WO2020037918A1 - 基于预测模型的风险控制策略的确定方法及相关装置 - Google Patents

基于预测模型的风险控制策略的确定方法及相关装置 Download PDF

Info

Publication number
WO2020037918A1
WO2020037918A1 PCT/CN2018/123513 CN2018123513W WO2020037918A1 WO 2020037918 A1 WO2020037918 A1 WO 2020037918A1 CN 2018123513 W CN2018123513 W CN 2018123513W WO 2020037918 A1 WO2020037918 A1 WO 2020037918A1
Authority
WO
WIPO (PCT)
Prior art keywords
risk
target
risk decision
user
decision rule
Prior art date
Application number
PCT/CN2018/123513
Other languages
English (en)
French (fr)
Inventor
王智浩
杨冬艳
刘玉洁
董晓琼
曹洋
秦威
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2020037918A1 publication Critical patent/WO2020037918A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"

Definitions

  • the present application relates to the field of communication technologies, and in particular, to a method for determining a risk control strategy based on a prediction model and a related device.
  • Risk control is that risk managers use various measures and / or methods to eliminate or reduce the possibility of risk events, or risk managers use various measures and / or methods to reduce the losses caused by risk events.
  • Risk control can also be called risk control, and the measures and / or methods used in the risk control process can also be called risk control strategies.
  • Risk decision-making refers to the process of selecting two or more risk control strategies under the influence of various uncertain factors. Due to the existence of uncertain factors, the profit and loss value brought about by the implementation of the risk control strategy cannot be determined in advance. Therefore, certain rules are required as the basis for selecting the risk control strategy in the risk decision process. In the face of multiple uncertain factors, the corresponding risk control strategy is selected from multiple risk control strategies for risk control based on the rule. The rules on which risk decisions are based are called risk decision rules.
  • the risk control system of risk control is mainly based on the risk decision rules made by experts to select risk control strategies.
  • the formulation of risk decision rules completely depends on the subjective experience of experts to determine and formulate.
  • the style decision rules are single and lacking.
  • Theoretical support and data basis are more arbitrary.
  • the randomness of risk decision rules based on experts also makes the risk control strategies determined based on risk decision rules have poor reliability and a small scope of application.
  • the embodiments of the present application provide a method and a related device for determining a risk control strategy based on a prediction model, which can enhance the flexibility of the determination method of the risk control strategy, improve the reliability of the risk control strategy, and have higher applicability.
  • an embodiment of the present application provides a method for determining a risk control strategy based on a prediction model, including:
  • the target risk decision rule generation model corresponding to the first user data is generated based on the target risk decision rule generation model.
  • the target risk decision rule generation model is trained based on the sample data corresponding to the target business type.
  • the sample data includes at least the first rule sample.
  • Data and second rule sample data the first rule sample data includes a first risk decision rule and its corresponding first sample user data, and the second rule sample data includes a second risk decision rule and its corresponding first sample data Two sample user data;
  • a target risk control strategy for performing risk control on the target business is determined according to the first user data and the target risk decision rule.
  • an embodiment of the present application provides a device for determining a risk control strategy based on a prediction model.
  • the device for determining includes:
  • An obtaining unit configured to obtain first user data and determine a target service associated with the first user data
  • a model determining unit configured to determine a target risk decision rule generation model associated with the target service according to the target service type of the target service obtained by the obtaining unit;
  • a rule generation unit configured to generate a target risk decision rule corresponding to the first user data obtained by the obtaining unit based on the target risk decision rule generation model determined by the model determination unit, and the target risk decision rule generation model is based on the target business type
  • the corresponding sample data is obtained through training.
  • the above sample data includes at least first rule sample data and second rule sample data.
  • the first rule sample data includes the first risk decision rule and its corresponding first sample user data.
  • the second rule sample data includes a second risk decision rule and its corresponding second sample user data;
  • the policy determining unit is configured to determine a target risk control strategy for performing risk control on the target service according to the first user data obtained by the obtaining unit and the target risk decision rule determined by the rule generating unit.
  • an embodiment of the present application provides a terminal device.
  • the terminal device includes a processor and a memory, and the processor and the memory are connected to each other.
  • the memory is configured to store a computer program that supports the terminal device to execute the method provided in the first aspect and / or any possible implementation manner of the first aspect.
  • the computer program includes program instructions, and the processor is configured to call the foregoing.
  • a program instruction executes the first aspect and / or the method provided in any possible implementation manner of the first aspect.
  • an embodiment of the present application provides a computer-readable storage medium.
  • the computer-readable storage medium stores a computer program, where the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the instructions.
  • the reliability of generating risk decision rules based on the target risk decision rule generation model is stronger, the determination method of the risk control strategy is more flexible, and the reliability of the risk control strategy generated by the target risk decision rules can be enhanced. Thereby, the effectiveness of risk control of the target business can be improved, and the applicability is higher.
  • FIG. 1 is a schematic flowchart of a method for determining a risk control strategy based on a prediction model according to an embodiment of the present application
  • FIG. 2 is a schematic flowchart of a method for constructing a risk decision rule generation model according to an embodiment of the present application
  • FIG. 3 is another schematic flowchart of a method for determining a risk control strategy based on a prediction model according to an embodiment of the present application
  • FIG. 4 is a schematic structural diagram of a device for determining a risk control strategy based on a prediction model according to an embodiment of the present application
  • FIG. 5 is a schematic structural diagram of a terminal device according to an embodiment of the present application.
  • the method for determining a risk control strategy based on a prediction model provided in the embodiment of the present application can be applied to multiple application fields such as investment, banking, insurance, securities, and e-commerce Of any application.
  • the application scenarios involved may include, but are not limited to, login, registration, pre-lending, during-lending, after-lending, holiday activities, or promotional activities.
  • the method provided in the embodiment of the present application may be applicable to the generation of risk decision rules of any service type in any of the foregoing application scenarios.
  • the above risk decision rules are the rules on which a risk control strategy is specifically selected in the risk decision process.
  • the risk decision rules will be described below as an example.
  • different risk decision rule generation models can be constructed. Different risk decision rule generation models can be applied to different application scenarios and the generation of risk decision rules for multiple businesses in various application scenarios with high flexibility.
  • the risk-based decision rule generation model can output real-time user data of a specific business in a specific application scenario to output risk-based decision rules for risk control of the business, thereby improving the relevance and applicability of risk-based decision rules to the business. Strong.
  • the risk decision rules are based on user data generation and have strong reliability, which enhances the accuracy and applicability of risk decisions made based on the risk control strategies determined by the risk decision rules.
  • the above-mentioned services may specifically be various services provided to users in various application fields such as investment, banking, insurance, securities, and e-commerce, such as insurance and loans.
  • the corresponding application scenarios under this service may include, but are not limited to, account registration, account login, insurance application, insurance approval, policy generation, and maintenance.
  • the above application scenarios are merely examples, not exhaustive.
  • the specific application scenarios may be determined according to actual application scenarios, and are not limited herein.
  • the method provided in the embodiment of the present application can generate risk decision rules corresponding to each business type based on sample data obtained from various data acquisition paths for each business type.
  • the model determines the risk control strategy based on the risk decision rules generated by the risk decision rule generation model.
  • a risk decision rule generation model suitable for generating risk decision rules corresponding to each business type can be constructed, in other words, for samples associated with multiple business types The data can be trained to generate a risk decision rule generation model.
  • multiple sets of network sets in a risk decision rule generation model can be constructed, where a set of network parameters of the risk decision rule generation model can be adapted to generate one Risk decision rules associated with business types.
  • a set of network parameters of the risk decision rule generation model can be adapted to generate one Risk decision rules associated with business types.
  • the embodiment of the present application will take the target business as an example for illustration.
  • the above risk decision rule generation model can be described by taking the risk decision rule generation model associated with the target business as an example, and there is no limitation here.
  • the method provided in the embodiment of the present application may build a risk decision rule generation model based on relevant sample data in historical risk control records of target businesses in specific application scenarios.
  • Based on the historical risk control records of the target business stored in the user database of the target business at least two risk decision rules adopted for the target business can be determined from the historical risk control records.
  • historical risk control records of each business can be obtained from the user group database of other businesses, and at least two risk decision rules adopted for each business are determined from the historical risk control records of each business, Other services include one or more services of the same type as the target service.
  • the sample user data can also be used to build a risk decision rule generation model, which can be determined according to the actual application scenario. Make restrictions.
  • the above sample user data includes, but is not limited to, the user's business account information, the user's page operation data, the user's business access duration, the user's business access frequency, the user's terminal device identification information, and the user's location information.
  • the actual application scenario is determined, and there is no limitation here.
  • the risk decision rule may represent a rule associated with user data, or a combination rule composed of various rules among multiple rules associated with user data.
  • the combination rule composed of one rule and / or multiple rules may also be referred to as an association rule, which may be specifically determined according to an actual application scenario, and is not limited herein.
  • the above-mentioned risk decision-related rule generation model may be called an association rule model, which is not limited herein.
  • the following will describe the construction of a risk decision rule generation model corresponding to any business, and the implementation method of outputting the risk decision rule corresponding to the business based on the constructed risk decision rule generation model, and describe the implementation method provided in the embodiment of the present application as an example.
  • any service here can be described by using a target service as an example.
  • the risk decision rule generation model corresponding to the target business can be described by taking the target risk decision rule generation model as an example.
  • the method and device provided in the embodiments of the present application will be described below with reference to FIGS. 1 to 5 respectively.
  • the methods provided in the embodiments of the present application may include data processing stages such as construction of a target risk decision rule generation model, generation of a risk decision rule based on the target risk decision rule generation model, and determination of a risk decision strategy based on the risk decision rule.
  • data processing stages such as construction of a target risk decision rule generation model, generation of a risk decision rule based on the target risk decision rule generation model, and determination of a risk decision strategy based on the risk decision rule.
  • FIG. 1 is a schematic flowchart of a method for determining a risk control strategy based on a prediction model according to an embodiment of the present application.
  • the target risk decision rule generation model corresponding to the target service is taken as an example for description, and the details are not described below.
  • the method provided in the embodiment of the present application may include the following steps S1 to S3:
  • the construction of the target risk decision rule generation model may include the modeling data (i.e., sample data) of the target risk decision rule generation model, the training of the target risk decision rule generation model, and the target risk decision rule generation. Data processing stages such as testing of models. Please refer to FIG. 2 together.
  • FIG. 2 is a schematic flowchart of a method for constructing a risk decision rule generation model according to an embodiment of the present application. The construction of the risk decision rule generation model provided in the embodiment of the present application can be described by the implementation manners provided in the following steps S11 to S13.
  • the modeling data of the target risk decision rule generation model may be sourced from user data stored in a user database of the target business, or may be derived from big data analysis to obtain user data associated with the target business. It is determined according to the actual application scenario, and there is no limitation here.
  • Based on the historical risk control records of the target business stored in the user database of the target business at least two risk decision rules adopted for the target business can be determined from the historical risk control records.
  • a target risk decision rule generation model corresponding to the target business is constructed.
  • historical risk control records of each business can be obtained from the user group database of other businesses, and at least two risk decision rules adopted for each business are determined from the historical risk control records of each business, Other services include one or more services of the same type as the target service.
  • the sample user data can also be used to build a target risk decision rule generation model, which can be determined according to the actual application scenario. No restrictions.
  • the modeling data of the target risk decision rule generation model used to build the target business may collect sample user data corresponding to the risk decision rule of the target business in a certain period of time, including but not limited to The sample user behavior data of the target service and the terminal device identification information associated with the sample user behavior data are not limited here.
  • the collection time period of the modeling data of the above-mentioned target risk decision rule generation model may be determined by factors such as the effective duration of the predefined risk decision rule of the target business, or policy changes (changes in local policies), and there is no limitation here. .
  • the collection period of the modeling data of the target risk decision rule generation model may be sample user data recorded one day before the current time. If the validity period of the target risk decision rule generated based on the target risk decision rule generation model is one week, the collection period of the modeling data of the target risk decision rule generation model may be the sample user data recorded in the previous week of the current time .
  • the target risk decision rule generation model can be used to regenerate the risk decision rule applicable to the next valid duration period, thereby realizing the risk decision rule.
  • Regular updates can ensure the timeliness of risk decision rules and enhance the reliability of risk decisions such as the selection of risk control strategies based on risk decision rules.
  • real-time updated sample user data can be used to generate corresponding risk decision rules, which can be applicable to emergency mechanisms and further ensure the generation of target risk decision rule generation model generation.
  • the risk decision rules are more reliable and applicable.
  • the sample user data collected during the collection of the modeling data includes, but is not limited to, the user's business account information, the user's page operation data, the user's business access duration, the user's business access frequency,
  • the terminal device identification information of the user and the area information of the user are not limited here.
  • the data of the user's page operation data, the user's business access duration, and the user's business access frequency can also be referred to as sample user behavior data, which is not limited herein.
  • the terminal device identification information of the user may be terminal device information associated with the sample user behavior data of the foregoing various forms of expression. In other words, the terminal device identification information of the user device may be used when the user accesses a target service.
  • the device information of the terminal device is not limited here.
  • the foregoing terminal device information may be device information of a terminal device used by a user to log in to a page for processing a target service or register a service account of the target service, and is not limited herein.
  • the device information of the terminal device may include a medium access control (MAC) address of the terminal device, an international mobile equipment identity (IMEI), an Internet protocol (IP) address, and a direct inward direction. Direct dialing (DID), the display resolution of the terminal device used by the user, and the user's contact information (mobile phone number, etc.) bound to the terminal device are not limited here.
  • MAC medium access control
  • IMEI international mobile equipment identity
  • IP Internet protocol
  • DID Direct dialing
  • the display resolution of the terminal device used by the user and the user's contact information (mobile phone number, etc.) bound to the terminal device are not limited here.
  • the above-mentioned user's location information may be the geographic location where the user logs in to the target service page or the business account of the target service, and the corresponding information may also be the user's login to the target service page or the target service business account.
  • the location and geographic location of the terminal equipment used can be determined according to the actual application scenario, and there is no limitation here.
  • the modeling data includes sample user data corresponding to at least two risk decision rules, where at least first sample data and second rule sample data are included.
  • the first rule sample data includes the first risk decision rule and its corresponding first sample user data
  • the second rule sample data includes the second risk decision rule and its corresponding second sample user data.
  • the above modeling data is sample data of the target business, it is the sample user data corresponding to the historical risk strategy rules used for risk control of the target business.
  • the risk decision rules used for the sample user data of the target business Is a known parameter, such as a first risk decision rule and / or a second risk decision rule. Therefore, the sample user data for the target service may be existing labeled data or customizable labeled data. Among them, existing annotations or custom annotations can be used to mark risk decision rules corresponding to sample user data, which can be understood as labels for risk decision rules. Therefore, for sample user data, the corresponding risk decision rules are also known data.
  • the feature data corresponding to the sample user data of the target business and the labeling of this part of the user data that is, the label of the risk decision rule
  • the modeling data for the risk decision rule generation model training are the modeling data for the risk decision rule generation model training.
  • modeling data of different data types and / or data contents may be correspondingly trained to obtain a target risk decision rule generation model suitable for outputting the risk decision rule.
  • the modeling data includes sample user data corresponding to a certain risk decision rule corresponding to the target business, and a set of network parameters can be obtained by training in this part of the modeling data, so that a target with such a set of network parameters
  • the risk decision rule generation model can correspondingly output the risk decision rules according to the input user data.
  • the risk decision rules included in the above modeling data may exist in the form of labels or thresholds. Different risk decision rules based on modeling data can be labeled with different values.
  • risk decision rules in the modeling data is a label
  • Rule 4 so that multiple sets of network parameters can be obtained by training with modeling data containing multiple risk decision rules, so that the target risk decision rule generation model can predict the target risk decision rules for the user data of the input target business.
  • data other than the risk decision rule in the modeling data of the target risk decision rule generation model may be referred to as user data.
  • user data For example, for sample user data whose risk decision rule is the first risk decision rule, in addition to including the first risk decision rule, there is also user data that uses the first risk decision rule for risk control for the target business. The first sample user data is described as an example).
  • the sample user data whose risk decision rule is the second risk decision rule in addition to including the second risk decision rule, there is also user data that uses the second risk decision rule for risk control for the target business (for convenience of description) (The second sample user data is taken as an example for description.) The details are not described below.
  • the model data of the above-mentioned target risk decision rule generation model includes the first sample user data and the second sample user data, and the sample user data corresponding to any risk decision rule includes, but is not limited to, the user's business account information, the user's Page operation data, user's service access duration, user's service access frequency, user's terminal equipment identification information, and user's location information.
  • a sample user feature pair may be constructed.
  • each risk decision rule can be labeled in the form of a label, and based on the label, a sample user characteristic corresponding to the risk decision rule is generated.
  • the first risk decision rule can be labeled based on label 1
  • the second risk decision rule can be labeled based on label 2
  • the feature corresponding to the first risk decision rule (such as the feature represented by a character "0") can be generated based on label 1.
  • the feature corresponding to the second risk decision rule (for example, a feature represented by a character “1”) is generated based on the label 2 and is not limited herein.
  • a sample user feature pair includes a first sample user feature corresponding to a first risk decision rule and a second sample user feature corresponding to a second risk decision rule.
  • the first sample user characteristics corresponding to the first risk decision rule may be constructed from the first sample user data corresponding to the first risk decision rule.
  • the second sample user characteristics corresponding to the second risk decision rule may be constructed from the second sample user data corresponding to the second risk decision rule.
  • the first sample user feature corresponding to the first risk decision rule may be used as the positive sample feature in the sample user feature pair
  • the second sample user feature corresponding to the second risk decision rule may be used as the sample user.
  • the negative sample features in the feature pair or vice versa
  • the initial network model of the risk decision rule prediction can be trained based on a positive and negative sample feature to obtain the predicted risk decision rule as the first risk decision rule or the second risk decision Rule Capability Decision Model for Target Risk Decision Making.
  • the above-mentioned target risk decision rule generation model predicts the output risk decision rule as the first risk decision rule or the second risk decision rule as examples, including but not limited to the first risk decision rule and the second risk decision rule.
  • the actual application scenario is determined, and there is no limitation here.
  • an abstract feature representation of sample user data corresponding to each risk decision rule can be obtained, and then according to the risks
  • the abstract feature of the sample user data corresponding to the decision rule is composed of the sample user feature corresponding to the risk decision rule.
  • a multi-character feature vector may be used to represent, and the feature vector may be composed of six partial features.
  • the above-mentioned six partial features may include the user's business account information in the first sample user data, the user's page operation data, the user's business access duration, the user's business access frequency, the user's terminal device identification information, and the user's identity.
  • any one of the above six partial features may be composed of one or more characters, one or more sets of characters, and / or one or more dimensions of characters, and is not limited herein.
  • the user's business account information includes four dimensions of business account information, the contacts bound to the business account, the contact methods bound to the business account, and the business area information to which the business account belongs, four characters can be used. (Or 4 groups of characters or 4 dimensions of characters, etc., which are not limited here) are used to represent the features abstracted by the user's business account information. Wherein, each of the above 4 characters can represent information of one dimension.
  • the information of each dimension can be classified separately, and different types of information (such as 0 or 1) are used to mark the information of different categories, and then the corresponding identifier of the information of each dimension can be obtained, so that the information of each dimension can be correspondingly
  • the combination of the identifiers yields a four-character business account feature. For example, if the contact person bound to the business account has a certain person, it can be marked with an identifier. If the contact person bound to the business account is empty, it can be marked with another identifier. Characteristics of the character corresponding to the bound contact. By analogy, the characteristics of each character in the above-mentioned four-character business account characteristics can be determined, so that the four-character business account characteristics can be obtained.
  • the characters corresponding to the dimensions may be filled in blanks, etc. to construct the features corresponding to the dimensions, which is not limited here.
  • the user's page operation data includes the page operation area (the area can be divided according to the page, and an identifier is used for classification and identification for each area, etc.), the page operation time (such as the classification and identification according to the length of time, etc.), and Page operation trajectory (can be classified and identified according to the type of trajectory) of the three dimensions of information, you can use 3 characters (or 3 groups of characters or 3 dimensions of characters, etc., without limitation here) to indicate the user's page operations
  • the features abstracted by the data are simply referred to as the user's page operation features for the convenience of description, and are not limited here.
  • the implementation process of the page operation characteristics of the user abstracted from the page operation data of the user reference may be made to the implementation manner corresponding to the business account characteristics of the user, which is not limited herein.
  • the user's business visit duration can be classified and identified according to the duration. Therefore, 1 character (or a group of characters or 1-dimensional characters, etc., is not limited here) can be used to represent the abstract feature of the user's business visit duration. For the convenience of describing the service access duration characteristics of the user, it is not limited here.
  • the user's service access frequency (can be classified according to frequency segmentation, etc.), the user's terminal device identification information (can be classified according to the amount of terminal device identification information, etc.), and the user's geographical information (can be based on Geographical classification, etc.)
  • User data such as one or more characters, can be used to represent its abstracted characteristics.
  • the user's service access frequency characteristics, the user's terminal equipment identification characteristics, and the user's geographical characteristics can be described separately.
  • the implementation process of the corresponding features abstracted by the foregoing user data refer to the implementation manner corresponding to the above-mentioned user's service account feature, which is not limited herein.
  • the sample user feature pairs may be input into the initial network model of the target risk decision rule generation model and passed through the initial network.
  • the model learns the sample user data features included in the input sample user feature pair and the label features of the corresponding risk decision rules to obtain a target risk decision rule generation model that has the ability to output risk decision rules corresponding to any user data feature.
  • the initial network model of the above-mentioned target risk decision rule generation model may be a backpropagation (BP) neural network model, or other types of neural network models, which are not limited herein.
  • BP backpropagation
  • the activation function of the above-mentioned target risk decision rule generation model may be a sigmoid function, etc., which may be specifically determined according to an actual application scenario, and is not limited herein.
  • the output of the above-mentioned target risk decision rule generation model is a label or threshold corresponding to each risk decision rule, and further, a specific risk decision rule may be determined based on the label or threshold corresponding to each risk decision rule. , which can be determined according to the actual application scenario, and is not limited here.
  • the test data generated by the risk decision rule includes at least one kind of test data for the risk decision rule, and further, at least one sample user characteristic may be constructed based on the test data of the at least one risk decision rule. Based on the test data generated by the above risk decision rules, a sample user test feature is constructed, and then the prediction accuracy of the risk decision rule of the target risk decision rule generation model can be tested based on the constructed sample user test feature.
  • the data types (or data dimensions) included in the sample user data included in the test data generated by the risk decision rules may be the same as the data types (or data) included in the modeling data of the target risk decision rule generation model. Dimensions) are the same, and no restrictions are made here, which can ensure the test effectiveness of the target risk decision rule generation model, improve the accuracy of the test results of the target risk decision rule generation model, and enhance the applicability of the target risk decision rule generation model.
  • the model based on the target risk decision rule generation model may learn the sample user test characteristics constructed by the test data generated by the risk decision rule generation above, and output a corresponding risk decision rule.
  • the risk decision rule generated by the target risk decision rule generation model can be combined with the test data corresponding to the known risk decision rule to calculate the loss value of the output of the target risk decision rule generation model (such as the label corresponding to the risk decision rule and / Or the difference between the thresholds, etc., which is not limited here).
  • the output loss value of the target risk decision rule generation model can be fed back to the target risk decision rule generation model. Based on the loss value, the network parameters of the target risk decision rule generation model are modified and optimized, which can improve the target risk decision rule.
  • the prediction accuracy of the generated model is stronger.
  • the training samples are from sample user data corresponding to multiple risk decision rules. Based on the sample user data corresponding to multiple risk decision rules, it can be constructed for model training. Sample user feature pairs for training the target risk decision rule generation model so that the model has the ability to output risk decision rules corresponding to the user characteristics corresponding to any user data, so that the model can be used to predict any user data based on the target risk decision rule generation model Corresponding risk decision rules realize effective prediction of risk decision rules, which can enhance the reliability of generating risk decision rules based on target risk decision rule generation models. At the same time, risk decision rules generated based on target risk decision rule generation models have users. The data support of the data further improves the applicability of target risk decision rules.
  • the process of generating a target risk decision rule based on a target risk decision rule generation model may include obtaining user data, determining a target risk decision rule generation model, and generating target risk decision rule data. stage. Further, a target risk control strategy for risk control of a target business can be determined based on target risk decision rules. Please refer to FIG. 3 together.
  • FIG. 3 is another schematic flowchart of a method for determining a risk control strategy based on a prediction model according to an embodiment of the present application. The method for determining a risk strategy provided in the embodiment of the present application may include the following steps S21 to S24.
  • S21 Obtain first user data, and determine a target service associated with the first user data.
  • a user when a user needs to handle a target service on a service operation page of a browser and / or client corresponding to the target service, it may be obtained based on user operations on the service operation page of the browser and / or client.
  • User data (for convenience of description, the first user data may be used as an example for description), and a target service may be determined according to the first user data.
  • the types of data included in the first user data include, but are not limited to, the user's business account information, the user's page operation data, the user's terminal device identification information, and the user's location information, which are not limited here.
  • the page operation data included in the first user data may include a page start position, a page link input, a page connection path, a page operation duration, and a page operation frequency, and the like is not limited herein.
  • derived data such as the user's business access duration and user's business access frequency can be calculated based on the above-mentioned page operation data, so that the derived data can be determined as a part of the first user data, which can be determined according to the actual application scenario. Make restrictions.
  • the target service associated with the first user data may be determined based on the user's business account information, page start location, page link input, and page connection path in the first user data. For example, when a user needs to log in to a business account of a certain service or register a business account of a certain service, he can click the icon of a browser and / or client for handling the service through a mouse or a finger, etc., so that the browser can be opened. And / or client's business operations page. Enter the existing business account information on the business operation page, or fill in the business account information to be registered.
  • the user data that the user operates on the business operation interface can be collected, including the user's business account information and page operations Data and more.
  • the business associated with the first user data can be determined, that is, the target business, so that a risk decision for the target business can be determined based on the target risk decision rule generation model corresponding to the target business. rule.
  • the user's service access duration can be monitored in real time, and the user's service access frequency can also be determined in combination with the user's historical record of access. It is determined according to the actual application scenario, and there is no limitation here.
  • the identification information of the terminal equipment used by the user and the location information of the user can be collected in real time, which can be determined according to the actual application scenario, and is not limited here.
  • the data type and / or data content of the first user data refer to the data type and / or data content included in any of the sample user data in the sample user data in the implementation manner provided in the steps S11 to S13 above. I will not repeat them here.
  • a target service can be determined based on the first user data, and the business type (i.e., target service type) of the target service is associated with each risk decision rule generation model included in the risk decision rule generation model set. Match the business types of the two, and determine the risk decision rule generation model corresponding to the target business type from the above risk decision rule generation model set.
  • the risk decision rule generation model set here can also include other risk decision rule generation models associated with its business type other than the target business type.
  • the other risk decision The rule generation model is trained from sample data associated with other business types.
  • the risk decision rule generation model associated with the target business is obtained from the above risk decision rule generation model set by matching, and then the user data (that is, the first user data) collected in real time can be performed based on the risk decision rule generation model. Learning to generate target risk decision rules corresponding to the first user data.
  • the terminal device may also match the risk decision rule generation model of the network parameter associated with the target service type from the above set of risk decision rule generation model sets based on the target service type, and may further based on the risk decision rule having the network parameter
  • the generating model learns the first user data collected in real time to generate target risk decision rules corresponding to the first user data. It can be determined according to the actual application scenario, and is not limited here.
  • the implementation manner of constructing the first user characteristics based on the first user data of the target service may be the same as the construction method of the sample user characteristics in the modeling data of the target risk decision rule generation model. Refer to the implementation manner provided in step S21 above, which is not repeated here.
  • the first user feature may be input into a target risk decision rule generation model, and the first user feature is learned based on the target risk decision rule generation model to generate the first user feature.
  • Target risk decision rules for a user's data may be input into a target risk decision rule generation model, and the first user feature is learned based on the target risk decision rule generation model to generate the first user feature.
  • the target risk decision rule generation model may output the label or threshold of the target risk decision rule corresponding to the first user data, and then may be determined based on the label or threshold of the target risk decision rule.
  • Target risk decision rules For example, after the target risk decision rule generation model learns the characteristics of the first user, the risk decision rule label that can be output correspondingly is label 1 or the risk decision rule threshold is threshold value 1, then the target risk decision generated for the first user data may be determined The rule is the first risk decision rule.
  • the target risk decision rule generation model learns the characteristics of the first user, the output risk decision rule label corresponding to the label 2 or the risk decision rule threshold value is the threshold value 2, then the target risk decision rule generated for the first user data may be determined Is the second risk decision rule.
  • the target risk decision rule generated for the first user data may be determined Is the second risk decision rule.
  • the data type and / or data content corresponding to the first user characteristic of the target risk decision rule generation model is obtained by filtering and input based on the first user data, and may be used for training and / or training of the target risk decision rule generation model.
  • the data types and / or data contents corresponding to the modeled data and / or user characteristics constructed by the test data input during the test phase are the same.
  • the data types and / or data content collected and filtered during the training, testing, and use phases of the target risk decision rule generation model are the same, so that the target risk decision rule generation model can better utilize the input user characteristics to learn And output the corresponding risk decision rules, which can increase the accuracy and reliability of the target risk decision rule generation model to generate risk decision rules, and have stronger applicability.
  • the target risk decision rule is a combination of one or more rules including the first risk decision rule and / or the second risk decision rule, and the one or more rules include:
  • Rule 1 When the device identification contained in the user's terminal device identification information is greater than or equal to the device identification threshold, the first risk control strategy is used to control the risk of the business. When the device identification contained in the user's terminal device identification information is less than the device identification threshold, a second risk control strategy is used to control the risk of the business.
  • the determination condition of the above rule 1 may be: the number of IPs is greater than the IP number threshold.
  • the above rule 1 indicates that: if the number of IPs used by the same user in the first user data is greater than the threshold for the number of IPs, a first risk control strategy (such as risk control strategy 1) may be selected when making risk decisions based on rule 1, and The target business corresponding to the user data performs risk control. If the number of IPs used by the same user in the user data is less than or equal to the IP number threshold, a second risk control strategy (such as risk control strategy 2) can be selected when making risk decisions based on rule 1, and the user corresponding to the user data Conduct risk control.
  • the first risk control strategy and / or the second risk control strategy are merely examples, and are not limited herein.
  • the determination condition of the above rule 1 may also be: the number of MACs is greater than a threshold of the number of MACs.
  • the above rule 1 may indicate that: if the number of MACs of the terminal device used by the same user in the first user data is greater than the threshold of the number of MACs, a first risk control strategy may be selected when making risk decisions based on rule 1 to the user The target business corresponding to the data is subject to risk control. If the MAC of the terminal device used by the same user in the user data is less than or equal to the MAC number threshold, a second risk control strategy may be selected when making a risk decision based on Rule 1, to perform risk control on the target service corresponding to the user data.
  • the thresholds (including the number of IP thresholds and / or the number of MAC thresholds) in the determination conditions of Rule 1 above may be generated after learning the sample user data by the target risk decision rule generation model, thereby increasing the target risk-based decision
  • the rule generation model generates applicability of target risk decision rules.
  • the thresholds in the following judgment conditions of rules 2, 3, and 4 can also be generated by the target risk decision rule generation model, which will not be described in detail below.
  • Rule 2 When the user's service access duration is greater than or equal to the access duration threshold, a third risk control strategy is used to control the risk of the business. When the user's service access duration is less than the access duration threshold, a fourth risk control strategy is adopted to control the risk of the business.
  • the determination condition of the above rule 2 may be that the user access amount per unit time is greater than the user access amount threshold.
  • Rule 2 may indicate that: if it is determined based on the first user data that the number of user visits of the same user or multiple users to the target service per unit time is greater than the user visit threshold, the third risk control may be selected when making risk decisions based on Rule 2.
  • Strategy such as risk control strategy 3, etc.
  • a fourth risk control method (such as risk Decision strategy 4): Perform risk control on the target service corresponding to the first user data.
  • Rule 3 When the user's service access frequency is greater than or equal to the access frequency threshold, the fifth risk control strategy is used to control the risk of the business. When the user's service access frequency is less than the access duration threshold, a sixth risk control strategy is used to control the risk of the service.
  • the determination condition of Rule 3 above may be: the service access frequency per unit time is greater than the service access frequency threshold.
  • Rule 3 may indicate that: if it is determined based on the first user data that the service access frequency of the same user or multiple users to the target service per unit time is greater than the service access frequency threshold, a fifth risk control may be selected when making risk decisions based on Rule 3.
  • a strategy such as a risk control strategy 5
  • a sixth risk control method (such as risk Decision strategy 6): Perform risk control on the target service corresponding to the first user data.
  • the fifth risk control strategy and the sixth risk control strategy are just examples, and are not limited herein.
  • the above-mentioned target risk decision rule may also be one or more combinations of rule 1, rule 2, and rule 3.
  • rule 4 may be used as an example.
  • Rule 4 When the device identification contained in the user's terminal device identification information is greater than or equal to the device identification threshold, and the user's service access duration is greater than or equal to the access duration threshold, the first risk control strategy is used to control the risk of the service, otherwise Adopting a second risk control strategy for business risk control.
  • the determination condition of rule 4 may be: IP number> IP number threshold
  • rule 4 may indicate that the association rule is a combination of the above rules 1 and 2. Only when the condition judgments of the above rules 1 and 2 are satisfied at the same time, the first risk control strategy can be selected when making risk decisions based on rule 4, and The target service corresponding to a user data is subject to risk control, otherwise a second risk control strategy may be selected when making a risk decision based on Rule 4 to perform risk control on the target service corresponding to the first user data, and there is no limitation here.
  • various parameters included in the first user data and the target risk decision rule may be The included judgment conditions are compared to determine the target risk control strategy for risk control of the target business. For example, if the first user data is learned based on the target risk decision rule generation model and the target risk decision rule is generated as rule 1, based on the first user data, the number of IPs included in the first user data is greater than the target. The threshold of the number of IPs generated by the risk decision rule generation model can determine that the target risk control strategy is the first risk control strategy.
  • the target risk decision rule is any combination of one or more rules other than rule 1
  • the target risk can be determined based on the judgment conditions included in each rule and the data type included in the first user data.
  • the control strategy can be determined according to the actual application scenario, and is not limited here.
  • the above-mentioned implementation manner of determining the risk control strategy based on the risk decision rule may be determined based on the target risk decision rule as an example.
  • step S24 which is determined based on the target risk decision rule.
  • the implementation method of the target risk control strategy is not repeated here.
  • a target risk decision rule generation model may be determined based on the target business associated with the first user data, and a target risk decision rule generation model may be generated based on the target risk decision rule generation model.
  • the target risk decision rule generation model is trained based on the sample data corresponding to the target business, making the risk decision rule generation based on the target risk decision rule generation model more reliable.
  • the target risk decision rule generated by the target risk decision rule generation model has The data support of the first user data further improves the applicability of target risk decision rules.
  • the target risk control strategy for risk control of the target business can be determined based on the target risk decision rule.
  • the determination method of the risk control strategy is more flexible, and the reliability of the target risk decision rule can be enhanced to The reliability of the risk control strategy generated by the target risk decision rules can improve the effectiveness and risk applicability of risk control of the target business.
  • FIG. 4 is a schematic structural diagram of a device for determining a risk control strategy based on a prediction model according to an embodiment of the present application.
  • An apparatus for determining a risk control strategy based on a prediction model provided in an embodiment of the present application includes:
  • the obtaining unit 41 is configured to obtain first user data and determine a target service associated with the first user data.
  • the model determining unit 42 is configured to determine a target risk decision rule generation model associated with the target service according to the target service type of the target service obtained by the obtaining unit 41.
  • a rule generation unit 43 is configured to generate a target risk decision rule corresponding to the first user data obtained by the obtaining unit 41 based on the target risk decision rule generation model determined by the model determination unit 42, and the target risk decision rule generation model is based on the foregoing
  • the sample data corresponding to the target business type is obtained by training.
  • the above sample data includes at least the first rule sample data and the second rule sample data.
  • the first rule sample data includes the first risk decision rule and its corresponding first sample user.
  • the second rule sample data includes a second risk decision rule and its corresponding second sample user data.
  • a policy determining unit 44 is configured to determine a target risk control strategy for performing risk control on the target service according to the first user data obtained by the obtaining unit and the target risk decision rule determined by the rule generating unit.
  • the model determining unit 42 is configured to:
  • the set of risk decision rule models also includes other risk decision rule models associated with other business types other than the target business type.
  • the other risk decision rule models are obtained by training the sample user data corresponding to the other business types.
  • the confirmation device further includes:
  • a model construction unit 45 is configured to obtain sample data corresponding to at least two risk decision rules used to construct the target risk decision rule generation model.
  • the at least two risk decision rules include at least the first risk decision rule and the second risk.
  • Decision rule the sample data includes at least the first rule sample data and the second rule sample data;
  • the model construction unit 45 is further configured to construct at least one sample user feature pair according to the sample data corresponding to the at least two risk decision rules, and generate a target risk decision rule based on the at least one sample user feature pair.
  • model building unit 45 is configured to:
  • model building unit 45 is configured to:
  • the data types included in any of the sample user data in the first user data and / or the sample data corresponding to the target service type include: a user's business account information, a user's page operation data, One or more of the user's service access duration, the user's service access frequency, the user's terminal equipment identification information, and the user's location information;
  • the business account information of the user and / or page operation data of the user are used to determine a service and / or a service type.
  • the target risk decision rule is a combination of one or more rules including the first risk decision rule and / or the second risk decision rule, and the one or more rules include:
  • Rule 1 When the device identification included in the user's terminal device identification information is greater than or equal to the device identification threshold, the first risk control strategy is used to control the risk of the business. When the device identification included in the user's terminal device identification information is used When the value is less than the above equipment identification threshold, the second risk control strategy is adopted for risk control of the business;
  • Rule 2 When the user's business access duration is greater than or equal to the access duration threshold, the third risk control strategy is used to control the business's risk. When the user's business access duration is less than the above access duration threshold, the fourth risk control strategy is used. Business risk control;
  • Rule 3 When the user's service access frequency is greater than or equal to the access frequency threshold, the fifth risk control strategy is used to control the risk of the business. When the user's business access frequency is less than the above access duration threshold, the sixth risk control strategy is used. Business risk control.
  • the above-mentioned device for determining a risk control strategy based on a prediction model may implement the implementation manners provided by the steps in FIG. 1 to FIG. 3 described above through each of its built-in function modules.
  • the obtaining unit 41 may be configured to perform operations such as obtaining first user data and / or sample user data in the foregoing steps.
  • operations such as obtaining first user data and / or sample user data in the foregoing steps.
  • the above-mentioned model determining unit 42 may be used to execute an implementation manner of generating a risk decision rule based on a business type in each of the steps described above, and the above-mentioned rule generation module 43 may be used to implement an implementation manner of generating a risk decision rule based on a risk decision rule generation model in each of the above steps
  • the above-mentioned policy determination unit 44 may be used to execute the implementation method of the risk control strategy based on the risk decision rules and user data in the above-mentioned steps, and the above-mentioned model construction unit 45 may be used to execute the correlation in the construction of the target risk decision-rule generation model in each of the above steps.
  • the implementation manners described in the steps reference may be made to the implementation manners provided in the foregoing steps, and details are not described herein again.
  • the target risk decision rule generation model is trained based on the sample data corresponding to the target business.
  • the risk decision rule generation based on the target risk decision rule generation model is more reliable and improves the applicability of the target risk decision rule.
  • the target risk control strategy for risk control of the target business can be determined.
  • the determination method of the risk control strategy is more flexible, which can enhance the reliability of the risk control strategy generated by the target risk decision rules and improve the target business. Effective risk control and higher applicability.
  • FIG. 5 is a schematic structural diagram of a terminal device according to an embodiment of the present application.
  • the terminal device in this embodiment may include one or more processors 501 and a memory 502.
  • the processor 501 and the memory 502 are connected via a bus 503.
  • the memory 502 is configured to store a computer program.
  • the computer program includes program instructions.
  • the processor 501 is configured to call the program instructions stored in the memory 502 and perform the following operations:
  • the target risk decision rule generation model corresponding to the first user data is generated based on the target risk decision rule generation model.
  • the target risk decision rule generation model is trained based on the sample data corresponding to the target business type.
  • the sample data includes at least the first rule sample.
  • Data and second rule sample data the first rule sample data includes a first risk decision rule and its corresponding first sample user data, and the second rule sample data includes a second risk decision rule and its corresponding first sample data Two sample user data;
  • a target risk control strategy for performing risk control on the target business is determined according to the first user data and the target risk decision rule.
  • the foregoing processor 501 is configured to:
  • the set of risk decision rule models also includes other risk decision rule models associated with other business types other than the target business type.
  • the other risk decision rule models are obtained by training the sample user data corresponding to the other business types.
  • the foregoing processor 501 is further configured to:
  • sample data corresponding to at least two risk decision rules used to build the target risk decision rule generation model includes at least the first risk decision rule and the second risk decision rule.
  • the sample data includes: Including at least the first rule sample data and the second rule sample data;
  • the foregoing processor 501 is configured to:
  • the foregoing processor 501 is configured to:
  • the data types included in any of the sample user data in the first user data and / or the sample data corresponding to the target service type include: a user's business account information, a user's page operation data, One or more of the user's service access duration, the user's service access frequency, the user's terminal equipment identification information, and the user's location information;
  • the business account information of the user and / or page operation data of the user are used to determine a service and / or a service type.
  • the target risk decision rule is a combination of one or more rules including the first risk decision rule and / or the second risk decision rule, and the one or more rules include:
  • Rule 1 When the device identification included in the user's terminal device identification information is greater than or equal to the device identification threshold, the first risk control strategy is used to control the risk of the business. When the device identification included in the user's terminal device identification information is used When the value is less than the above equipment identification threshold, the second risk control strategy is adopted for risk control of the business;
  • Rule 2 When the user's business access duration is greater than or equal to the access duration threshold, the third risk control strategy is used to control the business's risk, and when the user's business access duration is less than the access duration threshold, the fourth risk control strategy is used Business risk control;
  • Rule 3 When the user's service access frequency is greater than or equal to the access frequency threshold, the fifth risk control strategy is used to control the risk of the business. When the user's business access frequency is less than the above access duration threshold, the sixth risk control strategy is used. Business risk control.
  • the processor 501 may be a central processing unit (CPU), and the processor may also be another general-purpose processor or a digital signal processor (DSP). , Application specific integrated circuit (ASIC), ready-made programmable gate array (field programmable gate array), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general-purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the memory 502 may include a read-only memory and a random access memory, and provide instructions and data to the processor 501. A part of the memory 502 may further include a non-volatile random access memory. For example, the memory 502 may also store information of a device type.
  • the above-mentioned terminal device may implement the implementation manners provided by the steps in FIG. 1 to FIG. 3 described above through its built-in functional modules.
  • the implementation manners provided in the foregoing steps and details are not described herein again.
  • the target risk decision rule generation model is trained based on the sample data corresponding to the target business.
  • the risk decision rule generation based on the target risk decision rule generation model is more reliable and improves the applicability of the target risk decision rule.
  • the target risk control strategy for risk control of the target business can be determined.
  • the determination method of the risk control strategy is more flexible, which can enhance the reliability of the risk control strategy generated by the target risk decision rules and improve the target business. Effective risk control and higher applicability.
  • An embodiment of the present application further provides a computer-readable storage medium.
  • the computer-readable storage medium stores a computer program.
  • the computer program includes program instructions. When the program instructions are executed by a processor, each step in FIG. 1 to FIG. 3 is implemented.
  • the computer-readable storage medium may be an apparatus for determining a risk control strategy based on a prediction model provided in any of the foregoing embodiments, or an internal storage unit of the terminal device, such as a hard disk or a memory of an electronic device.
  • the computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, Flash card, etc.
  • the computer-readable storage medium may include both an internal storage unit and an external storage device of the electronic device.
  • the computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device.
  • the computer-readable storage medium can also be used to temporarily store data that has been or will be output.

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Educational Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种基于预测模型的风险控制策略的确定方法及装置，该方法包括：获取第一用户数据，确定上述第一用户数据所关联的目标业务（S21）；根据目标业务的目标业务类型确定出上述目标业务所关联的目标风险决策规则生成模型（S22）；基于上述目标风险决策规则生成模型生成所述第一用户数据对应的目标风险决策规则（S23）；根据上述第一用户数据和目标风险决策规则确定出对上述目标业务进行风险控制的目标风险控制策略（S24）。该方法可增强基于机器学习确定的风险控制策略与用户业务数据的关联密切性，提高风险控制策略的可靠性，适用性更强。

Description

基于预测模型的风险控制策略的确定方法及相关装置
本申请要求于2018年8月22日提交中国专利局、申请号为201810963590.6、申请名称为“基于预测模型的风险控制策略的确定方法及相关装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及通信技术领域，尤其涉及一种基于预测模型的风险控制策略的确定方法及相关装置。
背景技术
风险控制是风险管理者采用各种措施和/或方法来消灭或者减少风险事件发生的可能性，或者风险管理者采用各种措施和/或方法来减少风险事件发生时造成的损失。风险控制也可称为风控，风险控制过程中所采用的措施和/或方法也可称为风险控制策略。风险决策是指在多种不定因素的作用下，对两种或者两种以上的风险控制策略进行选择的过程。由于有不定因素存在，风险控制策略的实施所带来的损益值是无法预先确定的，因此需要一定的规则作为风险决策过程中选择风险控制策略的依据。在面对多种不定因素的情况下基于该规则从多种风险控制策略中选择相应的风险控制策略用于风险控制，风险决策所依据的规则称为风险决策规则。
现有技术中，风险控制类的风险决策***主要是基于专家制定的风险决策规则进行风险控制策略的选择，风险决策规则的制定完全依赖于专家的主观经验判断并制定，风采决策规则单一并且缺少理论支持和数据依据，随意性较强。同时，风险决策规则基于专家制定的随意性也使得基于风险决策规则确定的风险控制策略的可靠性差，适用范围小。
发明内容
本申请实施例提供一种基于预测模型的风险控制策略的确定方法及相关装置，可增强风险控制策略的确定方式的灵活性，提高风险控制策略的可靠性，适用性更高。
第一方面，本申请实施例提供了一种基于预测模型的风险控制策略的确定方法，包括：
获取第一用户数据，确定上述第一用户数据所关联的目标业务；
根据上述目标业务的目标业务类型确定出上述目标业务所关联的目标风险决策规则生成模型；
基于上述目标风险决策规则生成模型生成上述第一用户数据对应的目标风险决策规则，上述目标风险决策规则生成模型基于上述目标业务类型对应的样本数据训练得到，上述样本数据中至少包括第一规则样本数据和第二规则样本数据，上述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，上述第二规则样本数据中包括第二风险决策规则及其对应的第二样本用户数据；
根据上述第一用户数据和上述目标风险决策规则确定出对上述目标业务进行风险控制的目标风险控制策略。
第二方面，本申请实施例提供了一种基于预测模型的风险控制策略的确定装置，该确定装置包括：
获取单元，用于获取第一用户数据，确定上述第一用户数据所关联的目标业务；
模型确定单元，用于根据上述获取单元获取的上述目标业务的目标业务类型确定出上 述目标业务所关联的目标风险决策规则生成模型；
规则生成单元，用于基于上述模型确定单元确定的上述目标风险决策规则生成模型生成上述获取单元获取的上述第一用户数据对应的目标风险决策规则，上述目标风险决策规则生成模型基于上述目标业务类型对应的样本数据训练得到，上述样本数据中至少包括第一规则样本数据和第二规则样本数据，上述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，上述第二规则样本数据中包括第二风险决策规则及其对应的第二样本用户数据；
策略确定单元，用于根据上述获取单元获取的上述第一用户数据和上述规则生成单元确定的上述目标风险决策规则确定出对上述目标业务进行风险控制的目标风险控制策略。
第三方面，本申请实施例提供了一种终端设备，该终端设备包括处理器和存储器，该处理器和存储器相互连接。该存储器用于存储支持该终端设备执行上述第一方面和/或第一方面任一种可能的实现方式提供的方法的计算机程序，该计算机程序包括程序指令，该处理器被配置用于调用上述程序指令，执行上述第一方面和/或第一方面任一种可能的实施方式所提供的方法。
第四方面，本申请实施例提供了一种计算机可读存储介质，该计算机可读存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令当被处理器执行时使该处理器执行上述第一方面和/或第一方面任一种可能的实施方式所提供的方法。
采用本申请实施例中，基于目标风险决策规则生成模型生成风险决策规则的可靠性更强，风险控制策略的确定方式更灵活性，可增强以目标风险决策规则生成的风险控制策略的可靠性，从而可提高目标业务的风险控制有效性，适用性更高。
附图说明
为了更清楚地说明本申请实施例技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍。
图1是本申请实施例提供的基于预测模型的风险控制策略的确定方法的一流程示意图；
图2是本申请实施例提供的风险决策规则生成模型的构建方法的流程示意图；
图3是本申请实施例提供的基于预测模型的风险控制策略的确定方法的另一流程示意图；
图4是本申请实施例提供的基于预测模型的风险控制策略的确定装置的结构示意图；
图5是本申请实施例提供的终端设备的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述。
本申请实施例提供的基于预测模型的风险控制策略的确定方法(为方便描述，可简称本申请实施例提供的方法)可适用于投资、银行、保险、证券以及电商等多个应用领域中的任一应用领域。在各个应用领域中，所涉及的应用场景可包括但不限于登录、注册、贷前、贷中、贷后、节假日活动或者促销活动等。本申请实施例提供的方法可适用于上述任一应用场景的任一业务类型的风险决策规则的生成。这里，上述风险决策规则为风险决策过程中，具体选择哪一种风险控制策略所依据的规则。为方便描述，下面将以风险决策规则为例进行描述。针对不同应用场景、不同业务所关联的用户数据等，可构建不同的风险决策规则生成模型。不同的风险决策规则生成模型可适用于不同的应用场景，以及各种应用场景下的多种业务的风险决策规则的生成，灵活性高。基于风险决策规则生成模型可根 据具体应用场景下的具体业务的实时用户数据输出针对该业务进行风险控制所依据的风险决策规则，从而可提高用于风险决策规则与业务的关联密切性，适用性强。同时风险决策规则基于用户数据生成，可靠性强，增强基于该风险决策规则确定的风险控制策略进行的风险决策的准确性，适用性更强。这里，上述业务具体可为投资、银行、保险、证券以及电商等多个应用领域中，向用户提供的各种业务，例如，投保以及贷款等。对应的，以投保为例，在该业务下对应的应用场景可包括但不限于账户注册、账户登录、投保申请、投保审批、保单生成以及维持等。其中，上述应用场景仅是举例，而非穷举，具体可根据实际应用场景确定，在此不做限制。
在本申请实施例中，为了提高风险控制策略的确定准确率，本申请实施例提供的方法可基于从多种数据获取路径获取得到各个业务类型的样本数据构建各业务类型对应的风险决策规则生成模型，基于风险决策规则生成模型生成的风险决策规则确定风险控制策略。在本申请实施例中，基于不同业务类型所关联的样本数据可构建得到适用于生成各业务类型对应的风险决策规则的风险决策规则生成模型，换句话说，针对多种业务类型所关联的样本数据可训练得到一个风险决策规则生成模型。或者，可选的，基于不同业务类型所关联的样本数据可构建得到一个风险决策规则生成模型中的多组网络套数，其中，该风险决策规则生成模型的一组网络参数可适用于生成一种业务类型所关联的风险决策规则。为方便描述，本申请实施例将以目标业务为例进行说明，上述风险决策规则生成模型可以目标业务所关联的风险决策规则生成模型为例进行说明，在此不做限制。
本申请实施例提供的方法可基于具体应用场景中目标业务的历史风险控制记录中相关的样本数据构建风险决策规则生成模型。基于目标业务的用户数据库中存储的目标业务的历史风险控制记录，可从历史风险控制记录中确定出针对目标业务所采取的至少两种风险决策规则。从用户数据库中获取至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为目标业务类型对应的样本数据以构建目标业务对应的风险决策规则生成模型。可选的，基于大数据分析可从其他业务的用户群数据库中获取各业务的历史风险控制记录，从各业务的历史风险控制记录中确定出针对各业务所采取的至少两种风险决策规则，其他业务包括与目标业务为相同类型业务的一个或者多个业务。从各业务的用户数据库中获取至少两种风险决策规则中各风险决策规则对应的样本用户数据，该样本用户数据也可用于构建风险决策规则生成模型，具体可根据实际应用场景确定，在此不做限制。其中，上述样本用户数据包括但不限于用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息，具体可根据实际应用场景确定，在此不做限制。
在本申请实施例中，风险决策规则既可表示与用户数据关联的一种规则，也可表示与用户数据关联的多种规则中各规则组成的组合规则。为方便描述，上述一个规则和/或多种规则组成的组合规则也可称为关联规则，具体可根据实际应用场景确定，在此不做限制。对应的，上述风险决策关规则生成模型可以称为关联规则模型，在此不做限制。下面将对任一业务对应的风险决策规则生成模型的构建，以及基于构建的该风险决策规则生成模型输出该业务对应的风险决策规则的实现方式为例，对本申请实施例提供的实现方式进行描述。为方便描述，这里的任一业务可以目标业务为例进行说明。对应的，目标业务对应的风险决策规则生成模型可以目标风险决策规则生成模型为例进行说明。
下面将结合图1至图5分别对本申请实施例提供的方法及装置进行说明。本申请实施例提供的方法中可包括用于目标风险决策规则生成模型的构建、基于目标风险决策规则生成模型的风险决策规则的生成以及基于风险决策规则确定风险决策策略等数据处理阶段。其中，上述各个数据处理阶段的实现方式可参见如下图1至图3所示的实现方式。
参见图1，图1是本申请实施例提供的基于预测模型的风险控制策略的确定方法的一流程示意图。为方便描述，下面以目标业务对应的目标风险决策规则生成模型为例进行说明，下面不再赘述。本申请实施例提供的方法可包括如下步骤S1至S3：
S1、构建目标业务所关联的目标风险决策规则生成模型。
在一些可行的实施方式中，目标风险决策规则生成模型的构建可包括目标风险决策规则生成模型的建模数据(即样本数据)采集，目标风险决策规则生成模型的训练，以及目标风险决策规则生成模型的测试等数据处理阶段。请一并参见图2，图2是本申请实施例提供的风险决策规则生成模型的构建方法的流程示意图。本申请实施例提供的风险决策规则生成模型的构建可通过如下步骤S11至S13提供的实现方式进行说明。
S11、目标风险决策规则生成模型的建模数据采集。
在一些可行的实施方式中，目标风险决策规则生成模型的建模数据可来源与目标业务的用户数据库中存储的用户数据，也可来源于大数据分析得到与目标业务关联的用户数据，具体可根据实际应用场景确定，在此不做限制。基于目标业务的用户数据库中存储的目标业务的历史风险控制记录，可从历史风险控制记录中确定出针对目标业务所采取的至少两种风险决策规则。从用户数据库中获取至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为目标业务类型对应的建模数据(即样本数据)以构建目标业务对应的目标风险决策规则生成模型。可选的，基于大数据分析可从其他业务的用户群数据库中获取各业务的历史风险控制记录，从各业务的历史风险控制记录中确定出针对各业务所采取的至少两种风险决策规则，其他业务包括与目标业务为相同类型业务的一个或者多个业务。从各业务的用户数据库中获取至少两种风险决策规则中各风险决策规则对应的样本用户数据，该样本用户数据也可用于构建目标风险决策规则生成模型，具体可根据实际应用场景确定，在此不做限制。
在一些可行的实施方式中，，上述用于构建目标业务的目标风险决策规则生成模型的建模数据可采集某一个时间段内目标业务的风险决策规则所对应的样本用户数据，包括但不限于目标业务的样本用户行为数据以及跟这些样本用户行为数据所关联的终端设备标识信息等，在此不做限制。其中，上述目标风险决策规则生成模型的建模数据的采集时间段可由预定义的该目标业务的风险决策规则的有效时长，或者政策变化(地方政策的变化)等因素确定，在此不做限制。例如，若基于目标风险决策规则生成模型生成的风险决策规则的有效时长为一天，则目标风险决策规则生成模型的建模数据的采集时段可为当前时间的前一天所记录的样本用户数据。若基于目标风险决策规则生成模型生成的目标风险决策规则的有效时长为一周，则目标风险决策规则生成模型的建模数据的采集时段可为当前时间的前一周的时间内所记录的样本用户数据。这里，通过预定义关联规则的有效时长，在当前使用的风险决策规则失效时则可基于目标风险决策规则生成模型重新生成下一个有效时长周期所适用的风险决策规则，从而可实现风险决策规则的定期更新，从而可保证风险决策规则的时效性，增强基于风险决策规则进行风险控制策略的选择等风险决策的可靠性。面对目标业务的突发风险事项，或者政策变化等特殊情况，则可通过实时更新的样本用户数据生成相应的风险决策规则，从而可适用于应急机制，进一步保障基于目标风险决策规则生成模型生成的风险决策规则进行风险决策的可靠性，适用性更强。
在一些可行的实施方式中，上述建模数据的采集过程中所采集的样本用户数据包括但不限于用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息，在此不做限制。其中，上述用户的页面操作数据、用户的业务访问时长以及用户的业务访问频率等数据也可称为样本用户行为数据，在此不做限制。其中，上述用户的终端设备标识信息可为上述各种表现形式 的样本用户行为数据所关联的终端设备信息，换句话说，上述用户设备的终端设备标识信息可以为用户访问目标业务时所使用的终端设备的设备信息等，在此不做限制。例如，上述终端设备信息可以是用户登录办理目标业务的页面或者注册目标业务的业务账号时所使用的终端设备的设备信息等，在此不做限制。其中，上述终端设备的设备信息可包括终端设备的媒体访问控制(medium access control，MAC)地址、国际移动设备标识(international mobile equipment identity,IMEI)、互联网协议(internet protocol，IP)地址、直接内向拨号(direct inward dialling,DID)、用户所使用终端设备的显示屏分辨率、以及终端设备所绑定的用户联系方式(手机号等)等，在此不做限制。上述用户所处地域信息可以是用户登录办理目标业务的页面或者注册目标业务的业务账号时所处的地理位置，对应的也可以是用户登录办理目标业务的页面或者注册目标业务的业务账号时所使用的终端设备的定位地理位置等等，具体可根据实际应用场景确定，在此不做限制。
在一些可行的实施方式中，上述建模数据中包括至少两个风险决策规则对应的样本用户数据，其中，至少包括第一规则样本数据和第二规则样本数据。其中，上述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，上述第二规则样本数据中包括第二风险决策规则及其对应的第二样本用户数据，具体可根据实际应用场景确定，在此不做限制。为方便描述下面可以以第一风险决策规则对应的第一样本用户数据和第二规则对应的第二样本用户数据为例进行说明。获取得到上述建模数据之后，则可对采集到的上述建模数据进行数据清洗、特征筛选，最终生成用于构建风险决策规则生成模型所需的特征数据。由于上述建模数据为目标业务的样本数据，是针对目标业务进行风险控制所采用的历史风险策略规则所对应的样本用户数据，换句话说，针对目标业务的样本用户数据所采用的风险决策规则是已知参数，比如第一风险决策规则和/或第二风险决策规则。因此，对于目标业务的样本用户数据可为已有标注数据，也可为可自定义标注数据。其中，已有标注或自定义标注均可用于标记样本用户数据对应的风险决策规则，可以理解为风险决策规则的标签。因此，对于样本用户数据，其所对应的风险决策规则也为已知数据。在风险决策规则生成模型的训练阶段，目标业务的样本用户数据对应的特征数据，以及这部分样本用户数据的标注(即风险决策规则的标签)均为风险决策规则生成模型训练的建模数据。
在一些可行的实施方式中，在目标风险决策规则生成模型的训练过程中，不同数据类型和/或数据内容的建模数据可对应训练得到适用于输出风险决策规则的目标风险决策规则生成模型。换句话说，就是建模数据中包括目标业务对应的某一风险决策规则对应的样本用户数据，则可通过这部分的建模数据训练得到一套网络参数，使得具备这样一套网络参数的目标风险决策规则生成模型可根据输入的用户数据对应输出风险决策规则的能力。其中，上述建模数据中包括的风险决策规则可通过标签或者阈值的方式存在。基于建模数据中不同风险决策规则可采用不同取值的标签来标记。例如，假设风险决策规则在建模数据中存在的形式是标签，则可通过标签的取值0、1、2或者3分别标记风险决策规则1、风险决策规则2、风险决策规则3以及风险决策规则4，从而可实现通过包含多种风险决策规则的建模数据训练得到多套网络参数，从而可使得目标风险决策规则生成模型可对输入的目标业务的用户数据预测得到目标风险决策规则。
在一些可行的实施方式中，上述目标风险决策规则生成模型的建模数据中除了风险决策规则之外的数据可称为用户数据。例如，对于风险决策规则为第一风险决策规则的样本用户数据，其中除了包括第一风险决策规则之外，还有针对目标业务采用第一风险决策规则进行风险控制的用户数据(为方便描述可以第一样本用户数据为例进行说明)。同理，对于风险决策规则为第二风险决策规则的样本用户数据，其中除了包括第二风险决策规则之外，还有针对目标业务采用第二风险决策规则进行风险控制的用户数据(为方便描述可以 第二样本用户数据为例进行说明)，下面不再赘述。上述目标风险决策规则生成模型的建模数据中包括第一样本用户数据和第二样本用户数据在内的任一风险决策规则对应的样本用户数据包括但不限于用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息。
在一些可行的实施方式中，基于上述各风险决策规则对应的样本用户数据所组成的建模数据，可构建样本用户特征对。在样本用户特征的构建过程中，各个风险决策规则可以标签的形式标记，并基于该标签生成该风险决策规则对应的样本用户特征。例如，第一风险决策规则可基于标签1标记，第二风险决策规则可基于标签2标记，进而可基于标签1生成第一风险决策规则对应的特征(例如一个字符“0”表示的特征)，基于标签2生成第二风险决策规则对应的特征(例如一个字符“1”表示的特征)，在此不做限制。再比如，一个样本用户特征对中包括第一风险决策规则对应的第一样本用户特征和第二风险决策规则对应的第二样本用户特征。其中，上述第一风险决策规则对应的第一样本用户特征可由第一风险决策规则对应的第一样本用户数据构建。对应的，上述第二风险决策规则对应的第二样本用户特征可由上述第二风险决策规则对应的第二样本用户数据构建。换句话说，可以理解，上述第一风险决策规则对应的第一样本用户特征可作为样本用户特征对中的正样本特征，上述第二风险决策规则对应的第二样本用户特征可作为样本用户特征对中的负样本特征(反之也可)，进而可基于一正一负的样本特征训练风险决策规则预测的初始网络模型以得到具备预测风险决策规则为第一风险决策规则或者第二风险决策规则的能力的目标风险决策规则生成模型。其中，上述基于目标风险决策规则生成模型预测输出风险决策规则为第一风险决策规则或者第二风险决策规则仅是示例，包括但不限于第一风险决策规则和第二风险决策规则，具体可根据实际应用场景确定，在此不做限制。
在一些可行的实施方式中，基于任一风险决策规则对应的样本用户数据，按照预设的用户数据抽象规则可得到各风险决策规则对应的样本用户数据的抽象特征表示，进而可根据该各风险决策规则对应的样本用户数据的抽象特征组成得到该风险决策规则对应的样本用户特征。例如，对于第一样本用户特征可采用一个多字符的特征向量表示，该特征向量中可由六个部分特征组成。其中，上述六个部分特征可包括第一样本用户数据中的用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息的六部分的用户数据抽象得到。其中，上述六个部分特征中任一部分特征均可由一个或者多个字符、一组或者多组字符，和/或一个或者多个维度的字符等字符组成，在此不做限制。例如，假设上述用户的业务账号信息包括业务账号、业务账号所绑定的联系人、业务账号所绑定的联系方式以及业务账号所属业务地域信息的4个维度的信息，则可采用4个字符(或者4组字符或者4个维度的字符等，在此不做限制)用于表示用户的业务账号信息所抽象出来的特征。其中，上述4个字符中每个字符可表示一个维度的信息。其中，每个维度的信息可分别进行分类，并采用不同的标识(例如0或1)标记不同的类别的信息，进而可得到各个维度的信息对应的标识，从而可将各个维度的信息对应的标识组合得到包含4个字符的业务账号特征。例如，假设业务账号所绑定的联系人有某一个人则可采用一个标识进行标记，若业务账号所绑定的联系人为空，则可采用另外一个标识进行标记，进而可在上述业务账号所绑定的联系人对应的字符的特征。以此类推，可确定上述4个字符的业务账号特征中各个字符的特征，从而可得到4个字符的业务账号特征。这里，对于上述4个维度的信息中缺省信息的维度，该维度所对应的字符可填充为空等以构建该维度对应的特征，在此不做限制。
同理，假设用户的页面操作数据中包括页面操作区域(可按照页面划分区域，并针对每个区域采用一个标识进行分类标识等)、页面操作时长(可按照时长分段进行分类标识等) 以及页面操作轨迹(可按照轨迹类型进行分类标识等)的3个维度的信息，则可采用3个字符(或者3组字符或者3个维度的字符等，在此不做限制)表示用户的页面操作数据所抽象出来的特征，为方便描述可简称为用户的页面操作特征等，在此不做限制。其中，上述用户的页面操作特征由上述用户的页面操作数据抽象得到的实现过程可参见上述用户的业务账号特征对应的实现方式，在此不做限制。上述用户的业务访问时长可按照时长进行分类标识，因此可采用1个字符(或者1组字符或者1个维度的字符等，在此不做限制)表示用户的业务访问时长所抽象出来的特征，为方便描述可简称为用户的业务访问时长特征等，在此不做限制。同理，上述用户的业务访问频率(可按照频率分段进行分类标识等)、用户的终端设备标识信息(可按照终端设备标识信息的数量进行分类标识等)以及用户所处地域信息(可按照地域进行分类标识等)等用户数据，可采用一个或者多个字符表示其所抽象出来的特征，为方便描述可分别以用户的业务访问频率特征、用户的终端设备标识特征以及用户所处地域特征为例进行说明。其中，上述各个用户数据所抽象得到相应的特征的实现过程可参见上述用户的业务账号特征对应的实现方式，在此不做限制。
S12，目标风险决策规则生成模型的训练。
基于上述步骤S21获取得到目标风险决策规则生成模型的建模数据所构建的样本用户特征对之后，则可将上述样本用户特征对输入目标风险决策规则生成模型的初始网络模型中，通过上述初始网络模型对输入的样本用户特征对中包括的样本用户数据特征及其对应的风险决策规则的标签特征进行学习，得到具备输出任一用户数据特征对应的风险决策规则的能力的目标风险决策规则生成模型。这里，上述目标风险决策规则生成模型的初始网络模型可采用反向传播(back propagation，BP)神经网络模型，或者其他更多类型的神经网络模型，在此不做限制。其中，上述目标风险决策规则生成模型的激活函数可为sigmoid函数等，具体可根据实际应用场景确定，在此不做限制。可选的，在一些可行的实施方式中，上述目标风险决策规则生成模型的输出是各风险决策规则对应的标签或者阈值，进而可基于各风险决策规则对应的标签或者阈值确定具体的风险决策规则，具体可根据实际应用场景确定，在此不做限制。
S13，目标风险决策规则生成模型的测试。
在一些可行的实施方式中，目标风险决策规则生成模型构建完成之后，可在上述目标风险决策规则生成模型的建模数据的采集所选择的时间段之后，选择距离当前时间最近的一段时间内针对目标业务进行风险控制对应的用户数据作为风险决策规则生成的测试数据。这里，上述风险决策规则生成的测试数据中至少包括一种风险决策规则的测试数据，进而可基于上述至少一种风险决策规则的测试数据构建至少一个样本用户特征。通过上述风险决策规则生成的测试数据构建样本用户测试特征，进而可基于构建的样本用户测试特征对目标风险决策规则生成模型的风险决策规则预测精度进行测试。其中，上述风险决策规则生成的测试数据中所包括的样本用户数据所包括的数据类型(或称数据维度)可与目标风险决策规则生成模型的建模数据中所包括的数据类型(或称数据维度)相同，在此不做限制，可保证目标风险决策规则生成模型的测试有效性，提高目标风险决策规则生成模型的测试结果的准确性，增强目标风险决策规则生成模型的适用性。
在一些可行的实施方式中，基于上述目标风险决策规则生成模型可对上述生成风险决策规则的测试数据所构建的样本用户测试特征进行学习并输出对应的风险决策规则。进而可根据上述目标风险决策规则生成模型生成的风险决策规则，结合上述测试数据对应已知的风险决策规则，计算目标风险决策规则生成模型的输出的损失值(例如风险决策规则对应的标签和/或阈值的差值等，在此不做限制)。上述目标风险决策规则生成模型的输出的损失值可反馈至目标风险决策规则生成模型中，基于上述损失值对上述目标风险决策规则 生成模型的网络参数进行修正等优化处理，可提高目标风险决策规则生成模型的预测精度，适用性更强。
在本申请实施例中，目标风险决策规则生成模型的训练过程中，训练样本来自于多种风险决策规则对应的样本用户数据，基于多种风险决策规则对应的样本用户数据可构建用于模型训练的样本用户特征对，用于训练目标风险决策规则生成模型使得模型具备针对任一用户数据对应的用户特征对应输出各风险决策规则的能力，从而可基于目标风险决策规则生成模型预测任一用户数据对应的风险决策规则，实现了风险决策规则的有效预测，可增强基于目标风险决策规则生成模型生成风险决策规则的可靠性更强，同时基于目标风险决策规则生成模型所生成的风险决策规则有着用户数据的数据支持，进一步提高了目标风险决策规则的适用性。
S2、基于目标风险决策规则生成模型的目标风险决策规则的生成。
在一些可行的实施方式中，基于目标风险决策规则生成模型生成目标风险决策规则的过程中，可包括用户数据的获取、目标风险决策规则生成模型的确定，以及目标风险决策规则的生成等数据处理阶段。进一步的，基于目标风险决策规则可确定出对目标业务进行风险控制的目标风险控制策略。请一并参见图3，图3是本申请实施例提供的基于预测模型的风险控制策略的确定方法的另一流程示意图。本申请实施例提供的风险策略的确定方法可包括如下步骤S21至S24。
S21、获取第一用户数据，确定上述第一用户数据所关联的目标业务。
在一些可行的实施方式中，在用户需要在目标业务对应的浏览器和/或客户端的业务操作页面上办理目标业务时，可基于上述浏览器和/或客户端的业务操作页面上的用户操作获取用户数据(为方便描述可以第一用户数据为例进行说明)，并可根据上述第一用户数据确定目标业务。其中，上述第一用户数据中所包括的数据类型包括但不限于：用户的业务账号信息、用户的页面操作数据、用户的终端设备标识信息以及用户所处地域信息，在此不做限制。可选的，上述第一用户数据中所包括的页面操作数据可包括页面启动位置、页面链接输入、页面连接路径、页面操作时长以及页面操作频率等，在此不做限制。此外，基于上述页面操作数据还可计算用户的业务访问时长以及用户的业务访问频率等衍生数据，从而可将衍生数据确定为第一用户数据的一部分，具体可根据实际应用场景确定，在此不做限制。
在一些可行的实施方式中，基于上述第一用户数据中的用户的业务账号信息、页面启动位置、页面链接输入以及页面连接路径等信息可确定上述第一用户数据所关联的目标业务。例如，当用户需要登陆某一个业务的业务账号或者注册某一个业务的业务账号时，可通过鼠标或者手指等途径点击用于办理该业务的浏览器和/或客户端的图标，从而可打开浏览器和/或客户端的业务操作页面。在上述业务操作页面上输入已有的业务账号信息，或者填写待注册的业务账号信息。其中，当用户在业务操作页面上输入已有的业务账号信息，或者填写待注册的业务账号信息时，可采集上述用户在业务操作界面上操作的用户数据，包括用户的业务账号信息以及页面操作数据等等。基于上述用户的业务账号信息和/或页面操作数据可确定出第一用户数据所关联的业务，即目标业务，从而可基于目标业务对应的目标风险决策规则生成模型确定出针对目标业务进行风险决策规则。
可选的，在一些可行的实施方式中，在用户在业务操作界面上述操作的过程中，可实时监控用户的业务访问时长，还可结合用户访问的历史记录确定用户的业务访问频率，具体可根据实际应用场景确定，在此不做限制。此外，基于用户在业务操作界面上的操作可实时采集用户所使用的终端设备的标识信息以及用户所处地域信息等等，具体可根据实际应用场景确定，在此不做限制。其中，上述第一用户数据的数据类型和/或数据内容可参见 上述步骤S11至S13中各个步骤所提供的实现方式中样本用户数据中任一样本用户数据包括的数据类型和/或数据内容，在此不再赘述。
S22、根据上述目标业务的目标业务类型确定出上述目标业务所关联的目标风险决策规则生成模型。
在一些可行的实施方式中，基于上述第一用户数据可确定出目标业务，将目标业务的业务类型(即目标业务类型)与风险决策规则生成模型集合中包括的各风险决策规则生成模型所关联的业务类型进行匹配，从上述风险决策规则生成模型集合中确定出目标业务类型对应的风险决策规则生成模型。可以理解，这里风险决策规则生成模型集合中还可包括目标业务类型之外的其业务类型所关联的其他风险决策规则生成模型，如上述步骤S11至S13所提供的实现方式可知，上述其他风险决策规则生成模型由其他业务类型关联的样本数据训练得到。基于目标业务类型从上述风险决策规则生成模型集合中匹配得到目标业务所关联的风险决策规则生成模型，进而可基于该风险决策规则生成模型对实时采集到的用户数据(即第一用户数据)进行学习以生成上述第一用户数据对应的目标风险决策规则。可选的，终端设备也可基于目标业务类型从上述风险决策规则生成模型集合中匹配得到目标业务类型所关联的网络参数的风险决策规则生成模型，进而可基于具备该网络参数的该风险决策规则生成模型对实时采集到的第一用户数据进行学习以生成上述第一用户数据对应的目标风险决策规则。具体可根据实际应用场景确定，在此不做限制。
S23、基于上述目标风险决策规则生成模型生成所述第一用户数据对应的目标风险决策规则。
在一些可行的实施方式中，上述基于目标业务的第一用户数据可构建第一用户特征的实现方式可与上述目标风险决策规则生成模型的建模数据中样本用户特征的构建方式相同，具体可参见上述步骤S21所提供的实现方式，在此不再赘述。基于上述第一用户数据构建得到第一用户特征之后，则可将上述第一用户特征输入目标风险决策规则生成模型，基于上述目标风险决策规则生成模型对上述第一用户特征进行学习，生成上述第一用户数据对应的目标风险决策规则。可选的，目标风险决策规则生成模型对第一用户特征进行学习之后可对应输出第一用户数据对应的目标风险决策规则的标签或者阈值，进而可基于上述目标风险决策规则的标签或者阈值确定出目标风险决策规则。例如，当目标风险决策规则生成模型对第一用户特征进行学习之后可对应输出的风险决策规则标签为标签1或者风险决策规则阈值为阈值1，则可确定针对第一用户数据生成的目标风险决策规则为是第一风险决策规则。或者当目标风险决策规则生成模型对第一用户特征进行学习之后可对应输出的风险决策规则标签为标签2或者风险决策规则阈值为阈值2，则可确定针对第一用户数据生成的目标风险决策规则是第二风险决策规则。由此类推，可基于目标风险决策规则生成模型输出的风险决策规则标签确定第一用户数据对应的目标风险决策规则。
在本申请实施例中，基于第一用户数据筛选得到并输入目标风险决策规则生成模型的第一用户特征对应的数据类型和/或数据内容，可与目标风险决策规则生成模型的训练和/或测试阶段所输入的建模数据和/或测试数据构建的用户特征对应的数据类型和/或数据内容相同。在目标风险决策规则生成模型的训练阶段、测试阶段以及使用阶段所采集以及筛选的数据类型和/或数据内容相同，从而可更好地利用该目标风险决策规则生成模型对输入的用户特征进行学习并输出相应的风险决策规则，可增加目标风险决策规则生成模型生成风险决策规则的准确率和可靠性，适用性更强。
在一些可行的实施方式中，上述目标风险决策规则为包括上述第一风险决策规则和/或上述第二风险决策规则在内的一个或者多个规则的组合，上述一个或者多个规则包括：
规则1：当用户的终端设备标识信息中所包含的设备标识大于或者等于设备标识阈值 时，采用第一风险控制策略进行业务的风险控制。当用户的终端设备标识信息中所包含的设备标识小于设备标识阈值时，采用第二风险控制策略进行业务的风险控制。
例如，上述规则1的判断条件可为：IP个数大于IP个数阈值。
上述规则1表示：若上述第一用户数据中同一用户所使用的IP个数大于IP个数阈值，则基于规则1进行风险决策时可选择第一风险控制策略(例如风险控制策略1)，对该用户数据对应的目标业务进行风险控制。若用户数据中同一用户所使用的IP个数小于或者等于IP个数阈值，则基于规则1进行风险决策时可选择第二风险控制策略(例如风险控制策略2)，对该用户数据对应的用户行为进行风险控制。其中，上述第一风险控制策略和/或第二风险控制策略仅是举例，在此不做限制。
可选的，上述规则1的判断条件也可为：MAC个数大于MAC个数阈值。
这里，上述规则1可表示：若上述第一用户数据中同一用户所使用终端设备的MAC个数大于MAC个数阈值，则基于规则1进行风险决策时可选择第一风险控制策略，对该用户数据对应的目标业务进行风险控制。若用户数据中同一用户所使用终端设备的MAC小于或者等于MAC个数阈值，则基于规则1进行风险决策时可选择第二风险控制策略，对该用户数据对应的目标业务进行风险控制。
可选的，上述规则1的判断条件中的阈值(包括IP个数阈值和/或MAC个数阈值)可由目标风险决策规则生成模型对样本用户数据进行学习之后生成，从而可增加基于目标风险决策规则生成模型生成目标风险决策规则的适用性。同理，下面规则2、规则3和规则4等规则的判断条件中的阈值也可有目标风险决策规则生成模型生成，下面不再赘述。
规则2：当用户的业务访问时长大于或者等于访问时长阈值时，采用第三风险控制策略进行业务的风险控制。当用户的业务访问时长小于访问时长阈值时，采用第四风险控制策略进行业务的风险控制。
例如，上述规则2的判断条件可为：单位时间的用户访问量大于用户访问量阈值。
上述规则2可表示：若基于上述第一用户数据确定单位时间内同一用户或者多个用户对目标业务的用户访问量大于用户访问量阈值，则基于规则2进行风险决策时可选择第三风险控制策略(例如风险控制策略3等)，对第一用户数据对应的目标业务进行风险控制。若基于上述第一用户数据确定单位时间内同一用户或者多个用户对目标业务的用户访问量小于或者等于用户访问量阈值，则基于规则2进行风险决策时可选择第四风险控制方式(例如风险决策策略4)，对该第一用户数据对应的目标业务进行风险控制。其中，上述第三风险控制策略和第四风险控制策略仅是举例，在此不做限制。
规则3：当用户的业务访问频率大于或者等于访问频率阈值时，采用第五风险控制策略进行业务的风险控制。当用户的业务访问频率小于访问时长阈值时，采用第六风险控制策略进行业务的风险控制。
例如，上述规则3的判断条件可为：单位时间的业务访问频率大于业务访问频率阈值。
上述规则3可表示：若基于上述第一用户数据确定单位时间内同一用户或者多个用户对目标业务的业务访问频率大于业务访问频率阈值，则基于规则3进行风险决策时可选择第五风险控制策略(例如风险控制策略5等)，对第一用户数据对应的目标业务进行风险控制。若基于上述第一用户数据确定单位时间内同一用户或者多个用户对目标业务的业务访问频率小于或者等于业务访问频率阈值，则基于规则3进行风险决策时可选择第六风险控制方式(例如风险决策策略6)，对该第一用户数据对应的目标业务进行风险控制。其中，上述第五风险控制策略和第六风险控制策略仅是举例，在此不做限制。
可选的，上述目标风险决策规则也可为规则1、规则2以及规则3中的一个或者多个组合，为方便描述可以规则4为例进行说明。
规则4：当用户的终端设备标识信息中所包含的设备标识大于或者等于设备标识阈值，且用户的业务访问时长大于或者等于访问时长阈值时，采用第一风险控制策略进行业务的风险控制，否则采用第二风险控制策略进行业务的风险控制。
例如，规则4的判断条件可为：IP个数>IP个数阈值||MAC个数>MAC个数阈值||单位时间的用户访问量>用户访问量阈值。
这里，规则4可表示关联规则为上述规则1和规则2的组合，只有在上述规则1和规则2的条件判断同时成立时，基于规则4进行风险决策时可选择第一风险控制策略，对第一用户数据对应的目标业务进行风险控制，否则基于规则4进行风险决策时可选择第二风险控制策略，对第一用户数据对应的目标业务进行风险控制，在此不做限制。
S24、根据上述第一用户数据和上述目标风险决策规则确定出对上述目标业务进行风险控制的目标风险控制策略。
在一些可行的实施方式中，基于上述目标风险决策规则生成模型对第一用户数据进行学习并生成目标风险决策规则之后，可将第一用户数据中所包括的各项参数与目标风险决策规则中所包括的判断条件进行比对，从而可确定出对目标业务进行风险控制的目标风险控制策略。例如，若基于上述目标风险决策规则生成模型对第一用户数据进行学习并生成目标风险决策规则为规则1，基于上述第一用户数据中确定得到第一用户数据中所包括的IP个数大于目标风险决策规则生成模型所生成的IP个数阈值，则可确定目标风险控制策略为第一风险控制策略。同理，若目标风险决策规则为规则1之外的任意一个或者多个规则的组合，则可基于各个规则中所包括的判断条件，结合第一用户数据中所包括的数据类型确定出目标风险控制策略，具体可根据实际应用场景确定，在此不做限制。
S3、基于风险决策规则确定风险控制策略。
在一些可行的实施方式中，上述基于风险决策规则确定风险控制策略的实现方式可以基于目标风险决策规则确定目标风险控制策略为例，具体实现方式可参见上述步骤S24中，基于目标风险决策规则确定目标风险控制策略的实现方式，在此不再赘述。
在本申请实施例中，基于第一用户数据所关联的目标业务可确定出目标风险决策规则生成模型，基于目标风险决策规则生成模型可生成第一用户数据对应的目标风险决策规则。目标风险决策规则生成模型基于目标业务对应的样本数据训练得到，使得基于目标风险决策规则生成模型生成风险决策规则的可靠性更强，同时基于目标风险决策规则生成模型所生成的目标风险决策规则有着第一用户数据的数据支持，进一步提高了目标风险决策规则的适用性。此外，在本申请实施例中，基于目标风险决策规则可确定出对目标业务进行风险控制的目标风险控制策略，风险控制策略的确定方式更灵活性，同时目标风险决策规则的可靠性可增强以目标风险决策规则生成的风险控制策略的可靠性，从而可提高目标业务的风险控制有效性，适用性更高。
参见图4，图4是本申请实施例提供的基于预测模型的风险控制策略的确定装置的结构示意图。本申请实施例提供的基于预测模型的风险控制策略的确定装置包括：
获取单元41，用于获取第一用户数据，确定上述第一用户数据所关联的目标业务。
模型确定单元42，用于根据上述获取单元41获取的上述目标业务的目标业务类型确定出上述目标业务所关联的目标风险决策规则生成模型。
规则生成单元43，用于基于上述模型确定单元42确定的上述目标风险决策规则生成模型生成上述获取单元41获取的上述第一用户数据对应的目标风险决策规则，上述目标风险决策规则生成模型基于上述目标业务类型对应的样本数据训练得到，上述样本数据中至少包括第一规则样本数据和第二规则样本数据，上述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，上述第二规则样本数据中包括第二风险决策规则及 其对应的第二样本用户数据。
策略确定单元44，用于根据上述获取单元获取的上述第一用户数据和上述规则生成单元确定的上述目标风险决策规则确定出对上述目标业务进行风险控制的目标风险控制策略。
在一些可行的实施方式中，上述模型确定单元42用于：
将上述目标业务的目标业务类型与风险决策规则模型集合中包括的各风险决策规则模型所关联的业务类型进行匹配，从上述风险决策规则模型集合中确定出上述目标业务类型对应的目标风险决策规则生成模型；
其中，上述风险决策规则模型集合中还包括上述目标业务类型之外的其他业务类型所关联的其他风险决策规则模型，上述其他风险决策规则模型由上述其他业务类型对应的样本用户数据训练得到。
在一些可行的实施方式中，上述确认装置还包括：
模型构建单元45，用于获取用于构建上述目标风险决策规则生成模型的至少两种风险决策规则对应的样本数据，上述至少两种风险决策规则至少包括上述第一风险决策规则和上述第二风险决策规则，上述样本数据中至少包括上述第一规则样本数据和上述第二规则样本数据；
上述模型构建单元45，还用于根据上述至少两种风险决策规则对应的上述样本数据构建至少一个样本用户特征对，根据上述至少一个样本用户特征对构建目标风险决策规则生成模型。
在一些可行的实施方式中，上述模型构建单元45用于：
从上述目标业务的用户数据库中获取上述目标业务的历史风险控制记录，从上述历史风险控制记录中确定出针对上述目标业务所采取的至少两种风险决策规则；
从上述用户数据库中获取上述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为上述目标业务类型对应的样本数据。
在一些可行的实施方式中，上述模型构建单元45用于：
基于大数据分析从其他业务的用户群数据库中获取各业务的历史风险控制记录，从上述各业务的历史风险控制记录中确定出针对上述各业务所采取的至少两种风险决策规则，上述其他业务包括与上述目标业务为相同类型业务的一个或者多个业务；
从上述各业务的用户数据库中获取上述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为上述目标业务类型对应的样本数据。
在一些可行的实施方式中，上述第一用户数据和/或上述目标业务类型对应的样本数据中任一样本用户数据中所包含的数据类型包括：用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息中的一种或者多种；
其中，上述用户的业务账号信息和/或上述用户的页面操作数据用于确定业务和/或业务类型。
在一些可行的实施方式中，上述目标风险决策规则为包括上述第一风险决策规则和/或上述第二风险决策规则在内的一个或者多个规则的组合，上述一个或者多个规则包括：
规则1：当用户的终端设备标识信息中所包含的设备标识大于或者等于设备标识阈值时，采用第一风险控制策略进行业务的风险控制，当上述用户的终端设备标识信息中所包含的设备标识小于上述设备标识阈值时，采用第二风险控制策略进行业务的风险控制；
规则2：当用户的业务访问时长大于或者等于访问时长阈值时，采用第三风险控制策 略进行业务的风险控制，当上述用户的业务访问时长小于上述访问时长阈值时，采用第四风险控制策略进行业务的风险控制；
规则3：当用户的业务访问频率大于或者等于访问频率阈值时，采用第五风险控制策略进行业务的风险控制，当上述用户的业务访问频率小于上述访问时长阈值时，采用第六风险控制策略进行业务的风险控制。
具体实现中，上述基于预测模型的风险控制策略的确定装置可通过其内置的各个功能模块执行如上述图1至图3中各个步骤所提供的实现方式。例如，上述获取单元41可用于执行上述各个步骤中第一用户数据和/或样本用户数据的获取等操作，具体可参见上述各个步骤所提供的实现方式，在此不再赘述。上述模型确定单元42可用于执行上述各个步骤中基于业务类型确定风险决策规则生成模型的实现方式，上述规则生成模块43可用于执行上述各个步骤中基于风险决策规则生成模型生成风险决策规则的实现方式，上述策略确定单元44可用于执行上述各个步骤中基于风险决策规则和用户数据确定风险控制策略的实现方式，上述模型构建单元45可用于执行上述各个步骤中目标风险决策规则生成模型的构建中相关步骤所描述的实现方式，具体可参见上述各个步骤所提供的实现方式，在此不再赘述。
在本申请实施例中，目标风险决策规则生成模型基于目标业务对应的样本数据训练得到，基于目标风险决策规则生成模型生成风险决策规则的可靠性更强，提高了目标风险决策规则的适用性。基于目标风险决策规则可确定出对目标业务进行风险控制的目标风险控制策略，风险控制策略的确定方式更灵活性，可增强以目标风险决策规则生成的风险控制策略的可靠性，提高目标业务的风险控制有效性，适用性更高。
参见图5，图5是本申请实施例提供的终端设备的结构示意图。如图5所示，本实施例中的终端设备可以包括：一个或多个处理器501和存储器502。上述处理器501和存储器502通过总线503连接。存储器502用于存储计算机程序，该计算机程序包括程序指令，处理器501用于调用存储器502存储的程序指令，执行如下操作：
获取第一用户数据，确定上述第一用户数据所关联的目标业务；
根据上述目标业务的目标业务类型确定出上述目标业务所关联的目标风险决策规则生成模型；
基于上述目标风险决策规则生成模型生成上述第一用户数据对应的目标风险决策规则，上述目标风险决策规则生成模型基于上述目标业务类型对应的样本数据训练得到，上述样本数据中至少包括第一规则样本数据和第二规则样本数据，上述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，上述第二规则样本数据中包括第二风险决策规则及其对应的第二样本用户数据；
根据上述第一用户数据和上述目标风险决策规则确定出对上述目标业务进行风险控制的目标风险控制策略。
在一些可行的实施方式中，上述处理器501用于：
将上述目标业务的目标业务类型与风险决策规则模型集合中包括的各风险决策规则模型所关联的业务类型进行匹配，从上述风险决策规则模型集合中确定出上述目标业务类型对应的目标风险决策规则生成模型；
其中，上述风险决策规则模型集合中还包括上述目标业务类型之外的其他业务类型所关联的其他风险决策规则模型，上述其他风险决策规则模型由上述其他业务类型对应的样本用户数据训练得到。
在一些可行的实施方式中，上述处理器501还用于：
获取用于构建上述目标风险决策规则生成模型的至少两种风险决策规则对应的样本数 据，上述至少两种风险决策规则至少包括上述第一风险决策规则和上述第二风险决策规则，上述样本数据中至少包括上述第一规则样本数据和上述第二规则样本数据；
根据上述至少两种风险决策规则对应的上述样本数据构建至少一个样本用户特征对，根据上述至少一个样本用户特征对构建目标风险决策规则生成模型。
在一些可行的实施方式中，上述处理器501用于：
从上述目标业务的用户数据库中获取上述目标业务的历史风险控制记录，从上述历史风险控制记录中确定出针对上述目标业务所采取的至少两种风险决策规则；
从上述用户数据库中获取上述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为上述目标业务类型对应的样本数据。
在一些可行的实施方式中，上述处理器501用于：
基于大数据分析从其他业务的用户群数据库中获取各业务的历史风险控制记录，从上述各业务的历史风险控制记录中确定出针对上述各业务所采取的至少两种风险决策规则，上述其他业务包括与上述目标业务为相同类型业务的一个或者多个业务；
从上述各业务的用户数据库中获取上述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为上述目标业务类型对应的样本数据。
在一些可行的实施方式中，上述第一用户数据和/或上述目标业务类型对应的样本数据中任一样本用户数据中所包含的数据类型包括：用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息中的一种或者多种；
其中，上述用户的业务账号信息和/或上述用户的页面操作数据用于确定业务和/或业务类型。
在一些可行的实施方式中，上述目标风险决策规则为包括上述第一风险决策规则和/或上述第二风险决策规则在内的一个或者多个规则的组合，上述一个或者多个规则包括：
规则1：当用户的终端设备标识信息中所包含的设备标识大于或者等于设备标识阈值时，采用第一风险控制策略进行业务的风险控制，当上述用户的终端设备标识信息中所包含的设备标识小于上述设备标识阈值时，采用第二风险控制策略进行业务的风险控制；
规则2：当用户的业务访问时长大于或者等于访问时长阈值时，采用第三风险控制策略进行业务的风险控制，当上述用户的业务访问时长小于上述访问时长阈值时，采用第四风险控制策略进行业务的风险控制；
规则3：当用户的业务访问频率大于或者等于访问频率阈值时，采用第五风险控制策略进行业务的风险控制，当上述用户的业务访问频率小于上述访问时长阈值时，采用第六风险控制策略进行业务的风险控制。
应当理解，在一些可行的实施方式中，上述处理器501可以是中央处理单元(central processing unit，CPU)，该处理器还可以是其他通用处理器、数字信号处理器(digital signal processor，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现成可编程门阵列(field programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。该存储器502可以包括只读存储器和随机存取存储器，并向处理器501提供指令和数据。存储器502的一部分还可以包括非易失性随机存取存储器。例如，存储器502还可以存储设备类型的信息。
具体实现中，上述终端设备可通过其内置的各个功能模块执行如上述图1至图3中各 个步骤所提供的实现方式，具体可参见上述各个步骤所提供的实现方式，在此不再赘述。
在本申请实施例中，目标风险决策规则生成模型基于目标业务对应的样本数据训练得到，基于目标风险决策规则生成模型生成风险决策规则的可靠性更强，提高了目标风险决策规则的适用性。基于目标风险决策规则可确定出对目标业务进行风险控制的目标风险控制策略，风险控制策略的确定方式更灵活性，可增强以目标风险决策规则生成的风险控制策略的可靠性，提高目标业务的风险控制有效性，适用性更高。
本申请实施例还提供一种计算机可读存储介质，该计算机可读存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令被处理器执行时实现图1至图3中各个步骤所提供的基于预测模型的风险控制策略的确定方法，具体可参见上述各个步骤所提供的实现方式，在此不再赘述。
上述计算机可读存储介质可以是前述任一实施例提供的基于预测模型的风险控制策略的确定装置或者上述终端设备的内部存储单元，例如电子设备的硬盘或内存。该计算机可读存储介质也可以是该电子设备的外部存储设备，例如该电子设备上配备的插接式硬盘，智能存储卡(smart media card,SMC)，安全数字(secure digital,SD)卡，闪存卡(flash card)等。进一步地，该计算机可读存储介质还可以既包括该电子设备的内部存储单元也包括外部存储设备。该计算机可读存储介质用于存储该计算机程序以及该电子设备所需的其他程序和数据。该计算机可读存储介质还可以用于暂时地存储已经输出或者将要输出的数据。
本申请的权利要求书和说明书及附图中的术语“第一”、“第二”、“第三”、“第四”等是用于区别不同对象，而不是用于描述特定顺序。此外，术语“包括”和“具有”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、***、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。在本文中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置展示该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本文所描述的实施例可以与其它实施例相结合。

Claims (20)

  1. 一种基于预测模型的风险控制策略的确定方法，其特征在于，所述方法包括：
    获取第一用户数据，确定所述第一用户数据所关联的目标业务；
    根据所述目标业务的目标业务类型确定出所述目标业务所关联的目标风险决策规则生成模型；
    基于所述目标风险决策规则生成模型生成所述第一用户数据对应的目标风险决策规则，所述目标风险决策规则生成模型基于所述目标业务类型对应的样本数据训练得到，所述样本数据中至少包括第一规则样本数据和第二规则样本数据，所述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，所述第二规则样本数据中包括第二风险决策规则及其对应的第二样本用户数据；
    根据所述第一用户数据和所述目标风险决策规则确定出对所述目标业务进行风险控制的目标风险控制策略。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述目标业务的目标业务类型确定出所述目标业务所关联的目标风险决策规则生成模型包括：
    将所述目标业务的目标业务类型与风险决策规则模型集合中包括的各风险决策规则模型所关联的业务类型进行匹配，从所述风险决策规则模型集合中确定出所述目标业务类型对应的目标风险决策规则生成模型；
    其中，所述风险决策规则模型集合中还包括所述目标业务类型之外的其他业务类型所关联的其他风险决策规则模型，所述其他风险决策规则模型由所述其他业务类型对应的样本用户数据训练得到。
  3. 根据权利要求1或2所述的方法，其特征在于，所述方法还包括：
    获取用于构建所述目标风险决策规则生成模型的至少两种风险决策规则对应的样本数据，所述至少两种风险决策规则至少包括所述第一风险决策规则和所述第二风险决策规则，所述样本数据中至少包括所述第一规则样本数据和所述第二规则样本数据；
    根据所述至少两种风险决策规则对应的所述样本数据构建至少一个样本用户特征对，根据所述至少一个样本用户特征对构建目标风险决策规则生成模型。
  4. 根据权利要求3所述的方法，其特征在于，所述获取用于构建所述目标风险决策规则生成模型的至少两种风险决策规则对应的样本数据包括：
    从所述目标业务的用户数据库中获取所述目标业务的历史风险控制记录，从所述历史风险控制记录中确定出针对所述目标业务所采取的至少两种风险决策规则；
    从所述用户数据库中获取所述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为所述目标业务类型对应的样本数据。
  5. 根据权利要求3所述的方法，其特征在于，所述获取用于构建所述目标风险决策规则生成模型的至少两种风险决策规则对应的样本数据包括：
    基于大数据分析从其他业务的用户群数据库中获取各业务的历史风险控制记录，从所述各业务的历史风险控制记录中确定出针对所述各业务所采取的至少两种风险决策规则，所述其他业务包括与所述目标业务为相同类型业务的一个或者多个业务；
    从所述各业务的用户数据库中获取所述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为所述目标业务类型对应的样本数据。
  6. 根据权利要求1-5任一项所述的方法，其特征在于，所述第一用户数据和/或所述 目标业务类型对应的样本数据中任一样本用户数据中所包含的数据类型包括：用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息中的一种或者多种。
  7. 根据权利要求6所述的方法，其特征在于，所述目标风险决策规则为包括所述第一风险决策规则和/或所述第二风险决策规则在内的一个或者多个规则的组合，所述一个或者多个规则包括：
    规则1：当用户的终端设备标识信息中所包含的设备标识大于或者等于设备标识阈值时，采用第一风险控制策略进行业务的风险控制，当所述用户的终端设备标识信息中所包含的设备标识小于所述设备标识阈值时，采用第二风险控制策略进行业务的风险控制；
    规则2：当用户的业务访问时长大于或者等于访问时长阈值时，采用第三风险控制策略进行业务的风险控制，当所述用户的业务访问时长小于所述访问时长阈值时，采用第四风险控制策略进行业务的风险控制；
    规则3：当用户的业务访问频率大于或者等于访问频率阈值时，采用第五风险控制策略进行业务的风险控制，当所述用户的业务访问频率小于所述访问时长阈值时，采用第六风险控制策略进行业务的风险控制。
  8. 一种基于预测模型的风险控制策略的确定装置，其特征在于，所述确定装置包括：
    获取单元，用于获取第一用户数据，确定所述第一用户数据所关联的目标业务；
    模型确定单元，用于根据所述获取单元获取的所述目标业务的目标业务类型确定出所述目标业务所关联的目标风险决策规则生成模型；
    规则生成单元，用于基于所述模型确定单元确定的所述目标风险决策规则生成模型生成所述获取单元获取的所述第一用户数据对应的目标风险决策规则，所述目标风险决策规则生成模型基于所述目标业务类型对应的样本数据训练得到，所述样本数据中至少包括第一规则样本数据和第二规则样本数据，所述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，所述第二规则样本数据中包括第二风险决策规则及其对应的第二样本用户数据；
    策略确定单元，用于根据所述获取单元获取的所述第一用户数据和所述规则生成单元确定的所述目标风险决策规则确定出对所述目标业务进行风险控制的目标风险控制策略。
  9. 根据权利要求8所述的装置，其特征在于，所述模型确定单元用于：
    将所述目标业务的目标业务类型与风险决策规则模型集合中包括的各风险决策规则模型所关联的业务类型进行匹配，从所述风险决策规则模型集合中确定出所述目标业务类型对应的目标风险决策规则生成模型；
    其中，所述风险决策规则模型集合中还包括所述目标业务类型之外的其他业务类型所关联的其他风险决策规则模型，所述其他风险决策规则模型由所述其他业务类型对应的样本用户数据训练得到。
  10. 根据权利要求8或9所述的装置，其特征在于，所述确认装置还包括：
    模型构建单元，用于获取用于构建所述目标风险决策规则生成模型的至少两种风险决策规则对应的样本数据，所述至少两种风险决策规则至少包括所述第一风险决策规则和所述第二风险决策规则，所述样本数据中至少包括所述第一规则样本数据和所述第二规则样本数据；
    所述模型构建单元，还用于根据所述至少两种风险决策规则对应的所述样本数据构建至少一个样本用户特征对，根据所述至少一个样本用户特征对构建目标风险决策规则生成模型。
  11. 根据权利要求10所述的装置，其特征在于，所述模型构建单元用于：
    从所述目标业务的用户数据库中获取所述目标业务的历史风险控制记录，从所述历史风险控制记录中确定出针对所述目标业务所采取的至少两种风险决策规则；
    从所述用户数据库中获取所述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为所述目标业务类型对应的样本数据。
  12. 根据权利要求10所述的装置，其特征在于，所述模型构建单元用于：
    基于大数据分析从其他业务的用户群数据库中获取各业务的历史风险控制记录，从所述各业务的历史风险控制记录中确定出针对所述各业务所采取的至少两种风险决策规则，所述其他业务包括与所述目标业务为相同类型业务的一个或者多个业务；
    从所述各业务的用户数据库中获取所述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为所述目标业务类型对应的样本数据。
  13. 根据权利要求8-12任一项所述的装置，其特征在于，所述第一用户数据和/或所述目标业务类型对应的样本数据中任一样本用户数据中所包含的数据类型包括：用户的业务账号信息、用户的页面操作数据、用户的业务访问时长、用户的业务访问频率、用户的终端设备标识信息以及用户所处地域信息中的一种或者多种；
    其中，所述用户的业务账号信息和/或所述用户的页面操作数据用于确定业务和/或业务类型。
  14. 根据权利要求13所述的装置，其特征在于，所述目标风险决策规则为包括所述第一风险决策规则和/或所述第二风险决策规则在内的一个或者多个规则的组合，所述一个或者多个规则包括：
    规则1：当用户的终端设备标识信息中所包含的设备标识大于或者等于设备标识阈值时，采用第一风险控制策略进行业务的风险控制，当所述用户的终端设备标识信息中所包含的设备标识小于所述设备标识阈值时，采用第二风险控制策略进行业务的风险控制；
    规则2：当用户的业务访问时长大于或者等于访问时长阈值时，采用第三风险控制策略进行业务的风险控制，当所述用户的业务访问时长小于所述访问时长阈值时，采用第四风险控制策略进行业务的风险控制；
    规则3：当用户的业务访问频率大于或者等于访问频率阈值时，采用第五风险控制策略进行业务的风险控制，当所述用户的业务访问频率小于所述访问时长阈值时，采用第六风险控制策略进行业务的风险控制。
  15. 一种终端设备，其特征在于，包括处理器和存储器，所述处理器和存储器相互连接；
    所述存储器用于存储计算机程序，所述计算机程序包括程序指令，所述处理器被配置用于调用所述程序指令执行如下操作：
    获取第一用户数据，确定所述第一用户数据所关联的目标业务；
    根据所述目标业务的目标业务类型确定出所述目标业务所关联的目标风险决策规则生成模型；
    基于所述目标风险决策规则生成模型生成所述第一用户数据对应的目标风险决策规则，所述目标风险决策规则生成模型基于所述目标业务类型对应的样本数据训练得到，所述样本数据中至少包括第一规则样本数据和第二规则样本数据，所述第一规则样本数据中包括第一风险决策规则及其对应的第一样本用户数据，所述第二规则样本数据中包括第二风险决策规则及其对应的第二样本用户数据；
    根据所述第一用户数据和所述目标风险决策规则确定出对所述目标业务进行风险控制 的目标风险控制策略。
  16. 根据权利要求15所述的终端设备，其特征在于，所述处理器用于：
    将所述目标业务的目标业务类型与风险决策规则模型集合中包括的各风险决策规则模型所关联的业务类型进行匹配，从所述风险决策规则模型集合中确定出所述目标业务类型对应的目标风险决策规则生成模型；
    其中，所述风险决策规则模型集合中还包括所述目标业务类型之外的其他业务类型所关联的其他风险决策规则模型，所述其他风险决策规则模型由所述其他业务类型对应的样本用户数据训练得到。
  17. 根据权利要求15所述的终端设备，其特征在于，所述处理器还用于：
    获取用于构建所述目标风险决策规则生成模型的至少两种风险决策规则对应的样本数据，所述至少两种风险决策规则至少包括所述第一风险决策规则和所述第二风险决策规则，所述样本数据中至少包括所述第一规则样本数据和所述第二规则样本数据；
    根据所述至少两种风险决策规则对应的所述样本数据构建至少一个样本用户特征对，根据所述至少一个样本用户特征对构建目标风险决策规则生成模型。
  18. 根据权利要求15所述的终端设备，其特征在于，所述处理器用于：
    从所述目标业务的用户数据库中获取所述目标业务的历史风险控制记录，从所述历史风险控制记录中确定出针对所述目标业务所采取的至少两种风险决策规则；
    从所述用户数据库中获取所述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为所述目标业务类型对应的样本数据。
  19. 根据权利要求15所述的终端设备，其特征在于，所述处理器用于：
    基于大数据分析从其他业务的用户群数据库中获取各业务的历史风险控制记录，从所述各业务的历史风险控制记录中确定出针对所述各业务所采取的至少两种风险决策规则，所述其他业务包括与所述目标业务为相同类型业务的一个或者多个业务；
    从所述各业务的用户数据库中获取所述至少两种风险决策规则中各风险决策规则对应的样本用户数据，并将各风险决策规则及其对应的样本用户数据确定为所述目标业务类型对应的样本数据。
  20. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机程序，所述计算机程序包括程序指令，所述程序指令当被处理器执行时使所述处理器执行如权利要求1-7任一项所述的方法。
PCT/CN2018/123513 2018-08-22 2018-12-25 基于预测模型的风险控制策略的确定方法及相关装置 WO2020037918A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810963590.6A CN109034660B (zh) 2018-08-22 2018-08-22 基于预测模型的风险控制策略的确定方法及相关装置
CN201810963590.6 2018-08-22

Publications (1)

Publication Number Publication Date
WO2020037918A1 true WO2020037918A1 (zh) 2020-02-27

Family

ID=64628077

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/123513 WO2020037918A1 (zh) 2018-08-22 2018-12-25 基于预测模型的风险控制策略的确定方法及相关装置

Country Status (2)

Country Link
CN (1) CN109034660B (zh)
WO (1) WO2020037918A1 (zh)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112700329A (zh) * 2021-01-27 2021-04-23 永辉云金科技有限公司 一种风控规则引擎的响应方法和风控规则引擎
CN112767133A (zh) * 2021-01-26 2021-05-07 北京健康之家科技有限公司 业务决策方法及装置、存储介质、计算机设备
CN112950352A (zh) * 2021-02-08 2021-06-11 北京淇瑀信息科技有限公司 用户筛选策略生成方法、装置及电子设备
CN113657779A (zh) * 2021-08-20 2021-11-16 杭州时趣信息技术有限公司 一种动态配置化的风控决策方法、***、设备和存储介质
CN113673844A (zh) * 2021-08-04 2021-11-19 支付宝(杭州)信息技术有限公司 一种信息反馈方法、装置及设备
CN113781239A (zh) * 2021-09-10 2021-12-10 未鲲(上海)科技服务有限公司 一种策略确定方法、装置、电子设备以及存储介质
CN116823437A (zh) * 2023-06-14 2023-09-29 广东企企通科技有限公司 基于配置化风控策略的准入方法、装置、设备及介质

Families Citing this family (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109034660B (zh) * 2018-08-22 2023-07-14 平安科技（深圳）有限公司 基于预测模型的风险控制策略的确定方法及相关装置
CN109636607B (zh) * 2018-12-18 2024-03-15 平安科技（深圳）有限公司 基于模型部署的业务数据处理方法、装置和计算机设备
CN109767110A (zh) * 2019-01-04 2019-05-17 中国银行股份有限公司 一种风险控制***优化方法、装置、设备及存储介质
CN111861521A (zh) * 2019-04-26 2020-10-30 财付通支付科技有限公司 数据处理方法、装置、计算机可读介质及电子设备
CN110111903A (zh) * 2019-04-28 2019-08-09 中国疾病预防控制中心 一种突发公共卫生事件预测的方法及设备
CN117421702A (zh) * 2019-05-14 2024-01-19 创新先进技术有限公司 一种用户数据的处理方法和装置
CN110188940B (zh) * 2019-05-24 2023-09-05 深圳市腾讯计算机***有限公司 业务处理方法及装置
CN110348672A (zh) * 2019-05-24 2019-10-18 深圳壹账通智能科技有限公司 业务决策方法、装置、计算设备和计算机可读存储介质
CN110378780A (zh) * 2019-06-17 2019-10-25 深圳壹账通智能科技有限公司 风控策略调整方法、装置、计算机设备和存储介质
CN110348999B (zh) * 2019-06-29 2023-12-22 北京淇瑀信息科技有限公司 金融风险敏感用户识别方法、装置及电子设备
CN110598996A (zh) * 2019-08-15 2019-12-20 平安普惠企业管理有限公司 一种风险处理方法、装置、电子设备及存储介质
CN112633619A (zh) * 2019-10-08 2021-04-09 阿里巴巴集团控股有限公司 风险评估方法及装置
CN110796557A (zh) * 2019-11-04 2020-02-14 泰康保险集团股份有限公司 数据处理方法及装置、电子设备和计算机可读存储介质
CN111209930B (zh) * 2019-12-20 2023-08-11 上海淇玥信息技术有限公司 一种生成授信策略的方法、装置和电子设备
CN111724069A (zh) * 2020-06-22 2020-09-29 百度在线网络技术（北京）有限公司 用于处理数据的方法、装置、设备及存储介质
CN111786824A (zh) * 2020-06-23 2020-10-16 中国电力科学研究院有限公司 数据中心能效比优化方法、***、设备及可读存储介质
CN111861703B (zh) * 2020-07-10 2024-05-10 深圳无域科技技术有限公司 数据驱动的风控策略规则生成方法及***、风险控制方法及***
CN112016788A (zh) * 2020-07-14 2020-12-01 北京淇瑀信息科技有限公司 风险控制策略生成及风险控制方法、装置和电子设备
CN112184143B (zh) * 2020-09-07 2022-04-29 支付宝(杭州)信息技术有限公司 一种合规审核规则中的模型训练方法、装置和设备
CN112184304A (zh) * 2020-09-25 2021-01-05 中国建设银行股份有限公司 一种辅助决策的方法、***、服务器和存储介质
CN112348661B (zh) * 2020-10-21 2024-04-23 上海淇玥信息技术有限公司 基于用户行为轨迹的服务策略分配方法、装置及电子设备
CN112465632A (zh) * 2020-12-17 2021-03-09 深圳微米信息服务有限公司 一种新金融ai智能风控决策方法和***
CN114662824A (zh) * 2020-12-23 2022-06-24 顺丰恒通支付有限公司 风控策略切换方法、装置、计算机设备和存储介质
CN112508698B (zh) * 2021-02-07 2024-04-26 北京淇瑀信息科技有限公司 用户策略触发方法、装置及电子设备
CN112508631A (zh) * 2021-02-07 2021-03-16 北京淇瑀信息科技有限公司 用户策略分配方法、装置及电子设备
CN112950221B (zh) * 2021-03-26 2022-07-26 支付宝(杭州)信息技术有限公司 建立风控模型的方法和装置以及风险控制方法和装置
CN113179266A (zh) * 2021-04-26 2021-07-27 口碑(上海)信息技术有限公司 业务请求处理方法及装置、电子设备、存储介质
CN113409050B (zh) * 2021-05-06 2022-05-17 支付宝(杭州)信息技术有限公司 基于用户操作判断业务风险的方法和装置
CN113112352A (zh) * 2021-05-27 2021-07-13 中国工商银行股份有限公司 风险业务检测模型训练方法、风险业务检测方法及装置
CN113986843A (zh) * 2021-11-02 2022-01-28 青岛海尔工业智能研究院有限公司 数据风险预警处理方法、装置及电子设备

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376397A (zh) * 2014-10-15 2015-02-25 中国农业银行股份有限公司 一种数据实时分析方法及装置
CN106656932A (zh) * 2015-11-02 2017-05-10 阿里巴巴集团控股有限公司 一种业务处理方法及装置
CN108364132A (zh) * 2018-02-11 2018-08-03 深圳市快付通金融网络科技服务有限公司 一种风控方法、装置、计算机装置及计算机可读存储介质
CN109034660A (zh) * 2018-08-22 2018-12-18 平安科技（深圳）有限公司 基于预测模型的风险控制策略的确定方法及相关装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104144166B (zh) * 2014-08-18 2017-07-21 中国人民解放军信息工程大学 面向可重构服务承载网的安全管控模型建立方法
CN107025509B (zh) * 2016-02-01 2021-06-18 腾讯科技（深圳）有限公司 基于业务模型的决策***和方法
CN111507638B (zh) * 2016-03-25 2024-03-05 创新先进技术有限公司 一种风险信息输出、风险信息构建方法及装置
CN107424069B (zh) * 2017-08-17 2020-11-17 创新先进技术有限公司 一种风控特征的生成方法、风险监控方法及设备

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376397A (zh) * 2014-10-15 2015-02-25 中国农业银行股份有限公司 一种数据实时分析方法及装置
CN106656932A (zh) * 2015-11-02 2017-05-10 阿里巴巴集团控股有限公司 一种业务处理方法及装置
CN108364132A (zh) * 2018-02-11 2018-08-03 深圳市快付通金融网络科技服务有限公司 一种风控方法、装置、计算机装置及计算机可读存储介质
CN109034660A (zh) * 2018-08-22 2018-12-18 平安科技（深圳）有限公司 基于预测模型的风险控制策略的确定方法及相关装置

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112767133A (zh) * 2021-01-26 2021-05-07 北京健康之家科技有限公司 业务决策方法及装置、存储介质、计算机设备
CN112767133B (zh) * 2021-01-26 2024-02-27 北京水滴科技集团有限公司 业务决策方法及装置、存储介质、计算机设备
CN112700329A (zh) * 2021-01-27 2021-04-23 永辉云金科技有限公司 一种风控规则引擎的响应方法和风控规则引擎
CN112950352A (zh) * 2021-02-08 2021-06-11 北京淇瑀信息科技有限公司 用户筛选策略生成方法、装置及电子设备
CN113673844A (zh) * 2021-08-04 2021-11-19 支付宝(杭州)信息技术有限公司 一种信息反馈方法、装置及设备
CN113673844B (zh) * 2021-08-04 2024-02-23 支付宝(杭州)信息技术有限公司 一种信息反馈方法、装置及设备
CN113657779A (zh) * 2021-08-20 2021-11-16 杭州时趣信息技术有限公司 一种动态配置化的风控决策方法、***、设备和存储介质
CN113657779B (zh) * 2021-08-20 2024-01-09 杭州时趣信息技术有限公司 一种动态配置化的风控决策方法、***、设备和存储介质
CN113781239A (zh) * 2021-09-10 2021-12-10 未鲲(上海)科技服务有限公司 一种策略确定方法、装置、电子设备以及存储介质
CN116823437A (zh) * 2023-06-14 2023-09-29 广东企企通科技有限公司 基于配置化风控策略的准入方法、装置、设备及介质

Also Published As

Publication number Publication date
CN109034660A (zh) 2018-12-18
CN109034660B (zh) 2023-07-14

Similar Documents

Publication Publication Date Title
WO2020037918A1 (zh) 基于预测模型的风险控制策略的确定方法及相关装置
CN109241711B (zh) 基于预测模型的用户行为识别方法及装置
Ring et al. Flow-based network traffic generation using generative adversarial networks
US10691494B2 (en) Method and device for virtual resource allocation, modeling, and data prediction
WO2019114344A1 (zh) 一种基于图结构模型的异常账号防控方法、装置以及设备
CN110442712B (zh) 风险的确定方法、装置、服务器和文本审理***
JP2019510320A (ja) 問題予測方法及びシステム
US20210385251A1 (en) System and methods for integrating datasets and automating transformation workflows using a distributed computational graph
US11570214B2 (en) Crowdsourced innovation laboratory and process implementation system
CN112580952A (zh) 用户行为风险预测方法、装置、电子设备及存储介质
CN111552633A (zh) 接口的异常调用测试方法、装置、计算机设备及存储介质
US11546380B2 (en) System and method for creation and implementation of data processing workflows using a distributed computational graph
CN112966865B (zh) 携号转网预测方法、装置及设备
Chen et al. Cognitive computing on unstructured data for customer co-innovation
CN108521405A (zh) 一种风险管控方法、装置及存储介质
CN107622197A (zh) 设备识别方法及装置、用于设备识别的权重计算方法及装置
WO2022237175A1 (zh) 图数据的处理方法、装置、设备、存储介质及程序产品
CN112328802A (zh) 数据处理方法、装置和服务器
CN104580109B (zh) 生成点选验证码的方法及装置
CN112347457A (zh) 异常账户检测方法、装置、计算机设备和存储介质
US11755848B1 (en) Processing structured and unstructured text to identify sensitive information
CN109544165A (zh) 资源转移处理方法、装置、计算机设备和存储介质
CN116318974A (zh) 站点风险识别方法、装置、计算机可读介质及电子设备
CN110401626A (zh) 一种黑客攻击分级检测方法及装置
RU2745362C1 (ru) Система и способ формирования индивидуального содержимого для пользователя сервиса

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18930845

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18930845

Country of ref document: EP

Kind code of ref document: A1