular design, strong dependency transmission, and structured output.

[0031] (1) L1 Strategic Layer: Decoding Business Intent Prompt Objective: To transform natural language business requirements into structured Kanban meta-objectives.

[0032] Design considerations: • Employ a meta-problem strategy, first enabling the model to understand "what type of problem this task needs to solve"; • Forced output of JSON provides clear metrics for building boundaries for L2; • Three golden rules for the layered display of the risk control dashboard: Overview layer (one page for senior executives) → Analysis layer (multi-dimensional analysis) → Detailed layer (drill down).

[0033] (2) L2 Analysis Layer: Construction of Indicator System Prompt Objective: Based on the meta-objective output by L1, automatically construct a complete risk indicator tree and early warning rules.

[0034] Design considerations: • Single-focused subtasks: Each metric is built independently to avoid information clutter; • Introducing intelligent threshold design: Instead of relying on fixed thresholds, it is based on historical fluctuation ranges (e.g., the normal historical fluctuation range of the delinquency rate is 1.8% ± 0.3%, and an alert is issued if it exceeds this range). • Tiered early warning mechanism: Red (immediate action), Orange (same-day action), Yellow (attention), corresponding to the three key aspects of alarm design; This layer is the core engine of the entire alert chain, directly determining the information density and early warning capabilities of the dashboard.

[0035] (3) L3 Execution Layer: Kanban Visualization and Prompt Generation Objective: To transform the L2 metric tree into a directly renderable dashboard configuration (front-end components + alarm linkage).

[0036] Design considerations: • Structured output is mandatory: The output must be machine-readable JSON, directly interfaced with BI platforms (such as FineBI / FineReport) or front-end dashboard frameworks; • Built-in closed-loop tracking design: alarm → processing → feedback → optimization, all processes are electronically recorded; • Layered update frequency: core indicators at the minute level, key indicators at the hour level, and all indicators at the daily level, solving the fatal flaw of the traditional "T+1" reporting system.

[0037] In step S203, the risk control dashboard requirements are structured and parsed according to the three-level prompt chain to obtain standardized generation instructions.

[0038] In this embodiment, the application utilizes the rules and parameters defined in a three-level hint chain to parse the risk control dashboard request sent by the user layer by layer. Starting with the first-level hint chain, the general category of the request is determined; then, the specific content is further clarified based on the second-level hint chain; and finally, detailed parameter information is obtained based on the third-level hint chain. During the parsing process, the system understands and transforms the semantics of the request, converting the natural language description of the request into structured data that a computer can recognize and process. Specifically: Step 1: Input preprocessing (Parser layer) Objective: To transform users' natural language into L1-level structured input. For example, if a user inputs: "I want to check the recent overdue status of consumer loans, focusing on the East China region, for the director," the parsed result would be: {intent:"Post-loan monitoring - overdue analysis",entities:{product:"consumer loans",region:"East China"},constraints:{audience:"director",freq:"daily level"}}; Step 2: L1 Strategic Layer Analysis Objective: To parse the output of L1Prompt and extract the Kanban metatarget. For example, the original L1 output is: {"dashboard_meta":{"business_type":"post-loan","target_audience":"risk control director",...}}. The parsed structured data is directly passed as the context of L2. Step 3: Format conversion from L1 to L2 This is the most error-prone part, and explicit format conversion is required; Step 4: L2 analysis layer resolution Objective: To analyze the indicator tree, which contains the most information. Step 5: L2 to L3 format conversion Step 6: L3 Execution Layer Resolution In this embodiment, the parsing of the three-level hint chain is essentially a four-step loop of "input structuring → inter-level translation → output validation → failure fallback". Each step has clear parsing rules and fallback strategies to ensure that the entire chain from natural language to the final Kanban JSON is reliable and controllable.

[0039] In this embodiment, the structured parsed data is integrated and encapsulated to generate standardized generation instructions. These instructions contain all the information needed to generate a risk control dashboard, such as data metrics, data sources, calculation methods, display methods, and time ranges. The standardized generation instructions adopt a unified format and specifications, facilitating subsequent processing and execution by the high-performance computing engine.

[0040] In step S204, the standardized generation instructions are input into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard.

[0041] In this embodiment, a high-performance computing engine with efficient computing power and powerful data processing capabilities, such as Apache Spark, FineBi, or Flink, is selected. These engines can quickly process large-scale data and support complex computing tasks, meeting the requirements for real-time data and accuracy in the risk control dashboard generation process.

[0042] In this embodiment, after receiving standardized generation instructions, the high-performance computing engine retrieves relevant data from a data source based on the information in the instructions. The data source can be a database, data warehouse, or real-time data stream, etc. After retrieving the data, the engine processes and analyzes the data according to the calculation method defined in the instructions to calculate the required risk indicators. Then, according to the display requirements in the instructions, a visualization library (such as Matplotlib, ECharts, etc.) is used to generate corresponding charts and graphs from the calculation results, and these are combined into a complete risk control dashboard. Finally, the generated target risk control dashboard is returned to the user terminal for display.

[0043] In practical applications, users access the system interface through a computer browser and enter the following text requirement in the requirement input box: "Display the delinquency rate and non-performing loan rate of credit card business over the past week, presented in the form of bar charts and pie charts, with data sourced from the business database." The system receives this requirement through the request processing mechanism of the Django framework and stores it in a temporary variable. Then, this application performs the following operations: (1) Obtain the three-level hint chain: ① First-level suggestion chain construction: Define a requirement classification function in the Django framework. Through keyword matching and semantic analysis algorithms, identify the major categories of user requirements as "data indicator category" and "display method category".

[0044] ②Construction of a secondary prompt chain: For "data indicator category", further define specific indicators as "overdue rate" and "non-performing loan rate"; for "display method category", define specific display methods as "bar chart" and "pie chart".

[0045] ③ Three-level notification chain construction: For the "delinquency rate" indicator, the data source is defined as "business database", the calculation method is "number of delinquent accounts / total number of accounts", and the statistical period is "the past week"; for the "non-performing loan rate" indicator, the data source is also defined as "business database", the calculation method is "non-performing loan balance / total loan balance", and the statistical period is "the past week". For "bar charts" and "pie charts", define the chart titles, colors, labels, and other display parameters; (2) Based on the three-level prompt chain, the requirements of the risk control dashboard are structured and parsed to obtain standardized generation instructions: The system calls the requirement parsing function, parsing user requirements layer by layer according to the three-level suggestion chain rule. First, the first-level suggestion chain determines the major requirement category; then, the second-level suggestion chain clarifies the specific indicators and display methods; finally, the third-level suggestion chain retrieves detailed parameter information. The parsed data is then integrated to generate standardized generation instructions. (3) Input the standardized generation instructions into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard: The system sends standardized generation instructions to the Apache Spark computing engine. The Spark engine first connects to a MySQL database based on the data source information in the instructions to retrieve relevant data on credit card business over the past week. Then, it calculates the delinquency rate and non-performing loan rate according to the calculation methods defined in the instructions. Next, it uses Spark's visualization library (custom-developed in conjunction with Matplotlib) to generate bar charts and pie charts according to the display requirements in the instructions. Finally, the generated charts are combined into a complete risk control dashboard and returned to the user's browser via the Django framework for display.

[0046] This application provides a method for automatically generating risk control dashboards, comprising: receiving risk control dashboard requests sent by a user terminal; obtaining a three-level prompt chain; performing structured parsing of the risk control dashboard requests based on the three-level prompt chain to obtain standardized generation instructions; and inputting the standardized generation instructions into a high-performance computing engine to build the dashboard, thereby obtaining the target risk control dashboard. Compared with existing technologies, this application, by receiving risk control dashboard requests sent by a user terminal, utilizing an open-source programming framework to construct a three-level prompt chain to perform structured parsing of the requests to obtain standardized generation instructions, and then automatically building the target risk control dashboard with the help of a high-performance computing engine, achieves rapid, accurate, and standardized generation of risk control dashboards, improving the efficiency and flexibility of risk control work.

[0047] In some optional implementations of the embodiments of this application, the step of performing structured parsing of the risk control dashboard requirements to obtain standardized generation instructions specifically includes the following steps: (1) Call the large language model and input the risk control dashboard requirements into the large language model for scene recognition to obtain scene label data; (2) Read the risk control knowledge base and retrieve the template matching results corresponding to the scene tag data in the risk control knowledge base; (3) Generate standardized generation instructions based on template matching results.

[0048] In this application embodiment, a large language model with powerful natural language processing capabilities and pre-training or fine-tuning experience in the financial field is selected, such as the GPT series models fine-tuned with financial risk control data, and Wenxin Yiyan, etc. These models can understand complex natural language text and accurately identify its semantics and potential scenarios.

[0049] In this embodiment, the user-submitted risk control dashboard request text is used as input and passed to a large language model. The large language model uses its internal semantic understanding algorithm and knowledge graph to analyze and reason about the request text, identifying the risk control scenario to which the request belongs. For example, if the request text mentions "displaying the delinquency rate and default rate of a certain type of credit product over the past month," the large language model can identify the scenario as "displaying credit product risk indicators." After identification, corresponding scenario tag data is generated. The scenario tag data can be one or more keywords used to accurately describe the scenario characteristics of the request.

[0050] In this embodiment, the risk control knowledge base is a database containing rich risk control business knowledge and rules, storing requirement templates for various risk control scenarios. These templates are summarized and refined based on historical risk control dashboard requirements and practical business experience, and include standard information such as risk indicators, data sources, calculation methods, and display formats required for different scenarios. For example, for the "credit product risk indicator display" scenario, the knowledge base may store display templates containing indicators such as delinquency rate, non-performing loan rate, and loan balance.

[0051] In this embodiment, the application reads a risk control knowledge base and uses the obtained scene tag data as search keywords to perform exact or fuzzy matching within the knowledge base. By setting an appropriate matching algorithm and similarity threshold, the application finds the requirement template that best matches the scene tag data. If multiple matching results exist, they can be sorted according to their similarity level, and the most suitable template can be selected as the matching result.

[0052] In this embodiment, after obtaining the template matching result, the system extracts and organizes the information in the template. The risk indicators, data sources, calculation methods, display formats, and other parameters defined in the template are integrated and adjusted with the specific information (if any) in the user's requirements. For example, if the overdue rate calculation method defined in the template is "number of overdue accounts / total number of accounts," and the user's requirements do not specify other calculation methods, then that calculation method is directly adopted; if the user's requirements specify a data source, such as specifying data from a specific database, then the data source in the template is replaced with the user-specified source. After integration and adjustment, a standardized generation instruction is generated. This instruction adopts a unified format and specification, containing all the key information required to generate the risk control dashboard, facilitating subsequent processing and execution by the risk control dashboard generation system.

[0053] Compared with existing technologies, this application uses a large language model for scene recognition, combines the template matching results retrieved from the risk control knowledge base, and finally generates standardized generation instructions to achieve automated, accurate, and standardized parsing of risk control dashboard requirements, thereby improving the efficiency and quality of risk control dashboard generation.

[0054] In some optional implementations of the embodiments of this application, the step of inputting the risk control dashboard requirements into a large language model for scene recognition to obtain scene label data specifically includes the following steps: (1) Preprocess the risk control dashboard requirements to obtain preprocessed dashboard requirements; (2) Perform a complete semantic analysis of the preprocessing Kanban requirements based on the large language model to obtain business target data; (3) Based on the pre-constructed risk control domain knowledge graph, the business target data is matched with keywords and scenarios to obtain scenario tag data.

[0055] In this embodiment of the application, preprocessing includes data cleaning, word segmentation, part-of-speech tagging, and stop word filtering, specifically: (1) Data cleaning: Remove irrelevant characters, special symbols, spaces, etc. from the risk control dashboard requirement text to ensure the cleanliness and standardization of the text. For example, the requirement text may contain some garbled characters or redundant punctuation marks, which are removed through data cleaning to avoid interfering with subsequent processing; (2) Word segmentation: Use appropriate word segmentation algorithms to segment the cleaned text. In the field of financial risk control, it is necessary to use specialized word segmentation tools or models for this field to accurately identify professional terms and vocabulary. For example, for text such as "display of credit risk assessment indicators", it can be accurately segmented into words such as "credit", "risk assessment", "indicators", and "display". (3) Part-of-speech tagging: Perform part-of-speech tagging on the segmented words to clarify the part of speech of each word, such as nouns, verbs, adjectives, etc. This helps the subsequent large language model better understand the semantic structure of the text. For example, tagging "credit" as a noun and "display" as a verb enables the model to clearly grasp the grammatical relationships in the text; (4) Stop word filtering: Remove stop words in the requirement text, such as common words like "de", "shi", "zai" that contribute little to semantic understanding, so as to reduce data redundancy and improve processing efficiency. After the above preprocessing steps, a preprocessed dashboard requirement text is obtained, which prepares for the subsequent semantic analysis.

[0056] In the embodiments of the present application, a large language model with powerful natural language processing capabilities and a certain training basis in the financial field is selected in the present application, such as a GPT series model fine-tuned on financial texts or ERNIE Bot, etc. The semantic analysis interface of this model is called, and the preprocessed dashboard requirement text is transmitted to the model as input.

[0057] In the embodiments of the present application, the large language model uses its internal deep learning algorithm and huge knowledge reserve to conduct a comprehensive and in-depth semantic analysis on the input preprocessed dashboard requirement text. The model not only understands the meaning of each word in the text, but also can grasp the semantic relationships between words, sentence structures and the overall context. For example, for the requirement text "display the change trend of the overdue rate and non-performing rate of a certain type of credit product in the past month", the model can understand that its core business goal is to display the change trend of the risk indicators of the credit product, including the overdue rate and the non-performing rate, with the time range being the past month.

[0058] In the embodiments of the present application, after semantic analysis, key business target data is extracted from the output of the model in the present application. These data are presented in a structured form, clearly containing information such as the business field involved in the requirement, specific goals, and key indicators. For example, the extracted business target data can be expressed as: {Business field: Credit business; Business goal: Display the change trend of risk indicators; Key indicators: Overdue rate, non-performing rate; Time range: Past month}.

[0059] In the embodiments of the present application, the risk control domain knowledge graph is a database that stores financial risk control domain knowledge in the form of a graph structure, which contains information such as various risk control scenarios, business indicators, risk types, and association relationships. A comprehensive and accurate risk control domain knowledge graph is constructed by collecting and sorting out materials such as professional literature, historical data, and expert experience in the financial risk control field. For example, the knowledge graph may contain association relationships such as "Credit business - Risk indicator - Overdue rate" and "Credit business - Risk scenario - Overdue risk".

[0060] In this embodiment, keywords such as business domains and key indicators are extracted from business target data. These keywords are then matched with nodes in a risk control domain knowledge graph. By calculating the similarity between keywords and nodes, the most relevant nodes and association paths are found. For example, the extracted keywords "credit business" and "delinquency rate" can be matched with the relevant risk scenario node "delinquency risk" in the knowledge graph.

[0061] In this embodiment of the application, scenario tag data is generated based on the matching results. Scenario tag data can be one or more keywords used to accurately describe the risk control scenario to which the business objective belongs. For example, in the above example, the generated scenario tag data could be "credit business overdue risk scenario".

[0062] Compared with existing technologies, this application preprocesses the risk control dashboard requirements, uses a large language model to perform complete semantic analysis to obtain business target data, and then combines a pre-built risk control domain knowledge graph to match keywords and scenarios, thereby accurately generating scenario tag data. This solves the problems of low efficiency, limited accuracy, and lack of systematicity in existing technologies.

[0063] In some optional implementations of the embodiments of this application, after the step of inputting standardized generation instructions into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard, the following steps are also included: Acquire multi-source data corresponding to the requirements of the risk control dashboard; Identify high-priority risk nodes from multi-source data; The available space distribution is obtained from the limited interface layout, and the visual arrangement of high-priority risk nodes is optimized by force-directed layout algorithm to obtain the integrated multi-dimensional information view. Extract the path sequences of mutual influence between data through the integrated multidimensional information view; Real-time impact simulation results are obtained based on the path sequence; A visually intuitive color coding scheme is obtained based on the real-time impact simulation results; The color coding schemes are grouped according to the k-means clustering algorithm to determine the visualization rendering parameters; The target risk control dashboard is optimized by visualizing the rendering parameters to obtain the optimized target risk control dashboard.

[0064] In this application embodiment, the scope and source of data involved in the risk control dashboard requirement are determined, including but not limited to financial transaction data, corporate financial data, market data, industry dynamic data, etc.

[0065] In this application embodiment, relevant data is obtained from multiple data sources through data interfaces, database queries, web crawlers, etc., to ensure the integrity and accuracy of the data. For example, transaction records are obtained from financial transaction systems, financial statement data is obtained from corporate financial reports, and market conditions and industry dynamics information are obtained from financial news websites.

[0066] In the embodiments of this application, the acquired multi-source data is cleaned and preprocessed to remove noisy and duplicate data and convert the data into a format suitable for analysis.

[0067] In some optional implementations of the embodiments of this application, this application can extract key data relationships from multi-source data; process the key data relationships according to the graph neural network algorithm to obtain a risk association map; obtain the dynamic change trend between multi-dimensional information in multi-source data according to the risk association map; and mark multi-source data whose dynamic change trend exceeds a preset change trend threshold as potential impact chains, thereby determining high-priority risk nodes.

[0068] In this application embodiment, data mining algorithms, such as association rule mining and cluster analysis, are used to extract key data relationships from multi-source data. For example, association rule mining is used to discover the correlation between corporate financial indicators and market conditions, and cluster analysis is used to classify companies with similar risk characteristics.

[0069] In the embodiments of this application, the extracted key data relationships are constructed into a graph structure, where nodes represent data entities (such as enterprises, financial products, etc.) and edges represent relationships between data (such as association relationships, influence relationships, etc.).

[0070] In this embodiment, a graph neural network algorithm is used to process the constructed graph structure, learning the complex relationships and feature representations between nodes. Through training and optimization of multi-layer neural networks, a risk correlation graph that accurately reflects risk relationships is obtained. For example, in the risk correlation graph, it is clear how the risk of a certain enterprise propagates to other enterprises through supply chain relationships.

[0071] In this embodiment of the application, a time window is set to periodically update and analyze the risk association map.

[0072] In this application embodiment, the application monitors the attribute changes of nodes and edges in the risk association graph and analyzes the dynamic trends of changes between multi-dimensional information. For example, it observes the changes in a company's financial indicators at different points in time and how these changes affect its risk association with other companies.

[0073] In this application embodiment, a dynamic change trend threshold is preset, and a reasonable threshold range is set according to different data dimensions and risk types.

[0074] In this application embodiment, multi-source data whose dynamic change trend exceeds a preset threshold are marked as potential impact chains, which may indicate the spread and diffusion of risks.

[0075] In this application embodiment, high-priority risk nodes are identified based on the importance and scope of the potential impact chain. For example, in supply chain finance, changes in the risk of a core enterprise may have a significant impact on the entire supply chain; therefore, the core enterprise is identified as a high-priority risk node.

[0076] In this application embodiment, the application analyzes the limited interface layout of the target risk control dashboard to determine the distribution of available space, including space size, shape and location.

[0077] In this embodiment, a force-directed layout algorithm is used to visually optimize the arrangement of high-priority risk nodes. The force-directed layout algorithm simulates forces and motions in the physical world, calculating the attractive and repulsive forces between nodes to achieve a reasonable distribution of nodes within the available space, avoiding node overlap and excessive density.

[0078] In this embodiment, other relevant data and information are integrated with high-priority risk nodes according to certain logical relationships to obtain an integrated multi-dimensional information view. For example, information such as relevant risk indicators and impact paths are displayed around high-priority risk nodes.

[0079] In the embodiments of this application, the application uses path search algorithms, such as depth-first search and breadth-first search, to extract the path sequences that influence each other in the integrated multidimensional information view.

[0080] In this embodiment, the extracted path sequences are analyzed and filtered to remove redundant and invalid paths while retaining key path information. For example, the shortest impact path from a high-priority risk node to other important nodes is identified.

[0081] In this embodiment of the application, a risk impact simulation model is established, which uses the path sequence as input parameter to simulate the propagation and diffusion process of risk on different paths.

[0082] In this application embodiment, various factors are considered, such as the speed of risk propagation, the degree of impact, and the attenuation coefficient, to obtain real-time risk impact simulation results. For example, the simulation measures the impact of a certain risk event on other enterprises and financial products at different points in time.

[0083] In this application embodiment, an intuitive color coding scheme is designed based on the magnitude and degree of risk in the real-time impact simulation results. For example, red is used to represent high risk, yellow to represent medium risk, and green to represent low risk.

[0084] In this application embodiment, the mapping relationship between color and risk indicators is determined so that color can accurately reflect changes in risk. For example, the intensity of color can indicate the degree of risk.

[0085] In this embodiment, the k-means clustering algorithm is used to group color coding schemes, classifying colors with similar risk characteristics into one category. Clustering analysis allows for better organization and management of color information, improving visualization effectiveness.

[0086] In this application embodiment, the application determines visualization rendering parameters such as color transition, transparency, and border style based on clustering results, so as to make the display of the target risk control dashboard more beautiful and clear.

[0087] In this embodiment of the application, the determined visualization rendering parameters are applied to the target risk control dashboard to re-render and re-layout each element in the dashboard.

[0088] In this embodiment, the optimized target risk control dashboard is evaluated and tested to ensure it meets the requirements of a risk control dashboard and can accurately and intuitively display risk information. For example, financial risk control personnel are invited to try it out and provide feedback, and further adjustments and optimizations are made based on the feedback.

[0089] Compared with existing technologies, this application obtains multi-source data corresponding to the risk control dashboard requirements, deeply analyzes the relationships and dynamic trends between data, identifies high-priority risk nodes, optimizes the visualization layout, and performs personalized visualization optimization to obtain an optimized target risk control dashboard. This solves the problems of insufficient data utilization, inadequate capture of dynamic risk changes, poor visualization effects, and lack of personalized optimization in existing technologies.

[0090] In some optional implementations of the embodiments of this application, after the step of inputting standardized generation instructions into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard, the following steps are also included: Acquire raw, multi-source data corresponding to the requirements of the risk control dashboard; The real-time computing engine is invoked to perform data cleaning on the original multi-source data, resulting in cleaned multi-source data. Real-time data from cleaned multi-source data is stored in a distributed cluster. The historical data of real-time cleaned multi-source data is stored in a columnar database management system; The model calls the gateway to synchronize real-time and historical data to the dashboard data source of the high-performance computing engine.

[0091] In this embodiment, the application acquires raw, multi-source data corresponding to the risk control dashboard requirements. This raw data may come from different data sources, such as internal systems of financial institutions and external market data providers, covering various types including transaction data, customer information, and market data. By comprehensively and accurately acquiring raw, multi-source data, a rich data foundation is provided for subsequent data processing and analysis.

[0092] In this embodiment, the application invokes a real-time computing engine to clean the original multi-source data, resulting in cleaned multi-source data. The Flink real-time computing engine is characterized by high efficiency and low latency, enabling real-time processing of raw data to remove duplicates, correct errors, and fill in missing data, thereby improving data quality and accuracy. Data cleaning is a crucial step in ensuring the reliability of subsequent data analysis and risk control decisions.

[0093] In this embodiment, the real-time data from cleaned multi-source data is stored in a distributed cluster. Redis, a high-performance key-value storage system, features fast read / write speeds and persistence, making it suitable for storing real-time data. By storing real-time data in a distributed cluster, rapid data access and real-time updates can be achieved, meeting the high requirements of risk control dashboards for real-time data.

[0094] In this embodiment, the historical data of real-time cleaned multi-source data is stored in a columnar database management system. A columnar database management system is an open-source system with high query performance and powerful data analysis capabilities, making it suitable for storing large-scale historical data. By storing historical data in a columnar database management system, rich historical data support can be provided for subsequent data analysis and mining.

[0095] In this embodiment, the application synchronizes real-time and historical data to the dashboard data source of the high-performance computing engine via a model invocation gateway. The model invocation gateway, acting as a bridge for data synchronization, ensures accurate and timely synchronization of real-time and historical data to the high-performance computing engine, providing comprehensive and accurate data support for the risk control dashboard. Based on the synchronized data, the high-performance computing engine updates the risk control dashboard in real time, displaying the latest risk status and analysis results.

[0096] Compared with existing technologies, this application improves the accuracy and timeliness of risk control dashboards, providing strong support for financial risk control.

[0097] In some optional implementations of the embodiments of this application, the step of calling the real-time computing engine to perform data cleaning on the original multi-source data to obtain cleaned multi-source data specifically includes the following steps: The standard score algorithm is used to identify outliers in transaction amounts from the original multi-source data, and these outliers are corrected to obtain cleaned multi-source data; and / or Data deduplication is performed on the original multi-source data based on transaction identifier information to obtain cleaned multi-source data; and / or The original multi-source data is semantically standardized based on a unified index naming convention to obtain cleaned multi-source data.

[0098] In this embodiment, the application invokes the Flink real-time computing engine to process the transaction amount field in the original multi-source data. First, the mean (μ) and standard deviation (σ) of the transaction amount data are calculated. Then, based on the Z-score algorithm, for each transaction amount data point x, its Z-score value is calculated, with the formula Z = σx. μ. Set a threshold (e.g., Z=3). When |Z|>3, the transaction amount data point is determined to be an outlier.

[0099] In this application embodiment, for the identified outlier transaction amounts, this application adopts an appropriate repair strategy. For example, the outlier can be replaced by the average or median of other normal transaction amounts within the time period in which the data is located, or by a default value set according to business rules, thereby obtaining repaired transaction amount data, and finally obtaining cleaned multi-source data containing the repaired transaction amounts.

[0100] In this embodiment, the application utilizes the Flink real-time computing engine to extract transaction identification information from the original multi-source data. The transaction identification information can be a field that can uniquely identify a transaction, such as a transaction order number or a transaction serial number.

[0101] In this embodiment, the original multi-source data is deduplicated based on the extracted transaction identifier information. The Flink real-time computing engine can maintain a state that records the transaction identifiers that have been processed. When new data arrives, it checks whether its transaction identifier already exists in the state. If it does, the data is considered duplicate data and is discarded; if it does not exist, the data is retained and the state is updated. In this way, clean multi-source data free of duplicate data is obtained.

[0102] In this application embodiment, a unified indicator naming rule is pre-defined, and a mapping relationship table between the original indicator names and the unified indicator names is established. For example, the original data may contain indicators with different names such as "transaction amount", "transaction amount", and "sales amount" but actually represent the same meaning. In the mapping relationship table, they are all mapped to the unified indicator name "transaction amount".

[0103] In this embodiment, the application invokes the Flink real-time computing engine to iterate through the indicator names in the original multi-source data. Based on a pre-defined mapping table, the original indicator names are replaced with corresponding unified indicator names. For indicator names that do not have a corresponding relationship in the mapping table, further processing can be performed according to business logic, such as retaining the original name or manual intervention. After semantic standardization, cleaned multi-source data with unified indicator naming is obtained.

[0104] Compared with existing technologies, this application can perform outlier repair, data deduplication, and semantic standardization on raw multi-source data in real time and efficiently, thereby improving data quality and meeting the needs of real-time data analysis.

[0105] In some optional implementations of the embodiments of this application, after the step of inputting standardized generation instructions into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard, the following steps are also included: Obtain the indicator data of the target risk control dashboard, which includes high-priority indicators and medium- and low-priority indicators; Data corresponding to high-priority indicators is retrieved from the distributed cluster and refreshed according to the first preset refresh frequency; Data corresponding to low- and medium-priority indicators is obtained and refreshed according to the second preset refresh frequency.

[0106] In this embodiment, the application obtains the indicator data of the target risk control dashboard from relevant data sources. These indicator data are categorized into high-priority indicators and medium-to-low-priority indicators based on their importance to risk control operations and their frequency of change. High-priority indicators are typically those that have a critical impact on real-time risk monitoring and decision-making, such as abnormal fluctuations in real-time transaction amounts and the number of major risk warnings. Medium-to-low-priority indicators include those used for long-term risk trend analysis and statistical summarization, such as monthly risk loss rates and quarterly changes in risk exposure.

[0107] In this embodiment, the application retrieves data corresponding to high-priority indicators from a Redis cluster according to a pre-set first preset refresh frequency, and refreshes the display content of high-priority indicators in the target risk control dashboard. The Redis cluster has advantages such as high-speed read / write and data persistence, which can meet the real-time requirements of high-priority indicators. The first preset refresh frequency can be flexibly set according to the changing characteristics of high-priority indicators and business needs, for example, set to refresh once per second or once every 5 seconds, to ensure that risk control personnel can obtain the latest key risk information in a timely manner.

[0108] In this embodiment, a scheduled task can be established to trigger a data retrieval operation according to a first preset refresh frequency. A data query request is sent via an interface connection to the Redis cluster to obtain the latest data for high-priority indicators. Then, the obtained data is compared and updated with the existing high-priority indicator data in the target risk control dashboard to ensure that the data displayed on the dashboard is always up-to-date.

[0109] In this embodiment, the application acquires data corresponding to low- and medium-priority indicators according to a pre-set second preset refresh frequency, and refreshes the display content of low- and medium-priority indicators in the target risk control dashboard. Since low- and medium-priority indicators change relatively slowly, the second preset refresh frequency can be set relatively low, such as once per hour or once per day, to reduce unnecessary consumption of computing resources.

[0110] In this embodiment, different methods can be used to obtain low- and medium-priority indicator data depending on the data source. If the data is stored in a database, it can be obtained through a database query; if the data comes from an external data interface, it can be obtained by calling the corresponding interface. After obtaining the data, it is also compared and updated with the existing low- and medium-priority indicator data in the target risk control dashboard.

[0111] Compared with existing technologies, this application can dynamically update the indicator data of the target risk control dashboard by using different refresh frequencies according to the priority differences of the indicators, thereby improving the efficiency and rationality of data refresh and ensuring that the risk control dashboard can reflect the risk status in a timely and accurate manner, providing strong support for risk control decisions.

[0112] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0113] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0114] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0115] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0116] Further reference Figure 3 As a response to the above Figure 2 The implementation of the method shown in this application provides an embodiment of an automatic risk control dashboard generation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0117] like Figure 3 As shown, the risk control dashboard automatic generation device 200 of this application embodiment includes: The demand receiving module 210 is used to receive risk control dashboard demands sent by user terminals; Hint chain construction module 220 is used to obtain a three-level hint chain; The structured parsing module 230 is used to perform structured parsing of the risk control dashboard requirements based on the three-level prompt chain to obtain standardized generation instructions; The Kanban building module 240 is used to input standardized generation instructions into the high-performance computing engine to build the Kanban and obtain the target risk control Kanban.

[0118] In this embodiment, an automatic risk control dashboard generation device 200 is provided, comprising: a demand receiving module 210 for receiving risk control dashboard demands sent by a user terminal; a prompt chain construction module 220 for obtaining a three-level prompt chain; a structured parsing module 230 for performing structured parsing of the risk control dashboard demands based on the three-level prompt chain to obtain standardized generation instructions; and a dashboard building module 240 for inputting the standardized generation instructions into a high-performance computing engine to build the dashboard and obtain the target risk control dashboard. Compared with the prior art, this application, by receiving risk control dashboard demands sent by a user terminal, using an open-source programming framework to construct a three-level prompt chain to perform structured parsing of the demands to obtain standardized generation instructions, and then using a high-performance computing engine to automatically build the target risk control dashboard, achieves rapid, accurate, and standardized generation of risk control dashboards, improving the efficiency and flexibility of risk control work.

[0119] In some optional implementations of the embodiments of this application, the above-mentioned structured parsing module includes: The scene recognition submodule is used to call the large language model and input the risk control dashboard requirements into the large language model for scene recognition to obtain scene label data; The template matching submodule is used to read the risk control knowledge base and retrieve the template matching results corresponding to the scenario tag data from the risk control knowledge base; The standardized instruction generation submodule is used to generate standardized instructions based on the template matching results.

[0120] In some optional implementations of the embodiments of this application, the above-mentioned scene recognition submodule includes: The preprocessing unit is used to preprocess the risk control dashboard requirements to obtain preprocessed dashboard requirements; The complete semantic analysis unit is used to perform complete semantic analysis on the preprocessing dashboard requirements based on the large language model to obtain business target data. The keyword and scenario matching unit is used to match keywords and scenarios with business target data based on a pre-built risk control domain knowledge graph to obtain scenario tag data.

[0121] In some optional implementations of the embodiments of this application, the above-mentioned automatic risk control dashboard generation device 200 further includes: The multi-source data acquisition module is used to acquire multi-source data corresponding to the requirements of the risk control dashboard; The high-priority risk node identification module is used to identify high-priority risk nodes from multi-source data. The multidimensional information view acquisition module is used to obtain the available space distribution from the limited interface layout, and to optimize the visual arrangement of high-priority risk nodes using a force-directed layout algorithm to obtain the integrated multidimensional information view. The path sequence extraction module is used to extract the path sequences that influence each other between data through the integrated multidimensional information view; The real-time impact simulation result acquisition module is used to obtain real-time impact simulation results based on the path sequence; The color coding scheme acquisition module is used to obtain an intuitive color coding scheme based on the real-time impact simulation results. The visualization rendering parameter determination module is used to group color encoding schemes according to the k-means clustering algorithm and determine visualization rendering parameters. The visualization optimization module is used to optimize the target risk control dashboard based on the visualization rendering parameters, resulting in an optimized target risk control dashboard.

[0122] In some optional implementations of the embodiments of this application, the above-mentioned automatic risk control dashboard generation device 200 further includes: The raw multi-source data acquisition module is used to acquire raw multi-source data corresponding to the requirements of the risk control dashboard; The data cleaning module is used to call the real-time computing engine to clean the raw multi-source data and obtain cleaned multi-source data. The real-time data storage module is used to store the cleaned multi-source data in real time to a distributed cluster; The historical data storage module is used to store the historical data of real-time cleaned multi-source data into a columnar database management system; The data synchronization module is used to synchronize real-time and historical data to the Kanban data source of the high-performance computing engine by calling the gateway according to the model.

[0123] In some optional implementations of the embodiments of this application, the data cleaning module includes: The outlier handling submodule is used to identify outliers in transaction amounts from the original multi-source data based on a standard score algorithm, and to perform anomaly repair on these outliers to obtain cleaned multi-source data; and / or The data deduplication submodule is used to deduplicate the original multi-source data based on transaction identifier information to obtain cleaned multi-source data; and / or The semantic standardization submodule is used to perform semantic standardization on the original multi-source data according to a unified index naming, so as to obtain cleaned multi-source data.

[0124] In some optional implementations of the embodiments of this application, the above-mentioned automatic risk control dashboard generation device 200 further includes: The indicator data acquisition module is used to acquire indicator data from the target risk control dashboard. The indicator data includes high-priority indicators and medium- and low-priority indicators. The first data refresh module is used to pull data corresponding to high-priority indicators from the distributed cluster and refresh it according to a first preset refresh frequency. The second data refresh module is used to obtain and refresh the data corresponding to the medium and low priority indicators according to the second preset refresh frequency.

[0125] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of a computer device according to an embodiment of this application.

[0126] Computer device 300 includes a memory 310, a processor 320, and a network interface 330 that are interconnected via a system bus. It should be noted that only computer device 300 with components 310-330 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0127] Computer devices can include desktop computers, laptops, handheld computers, and cloud servers. These devices allow for human-computer interaction with users through keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0128] The memory 310 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 310 may be an internal storage unit of the computer device 300, such as the hard disk or memory of the computer device 300. In other embodiments, the memory 310 may also be an external storage device of the computer device 300, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 300. Of course, the memory 310 may include both internal storage units and external storage devices of the computer device 300. In the embodiments of this application, the memory 310 is typically used to store the operating system and various application software installed on the computer device 300, such as computer-readable instructions for the automatic generation method of risk control dashboards. In addition, the memory 310 can also be used to temporarily store various types of data that have been output or will be output.

[0129] In some embodiments, processor 320 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. Processor 320 is typically used to control the overall operation of computer device 300. In embodiments of this application, processor 320 is used to execute computer-readable instructions stored in memory 310 or process data, such as executing computer-readable instructions for an automatic risk control dashboard generation method.

[0130] The network interface 330 may include a wireless network interface or a wired network interface, which is typically used to establish a communication connection between the computer device 300 and other electronic devices.

[0131] The computer equipment provided in this application receives risk control dashboard requests sent by user terminals, uses an open-source programming framework to construct a three-level prompt chain to perform structured parsing of the requests, obtains standardized generation instructions, and then uses a high-performance computing engine to automatically build the target risk control dashboard, thereby achieving rapid, accurate, and standardized generation of risk control dashboards and improving the efficiency and flexibility of risk control work.

[0132] This application also provides another implementation method, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the risk control dashboard automatic generation method described above.

[0133] The computer-readable storage medium provided in this application receives risk control dashboard requests sent by user terminals, uses an open-source programming framework to construct a three-level prompt chain to perform structured parsing of the requests, obtains standardized generation instructions, and then uses a high-performance computing engine to automatically build the target risk control dashboard, thereby achieving rapid, accurate, and standardized generation of risk control dashboards and improving the efficiency and flexibility of risk control work.

[0134] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of this application.

[0135] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

[0136] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

Claims

1. A method for automatically generating risk control dashboards, characterized in that, Includes the following steps: Receive risk control dashboard requests sent from user terminals; Obtain the three-level hint chain; The risk control dashboard requirements are structured and parsed based on the three-level prompt chain to obtain standardized generation instructions; The standardized generation instructions are input into the high-performance computing engine to build the dashboard, resulting in the target risk control dashboard.

2. The method for automatically generating risk control dashboards according to claim 1, characterized in that, The step of performing structured parsing of the risk control dashboard requirements to obtain standardized generation instructions specifically includes the following steps: The large language model is invoked, and the risk control dashboard requirements are input into the large language model for scene recognition to obtain scene label data; Read the risk control knowledge base and retrieve the template matching results corresponding to the scenario tag data from the risk control knowledge base; The standardized generation instruction is generated based on the template matching result.

3. The method for automatically generating risk control dashboards according to claim 2, characterized in that, The step of inputting the risk control dashboard requirements into the large language model for scene recognition to obtain scene label data specifically includes the following steps: The risk control dashboard requirements are preprocessed to obtain preprocessed dashboard requirements; Based on the large language model, a complete semantic analysis is performed on the preprocessing Kanban requirements to obtain business target data; The business target data is matched with keywords and scenarios based on a pre-constructed risk control domain knowledge graph to obtain the scenario tag data.

4. The method for automatically generating risk control dashboards according to claim 1, characterized in that, After the step of inputting the standardized generation instructions into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard, the following steps are also included: Acquire multi-source data corresponding to the requirements of the risk control dashboard; Identify high-priority risk nodes from the multi-source data; The available space distribution is obtained from the limited interface layout, and the visual arrangement of the high-priority risk nodes is optimized by the force-directed layout algorithm to obtain the integrated multi-dimensional information view. The path sequences of mutual influence between data are extracted from the integrated multidimensional information view; The real-time impact simulation results are obtained based on the path sequence; Based on the real-time impact simulation results, an intuitive color coding scheme is obtained; The color encoding scheme is grouped according to the k-means clustering algorithm to determine the visualization rendering parameters; The target risk control dashboard is visualized and optimized based on the visualization rendering parameters to obtain the optimized target risk control dashboard.

5. The method for automatically generating risk control dashboards according to claim 1, characterized in that, After the step of inputting the standardized generation instructions into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard, the following steps are also included: Obtain the original multi-source data corresponding to the risk control dashboard requirements; The real-time computing engine is invoked to perform data cleaning on the original multi-source data to obtain cleaned multi-source data. The real-time data from the cleaned multi-source data is stored in a distributed cluster. The historical data of real-time cleaned multi-source data is stored in a columnar database management system; The model calls the gateway to synchronize real-time data and historical data to the Kanban data source of the high-performance computing engine.

6. The method for automatically generating risk control dashboards according to claim 5, characterized in that, The step of calling the real-time computing engine to perform data cleaning on the original multi-source data to obtain cleaned multi-source data specifically includes the following steps: The abnormal transaction amounts in the original multi-source data are identified using a standard score algorithm, and the abnormal transaction amounts are corrected to obtain the cleaned multi-source data. and / or The original multi-source data is deduplicated based on the transaction identifier information to obtain the cleaned multi-source data; and / or The original multi-source data is semantically standardized according to a unified index naming convention to obtain the cleaned multi-source data.

7. The method for automatically generating risk control dashboards according to claim 1, characterized in that, After the step of inputting the standardized generation instructions into the high-performance computing engine to build the dashboard and obtain the target risk control dashboard, the following steps are also included: Obtain the indicator data of the target risk control dashboard, wherein the indicator data includes high-priority indicators and medium- and low-priority indicators; Data corresponding to the high-priority indicators is retrieved from the distributed cluster and refreshed according to the first preset refresh frequency; Data corresponding to the low-to-medium priority indicators is obtained and refreshed according to the second preset refresh frequency.

8. An automatic risk control dashboard generation device, characterized in that, include: The request receiving module is used to receive risk control dashboard requests sent by user terminals; The hint chain building module is used to obtain a three-level hint chain; The structured parsing module is used to perform structured parsing of the risk control dashboard requirements based on the three-level prompt chain to obtain standardized generation instructions; The Kanban building module is used to input the standardized generation instructions into the high-performance computing engine to build the Kanban and obtain the target risk control Kanban.

9. A computer device, comprising a memory and a processor, characterized in that, The memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, it implements the steps of the automatic risk control dashboard generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the automatic risk control dashboard generation method as described in any one of claims 1 to 7.

Images