CN117785157B - Decision engine based on financial wind control business rule scene and implementation method - Google Patents

Abstract

The invention relates to the field of wind control distributed micro-service architecture, and provides a decision engine based on a financial wind control business rule scene and an implementation method thereof. The method aims to solve the problems that the prior investment entity scoring card model configuration is poor in data readability, maintainability and the like. The main scheme comprises that a scoring data model module creates a data model through scoring rules and a real-time data structure, and provides a data set for a scoring card. The scoring card module binds the scoring card with the data model and configures scoring rules. The decision flow module creates a decision flow, configures the order of the scoring cards and provides a calling template. And the simulation test module performs simulation test on the decision flow to obtain a scoring result. The test list module defines test objects, fills in test data, verifies decision flow according to the grading result, and stores the test objects. And the deployment configuration module binds the decision flow with the server, submits approval and obtains a rating model to be approved. The rule approval module verifies rule configuration and passes through the post-deployment model.

Description

Decision engine based on financial wind control business rule scene and implementation method

Technical Field

The invention relates to the field of wind control distributed micro-service architecture, and provides a decision engine based on a financial wind control business rule scene and an implementation method thereof.

Background

Traditional technical architecture constraints

At present, the regular arrangement and calculation scene of the open department unique wind control system are solidified, the regular configuration is mainly carried out through the fixed page development, the system still adopts the traditional single technical framework of most bank development technical systems, the system is provided with a plurality of functional modules such as a financing main body, wind control model management, early warning signal generation, early warning signal treatment flow, wind control rule configuration and the like, all externally exposed services are assembled through the calling among the modules, the micro-service among the modules is not carried out, and the technical support is carried out for the financing service of the bank clients of the open department on the framework through the financial market personal service and the management system framework of the traditional bank.

Business development and trading pressure

Starting from the last year, along with the rapid service mode of 'gold market service promotion' component platformization proposed by a certain city business and the rapid increase of the line investment and financing service volume, the system is suitable for the characteristics of short, flat, fast and miscellaneous supporting funds service, and although along with the continuous development of the service, the continuous introduction of various metering tools and the continuous patch upgrading of the system for supporting the development of the service, various problems are exposed, especially after early warning, the timeliness is poor, the functions are not aggregated enough, and the integrated comparison of the system is lacking, so that the comprehensive platform of the financial market service always does not reach the ideal of service top-level planning.

System reconstruction urgency brought by quick response service

However, in order to rapidly meet the development of high-speed business, the system is upgraded at high speed, the defects of the traditional monomer architecture are gradually shown, and the real-time performance of the wind control early warning of the market financial business is improved by introducing the Internet basic middleware such as redis, mq and the like to improve the timeliness of data transmission and sharing in the upgrading process, but the self architecture characteristics are not radically cured.

Instead, with the rapid iteration of the service, the logic complexity of the service code is increasing. The development of new business, and the maintenance of old systems, has become increasingly more barriers. Therefore, a set of decision computing technology frameworks capable of supporting large-volume commercial banking transaction volume impact and unique "componentized, high-response" business models are in need.

Disclosure of Invention

The method aims to solve the problems that the prior investment entity scoring card model configuration is poor in data readability, maintainability and the like.

In order to achieve the purpose, the engine adopts the following technical design scheme:

the invention provides a decision engine based on a financial wind control business rule scene, which comprises the following modules:

score facts data model declaration module:

Creating a scoring rule fact data model through a scoring rule fact data table or a real-time fact data structure, defining input variables and output variables through the model, creating rule constants and an aggregate calculation function, and providing a fact data set for subsequent configuration scoring cards;

rule scoring card declaration module:

Creating a rule scoring card, binding the rule scoring card with the rule fact data model created by the scoring fact data model declaration module, configuring the scoring card through input variables, output variables, rule constants and aggregation calculation functions defined by the rule fact data model created by the scoring fact data model declaration module, and providing scoring rules of fact data for subsequent rule decision flows;

rule decision flow declaration module:

Creating a rule decision flow, configuring the execution sequence of rule scoring cards created by each rule scoring card declaration module in the decision flow in a dragging mode to obtain a scoring rule fact data table or a scoring rule set template of a real-time fact data structure, and providing a calling template for a subsequent simulation test object to arrange a series or parallel decision model in a flow mode;

simulation test object declaration module:

Creating a simulation test object, binding the rule decision flow created by the rule decision flow declaration module with the rule test object, inputting scoring rule indexes of all the fact data in the rule decision flow created by the rule scoring card declaration module, and obtaining scoring results of the fact data of all the rule scoring cards bound by the decision flow;

Sub-step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking;

deploying a configuration object declaration module:

Creating a deployment configuration object, binding a rule decision flow created by a rule decision flow declaration module with a server, submitting approval by a main body grading model, and finally obtaining an investment main body grading model to be approved;

Rule approval module:

And (3) using the model to approve the post role user, carrying out rule configuration correctness verification and recheck approval, and carrying out on-line deployment on the investment entity rating model by the approval.

In the above decision engine, the scoring fact data model declaration module specifically includes the following steps:

Step S1.1: creating a rule fact data model, defining input variables and output variables through the model, creating rule constants and an aggregate calculation function, and providing a fact data set for a subsequent configuration grading card;

Substep s1.1.1: defining a data model list, pre-defining a data model number and a data model name after model data analysis, adding the pre-defined data model number and the data model name to the data model list, and binding the fact data list with the data model name in the data model list according to rules related to the bisector model.

Substep S1.1.2: defining a variable management list, selecting a grading rule fact data model, dragging according to a data structure field corresponding to the data model to achieve standard definition of variable names and variable types, wherein the grading rule fact data model is mapped by a data table, variables are mapped by data table fields, the variable types are converted into Java data types which can be identified by Java language codes according to definition of the data table structure field types, and finally the variable names and the variable types are added into the variable management list;

Substep S1.1.3: defining a rule constant management list, adding constant dictionary codes and constant dictionary names to the constant management list to obtain a constant dictionary in the rule constant management list, and then adding constant value information to the constant dictionary, wherein the constant value information comprises constant actual values, constant display values and data types;

Substep S1.1.4: defining an aggregation calculation function management list, adding a function code, a function name and a return value type into the aggregation calculation function management list, adding function parameter content according to requirements, inputting the parameter code, the parameter name and the parameter type, writing a statement of a function, compiling the function statement after writing, and storing the function after compiling.

In the above decision engine, the implementation of the rule scoring card declaration module specifically includes the following steps:

Substep S2.1: defining a scoring card list, adding a decision code and a decision name to the scoring card list, and selecting a rule fact data model to bind with the decision code of the rule scoring card;

Substep S2.2: the configuration rule scoring card specifically comprises:

adding scoring rules to variable, constant and function configuration of the rule fact data model bound by the scoring card;

Selecting variables, functions and constants to assign to categories, major factors and sub-category factors in a scoring rule, wherein the categories, major factors and sub-category factors are in progressive subdivision relation according to category levels;

filling in index weight, index options, conditions and index scores;

Wherein the variables refer to input variables and output variables, the constants are regular constants, and the functions refer to an aggregation calculation function;

substep S2.3: testing rule scoring cards, filling in each input item data contained in the scoring card model, calculating scoring results, and checking rule scoring cards according to the scoring results;

Substep S2.4: and issuing rule scoring cards, and selecting the checked rule scoring cards from the scoring card list for issuing.

In the above decision engine, the implementation of the rule decision flow declaration module specifically includes the following steps:

step S3.1: defining a decision flow list, and adding a decision code and a decision name filled in the form into the decision flow list;

step S3.2: configuring a rule decision flow, editing and dragging the issued rule scoring card to a work display area of the rule decision flow, and sequentially linking different nodes by using connecting lines according to an executed logic flow to obtain the rule decision flow;

step S3.3: testing rule decision flow, filling in the input item data in each scoring index in the rule decision flow, calculating scoring results after filling in, and checking the rule decision flow according to the scoring results;

Step S3.4: and issuing the rule decision flow, and selecting the checked rule decision flow from the decision flow list for issuing.

In the above decision engine, the implementation of the simulation test object declaration module includes the following steps:

Sub-step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking.

In the above decision engine, the deployment configuration object declaration module implementation package includes the following steps:

Substep S5.1: creating a deployment configuration list, creating a deployment configuration object in the deployment configuration list, filling in a release name and a server URL address, binding a released rule decision stream, then performing batch simulation test, storing the deployment configuration object after the test is successful to obtain a deployment task of an investment main body rating model, realizing distributed hot deployment on a production mode through management of a deployment task version, supporting seamless rollback and upgrading deployment iteration of the version, and distributing the deployment object to a service code through the server URL address for calling;

sub-step S5.2: submitting the investment entity rating model for approval.

In the above decision engine, the rule approval module sub-step S6.1: and examining and approving the investment entity rating model in the task list to be examined and approved, carrying out examination and approval double-post rechecking on the correctness of the investment entity rating model, and carrying out online deployment on the investment entity rating model through a post point.

The invention also provides a method for realizing the decision engine based on the financial wind control business rule scene, which comprises the following steps:

s1, scoring a fact data model declaration:

Creating a scoring rule fact data model through a scoring rule fact data table or a real-time fact data structure, defining input variables and output variables through the model, creating rule constants and an aggregate calculation function, and providing a fact data set for subsequent configuration scoring cards;

step S1.1: firstly loading a background grading fact data table in a business readable and understandable form at the front end, then creating a rule fact data model in a draggable and visual form at a front end page, defining input variables and output variables through the model after a required data model is obtained, creating a rule constant and an aggregation calculation function, and providing a fact data set for a follow-up configuration grading card;

Substep s1.1.1: defining a data model list, adding a predefined data model number and a predefined data model name after model data analysis to the data model list, and binding the fact data table with the data model names in the data model list according to a scoring model rule.

Substep S1.1.2: defining a variable management list, selecting a grading rule fact data model, dragging according to a data structure field corresponding to the data model to achieve standard definition of variable names and variable types, wherein the grading rule fact data model is mapped by a data table, variables are mapped by data table fields, variable types are converted into Java data types which can be identified by Java language codes according to definition of the data table structure field types, technical errors that grading model verification is not passed, java data types are inconsistent with database table data types and the like in the later stage of the grading model due to non-standardization caused by manually inputting model variables in the past are avoided, and finally the variable names and the variable types are added into the variable management list;

substep S1.1.3: defining a regular constant management list, adding constant dictionary codes and constant dictionary names to the constant management list to obtain a constant dictionary in the regular constant management list, and then adding constant value information to the constant dictionary, wherein the constant value information comprises constant actual values, constant display values and data types (Long, double, boolen, date, string).

Substep S1.1.4: defining an aggregation calculation function management list, adding a function code, a function name and a return value type (Long, double, boolen, date, string) into the aggregation calculation function management list, adding function parameter content according to requirements, inputting the parameter code, the parameter name and the parameter type, writing a sentence of a function, compiling the function sentence after writing, and storing the function after compiling;

s2, rule scoring card declares:

Creating a rule scoring card, binding the rule scoring card with the rule fact data model created in the step S1, configuring the scoring card through input variables, output variables, rule constants and aggregation calculation functions defined by the rule fact data model created in the step S1, and providing scoring rules of fact data for subsequent rule decision flows;

Substep S2.1: defining a scoring card list, adding a decision code and a decision name to the scoring card list, and selecting a rule fact data model to bind with the decision code of the rule scoring card;

Substep S2.2: the configuration rule scoring card specifically comprises:

adding scoring rules to variable, constant and function configuration of the rule fact data model bound by the scoring card;

Selecting variables, functions and constants to assign to categories, major factors and sub-category factors in a scoring rule, wherein the categories, major factors and sub-category factors are in progressive subdivision relation according to category levels;

fill in index weight, index options (A-Z), conditions (greater than, less than, equal to, among the collection); filling in index scores;

Wherein the variables refer to input variables and output variables, the constants are regular constants, and the functions refer to an aggregation calculation function;

substep S2.3: testing rule scoring cards, filling in each input item data contained in the scoring card model, calculating scoring results, and checking rule scoring cards according to the scoring results;

Substep S2.4: and issuing rule scoring cards, and selecting the checked rule scoring cards from the scoring card list for issuing.

S3, rule decision flow declaring:

Creating a rule decision flow, configuring the execution sequence of rule scoring cards created in each step S2 in the decision flow in a dragging mode, obtaining a scoring rule fact data table or a scoring rule set template of a real-time fact data structure, providing a calling template for a subsequent simulation test object, and arranging a series or parallel decision model in a flow mode;

step S3.1: defining a decision flow list, and adding a decision code and a decision name filled in the form into the decision flow list;

step S3.2: configuring a rule decision flow, editing and dragging the issued rule scoring card to a work display area of the rule decision flow, and sequentially linking different nodes by using connecting lines according to an executed logic flow to obtain the rule decision flow;

step S3.3: testing rule decision flow, filling in the input item data in each scoring index in the rule decision flow, calculating scoring results after filling in, and checking the rule decision flow according to the scoring results;

step S3.4: issuing rule decision flows, and selecting the checked rule decision flows from the decision flow list for issuing;

s4, declaring by a simulation test object:

Creating a simulation test object, binding the rule decision flow created in the step S3 with the rule test object, inputting scoring rule indexes of all the fact data in the rule decision flow created in the step S2, and obtaining scoring results of the fact data of all the rule scoring cards bound by the decision flow.

Sub-step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking.

S5, deploying a configuration object declaration:

Creating a deployment configuration object, binding the rule decision flow created in the step S3 with a server, submitting and approving the main body rating model, and finally obtaining an investment main body rating model to be approved;

Substep S5.1: creating a deployment configuration list, creating a deployment configuration object in the deployment configuration list, filling in a release name and a server URL address, binding a released rule decision stream, then performing batch simulation test, storing the deployment configuration object after the test is successful to obtain a deployment task of the investment main body rating model, realizing distributed hot deployment on a production mode through management of a deployment task version, supporting seamless rollback and upgrading deployment iteration of the version, and distributing the deployment object to a service code through the server URL address for calling.

Sub-step S5.2: submitting the investment entity rating model for approval.

S6, rule approval:

And (3) using the model to approve the post role user, carrying out rule configuration correctness verification and recheck approval, and carrying out on-line deployment on the investment entity rating model by the approval.

Substep S6.1: and examining and approving the investment entity rating model in the task list to be examined and approved, carrying out examination and approval double-post rechecking on the correctness of the investment entity rating model, and carrying out online deployment on the investment entity rating model through a post point.

Because the invention adopts the technical means, the invention has the following beneficial effects:

1. The decision engine is constructed based on a micro-service and distributed computing architecture, and can efficiently support a high-concurrency and high-throughput market investment transaction scene under financial market wind control business;

2. a streaming orchestration rules component:

In step S3, a rule decision flow declaration step, a flow type arrangement decision model is introduced, rule components of various scenes of a decision engine are combined together, the series or parallel combination is carried out through decision flows, and service codes are called and configured through the defined decision flows. By the method, the service flow calling instruction can be quickly arranged and realized, and service codes can call various models in a decision engine when data is loaded. The method provides a visual rule flexible configuration and maintenance mode for subsequent service change and modification, decouples the original hard coding mode and the flow mode, can improve the flexibility and responsiveness of the service change of the system, and solves the problem that the decision component can only be independently called in the traditional mode to cause inefficiency.

3. High-speed arithmetic processing capability:

In step S5, deployment configuration object declaration: the proposal applies for carrying out calculation microservice of the decision engine based on the microservice architecture, is completely compatible with the microservice architecture, fully utilizes the distributed framework and the elastic calculation capability of the microservice to carry out multi-node deployment and provide distributed calculation services, fully utilizes the memory lifting decision algorithm on the basis, and provides high-efficiency, high-speed and stable service processing capability for financial market service wind control.

4. Pure memory engine, independent of database:

in the step S1, scoring the fact data model declaration step: compared with a general rule engine on the market, the decision engine designed by the proposal application has high operation processing rule triggering capability, and is characterized in that the decision engine finishes the necessary condition that the decision engine needs to be loaded and calculated by utilizing the mode of loading the real-time data transmission and structure formed fact data and scoring rules into the memory of a computer when the main body scores, and carries out high-speed operation circulation of the fact data and the scoring model in the pure memory.

5. Asynchronous processing framework:

Through step S5, in the deployment configuration object declaration step, the substep S5.1 confirms that the proposal application provides a sound asynchronous scheduling framework, and the service code scheduling framework and the URL deployment calling interface can be subjected to centralized allocation and access to perform asynchronous calling processing on the scoring card regular configuration model through configuration of an asynchronous Mq mode. The rating producer generates a scheduling message, and after the consumer scheduling distributed decision engine calculates that the micro-service asynchronous polling frame consumes the asynchronous message, the decision engine is called to start and load the asynchronous decision model library at the same time, so that the subsequent decision engine scheduling task can be processed.

6. Perfect fault tolerance compensation capability:

Through the step S4, the simulation test object declaration step and the step S6, the rule approval step, besides having a perfect fault-tolerant mechanism, not only supports fault-tolerant checking in the process of calling a decision engine process, but also can realize the configuration correctness of a grading card model through double post rechecking, if the application of a decision engine micro-service of a certain deployment model is wrong, the sub-step S5.1 in the deployment configuration object declaration step can be matched with the fault-tolerant scheme configured in advance immediately according to the fault condition detected in advance, and the available decision engine micro-service can be switched immediately and the emergency process is started.

7. Fully applied hot deployment mode:

Through the step S5, in the deployment configuration object declaration step, the substep S5.1, the decision engine deployment of the proposal introduces a comprehensive model heat deployment mode, and records of state front-back change are carried out on each model deployment through the heat deployment mode before and after each wind control model deployment, thereby ensuring the model version consistency in the model deployment and iterative upgrading process and being close to the customized actual use scene.

In summary, the problems of poor data readability, poor maintainability and the like in the prior investment entity score card model configuration are often caused, the key problems that the score model can be updated by a technician rather than a wind control service model manager, the model iteration update period is long, the accuracy cannot be ensured and the like are caused when the score model is iteratively updated, the decision engine score card assembly is designed and mainly used for solving the technical problems that the service personnel can autonomously configure the model, the score card rule deployment and update iteration can be rapidly, flexibly, simply and clearly carried out, the accuracy of the score model configuration can be ensured by the score card model simulation test, the accuracy of the investment entity score calculation result and the like can be ensured on the accuracy of the model configuration, the method has the advantages that the financial market business is realized, the configuration is complicated and inflexible, the transition of science and technology personnel is seriously hindered, the development of the investment business is supported by updating an iterative grading model for the wind control business, the flexible operation and the quick response of the business is realized, the low-efficiency realization of the repeated response speed code writing of the very slow grading model is avoided, meanwhile, great convenience is brought to wind control business departments and even later maintenance personnel of the science and technology development departments for quickly carding the grading model and the grading business flow, the configuration deployment period of a grading card rule model is reduced, the later iterative maintenance and the increase of code quantity of the grading card rule model due to frequent grading model can be prevented, the confusion and the error of internal code logic caused by manual modification operation are prevented, the massive consumption of service hardware resources is reduced, and meanwhile, the unnecessary operation risks caused by the rapid increase of maintenance cost, the non-standardization of maintenance operation and the like are also reduced.

Drawings

FIG. 1 is a schematic diagram of a rule fact data model declaration flow;

FIG. 2 is a schematic diagram of a rule scoring card declaration process;

FIG. 3 is a schematic diagram of a rule decision flow declaration process;

FIG. 4 is a schematic diagram of a simulation test object declaration flow;

FIG. 5 is a schematic diagram of a deployment configuration object declaration flow;

Fig. 6 is a schematic diagram of a rule approval process.

Detailed Description

Hereinafter, embodiments of the present invention will be described in detail. While the invention will be described and illustrated in conjunction with certain specific embodiments, it will be understood that it is not intended to limit the invention to these embodiments alone. On the contrary, the invention is intended to cover modifications and equivalent arrangements included within the scope of the appended claims.

In addition, numerous specific details are set forth in the following description in order to provide a better illustration of the invention. It will be understood by those skilled in the art that the present invention may be practiced without these specific details.

The decision engine platform provides rule configuration and related functions for rule management. And managing the rules of the business system outside the business system, and managing the definition, deployment and execution of the rules.

The invention provides a decision engine-based investment body rating model configuration and calculation mode wind control decision method, which comprises gradually deriving rule variables contained in a rating model through data model definition, defining the rating grade of the rating rule through the variables, pushing data of the main body rating to a rating card calculation core component to calculate a main body rating result, innovatively improving the rating card model configuration characteristic design to the rating rule recognizable by service personnel in a mode of only machine recognition compared with the traditional configuration mode, designing a recognition mode capable of realizing man-machine sharing and common reading, changing the lag mode of the past rating card with poor readability, poor maintainability, low configuration flexibility, long iteration cycle time, high labor cost consumption and single calculation node, the decision engine-based investment main body rating model configuration and calculation mode wind control decision method makes a large number of grading card configuration verification on flexibility and visual design, obtains a method which can realize independent configuration of business personnel, is convenient for the business personnel to understand, has strong readability, high flexibility and short configuration period, iterates and updates the lightweight main body rating model configuration with low cost, has been applied to the technical design realization, has fully performed deep optimization on the situation of memory leakage and even downtime caused by JVM accumulation in a large transaction data volume scene, has changed Java objects into a backward scene of memory overflow caused by difficult recovery in a large data volume scene through a lightweight Map data result, realizes the innovative technical realization of lightweight loading of a decision engine grading card calculation component to factual data and a grading model library and ensures the stable operation of the memory,

The proposal application mainly comprises six modules of rule fact data model declaration, rule score card declaration, rule decision flow declaration, simulation test object declaration, deployment configuration object declaration and rule approval, wherein:

The rule fact data model declares that in a rule item, in order to describe the rule, a business system is firstly modeled, the data of the business system is refined, which can be used as the entry for describing the rule is confirmed, and the entries for describing the rule are integrated in a structured way, so that the rule fact data model in the rule item is formed. And a variable management function for providing the type of detailed data, input and output settings and the like. In the rule configuration process, certain indexes exist in the form of variables, and the variables need to be defined so as to be used in rule definition and configuration. And a constant management function for managing global constants and forming rules by using different constants according to service requirements. And the function management function is used for writing a function for custom functions in rules, and mainly performs complex business logic processing.

The rule scoring card declares that the rule scoring card is a scoring problem in the business logic. Different scoring parameters in the service scoring logic correspond to different attributes of the data object, different segmentation values of each scoring parameter correspond to different scores, the scores of all the parameters are summed up according to the actual incoming values, and finally the scoring result, and the proportion of the parameters in the total score can be set in the scoring model.

Rule decision flow declares that the role of rule decision flow has two roles in the overall rule base: firstly, arranging the execution sequence of rule components, and secondly, taking the rule components as an entry for rule calling. The same rule component can be mounted for multiple times in the decision flow, and the multiplexing effect is achieved for some general rules. The editing page of the decision flow component finishes editing the decision flow structure in a dragging mode, and the mode more intuitively reflects the decision flow logic and facilitates the writing of the decision flow.

The simulation test object declares that the simulation test object creates test objects in the form of a test task for a decision stream. After the test object is selected, the system analyzes the engine underlying code and adds the data object attributes involved to the data settings column. Other undisplayed fields in the data model may be assigned values according to default values. Setting a rule execution expectation value in a rule expectation column. Clicking to execute the test, the system submits the data to the engine bottom code execution environment for execution, and returns the execution result. The test task is saved for the next execution.

The deployment configuration object declares that the new rule needs to be developed and is correspondingly established, and the rule content is released to the execution server by executing the deployment task.

And before the rule decision flow in the rule model is deployed, rule configuration correctness verification and rechecking approval are required, so that the rule can be operated and executed, and the rule can be deployed in a server on line after the approval is passed.

Example 1

The invention provides a decision engine based on a financial wind control business rule scene, which comprises the following modules:

score facts data model declaration module:

Creating a scoring rule fact data model through a scoring rule fact data table or a real-time fact data structure, defining input variables and output variables through the model, creating rule constants and an aggregate calculation function, and providing a fact data set for subsequent configuration scoring cards;

rule scoring card declaration module:

Creating a rule scoring card, binding the rule scoring card with the rule fact data model created by the scoring fact data model declaration module, configuring the scoring card through input variables, output variables, rule constants and aggregation calculation functions defined by the rule fact data model created by the scoring fact data model declaration module, and providing scoring rules of fact data for subsequent rule decision flows;

rule decision flow declaration module:

Creating a rule decision flow, configuring the execution sequence of rule scoring cards created by each rule scoring card declaration module in the decision flow in a dragging mode to obtain a scoring rule set template of a fact data table or instant fact data, and providing a calling template for a subsequent simulation test object to arrange a series or parallel decision model in a flow mode;

simulation test object declaration module:

Creating a simulation test object, binding the rule decision flow created by the rule decision flow declaration module with the rule test object, inputting scoring rule indexes of all the fact data in the rule decision flow created by the rule scoring card declaration module, and obtaining scoring results of the fact data of all the rule scoring cards bound by the decision flow;

Sub-step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking;

deploying a configuration object declaration module:

Creating a deployment configuration object, binding a rule decision flow created by a rule decision flow declaration module with a server, submitting approval by a main body grading model, and finally obtaining an investment main body grading model to be approved;

Rule approval module:

And (3) using the model to approve the post role user, carrying out rule configuration correctness verification and recheck approval, and carrying out on-line deployment on the investment entity rating model by the approval.

In the above decision engine, the scoring fact data model declaration module specifically includes the following steps:

step S1.1: firstly loading a background grading fact data table in a business readable and understandable form at the front end, then creating a rule fact data model in a draggable and visual form at a front end page, defining input variables and output variables through the model after a required data model is obtained, creating a rule constant and an aggregation calculation function, and providing a fact data set for a follow-up configuration grading card;

Substep s1.1.1: defining a data model list, adding a predefined data model number and a predefined data model name after model data analysis to the data model list, and binding the fact data table with the data model names in the data model list according to a scoring model rule.

Substep S1.1.2: defining a variable management list, selecting a grading rule fact data model, dragging according to a data structure field corresponding to the data model to achieve standard definition of variable names and variable types, wherein the grading rule fact data model is mapped by a data table, variables are mapped by data table fields, the variable types are converted into Java data types which can be identified by Java language codes according to definition of the data table structure field types, and finally the variable names and the variable types are added into the variable management list;

Substep S1.1.3: defining a rule constant management list, adding constant dictionary codes and constant dictionary names to the constant management list to obtain a constant dictionary in the rule constant management list, and then adding constant value information to the constant dictionary, wherein the constant value information comprises constant actual values, constant display values and data types;

Substep S1.1.4: defining an aggregation calculation function management list, adding a function code, a function name and a return value type into the aggregation calculation function management list, adding function parameter content according to requirements, inputting the parameter code, the parameter name and the parameter type, writing a statement of a function, compiling the function statement after writing, and storing the function after compiling.

In the above decision engine, the implementation of the rule scoring card declaration module specifically includes the following steps:

Substep S2.1: defining a scoring card list, adding a decision code and a decision name to the scoring card list, and selecting a rule fact data model to bind with the decision code of the rule scoring card;

Substep S2.2: the configuration rule scoring card specifically comprises:

adding scoring rules to variable, constant and function configuration of the rule fact data model bound by the scoring card;

Selecting variables, functions and constants to assign to categories, major factors and sub-category factors in a scoring rule, wherein the categories, major factors and sub-category factors are in progressive subdivision relation according to category levels;

filling in index weight, index options, conditions and index scores;

Wherein the variables refer to input variables and output variables, the constants are regular constants, and the functions refer to an aggregation calculation function;

substep S2.3: testing rule scoring cards, filling in each input item data contained in the scoring card model, calculating scoring results, and checking rule scoring cards according to the scoring results;

Substep S2.4: and issuing rule scoring cards, and selecting the checked rule scoring cards from the scoring card list for issuing.

In the above decision engine, the implementation of the rule decision flow declaration module specifically includes the following steps:

step S3.1: defining a decision flow list, and adding a decision code and a decision name filled in the form into the decision flow list;

step S3.2: configuring a rule decision flow, editing and dragging the issued rule scoring card to a work display area of the rule decision flow, and sequentially linking different nodes by using connecting lines according to an executed logic flow to obtain the rule decision flow;

step S3.3: testing rule decision flow, filling in the input item data in each scoring index in the rule decision flow, calculating scoring results after filling in, and checking the rule decision flow according to the scoring results;

Step S3.4: and issuing the rule decision flow, and selecting the checked rule decision flow from the decision flow list for issuing.

In the above decision engine, the implementation of the simulation test object declaration module includes the following steps:

Sub-step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking.

In the above decision engine, the deployment configuration object declaration module implementation package includes the following steps:

Substep S5.1: creating a deployment configuration list, creating a deployment configuration object in the deployment configuration list, filling in a release name and a server URL address, binding a released rule decision stream, then performing batch simulation test, storing the deployment configuration object after the test is successful to obtain a deployment task of an investment main body rating model, realizing distributed hot deployment on a production mode through management of a deployment task version, supporting seamless rollback and upgrading deployment iteration of the version, and distributing the deployment object to a service code through the server URL address for calling;

sub-step S5.2: submitting the investment entity rating model for approval.

In the above decision engine, the rule approval module sub-step S6.1: and examining and approving the investment entity rating model in the task list to be examined and approved, carrying out examination and approval double-post rechecking on the correctness of the investment entity rating model, and carrying out online deployment on the investment entity rating model through a post point.

Example 2

The invention also provides a method for realizing the decision engine based on the financial wind control business rule scene, which comprises the following steps:

s1, scoring a fact data model declaration:

Creating a scoring rule fact data model through a scoring rule fact data table or a real-time fact data structure, defining input variables and output variables through the model, creating rule constants and an aggregate calculation function, and providing a fact data set for subsequent configuration scoring cards;

step S1.1: firstly loading a background grading fact data table in a business readable and understandable form at the front end, then creating a rule fact data model in a draggable and visual form at a front end page, defining input variables and output variables through the model after a required data model is obtained, creating a rule constant and an aggregation calculation function, and providing a fact data set for a follow-up configuration grading card;

Substep s1.1.1: defining a data model list, adding a predefined data model number and a predefined data model name after model data analysis to the data model list, and binding the fact data table with the data model names in the data model list according to a scoring model rule.

Substep S1.1.2: defining a variable management list, selecting a grading rule fact data model, dragging according to a data structure field corresponding to the data model to achieve standard definition of variable names and variable types, wherein the grading rule fact data model is mapped by a data table, variables are mapped by data table fields, variable types are converted into Java data types which can be identified by Java language codes according to definition of the data table structure field types, technical errors that grading model verification is not passed, java data types are inconsistent with database table data types and the like in the later stage of the grading model due to non-standardization caused by manually inputting model variables in the past are avoided, and finally the variable names and the variable types are added into the variable management list;

substep S1.1.3: defining a regular constant management list, adding constant dictionary codes and constant dictionary names to the constant management list to obtain a constant dictionary in the regular constant management list, and then adding constant value information to the constant dictionary, wherein the constant value information comprises constant actual values, constant display values and data types (Long, double, boolen, date, string).

Substep S1.1.4: defining an aggregation calculation function management list, adding a function code, a function name and a return value type (Long, double, boolen, date, string) into the aggregation calculation function management list, adding function parameter content according to requirements, inputting the parameter code, the parameter name and the parameter type, writing a sentence of a function, compiling the function sentence after writing, and storing the function after compiling;

s2, rule scoring card declares:

Creating a rule scoring card, binding the rule scoring card with the rule fact data model created in the step S1, configuring the scoring card through input variables, output variables, rule constants and aggregation calculation functions defined by the rule fact data model created in the step S1, and providing scoring rules of fact data for subsequent rule decision flows;

Substep S2.1: defining a scoring card list, adding a decision code and a decision name to the scoring card list, and selecting a rule fact data model to bind with the decision code of the rule scoring card;

Substep S2.2: the configuration rule scoring card specifically comprises:

adding scoring rules to variable, constant and function configuration of the rule fact data model bound by the scoring card;

Selecting variables, functions and constants to assign to categories, major factors and sub-category factors in a scoring rule, wherein the categories, major factors and sub-category factors are in progressive subdivision relation according to category levels;

fill in index weight, index options (A-Z), conditions (greater than, less than, equal to, among the collection); filling in index scores;

Wherein the variables refer to input variables and output variables, the constants are regular constants, and the functions refer to an aggregation calculation function;

substep S2.3: testing rule scoring cards, filling in each input item data contained in the scoring card model, calculating scoring results, and checking rule scoring cards according to the scoring results;

Substep S2.4: and issuing rule scoring cards, and selecting the checked rule scoring cards from the scoring card list for issuing.

S3, rule decision flow declaring:

Creating a rule decision flow, configuring the execution sequence of rule scoring cards created in each step S2 in the decision flow in a dragging mode to obtain a scoring rule set template of a fact data table or instant fact data, and providing a calling template for a subsequent simulation test object to arrange a series or parallel decision model in a flow mode;

step S3.1: defining a decision flow list, and adding a decision code and a decision name filled in the form into the decision flow list;

step S3.2: configuring a rule decision flow, editing and dragging the issued rule scoring card to a work display area of the rule decision flow, and sequentially linking different nodes by using connecting lines according to an executed logic flow to obtain the rule decision flow;

step S3.3: testing rule decision flow, filling in the input item data in each scoring index in the rule decision flow, calculating scoring results after filling in, and checking the rule decision flow according to the scoring results;

step S3.4: issuing rule decision flows, and selecting the checked rule decision flows from the decision flow list for issuing;

s4, declaring by a simulation test object:

Creating a simulation test object, binding the rule decision flow created in the step S3 with the rule test object, inputting scoring rule indexes of all the fact data in the rule decision flow created in the step S2, and obtaining scoring results of the fact data of all the rule scoring cards bound by the decision flow.

Sub-step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking.

S5, deploying a configuration object declaration:

Creating a deployment configuration object, binding the rule decision flow created in the step S3 with a server, submitting and approving the main body rating model, and finally obtaining an investment main body rating model to be approved;

Substep S5.1: creating a deployment configuration list, creating a deployment configuration object in the deployment configuration list, filling in a release name and a server URL address, binding a released rule decision stream, then performing batch simulation test, storing the deployment configuration object after the test is successful to obtain a deployment task of the investment main body rating model, realizing distributed hot deployment on a production mode through management of a deployment task version, supporting seamless rollback and upgrading deployment iteration of the version, and distributing the deployment object to a service code through the server URL address for calling.

Sub-step S5.2: submitting the investment entity rating model for approval.

S6, rule approval:

And (3) using the model to approve the post role user, carrying out rule configuration correctness verification and recheck approval, and carrying out on-line deployment on the investment entity rating model by the approval.

Substep S6.1: and examining and approving the investment entity rating model in the task list to be examined and approved, carrying out examination and approval double-post rechecking on the correctness of the investment entity rating model, and carrying out online deployment on the investment entity rating model through a post point.

Claims (3)

1. A decision engine system based on a financial wind control business rule scene is characterized by comprising the following modules:

score facts data model declaration module:

Creating a scoring rule fact data model through a scoring rule fact data table or a real-time fact data structure, defining input variables and output variables through the model, creating rule constants and an aggregate calculation function, and providing a fact data set for subsequent configuration scoring cards;

the scoring fact data model declaration module specifically comprises the following steps:

step S1.1: firstly loading a background grading fact data table in a business readable and understandable form at the front end, then creating a rule fact data model in a draggable and visual form at a front end page, defining input variables and output variables through the model after a required data model is obtained, creating a rule constant and an aggregation calculation function, and providing a fact data set for a follow-up configuration grading card;

Step s1.1.1: defining a data model list, adding a predefined data model number and a predefined data model name after model data analysis to the data model list, and binding the fact data table with the data model names in the data model list according to a grading card rule

Step S1.1.2: defining a variable management list, selecting a grading rule fact data model, dragging according to a data structure field corresponding to the data model to achieve standard definition of variable names and variable types, wherein the grading rule fact data model is mapped by a data table, variables are mapped by data table fields, the variable types are converted into Java data types which can be identified by Java language codes according to definition of the data table structure field types, and finally the variable names and the variable types are added into the variable management list;

step S1.1.3: defining a rule constant management list, adding constant dictionary codes and constant dictionary names to the constant management list to obtain a constant dictionary in the rule constant management list, and then adding constant value information to the constant dictionary, wherein the constant value information comprises constant actual values, constant display values and data types;

Step S1.1.4: defining an aggregation calculation function management list, adding a function code, a function name and a return value type into the aggregation calculation function management list, adding function parameter content according to requirements, inputting the parameter code, the parameter name and the parameter type, writing a sentence of a function, compiling the function sentence after writing, and storing the function after compiling;

rule scoring card declaration module:

Creating a rule scoring card, binding the rule scoring card with the rule fact data model created by the scoring fact data model declaration module, configuring the scoring card through input variables, output variables, rule constants and aggregation calculation functions defined by the rule fact data model created by the scoring fact data model declaration module, and providing scoring rules of fact data for subsequent rule decision flows;

rule decision flow declaration module:

Creating a rule decision flow, configuring the execution sequence of rule scoring cards created by each rule scoring card declaration module in the decision flow in a dragging mode to obtain a scoring rule fact data table or a scoring rule set template of a real-time fact data structure, and providing a calling template for a subsequent simulation test object to arrange a series or parallel decision model in a flow mode;

the implementation of the rule decision flow declaration module specifically comprises the following steps:

step S3.1: defining a decision flow list, and adding a decision code and a decision name filled in the form into the decision flow list;

step S3.2: configuring a rule decision flow, editing and dragging the issued rule scoring card to a work display area of the rule decision flow, and sequentially linking different nodes by using connecting lines according to an executed logic flow to obtain the rule decision flow;

step S3.3: testing rule decision flow, filling in the input item data in each scoring index in the rule decision flow, calculating scoring results after filling in, and checking the rule decision flow according to the scoring results;

step S3.4: issuing rule decision flows, and selecting the checked rule decision flows from the decision flow list for issuing;

simulation test object declaration module:

Creating a simulation test object, binding the rule decision flow created by the rule decision flow declaration module with the rule test object, inputting scoring rule indexes of all the fact data in the rule decision flow created by the rule scoring card declaration module, and obtaining scoring results of the fact data of all the rule scoring cards bound by the decision flow;

Step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking;

deploying a configuration object declaration module:

Creating a deployment configuration object, binding a rule decision flow created by a rule decision flow declaration module with a server, submitting approval by a main body grading model, and finally obtaining an investment main body grading model to be approved;

The deployment configuration object declaration module implementation package comprises the following steps:

step S5.1: creating a deployment configuration list, creating a deployment configuration object in the deployment configuration list, filling in a release name and a server URL address, binding a released rule decision stream, then performing batch simulation test, storing the deployment configuration object after the test is successful to obtain a deployment task of an investment main body rating model, realizing distributed hot deployment on a production mode through management of a deployment task version, supporting seamless rollback and upgrading deployment iteration of the version, and distributing the deployment object to a service code through the server URL address for calling;

Step S5.2: submitting the investment subject rating model for approval;

rule approval module:

Performing rule configuration correctness verification and rechecking approval by using a model approval post role user, wherein the approval is performed by online deployment of an investment main grading model;

the rule scoring card declaration module comprises the following steps:

step S2.1: defining a scoring card list, adding a decision code and a decision name to the scoring card list, and selecting a rule fact data model to bind with the decision code of the rule scoring card;

step S2.2: the configuration rule scoring card specifically comprises:

adding scoring rules to variable, constant and function configuration of the rule fact data model bound by the scoring card;

Selecting variables, functions and constants to assign to categories, major factors and sub-category factors in a scoring rule, wherein the categories, major factors and sub-category factors are in progressive subdivision relation according to category levels;

filling in index weight, index options, conditions and index scores;

Wherein the variables refer to input variables and output variables, the constants are regular constants, and the functions refer to an aggregation calculation function;

step S2.3: testing rule scoring cards, filling in each input item data contained in the scoring card model, calculating scoring results, and checking rule scoring cards according to the scoring results;

step S2.4: issuing rule scoring cards, wherein the rule scoring cards which are verified are selected from a scoring card list to be issued;

the deployment configuration object declaration module performs multi-node deployment and provides distributed computing services based on the micro-service architecture;

The decision engine is a pure memory engine, does not depend on a database, and completes the necessary conditions that the decision engine needs to load and calculate by utilizing the real-time data transmission and the structure formed by the main body score and the way that the scoring rule is loaded into the memory of the computer, and carries out high-speed operation circulation of the real-time data and the scoring model in the pure memory;

Step 5.1, providing an asynchronous scheduling framework, carrying out centralized allocation and access on a service code scheduling framework and a URL deployment calling interface to carry out asynchronous call processing on a scoring card rule configuration model by configuring an asynchronous Mq mode, generating scheduling information by a rating producer, calling a decision engine and simultaneously starting and loading an asynchronous decision model library after a consumer schedules a distributed decision engine to calculate asynchronous polling frame of micro service to consume the asynchronous information, and processing subsequent decision engine scheduling tasks;

Performing fault tolerance check in the process of invoking a decision engine process, realizing the configuration correctness of a grading card model through double-post rechecking, if the application of a decision engine micro-service of a certain deployment model is wrong, immediately matching a fault tolerance scheme configured in advance according to the error condition detected in advance through step S5.1, immediately switching the available decision engine micro-service and starting an emergency flow;

and S5.1, before and after each wind control model deployment, recording the state change of each model deployment through a heat deployment mode.

2. The decision engine system based on the financial wind control business rule scenario of claim 1, wherein the rule approval module step S6.1: and examining and approving the investment entity rating model in the task list to be examined and approved, carrying out examination and approval double-post rechecking on the correctness of the investment entity rating model, and carrying out online deployment on the investment entity rating model through a post point.

3. The implementation method of the decision engine based on the financial wind control business rule scene is characterized by comprising the following steps:

S1, scoring a fact data model declaration:

Creating a scoring rule fact data model through a scoring rule fact data table or a real-time fact data structure, defining input variables and output variables through the model, creating rule constants and an aggregate calculation function, and providing a fact data set for subsequent configuration scoring cards;

Step S1.1: when a scoring rule fact data model is created, a scoring rule fact data table is created, then a fact data model corresponding to the fact data table is created through business readability loading of the table on a page, then an input variable and an output variable are defined in a pull-down dragging mode through the model, a rule constant and an aggregation calculation function are created, and a fact data set is provided for subsequent configuration scoring cards;

Step s1.1.1: defining a data model list, adding a predefined data model number and a predefined data model name after model data analysis to the data model list, and binding the fact data table with the data model names in the data model list according to a grading card rule

Step S1.1.2: defining a variable management list, selecting a grading rule fact data model, dragging according to a data structure field corresponding to the data model to achieve standard definition of variable names and variable types, wherein the grading rule fact data model is mapped by a data table, variables are mapped by data table fields, the variable types are converted into Java data types which can be identified by Java language codes according to definition of the data table structure field types, and finally the variable names and the variable types are added into the variable management list;

step S1.1.3: defining a rule constant management list, adding constant dictionary codes and constant dictionary names to the constant management list to obtain a constant dictionary in the rule constant management list, and then adding constant value information to the constant dictionary, wherein the constant value information comprises constant actual values, constant display values and data types;

Step S1.1.4: defining an aggregation calculation function management list, adding a function code, a function name and a return value type into the aggregation calculation function management list, adding function parameter content according to requirements, inputting the parameter code, the parameter name and the parameter type, writing a sentence of a function, compiling the function sentence after writing, and storing the function after compiling;

S2, rule scoring card declares:

Creating a rule scoring card, binding the rule scoring card with the rule fact data model created in the step S1, configuring the scoring card through input variables, output variables, rule constants and aggregation calculation functions defined by the rule fact data model created in the step S1, and providing scoring rules of fact data for subsequent rule decision flows;

s3, rule decision flow declaring:

Creating a rule decision flow, configuring the execution sequence of rule scoring cards created in each step S2 in the decision flow in a dragging mode, obtaining a scoring rule fact data table or a scoring rule set template of a real-time fact data structure, providing a calling template for a subsequent simulation test object, and arranging a series or parallel decision model in a flow mode;

step S3.1: defining a decision flow list, and adding a decision code and a decision name filled in the form into the decision flow list;

step S3.2: configuring a rule decision flow, editing and dragging the issued rule scoring card to a work display area of the rule decision flow, and sequentially linking different nodes by using connecting lines according to an executed logic flow to obtain the rule decision flow;

step S3.3: testing rule decision flow, filling in the input item data in each scoring index in the rule decision flow, calculating scoring results after filling in, and checking the rule decision flow according to the scoring results;

step S3.4: issuing rule decision flows, and selecting the checked rule decision flows from the decision flow list for issuing;

s4, simulating test objects to declare:

Creating a simulation test object, binding the rule decision flow created in the step S3 with the rule test object, inputting scoring rule indexes of each piece of fact data in the rule decision flow created in the step S2, and obtaining scoring results of the fact data of each rule scoring card bound by the decision flow;

Step S4.1: defining a simulation test list, creating a simulation test object in the simulation test list, filling in the name of a score card model to be tested, binding a rule decision stream, filling in the input item data in each score index in the rule decision stream, calculating a score result through a simulation test after filling in, checking the rule decision stream according to the score result, and storing the simulation test object after checking;

s5, deploying a configuration object declaration:

Creating a deployment configuration object, binding the rule decision flow created in the step S3 with a server, submitting and approving the main body rating model, and finally obtaining an investment main body rating model to be approved;

step S5.1: creating a deployment configuration list, creating a deployment configuration object in the deployment configuration list, filling in a release name and a server URL address, binding a released rule decision stream, then performing batch simulation test, storing the deployment configuration object after the test is successful to obtain a deployment task of an investment main body rating model, realizing distributed hot deployment on a production mode through management of a deployment task version, supporting seamless rollback and upgrading deployment iteration of the version, and distributing the deployment object to a service code through the server URL address for calling;

Step S5.2: submitting the investment subject rating model for approval;

s6, rule approval:

Performing rule configuration correctness verification and rechecking approval by using a model approval post role user, wherein the approval is performed by online deployment of an investment main grading model;

s2, the rule scoring card declaration specifically comprises the following steps:

step S2.1: defining a scoring card list, adding a decision code and a decision name to the scoring card list, and selecting a rule fact data model to bind with the decision code of the rule scoring card;

step S2.2: the configuration rule scoring card specifically comprises:

adding scoring rules to variable, constant and function configuration of the rule fact data model bound by the scoring card;

Selecting variables, functions and constants to assign to categories, major factors and sub-category factors in a scoring rule, wherein the categories, major factors and sub-category factors are in progressive subdivision relation according to category levels;

filling in index weight, index options, conditions and index scores;

Wherein the variables refer to input variables and output variables, the constants are regular constants, and the functions refer to an aggregation calculation function;

step S2.3: testing rule scoring cards, filling in each input item data contained in the scoring card model, calculating scoring results, and checking rule scoring cards according to the scoring results;

step S2.4: issuing rule scoring cards, wherein the rule scoring cards which are verified are selected from a scoring card list to be issued;

step S5, performing multi-node deployment and providing distributed computing services based on a micro-service architecture;

The decision engine is a pure memory engine, does not depend on a database, and completes the necessary conditions that the decision engine needs to load and calculate by utilizing the real-time data transmission and the structure formed by the main body score and the way that the scoring rule is loaded into the memory of the computer, and carries out high-speed operation circulation of the real-time data and the scoring model in the pure memory;

Step S5.1, providing an asynchronous scheduling framework, carrying out centralized allocation and access on a service code scheduling framework and a URL deployment calling interface to carry out asynchronous call processing on a grading card rule configuration model through configuration of an asynchronous Mq mode, generating scheduling information by a grading producer, calling a decision engine to start and load an asynchronous decision model library simultaneously after a consumer scheduling distributed decision engine calculates asynchronous polling frames of micro services to consume the asynchronous information, and processing subsequent decision engine scheduling tasks;

Performing fault tolerance check in the process of invoking a decision engine process, realizing the configuration correctness of a grading card model through double-post rechecking, if the application of a decision engine micro-service of a certain deployment model is wrong, immediately matching a fault tolerance scheme configured in advance according to the error condition detected in advance through step S5.1, immediately switching the available decision engine micro-service and starting an emergency flow;

and S5.1, before and after each wind control model deployment, recording the state change of each model deployment through a heat deployment mode.