A method and device for determining a risk control feature blood relationship

Abstract

The present application relates to the field of risk control features, and provides a method and device for determining the blood relationship of risk control features, wherein the risk control features include data sources and derivative features of risk control, the method comprises: mining and cleaning the feature code to obtain a first dependency relationship between the risk control features, wherein the feature code reflects the reference relationship between the risk control features; determining the event steps corresponding to the risk control features in the first dependency relationship by using a first recursive stack algorithm, establishing a second dependency relationship between the risk control features and the corresponding event steps; and constructing the blood relationship of the risk control features from the first dependency relationship and the second dependency relationship. The present application is suitable for determining the full blood relationship in a complex data scenario with multiple data sources.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of risk control features, in particular to a method and device for determining the blood relationship of risk control features. BACKGROUND

[0002] With the continuous promotion and innovation of business, there are various forms of risk control feature data on the line, including data sources, standard features, and derivative features. The data sources include collected data of self-owned users, such as user short messages, address books, and device information. The standard features refer to features that are calculated based on one data source. The derivative features can be calculated based on the output results of multiple data sources and multiple standard features or derivative features. Various models and strategies can use feature data. Due to the large number and variety of feature data, the relationship of feature data used by various models and strategies is chaotic and difficult to maintain. In order to track the flow of data in a large number of features and data sources, view the source of upstream data and the destination of downstream data of risk data in the data system, and complete a crucial link in data governance, the establishment of blood relationship is indispensable.

[0003] In the prior art, a data blood relationship arrangement method is as follows: extracting regularized SQL statements from a script file containing SQL code, cleaning the SQL statements; performing lexical analysis on the regularized SQL statements, generating and traversing an abstract syntax tree to perform syntax analysis on the SQL statements; then performing blood relationship analysis on the SQL statements based on the abstract syntax tree to obtain blood relationship analysis results; and finally drawing a data blood relationship diagram of the SQL statements based on the blood relationship analysis results and performing visual display. This data blood relationship arrangement method has the following defects: only SQL is considered, and the blood relationship of data applications (such as risk control event steps) is not combed, which cannot meet the scene of exploring the dependency relationship between features.

[0004] In the prior art, another data blood relationship arrangement method is as follows: parsing SQL statements to form a syntax tree, storing the collected node data relationship in a neo4j graph database, and storing the necessary information collected in an HBase; introducing the information of heterogeneous data sources into the blood relationship system to form the blood relationship. This data blood relationship arrangement method has the following defects: only SQL is considered, and the blood relationship of data applications is not combed, which cannot timely discover data not in use. SUMMARY

[0005] The present application is used to solve the problem in the prior art that the determination of feature data blood relationship depends on SQL and is not applicable to data scenes with multiple data sources and complex blood relationships.

[0006] To solve the above technical problems, the first aspect of the present application provides a method for determining the blood relationship of risk control features, wherein the risk control features include data sources and derivative features of risk control, and the method comprises the following steps:

[0007] mining and cleaning the feature code to obtain a first dependency relationship between the risk control features, wherein the feature code reflects the reference relationship between the risk control features;

[0008] determining the event steps corresponding to the risk control features in the first dependency relationship by using a first recursive stack algorithm, and establishing a second dependency relationship between the risk control features and the corresponding event steps;

[0009] composing the blood relationship of risk control features from the first dependency relationship and the second dependency relationship.

[0010] As a further embodiment of the present application, the mining and cleaning of the feature code to obtain the first dependency relationship between the risk control features comprises the following steps:

[0011] extracting information from the feature code to obtain a plurality of vector edges, wherein each node on each vector edge is a risk control feature;

[0012] analyzing the vector edges by using a second recursive stack algorithm to obtain the first dependency relationship between the risk control features.

[0013] As a further embodiment of the present application, the analysis of the vector edges by using the second recursive stack algorithm to obtain the first dependency relationship between the risk control features comprises the following steps:

[0014] selecting one of the risk control features from all the risk control features of the vector edges as a recursive root node;

[0015] determining the parent nodes of the recursive root node according to all the vector edges, and storing them in a second parent stack;

[0016] determining the child nodes of the recursive root node according to all the vector edges, and storing them in a second child stack;

[0017] executing a second parent recursive process: taking out the top node of the second parent stack, judging whether the taken-out node is a data source, if yes, not processing, if not, querying the parent nodes of the taken-out node from all the vector edges, and storing them in the second parent stack, repeating the above process until the second parent stack is empty;

[0018] executing a second child recursive process: taking out the top node of the second child stack, querying the child nodes of the taken-out node from all the vector edges, and storing them in the second child stack, repeating the above process until the second child stack is empty;

[0019] The second parent recursive process and the second child recursive process record the dependency relationship between the parent nodes and the child nodes, which is the first dependency relationship between the risk control feature values;

[0020] The second parent recursive process and the second child recursive process are performed in parallel.

[0021] As a further embodiment herein, the first recursive stack algorithm is used to determine the event steps corresponding to the risk control features in the first dependency relationship, and the second dependency relationship between the risk control features and the corresponding event steps is established, including:

[0022] For each target event, the second dependency relationship between the risk control features and the corresponding target event steps in the target event is established according to the following process:

[0023] From the first dependency relationship, the risk control feature of a step in the target event is selected as the recursive root node;

[0024] According to the first dependency relationship, the parent node of the recursive root node is determined and stored in the first parent stack;

[0025] According to the first dependency relationship, the child node of the recursive root node is determined and stored in the first child stack;

[0026] The first parent recursive process is executed: the top node in the first parent stack is taken out, each online step corresponding to the taken-out node is scored, the target event step corresponding to the taken-out node is determined according to the scoring result, the parent node of the taken-out node is stored in the first parent stack, and the above process is repeated until the first parent stack is empty;

[0027] The first child recursive process is executed: the top node in the first child stack is taken out, each online step corresponding to the taken-out node is scored, and whether the taken-out node has a target event step is determined according to the scoring result, if it has, the child node of the taken-out node is stored in the first child stack, and the above process is repeated until the first child stack is empty;

[0028] The first parent recursive process and the first child recursive process record the second dependency relationship between the risk control features and the corresponding target event steps;

[0029] The first parent recursive process and the first child recursive process are performed in parallel.

[0030] As a further embodiment herein, the first parent recursive process scores each online step corresponding to the taken-out node, and determines the target event step corresponding to the taken-out node according to the scoring result, including:

[0031] The consistency of each online step corresponding to the taken-out node and the target event step corresponding to the child node of the taken-out node is determined;

[0032] score each online step corresponding to the take-out node according to consistency between each online step corresponding to the take-out node and a target event step corresponding to a child node of the take-out node;

[0033] score each online step corresponding to the take-out node according to consistency between each online step corresponding to the take-out node and a target event step corresponding to a child node of the take-out node;

[0034] As a further embodiment herein, the scoring of each online step corresponding to the take-out node in the first child recursive procedure, and the determination of whether the take-out node has a target event step according to the scoring result, include:

[0035] determining consistency between each online step corresponding to the take-out node and a target event step corresponding to a parent node of the take-out node;

[0036] score each online event step corresponding to the take-out node according to consistency between each online step corresponding to the take-out node and a target event step corresponding to a parent node of the take-out node;

[0037] if the score of each online step corresponding to the take-out node is a predetermined value, it is determined that the take-out node has no target event step;

[0038] if at least one of the scores of the online steps corresponding to the take-out node is greater than the predetermined value, the online step with the score greater than the predetermined value is taken as the target event step corresponding to the take-out node.

[0039] As a further embodiment herein, the consistency between the steps of the nodes includes: consistency between the first dependency relationship between the nodes and the order of the steps between the nodes, and consistency between the step names between the nodes.

[0040] As a further embodiment herein, it further includes:

[0041] For each risk control feature under each event step, it is determined whether the risk control feature has a derivative feature and whether the risk control feature is referenced in the policy file of the event step, and if the determination result is no, the risk control feature is offline processed in the event step.

[0042] As a further embodiment herein, it further includes:

[0043] all online in-use policy files are obtained, and each policy file includes at least one decision file;

[0044] according to the naming rule of the risk control feature, the in-use risk control features in each decision file are determined;

[0045] a third dependency relationship between the decision file and the in-use risk control features thereof is established.

[0046] As a further embodiment herein, the method further comprises: displaying the blood relationship of the risk control features in a tree diagram, wherein nodes in the tree diagram are the risk control features and event steps corresponding to the risk control features.

[0047] A second aspect of the present disclosure provides a device for determining blood relationship of risk control features, the risk control features including data sources of risk control and derived features, the device comprising:

[0048] a first relationship determining unit configured to mine and clean the feature code to obtain a first dependency relationship between the risk control features, the feature code reflecting reference relationships between the risk control features;

[0049] a second relationship determining unit configured to determine event steps corresponding to the risk control features in the first dependency relationship by using a first recursive stack algorithm, and establish a second dependency relationship between the risk control features and the corresponding event steps;

[0050] a combining unit configured to combine the first dependency relationship and the second dependency relationship to form the blood relationship of the risk control features.

[0051] A third aspect of the present disclosure provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method of any one of the preceding aspects when executing the computer program.

[0052] A fourth aspect of the present disclosure provides a computer storage medium having a computer program stored thereon, wherein the computer program is executable by a processor of a computer device to implement the method of any one of the preceding embodiments.

[0053] A fifth aspect of the present disclosure provides a computer program product, comprising a computer program executable by a processor of a computer device to implement the method of any one of the preceding embodiments.

[0054] The method and device for determining blood relationship of risk control features provided in the present disclosure, in addition to mining and cleaning the feature code to obtain a first dependency relationship between the risk control features, also determine event steps corresponding to the risk control features in the first dependency relationship by using a first recursive stack algorithm, establish a second dependency relationship between the risk control features and the corresponding event steps, and combine the first dependency relationship and the second dependency relationship to form the blood relationship of the risk control features. The present disclosure is applicable to complex blood relationship analysis of risk control features online in multiple event steps, and can improve the comprehensiveness of blood relationship analysis.

[0055] In order to make the above and other purposes, features and advantages of the present disclosure more obvious and easy to understand, the following preferred embodiments are described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description only constitute some embodiments of the present disclosure, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0057] Figure 1 A first flowchart of a method for determining a blood relationship of a risk control feature according to an embodiment of the present disclosure is shown;

[0058] Figure 2 A flowchart of a first dependency relationship determination process according to an embodiment of the present disclosure is shown;

[0059] Figure 3 A flowchart of a second recursive stack algorithm execution process according to an embodiment of the present disclosure is shown;

[0060] Figure 4 A flowchart of a second dependency relationship determination process according to an embodiment of the present disclosure is shown;

[0061] Figure 5 A flowchart of a node step scoring process according to an embodiment of the present disclosure is shown;

[0062] Figure 6 A flowchart of a node target event step determination process according to an embodiment of the present disclosure is shown;

[0063] Figure 7 A second flowchart of a method for determining a blood relationship of a risk control feature according to an embodiment of the present disclosure is shown;

[0064] Figure 8 A first structural diagram of a device for determining a blood relationship of a risk control feature according to an embodiment of the present disclosure is shown;

[0065] Figure 9 A second structural diagram of a device for determining a blood relationship of a risk control feature according to an embodiment of the present disclosure is shown;

[0066] Figure 10A A schematic diagram of a dependency relationship of a risk control feature according to an embodiment of the present disclosure is shown;

[0067] Figure 10B A schematic diagram of a dependency relationship of an event step between risk control features according to an embodiment of the present disclosure is shown;

[0068] Figure 11 A structural diagram of a computer device according to an embodiment of the present disclosure is shown.

[0069] Explanation of drawing symbols:

[0070] 801, a first relationship determination unit;

[0071] 802, a second relationship determination unit;

[0072] 803, combination unit;

[0073] 804, third relationship determining unit;

[0074] 805, fourth relationship determining unit;

[0075] 1102, computer device;

[0076] 1104, processor;

[0077] 1106, memory;

[0078] 1108, driving mechanism;

[0079] 1110, input / output module;

[0080] 1112, input device;

[0081] 1114, output device;

[0082] 1116, presentation device;

[0083] 1118, graphical user interface;

[0084] 1120, network interface;

[0085] 1122, communication link;

[0086] 1124, communication bus. DETAILED DESCRIPTION

[0087] The technical solutions in the embodiments herein will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments herein. Obviously, the described embodiments are only part of the embodiments herein, rather than all the embodiments. Based on the embodiments herein, all the other embodiments obtained by those skilled in the art without any creative work under the premise that the embodiments herein are within the scope of protection herein.

[0088] It should be noted that the terms "first", "second", and the like in the specification and claims herein and the above-described accompanying drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments herein described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, device, product, or apparatus that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to the process, method, product, or apparatus.

[0089] The specification provides method operation steps as described in the embodiments or flow charts, but can include more or less steps based on the implementation or the nature of the processing. The order of the steps listed in the embodiments is merely one of many possible execution sequences, and does not represent the only execution sequence. In actual system or device product implementation, the method steps can be executed in sequence or in parallel according to the method sequence shown in the embodiments or drawings.

[0090] It should be noted that the risk control feature blood relationship determination method and device provided in the present application can be used in the financial field, such as a banking system, and can also be used in any field other than the financial field. The application field of the risk control feature blood relationship determination method and device provided in the present application is not limited.

[0091] It should be noted that the risk control feature data (including but not limited to data for analysis, stored data, and displayed data) involved in the present application is data authorized by the user or data authorized by all parties.

[0092] In an embodiment, a risk control feature blood relationship determination method is provided to solve the problem that the existing feature blood relationship determination method is not comprehensive and is not suitable for the scenario of putting the risk control feature online in multiple event steps. Specifically, as shown in Figure 1 The risk control feature blood relationship determination method includes the following steps:

[0093] In step 101, the feature code is mined and cleaned to obtain a first dependency relationship between the risk control features. The feature code reflects the reference relationship between the risk control features.

[0094] In detail, the risk control features described in the present application include data sources and derived features of risk control, and can be divided into data sources, standard features, and derived features. The standard feature is a special derived feature. The data source is all data used for model training and decision-making accessed by risk control, including self-owned data (user data) and third-party data. The standard feature includes a feature (for example, a feature ending with sd) obtained by relying on a data source. The derived feature is a new derived feature (for example, a feature ending with dr) generated by relying on multiple data sources or existing derived features.

[0095] Each derived feature corresponds to a feature code, and the feature code generates some feature output items by referencing other risk control features. The feature output items are used for the strategy file to make decisions. The strategy file can be determined according to the business scenario, and the internal logic thereof is not limited herein. For example, the feature code A of the derived feature A references the output item of another derived feature B, and the derived feature A depends on the derived feature B. For another example, the feature code of the feature sms_report_sd references the short message data source SmsReportService, and can generate feature output items such as whether the user is expected, salary, and the like. Therefore, the sms_report_sd depends on the SmsReportService.

[0096] The first dependency relationship is the dependency relationship between the data source and the feature and the dependency relationship between the features. The dependency relationship can be displayed in the form of a tree diagram. The nodes in the tree diagram are the risk control features, and the directed connection edges in the tree diagram represent the dependency relationship between the nodes, which is from the parent node to the child node.

[0097] In step 102, the first recursive stack algorithm is used to determine the event steps corresponding to the risk control features in the first dependency relationship, and the second dependency relationship between the risk control features and the corresponding event steps is established.

[0098] In detail, the second dependency relationship is the dependency relationship between the data source and the event step and the dependency relationship between the features and the event step, which is displayed in the form of a tree diagram. The risk control features in the present disclosure are online in the event steps, and the running of the event steps is determined according to the strategy file. Specifically, the event step refers to the event step of risk control, which can be triggered by various businesses or various events triggered by risk control itself, such as user borrowing, user lending, user repayment, and item submission. The event runs in the form of a flowchart, in which the running nodes are named as A, B, C, and are called event steps. Each event step contains data nodes (mainly responsible for obtaining data sources), feature nodes (obtained from the data sources), and strategy nodes (calculate the strategy results according to the features, so as to determine the running of the subsequent process).

[0099] In step 103, the risk control feature blood relationship is constituted by the first dependency relationship and the second dependency relationship.

[0100] In this step, the first dependency relationship and the second dependency relationship can be integrated, and the event steps in the second dependency relationship are added to the tree diagram of the first dependency relationship as the attributes of the risk control features. The risk control feature blood relationship is finally obtained from the tree diagram. Through the second dependency relationship, the misjudgment of the risk feature event step blood relationship can be avoided.

[0101] In practice, the first and second dependencies are stored in the cache in JSON format to ensure that all upper-layer query applications only interact with the cache, and the cached data is updated in real time through MySQL binlog monitoring.

[0102] The data model in the cache is as follows:

[0103] {

[0104] "type":"Node type: serviceNode (data node), featureNode (risk feature node)",

[0105] "id":"Unique ID in the tree diagram",

[0106] "name": "Node name, including 1. data source 2. standard features 3. derived features",

[0107] "step":"Risk characteristics correspond to event steps",

[0108] "state": "The position of the current node relative to its parent node, where left indicates the parent node and right indicates the child node",

[0109] "level": The level corresponding to the node.

[0110] "children":Array[]"Next-level nodes, fields are similar to the current node",

[0111] }

[0112] In addition to mining and cleaning the feature codes to obtain the first dependency relationship between risk control features, this embodiment also uses the first recursive stack algorithm to determine the event steps corresponding to the risk control features in the first dependency relationship, and establishes the second dependency relationship between the risk control features and the corresponding event steps. The first dependency relationship and the second dependency relationship constitute the lineage relationship of risk control features. This paper is applicable to the complex lineage relationship sorting of risk control features that are online in multiple event steps, and can improve the comprehensiveness of lineage relationship sorting.

[0113] In one embodiment of this article, such as Figure 2 As shown, step 101 above performs feature code mining and cleaning to obtain the first dependency relationships between risk control features, including:

[0114] Step 201: Extract information from the feature code to obtain multiple vector edges, and the nodes on each vector edge are risk control features.

[0115] In detail, the feature code contains referenced risk control features, and the referenced risk control features are expressed in a preset rule, for example, surrounded by single quotation marks or double quotation marks, so that a plurality of vector edges representing reference relationships can be extracted, and each vector edge is a dependency relationship between risk control features.

[0116] In step 202, the second recursive stack algorithm is used to analyze the vector edges to obtain a first dependency relationship between the risk control features.

[0117] In this step, in order to improve the determination efficiency of the first dependency relationship, the second recursive stack algorithm determines the first dependency relationship between the risk control features in a parallel execution mode of the parent stack and the child stack.

[0118] Specifically, as shown in Figure 3 The specific execution process of step 202 includes:

[0119] In step 301, one of the risk control features in all vector edges is selected as a recursive root node.

[0120] In this step, one of the risk control features can be randomly selected as the recursive root node, and the specific recursive root node is not limited in this paper.

[0121] In step 302, the parent nodes of the recursive root node are determined according to all vector edges, and are stored in the second parent stack.

[0122] In this step, first, the vector edges pointing to the recursive root node are determined, and the non-recursive root nodes in these vector edges are the parent nodes. The node data in the second parent stack is obtained by using the first-in-first-out rule.

[0123] In step 303, the child nodes of the recursive root node are determined according to all vector edges, and are stored in the second child stack.

[0124] In this step, first, the vector edges including the recursive root node pointing to other nodes are determined, and the non-recursive root nodes in these vector edges are the child nodes. Similarly, the node data in the second child stack is also obtained by using the first-in-first-out rule.

[0125] In step 304, the second parent recursive process is executed: the top node in the second parent stack is taken out, it is judged whether the taken-out node is a data source, if yes, no processing is performed, if not, the parent nodes of the taken-out node are queried from all vector edges and are stored in the second parent stack, and the above process is repeated until the second parent stack is empty.

[0126] When the taken-out node is a data source, it means that the taken-out node no longer has leaf nodes, if not, the taken-out node and the edges between the taken-out node are recorded in the tree diagram.

[0127] Step 305, a second child recursive process is performed: the top node in the second child stack is taken out, the child nodes of the taken-out node are queried from all vector edges, and are stored into the second child stack, and the above process is repeated until the second child stack is empty.

[0128] The nodes in the second child stack are only features, not data sources, so the data sources are not allowed to be judged in the second child recursive process.

[0129] Step 306, the dependency relationship between the parent nodes and the child nodes in the second parent recursive process and the second child recursive process is recorded, and the dependency relationship is the first dependency relationship between the risk control features.

[0130] The second parent recursive process and the second child recursive process are performed in parallel.

[0131] In an embodiment of the present application, as shown in FIG. 2, the step 102 determines the event steps corresponding to the risk control features in the first dependency relationship by using the first recursive stack algorithm, and establishes the second dependency relationship between the risk control features and the corresponding event steps, which includes: Figure 4

[0132] For each target event, the second dependency relationship between the risk control features and the corresponding target event steps in the target event is established according to the following process.

[0133] The target event described herein can be set by the user according to the needs, and can also be each event in the online existing events. If the online existing events are 100, 100 second dependency relationships will be obtained.

[0134] Step 401, from the first dependency relationship, the risk control feature of a step in the target event is selected as the recursive root node.

[0135] In the implementation of this step, a step can be randomly determined or specified by a user, which is not limited herein.

[0136] Step 402, according to the first dependency relationship, the parent node of the recursive root node is determined and stored in the first parent stack.

[0137] In the implementation of this step, the node pointing to the recursive root node in the first dependency relationship is taken as the parent node of the recursive root node. The first parent stack uses the first-in first-out method to obtain data.

[0138] Step 403, according to the first dependency relationship, the child node of the recursive root node is determined and stored in the first child stack.

[0139] In the implementation of this step, the node executed by the recursive root node in the first dependency relationship is taken as the child node of the recursive root node. In addition, the first child stack uses the first-in first-out method to obtain data.​

[0140] Step 404, a first parent recursive process is executed: a top node in the first parent stack is taken out, each online step corresponding to the taken-out node is scored, a target event step corresponding to the taken-out node is determined according to a scoring result, a parent node of the taken-out node is stored in the first parent stack, and the above process is repeated until the first parent stack is empty.

[0141] In this step, each online step corresponding to the taken-out node can be determined according to the configuration information of the online feature in the step.

[0142] Step 405, a first child recursive process is executed: a top node in the first child stack is taken out, each online step corresponding to the taken-out node is scored, and whether a target event step exists is determined according to a scoring result, if the target event step exists, a child node of the taken-out node is stored in the first child stack, and the above process is repeated until the first child stack is empty.

[0143] Step 406, a second dependency relationship between a risk control feature and a corresponding target event step is recorded in the first parent recursive process and the first child recursive process.

[0144] The first parent recursive process and the first child recursive process are performed in parallel.

[0145] In this embodiment, the first parent stack and the second child stack are executed in parallel, so that the determination efficiency of the second dependency relationship can be improved.

[0146] In an embodiment, as shown in FIG. 4, the step 404 of scoring each online step corresponding to the taken-out node according to the scoring result to determine the target event step corresponding to the taken-out node includes: Figure 5

[0147] Step 501, consistency between each online step corresponding to the taken-out node and a target event step corresponding to a child node of the taken-out node is determined.

[0148] The evaluation parameters of the consistency between the steps of the nodes include consistency between a first dependency relationship between the nodes and a step sequence order between the nodes, and consistency between step names between the nodes.

[0149] Step 502, each online step corresponding to the taken-out node is scored according to the consistency between each online step corresponding to the taken-out node and a target event step corresponding to a child node of the taken-out node.

[0150] Specifically, if the evaluation parameter result of the consistency between the steps of the nodes is consistent, a first predetermined value (for example, 1) is added to the score, and if the evaluation result is inconsistent, a second predetermined value (for example, 0) is added to the score.

[0151] ​In one embodiment, a certain feature sms_dr at the lending event step Lend_B, the feature sms_dr depends on the feature sms_sd at all online event steps, i.e., the deposit event step A and the lending event step Lend_A.

[0152] The feature sms_sd is a parent node and is a node taken out from the parent stack, and correspondingly, the child node of the node sms_sd is the feature sms_dr. The step 501 determines the consistency, i.e., determines the consistency of the deposit event step A of the feature sms_sd and the lending event step Lend_B of the feature sms_dr, and determines the consistency of the deposit event step A of the feature sms_sd and the lending event step Lend_A of the feature sms_dr.

[0153] When the step 502 scores the deposit event step A of the feature sms_sd, because the feature sms_sd belongs to the event step Lend_B and the event step A does not have a dependency relationship and the step names are not consistent, the deposit event step A of the feature sms_sd is scored as zero.

[0154] When the step 502 scores the event step Lend_A of the feature sms_sd, because the feature sms_sd belongs to the event step Lend_B and the event step Lend_A has a dependency relationship (determined according to the order of A and B) and belongs to the same event, and the step names of the event step Lend_B and the event step Lend_A are consistent, the event step Lend_A of the feature sms_sd is scored as two (increased by one each time the parameter is consistent).

[0155] The step 503 takes the online step with the highest score as the target event step corresponding to the taken-out node.

[0156] Continuing the previous embodiment, the target event step of the feature sms_sd is finally determined as Lend_A. Thus, the following dependency relationship sms_sd-Lend_A<-sms_dr-Lend_B is obtained.

[0157] In one embodiment of the present application, as shown in Figure 6 The step 405 scores each online step corresponding to the taken-out node, and determines whether the taken-out node has a target event step according to the scoring result, which includes:

[0158] The step 601 determines the consistency of each online step corresponding to the taken-out node and the event step corresponding to the parent node of the taken-out node.

[0159] The evaluation parameters of the consistency of the steps between nodes include: the consistency of the first dependency relationship between nodes and the order of the steps between nodes, and the consistency of the step names between nodes.

[0160] Step 602, score each online event step corresponding to the taking-out node according to the consistency of each online step corresponding to the taking-out node and the event step corresponding to the parent node of the taking-out node.

[0161] When the first dependency relationship between nodes is consistent with the order of steps between nodes, the score is increased by a first predetermined value, and if not consistent, the score is a second predetermined value. When the step name consistency between nodes is consistent, the score is increased by a first predetermined value, and if not consistent, the score is a second predetermined value.

[0162] Step 603, if the score of each online step corresponding to the taking-out node is a predetermined value, it is determined that the taking-out node does not exist target event step.

[0163] Step 604, if at least one of the scores of the online steps corresponding to the taking-out node is greater than the predetermined value, the online step with the score greater than the predetermined value is taken as the target event step corresponding to the taking-out node.

[0164] In a specific embodiment, assuming that the child nodes of the feature sms_dr are sms_dr2 and sms_dr3, the current taking-out node is the child node sms_dr2, and the corresponding feature sms_dr is the parent node of the taking-out node sms_dr2. The next taking-out node is sms_dr3, and the corresponding feature sms_dr is the parent node of the taking-out node sms_dr3. Step 602 is to score the event steps of sms_dr2 / sms_dr3 according to the consistency of the event steps of the feature sms_dr.

[0165] Assuming that all event steps of the feature sms_dr2 are Repay_A, and the event steps of the feature sms_dr3 are Lend_C. The event step of the feature sms_dr is Lend_B.

[0166] Through step 602, process sms_dr2, query all steps, RepayA and A events are inconsistent with the loan event of sms_dr, so the score is zero, through step 603 to determine that sms_dr2 does not exist target event step. Process sms_dr3, the online steps Lend_C and Lend_B belong to the same event, the score is 1, through step 604 to determine that sms_dr3 exists target event step, establish the dependency relationship between the event step Lend_B of sms_dr and the event step Lend_C of sms_dr3.

[0167] In the implementation of the embodiments described herein, the second dependency relationship can be determined on the basis of the first dependency relationship. Specifically, in the initialization, the first dependency relationship is regarded as the second dependency relationship, and the event step of each feature is determined according to steps 401 to 406. In the execution of steps 401 to 406, the node corresponding to the current feature is replaced by the event step of the feature, and when the child node of the feature does not have an event step associated with the node event step (i.e., the score is the second predetermined value), the related edge is deleted in the second dependency relationship.

[0168] The embodiments can determine the second dependency relationship through the first dependency relationship, and achieve the effect of reducing branches.

[0169] In an embodiment described herein, for each risk control feature under an event step, it is determined whether the risk control feature has a derivative feature and whether the risk control feature is referenced in the strategy file of the event step. If the determination result is no, the risk control feature is determined as a useless feature, and the risk control feature is offline processed in the event step.

[0170] This step can timely find useless features and timely perform offline processing, thereby avoiding online empty running and consuming limited computing resources.

[0171] In an embodiment described herein, as shown in Figure 7 The method for determining the blood relationship of the risk control feature further includes:

[0172] In step 701, all online strategy files are obtained, and each strategy file includes at least one decision file.

[0173] Specifically, the decision file is stored in the strategy file in the form of code, including nearly ten types of files such as decision flow, decision script, rule set, and decision table. The decision file step includes python code and a rule set in json format.

[0174] In step 702, the in-use risk control features in each decision file are determined according to a risk control feature naming rule.

[0175] In some embodiments, the end of the risk control feature naming, for example, _sd or _dr, is used to distinguish the risk control features, and the risk control features in the decision file are identified according to the reference rule of the risk control features in the decision file, for example, the risk control features are surrounded by single quotation marks or double quotation marks. In the specific implementation, a set of rules for extracting features can be customized according to the risk control feature naming rule, and the in-use risk control features in the decision file are mined by using the rules.

[0176] In step 703, a third dependency relationship between the decision file and the in-use risk control features thereof is established.

[0177] In detail, the third dependency relationship is the dependency relationship between the feature->decision file. The third dependency relationship is established in accordance with the inverted index mode.

[0178] When the embodiment is implemented, the index is stored in the cache, and it is ensured that all upper-layer query applications only interact with the cache.

[0179] The data model of the cache is, for example:

[0180] {

[0181] "code":"feature name",

[0182] "strategyName":"strategy name",

[0183] "projectName":"project name",

[0184] "blockType":"decision file type",

[0185] "fileName":"decision file name",

[0186] "comment":"Chinese name of strategy",

[0187] "lines":"position of feature"

[0188] }。

[0189] In step 704, the third dependency relationship is added to the blood relationship of the risk control feature.

[0190] Further, the method further includes: establishing a dependency relationship between the metadata of the determined data source and the data source, and adding the dependency relationship to the blood relationship of the risk control feature.

[0191] The metadata described herein is a data table stored by the data source, including various online data tables, such as hive, hbase, mysql, and mongo tables.

[0192] The embodiment can further upgrade the blood relationship, and provide full-amount data for subsequent query data.

[0193] Based on the same inventive concept, the present disclosure also provides a device for determining a blood relationship of a risk control feature, as described in the following embodiments. Since the device for determining the blood relationship of the risk control feature has a similar problem-solving principle to the method for determining the blood relationship of the risk control feature, the implementation of the device for determining the blood relationship of the risk control feature can be referred to the method for determining the blood relationship of the risk control feature, and the repeated parts will not be described herein.

[0194] Specifically, as shown in Figure 8 the device for determining the blood relationship of the risk control feature includes:

[0195] The first relationship determining unit 801 is configured to mine and clean the feature code to obtain a first dependency relationship between the risk control features, and the feature code reflects a reference relationship between the risk control features.

[0196] The second relationship determining unit 802 is configured to determine an event step corresponding to the risk control features in the first dependency relationship by using a first recursive stack algorithm, and establish a second dependency relationship between the risk control features and the corresponding event step.

[0197] The combination unit 803 is configured to constitute a risk control feature blood relationship by using the first dependency relationship and the second dependency relationship.

[0198] The embodiment mines and cleans the feature code to obtain the first dependency relationship between the risk control features, and determines the event step corresponding to the risk control features in the first dependency relationship by using the first recursive stack algorithm, and establishes the second dependency relationship between the risk control features and the corresponding event step, and constitutes the risk control feature blood relationship by using the first dependency relationship and the second dependency relationship. The present embodiment is applicable to the complex blood relationship of the risk control features online in multiple event steps, and can improve the comprehensiveness of the blood relationship.

[0199] Further, as shown in Figure 9 The determining device of the risk control feature blood relationship further includes:

[0200] The third relationship determining unit 804 is configured to determine an in-use risk control feature from a decision file of a policy file, and establish a dependency relationship between the decision file and the in-use risk control feature.

[0201] The fourth relationship determining unit 805 is configured to establish a dependency relationship between a data source and metadata.

[0202] The combination unit 803 is further configured to add the third relationship determining unit 804 and the fourth relationship determining unit 805 to the blood relationship.

[0203] After the blood relationship is determined by the above embodiment, the user can query the dependency relationship between the risk control features, the dependency relationship between the event steps of the risk control features, and the dependency relationship between the policy files of the risk control features. The query conditions include: specifying the risk control features and the query content, and the query content includes: the dependency relationship between the risk control features, the dependency relationship between the event steps of the risk control features, and the dependency relationship between the policy files of the risk control features. The following three embodiments illustrate the query results.

[0204] Case one: query the dependency relationship between the risk control features

[0205] The query condition is a data source, for example, A. When querying in detail, A is taken as a starting point, and risk control features (i.e., standard features and derived features) dependent on the data source are queried one by one, and are displayed through a tree diagram. As shown in Figure 10A , the gray circles represent data sources, and the white circles represent risk control features dependent on the data sources. The left direction in the figure represents a parent node, and the right direction represents a subset node.

[0206] Case two: dependency relationship between event steps of risk control features

[0207] The query condition is a user-specified risk control feature B and an event step HH_tfix_tag_A. As shown in Figure 10B , the gray circles represent the query nodes of the risk control feature B. The left direction represents data sources dependent on the risk control feature, and the right direction represents derived features (child nodes) of the risk control feature in the HH_tfix_tag_A step and event steps of the derived features.

[0208] The user-specified risk control feature is taken as a query starting point, and the data source of the risk control feature is traced upward, and derived features in event steps dependent on the risk control feature are searched downward.

[0209] Case three: dependency relationship between strategy files of risk control features

[0210] The query condition is a user-specified risk control feature. The use of the risk control feature in decision files in all online strategy files is queried, and the query result is displayed in a table form.

[0211] The method and device for determining the blood relationship of risk control features provided in the present text can determine the blood relationship of the entire link of a data source- derived feature-decision file and metadata. The data flow of user queries is displayed through a visual chart, the risk control features are traceable, the data value is easy to evaluate, and the unused risk control features can be found in time and processed offline, so that the computing resources are saved.

[0212] In an embodiment of the present text, a computer device is further provided, as shown in Figure 11As shown, computer device 1102 can include one or more processors 1104, such as one or more central processing units (CPUs), each of which can implement one or more hardware threads. Computer device 1102 can also include any memory 1106 for storing any kind of information such as code, settings, data, etc. Without limitation, for example, memory 1106 can include any one or combination of: any type of RAM, any type of ROM, a flash memory device, a hard disk, an optical disk, etc. More generally, any memory can use any technology for storing information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory can represent a fixed or removable component of computer device 1102. In one case, computer device 1102 can perform any operation of the associated instructions when processor 1104 executes the associated instructions stored in any memory or combination of memories. Computer device 1102 can also include one or more drive mechanisms 1108 for interacting with any memory, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.

[0213] Computer device 1102 can also include an input / output module 1110 (I / O) for receiving various inputs (via input device 1112) and for providing various outputs (via output device 1114). One particular output mechanism can include a presentation device 1116 and associated graphical user interface 1118 (GUI). In other embodiments, input / output module 1110 (I / O), input device 1112, and output device 1114 can also not be included, just as a computer device in a network. Computer device 1102 can also include one or more network interfaces 1120 for exchanging data with other devices via one or more communication links 1122. One or more communication buses 1124 couple the above-described components together.

[0214] Communication links 1122 can be implemented in any manner, such as through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. Communication links 1122 can include any combination of hardwired links, wireless links, routers, gateway functionality, name servers, etc., governed by any protocol or combination of protocols.

[0215] The embodiments herein also provide a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, performs the steps of the above method.

[0216] The embodiments herein also provide a computer readable instruction, wherein when a processor executes the instruction, the program therein causes the processor to perform the method described in the foregoing embodiments.

[0217] It should be understood that the size of the sequence number of the above processes in various embodiments herein does not mean the order of execution, and the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments herein.

[0218] It should also be understood that in the embodiments herein, the term "and / or" is only a description of the association relationship of the associated objects, which means that there can be three relationships. For example, A and / or B can represent three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper generally represents that the front and rear associated objects are in an "or" relationship.

[0219] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software, or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in the above description in a general manner. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this paper.

[0220] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0221] In several embodiments provided herein, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, and the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interface, device or unit, and can also be electrical, mechanical or other form of connection.

[0222] The units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on multiple network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiments herein.

[0223] In addition, each functional unit in each embodiment herein can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0224] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions herein, essentially or in the form of a contribution to the prior art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment herein. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various other media that can store program codes.

[0225] The principles and implementation manners of the present application are described herein by using specific embodiments, and the above description of the embodiments is only for the purpose of helping to understand the methods and core ideas thereof; meanwhile, for those skilled in the art, according to the ideas herein, the specific implementation manners and application ranges can be changed, and the above description of the specification should not be understood as limiting the present application.

Claims

1. A method for determining the blood relationship of risk control characteristics, characterized in that, Risk control features include data sources and derived features, and the method includes: The feature codes are mined and cleaned to obtain the first dependency relationship between risk control features, and the feature codes reflect the reference relationship between risk control features; The first recursive stack algorithm is used to determine the event steps corresponding to the risk control features in the first dependency relationship, and a second dependency relationship is established between the risk control features and the corresponding event steps. The risk control feature lineage is constituted by the first dependency relationship and the second dependency relationship; Specifically, the process of determining the event steps corresponding to the risk control features in the first dependency relationship using the first recursive stack algorithm, and establishing the second dependency relationship between the risk control features and the corresponding event steps, includes: For each target event, establish a second dependency relationship between the risk control features under that target event and the corresponding target event steps according to the following process: From the first dependency relationship, select the risk control feature of a certain step in the target event as the recursive root node; Based on the first dependency relationship, determine the parent node of the recursive root node and store it in the first parent stack; Based on the first dependency relationship, determine the child nodes of the recursive root node and store them in the first child stack; Execute the first parent recursive process: Remove the top node from the first parent stack, score the steps on each line corresponding to the removed node, determine the target event step corresponding to the removed node based on the scoring results, store the parent node of the removed node into the first parent stack, and repeat the above process until the first parent stack is empty. Execute the first child recursive process: Remove the top node from the first child stack, score the steps on each line corresponding to the removed node, determine whether the removed node has a target event step based on the score results, and if it does, store the child nodes of the removed node into the first child stack. Repeat the above process until the first child stack is empty. The first parent recursive process and the first child recursive process record the second dependency relationship between risk control features and corresponding target event steps; The first parent recursive process and the first child recursive process are carried out in parallel.

2. The method as described in claim 1, characterized in that, After mining and cleaning the feature codes, the first dependencies between risk control features are obtained, including: Information is extracted from the feature code to obtain multiple vector edges, and the nodes on each vector edge are risk control features; The first dependency relationship between risk control features is obtained by analyzing the vector edges using the second recursive stack algorithm.

3. The method as described in claim 2, characterized in that, Using the second stack recursive algorithm, the vector edges are analyzed to obtain the first dependencies between risk control features, including: From all the risk control features of the vector edges, select one of them as the recursive root node; Based on all vector edges, determine the parent node of the recursive root node and store it in the second parent stack; Based on all vector edges, determine the child nodes of the recursive root node and store them in the second child stack; Execute the second parent recursive process: Remove the top node from the second parent stack, determine whether the removed node is the data source, if so, do not process it, if not, query the parent node of the removed node from all vector edges and store it in the second parent stack, repeat the above process until the second parent stack is empty. Execute the second sub-level recursive process: Remove the top node from the second sub-level stack, query the child nodes of the removed node from all vector edges, and store them into the second sub-level stack. Repeat the above process until the second sub-level stack is empty. The second parent recursive process and the second child recursive process record the dependency relationship between the parent node and the child node. This dependency relationship is the first dependency relationship between the risk control feature values. The second parent recursive process and the second child recursive process are carried out in parallel.

4. The method as described in claim 1, characterized in that, In the first parent recursive process, each online step corresponding to the extracted node is scored, and the target event steps corresponding to the extracted node are determined based on the scoring results, including: Determine the consistency between each online step corresponding to the extracted node and the target event steps corresponding to the child nodes of the extracted node. Each online step corresponding to the extracted node is scored based on its consistency with the target event steps corresponding to the child nodes of the extracted node. The online step with the highest score is selected as the target event step for the node to be extracted.

5. The method as described in claim 1, characterized in that, In the first-level recursive process, each online step corresponding to the retrieved node is scored. The steps to determine whether the retrieved node contains the target event based on the scoring results include: Determine the consistency between each online step corresponding to the extracted node and the target event step corresponding to the parent node of the extracted node. The online event steps corresponding to the extracted node are scored based on their consistency with the target event steps corresponding to the parent node of the extracted node. If the score of each online step corresponding to the extracted node is a predetermined value, then it is determined that the extracted node does not have a target event step. If at least one of the scores of the online steps corresponding to the extracted node is greater than a predetermined value, then the online steps with scores greater than the predetermined value will be taken as the target event steps corresponding to the extracted node.

6. The method as described in claim 4 or 5, characterized in that, Consistency between steps includes: consistency of the first dependency between nodes, consistency of the order of steps between nodes, and consistency of step names between nodes.

7. The method as described in claim 1, characterized in that, Also includes: For each risk control feature under each event step, determine whether the risk control feature has derived features and whether the risk control feature is referenced in the strategy file of the event step. If the determination result is no, the risk control feature is taken offline in the event step.

8. The method as described in claim 1, characterized in that, Also includes: Obtain all online policy files in use, with each policy file containing at least one decision file; Based on the naming rules for risk control features, the in-use risk control features in each decision document are determined. Establish a third dependency relationship between decision-making documents and their in-use risk control features; Add a third dependency relationship to the lineage of risk control features.

9. The method as described in claim 1, characterized in that, Also includes: The lineage of risk control features is displayed in a tree diagram, where nodes in the tree diagram represent risk control features and their corresponding event steps.

10. A device for determining the blood relationship of risk control characteristics, characterized in that, Risk control features include data sources and derived features, and the device includes: The first relationship determination unit is used to mine and clean the feature code to obtain the first dependency relationship between risk control features, wherein the feature code reflects the reference relationship between risk control features; The second relation determination unit is used to determine the event steps corresponding to the risk control features in the first dependency relationship using the first recursive stack algorithm, and to establish the second dependency relationship between the risk control features and the corresponding event steps. A combination unit is used to form a risk control feature lineage relationship from the first dependency relationship and the second dependency relationship; Specifically, the process of determining the event steps corresponding to the risk control features in the first dependency relationship using the first recursive stack algorithm, and establishing the second dependency relationship between the risk control features and the corresponding event steps, includes: For each target event, establish a second dependency relationship between the risk control features under that target event and the corresponding target event steps according to the following process: From the first dependency relationship, select the risk control feature of a certain step in the target event as the recursive root node; Based on the first dependency relationship, determine the parent node of the recursive root node and store it in the first parent stack; Based on the first dependency relationship, determine the child nodes of the recursive root node and store them in the first child stack; Execute the first parent recursive process: Remove the top node from the first parent stack, score the steps on each line corresponding to the removed node, determine the target event step corresponding to the removed node based on the scoring results, store the parent node of the removed node into the first parent stack, and repeat the above process until the first parent stack is empty. Execute the first child recursive process: Remove the top node from the first child stack, score the steps on each line corresponding to the removed node, determine whether the removed node has a target event step based on the score results, and if it does, store the child nodes of the removed node into the first child stack. Repeat the above process until the first child stack is empty. The first parent recursive process and the first child recursive process record the second dependency relationship between risk control features and corresponding target event steps; The first parent recursive process and the first child recursive process are carried out in parallel.

11. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 9.

12. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor of the computer device, it implements the method of any one of claims 1 to 9.

13. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor of a computer device, implements the method of any one of claims 1 to 9.

Images