Risk control system based on multi-modal rule engine

Abstract

The application discloses a risk control system based on a multi-modal rule engine, comprising a multi-modal rule engine architecture, a dynamic risk control model calculation system and a high-performance distributed architecture, wherein the multi-modal rule engine architecture comprises a scene center, a decision center and a system center, the dynamic risk control model calculation system comprises a timing task scheduling module, a central control application integration module and an automatic model optimization module, and the high-performance distributed architecture comprises a load balancing layer, an application layer and a data layer. The application can realize efficient calculation and real-time decision of risk control indexes through multi-dimensional rule configuration, dynamic workflow arrangement and intelligent model optimization.

Description

Technical Field

[0001] This invention relates to the field of computer risk control technology, specifically a risk control system based on a multimodal rule engine. Background Technology

[0002] Traditional risk control systems rely on static rule bases, making it difficult to cope with complex and ever-changing business scenarios (such as customs audits and enterprise assessments), and have the following drawbacks: 1. Insufficient flexibility in rule configuration: Relies on hard coding or a single visual editor, which cannot meet the needs of non-technical personnel and complex logic.

[0003] 2. Low data integration efficiency: Multi-source heterogeneous data (such as transaction data and equipment information) requires manual cleaning and transformation, which leads to decision-making delays.

[0004] 3. Lagging model updates: Rule iteration requires manual intervention and cannot be dynamically optimized based on real-time data, affecting the accuracy of risk control.

[0005] 4. Poor system scalability: Centralized architecture is difficult to support high-concurrency scenarios, and the permission management is rudimentary, which poses a risk of data leakage. Summary of the Invention

[0006] The technical objective of this invention is to address the above-mentioned shortcomings by providing a risk control system based on a multimodal rule engine. Through multi-dimensional rule configuration, dynamic workflow orchestration, and intelligent model optimization, this system enables efficient calculation and real-time decision-making of risk control indicators.

[0007] This invention provides a risk control system based on a multimodal rule engine, comprising: a multimodal rule engine architecture, a dynamic risk control model calculation system, and a high-performance distributed architecture, wherein... The multimodal rule engine architecture includes a scene center, a decision center, and a system center. The scene center supports navigation-style and script-style rule editing, provides multiple rule models, and has a built-in method library. The built-in method library supports dynamic assignment and conversion of variables, parameters, and constants. The decision center is used to build decision flow workflows and provides decision services to the outside world through interfaces. The system center implements organization-level permission isolation based on the RBAC model. The dynamic risk control model calculation system includes a timed task scheduling module, a central control application integration module, and an automated model optimization module. The timed task scheduling module is used to trigger a big data program to calculate enterprise risk control indicators and store the results in an indicator library. The central control application integration module is used to call the rule engine through an interface, obtain the scoring results, and persist them. The automated model optimization module is used to train and evaluate the model based on historical data and dynamically adjust the rule weights and thresholds. The high-performance distributed architecture includes a load balancing layer, an application layer, and a data layer. The application layer is divided into Web services, engine services, and static resource services. The data layer integrates structured data, caching, and service discovery.

[0008] The navigation method includes a drag-and-drop GUI, the script method includes Groovy scripts, and the multiple rule models include at least a normal rule model, a cyclic rule model, a scorecard model, and a decision table model.

[0009] The built-in method library contains built-in functions and custom methods, the variables include business characteristics, the parameters include intermediate variables, and the constants include enumerated values.

[0010] The decision center constructs a decision workflow that supports nested nodes and mutually exclusive / parallel execution.

[0011] The system center supports fine-grained management of users, roles, and organizations.

[0012] The scheduled task module triggers a big data program to calculate enterprise risk control indicators according to a predetermined cycle. The enterprise risk control indicators include at least transaction frequency and equipment risk value.

[0013] The central control application integration module calls the rule engine through an HTTP interface, passing in the enterprise ID, model type, and periodic parameters to obtain and persist the scoring results.

[0014] The application layer supports horizontal scaling.

[0015] The risk control system based on a multimodal rule engine of this invention has the following advantages: 1. Hybrid rule editing mode: navigation-style editing lowers the barrier to entry for non-technical personnel, while script-style editing supports complex logic (such as regular expression matching) to meet the needs of multiple scenarios; 2. Dynamic workflow orchestration: the decision flow supports node-level parallel and mutually exclusive configuration, and flexible scheduling of the risk control process is achieved through DAG (Directed Acyclic Graph); 3. Data-driven model optimization: an automated evaluation module is introduced, and rule priorities are dynamically adjusted based on A / B test results, reducing manual intervention; 4. Institutional-level data isolation: Redis sharding and database schema isolation ensure that data from different institutions is stored and computed independently. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] The invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 This is a logical structure block diagram of a risk control system based on a multimodal rule engine according to an embodiment of the present invention; Figure 2 This is an overall architecture diagram of a risk control system based on a multimodal rule engine according to an embodiment of the present invention; Figure 3 This is a flowchart of the scoring process of a risk control system based on a multimodal rule engine according to an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments are not intended to limit the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0020] It should be understood that in the description of the embodiments of the present invention, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance, nor as indicating or implying order. In the embodiments of the present invention, "multiple" refers to two or more.

[0021] In this embodiment of the invention, "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, B existing alone, or both A and B existing simultaneously. Furthermore, in this document, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0022] Definitions: Drag-and-drop GUI: A visual rule configuration interface.

[0023] Groovy scripts: An embedded dynamic script execution engine that allows users (usually risk control experts or developers) to directly write scripts to define complex business logic.

[0024] RESTful interface: A standardized web service interface based on the HTTP protocol and following the REST (Representational State Transfer) architectural style.

[0025] RBAC (Role-Based Access Control) is an access control model that manages user permissions through roles. Its core principle is to assign permissions to roles rather than users, simplifying the permission management process. Tomcat container: is the core open-source Java web application server of the Apache Software Foundation's Jakarta project.

[0026] PostgreSQL: A feature-rich free software object-relational database management system (ORDBMS) based on POSTGRES version 4.2 developed by the Computer Science Department of the University of California.

[0027] Redis (Remote Dictionary Server) is an open-source, ANSI C-written, network-enabled, in-memory or persistent, log-structured, key-value database that provides APIs for multiple languages.

[0028] ZooKeeper is a distributed, open-source distributed application coordination service. It is an open-source implementation of Google's Chubby and a crucial component of Hadoop and HBase. It provides consistency services for distributed applications, offering features such as configuration maintenance, domain name service, distributed synchronization, and group services.

[0029] like Figure 1 As shown, the risk control system based on a multimodal rule engine provided in this embodiment can serve as an intelligent risk control rule engine system and its dynamic decision-making method for customs operations.

[0030] This embodiment provides a risk control system based on a multimodal rule engine, which may include a multimodal rule engine architecture, a dynamic risk control model calculation system, and a high-performance distributed architecture.

[0031] The multimodal rule engine architecture includes three core modules: scenario center, decision center, and system center. The scenario center supports navigation-style and script-style rule editing, provides multiple rule models, and has a built-in method library that supports dynamic assignment and conversion of variables, parameters, and constants. The decision center is used to build decision workflows and provides decision services to the outside world through interfaces. The system center implements organization-level permission isolation based on the RBAC model.

[0032] The Scenario Center offers two rule editing methods: a navigation-based (drag-and-drop GUI) and a script-based (Groovy script) approach. It supports users in flexibly customizing business rule logic and integrates various preset rule models, covering ordinary rules, loop rules, scorecards (simple / complex), and decision tables (ordinary / cross-tabulation), significantly improving rule configuration efficiency. The Scenario Center also includes a rich built-in method library (containing built-in functions and custom methods), supporting dynamic assignment and type conversion of variables, parameters, and constants, enhancing the adaptability and extensibility of rule expression. Variables can include business characteristics, parameters can include intermediate variables, and constants can include enumerated values.

[0033] The decision center constructs decision workflows, supporting nested nodes and mutually exclusive / parallel execution. It supports the visual orchestration of complex business decision paths and provides efficient and stable decision services to the outside world through standardized interfaces (RESTful interfaces), facilitating system integration and function reuse.

[0034] The system center is based on the RBAC permission model, supporting fine-grained control over users, roles, and organizations. It enables fine-grained permission control for different organizations, departments, and users, ensuring data security and operational isolation in multi-user collaborative environments.

[0035] The dynamic risk control model calculation system includes a timed task scheduling module, a central control application integration module, and an automated model optimization module. The timed task scheduling module is used to trigger big data programs to calculate enterprise risk control indicators and store the results in the indicator library. The central control application integration module is used to call the rule engine through a standard interface and obtain the scoring results. At the same time, the automated model optimization module is used to continuously train and evaluate the model based on historical data, dynamically adjust the weight configuration and judgment threshold in the rule engine, thereby continuously improving the accuracy and adaptability of the system.

[0036] The time-based task scheduling module triggers a big data program to calculate enterprise risk control indicators according to a predetermined cycle. These indicators include at least transaction frequency and equipment risk value. The predetermined cycle can be the 1st of each month.

[0037] The central control application integration module calls the rule engine service via an HTTP interface, passing in the enterprise ID, model type, and periodic parameters to obtain real-time scoring results and persistently store these scoring results.

[0038] The automated model optimization module analyzes accumulated historical data, continuously trains and evaluates various prediction models (such as logistic regression and random forest) based on the historical data, and dynamically adjusts rule weights and decision thresholds to improve the model's prediction accuracy.

[0039] A high-performance distributed architecture mainly includes a load balancing layer, an application layer, and a data layer. The application layer is divided into web services, engine services, and static resource services, while the data layer integrates structured data, caching, and service discovery.

[0040] The high-performance distributed architecture deployment solution uses Nginx as the load balancing component to uniformly receive external requests and efficiently distribute them. The application layer is service-splittered, deploying web services (based on a Tomcat container), engine services (Tomcat), and static resource services, supporting horizontal scaling. The data layer integrates multiple storage technologies: PostgreSQL for storing structured business data, Redis for high-speed caching to improve response performance, and ZooKeeper for service registration and discovery, ensuring high availability and scalability.

[0041] The risk control system based on a multimodal rule engine according to the present invention has been described above by way of example with reference to the accompanying drawings. However, those skilled in the art should understand that various modifications can be made to the risk control system based on a multimodal rule engine proposed in the present invention without departing from the scope of the invention. Therefore, the scope of protection of the present invention should be determined by the contents of the appended claims.

Claims

1. A risk control system based on a multimodal rule engine, characterized in that, include: The architecture includes a multimodal rules engine, a dynamic risk control model calculation system, and a high-performance distributed architecture. The multimodal rule engine architecture includes a scene center, a decision center, and a system center. The scene center supports navigation-style and script-style rule editing, provides multiple rule models, and has a built-in method library. The built-in method library supports dynamic assignment and conversion of variables, parameters, and constants. The decision center is used to build decision flow workflows and provides decision services to the outside world through interfaces. The system center implements organization-level permission isolation based on the RBAC model. The dynamic risk control model calculation system includes a timed task scheduling module, a central control application integration module, and an automated model optimization module. The timed task scheduling module is used to trigger a big data program to calculate enterprise risk control indicators and store the results in an indicator library. The central control application integration module is used to call the rule engine through an interface, obtain the scoring results, and persist them. The automated model optimization module is used to train and evaluate the model based on historical data and dynamically adjust the rule weights and thresholds. The high-performance distributed architecture includes a load balancing layer, an application layer, and a data layer. The application layer is divided into Web services, engine services, and static resource services. The data layer integrates structured data, caching, and service discovery.

2. The risk control system based on a multimodal rule engine according to claim 1, characterized in that, The navigation method includes a drag-and-drop GUI, the script method includes Groovy scripts, and the multiple rule models include at least a normal rule model, a cyclic rule model, a scorecard model, and a decision table model.

3. The risk control system based on a multimodal rule engine according to claim 1, characterized in that, The built-in method library contains built-in functions and custom methods, the variables include business characteristics, the parameters include intermediate variables, and the constants include enumerated values.

4. The risk control system based on a multimodal rule engine according to claim 1, characterized in that, The decision center constructs a decision workflow that supports nested nodes and mutually exclusive / parallel execution.

5. The risk control system based on a multimodal rule engine according to claim 1, characterized in that, The system center supports fine-grained management of users, roles, and organizations.

6. The risk control system based on a multimodal rule engine according to claim 1, characterized in that, The scheduled task module triggers a big data program to calculate enterprise risk control indicators according to a predetermined cycle. The enterprise risk control indicators include at least transaction frequency and equipment risk value.

7. The risk control system based on a multimodal rule engine according to claim 1, characterized in that, The central control application integration module calls the rule engine through an HTTP interface, passing in the enterprise ID, model type, and periodic parameters to obtain and persist the scoring results.

8. The risk control system based on a multimodal rule engine according to claim 1, characterized in that, The application layer supports horizontal scaling.

Images