A model processing method, device and equipment

Abstract

The embodiments of this specification disclose a model processing method, apparatus, and device, the method comprising: obtaining a resource loss objective function and a group trust score constraint function of a risk prevention and control model applied in a trusted scenario; constructing a corresponding optimization function based on the resource loss objective function and a corresponding first weight, and the group trust score constraint function and a corresponding second weight; determining optimization target information corresponding to the resource loss objective function and the group trust score constraint function based on the constructed optimization function, the resource loss objective function, and the group trust score constraint function; and performing model training on the risk prevention and control model applied in the trusted scenario based on the determined optimization target information and through training samples to obtain a trained risk prevention and control model.

Description

Technical Field

[0001] This document relates to the field of computer technology, and in particular to a model processing method, device and equipment. Background Art

[0002] When users use designated applications to transfer resources (such as payments or transfers), they often need to conduct risk analysis on the transaction. The goal of risk control models in trusted scenarios is to identify risk-free business data and quickly release it. Due to the critical role of risk control models in trusted scenarios, they must treat different user groups fairly. If risk control models cannot guarantee fairness during the trusted release of business data, ensuring that they treat different user groups equally becomes a critical issue that needs to be addressed.

[0003] Generally, the unfairness of risk prevention and control models is mainly caused by data imbalance. Therefore, the data distribution can be adjusted to ensure balanced distribution of different subgroups. However, the above method only relies on data distribution and cannot ensure that the risk prevention and control model itself guarantees fairness during the learning process. In addition, in trusted scenarios, for some subgroups, unfairness may not be caused by uneven distribution, but by other factors. In addition, a risk prevention and control model can be constructed for each group separately. However, if there are many group categories, multiple different risk prevention and control models need to be constructed, which will result in high modeling costs and complex maintenance. For this reason, it is necessary to provide a technical solution that can enable the risk prevention and control model to treat different user groups equally during the trusted release process, simplify the model construction process, and improve the model maintenance efficiency. Summary of the Invention

[0004] The purpose of the embodiments of this specification is to provide a technical solution that enables a risk prevention and control model to treat different user groups equally during the trusted release process, and to simplify the model construction process and improve the efficiency of model maintenance.

[0005] In order to implement the above technical solution, the embodiments of this specification are implemented as follows:

[0006] An embodiment of this specification provides a model processing method, the method comprising: obtaining a resource loss objective function and a group credibility score constraint function of a risk prevention and control model applied in a trusted scenario. Based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, a corresponding optimization function is constructed. Based on the constructed optimization function, the resource loss objective function and the group credibility score constraint function, the optimization target information corresponding to the resource loss objective function and the group credibility score constraint function is determined. Based on the determined optimization target information, the risk prevention and control model applied in the trusted scenario is trained through training samples to obtain a trained risk prevention and control model.

[0007] An embodiment of this specification provides a model processing device, which includes: a function acquisition module, which acquires a resource loss objective function and a group trust score constraint function of a risk prevention and control model applied in a trusted scenario. An optimization function determination module, which constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group trust score constraint function and the corresponding second weight. An optimization target determination module, which determines the optimization target information corresponding to the resource loss objective function and the group trust score constraint function based on the constructed optimization function, the resource loss objective function and the group trust score constraint function. A training module, which performs model training on the risk prevention and control model applied in the trusted scenario based on the determined optimization target information and through training samples to obtain a trained risk prevention and control model.

[0008] An embodiment of this specification provides a model processing device, the model processing device comprising: a processor; and a memory arranged to store computer-executable instructions, wherein the executable instructions, when executed, cause the processor to: obtain a resource loss objective function and a group trust score constraint function of a risk prevention and control model applied in a trusted scenario. Based on the resource loss objective function and the corresponding first weight, and the group trust score constraint function and the corresponding second weight, a corresponding optimization function is constructed. Based on the constructed optimization function, the resource loss objective function and the group trust score constraint function, the optimization target information corresponding to the resource loss objective function and the group trust score constraint function is determined. Based on the determined optimization target information, and through training samples, the risk prevention and control model applied in the trusted scenario is trained to obtain a trained risk prevention and control model.

[0009] An embodiment of this specification also provides a storage medium for storing computer-executable instructions, which implement the following process when executed by a processor: obtaining a resource loss objective function and a group trust score constraint function of a risk prevention and control model applied in a trusted scenario. Based on the resource loss objective function and the corresponding first weight, and the group trust score constraint function and the corresponding second weight, a corresponding optimization function is constructed. Based on the constructed optimization function, the resource loss objective function and the group trust score constraint function, the optimization target information corresponding to the resource loss objective function and the group trust score constraint function is determined. Based on the determined optimization target information, the risk prevention and control model applied in the trusted scenario is trained through training samples to obtain a trained risk prevention and control model. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0011] Figure 1 This is an embodiment of a processing method for a model of this specification;

[0012] Figure 2 This is another example of a processing method for this specification;

[0013] Figure 3 A schematic diagram of the processing process of a model in this specification;

[0014] Figure 4 This is a processing device embodiment of a model of this specification;

[0015] Figure 5 This is a processing device embodiment of a model in this specification. DETAILED DESCRIPTION

[0016] The embodiments of this specification provide a model processing method, device and equipment.

[0017] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments derived by those skilled in the art based on the embodiments in this specification without creative effort shall fall within the scope of protection of this specification.

[0018] Example 1

[0019] like Figure 1 As shown, the embodiments of this specification provide a model processing method, the execution subject of the method can be a terminal device or a server, etc., wherein the terminal device can be a mobile terminal device such as a mobile phone, a tablet computer, or a computer device such as a laptop or desktop computer, or an IoT device (specifically such as a smart watch, a car device, etc.), etc., wherein the server can be an independent server, or a server cluster composed of multiple servers, etc., and the server can be a backend server for a financial service or an online shopping service, etc., or a backend server for an application, etc. The method can specifically include the following steps:

[0020] In step S102, a resource loss objective function and a group credibility score constraint function of a risk prevention and control model applied in a credible scenario are obtained.

[0021] Among them, a trusted scenario can be a scenario where the probability of risk occurrence is relatively stable and far lower than the probability of overall market risk. The accumulation of trusted data in a trusted scenario can help release low-risk transaction events and reduce the amount of analysis at the identification layer. A risk prevention and control model can be a model used to prevent and control risks for a certain business (such as a payment business, instant messaging business, or information recommendation business). The risk prevention and control model can be constructed using a variety of different algorithms. For example, the risk prevention and control model can be constructed using a neural network model, or the risk prevention and control model can be constructed using a classification algorithm, etc. The specific setting can be based on actual conditions and is not limited in this embodiment of the specification. The risk prevention and control model in this embodiment can be a trusted model applied to a trusted scenario, wherein it can be lightweight in engineering deployment and the probability of risk occurrence is stable and far lower than the probability of overall market risk. The model used in the above-mentioned model is a trusted model. The resource loss objective function can be an objective function involved in the risk prevention and control model, wherein the resource loss can be a financial loss, etc., and can be specifically set according to actual conditions and is not limited in this embodiment of the specification. The group trust score constraint function can be a function corresponding to the group trust score as a constraint. The group trust score constraint function can be determined by constraints related to the variables in the resource loss objective function.

[0022] In practice, when users use designated applications to transfer resources (such as payments or transfers), they often need to conduct risk analysis on the transaction. Typically, risks associated with a transaction may include theft, fraud, and designated illegal financial activities. The goal of risk prevention and control models in trusted scenarios is to identify risk-free business data and quickly release it. This reduces disruption to users while conserving system computing resources. Generally speaking, risk prevention and control models applied in trusted scenarios can release over 90% of the business data traffic in the risk prevention and control system, leaving only around 10% of the traffic flowing to the deep analysis layer for refined analysis. Due to the key position of risk prevention and control models in trusted scenarios in the risk prevention and control system, risk prevention and control models are required to treat different user groups fairly. If the risk prevention and control models cannot guarantee fairness during the trusted release of business data (fairness has different interpretations in different scenarios. In trusted scenarios, fairness refers to group fairness. Group fairness means dividing different users into predefined groups based on sensitive or protected attributes (such as ethnicity and gender), and ensuring that the statistical measurement standards of the results are equal among all groups. If there are groups divided by gender, similar decisions should be made for the entire group, and no one gender should be favored), it will also cause trouble for users' business activities. Therefore, how to ensure that the risk prevention and control model can treat different user groups equally has become an important issue that needs to be addressed. Generally, the unfairness of the risk prevention and control model is mainly caused by the imbalance of data. Therefore, the data distribution can be adjusted to ensure the balanced distribution of different subgroups. However, the above method only relies on the data distribution and cannot ensure that the risk prevention and control model itself guarantees fairness during the learning process. In addition, in a trusted scenario, for some subgroups, unfairness may not be caused by the imbalance of distribution, but by the different distribution of trust scores given to the subgroups by the risk prevention and control model. In addition, a risk prevention and control model can be constructed for each group separately. However, if there are many group categories, it is necessary to construct multiple different risk prevention and control models, which will lead to high modeling costs and complex maintenance. For this reason, it is necessary to provide a technical solution that can enable the risk prevention and control model to treat different user groups equally during the trusted release process, simplify the model construction process, and improve the maintenance efficiency of the model. The embodiment of this specification provides a feasible technical solution, which can specifically include the following content.

[0023] When it is necessary to construct and train a risk prevention and control model for use in a trusted scenario, one or more different algorithms can be pre-selected according to the actual situation to construct the risk prevention and control model. For example, the risk prevention and control model can be constructed based on a convolutional neural network model, or the risk prevention and control model can be constructed based on a k-nearest neighbor classification algorithm, a decision tree classification algorithm, or a naive Bayes classification algorithm, etc. The specific setting can be based on the actual situation. After the risk prevention and control model is constructed in the above manner, the risk prevention and control model can be trained. During the model training process, in order to enable the risk prevention and control model to converge to achieve the purpose of being applied to a specified business, an optimization goal can be set for the risk prevention and control model. In order to obtain the optimization goal, the resource loss objective function of the risk prevention and control model applied to the trusted scenario can be obtained, and the group trust score can be used as a constraint to determine the corresponding constraint function, that is, the group trust score constraint function. In this embodiment, the resource loss objective function and the group trust score constraint function of the risk prevention and control model applied to the trusted scenario obtained can include:

[0024]

[0025] Among them, min θ∈Θ g0(θ) is the resource loss objective function, g i (θ)≤0 is the group credibility score constraint function, θ is a variable, that is, it can be the model parameter of the risk prevention and control model, i represents the i-th constraint condition, Θ is the value range of the variable θ, and m represents the maximum value of the constraint condition.

[0026] In step S104, a corresponding optimization function is constructed based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight.

[0027] The first weight may be the weight corresponding to the resource loss objective function, which may be an unknown variable to be determined. In practical applications, the value or value range of the first weight may be pre-set based on actual conditions, and may be specifically set based on actual conditions. The second weight may be the weight corresponding to the group credibility score constraint function, which may be an unknown variable to be determined. In practical applications, the value or value range of the second weight may be pre-set based on actual conditions, and may be specifically set based on actual conditions.

[0028] In implementation, corresponding weights can be set for the resource loss objective function and the group credibility score constraint function according to actual conditions, that is, a first weight is set for the resource loss objective function, and a second weight is set for the group credibility score constraint function. Then, the resource loss objective function can be multiplied by the first weight, and the group credibility score constraint function can be multiplied by the second weight. The two multiplied parts are added together (that is, the resource loss objective function and the group credibility score constraint function are weighted separately, and then added together) to obtain a corresponding expression, which can be an optimization function. For example, based on the example of the resource loss objective function and the group credibility score constraint function, the optimization function can be

[0029]

[0030] Among them, λ0 is the first weight, λ i is the second weight, and Y(θ,λ) is the optimization function.

[0031] In step S106 , based on the constructed optimization function, resource loss objective function and group credibility score constraint function, optimization target information corresponding to the resource loss objective function and the group credibility score constraint function is determined.

[0032] In implementation, after obtaining the corresponding optimization function through the above method, the corresponding optimization target information can be determined by combining the resource loss objective function and the group credibility score constraint function. Specifically, the optimization function is

[0033]

[0034] The resource loss objective function and group credibility score constraint function are:

[0035]

[0036] Then the optimization target information corresponding to the resource loss objective function and the group credibility score constraint function can be the combination of the above two expressions, that is, as follows

[0037]

[0038] If λ0min θ∈Θ If g0(θ) holds, then we need Minimum, that is Maximum, at this time, it becomes calculation The maximum value of Then calculate The maximum value and It can be used as the final optimization target information.

[0039] In step S108, based on the determined optimization target information and using training samples, the risk prevention and control model applied in the trusted scenario is trained to obtain a trained risk prevention and control model.

[0040] During implementation, a certain amount of historical business data can be obtained for a specified business. This data can be used as training samples. This training sample can then be input into a pre-built risk prevention and control model to obtain the corresponding results. Based on this result and using the optimization target information, the parameters in the risk prevention and control model can be adjusted. The risk prevention and control model can then be trained using the training samples and the optimization target information. Ultimately, a converged trained risk prevention and control model can be obtained. This risk prevention and control model can then be deployed in the specified business. Since the group credibility score is used as a constraint function to constrain its fairness, the fairness of the risk prevention and control model for subgroups is improved.

[0041] The embodiment of this specification provides a model processing method, which obtains the resource loss objective function and the group trust score constraint function of the risk prevention and control model applied in the trustworthy scenario, and constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, as well as the group trust score constraint function and the corresponding second weight. Then, based on the constructed optimization function, resource loss objective function and group trust score constraint function, the optimization target information corresponding to the resource loss objective function and the group trust score constraint function is determined. Finally, based on the determined optimization target information and through the training samples, the risk prevention and control model applied in the trustworthy scenario can be trained to obtain the trained risk prevention and control model. In this way, it does not depend on the data distribution and does not need to be constructed separately for each group. The risk prevention and control model is not based on the resource loss objective function and the group trust score constraint function, and combined with the corresponding solution method, which improves the fairness of the risk prevention and control model to sub-groups in the trusted scenario, and can be extended to other fairness issues. In addition, for the fairness of trusted release, the group trust score constraint fairness is used to improve the distribution of trust scores, thereby ensuring fairness in trusted release. In addition, since there is no need to construct a risk prevention and control model for each group separately, there is no need to construct multiple different risk prevention and control models, thereby reducing the modeling cost and simplifying the maintenance process. Based on the above processing, this technical solution can enable the risk prevention and control model to treat different user groups equally in the trusted release process, and can simplify the model construction process and improve the model maintenance efficiency.

[0042] Example 2

[0043] like Figure 2As shown, the embodiments of this specification provide a model processing method, the execution subject of the method can be a terminal device or a server, etc., wherein the terminal device can be a mobile terminal device such as a mobile phone, a tablet computer, or a computer device such as a laptop or desktop computer, or an IoT device (specifically such as a smart watch, a car device, etc.), etc., wherein the server can be an independent server, or a server cluster composed of multiple servers, etc., and the server can be a backend server for a financial service or an online shopping service, etc., or a backend server for an application, etc. The method can specifically include the following steps:

[0044] In step S202, a resource loss objective function and a group credibility score constraint function of a risk prevention and control model applied in a credible scenario are obtained.

[0045] The risk prevention and control model can be constructed using a neural network model. The risk prevention and control model can be used to obtain data from business data of preset businesses that does not contain preset risks, where the preset risks may include one or more of the following: fraud risk, theft risk, and illegal financial activity risk.

[0046] During implementation, the resource loss objective function and group credibility score constraint function of the risk prevention and control model applied to the trusted scenario may include:

[0047]

[0048] In step S204, based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, a corresponding Lagrangian function is constructed, and the constructed Lagrangian function is used as the optimization function.

[0049] In implementation, corresponding weights can be set for the resource loss objective function and the group credibility score constraint function according to actual conditions, that is, a first weight is set for the resource loss objective function, and a second weight is set for the group credibility score constraint function. Then, as Figure 3 As shown, the corresponding Lagrangian function can be constructed based on the above resource loss objective function and the corresponding first weight, as well as the above group credibility score constraint function and the corresponding second weight. For example, based on the above resource loss objective function and group credibility score constraint function, the Lagrangian function can be

[0050]

[0051] in, is a Lagrangian function, which can be used as an optimization function. In this way, the optimization problem of the risk prevention and control model can be transformed into a Lagrangian function.

[0052] In step S206 , based on the constructed optimization function, resource loss objective function and group credibility score constraint function, optimization target information corresponding to the resource loss objective function and the group credibility score constraint function is determined.

[0053] The optimization target information may include first sub-optimization target information and second sub-optimization target information. The first sub-optimization target information may be optimization target information for the training sample, and the second sub-optimization target information may be optimization target information for the second weight.

[0054] In implementation, after obtaining the corresponding optimization function in the above manner, the constrained optimization problem of the above risk prevention and control model can be transformed into a game problem by combining the resource loss objective function and the group credibility score constraint function, thereby obtaining the optimization target information (including the first sub-optimization target information and the second sub-optimization target information) corresponding to the resource loss objective function and the group credibility score constraint function. The constrained optimization problem is a type of mathematical optimization problem, which consists of two parts: the objective function and the constraints related to the variables in the objective function. The optimization process is to optimize (maximize or minimize) the objective function under the constraints. The game problem considers the predicted behavior and actual behavior of individuals in the game and studies their optimization strategies. Behaviors with competitive or confrontational nature are called game behaviors. In the above behaviors, the parties participating in the struggle or competition each have different goals or interests. In order to achieve their respective goals and interests, each party must consider the opponent's various possible action plans and strive to select the most beneficial or reasonable plan for themselves.

[0055] like Figure 3 As shown, based on the example of the optimization function (i.e., Lagrangian function) constructed above, the optimization target information corresponding to the resource loss objective function and the group credibility score constraint function may include first sub-optimization target information and second sub-optimization target information, wherein the first sub-optimization target information may be

[0056]

[0057] The second sub-optimization target information can be

[0058]

[0059] in, for Figure 3 The optimization objective of the θ-player in the minimum problem is, for Figure 3 The optimization goal of the λ-player in the largest problem is, g iThe proxy constraint of (θ) can be understood through the following example: if a constraint condition is x<2, the constraint condition can be relaxed to obtain x<2+delta. The obtained x<2+delta is a proxy constraint. Similar processes can be called proxy constraints, and the specific settings can be based on actual conditions.

[0060] The result of solving the above game problem is to obtain the optimization target information corresponding to the resource loss objective function and the group credibility score constraint function.

[0061] In step S208, historical service data of users belonging to different subgroups in the preset service is obtained, and the obtained historical service data is used as training samples.

[0062] Among them, the preset services may include multiple types, for example, the preset services may be payment services, transfer services or instant messaging services, etc., which can be specifically set according to actual conditions and are not limited in this embodiment of the present specification.

[0063] In step S210, based on the determined optimization target information, the training sample is input into the risk prevention and control model, and the feasible solution corresponding to the optimization target information is determined by the risk prevention and control model. The risk prevention and control model selects the target feasible solution that meets the resource loss selection rule from the feasible solutions corresponding to the optimization target information according to the preset resource loss selection rule, and uses the target feasible solution as the output result of the risk prevention and control model.

[0064] The resource loss selection rule may be a capital selection rule. The resource loss selection rule may be set according to actual conditions, and the embodiments of this specification do not limit this.

[0065] In step S212, if the above output result does not meet the preset convergence condition, continue to train the risk prevention and control model applied to the trusted scenario based on the determined optimization target information and through training samples until the output result meets the preset convergence condition, thereby obtaining the trained risk prevention and control model.

[0066] The embodiment of this specification provides a model processing method, which obtains the resource loss objective function and the group trust score constraint function of the risk prevention and control model applied in the trustworthy scenario, and constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, as well as the group trust score constraint function and the corresponding second weight. Then, based on the constructed optimization function, resource loss objective function and group trust score constraint function, the optimization target information corresponding to the resource loss objective function and the group trust score constraint function is determined. Finally, based on the determined optimization target information and through the training samples, the risk prevention and control model applied in the trustworthy scenario can be trained to obtain the trained risk prevention and control model. In this way, it does not depend on the data distribution and does not need to be constructed separately for each group. The risk prevention and control model is not based on the resource loss objective function and the group trust score constraint function, and combined with the corresponding solution method, which improves the fairness of the risk prevention and control model to sub-groups in the trusted scenario, and can be extended to other fairness issues. In addition, for the fairness of trusted release, the group trust score constraint fairness is used to improve the distribution of trust scores, thereby ensuring fairness in trusted release. In addition, since there is no need to construct a risk prevention and control model for each group separately, there is no need to construct multiple different risk prevention and control models, thereby reducing the modeling cost and simplifying the maintenance process. Based on the above processing, this technical solution can enable the risk prevention and control model to treat different user groups equally in the trusted release process, and can simplify the model construction process and improve the model maintenance efficiency.

[0067] Example 3

[0068] The above is a processing method of the model provided in the embodiment of this specification. Based on the same idea, the embodiment of this specification also provides a processing device of the model, such as Figure 4 shown.

[0069] The processing device of the model includes: a function acquisition module 401, an optimization function determination module 402, an optimization target determination module 403 and a training module 404, wherein:

[0070] Function acquisition module 401, which acquires the resource loss objective function and group credibility score constraint function of the risk prevention and control model applied in the trustworthy scenario;

[0071] An optimization function determination module 402 constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight;

[0072] The optimization target determination module 403 determines the optimization target information corresponding to the resource loss target function and the group credibility score constraint function based on the constructed optimization function, the resource loss target function and the group credibility score constraint function;

[0073] The training module 404 performs model training on the risk prevention and control model applied in the trustworthy scenario based on the determined optimization target information and using training samples to obtain a trained risk prevention and control model.

[0074] In the embodiments of this specification, the risk prevention and control model is constructed by a neural network model.

[0075] In the embodiment of this specification, the risk prevention and control model is used to obtain data without preset risks from the business data of preset businesses, and the preset risks include one or more of fraud risks, theft risks, and illegal financial activity risks.

[0076] In the embodiment of this specification, the optimization function determination module 402 constructs a corresponding Lagrangian function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, and uses the constructed Lagrangian function as the optimization function.

[0077] In the embodiment of this specification, the device further includes:

[0078] The sample acquisition module acquires historical business data of users belonging to different subgroups in a preset business, and uses the acquired historical business data as the training samples.

[0079] In an embodiment of this specification, the optimization target information includes first sub-optimization target information and second sub-optimization target information. The first sub-optimization target information is the optimization target information for the training sample, and the second sub-optimization target information is the optimization target information for the second weight.

[0080] In the embodiment of this specification, the training module 404 includes:

[0081] an output unit, based on the determined optimization target information, inputting the training sample into the risk prevention and control model, determining a feasible solution corresponding to the optimization target information through the risk prevention and control model, and selecting a target feasible solution that satisfies the resource loss selection rule from the feasible solutions corresponding to the optimization target information through the risk prevention and control model according to a preset resource loss selection rule, and using the target feasible solution as an output result of the risk prevention and control model;

[0082] The training unit continues to train the risk prevention and control model applied in the trusted scenario based on the determined optimization target information and through training samples if the output result does not meet the preset convergence condition, thereby obtaining a trained risk prevention and control model.

[0083] The embodiment of this specification provides a model processing device, which obtains the resource loss objective function and the group trust score constraint function of the risk prevention and control model applied in the trustworthy scenario, and constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, as well as the group trust score constraint function and the corresponding second weight. Then, based on the constructed optimization function, resource loss objective function and group trust score constraint function, the optimization target information corresponding to the resource loss objective function and the group trust score constraint function is determined. Finally, based on the determined optimization target information and through the training samples, the risk prevention and control model applied in the trustworthy scenario can be trained to obtain the trained risk prevention and control model. In this way, it does not depend on the data distribution and does not need to be constructed separately for each group. The risk prevention and control model is not based on the resource loss objective function and the group trust score constraint function, and combined with the corresponding solution method, which improves the fairness of the risk prevention and control model to sub-groups in the trusted scenario, and can be extended to other fairness issues. In addition, for the fairness of trusted release, the group trust score constraint fairness is used to improve the distribution of trust scores, thereby ensuring fairness in trusted release. In addition, since there is no need to construct a risk prevention and control model for each group separately, there is no need to construct multiple different risk prevention and control models, thereby reducing the modeling cost and simplifying the maintenance process. Based on the above processing, this technical solution can enable the risk prevention and control model to treat different user groups equally in the trusted release process, and can simplify the model construction process and improve the model maintenance efficiency.

[0084] Example 4

[0085] The above is a processing device for the model provided in the embodiment of this specification. Based on the same idea, the embodiment of this specification also provides a processing device for the model, such as Figure 5 shown.

[0086] The processing device of the model may be arranged in the terminal device or the server provided in the above embodiment.

[0087] The processing device of the model may vary greatly due to different configurations or performances, and may include one or more processors 501 and memory 502, and the memory 502 may store one or more storage applications or data. Among them, the memory 502 may be a temporary storage or a persistent storage. The application stored in the memory 502 may include one or more modules (not shown in the figure), each module may include a series of computer-executable instructions in the processing device of the model. Furthermore, the processor 501 may be configured to communicate with the memory 502 to execute a series of computer-executable instructions in the memory 502 on the processing device of the model. The processing device of the model may also include one or more power supplies 503, one or more wired or wireless network interfaces 504, one or more input and output interfaces 505, and one or more keyboards 506.

[0088] Specifically, in this embodiment, the model processing 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 in the model processing device, and the one or more programs are configured to be executed by one or more processors, including computer-executable instructions for performing the following:

[0089] Obtain the resource loss objective function and group credibility score constraint function of the risk prevention and control model applied to the trustworthy scenario;

[0090] Constructing a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight;

[0091] Determining optimization target information corresponding to the resource loss objective function and the group credibility score constraint function based on the constructed optimization function, the resource loss objective function, and the group credibility score constraint function;

[0092] Based on the determined optimization target information, the risk prevention and control model applied in the trusted scenario is trained through training samples to obtain a trained risk prevention and control model.

[0093] In the embodiments of this specification, the risk prevention and control model is constructed by a neural network model.

[0094] In the embodiment of this specification, the risk prevention and control model is used to obtain data without preset risks from the business data of preset businesses, and the preset risks include one or more of fraud risks, theft risks, and illegal financial activity risks.

[0095] In the embodiment of this specification, the corresponding optimization function is constructed based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, including:

[0096] Based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, a corresponding Lagrangian function is constructed, and the constructed Lagrangian function is used as the optimization function.

[0097] In the embodiment of this specification, it also includes:

[0098] The historical service data of users belonging to different subgroups in the preset service are obtained, and the obtained historical service data are used as the training samples.

[0099] In an embodiment of this specification, the optimization target information includes first sub-optimization target information and second sub-optimization target information. The first sub-optimization target information is the optimization target information for the training sample, and the second sub-optimization target information is the optimization target information for the second weight.

[0100] In the embodiments of this specification, the risk prevention and control model applied to the trusted scenario is trained based on the determined optimization target information and through training samples to obtain the trained risk prevention and control model, including:

[0101] Based on the determined optimization target information, the training sample is input into the risk prevention and control model, a feasible solution corresponding to the optimization target information is determined by the risk prevention and control model, and a target feasible solution that satisfies the resource loss selection rule is selected from the feasible solutions corresponding to the optimization target information by the risk prevention and control model according to a preset resource loss selection rule, and the target feasible solution is used as an output result of the risk prevention and control model;

[0102] If the output result does not meet the preset convergence conditions, continue to train the risk prevention and control model applied to the trusted scenario based on the determined optimization target information and through training samples until the output result meets the preset convergence conditions, thereby obtaining the trained risk prevention and control model.

[0103] The embodiment of this specification provides a model processing device, which obtains the resource loss objective function and the group trust score constraint function of the risk prevention and control model applied in the trustworthy scenario, and constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, as well as the group trust score constraint function and the corresponding second weight. Then, based on the constructed optimization function, resource loss objective function and group trust score constraint function, the optimization target information corresponding to the resource loss objective function and the group trust score constraint function is determined. Finally, based on the determined optimization target information and through the training samples, the risk prevention and control model applied in the trustworthy scenario can be trained to obtain the trained risk prevention and control model. In this way, it does not depend on the data distribution and does not need to be constructed separately for each group. The risk prevention and control model is not based on the resource loss objective function and the group trust score constraint function, and combined with the corresponding solution method, which improves the fairness of the risk prevention and control model to sub-groups in the trusted scenario, and can be extended to other fairness issues. In addition, for the fairness of trusted release, the group trust score constraint fairness is used to improve the distribution of trust scores, thereby ensuring fairness in trusted release. In addition, since there is no need to construct a risk prevention and control model for each group separately, there is no need to construct multiple different risk prevention and control models, thereby reducing the modeling cost and simplifying the maintenance process. Based on the above processing, this technical solution can enable the risk prevention and control model to treat different user groups equally in the trusted release process, and can simplify the model construction process and improve the model maintenance efficiency.

[0104] Example 5

[0105] Furthermore, based on the above Figures 1 to 3 In one or more embodiments of the present specification, a storage medium is provided for storing computer-executable instruction information. In a specific embodiment, the storage medium may be a USB flash drive, an optical disk, a hard disk, etc. When the computer-executable instruction information stored in the storage medium is executed by a processor, the following process can be implemented:

[0106] Obtain the resource loss objective function and group credibility score constraint function of the risk prevention and control model applied to the trustworthy scenario;

[0107] Constructing a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight;

[0108] Determining optimization target information corresponding to the resource loss objective function and the group credibility score constraint function based on the constructed optimization function, the resource loss objective function, and the group credibility score constraint function;

[0109] Based on the determined optimization target information, the risk prevention and control model applied in the trusted scenario is trained through training samples to obtain a trained risk prevention and control model.

[0110] In the embodiments of this specification, the risk prevention and control model is constructed by a neural network model.

[0111] In the embodiment of this specification, the risk prevention and control model is used to obtain data without preset risks from the business data of preset businesses, and the preset risks include one or more of fraud risks, theft risks, and illegal financial activity risks.

[0112] In the embodiment of this specification, the corresponding optimization function is constructed based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, including:

[0113] Based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, a corresponding Lagrangian function is constructed, and the constructed Lagrangian function is used as the optimization function.

[0114] In the embodiment of this specification, it also includes:

[0115] The historical service data of users belonging to different subgroups in the preset service are obtained, and the obtained historical service data are used as the training samples.

[0116] In an embodiment of this specification, the optimization target information includes first sub-optimization target information and second sub-optimization target information. The first sub-optimization target information is the optimization target information for the training sample, and the second sub-optimization target information is the optimization target information for the second weight.

[0117] In the embodiments of this specification, the risk prevention and control model applied to the trusted scenario is trained based on the determined optimization target information and through training samples to obtain the trained risk prevention and control model, including:

[0118] Based on the determined optimization target information, the training sample is input into the risk prevention and control model, a feasible solution corresponding to the optimization target information is determined by the risk prevention and control model, and a target feasible solution that satisfies the resource loss selection rule is selected from the feasible solutions corresponding to the optimization target information by the risk prevention and control model according to a preset resource loss selection rule, and the target feasible solution is used as an output result of the risk prevention and control model;

[0119] If the output result does not meet the preset convergence conditions, continue to train the risk prevention and control model applied to the trusted scenario based on the determined optimization target information and through training samples until the output result meets the preset convergence conditions, thereby obtaining the trained risk prevention and control model.

[0120] The embodiment of this specification provides a storage medium, which obtains the resource loss objective function and the group trust score constraint function of the risk prevention and control model applied in the trusted scenario, and constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, as well as the group trust score constraint function and the corresponding second weight. Then, based on the constructed optimization function, the resource loss objective function and the group trust score constraint function, the optimization target information corresponding to the resource loss objective function and the group trust score constraint function is determined. Finally, based on the determined optimization target information and through the training samples, the risk prevention and control model applied in the trusted scenario can be trained to obtain the trained risk prevention and control model. In this way, it does not rely on the data distribution and does not need to construct a risk control model for each group separately. The risk prevention and control model is not based on the resource loss objective function and the group trust score constraint function, and combined with the corresponding solution method, which improves the fairness of the risk prevention and control model to sub-groups in the trusted scenario and can be extended to other fairness issues. In addition, in terms of the fairness of trusted release, the group trust score is used to constrain fairness and improve the distribution of trust scores, thereby ensuring fairness in trusted release. In addition, since there is no need to construct a risk prevention and control model for each group separately, there is no need to construct multiple different risk prevention and control models, thereby reducing the modeling cost and simplifying the maintenance process. Based on the above processing, this technical solution can enable the risk prevention and control model to treat different user groups equally in the trusted release process, simplify the model construction process, and improve the model maintenance efficiency.

[0121] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0122] In the 1990s, technological improvements could be clearly distinguished as either hardware improvements (for example, improvements to circuit structures such as diodes, transistors, and switches) or software improvements (improvements to process flows). However, with the advancement of technology, many process flow improvements today can now be considered direct improvements to hardware circuit structures. Designers almost always create the corresponding hardware circuit structure by programming the improved process flow into the hardware circuit. Therefore, it cannot be said that a process flow improvement cannot be implemented using hardware modules. For example, a programmable logic device (PLD), such as a field programmable gate array (FPGA), is an integrated circuit whose logical function is determined by user programming. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to hire a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly done using "logic compiler" software. This is similar to the software compiler used when developing programs. Before compilation, the original code must also be written in a specific programming language, called a hardware description language (HDL). There is not just one HDL, but many, 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. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art will also understand that by simply programming the method flow in one of these hardware description languages ​​and then programming it into an integrated circuit, a hardware circuit that implements the logic method flow can be easily obtained.

[0123] The controller can be implemented in any suitable manner. For example, the controller can take the form of 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, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to implementing the controller in a purely computer-readable program code format, the controller can be implemented in the form of logic gates, switches, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the means for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the means for implementing various functions can be considered as both a software module implementing the method and a structure within the hardware component.

[0124] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, 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.

[0125] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0126] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0127] The embodiments of this specification are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable fraud case serial and parallel device to produce a machine, so that the instructions executed by the processor of the computer or other programmable fraud case serial and parallel device generate instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0128] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable fraud case serial and parallel device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, the instruction device being implemented in the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0129] These computer program instructions may also be loaded onto a computer or other programmable device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0130] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0131] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0132] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media 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 disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0133] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0134] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Thus, one or more embodiments of this specification may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0135] One or more embodiments of this specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. One or more embodiments of this specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0136] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0137] The foregoing is merely an example of the present invention and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this specification are intended to be included within the scope of the claims of this specification.

Claims

1. A method for processing a model, the method comprising: Obtaining a resource loss objective function and a group credibility score constraint function of a risk prevention and control model applied to a trusted scenario, wherein the group credibility score constraint function is a function corresponding to a group credibility score as a constraint, and the group credibility score constraint function is determined by constraints related to variables in the resource loss objective function; Constructing a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight; Determining optimization target information corresponding to the resource loss objective function and the group credibility score constraint function based on the constructed optimization function, the resource loss objective function, and the group credibility score constraint function; Based on the determined optimization target information, the risk prevention and control model applied in the trusted scenario is trained through training samples to obtain a trained risk prevention and control model.

2. According to the method of claim 1, the risk prevention and control model is constructed by a neural network model.

3. According to the method according to claim 2, the risk prevention and control model is used to obtain data without preset risks from the business data of the preset business, and the preset risks include one or more of fraud risk, theft risk, and illegal financial activity risk.

4. The method according to claim 1, wherein constructing a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, comprises: Based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight, a corresponding Lagrangian function is constructed, and the constructed Lagrangian function is used as the optimization function.

5. The method according to claim 3, further comprising: The historical service data of users belonging to different subgroups in the preset service are obtained, and the obtained historical service data are used as the training samples.

6. According to the method of claim 5, the optimization target information includes first sub-optimization target information and second sub-optimization target information, the first sub-optimization target information is the optimization target information for the training sample, and the second sub-optimization target information is the optimization target information for the second weight.

7. The method according to claim 5, wherein the step of training the risk prevention and control model applied to the trusted scenario based on the determined optimization target information and using training samples to obtain the trained risk prevention and control model comprises: Based on the determined optimization target information, the training sample is input into the risk prevention and control model, a feasible solution corresponding to the optimization target information is determined by the risk prevention and control model, and a target feasible solution that satisfies the resource loss selection rule is selected from the feasible solutions corresponding to the optimization target information by the risk prevention and control model according to a preset resource loss selection rule, and the target feasible solution is used as an output result of the risk prevention and control model; If the output result does not meet the preset convergence conditions, continue to train the risk prevention and control model applied to the trusted scenario based on the determined optimization target information and through training samples until the output result meets the preset convergence conditions, thereby obtaining the trained risk prevention and control model.

8. A model processing device, comprising: A function acquisition module, which acquires a resource loss objective function and a group credibility score constraint function of a risk prevention and control model applied to a trusted scenario. The group credibility score constraint function is a function corresponding to a group credibility score as a constraint, and the group credibility score constraint function is determined by constraints related to variables in the resource loss objective function; An optimization function determination module constructs a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight; An optimization target determination module, which determines optimization target information corresponding to the resource loss objective function and the group credibility score constraint function based on the constructed optimization function, the resource loss objective function and the group credibility score constraint function; The training module trains the risk prevention and control model applied in the trusted scenario based on the determined optimization target information and through training samples to obtain the trained risk prevention and control model.

9. A model processing device, comprising: processor; as well as a memory arranged to store computer-executable instructions which, when executed, cause the processor to: Obtaining a resource loss objective function and a group credibility score constraint function of a risk prevention and control model applied to a trusted scenario, wherein the group credibility score constraint function is a function corresponding to a group credibility score as a constraint, and the group credibility score constraint function is determined by constraints related to variables in the resource loss objective function; Constructing a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight; Determining optimization target information corresponding to the resource loss objective function and the group credibility score constraint function based on the constructed optimization function, the resource loss objective function, and the group credibility score constraint function; Based on the determined optimization target information, the risk prevention and control model applied in the trusted scenario is trained through training samples to obtain a trained risk prevention and control model.

10. A storage medium for storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the following process: Obtaining a resource loss objective function and a group credibility score constraint function of a risk prevention and control model applied to a trusted scenario, wherein the group credibility score constraint function is a function corresponding to a group credibility score as a constraint, and the group credibility score constraint function is determined by constraints related to variables in the resource loss objective function; Constructing a corresponding optimization function based on the resource loss objective function and the corresponding first weight, and the group credibility score constraint function and the corresponding second weight; Determining optimization target information corresponding to the resource loss objective function and the group credibility score constraint function based on the constructed optimization function, the resource loss objective function, and the group credibility score constraint function; Based on the determined optimization target information, the risk prevention and control model applied in the trusted scenario is trained through training samples to obtain a trained risk prevention and control model.

Images