CN117635151A - Risk control strategy determining method, apparatus, device and storage medium - Google Patents

Abstract

The embodiment of the application provides a risk control strategy determining method, device, equipment and storage medium, and relates to the field of computers, wherein the method comprises the following steps: acquiring account feature data of a target account; inputting the account feature data into a control probability prediction model, and acquiring control probability and non-control probability corresponding to a candidate risk control strategy; inputting the account characteristic data into a risk increment prediction model corresponding to a candidate risk control strategy, and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment; and determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, so that the risk index can be reduced by using lower cost.

Description

Risk control strategy determining method, apparatus, device and storage medium

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a storage medium for determining a risk control policy.

Background

In an account risk control scenario such as payment risk control, whether risk control is performed on an account identified as being at risk, and by what policy, risk control is performed, has a great influence on whether risk continues to occur in the future of the account. The selection of different risk control strategies may lead to different results. For example, for a gray account, a heavier risk control strategy may be more able to hold down potential risks, with a lower likelihood of future re-outbreaks, whereas a lighter risk control strategy may not be able to prevent further risk transactions. Therefore, how to determine a risk control policy for risk control of an account is a technical problem to be solved.

Disclosure of Invention

The embodiment of the application provides a risk control strategy determining method, device, equipment and storage medium, so as to decide the risk control strategy and determine the risk control strategy.

In a first aspect, an embodiment of the present application provides a risk control policy determining method, where the method includes:

acquiring account feature data of a target account;

inputting the account feature data into a control probability prediction model, and obtaining control probability and non-control probability corresponding to a candidate risk control strategy, wherein the control probability prediction model is used for predicting the control probability, the control probability corresponding to the candidate risk control strategy is the probability of performing risk control on the target account based on the candidate risk control strategy, and the non-control probability is the probability of not performing risk control on the target account;

Inputting the account characteristic data into a risk increment prediction model corresponding to a candidate risk control strategy, and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, wherein the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed;

determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy;

and determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, wherein the target risk control strategy is used for performing risk control on the target account.

In a second aspect, an embodiment of the present application provides a risk control policy determining apparatus, including:

the feature acquisition module is used for acquiring account feature data of the target account;

the control prediction module is used for inputting the account feature data into a control probability prediction model to obtain control probability and non-control probability corresponding to a candidate risk control strategy, wherein the control probability prediction model is used for predicting the control probability, the control probability corresponding to the candidate risk control strategy is the probability of performing risk control on the target account based on the candidate risk control strategy, and the non-control probability is the probability of not performing risk control on the target account;

the risk increment prediction module is used for inputting the account characteristic data into a risk increment prediction model corresponding to a candidate risk control strategy and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, wherein the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed;

The conversion increment determining module is used for determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy;

the target strategy determining module is used for determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, wherein the target risk control strategy is used for performing risk control on the target account.

In a third aspect, embodiments of the present application provide a computer device, the device comprising:

a processor; and

a memory arranged to store computer executable instructions that, when executed, cause the processor to perform the risk control policy determination method as provided by the previous embodiment.

In a fourth aspect, embodiments of the present application provide a storage medium for storing computer-executable instructions that, when executed by a processor, implement steps in a risk control policy determination method as provided by the foregoing embodiments.

According to the risk control strategy determining method, device, equipment and storage medium, the control probability and the non-control probability corresponding to the candidate risk control strategy are obtained, the first risk increment and the second risk increment corresponding to the candidate risk control strategy are obtained, the risk conversion increment corresponding to the candidate risk control strategy is obtained according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment, the target risk control strategy is determined according to the risk conversion increment corresponding to the candidate risk control strategy, and overall control cost and risk control can be optimized.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and that other drawings may be obtained according to these drawings without inventive effort to a person skilled in the art.

Fig. 1 is a schematic flow chart of a risk control policy determining method provided in an embodiment of the present application;

FIG. 2 is a schematic diagram of a training method of a risk increment prediction model in the risk control strategy determination method shown in FIG. 1;

FIG. 3 is an example training flow of the risk delta prediction model shown in FIG. 2;

fig. 4 is a schematic diagram of a process for obtaining a problem transformation increment in a risk control policy determination method according to an embodiment of the present application;

FIG. 5 is a schematic diagram of a training process of the problem delta prediction model in the schematic diagram of the acquisition process of the problem delta shown in FIG. 4;

fig. 6 is a schematic block diagram of a risk control policy determining apparatus according to an embodiment of the present application;

fig. 7 is a schematic structural diagram of a computer device according to an embodiment of the present application.

Detailed Description

In order to better understand the technical solutions in the embodiments of the present application, the following description will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present specification.

The terms first, second and the like in the description and in the claims, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the present application may be implemented in sequences other than those illustrated or described herein.

In the event of non-conflicting aspects, one or more of the embodiments of the present application and features of the embodiments may be combined with one another. Embodiments of the present application will be described in detail below with reference to the accompanying drawings in conjunction with the embodiments.

The embodiment of the application provides a risk control strategy determining method, device, equipment and storage medium, which are used for determining a risk control strategy. In the risk control policy determining method provided by the embodiment of the application, account feature data of a target account is obtained; inputting the account feature data into a control probability prediction model, and acquiring control probability and non-control probability corresponding to a candidate risk control strategy, wherein the control probability corresponding to the candidate risk control strategy is the probability of performing risk control on the target account based on the candidate risk control strategy, and the non-control probability is the probability of not performing risk control on the target account; inputting the account characteristic data into a risk increment prediction model corresponding to a candidate risk control strategy, and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, wherein the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed; determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy; and determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, wherein the target risk control strategy is used for performing risk control on the target account.

According to the risk control strategy determining method, device, equipment and storage medium, the control probability and the non-control probability corresponding to the candidate risk control strategy are obtained, the first risk increment and the second risk increment corresponding to the candidate risk control strategy are obtained, the risk conversion increment corresponding to the candidate risk control strategy is obtained according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment, the target risk control strategy is determined according to the risk conversion increment corresponding to the candidate risk control strategy, control with the lowest cost is facilitated, maximum risk index reduction is achieved, and overall control cost is optimized.

Based on the above embodiment, further, according to the risk conversion increment corresponding to the target risk control policy, the embodiment of the present application determines an account category to which the target account belongs, and determines whether to execute risk control on the target account using the target risk control policy according to the account category to which the target account belongs, so that risk control can be executed on the target account in a targeted manner, account conversion capability is improved, and input-output ratio is optimized.

On the basis of the above embodiment, further, according to the risk conversion increment corresponding to the candidate risk control strategy and the problem conversion increment corresponding to the candidate risk control strategy, the embodiment of the application determines the target risk control strategy, and can combine account risk and user experience to select a risk control strategy with better overall effect, so that overall control cost and risk control are more optimized.

Based on the above embodiment, further, according to the risk conversion increment and the problem conversion increment corresponding to the target risk control policy, the embodiment of the present application determines an account category to which the target account belongs, and determines whether to execute risk control on the target account by using the target risk control policy according to the account category to which the target account belongs, so that risk control can be executed on the target account in a targeted manner, the overall input-output ratio is further optimized, and the overall control cost and risk control are further optimized.

The method provided by the embodiment of the application can be used for deciding a risk control strategy and can be used in account risk control scenes such as payment risk control. Fig. 1 is a flowchart of a risk control policy determining method provided in an embodiment of the present application, and referring to fig. 1, a detailed description will be given of the risk control policy determining method provided in the embodiment of the present application.

As shown in fig. 1, the risk control policy determining method provided in the embodiment of the present application may include the following processes:

s102, acquiring account feature data of a target account.

The target account may be an account of which the risk control policy is to be determined, and the account feature data is used to represent features of the account, and may be extracted from the account data.

In one implementation, the account feature data may include: basic feature data, behavior feature data, relationship feature data, etc. of the account. The basic feature data is used for representing the attribute of the account, for example, whether the account is an enterprise account or a personal account, if the account is a personal account of a user, user information and the like; the behavior characteristic data is used for representing the behavior of the account, such as the number of recent payment times of the account, transaction amount, login behavior of the user to the account, decryption behavior and the like; the relationship characteristic data is used to represent the relationship of the account, for example, the communication relationship or community relationship between the account and other accounts, and the like.

It may be understood that the account feature data may be any one of feature data such as basic feature data, behavior feature data, and relationship feature data of an account, or may be any combination thereof, which is not limited in this embodiment.

S104, inputting the account feature data into a control probability prediction model, and acquiring the control probability and the non-control probability corresponding to the candidate risk control strategy.

The candidate risk control strategies are candidate strategies of target risk control strategies for performing risk control on the account, and one or more candidate strategies can be selected. When the candidate risk control strategies are one, the control probability and the non-control probability corresponding to the candidate risk control strategies can be obtained, and when the candidate risk control strategies are a plurality of, the control probability and the non-control probability corresponding to each candidate risk control strategy can be obtained. The control probability corresponding to the candidate risk control policy indicates a probability of controlling the account based on the candidate risk control policy, that is, a probability of selecting the candidate risk control policy as the control policy, and the non-control probability indicates a probability of not controlling the account.

For example, if the candidate risk control policy is candidate risk control policy 1, the control probability P1 and the non-control probability P0 corresponding to the candidate risk control policy 1 may be obtained; if the candidate risk control policies are candidate risk control policy 1, candidate risk control policy 2, and candidate risk control policy 3, the control probability P1 corresponding to the candidate risk control policy 1, the control probability P2 corresponding to the candidate risk control policy 2, the control probability P3 corresponding to the candidate risk control policy 3, and the non-control probability P0 may be obtained.

The control probability and the non-control probability corresponding to the candidate risk control strategy can be obtained through prediction of a control prediction model. The control prediction model is used for predicting the control probability corresponding to each risk control strategy according to the account characteristic data, namely predicting the probability that each risk control strategy is selected as the control strategy for controlling the account, and the non-control probability is a tendency model, and can predict the tendency that each risk control strategy is selected as the control strategy according to the account characteristic data and output a prediction result. After the account feature data is input into the control probability prediction model, the control probability and the non-control probability corresponding to the candidate risk control strategy can be obtained from the prediction result output by the control prediction model. And each risk control strategy which can be predicted by the control probability prediction model comprises a candidate risk control strategy. For example, if the candidate risk control policy is risk control policy 1, each risk control policy may include: risk control strategy 1, risk control strategy 2, risk control strategy 3, including candidate risk control strategy 1.

The control probability prediction model is a machine learning model, which can be trained in advance. In one implementation, a sample account for training a control probability prediction model may be obtained, where the sample account may include a sample account controlled by a risk control policy and a sample account not controlled by the risk control policy, each sample is labeled with a control tag that may identify whether the sample account is controlled, and if so, which risk control policy the control policy is, and then the sample account may be used to perform supervised training on the control probability prediction model to be trained, thereby obtaining a trained control probability prediction model. The control probability prediction model to be trained can be a general machine learning model, such as a tree model, a neural network model and the like.

It should be noted that in some risk control scenarios, for example, in a payment risk control scenario, a control probability prediction model already exists in a risk control system, and when the risk control policy determination method provided in this embodiment is executed, the existing control probability prediction model may also be directly obtained, and account feature data may be input into the control prediction model, so as to multiplex the existing control probability prediction model, thereby improving processing efficiency.

S106, inputting the account feature data into a risk increment prediction model corresponding to the candidate risk control strategy, and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy.

The first risk increment corresponding to the candidate risk control strategy indicates that risk control is not performed originally, and if risk control is performed based on the candidate risk control strategy, the probability increment of risk occurrence of the account is compared with the probability increment of risk occurrence of the account without the risk control; the second risk increment corresponding to the candidate risk control strategy represents a probability increment of risk occurrence of the account if the risk control is not performed, compared with the risk control is not performed.

The first risk increment and the second risk increment corresponding to the candidate risk control strategy can be predicted by a risk increment prediction model corresponding to the candidate risk control strategy. The risk increment prediction model is used for predicting a first risk increment and a second risk increment corresponding to the corresponding candidate risk control strategies based on the account feature data. The risk increment prediction model corresponding to the candidate risk control strategy can be obtained by performing inverse fact learning on the probability of risk occurrence of the account when the candidate risk control strategy is used for controlling the risk of the account.

In the actual scene of risk control decision, it is often difficult to control to which control mode is out, so that the risk can be effectively managed, the user experience is not influenced, and the possibility that the account continues to normally trade is killed. In general, a control policy is selected as a current control means, and after the control means is issued to the account, only the subsequent performance of the account under the current control means can be observed. If the account continues to be at risk after the current control is issued, it is difficult to determine if the risk can be managed using another control. Because of the history-based data presentation, there is no experience with other control means.

In order to be able to estimate whether the risk and experience problems can occur to the account under different control means, the embodiment of the application adopts a method of inverse fact learning to train a risk increment prediction model. The inverse fact learning is an algorithm or learning method in the fields of artificial intelligence and machine learning, and is mainly used for learning causal relationships or conditional causal relationships from observed data. And after the risk control is performed on the account based on the candidate risk control strategy, performing inverse fact learning on the probability of the risk occurrence of the account, and learning the causal relationship and conditional causal relationship between the probability of the risk control and the probability of the risk occurrence of the account based on the candidate risk control strategy, predicting the increment of the risk conversion of the account based on the candidate risk control strategy, so as to determine a target risk control strategy with better risk prevention effect based on the increment of the risk conversion of the account.

It should be noted that the candidate risk control strategies are in one-to-one correspondence with the risk increment prediction models. The candidate risk control strategies can be one or a plurality of candidate risk control strategies. When the candidate risk control strategy is one, the account feature data can be input into a risk increment prediction model corresponding to the candidate risk control strategy, and a first risk increment and a second risk increment corresponding to the candidate risk control strategy are predicted; when the candidate risk control strategies are multiple, account feature data can be respectively input into a risk increment prediction model corresponding to each candidate risk control strategy, and a first risk increment and a second risk increment corresponding to each candidate risk control strategy are predicted.

For example, if the candidate risk control policy is candidate risk control policy 1, the account feature data may be input into the risk increment prediction model 1 corresponding to candidate risk control policy 1, so as to predict the first risk increment D10 and the second risk increment D11 corresponding to candidate risk control policy 1.

For another example, if the candidate risk control policies are the candidate risk control policy 1, the candidate risk control policy 2, and the candidate risk control policy 3, the account feature data may be input into the risk increment prediction model 1 corresponding to the candidate risk control policy 1, so as to predict the first risk increment D10 and the second risk increment D11 corresponding to the candidate risk control policy 1; inputting the account characteristic data into a risk increment prediction model 2 corresponding to the candidate risk control strategy 2, and predicting a first risk increment D20 and a second risk increment D21 corresponding to the candidate risk control strategy 2; the account characteristic data are input into a risk increment prediction model 3 corresponding to the candidate risk control strategy 3, and a first risk increment D30 and a second risk increment D31 corresponding to the candidate risk control strategy 3 are predicted.

The specific training method of the risk increment prediction model corresponding to the candidate strategy will be described in detail later.

S108, determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy.

The first component of the risk conversion increment corresponding to the candidate risk control strategy can be determined according to the control probability corresponding to the candidate risk control strategy and the first risk increment corresponding to the candidate risk control strategy; determining a second component of the risk conversion increment corresponding to the candidate risk control strategy according to the uncontrolled probability and the second risk increment corresponding to the candidate risk control strategy; and then, determining the risk conversion increment corresponding to the candidate risk control strategy according to the first component of the risk conversion increment corresponding to the candidate risk control strategy and the second component of the risk conversion increment corresponding to the candidate risk control strategy.

In specific implementation, the control probability corresponding to the candidate risk control strategy can be used for weighting the corresponding first risk increment to obtain a first component of the risk conversion increment corresponding to the candidate risk control strategy, the uncontrolled probability is used for weighting the corresponding second risk increment of the candidate risk control strategy to obtain a second component of the risk conversion increment corresponding to the candidate risk control strategy, and the risk conversion increment corresponding to the candidate risk control strategy is obtained according to the sum of the first component of the risk conversion increment corresponding to the candidate risk control strategy and the second component of the risk conversion increment.

For example, for the candidate risk control policy 1, if the control probability corresponding to the candidate risk control policy 1 is P1, the uncontrolled probability is P0, the first risk increment corresponding to the candidate risk control policy 1 is D10, and the second risk increment is D11, the first component of the risk conversion increment corresponding to the candidate risk control policy 1 may be the product of the control probability corresponding to the candidate risk control policy 1 and the first risk increment corresponding to the candidate risk control policy 1, that is, P1×d10, the second component of the risk conversion increment corresponding to the candidate risk control policy 1 may be the second risk increment of the uncontrolled probability corresponding to the candidate risk control policy 1, that is, P0×d11, and the risk conversion increment corresponding to the candidate risk control policy 1 may be the sum of the first component of the risk conversion increment corresponding to the candidate risk control policy 1 and the second component of the risk conversion increment, that is, P1×d10+p0×d11.

For another example, for the candidate risk control policy 2, if the control probability corresponding to the candidate risk control policy 2 is P2, the uncontrolled probability is P0, the first risk increment corresponding to the candidate risk control policy 2 is D20, and the second risk increment is D21, the first component of the risk conversion increment corresponding to the candidate risk control policy 2 may be the product of the control probability corresponding to the candidate risk control policy 2 and the first risk increment corresponding to the candidate risk control policy 2, that is, P2×d20, the second component of the risk conversion increment corresponding to the candidate risk control policy 2 may be the second risk increment of the uncontrolled probability corresponding to the candidate risk control policy 2, that is, P0×d21, and the risk conversion increment corresponding to the candidate risk control policy 2 may be the sum of the first component of the risk conversion increment corresponding to the candidate risk control policy 2 and the second component of the risk conversion increment, that is, P2×d20+p0×d21.

S110, determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy.

The target risk control strategy is used for performing risk control on the target account, and the target risk control strategy can be determined according to a risk conversion increment corresponding to the candidate risk control strategy.

As described above, the candidate risk control strategies may be one or more.

In one implementation, when the candidate risk control policy is one, the risk conversion increment corresponding to the candidate risk control policy may be compared with a set risk conversion increment threshold, and whether to determine the candidate risk control policy as the target risk control policy is determined according to the comparison result. Wherein the risk transition increment threshold may be an empirical value, which may be a value not greater than 0. For example, the risk conversion increment threshold may be set to 0, and if the risk conversion increment corresponding to the candidate risk control policy is smaller than 0, it may be determined that the difference between the probability of risk occurrence when the account is controlled based on the candidate risk control policy and the probability of risk occurrence when the account is not controlled, that is, the probability of risk occurrence when the account is controlled based on the candidate risk control policy is increased negatively compared to when the account is not controlled, and therefore, the probability of risk occurrence when the account is controlled based on the candidate risk control policy can be reduced.

By determining the target risk control policy according to the risk conversion increment corresponding to the candidate risk control policy, it can be determined whether the candidate risk control policy is a risk control policy effective for controlling the risk of the account, and thus, the decline of the risk index can be ensured.

In one implementation manner, when the number of candidate risk control strategies is multiple, a risk conversion increment corresponding to each candidate risk control strategy may be obtained, and the target risk control strategy is determined from the multiple candidate risk control strategies according to the size of the risk conversion increment corresponding to each candidate risk control strategy. Wherein, the candidate risk control strategy with the smallest risk conversion increment can be determined as the target risk control strategy. For example, if the candidate risk control policies are candidate risk control policy 1, candidate risk control policy 2, and candidate risk control policy 3, the risk conversion increment corresponding to candidate risk control policy 1 is-0.5, the risk conversion increment corresponding to candidate risk control policy 2 is-0.7, the risk conversion increment corresponding to candidate risk control policy 3 is-0.3, and the risk conversion increments corresponding to candidate risk control policies 1, 2, and 3 are all negative values, which increase the probability of risk occurrence of the account negatively, but the absolute value of the risk conversion increment of candidate risk control policy 2 is the largest, and if the risk control is performed on the account based on candidate risk control policy 2, the degree of negative increase of the probability of risk occurrence of the account is the largest, that is, the probability of risk occurrence of the account is the largest, at this time, candidate risk control policy 2 may be determined as the target risk control policy.

The target risk control strategies are determined according to the risk conversion increment corresponding to different candidate risk control strategies, so that the risk control strategy with good control effect can be selected from the candidate risk control strategies according to the performance of the account when encountering different risk control strategies, the lowest control cost is convenient to use, and the maximum reduction of risk indexes is ensured.

In the above embodiment, the first risk increment and the second risk increment corresponding to the candidate risk control policy are predicted based on the risk increment prediction model corresponding to the candidate risk control policy, and the risk conversion increment corresponding to the candidate risk control policy is calculated based on the first risk increment and the second risk increment corresponding to the candidate risk control policy. Fig. 2 is a schematic diagram of a training method of a risk increment prediction model corresponding to a candidate risk control strategy, and a detailed description will be given of a specific training method of the risk increment prediction model with reference to fig. 2.

As shown in fig. 2, the risk increment prediction model corresponding to the candidate risk control policy may include a first risk increment prediction model and a second risk increment prediction model, where the first risk increment prediction model is used for predicting a first risk increment corresponding to the candidate risk control policy based on account feature data, that is, if risk control is performed on the account based on the candidate risk control policy, performing risk control is compared with performing no risk control on a probability increment of risk of the account; the second risk increment prediction model is used for predicting a second risk increment corresponding to the candidate risk control strategy based on the account feature data, namely, the risk control is originally performed on the account based on the candidate risk control strategy, and if the risk control is not performed, the probability increment of risk occurrence of the account is compared with the risk control is not performed.

Fig. 2 (a) is a flow chart of a training method of the first risk incremental prediction model. As shown in fig. 2 (a), the training process of the first risk delta prediction model may include the following processes:

s201, acquiring first sample data.

The first sample data comprises account feature data of a first sample account and first risk probability corresponding to the first sample account, wherein the first sample account is an account sample which does not carry out risk control, and the first risk probability corresponding to the first sample account is a real probability of risk of the account when the risk control is not carried out on the account, and is a real risk probability value of the first sample account.

S203, inputting the account feature data in the first sample data into a second risk prediction model corresponding to the candidate risk control strategy, and predicting a second risk probability corresponding to the first sample account.

The second risk probability corresponding to the first sample account is the probability of risk occurrence of the account when the candidate risk control strategy is used for risk control of the account, and the second risk probability is the inverse fact risk probability value of the first sample account. The second risk probability corresponding to the first sample account can be predicted by a second risk prediction model corresponding to the candidate risk control strategy.

The second risk prediction model corresponding to the candidate risk control strategy is used for predicting the probability of risk occurrence of the account when the candidate risk control strategy is used for carrying out risk control on the account according to the account feature data, namely predicting the second risk probability, and the second risk prediction model can be obtained by carrying out supervised training on second sample data in advance. The second sample account is an account sample for risk control based on the candidate risk control policy, and the second sample data may include account feature data of the second account and a second risk probability corresponding to the second account. In specific implementation, account feature data of a second sample account can be input into a risk prediction model to be trained, second risk probability corresponding to the second sample account is predicted, a predicted value of the second risk probability corresponding to the second sample account is obtained, then, according to an error between the predicted value of the second risk probability and an actual value, parameters of the risk prediction model to be trained are trained and optimized, and finally, the second risk prediction model is obtained through training.

S205, obtaining a first risk increment corresponding to the first sample account according to the difference value between the second risk probability corresponding to the first sample account and the first risk probability.

The first risk increment indicates that risk control is not performed originally, and if risk control is performed based on the candidate risk control strategy, the probability increment of risk occurrence of the risk control account is performed compared with that of the risk control account, which can be obtained according to the difference value between the second risk probability corresponding to the first sample account and the first risk probability.

S207, performing supervised training on the risk increment prediction model to be trained by using the account feature data of the first sample account and the first risk increment corresponding to the first sample account to obtain a first risk increment prediction model.

The risk increment prediction model to be trained can be a machine learning model, such as a tree model, a neural network model, and the like. In specific implementation, account feature data of a first sample account can be input into a risk increment prediction model to be trained, a first risk increment corresponding to the first sample account is predicted, a predicted value of the first risk increment corresponding to the first sample account is obtained, then parameters of the risk increment prediction model to be trained are trained and optimized according to errors between the predicted value of the first risk increment and an actual value of the first risk increment, and finally the first risk increment prediction model is obtained.

Fig. 2 (b) is a flow chart of a training method of the second risk incremental prediction model. As shown in fig. 2 (b), the training process of the second risk delta prediction model may include the following processes:

s202, second sample data is acquired.

The second sample data comprises account feature data of a second sample account and second risk probability corresponding to the second sample account, the second sample account is an account sample for performing risk control on the account based on a candidate risk control strategy, and the second risk probability corresponding to the second sample account is a real probability of risk of the account when the account is subjected to risk control based on the candidate risk control strategy, and is a real risk probability of the second sample account.

S204, inputting the account feature data in the second sample data into the first risk prediction model, and predicting the first risk probability corresponding to the second sample account.

The first risk probability is the probability of risk of the account when risk control is not performed, and is the inverse fact risk probability of the second sample account. The first risk probability corresponding to the second sample account can be predicted by the first risk prediction model.

The first risk prediction model is used for predicting the probability of risk occurrence of the account when the account is not subjected to risk control according to the account feature data, namely predicting the first risk probability, and can be obtained by performing supervised training on the first sample data in advance. The first sample account is an account sample without risk control, and the first sample data may include account feature data of the first account and a first risk probability corresponding to the first account. In specific implementation, the account feature data of the first sample account can be input into a risk prediction model to be trained, the first risk probability corresponding to the first sample account is predicted, the predicted value of the first risk probability corresponding to the first sample account is obtained, then, according to the error between the predicted value of the first risk probability and the actual value, the parameters of the risk prediction model to be trained are trained and optimized, and finally, the first risk prediction model is obtained through training.

S206, obtaining a second risk increment corresponding to the second sample account according to the difference value between the second risk probability corresponding to the second sample account and the first risk probability.

The second risk increment represents a probability increment of risk occurrence of the account compared with the risk occurrence of the account when the risk control is not performed if the risk control is not performed based on the candidate risk control strategy, and the probability increment can be obtained according to a difference value between the second risk probability corresponding to the second sample account and the first risk probability.

S208, performing supervised training on the risk increment prediction model to be trained by using the account feature data of the second sample account and the second risk increment corresponding to the second sample account to obtain a second risk increment prediction model.

The risk increment prediction model to be trained can be a machine learning model, such as a tree model, a neural network model, and the like. In specific implementation, account feature data of a second sample account can be input into a risk increment prediction model to be trained, a second risk increment corresponding to the second sample account is predicted, a predicted value of the second risk increment corresponding to the second sample account is obtained, and then parameters of the risk increment prediction model to be trained are trained and optimized according to errors between the predicted value of the second risk increment and an actual value of the second risk increment, so that the second risk increment prediction model is finally obtained.

From the above, it can be seen that the training methods of the first risk incremental prediction model and the second risk incremental prediction model are basically the same, and it can be understood that the first risk incremental prediction model and the second risk incremental prediction model are two modules in the risk incremental prediction model, and training of the two models can be performed simultaneously. Fig. 3 is a training example of a risk increment prediction model corresponding to a candidate risk control policy, and the training method of the risk increment prediction model provided in the embodiment of the present application will be further described with reference to fig. 3.

As shown in fig. 3, the training process of the risk increment prediction model corresponding to the candidate risk control strategy may include the following steps:

s302, first sample data and second sample data corresponding to the candidate risk control strategy are obtained.

The first sample data comprises account feature data of a first sample account and first risk probability corresponding to the first sample account, wherein the first sample account is an account sample which does not carry out risk control; the second sample data comprises account feature data of a second sample account and second risk probabilities corresponding to the second sample account, and the second sample account is an account sample for risk control based on the candidate risk control strategy.

S304, performing supervised training on a first risk prediction model to be trained by using the first sample data to obtain a first risk prediction model; and performing supervised training on the second risk prediction model to be trained by using the second sample data to obtain the second risk prediction model.

The first risk prediction model is used for predicting the probability of risk occurrence of the account under the condition that risk control is not performed according to the account feature data, namely predicting the first risk probability; the second risk prediction model is used for predicting the probability of risk occurrence of the account under the condition of performing risk control on the account based on the candidate risk control strategy according to the account feature data, namely predicting the second risk probability.

S306, inputting account feature data of the first sample account into a second risk prediction model, and predicting a second risk probability corresponding to the first sample account; and inputting the account characteristic data of the second sample account into the first risk prediction model, and predicting the first risk probability corresponding to the second sample account.

The first sample account is an account which does not perform risk control originally, the first risk probability in the first sample data is the fact risk probability, and the second risk probability predicted by the second risk prediction model is the anti-fact risk probability.

The second sample account is an account which is originally subjected to risk control based on a candidate risk control strategy, the second risk probability in the second sample data is the fact risk probability, and the second risk probability predicted by the second risk prediction model is the anti-fact risk probability.

S308, obtaining a first risk increment corresponding to the first sample account according to the difference value between the second risk probability corresponding to the first sample account and the first risk probability; and obtaining a second risk increment corresponding to the second sample account according to the difference value between the second risk probability corresponding to the second sample account and the first risk probability.

The first risk increment is a probability increment of risk occurrence of the account compared with the risk non-occurrence of the risk control if the risk control is performed based on the candidate risk control strategy.

The second risk increment is a probability increment of risk occurrence of the account compared with the risk control without the risk control based on the candidate risk control strategy if the risk control is not performed.

S310, performing supervised training on a first risk increment prediction model to be trained by using account feature data of a first sample account and a first risk increment corresponding to the first sample account to obtain a first risk increment prediction model corresponding to a candidate risk control strategy; and performing supervised training on the second risk increment prediction model to be trained by using the account feature data of the second sample account and the second risk increment corresponding to the second sample account to obtain a second risk increment prediction model corresponding to the candidate risk control strategy.

S312, obtaining a risk increment prediction model corresponding to the candidate risk control strategy according to the first risk increment prediction model corresponding to the candidate risk control strategy and the second risk increment prediction model corresponding to the candidate risk control strategy.

It can be understood that the training method is used for training to obtain a risk increment prediction model corresponding to a candidate risk control policy, and the method can be used for a case that one candidate risk control policy exists, or a case that a plurality of candidate risk control policies exist, when a plurality of candidate risk control policies exist, first sample data and second sample data corresponding to each candidate risk control policy can be obtained according to the method, and a risk increment prediction model corresponding to each candidate risk control policy is obtained based on the first sample data and the second sample data corresponding to each candidate risk control policy. After the risk increment prediction model corresponding to each candidate risk control strategy is obtained, the risk control strategy determining method shown in fig. 1 may be executed, so as to determine the target risk control strategy.

Further, on the basis of the above embodiment, in addition to determining the target risk control policy for performing risk control on the target account based on the risk conversion increment corresponding to the candidate risk control policy, the method provided in the embodiment of the present application further determines the category of the target account according to the risk conversion increment corresponding to the target risk policy. The class of target accounts may be used to further subdivide the user so that the control means acts precisely on a certain class of accounts.

Specifically, for the target account, when the risk conversion increment corresponding to the target risk control policy is smaller than the first increment threshold, determining that the target account belongs to the first type account; when the risk conversion increment corresponding to the target risk control strategy is larger than a second increment threshold, determining that the current account belongs to a second type of account; when the risk conversion increment corresponding to the target candidate risk control strategy is between a first increment threshold and a second increment threshold, acquiring a first risk probability and a second risk probability corresponding to the target risk control strategy for the target account, determining that the target account belongs to a third type account when the first risk probability and the second risk probability corresponding to the target risk control strategy are smaller than the first risk threshold, and determining that the target account belongs to a fourth type account when the first risk probability and the second risk probability corresponding to the target risk control strategy are larger than the second risk threshold. The first risk probability corresponding to the target risk control strategy is the probability of the risk of the account when the risk control is not performed, the probability can be obtained through prediction of a first risk prediction model corresponding to the target risk control strategy, the second risk probability corresponding to the target risk control strategy is the probability of the risk of the account when the risk control is performed based on the target risk control strategy, and the probability can be obtained through a second risk prediction model corresponding to the target risk control strategy. The first risk prediction model and the second risk prediction model corresponding to the target risk control strategy may be pre-trained based on the first sample data and the second sample data corresponding to the target risk control strategy.

The first increment threshold and the second increment threshold may be preset values, which may be empirical values, wherein the first increment threshold may be negative, the second increment threshold may be positive, for example, the first increment threshold may be-0.1, and the second increment threshold may be 0.1. Accordingly, the first type of account may be an account whose risk occurrence probability can be significantly reduced if the account is controlled based on the target risk control policy, the second type of account may be an account whose risk occurrence probability is significantly increased if the account is controlled based on the target risk control policy, the third type of account may be an account whose risk occurrence probability is smaller regardless of whether the account is controlled or not based on the target risk control policy, and the fourth type of account may be an account whose risk occurrence probability is greater regardless of whether the account is controlled or not based on the target risk control policy.

It may be understood that in the embodiment of the present application, there may be one or more target accounts, and when there is one target account, the corresponding target risk control policy may be determined for the target account based on the risk policy determining method provided in the embodiment of the present application, and then the account category to which the target account belongs may be determined according to the risk conversion increment of the target risk control policy corresponding to the target account. When there are multiple target accounts, the target risk control policy corresponding to each target account can be determined based on the risk policy determining method provided by the embodiment of the present application, and then the account category to which each target account belongs is determined according to the risk conversion increment of the target risk control policy corresponding to each target account, so that the multiple target accounts are classified.

By classifying the target accounts, the account with the highest input-output ratio can be screened out, so that the risk prevention and control can be executed in a targeted manner, the input-output ratio of the risk control is improved, and the overall punishment cost and the risk control are further optimized.

Further, after determining the account type of the target account according to the risk conversion increment corresponding to the target risk control policy, it may be determined whether to execute risk control on the target account using the target risk control policy according to the account type to which the target account belongs, thereby pertinently executing risk control,

for example, only the target account of the designated account type, for example, the first type account described above, that is, the account whose risk occurrence probability can be significantly reduced when the account is controlled based on the target risk control policy, compared with the account without risk control, may be subjected to risk control. Aiming at the target account, the target risk policy has the highest risk conversion capability and highest input-output ratio, and performs risk control on the target account by using the target risk control policy only when the target account belongs to the first account, so that a control means accurately acts on the first account, invalid or negative influence control is reduced, and risk control is optimized.

It will be appreciated that in the above embodiments, the first increment threshold is negative and the second increment threshold is positive, although in some cases the values of the first increment threshold and the second increment threshold may be equal, for example, they may each be 0. That is, the first delta threshold may be a value not greater than 0 and the second delta threshold may be a value not less than 0. Accordingly, the first type of account is, for example, an account whose risk occurrence probability can be reduced when the account is controlled based on the target risk control policy, the second type of account is, for example, an account whose risk occurrence probability is increased when the account is controlled based on the target risk control policy, the third type of account is, for example, an account whose risk occurrence probability is smaller regardless of whether the account is controlled or not based on the target risk control policy, and the fourth type of account is, for example, an account whose risk occurrence probability is larger regardless of whether the account is controlled or not based on the target risk control policy, compared to an account whose risk occurrence probability is reduced when the account is not controlled.

In the above embodiment, the target risk control policy of the target account is determined according to the risk conversion increment corresponding to the candidate risk control policy, where the risk is the risk of the account having a security problem. The selection of the risk control policy affects not only the probability of the account being at risk for security, but also the user experience, which in turn affects the likelihood that the account will continue to transact, and thus the likelihood that the account will change from bad. For example, for a gray account, controlling with different risk control strategies may lead to different results. A control strategy with a heavier penalty may be more robust against potential risks, reducing the likelihood of future outbreaks of the account, but may affect the user experience, throttling the account away from continuing normal transactions, and reducing the likelihood of turning it well. Conversely, a control strategy with a lighter penalty may not prevent the continuation of the risk transaction, but it does not interfere with the subsequent normal transactions of the account.

In order to further optimize overall risk control, in some embodiments, account risk problems and user experience problems are considered at the same time, and on the basis of determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, the risk conversion increment corresponding to the candidate risk control strategy and the problem conversion increment are further combined to determine the target risk control strategy.

In one implementation, the process S110 in the foregoing embodiment may further include: and determining a target risk control strategy according to the risk conversion increment and the problem conversion increment corresponding to the candidate risk control strategy.

Compared with the situation that risk control is not performed, the problem conversion increment corresponding to the candidate risk control strategy is used for performing risk control based on the candidate risk control strategy, and the probability increment of experience problems of a user is similar to the obtaining mode of the risk conversion increment corresponding to the candidate risk control strategy. Fig. 4 shows a schematic flow chart of obtaining a problem transformation increment corresponding to a candidate risk control strategy. As shown in fig. 4, on the basis of the embodiment shown in fig. 2, the following process may be further performed to obtain a problem transformation increment corresponding to the candidate risk control policy, thereby facilitating the subsequent determination of the target risk control policy according to the risk transformation increment corresponding to the candidate risk control policy and the problem transformation increment.

S406, inputting account feature data of the target account into a problem increment prediction model corresponding to the candidate risk control strategy, and predicting a first problem increment and a second problem increment corresponding to the candidate risk control strategy.

The first problem increment indicates that risk control is not performed originally, and if risk control is performed based on a candidate risk control strategy, the probability increment that the risk control is performed is compared with the probability increment that the risk control is not performed, and the account experiences problems; the second problem increment represents a probability increment that the risk control is originally performed based on the candidate risk control strategy, and if the risk control is not performed, the risk control is performed compared with the risk control is not performed, and the account experiences the problem.

The first problem increment and the second problem increment corresponding to the candidate risk control strategy can be predicted by a problem increment prediction model corresponding to the candidate risk control strategy. The problem increment prediction model corresponding to the candidate risk control strategy is used for predicting a first problem increment and a second problem increment corresponding to the candidate risk control strategy, and can be obtained through pre-training.

The problem increment prediction model corresponding to the candidate risk control strategy is similar to the risk increment prediction model corresponding to the candidate risk control strategy. For example, the problem delta prediction models corresponding to the candidate risk control strategies may include a first problem delta prediction model for predicting a first problem delta and a second problem delta prediction model for predicting a second problem delta.

The problem increment prediction model corresponding to the candidate risk control strategy is similar to the training method of the risk increment prediction model corresponding to the candidate risk control strategy. FIG. 5 shows a schematic training flow diagram of a problem delta prediction model. As shown in fig. 5, in one implementation, the problem increment prediction model corresponding to the candidate risk control policy may be obtained through training by the following method.

S502, obtaining third sample data and fourth sample data corresponding to the candidate risk control strategy.

The third sample data comprises account feature data of a third sample account and first problem probability corresponding to the third sample account; the fourth sample data comprises account feature data of a fourth sample account and a second problem probability corresponding to the fourth sample account. The third sample account is an account sample without risk control, and the fourth sample account is an account sample with risk control based on a candidate risk control policy. The first problem probability is the probability of experience problems of the account when risk control is not performed, and the second problem probability is the probability of experience problems of the account when risk control is performed based on a candidate risk control strategy.

S504, inputting account feature data of a third sample account into a second problem prediction model corresponding to the candidate risk control strategy, and predicting second problem probability corresponding to the third sample account; and inputting the account feature data of the fourth sample account into the first problem prediction model to predict the first problem probability corresponding to the fourth sample account.

The first problem prediction model is used for predicting first problem probability, namely probability of experience problems of the account when risk control is not performed, and the second problem prediction model corresponding to the candidate risk control strategy is used for predicting second problem probability, namely probability of experience problems of the account when risk control is performed based on the candidate risk control strategy.

The first problem prediction model and the second problem prediction model corresponding to the candidate risk control strategy may be trained in advance. In one implementation, the third sample data may be used to perform supervised training on the first problem prediction model to be trained to obtain the first problem prediction model; and performing supervised training on the second problem prediction model to be trained by using fourth sample data corresponding to the candidate risk control strategy to obtain the second problem prediction model corresponding to the candidate risk control strategy.

S506, obtaining a first problem increment corresponding to the third sample account according to the difference value between the second problem probability corresponding to the third sample account and the first problem probability; and obtaining a second problem increment corresponding to the fourth sample account according to the difference value between the second problem probability corresponding to the fourth sample account and the first problem probability.

The first problem increment is a probability increment that the risk control is not performed originally, and if the risk control is performed based on the candidate risk control strategy, compared with the risk control is not performed, the account experiences the problem.

The second problem increment is a probability increment that risk control is originally performed based on the candidate risk control strategy, and if risk control is not performed, the risk control is performed based on the candidate risk control strategy, compared with the risk control is not performed, the account experiences problems.

S508, performing supervised training on the first problem increment prediction model to be trained by using the account feature data of the third sample account and the first problem increment corresponding to the third sample account to obtain a first problem increment prediction model corresponding to the candidate risk control strategy; and performing supervised training on the second problem increment prediction model to be trained by using the account feature data of the fourth sample account and the second problem increment corresponding to the fourth sample account to obtain a second problem increment prediction model corresponding to the candidate risk control strategy.

S510, obtaining a problem increment prediction model corresponding to the candidate risk control strategy according to the first problem increment prediction model corresponding to the candidate risk control strategy and the second problem increment prediction model corresponding to the candidate risk control strategy.

After obtaining the problem increment prediction model corresponding to the candidate risk control strategy, the account feature data of the target account can be input into the problem increment prediction model corresponding to the candidate risk control strategy, and the first problem increment and the second problem increment corresponding to the candidate risk control strategy are predicted for the target account.

S408, determining a problem conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first problem increment and the second problem increment corresponding to the candidate risk control strategy.

The method comprises the steps of obtaining a risk conversion increment corresponding to a candidate risk control strategy, wherein similar to the risk conversion increment, the control probability and the non-control probability corresponding to the candidate risk control strategy can be used for carrying out weighted summation on a first problem increment and a second problem increment corresponding to the candidate risk control strategy, and the problem conversion increment corresponding to the candidate risk control strategy is obtained.

For example, the problem transformation delta corresponding to the candidate policy may be calculated based on the following formula:

Problem conversion increment = uncontrolled probability first problem increment + controlled probability second problem increment.

After obtaining the problem conversion increment corresponding to the candidate risk control strategy, further, the target risk control strategy can be determined based on the risk conversion increment corresponding to the candidate risk control strategy and the problem conversion increment.

In one implementation manner, the risk conversion increment and the problem conversion increment corresponding to the candidate risk control strategy may be weighted and summed according to preset risk weights and problem weights to obtain an increment coefficient corresponding to the candidate risk control strategy, and the target risk control strategy is determined according to the increment coefficient corresponding to the candidate risk control strategy.

For example, if the preset risk weight is 0.8, the problem weight is 0.3, the risk conversion increment corresponding to the candidate risk control policy 1 is-0.7, the problem conversion increment is +0.3, the risk conversion increment corresponding to the candidate risk control policy 2 is-0.6, and the problem conversion increment is 0, the increment coefficient corresponding to the candidate risk control policy 1 is 0.8 (-0.7) +0.3= -0.47, the increment coefficient corresponding to the candidate risk control policy 2 may be 0.8 (-0.6) +0.3= -0.48, and the increment coefficient corresponding to the candidate risk control policy 2 is smaller than the increment coefficient corresponding to the candidate risk control policy 1, which indicates that the candidate risk control policy 2 makes the account more likely to be well-rotated than the candidate risk control policy 1, and in this case, the candidate risk control policy 2 may be the target risk control policy.

In one implementation, the first candidate risk control policy may be determined according to a problem transformation increment corresponding to the candidate risk control policy, where the first candidate risk control policy may be a policy with a problem transformation increment not greater than 0, that is, a policy that does not cause an increase in probability that an account experiences a problem, and then, according to a risk transformation increment corresponding to the first candidate risk control policy, the target risk control policy is selected from the first candidate risk control policy. The target risk control policy may be a risk control policy with a minimum risk conversion increment in the first candidate risk control policy.

For example, if the risk conversion increment corresponding to the candidate risk control policy 1 is-0.9, the problem conversion increment is 0.2, the risk conversion increment corresponding to the candidate risk control policy 2 is-0.8, the problem conversion increment is 0.1, the risk conversion increment corresponding to the candidate risk control policy 3 is-0.7, and the problem conversion increment is 0, the candidate risk control policy 3 may be determined as the target risk control policy.

Further, on the basis of determining the target risk control strategy according to the risk conversion increment and the problem conversion increment corresponding to the candidate risk control strategy, the method and the device can determine the category of the target account according to the risk conversion increment and the problem conversion increment corresponding to the target risk control strategy so as to further subdivide the target account, and therefore the account with higher conversion capability and input and output can be screened out.

For example, the risk conversion increment and the problem conversion increment corresponding to the target risk control strategy can be compared with a set two-dimensional threshold grid, so that the account type of the target account can be determined. The two-dimensional threshold grid may be a grid divided by a hierarchy of risk conversion increments and a hierarchy of problem conversion increments, each network representing an account category.

Further, whether the target risk control strategy is used for risk control of the target account can be determined according to the category of the target account, so that the target risk control strategy is enabled to act on the target account more accurately, the overall input-output ratio is highest, and the overall penalty cost and the risk control are optimized.

For example, the target risk control policy may be used to perform risk control on the target account when the account class of the target account belongs to a specified class, e.g., for the target account, the risk conversion increment and the problem conversion increment corresponding to the target risk control policy belong to a specified network.

The embodiment of the application also provides a risk control strategy determining device based on the same technical conception. Fig. 6 is a schematic block diagram of a risk control policy determining apparatus according to an embodiment of the present application. As shown in fig. 6, the risk control policy determining apparatus provided in the embodiment of the present application may include:

A feature acquisition module 10, configured to acquire account feature data of a target account;

the control prediction module 20 is configured to input the account feature data into a control probability prediction model, and obtain a control probability and an uncontrolled probability corresponding to a candidate risk control policy, where the control probability prediction model is used for predicting the control probability, the control probability corresponding to the candidate risk control policy is a probability of performing risk control on the target account based on the candidate risk control policy, and the uncontrolled probability is a probability of not performing risk control on the target account;

the risk increment prediction module 30 is configured to input the account feature data into a risk increment prediction model corresponding to a candidate risk control policy, and predict a first risk increment and a second risk increment corresponding to the candidate risk control policy; the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, wherein the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed;

The conversion increment determining module 40 is configured to determine a risk conversion increment corresponding to the candidate risk control policy according to the control probability and the non-control probability corresponding to the candidate risk control policy, and the first risk increment and the second risk increment corresponding to the candidate risk control policy;

the target policy determining module 50 is configured to determine a target risk control policy according to the risk conversion increment corresponding to the candidate risk control policy, where the target risk control policy is used to perform risk control on the target account.

It should be noted that, the embodiments of the risk control policy determining apparatus and the embodiments of the risk control policy determining method in the present application are based on the same inventive concept, so that the specific implementation of the embodiments may refer to the implementation of the corresponding method embodiments, and the repetition is omitted.

The embodiment of the application further provides a computer device, based on the same technical concept, for executing the risk control policy determining method provided in the foregoing embodiment, as shown in fig. 7.

Computer devices may vary considerably in configuration or performance and may include one or more processors and memory in which one or more stored applications or data may be stored. Wherein the memory may be transient or persistent. The application program stored in the memory may include one or more modules, each of which may include a series of computer-executable instructions for use in a computer device. Still further, the processor may be configured to communicate with a memory and execute a series of computer-executable instructions in the memory on the computer device. The computer device may also include one or more power supplies, one or more wired or wireless network interfaces, one or more input/output interfaces, one or more keyboards, and the like.

In a particular embodiment, a computer device includes a memory, and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the computer device, and configured to be executed by one or more processors, the one or more programs comprising computer-executable instructions for:

acquiring account feature data of a target account;

inputting the account feature data into a control probability prediction model, and acquiring control probability and non-control probability corresponding to a candidate risk control strategy, wherein the control probability corresponding to the candidate risk control strategy is the probability of performing risk control on the target account based on the candidate risk control strategy, and the non-control probability is the probability of not performing risk control on the target account;

inputting the account characteristic data into a risk increment prediction model corresponding to a candidate risk control strategy, and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; the control probability prediction model is used for predicting control probability, the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed;

Determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy;

and determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, wherein the target risk control strategy is used for performing risk control on the target account.

It should be noted that, in the present application, the embodiment about the computer device and the embodiment about the risk control policy determining method in the present application are based on the same inventive concept, so the specific implementation of this embodiment may refer to the implementation of the foregoing corresponding method embodiment, and the repetition is not repeated.

The embodiment of the application also provides a storage medium for storing computer executable instructions based on the same technical conception;

in a specific embodiment, the storage medium may be a usb disk, an optical disk, a hard disk, etc., and the computer executable instructions stored on the storage medium include steps for performing the method of:

Acquiring account feature data of a target account;

inputting the account feature data into a control probability prediction model, and acquiring control probability and non-control probability corresponding to a candidate risk control strategy, wherein the control probability corresponding to the candidate risk control strategy is the probability of performing risk control on the target account based on the candidate risk control strategy, and the non-control probability is the probability of not performing risk control on the target account;

inputting the account characteristic data into a risk increment prediction model corresponding to a candidate risk control strategy, and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; the control probability prediction model is used for predicting control probability, the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed;

Determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy;

and determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, wherein the target risk control strategy is used for performing risk control on the target account.

It should be noted that, the embodiments related to the storage medium and the embodiments related to the risk control policy determining method in the present application are based on the same inventive concept, so that the implementation of this embodiment may refer to the implementation of the foregoing corresponding method embodiment, and the repetition is not repeated.

The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

In the 90 s of the 20 th century, improvements to one technology could clearly be distinguished as improvements in hardware (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software (improvements to the process flow). However, with the development of technology, many improvements of the current method flows can be regarded as direct improvements of hardware circuit structures. Designers almost always obtain corresponding hardware circuit structures by programming improved method flows into hardware circuits. Therefore, an improvement of a method flow cannot be said to be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device, PLD) (e.g., field programmable gate array (Field Programmable Gate Array, FPGA)) is an integrated circuit whose logic function is determined by the programming of the device by a user. A designer programs to "integrate" a digital system onto a PLD without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, such programming is mostly implemented by using "logic compiler" software, which is similar to the software compiler used in program development and writing, and the original code before the compiling is also written in a specific programming language, which is called hardware description language (Hardware Description Language, HDL), but not just one of the hdds, but a plurality of kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), lava, lola, myHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog are currently most commonly used. It will also be apparent to those skilled in the art that a hardware circuit implementing the logic method flow can be readily obtained by merely slightly programming the method flow into an integrated circuit using several of the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic controllers, and embedded microcontrollers, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, atmel AT91SAM, microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic of the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller may thus be regarded as a kind of hardware component, and means for performing various functions included therein may also be regarded as structures within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.

The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functionality of the units may be implemented in one or more software and/or hardware when implementing one or more embodiments of the present description.

It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

Embodiments of the present description are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the specification. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable fraud case serial-to-parallel device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable fraud case serial-to-parallel device, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.

It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

One or more embodiments of the present specification may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more embodiments of the present description may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.

The foregoing is merely exemplary of the present disclosure and is not intended to limit the present disclosure. Various modifications and alterations to this specification will become apparent to those skilled in the art. Any modifications, equivalent substitutions, improvements, or the like, which are within the spirit and principles of the present description, are intended to be included within the scope of the claims of the present description.

Claims (15)

1. A risk control policy determination method, the method comprising:

acquiring account feature data of a target account;

inputting the account feature data into a control probability prediction model, and acquiring control probability and non-control probability corresponding to a candidate risk control strategy; the control probability prediction model is used for predicting control probability, the control probability corresponding to the candidate risk control strategy is the probability of performing risk control on the target account based on the candidate risk control strategy, and the non-control probability is the probability of not performing risk control on the target account;

inputting the account characteristic data into a risk increment prediction model corresponding to the candidate risk control strategy, and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, wherein the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed;

Determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy;

determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy; the target risk control strategy is used for performing risk control on the target account.

2. The risk control policy determining method according to claim 1, wherein determining the risk conversion increment corresponding to the candidate risk control policy according to the control probability and the non-control probability corresponding to the candidate risk control policy and the first risk increment and the second risk increment corresponding to the candidate risk control policy includes:

weighting a first risk increment corresponding to the candidate risk control strategy by using the control probability corresponding to the candidate risk control strategy to obtain a first component of a risk conversion increment corresponding to the candidate risk control strategy;

weighting a second risk increment corresponding to the candidate risk control strategy by using the uncontrolled probability to obtain a second component of a risk conversion increment corresponding to the candidate risk control strategy;

And obtaining the risk conversion increment corresponding to the candidate risk control strategy according to the sum of the first component and the second component of the risk conversion increment corresponding to the candidate risk control strategy.

3. The risk control policy determining method according to claim 1, when the number of the candidate risk control policies is one, the determining a target risk control policy according to a risk conversion increment corresponding to the candidate risk control policy includes:

comparing the risk conversion increment corresponding to the candidate risk control strategy with a preset risk conversion increment threshold, and determining the candidate risk control strategy as the target risk control strategy if the risk conversion increment corresponding to the candidate risk control strategy is smaller than the risk conversion increment threshold; wherein the risk conversion increment threshold is a value no greater than 0.

4. The risk control policy determining method according to claim 1, when the number of candidate risk control policies is plural, the determining a target risk control policy according to a risk conversion increment corresponding to the candidate risk control policies includes:

and determining the candidate risk control strategy with the minimum risk conversion increment as a target risk control strategy according to the risk conversion increment corresponding to each candidate risk control strategy in the plurality of candidate risk control strategies.

5. The risk control strategy determination method according to claims 1-4, wherein the risk increment prediction model corresponding to the candidate risk control strategy includes: a first risk delta prediction model and a second risk delta prediction model, the method further comprising:

acquiring first sample data and second sample data corresponding to the candidate risk control strategy; the first sample data comprises account feature data of a first sample account and first risk probability corresponding to the first sample account, wherein the first risk probability is the probability of risk occurrence of the account under the condition that risk control is not performed on the account; the second sample data comprises account feature data of a second sample account and second risk probability corresponding to the second sample account, wherein the second risk probability is the probability of risk occurrence of the account when the candidate risk control strategy is used for risk control of the account;

inputting the account feature data of the first sample account into a second risk prediction model corresponding to the candidate risk control strategy, and predicting a second risk probability corresponding to the first sample data; inputting account feature data of the second sample account into a first risk prediction model, and predicting a first risk probability corresponding to the second sample account; the first risk prediction model is used for predicting a first risk probability corresponding to the account according to the account feature data, and the second risk prediction model is used for predicting a second risk probability corresponding to the account according to the account feature data;

Obtaining a first risk increment corresponding to the first sample data according to a difference value between a second risk probability corresponding to the first sample data and the first risk probability; obtaining a second risk increment corresponding to the second sample account according to the difference value between the second risk probability corresponding to the second sample account and the first risk probability;

performing supervised training on a first risk increment prediction model to be trained by using account feature data of the first sample account and a first risk increment corresponding to the first sample account to obtain a first risk increment prediction model corresponding to the candidate risk control strategy; and performing supervised training on a second risk increment prediction model to be trained by using the account feature data of the second sample account and a second risk increment corresponding to the second sample account to obtain a second risk increment prediction model corresponding to the candidate risk control strategy.

6. The risk control policy determination method of claim 5, the method further comprising:

performing supervised training on a first risk prediction model to be trained by using account feature data of the first sample account and a first risk probability corresponding to the first sample account to obtain the first risk prediction model;

And performing supervised training on the second risk prediction model to be trained by using the account feature data of the second sample account and the second risk probability corresponding to the second sample account to obtain a second risk prediction model corresponding to the candidate risk control strategy.

7. The risk control policy determination method of claim 1, the method further comprising:

and determining the account category to which the target account belongs according to the risk conversion increment corresponding to the target risk control strategy.

8. The risk control policy determining method according to claim 7, wherein determining, according to the risk conversion increment corresponding to the target risk control policy, an account category to which the target account belongs includes:

when the risk conversion increment corresponding to the target risk control strategy is smaller than a first increment threshold, determining that the target account belongs to a first type account; wherein the first delta threshold is not greater than 0, and the first type account is: if risk control is performed on the account based on the target risk control policy, the probability of risk occurrence can be reduced compared with an account in which risk control is not performed;

when the risk conversion increment corresponding to the target risk control strategy is larger than a second increment threshold, determining that the current account belongs to a second type of account; wherein the second increment threshold is not less than 0, and the second type account is: if risk control is performed on the account based on the target risk control policy, the probability of risk occurrence is increased instead compared with the account without risk control;

When the risk conversion increment corresponding to the target candidate risk control strategy is between the first increment threshold and the second increment threshold, acquiring a first risk probability and a second risk probability corresponding to the target risk control strategy for the target account; the first risk probability is the probability of risk occurrence of the account under the condition that risk control is not performed on the account, and the second risk probability corresponding to the target risk control strategy is the probability of risk occurrence of the account under the condition that risk control is performed on the account based on the target risk control strategy;

when the first risk probability and the second risk probability corresponding to the target risk control strategy are both smaller than a preset risk threshold value, determining that the target account belongs to a third class account; wherein the third type of account is: whether risk control is performed on accounts or not on accounts with low risk occurrence probability based on the target risk control strategy;

when the first risk probability and the second risk probability corresponding to the target risk control strategy are both larger than a preset risk threshold, determining that the target account belongs to a fourth type account; wherein the fourth type of account is: whether risk control is performed on accounts based on the target risk control strategy or not, accounts with high risk occurrence probability are subjected to risk control.

9. The risk control policy determination method of claim 8, the method further comprising:

and executing risk control on the target account by using the target risk control strategy when the target account belongs to the first type of account.

10. The risk control policy determining method according to claim 1, wherein the determining the target risk control policy according to the risk conversion increment corresponding to the candidate risk control policy includes:

determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy and the problem conversion increment corresponding to the candidate risk control strategy; the problem conversion increment represents a probability increment that the account experiences a problem when risk control is performed based on the candidate risk control strategy compared with when risk control is not performed.

11. The risk control policy determination method of claim 10, the method further comprising:

inputting the account characteristic data into a problem increment prediction model corresponding to the candidate risk control strategy, and predicting a first problem increment and a second problem increment corresponding to the candidate risk control strategy; the problem increment prediction model corresponding to the candidate risk control strategy is used for predicting a first problem increment and a second problem increment corresponding to the candidate risk control strategy, wherein the first problem increment represents a probability increment that an account experiences a problem when risk control is performed based on the candidate risk control strategy, and the risk control is performed compared with the risk control is not performed; the second risk increment represents a probability increment that the risk control is originally performed based on the candidate risk control strategy, and if the risk control is not performed, the risk control is performed compared with the risk control is not performed, and the account experiences problems;

And determining a problem conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first problem increment and the second problem increment corresponding to the candidate risk control strategy.

12. The risk control policy determination method of claim 10, the method further comprising:

determining the account category of the target account according to the risk conversion increment corresponding to the target risk control strategy and the problem conversion increment corresponding to the target risk control strategy;

and determining whether to execute risk control on the target account by using the target risk control strategy according to the account category to which the target account belongs.

13. A risk control policy determination device, the device comprising:

the feature acquisition module is used for acquiring account feature data of the target account;

the control prediction module is used for inputting the account feature data into a control probability prediction model to obtain control probability and non-control probability corresponding to a candidate risk control strategy, wherein the control probability prediction model is used for predicting the control probability, the control probability corresponding to the candidate risk control strategy is the probability of performing risk control on the target account based on the candidate risk control strategy, and the non-control probability is the probability of not performing risk control on the target account;

The risk increment prediction module is used for inputting the account characteristic data into a risk increment prediction model corresponding to the candidate risk control strategy and predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy; the risk increment prediction model corresponding to the candidate risk control strategy is used for predicting a first risk increment and a second risk increment corresponding to the candidate risk control strategy, wherein the first risk increment represents a probability increment that risk is generated in an account when risk control is performed based on the candidate risk control strategy and compared with risk control is not performed; the second risk increment represents a probability increment that risk is generated in the account when risk control is not performed compared with risk control is not performed if the risk control is not performed;

the conversion increment determining module is used for determining a risk conversion increment corresponding to the candidate risk control strategy according to the control probability and the non-control probability corresponding to the candidate risk control strategy and the first risk increment and the second risk increment corresponding to the candidate risk control strategy;

The target strategy determining module is used for determining a target risk control strategy according to the risk conversion increment corresponding to the candidate risk control strategy, wherein the target risk control strategy is used for performing risk control on the target account.

14. A computer apparatus, the apparatus comprising:

a processor; and

a memory arranged to store computer executable instructions which, when executed, cause the processor to perform the method of any of claims 1-12.

15. A storage medium storing computer executable instructions which, when executed by a processor, implement the steps in the method of any one of claims 1-12.