CN113409138A - Risk control decision engine device - Google Patents

Abstract

The invention discloses a risk control decision engine device. The system comprises a variable library, a component library, a strategy, a decision laboratory, a decision engine and a strategy report, wherein the variable library comprises a plurality of variables used in different scenes; the component library is a set of rules, each rule is a logical operation consisting of a series of variables and operators, and the variables in the variable library are used for rule configuration in the rules; the strategy is a scene for wind control, the workflow is a flow chart of strategy configuration, the workflow is configured by using a rule component in a component library, and the workflow is a task configuration center of the strategy; the decision laboratory is a verification process for performing decision calculation on the configured strategy; the decision engine is a process for calculating the configured strategy; the strategy report is a wind control decision report obtained after the strategy is calculated by the decision engine. The invention has the beneficial effects that: and flexibly transforming and adjusting according to the actual situation, and evaluating the dimensions and the weights of various elements of the user risk.

Description

Risk control decision engine device

Technical Field

The invention relates to the technical field of internet correlation, in particular to a risk control decision engine device.

Background

Internet finance has been rapidly developed along with the development of internet technology. Unlike the traditional financial industry, most of the internet finance is done online, not on-site. Traditional risk control in the financial industry relies more on offline auditing and background surveys. The offline auditing and background investigation can take a long time, so the period of the traditional financial industry wind control business is long. In addition, the content of background survey in the traditional financial industry is relatively fixed, and the same set of background survey items is almost used for all users.

Internet finance is quite different from the driving finance industry, for example, the internet finance emphasizes timeliness, a user can hope to reply as soon as possible, preferably in real time, when applying online, and the traditional finance industry is obviously not suitable for long offline auditing and investigation. In addition, as online activities increase and big data technology develops, more information about users can be obtained by using big data, and the online activities of users can provide more comprehensive, reliable and dynamic user information than a single background survey. In addition, the dimensions and weights of various elements for evaluating the user risk are also continuously changed, and the dimensions and weights for evaluating the user risk also need to be flexibly changed and adjusted according to actual conditions.

Disclosure of Invention

The invention provides a risk control decision engine device which can be flexibly changed and adjusted according to actual conditions in order to overcome the defects in the prior art.

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

a risk control decision engine device comprises a variable library, a component library, a strategy, a decision laboratory, a decision engine and a strategy report, wherein the variable library comprises a plurality of variables used in different scenes; the component library is a set of rules, each rule is a logical operation consisting of a series of variables and operators, and the variables in the variable library are used for rule configuration in the rules; the strategy is a scene for wind control, the workflow is a flow chart of strategy configuration, the workflow is configured by using a rule component in a component library, and the workflow is a task configuration center of the strategy; the decision laboratory is a verification process for performing decision calculation on the configured strategy; the decision engine is a process for calculating the configured strategy; the strategy report is a wind control decision report obtained after the strategy is calculated by a decision engine.

The variable library in the device comprises a plurality of variables used in different scenes, the variables are the most basic data indexes, each rule in the component library is a logical operation formed by a series of variables and operators, the variables in the variable library are used in the rule for configuring the rule, one strategy corresponds to one wind control scene, for example, before loan, in loan, after loan and the like, rule components in the component library are used in a workflow for configuring, corresponding variables, rules and the workflow can be flexibly changed and adjusted according to actual conditions, the configured strategy is verified, then strategy output, strategy online management and strategy release history are carried out, and after the strategy is calculated by a decision engine, a wind control decision report is obtained, and dimensions and weights of various elements for evaluating user risks are obtained.

Preferably, the variable library is a general term of a set of a series of variables, wherein the set of the series of variables comprises an original variable, a derivative variable, an output variable and an intermediate variable; the method comprises the following steps that an original variable is a minimum unit configured by a rule, the original variable is a carrier of data required by the rule, and the original variable comprises a visual configuration function, list display, single variable detail display, editing and deleting functions; the intermediate variables can not be configured like the original variables, the intermediate variables are the results of the rules, one or more results can be generated after the rules are executed by decision, the results can be used as the variables of the next rule, and the generated variables are called the intermediate variables; the derived variables are configurable by pages, and the logic of the derived variables is a simple expression formed by combining original variables and simple operation logic, so that a defined variable is also used in the rule; when the output variable is used for strategy output configuration, the user-defined strategy output variable is used.

Preferably, the component library comprises rule components, each rule component is composed of a series of condition groups, and each condition group is a logic formed by combining a series of variables and simple operation logic; the rule configuration logic may be understood as: if a certain rule is operated, if the condition is satisfied, the result of the rule is the result of the variable expression after the condition of the rule is satisfied, otherwise, the result is the result of the expression of a certain default rule; the rule is the minimum unit of the configuration strategy, and the conditional expression consists of variables and operation logic symbols.

Preferably, the policies include a standard policy and a merchant policy, the standard policy is a public policy, the merchant policy is introduced and copied based on the standard policy to produce a new merchant policy, and the merchant policy is configured by manually adding the policy, introducing and copying from the standard policy, and copying from another merchant policy.

Preferably, the workflow is a logic sequence diagram executed by a decision engine, the workflow includes a start node, a task node, a condition judgment node, a workflow edge, and a task end node, the start node is a start point of policy execution and is also a primary configuration start point of the workflow, and only one start node is in the workflow; the task node is a place for configuring the components in the workflow, the task node can configure all the components in the component library, the number of the configurable components of the task node is not limited, and the number of the task nodes in the workflow is not limited; the condition judgment node is a branch of workflow execution and is used for carrying out condition judgment by configuring a rule component in a component library; the task nodes and the condition judgment nodes are associated through a connecting line, the connecting line is called a workflow edge, a configuration rule component is arranged on the workflow edge, and the workflow edge is executed by default; the end node is the tail end of the workflow, and the workflow execution to the end node indicates the end of the workflow.

Preferably, the policy includes policy output, the policy output is a self-defined and configurable item, a user configures the output item of the policy by self-defining according to the own scene, the output variables are configurable in the policy output item, one policy can configure any multiple policy output variables, the output variables have default values, the configuration item can be output according to the type of the variables, after the policy output is configured, the final output of the decision engine can be configured according to the configured output variables, and the policy report under the special scene is specified.

Preferably, the policy includes policy online management, specifically: the strategy can be modified after being configured, and the operation can record an operator, operate specific rules, operate strategy output items comprising the strategy, edit the workflow in the strategy, add, delete and edit a certain node in the work and add, modify and edit functions of rule components comprising the workflow.

Preferably, the policy includes a policy issuing history, specifically: the strategy is configured and verified, then the strategy is released and on-line, the strategy keeps a release history record, the release history record comprises an operation release person, version information of the release strategy and release strategy time information, and the strategies before and after release are recorded, if the release abnormity occurs, the strategy can be rolled back to the version before release.

Preferably, the decision laboratory is an important link for verifying the configured strategy, the strategy is to be verified after the strategy configuration is completed, online release operation can be performed only after the verification is passed, the strategy verification does not pass the adjustment of the strategy, and the decision laboratory is equivalent to the last set of defense line released online until the strategy is adjusted to reach the expected result.

Preferably, the decision engine is a core of the whole policy calculation, specifically: after the strategy is released online, the decision engine executes the configured workflow, and the result of each step of the decision engine is obtained through arithmetic logic calculation in the executing process according to the execution of one node by one node; the process of the decision engine is operated by one node through a logic operation rule along with the acquisition of variable data, so that the result of the decision engine is obtained, and a wind control decision report is finally formed.

The invention has the beneficial effects that: and the strategy is calculated by a decision engine to obtain a wind control decision report and evaluate the dimensionality and weight of various elements of the user risk.

Detailed Description

The invention is further described with reference to specific embodiments.

A risk control decision engine device comprises a variable base, a component base, a strategy, a decision laboratory, a decision engine and a strategy report,

the variable library comprises a plurality of variables used in different scenes; the variable library is a general term of a set of a series of variables, wherein the set comprises original variables, derivative variables, output variables, intermediate variables, model variables and the like; the method comprises the following steps that an original variable is a minimum unit configured by a rule, the original variable is a carrier of data required by the rule, and the original variable comprises functions of visual configuration, list display, single variable detail display, editing, deleting and the like; the intermediate variable is a more special variable in the variable library, configuration operation can be carried out unlike the original variable, the intermediate variable is the result of the rule, one or more results can be generated after the rule is executed by decision, the result can be used as the variable of the next rule, and the generated variable is called the intermediate variable; the derived variable is one of a variable library, is also special and can be configured by pages, and the logic of the derived variable is a simple expression formed by combining an original variable and simple operation logic such as (+, -, +/-), so as to define a variable which is also used in the rule; the output variable is one of the variable libraries used for the user-defined policy output when configuring the policy output.

The component library is a set of rules, each rule is a logical operation consisting of a series of variables and operators, and the variables in the variable library are used for rule configuration in the rules; the component library comprises rule components, each rule component is composed of a series of condition groups, and each condition group is logic formed by combining a series of variables and simple operation logic such as (+, -, +,/, &, =, | =, > =, <, =, <' =, and, or); the rule configuration logic may be understood as: if a certain rule is operated, if the condition is satisfied, the result of the rule is the result of the variable expression after the condition of the rule is satisfied, otherwise, the result is the result of the expression of a certain default rule; if expressed by using a simple language, this can be expressed as follows: the rule is the minimum unit of the configuration strategy, the condition expression is composed of variables and operation logic symbols, the variables can have default values, and the variables can also be custom constants. The rule component is divided into a common rule component, a mathematical rule component, a model rule component and the like, wherein the mathematical rule component is used in a strategy requiring a mathematical operation scene, the mathematical operation component mainly comprises some mathematical functions, and the mathematical functions are as follows: adding, subtracting, multiplying, dividing, reducing, enlarging, averaging, rounding down, rounding up, logarithms, exponentiation, and the like.

A policy is a scenario for wind control, a workflow is a flowchart for policy configuration, the workflow is configured by using rule components in a component library, the workflow is a task configuration center of the policy, and generally, one policy corresponds to one scenario for wind control, for example: before, during, after, etc.; the strategy comprises a standard strategy and a merchant strategy, the standard strategy is a public strategy, the merchant strategy refers and copies to produce a new merchant strategy on the basis of the standard strategy, the use of the standard strategy reduces the cost and time for creating the merchant strategy, the reusability of the standard strategy can be increased, and high-efficiency strategy configuration is realized; the configuration ways of the merchant strategy are manually adding the strategy, quoting from the standard strategy, copying and copying from other merchant strategies.

The workflow is a logic sequence diagram executed by a decision engine, the workflow comprises a starting node, a task node, a condition judgment node, a workflow edge and a task ending node, the starting node is a starting point of strategy execution and is also a primary configuration starting point of the workflow, and only one starting node is in the workflow; the task node is a place for configuring the components in the workflow, the task node can configure all the components in the component library, the number of the configurable components of the task node is not limited, and the number of the task nodes in the workflow is not limited; the condition judgment node is a branch of workflow execution and is used for carrying out condition judgment by configuring a rule component in a component library; the task nodes and the condition judgment nodes are associated through a connecting line, the connecting line is called a workflow edge, a configuration rule component is arranged on the workflow edge, and the workflow edge is executed by default; the end node is the tail end of the workflow, and the workflow execution to the end node indicates the end of the workflow.

The strategy comprises strategy output, the strategy output is a self-defined and configurable item, a user carries out self-defined configuration on the strategy output item according to own scenes, output variables can be configured in the strategy output item, one strategy can configure any multiple strategy output variables, default values exist in the output variables, the configuration item can be output according to the types of the variables, after the strategy output is configured, the output of a final decision engine can be configured according to the configured output variables, and a strategy report under a special scene is specified.

The strategy comprises strategy online management, specifically comprising: the strategy can be modified after being configured, and the operation can record operators, operate specific rules, operate strategy output items comprising the strategy, edit the workflow in the strategy, add, delete and edit a certain node in the work and the like, and include the functions of adding, modifying and editing rule components in the workflow.

The strategy comprises a strategy release history, and specifically comprises the following steps: the strategy is configured and verified, then the strategy is released and on-line, the strategy keeps a release history record, the release history record comprises information such as an operation release person, version information of the release strategy, release strategy time and the like, and the strategies before and after release are recorded, if the release abnormity occurs, the strategy can be rolled back to the version before release.

The decision laboratory is a verification process for performing decision calculation on the configured strategy; the decision laboratory is an important link for verifying the configured strategy, the strategy is verified after the strategy configuration is completed, online release operation can be carried out only after the strategy passes the verification, the strategy verification does not pass the adjustment of the strategy, and the decision laboratory is equivalent to the last set of online release defense line until the strategy is adjusted to reach the expected result.

The decision engine is a process for calculating the configured strategy; the strategy report is a wind control decision report obtained after the strategy is calculated by the decision engine. The decision engine is the core of the whole strategy calculation, and specifically comprises the following steps: after the strategy is released online, the decision engine executes the configured workflow, and the result of each step of the decision engine is obtained through arithmetic logic calculation in the executing process according to the execution of one node by one node; the process of the decision engine is operated by one node through a logic operation rule along with the acquisition of variable data, so that the result of the decision engine is obtained, and a wind control decision report is finally formed.

Rules are stored in the rule component library, and each rule corresponds to the logic operation of a variable. Variables for rule operations are stored in the variable library. The decision engine is a rule operation engine, each strategy corresponds to one application scene, and for each strategy, the decision engine uses the rules in the component library and the variables in the variable library to perform operation. The whole device is also provided with an input/output interface for inputting and/or outputting data, the input/output interface receives user data and application scene data, a decision engine generates a strategy according to the application scene data, and rule settlement is carried out on the user under the strategy according to the user data.

Variables may include functions, input parameters, output parameters, temporary parameters. Conditional decisions can be implemented by logical operations, such as the introduction of operators and operators. The conclusion can be reached by the condition judgment of the variable, and the conclusion is the output of the rule. The rule operation functions provided by the component library include: the system comprises a rule editing function, a rule labeling function, a rule copying function, a rule stopping and starting function, a rule pushing strategy function, a rule checking and testing function and a rule and strategy correlation function. The variable library comprises a function library and a parameter library, the function library stores functions and provides function operation functions, the functions are used for rule operation, and the function operation functions provided by the function library comprise: a function adding function and a function editing function; the parameter library stores parameters and provides parameter operation functions, the parameters are used for rule operation, and the parameter operation functions provided by the parameter library comprise: the system comprises a parameter editing function, a draft function, a parameter importing and exporting function, a parameter and rule associating function and a parameter and strategy associating function. The function and the parameter are variables called by the rule, the function called by the rule can be from a function library, and the input parameter, the output parameter and the temporary parameter called by the rule can be from a parameter library.

The decision engine is a rule settlement engine and is realized by adopting a Drools rule engine. The decision engine is matched with the strategies for use, specifically, each strategy corresponds to one application scene, and for each strategy, the decision engine performs operation by using rules in the component library and variables in the variable library. The policy may invoke one or several rules according to the requirements of the reference scenario, where each rule is to make a conditional judgment on its respective variable and then to reach a conclusion. Each rule calls functions and parameters needed to be used in a function library and a parameter library respectively, judgment is carried out according to the conditions of each rule, and then a conclusion is obtained. And (4) carrying out rule settlement on each rule and conclusion by a decision engine (Drools rule engine) to obtain the final output of the strategy. Because the decision engine (Drools rule engine) can independently use the rules in the component library and the variables in the variable library to operate for each strategy, the strategy is very flexible to deploy and can be flexibly changed or adjusted according to different application scenes. Moreover, when needed, in the same strategy, different rules can be settled by calling different rules and/or variables for different users.

The decision engine also comprises a strategy base, the strategy base stores strategies and provides a strategy operation function, the strategies are generated according to the existing application scene data, each strategy corresponds to one application scene, and the decision engine performs rule settlement for each strategy. The policy operation functions provided by the policy repository include: the system comprises a strategy editing function, a strategy stopping and starting function, a strategy unloading and deploying function, a strategy refreshing function, a strategy shallow copy function, a strategy deep copy function and a strategy testing function. The strategy library is used for accelerating the real-time deployment capability of the strategy. For a policy which is frequently used and has better universality, the policy can be saved in a policy base after the decision engine completes rule settlement. In the subsequent use process, if an application scene matched with the existing strategy is encountered, the existing strategy can be directly called from the strategy library without recalculation, and the deployment time can be shortened. The operation function provided by the strategy base can modify and adjust the strategy stored in the strategy base.

The input-output interface performs input and/or output of data. The input and output interface receives the user data and the application scene data, the decision engine generates a strategy according to the application scene data, and rules settlement is carried out on the user according to the strategy and the user data. And when the application scene data received by the input and output interface is matched with the existing application scene data, directly calling the stored strategy from the strategy library. In some applications, different rules and/or variables may be used for different users under the same policy.

The decision engine of the present invention: and establishing mapping and association relations among the function library, the parameter library, the component library and the strategy library. The policy calls and settles the rules to settle the rules for the user under the policy. And the rules further call functions and parameters to realize the logic judgment of specific variables. After the association and mapping are established among the function library, the parameter library, the component library and the strategy library, the operation of one library is prompted in the rest libraries. For example, if a function in the library is modified, a rule in the component library associated with the function (calling the function) is prompted, and similarly, a policy in the policy library associated with the rule (calling the rule) is prompted. As another example, if a rule is adjusted, a policy associated with (invoking) the rule in the policy repository may be prompted.

The decision engine also includes a workflow engine, such as jBPM. A workflow engine such as jBPM has a conditional logic judgment function. Because the Drools rule engine does not have the condition logic (if-then), in order to realize the complete and comprehensive logic judgment of the rule, a workflow engine (jBPM) with a condition logic judgment function is introduced to assist the decision engine (Drools) to operate.

A specific application scenario is presented below to aid in understanding the decision engine of the present invention. In the field of internet finance, the most common application scenario is to apply for credit, for example, user a is applying for credit. The input/output interface receives application scenario data and user data, the application scenario data is credit application credit, and the user data is identity information of the user a. The decision engine (Drools rules engine) would then generate a policy from the application credit, which corresponds to the application scenario in which the credit is applied, such as may be named the application credit policy. The credit application credit policy is selectively invoked in the following rules: the system comprises a user identity identification rule, a basic information acquisition rule, a background investigation rule, a behavior investigation rule, a peer risk verification rule and the like. And the user identity identification rule, the basic information acquisition rule, the background investigation rule, the behavior investigation rule, the peer risk verification rule and other rules can call the functions or parameters required by the user respectively, and the user can calculate by utilizing respective logic judgment operation. It should be noted that the credit extension policy may choose to invoke different combinations of rules for different users. For example, for the user a, the user identity recognition rule and the basic information acquisition rule are called first to obtain the user identity and the basic information. For example, if the user identity information shows that the user a is a young person and the work history and credit records are few, the credit application policy may select to invoke behavior investigation rules, peer risk verification rules, and the like at this time, and perform rule calculation according to the information of the user a, such as network behavior, social range, credit records of other internet financial enterprises, and the like. And after the operation is finished, storing the credit application credit strategy into a strategy library. User B then also applies for credit. The input and output interface receives application scene data and user data, the application scene data is also credit application credit, and the user data is identity information of a user B. Because the credit application credit extension strategy is stored in the strategy library and is matched with the current application scene data, the credit application credit extension strategy is directly called from the strategy library to realize the rapid deployment of the strategy. The user identity information shows that the user B is a middle-aged person and has rich work records and credit records, the credit application strategy selects to call background investigation rules, peer risk verification rules and the like at the moment, and rule calculation is performed through the information of the background investigation of the user B, the credit records of other financial enterprises (including Internet financial enterprises) and the like. Therefore, different users can apply different rules to judge in the same strategy based on different user data.

Claims (10)

1. A risk control decision engine device is characterized by comprising a variable library, a component library, a strategy, a decision laboratory, a decision engine and a strategy report, wherein the variable library comprises a plurality of variables used in different scenes; the component library is a set of rules, each rule is a logical operation consisting of a series of variables and operators, and the variables in the variable library are used for rule configuration in the rules; the strategy is a scene for wind control, the workflow is a flow chart of strategy configuration, the workflow is configured by using a rule component in a component library, and the workflow is a task configuration center of the strategy; the decision laboratory is a verification process for performing decision calculation on the configured strategy; the decision engine is a process for calculating the configured strategy; the strategy report is a wind control decision report obtained after the strategy is calculated by a decision engine.

2. The risk control decision engine of claim 1, wherein the variable library is a collection of a series of variables, including original variables, derived variables, output variables, intermediate variables; the method comprises the following steps that an original variable is a minimum unit configured by a rule, the original variable is a carrier of data required by the rule, and the original variable comprises a visual configuration function, list display, single variable detail display, editing and deleting functions; the intermediate variables can not be configured like the original variables, the intermediate variables are the results of the rules, one or more results can be generated after the rules are executed by decision, the results can be used as the variables of the next rule, and the generated variables are called the intermediate variables; the derived variables are configurable by pages, and the logic of the derived variables is a simple expression formed by combining original variables and simple operation logic, so that a defined variable is also used in the rule; when the output variable is used for strategy output configuration, the user-defined strategy output variable is used.

3. The risk control decision engine of claim 1 wherein the library of components comprises rule components, a rule component comprising a set of conditions, each condition set comprising a logic of a set of variables plus simple arithmetic logic; the rule configuration logic may be understood as: if a certain rule is operated, if the condition is satisfied, the result of the rule is the result of the variable expression after the condition of the rule is satisfied, otherwise, the result is the result of the expression of a certain default rule; the rule is the minimum unit of the configuration strategy, and the conditional expression consists of variables and operation logic symbols.

4. The risk control decision engine of claim 1, wherein the policy comprises a standard policy and a merchant policy, the standard policy is a public policy, the merchant policy is referenced and copied based on the standard policy to produce a new merchant policy, and the merchant policy is configured by manually adding the policy, referencing from the standard policy, copying, and copying from another merchant policy.

5. The risk control decision engine apparatus according to claim 1, wherein the workflow is a logic sequence diagram executed by the decision engine, the workflow includes a content start node, a task node, a condition judgment node, a workflow edge, and a task end node, the start node is a start point of policy execution and is also a primary configuration start point of the workflow, and the start node is only one in the workflow; the task node is a place for configuring the components in the workflow, the task node can configure all the components in the component library, the number of the configurable components of the task node is not limited, and the number of the task nodes in the workflow is not limited; the condition judgment node is a branch of workflow execution and is used for carrying out condition judgment by configuring a rule component in a component library; the task nodes and the condition judgment nodes are associated through a connecting line, the connecting line is called a workflow edge, a configuration rule component is arranged on the workflow edge, and the workflow edge is executed by default; the end node is the tail end of the workflow, and the workflow execution to the end node indicates the end of the workflow.

6. A risk control decision engine apparatus as claimed in claim 1 or 5, wherein the policy includes policy output, the policy output is a customizable and configurable item, the user configures the policy output according to his/her own scenario, the policy output item has configurable output variables, a policy can configure any number of policy output variables, the output variables have default values, the configuration item can be output according to the type of the variables, after the policy output is configured, the final decision engine output can be configured according to the configured output variables, and the policy report under a special scenario is specified.

7. A risk control decision engine apparatus according to claim 1 or 5, wherein the policy comprises policy online management, specifically: the strategy can be modified after being configured, and the operation can record an operator, operate specific rules, operate strategy output items comprising the strategy, edit the workflow in the strategy, add, delete and edit a certain node in the work and add, modify and edit functions of rule components comprising the workflow.

8. The risk control decision engine apparatus according to claim 7, wherein the policy includes a policy issuance history, specifically: the strategy is configured and verified, then the strategy is released and on-line, the strategy keeps a release history record, the release history record comprises an operation release person, version information of the release strategy and release strategy time information, and the strategies before and after release are recorded, if the release abnormity occurs, the strategy can be rolled back to the version before release.

9. The risk control decision engine apparatus according to claim 1, wherein the decision laboratory is an important link for verifying the configured policy, the policy is to be verified each time the policy configuration is completed, the online release operation can be performed only after the verification is passed, the policy verification does not pass the adjustment of the policy, until the policy is adjusted to reach the expected result, and the decision laboratory is equivalent to the last set of defense line released online.

10. The risk control decision engine apparatus according to claim 7, wherein the decision engine is a core of the whole policy computation, and specifically comprises: after the strategy is released online, the decision engine executes the configured workflow, and the result of each step of the decision engine is obtained through arithmetic logic calculation in the executing process according to the execution of one node by one node; the process of the strategy engine is along with the acquisition of variable data, and the result of the decision engine is obtained through the operation of a node through a logic operation rule, and finally a wind control decision report is formed.