Business risk control method and device, electronic equipment and storage medium
By mapping the prediction results of the risk prediction model, the need to adjust the cutoff threshold after the model iterative update is solved, the risk control efficiency and stability are improved, and the model upgrade without perception on the business side is achieved.
Patent Information
- Application Number
- CN202310539952.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-12
- Publication Date
- 2024-11-12
AI Technical Summary
In the risk control business, the cutoff threshold needs to be adjusted after the model iteration and update, resulting in high labor and time costs, affecting the efficiency and stability of risk control.
By inputting the risk control feature data of the target object into the current model for risk prediction processing, and calling the prediction result mapping function to map the predicted risk probability, generating risk control results, and implementing the model based on the logarithmic probability alignment of the historical model with the current model The sorting of prediction results before and after the upgrade is consistent, avoiding adjusting the cutoff threshold.
It improves the efficiency of online business risk control, ensures the stability of risk control after model upgrade, and reduces the operation costs and error risks on the business side.
Smart Images

Figure CN118941374A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a business risk control method, device, electronic device and storage medium. Background Art
[0002] In some businesses involving risk control (abbreviated as risk control), risk prediction models are usually used to perform risk control on online businesses to ensure the security of online businesses. For example, in the risk control of credit business, the corresponding risk prediction model can be used to predict the probability that a user is a risky user. If the prediction result is higher than the cutoff (i.e., threshold) set in the credit business, the user's loan application will be rejected.
[0003] Taking into account that the risk control-related information of the risk control object (such as individuals or enterprises) will change with time and environment, such as changes in the credit status of users in credit business, which will lead to inaccurate prediction results of the risk prediction model, it is necessary to continuously iterate and update (or upgrade) the risk prediction model deployed online to ensure the prediction accuracy of the online model.
[0004] In the related technology, while iteratively updating or upgrading the risk prediction model deployed on the business line, the related business side is also required to readjust the business cutoff threshold, and then conduct observation and evaluation for 1-3 performance periods based on the adjusted business cutoff threshold. Only then can the upgraded model be officially used for online business risk control, which consumes a lot of manpower and time costs, not only affecting the efficiency of online business risk control, but also affecting the stability of online business risk control based on the upgraded model. Summary of the invention
[0005] In order to solve the problems of the prior art, the embodiments of the present application provide a business risk control method, device, electronic device and storage medium. The technical solution is as follows:
[0006] In one aspect, a business risk control method is provided, the method comprising:
[0007] In response to a business processing request of a target object for a risk control business, acquiring feature data of the target object;
[0008] The risk control feature data of the target object is input into the current risk prediction model corresponding to the current model application cycle for risk prediction processing to obtain a predicted risk probability; the predicted risk probability represents the possibility that the target object belongs to a preset risk category, the current risk prediction model is trained based on the risk performance sample data in the historical model application cycle, and the historical model application cycle performs risk prediction processing based on the historical risk prediction model;
[0009] Calling a prediction result mapping function to map the predicted risk probability to obtain a mapped risk probability; the prediction result mapping function characterizes a mapping relationship between the prediction result of the historical risk prediction model and the prediction result of the current risk prediction model, and the prediction result mapping function is obtained based on aligning the logarithmic probability of the prediction result of the historical risk prediction model with the logarithmic probability of the prediction result of the current risk prediction model;
[0010] Based on the comparison between the mapped risk probability and the preset risk control threshold corresponding to the risk control business, a risk control result is generated; the risk control result represents whether the business service requested by the business processing request is provided to the target object.
[0011] In another aspect, a business risk control device is provided, the device comprising:
[0012] A feature data acquisition module, configured to acquire risk control feature data of a target object in response to a business processing request of the target object for a risk control business;
[0013] A risk prediction processing model is used to input the characteristic data of the target object into the current risk prediction model corresponding to the current model application cycle for risk prediction processing to obtain a predicted risk probability; the predicted risk probability represents the possibility that the target object belongs to a preset risk category, the current risk prediction model is trained based on risk performance sample data within a historical model application cycle, and the historical model application cycle performs risk prediction processing based on a historical risk prediction model;
[0014] A prediction result mapping module, used for calling a prediction result mapping function to perform mapping processing on the predicted risk probability to obtain a mapped risk probability; the prediction result mapping function represents the mapping relationship between the prediction result of the historical risk prediction model and the prediction result of the current risk prediction model, and the prediction result mapping function is obtained based on aligning the logarithmic probability of the prediction result of the historical risk prediction model with the logarithmic probability of the prediction result of the current risk prediction model;
[0015] A risk control result generation module is used to generate a risk control result based on the comparison between the mapped risk probability and the preset risk control threshold corresponding to the risk control business; the risk control result represents whether the business service requested by the business processing request is provided to the target object.
[0016] In an exemplary embodiment, the apparatus further includes a mapping function generating module, wherein the mapping function generating module includes:
[0017] A first prediction module, used for inputting the risk performance sample data into the historical risk prediction model for risk prediction processing to obtain a first sample prediction result of the risk performance sample data;
[0018] A first function construction module is used to construct a first logarithmic probability regression function based on the first sample prediction result and label information of the risk performance sample data corresponding to the first sample prediction result, using the first sample prediction result as input data; the label information indicates whether it is the preset risk category;
[0019] A second prediction module is used to input the risk performance sample data into the current risk prediction model to perform risk prediction processing to obtain a second sample prediction result of the risk performance sample data;
[0020] A second function construction module is used to construct a second logarithmic probability regression function based on the second sample prediction result and label information of the risk performance sample data corresponding to the second sample prediction result, using the second sample prediction result as input data;
[0021] A logarithmic probability alignment module is used to perform logarithmic probability alignment on the first logarithmic probability regression function and the second logarithmic probability regression function to obtain the prediction result mapping function.
[0022] In an exemplary embodiment, the first function building module is specifically used to:
[0023] Constructing a first initial logistic regression model corresponding to the first logarithmic probability regression function; inputting each of the first sample prediction results into the first initial logistic regression model respectively to obtain a first logistic regression result corresponding to each of the first sample prediction results;
[0024] Based on the difference between the first logistic regression result corresponding to each of the first sample prediction results and the label information corresponding to the corresponding first sample prediction result, adjusting the parameters of the first initial logistic regression model until a preset end condition is reached, thereby obtaining the first logarithmic probability regression function;
[0025] The second function building module is specifically used for:
[0026] Constructing a second initial logistic regression model corresponding to the second logarithmic probability regression function; inputting each of the second sample prediction results into the second initial logistic regression model respectively to obtain a second logistic regression result corresponding to each of the second sample prediction results;
[0027] Based on the difference between the second logistic regression results corresponding to each of the second sample prediction results and the label information corresponding to the corresponding second sample prediction results, the parameters of the second initial logistic regression model are adjusted until the preset end condition is reached to obtain the second logarithmic probability regression function.
[0028] In an exemplary embodiment, the mapping function generation module further includes:
[0029] A sample data screening module, used to screen target risk performance sample data from the risk performance sample data based on the first sample prediction result of the risk performance sample data; the first sample prediction result of each target risk performance sample data covers a preset numerical range, and the first sample prediction result of the target risk performance sample data is used to construct the first logarithmic probability regression function;
[0030] The second prediction module is specifically used to: input the target risk performance sample data into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of the target risk performance sample data; the second sample prediction result of the target risk performance sample data is used to construct the second logarithmic probability regression function.
[0031] In an exemplary embodiment, the apparatus further comprises:
[0032] A risk performance sample acquisition module is used to acquire risk performance sample data within the historical model application cycle; each of the risk performance sample data includes sample risk control feature data and label information indicating whether it is the preset risk category;
[0033] A third prediction module is used to input the sample risk control feature data in each of the risk performance sample data into the risk prediction model to be trained for risk prediction processing, so as to obtain the training prediction results of each of the risk performance sample data;
[0034] A loss value determination module, used to determine the loss value based on the difference between the training prediction result of each risk performance sample data and the label information of the corresponding risk performance sample data;
[0035] A training submodule, used for adjusting the model parameters of the risk prediction model to be trained based on the loss value until a preset training end condition is met, thereby obtaining the current risk prediction model;
[0036] The model updating module updates the model used for risk prediction processing in the risk control business to the current risk prediction model and enters the current model application cycle.
[0037] In an exemplary embodiment, the model updating module is further used to: when the current model application cycle reaches a preset duration, use the current model application cycle as the historical model application cycle and the current risk prediction model as the historical risk prediction model.
[0038] In an exemplary embodiment, the risk control result generation module is specifically used to:
[0039] When the mapped risk probability exceeds a preset risk control threshold corresponding to the risk control business, a first risk control result is generated; the first risk control result indicates that the business service requested by the business processing request is refused to be provided to the target object;
[0040] When the mapped risk probability does not exceed the preset risk control threshold corresponding to the risk control business, a second risk control result is generated; the second risk control result indicates that the business service requested by the business processing request is provided to the target object.
[0041] On the other hand, an electronic device is provided, including a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement any of the business risk control methods described above.
[0042] On the other hand, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or the at least one program is loaded and executed by a processor to implement a business risk control method as described in any of the above aspects.
[0043] On the other hand, a computer program product or a computer program is provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device performs any of the above-mentioned business risk control methods.
[0044] The embodiment of the present application obtains predicted risk probability by inputting the risk control feature data of the target object into the current risk prediction model (i.e., the upgraded model) corresponding to the current model application cycle for risk prediction processing, and then calls the prediction result mapping function that characterizes the mapping relationship between the prediction results of the historical risk prediction model (i.e., the model before the upgrade) and the prediction results of the current risk prediction model to perform mapping processing on the above-mentioned predicted risk probability, so as to obtain the mapped risk probability falling within the range of the prediction results of the model before the upgrade, and generates the risk control result based on the comparison between the mapped risk probability and the preset risk control threshold of the risk control business, wherein the prediction result mapping function is obtained based on the logarithmic probability of the prediction results of the aligned historical risk prediction model and the logarithmic probability of the prediction results of the current risk prediction model, thereby achieving the ranking of the prediction results without changing before and after the model upgrade. Therefore, the above-mentioned technical solution not only does not require the business layer to change the cutoff threshold of its own risk control strategy, thereby improving the efficiency of online business risk control, but also ensures the stability of online business risk control after the certification model is upgraded. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0046] Figure 1 It is a schematic diagram of an implementation environment provided by an embodiment of the present application;
[0047] Figure 2 It is a flowchart of a business risk control method provided by an embodiment of the present application;
[0048] Figure 3 It is a flowchart of another business risk control method provided by an embodiment of the present application;
[0049] Figure 4 It is a flowchart of another business risk control method provided by an embodiment of the present application;
[0050] Figure 5 It is a flowchart of another business risk control method provided by an embodiment of the present application;
[0051] Figure 6 It is a structural block diagram of a business risk control device provided in an embodiment of the present application;
[0052] Figure 7 It is a hardware structure block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0053] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0054] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0055] It is understandable that in the specific implementation of this application, related data such as user information is involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.
[0056] See also Figure 1 , which is a schematic diagram of an implementation environment provided by an embodiment of the present application, the implementation environment includes a terminal 110 and a server 120, wherein the terminal 110 and the server 120 can communicate via a wired or wireless network connection.
[0057] The terminal 110 includes but is not limited to mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, aircraft, etc. The terminal 110 is installed with client software with business functions such as an application (Application, referred to as App), which can be an independent application or a subroutine in the application. In the embodiment of the present application, the above-mentioned business functions include functions related to the business that requires risk control (hereinafter referred to as risk control business), such as credit business functions. The user of the terminal 110 can initiate a business processing request for the risk control business through the business function of the application, for example, a loan request can be initiated.
[0058] The server 120 can provide background services for the application in the terminal 110. A risk prediction model for risk control business can be deployed on the server 120. When the server 120 receives a business processing request for a risk business, it can call the risk prediction model for the risk control business deployed thereon to perform risk prediction processing to obtain a predicted risk probability, and then generate a risk control result in combination with the cutoff threshold on the business side.
[0059] In an embodiment of the present application, in order to ensure the accuracy of the prediction results of the risk prediction model and improve the stability of risk control, the risk prediction model deployed on the server 120 needs to be continuously upgraded and updated, that is, a new risk prediction model needs to be trained every once in a while, and then the new risk prediction model is deployed to the business line for risk prediction processing to replace the old risk prediction model on the business line to achieve business risk control.
[0060] It should be noted that the server involved in the embodiments of the present application can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network) and big data and artificial intelligence platforms.
[0061] In an exemplary embodiment, the terminal 110 and the server can both be node devices in the blockchain system, and can share the acquired and generated information with other node devices in the blockchain system to achieve information sharing between multiple node devices. Multiple node devices in the blockchain system can be configured with the same blockchain, which is composed of multiple blocks, and the adjacent blocks have an association relationship, so that when the data in any block is tampered, it can be detected by the next block, thereby preventing the data in the blockchain from being tampered with, and ensuring the security and reliability of the data in the blockchain.
[0062] The embodiments of the present application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, assisted driving, etc.
[0063] Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Artificial intelligence technology is an interdisciplinary subject that covers a wide range of fields, including both hardware-level technology and software-level technology. The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operating / interactive systems, and mechatronics. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, as well as machine learning / deep learning, autonomous driving, and smart transportation.
[0064] Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.
[0065] See also Figure 2 , which is a flow chart of a business risk control method provided by an embodiment of the present application, which can be applied to Figure 1 Server in. It should be noted that this specification provides method operation steps as described in the embodiments or flow charts, but more or fewer operation steps may be included based on routine or non-creative work. The order of steps listed in the embodiments is only one way of executing the steps among many orders, and does not represent the only order of execution. When the actual system or product is executed, it can be executed in sequence or in parallel (for example, in a parallel processor or multi-threaded processing environment) according to the method shown in the embodiments or drawings. Specific examples Figure 2 As shown, the method may include:
[0066] S201, in response to a business processing request of a target object for a risk control business, obtaining risk control feature data of the target object.
[0067] Among them, risk control business refers to business involving risk control, such as credit business in financial scenarios.
[0068] The target object refers to the account that initiates the business processing request for the risk control business. The account can be a personal account or a corporate account.
[0069] The risk control feature data of the target object may include feature data of the target object related to the above-mentioned risk control business, specifically including attribute data and business behavior data. For example, the attribute data may include age, gender, salary, etc., and the business behavior data may include business behavior data associated with the risk control business. In a specific implementation, the business behavior data may be the business behavior data of the target object in a recent period of time, such as the last week or the last month, etc.
[0070] S203, inputting the risk control feature data of the target object into the current risk prediction model corresponding to the current model application cycle to perform risk prediction processing to obtain a predicted risk probability.
[0071] The predicted risk probability represents the possibility that the target object belongs to a preset risk category. For example, the preset risk category may refer to a bad customer with risks, etc. The preset risk category may be set based on actual needs.
[0072] Among them, the current risk prediction model refers to the risk prediction model currently deployed on the business line for risk prediction processing. The model application cycle refers to the period of time that the risk prediction model will be continuously applied after it is deployed online. During this period, risk prediction processing will be performed based on the risk prediction model. Therefore, the current model application cycle refers to the period of time that the current risk prediction model is continuously applied.
[0073] In an embodiment of the present application, the current risk prediction model is trained based on risk performance sample data within a historical model application cycle, and the historical model application cycle performs risk prediction processing based on the historical risk prediction model.
[0074] Among them, the historical model application cycle refers to a model application cycle before the current model application cycle, and the risk prediction model used for risk prediction processing in the historical model application cycle is called the historical risk prediction model.
[0075] For example, the risk prediction model A2 is currently deployed on the business line for risk prediction processing. Before the risk prediction model A2 was put online, the risk prediction model A1 was deployed on the business line for risk prediction processing, and the model application cycle of the risk prediction model A1 is 1 month (assuming it is March). Starting from April, the risk prediction model A2 is put online to replace the risk prediction model A1 for risk prediction processing. Then, the risk prediction model A2 is the current risk prediction model, April is the current model application cycle, and the risk prediction model A1 is the historical risk prediction model, and March is the historical model application cycle.
[0076] It should be noted that in the subsequent introduction of the embodiments of the present application, for the sake of convenience and brevity, the current risk prediction model will be referred to as the new model, and the above-mentioned historical risk prediction model will be referred to as the old model.
[0077] Among them, the risk performance sample data includes sample risk control feature data and label information indicating whether it is a preset risk category, for example, the label information is 1 indicating that it is a preset risk category such as a bad customer, and 0 indicates that it is not a preset risk category such as a good customer. The label information can be manually labeled based on the actual situation. The sample risk control feature data may include attribute data and business behavior data of the sample target object. For example, the attribute data may include age, gender, salary, etc. The business behavior data may include business behavior data associated with the risk control business. In a specific implementation, the business behavior data may be the business behavior data of the target object in the most recent period of time, such as the most recent week or the most recent month, etc.
[0078] Based on this, in some exemplary embodiments, Figure 3 As shown, before step S203, the method may further include:
[0079] S301, obtaining risk performance sample data within the historical model application cycle.
[0080] Among them, each risk performance sample data includes sample risk control feature data and label information indicating whether it is a preset risk category.
[0081] Taking the credit business as an example, the risk control feature data corresponding to users who actually have credit risks in the previous model application cycle can be obtained as positive risk performance sample data (label information is 1), and the risk control feature data corresponding to users who actually have no credit risks in the previous model application cycle can be obtained as negative risk performance sample data (label information is 0), so that multiple risk performance sample data can be obtained.
[0082] S303, inputting the sample risk control feature data in each risk performance sample data into the risk prediction model to be trained for risk prediction processing, and obtaining the training prediction results of each risk performance sample data.
[0083] The risk prediction model to be trained may be an initialized classification model, such as a binary classification model, which can predict the probability that the input data belongs to a preset risk category, such as the probability of the existence of credit risk.
[0084] In the specific implementation, the sample risk control feature data in each risk performance sample data are respectively input into the risk prediction model to be trained for risk prediction processing, and the corresponding training prediction results output by the risk prediction model to be trained are obtained. The training prediction results represent the probability that the corresponding sample risk control feature data belongs to the preset risk category.
[0085] S305: Determine a loss value based on the difference between the training prediction result of each risk performance sample data and the label information of the corresponding risk performance sample data.
[0086] Among them, the loss value can be calculated using a preset loss function, combining the difference between the training prediction results of each risk performance sample data and the corresponding label information. Exemplarily, the preset loss function can be a cross entropy loss function.
[0087] S307, adjusting the model parameters of the risk prediction model to be trained based on the loss value until a preset training end condition is met to obtain a current risk prediction model.
[0088] Specifically, when adjusting the model parameters of the risk prediction model to be trained based on the loss value, the gradient descent method can be used to optimize the model parameters, and the iterative training can be continued based on the optimized model parameters until the preset training end conditions are met and the training is terminated. The model corresponding to the model parameters at the end of the training is used as the current risk prediction model.
[0089] Among them, the preset training end condition can be that the number of iterations reaches the preset number of iterations, the loss value reaches the preset loss threshold, or the difference between the loss values corresponding to two adjacent iterations reaches the preset difference threshold. The specific preset number of iterations, preset loss threshold and preset difference threshold can all be set according to actual needs.
[0090] S309, updating the model used for risk prediction processing in the risk control business to the current risk prediction model, and entering the current model application cycle.
[0091] In an embodiment of the present application, after the current risk prediction model is trained, the current risk prediction model can be deployed online as a new risk prediction model, so that the current risk prediction model can be used for risk prediction processing in subsequent risk control business, thereby realizing the upgrade of the online risk prediction model.
[0092] In the above implementation, by training the current risk prediction model based on risk performance sample data of the historical model application cycle, the prediction accuracy of the current risk prediction model can be improved, which is beneficial to improving the accuracy of online risk control.
[0093] In some specific implementations, the method of the embodiment of the present application may also include: when the current model application cycle reaches a preset duration, using the current model application cycle as the historical model application cycle, and the current risk prediction model as the historical risk prediction model.
[0094] Among them, the preset duration can be set based on actual business needs, for example, it can be set to 1 month, 2 months, etc. Generally, the shorter the preset duration is, the more frequent the online risk prediction model is upgraded, the better the accuracy of risk control, but the cost consumption will also increase.
[0095] S205, calling the prediction result mapping function to perform mapping processing on the predicted risk probability to obtain a mapped risk probability.
[0096] Among them, the prediction result mapping function represents the mapping relationship between the prediction result of the historical risk prediction model and the prediction result of the current risk prediction model.
[0097] Among them, the prediction result mapping function is obtained based on aligning the logarithmic probability of the prediction result of the historical risk prediction model with the logarithmic probability of the prediction result of the current risk prediction model.
[0098] Odds refers to the ratio between the probability of an event occurring and the probability of not occurring. Assuming that the probability of an event occurring is p, the probability of the event not occurring is 1-p. The odds of the event is expressed as: Taking the logarithm of the natural base based on the probability, we can get the logarithmic probability of the event. In the embodiment of the present application, the prediction result of the risk prediction model represents the probability of belonging to a preset risk category, so the above-mentioned event refers to belonging to a preset risk category, such as a bad customer.
[0099] In the embodiment of the present application, the logarithmic probability of the prediction result of the risk prediction model can characterize the risk probability of the model prediction result. The risk probability of the model prediction result can also be understood as the risk scale of the model prediction result. In the credit business, the risk scale of the risk prediction model prediction result can also be called the default probability, which represents the possibility that the borrower will not be able to repay the principal and interest of the loan from the financial institution or perform related obligations as required by the contract within a certain period of time in the future.
[0100] Among them, aligning the logarithmic probability means making the logarithmic probability of the prediction results of the new and old models equal, and aligning the logarithmic probability of the prediction results of the new and old models so that the risk scales of the prediction results of the two models are aligned, and then the ranking of the mapped risk probability obtained after the mapping process based on the prediction result mapping function is consistent with the ranking of the predicted risk probability before the mapping process, that is, the KS value of the model before and after the mapping will not change. Among them, KS (Kolmogorov-Smirnov) refers to the difference (maximum value) of the cumulative distribution between good and bad samples, which is used to evaluate the risk differentiation ability of the model. Generally, the greater the cumulative difference between good and bad samples, the stronger the risk differentiation ability of the model, and the larger the KS index.
[0101] Exemplarily, the prediction result mapping function can be expressed as Score=A+B*_, where Score_new is the prediction result of the current risk prediction model, i.e., the predicted risk probability in step S203, Score is the mapped risk probability, and A and B are parameters determined by aligning the logarithmic probability of the prediction results of the new and old models.
[0102] Although the mapping relationship between the prediction results of the new and old models can be obtained by aligning the distribution of prediction results, the same mapping point of the prediction results of the new and old models will correspond to different default probabilities. Since the performance of the new model is improved compared to the old model, the mapping relationship between the prediction results of the new and old models obtained by directly aligning the distribution of prediction results will lead to too strict risk control, which is not conducive to the business development of the business side, and thus it is impossible to maximize the performance improvement of the new model over the old model, but will have an adverse effect on the business side. The embodiment of the present application realizes alignment from the perspective of the logarithmic probability (or risk scale) of the prediction results to obtain the mapping relationship between the prediction results of the new and old models, so that the sorting of the prediction results is not changed before and after the model upgrade, and the performance improvement of the new model over the old model can be maximized.
[0103] Based on this, in some possible implementations, after training to obtain the current risk prediction model, i.e., the new model, the embodiment of the present application may further include generating the above-mentioned prediction result mapping function, such as Figure 4 As shown, generating a prediction result mapping function may include:
[0104] S401, inputting the risk performance sample data into the historical risk prediction model for risk prediction processing to obtain a first sample prediction result of the risk performance sample data.
[0105] Specifically, each risk performance sample data within the historical model application period is respectively input into the historical risk prediction model for risk prediction processing, and a first sample prediction result of each risk performance sample data can be obtained. The first sample prediction result represents the probability that the corresponding risk performance sample data belongs to a preset risk category.
[0106] S403, using the first sample prediction result as input data, and constructing a first logarithmic probability regression function based on the first sample prediction result and label information of the risk performance sample data corresponding to the first sample prediction result.
[0107] The first logarithmic probability regression function is a function representation of the logarithmic probability of the prediction result of the old model. The label information of the risk performance sample data indicates whether it is a preset risk category, for example, the label information is 1, indicating that it is a preset risk category (such as a bad customer), and the label information is 0, indicating that it is not a preset risk category (such as a good customer).
[0108] In a specific implementation, when implementing the above step S403, a first initial logistic regression model corresponding to the first logarithmic probability regression function can be first constructed, and then each first sample prediction result can be respectively input into the first initial logistic regression model to obtain a first logistic regression result corresponding to each first sample prediction result, and then based on the difference between the first logistic regression result corresponding to each first sample prediction result and the label information corresponding to the corresponding first sample prediction result, the parameters of the first initial logistic regression model are adjusted until the preset end condition is reached to obtain the first logarithmic probability regression function.
[0109] Specifically, the first initial logistic regression model is expressed as: Among them, y1 is the first logistic regression result, which represents the probability that the input data belongs to the preset risk category; x1 is the prediction result of the first input sample, which is the prediction result of the old model; ω1 and b1 are the parameters of the first initial logistic regression model.
[0110] When adjusting the parameters of the first initial logistic regression model based on the difference between the first logistic regression results corresponding to each first sample prediction result and the label information corresponding to the corresponding first sample prediction result, the preset loss function can be first used to obtain the loss value based on the difference between the first logistic regression results corresponding to each first sample prediction result and the label information corresponding to the corresponding first sample prediction result, and then the parameters of the first initial logistic regression model are reversely adjusted based on the loss value, and the iteration is continued based on the adjusted parameters until the preset end condition is reached, so as to determine the parameter values of parameters ω1 and b1 in the first initial logistic regression model, hereinafter referred to as the first logistic regression model.
[0111] The preset loss function may be, but is not limited to, a cross entropy loss function. The preset end condition may be that the number of iterations reaches a preset number of iterations, or that the loss value reaches a preset loss threshold, or that the difference between the loss values corresponding to two adjacent iterations reaches a preset difference threshold. The specific preset number of iterations, preset loss threshold, and preset difference threshold may be set according to actual needs.
[0112] After determining the parameter values of the parameters ω1 and b1 in the first initial logistic regression model, that is, obtaining the first logistic regression model, the representation function of the first logistic regression model can be transformed to obtain the following function: This function is the first logarithmic probability regression function, where The log odds of predicting the outcome for the old model.
[0113] S405: Input the risk performance sample data into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of the risk performance sample data.
[0114] Specifically, each risk performance sample data within the historical model application cycle is respectively input into the current risk prediction model for risk prediction processing, and a second sample prediction result of each risk performance sample data can be obtained. The second sample prediction result represents the probability that the corresponding risk performance sample data belongs to a preset risk category.
[0115] S407, using the second sample prediction result as input data, and constructing a second logarithmic probability regression function based on the second sample prediction result and the label information of the risk performance sample data corresponding to the second sample prediction result.
[0116] The second log-odds regression function is a function representation of the log-odds of the prediction results of the new model.
[0117] In a specific implementation, when implementing the above step S407, a second initial logistic regression model corresponding to the second logarithmic probability regression function can be first constructed, and then each second sample prediction result can be respectively input into the second initial logistic regression model to obtain a second logistic regression result corresponding to each second sample prediction result, and then based on the difference between the second logistic regression result corresponding to each second sample prediction result and the label information corresponding to the corresponding second sample prediction result, the parameters of the second initial logistic regression model are adjusted until the preset end condition is reached to obtain the second logarithmic probability regression function.
[0118] Specifically, the second initial logistic regression model is expressed as: Among them, y2 is the second logistic regression result, which represents the probability that the input data belongs to the preset risk category; x2 is the second sample prediction result of the input, that is, the prediction result of the new model; ω2 and b2 are the parameters of the second initial logistic regression model.
[0119] When adjusting the parameters of the second initial logistic regression model based on the difference between the second logistic regression results corresponding to each second sample prediction result and the label information corresponding to the corresponding second sample prediction result, the preset loss function can be first used to obtain the loss value based on the difference between the second logistic regression results corresponding to each second sample prediction result and the label information corresponding to the corresponding second sample prediction result, and then the parameters of the second initial logistic regression model are reversely adjusted based on the loss value, and the iteration is continued based on the adjusted parameters until the preset end condition is reached, so as to determine the parameter values of parameters ω2 and b2 in the second initial logistic regression model, hereinafter referred to as the second logistic regression model.
[0120] The preset loss function may be, but is not limited to, a cross entropy loss function. The preset end condition may be that the number of iterations reaches a preset number of iterations, or that the loss value reaches a preset loss threshold, or that the difference between the loss values corresponding to two adjacent iterations reaches a preset difference threshold. The specific preset number of iterations, preset loss threshold, and preset difference threshold may be set according to actual needs.
[0121] After the parameter values of the parameters ω2 and b2 in the second initial logistic regression model are determined, that is, the second logistic regression model is obtained, the representation function of the second logistic regression model can be transformed to obtain the following function: This function is the second logarithmic probability regression function, where The log odds of predicting the outcome for the new model.
[0122] S409, performing logarithmic probability alignment on the first logarithmic probability regression function and the second logarithmic probability regression function to obtain a prediction result mapping function.
[0123] Specifically, the first logarithmic probability regression function is aligned with the second logarithmic probability regression function, that is, Then we can get the following linear relationship between x1 and x2: ω1x1+1=ω2x2+2. By transforming this linear relationship, we can get the following linear mapping function:
[0124]
[0125] It can be seen that the above linear mapping function represents the mapping relationship between the new model prediction result, i.e., x2, and the old model prediction result, i.e., x1. Then, by considering x1 in the above linear mapping function as the dependent variable Score and x2 as the independent variable Score_new, the prediction result mapping function Score=A+B*Score_new of the embodiment of the present application can be obtained, where
[0126] In the above implementation, by aligning the logarithmic probability of the prediction results of the new and old models to obtain the prediction result mapping function, the prediction results of the new model can be mapped to the prediction scale of the old model without changing the ranking of the prediction results, and the default probability can be kept consistent at each model quantile.
[0127] In some specific implementations, in order to confirm the distribution boundary of the old model as much as possible to improve the accuracy of the prediction result mapping function, such as Figure 5 As shown, before step S405, the method may further include:
[0128] S501, based on the first sample prediction result of the risk performance sample data, select target risk performance sample data from the risk performance sample data.
[0129] Among them, the first sample prediction result of each target risk performance sample data covers a preset numerical range, which is 0-1, that is, the distribution curve corresponding to the first sample prediction result of each target risk performance sample data covers the range of 0-1.
[0130] Based on this, continue to see Figure 5 The aforementioned step S405 is specifically implemented as S503, which inputs each target risk performance sample data into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of each target risk performance sample data.
[0131] Furthermore, the aforementioned step S403 is specifically S505 when constructing the first logarithmic probability regression function, and the first logarithmic probability regression function is constructed based on the first sample prediction result of the target risk performance sample data; the aforementioned step S407 is specifically S507 when constructing the second logarithmic probability regression function, and the second logarithmic probability regression function is constructed based on the second sample prediction result of the target risk performance sample data.
[0132] S207: Generate a risk control result based on a comparison between the mapped risk probability and a preset risk control threshold corresponding to the risk control business.
[0133] The risk control result indicates whether the business service requested by the business processing request is provided to the target object. Taking the credit business as an example, the risk control result may be providing the target object with a loan service or not providing the target object with a loan service.
[0134] Exemplarily, when the mapped risk probability exceeds a preset risk control threshold corresponding to the risk control business, a first risk control result is generated, and the first risk control result indicates that the business service requested by the business processing request of the target object is refused to be provided.
[0135] When the mapped risk probability does not exceed a preset risk control threshold corresponding to the risk control business, a second risk control result is generated, and the second risk control result indicates that the business service requested by its business processing request is provided to the target object.
[0136] Among them, the preset risk control threshold corresponding to the risk control business is the risk control threshold used in the historical model application cycle, that is, the business-side cutoff threshold (that is, the preset risk control threshold) used before and after the online risk prediction model upgrade does not need to be changed, which can realize the model upgrade without the business side's perception, and can also ensure the stability of the online business after the upgrade.
[0137] The embodiment of the present application aligns the logarithmic probability (also known as the risk scale) of the prediction results of the new and old models to obtain a mapping relationship between the prediction results of the new and old models, namely, a prediction result mapping function. In this way, when an online business processing request for risk control business obtains a prediction result based on the new model, namely, a predicted risk probability, the prediction result mapping function can be called imperceptibly to map it to the prediction scale of the old model, so that the business strategy, namely, the preset risk control threshold, does not need to be changed, and the advantage of the improved risk differentiation of the new model over the old model can be maximized, thereby accelerating the model iteration and updating speed, reducing the operations on the business side, and reducing the risk of operational errors, so that the business side (financial credit institutions) can achieve imperceptible processing during the model iteration and upgrade, thereby improving the efficiency of online business risk control and ensuring the stability of online business risk control after the certification model is upgraded.
[0138] When the technical solution of the embodiment of the present application is applied to business risk control in financial scenarios, the risk scale alignment method can enable financial institutions to improve business approval rates due to improved model performance while keeping risks unchanged, thereby also improving business processing capabilities.
[0139] Corresponding to the business risk control methods provided in the above-mentioned embodiments, an embodiment of the present application also provides a business risk control device. Since the business risk control device provided in the embodiment of the present application corresponds to the business risk control methods provided in the above-mentioned embodiments, the implementation methods of the aforementioned business risk control methods are also applicable to the business risk control device provided in this embodiment and will not be described in detail in this embodiment.
[0140] See also Figure 6 , which is a schematic diagram of the structure of a business risk control device provided in an embodiment of the present application, the device has the function of implementing the business risk control method in the above method embodiment, and the function can be implemented by hardware or by hardware executing corresponding software. Figure 6 As shown, the business risk control device 600 may include:
[0141] The feature data acquisition module 610 is used to acquire the risk control feature data of the target object in response to the business processing request of the target object for the risk control business;
[0142] The risk prediction processing model 620 is used to input the risk control feature data of the target object into the current risk prediction model corresponding to the current model application cycle to perform risk prediction processing to obtain a predicted risk probability; the predicted risk probability represents the possibility that the target object belongs to a preset risk category, the current risk prediction model is trained based on the risk performance sample data in the historical model application cycle, and the historical model application cycle performs risk prediction processing based on the historical risk prediction model;
[0143] A prediction result mapping module 630 is used to call a prediction result mapping function to perform mapping processing on the predicted risk probability to obtain a mapped risk probability; the prediction result mapping function represents the mapping relationship between the prediction result of the historical risk prediction model and the prediction result of the current risk prediction model, and the prediction result mapping function is obtained based on aligning the logarithmic probability of the prediction result of the historical risk prediction model with the logarithmic probability of the prediction result of the current risk prediction model;
[0144] The risk control result generation module 640 is used to generate a risk control result based on the comparison between the mapped risk probability and the preset risk control threshold corresponding to the risk control business; the risk control result represents whether the business service requested by the business processing request is provided to the target object.
[0145] In an exemplary embodiment, the apparatus 600 further includes a mapping function generating module, and the mapping function generating module includes:
[0146] A first prediction module, used for inputting the risk performance sample data into the historical risk prediction model for risk prediction processing to obtain a first sample prediction result of the risk performance sample data;
[0147] A first function construction module is used to construct a first logarithmic probability regression function based on the first sample prediction result and label information of the risk performance sample data corresponding to the first sample prediction result, using the first sample prediction result as input data; the label information indicates whether it is the preset risk category;
[0148] A second prediction module is used to input the risk performance sample data into the current risk prediction model to perform risk prediction processing to obtain a second sample prediction result of the risk performance sample data;
[0149] A second function construction module is used to construct a second logarithmic probability regression function based on the second sample prediction result and label information of the risk performance sample data corresponding to the second sample prediction result, using the second sample prediction result as input data;
[0150] A logarithmic probability alignment module is used to perform logarithmic probability alignment on the first logarithmic probability regression function and the second logarithmic probability regression function to obtain the prediction result mapping function.
[0151] In an exemplary embodiment, the first function building module is specifically used to:
[0152] Constructing a first initial logistic regression model corresponding to the first logarithmic probability regression function; inputting each of the first sample prediction results into the first initial logistic regression model respectively to obtain a first logistic regression result corresponding to each of the first sample prediction results;
[0153] Based on the difference between the first logistic regression result corresponding to each of the first sample prediction results and the label information corresponding to the corresponding first sample prediction result, adjusting the parameters of the first initial logistic regression model until a preset end condition is reached, thereby obtaining the first logarithmic probability regression function;
[0154] The second function building module is specifically used for:
[0155] Constructing a second initial logistic regression model corresponding to the second logarithmic probability regression function; inputting each of the second sample prediction results into the second initial logistic regression model respectively to obtain a second logistic regression result corresponding to each of the second sample prediction results;
[0156] Based on the difference between the second logistic regression results corresponding to each of the second sample prediction results and the label information corresponding to the corresponding second sample prediction results, the parameters of the second initial logistic regression model are adjusted until the preset end condition is reached to obtain the second logarithmic probability regression function.
[0157] In an exemplary embodiment, the mapping function generation module further includes:
[0158] A sample data screening module, used to screen target risk performance sample data from the risk performance sample data based on the first sample prediction result of the risk performance sample data; the first sample prediction result of each target risk performance sample data covers a preset numerical range, and the first sample prediction result of the target risk performance sample data is used to construct the first logarithmic probability regression function;
[0159] The second prediction module is specifically used to: input the target risk performance sample data into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of the target risk performance sample data; the second sample prediction result of the target risk performance sample data is used to construct the second logarithmic probability regression function.
[0160] In an exemplary embodiment, the apparatus 600 further includes:
[0161] A risk performance sample acquisition module is used to acquire risk performance sample data within the historical model application cycle; each of the risk performance sample data includes sample risk control feature data and label information indicating whether it is the preset risk category;
[0162] A third prediction module is used to input the sample risk control feature data in each of the risk performance sample data into the risk prediction model to be trained for risk prediction processing, so as to obtain the training prediction results of each of the risk performance sample data;
[0163] A loss value determination module, used to determine the loss value based on the difference between the training prediction result of each risk performance sample data and the label information of the corresponding risk performance sample data;
[0164] A training submodule, used for adjusting the model parameters of the risk prediction model to be trained based on the loss value until a preset training end condition is met, thereby obtaining the current risk prediction model;
[0165] The model updating module updates the model used for risk prediction processing in the risk control business to the current risk prediction model and enters the current model application cycle.
[0166] In an exemplary embodiment, the model updating module is further used to: when the current model application cycle reaches a preset duration, use the current model application cycle as the historical model application cycle and the current risk prediction model as the historical risk prediction model.
[0167] In an exemplary embodiment, the risk control result generation module 640 is specifically used to:
[0168] When the mapped risk probability exceeds a preset risk control threshold corresponding to the risk control business, a first risk control result is generated; the first risk control result indicates that the business service requested by the business processing request is refused to be provided to the target object;
[0169] When the mapped risk probability does not exceed the preset risk control threshold corresponding to the risk control business, a second risk control result is generated; the second risk control result indicates that the business service requested by the business processing request is provided to the target object.
[0170] It should be noted that the device provided in the above embodiment, when implementing its functions, is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.
[0171] An embodiment of the present application provides an electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement any one of the business risk control methods provided in the above method embodiments.
[0172] The memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory can also include a memory controller to provide the processor with access to the memory.
[0173] The method embodiments provided in the embodiments of the present application may be executed in a computer terminal, a server or a similar computing device, that is, the above-mentioned electronic device may include a computer terminal, a server or a similar computing device. Figure 7 is a hardware structure block diagram of an electronic device for running a business risk control method provided in an embodiment of the present application, such as Figure 7 As shown, the server 700 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 710 (the processor 710 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 730 for storing data, and one or more storage media 720 (such as one or more mass storage devices) for storing application programs 723 or data 722. Among them, the memory 730 and the storage medium 720 can be short-term storage or permanent storage. The program stored in the storage medium 720 may include one or more modules, each of which may include a series of instruction operations on the server. Furthermore, the central processing unit 710 can be configured to communicate with the storage medium 720 and execute a series of instruction operations in the storage medium 720 on the server 700. The server 700 may also include one or more power supplies 760, one or more wired or wireless network interfaces 750, one or more input and output interfaces 740, and / or one or more operating systems 721, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.
[0174] The input / output interface 740 may be used to receive or send data via a network. The specific example of the network may include a wireless network provided by a communication provider of the server 700. In one example, the input / output interface 740 includes a network adapter (Network Interface Controller, NIC), which may be connected to other network devices via a base station so as to communicate with the Internet. In one example, the input / output interface 740 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
[0175] It can be understood by those skilled in the art that Figure 7 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 7 More or fewer components as shown, or with Figure 7 Different configurations shown.
[0176] An embodiment of the present application also provides a computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to implementing a business risk control method. The at least one instruction or the at least one program is loaded and executed by the processor to implement any one of the business risk control methods provided by the above method embodiments.
[0177] The embodiments of the present application also provide a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. The processor of the electronic device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the electronic device executes any one of the business risk control methods provided in the above method embodiments.
[0178] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.
[0179] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of this specification are described. Other embodiments are within the scope of the attached claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0180] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0181] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.
[0182] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.
Claims
1. A business risk control method, characterized in that: The method comprises: In response to a business processing request of a target object for a risk control business, obtaining risk control feature data of the target object; The risk control feature data of the target object is input into the current risk prediction model corresponding to the current model application cycle for risk prediction processing to obtain a predicted risk probability; the predicted risk probability represents the possibility that the target object belongs to a preset risk category, the current risk prediction model is trained based on the risk performance sample data in the historical model application cycle, and the historical model application cycle performs risk prediction processing based on the historical risk prediction model; Calling a prediction result mapping function to map the predicted risk probability to obtain a mapped risk probability; the prediction result mapping function characterizes a mapping relationship between the prediction result of the historical risk prediction model and the prediction result of the current risk prediction model, and the prediction result mapping function is obtained based on aligning the logarithmic probability of the prediction result of the historical risk prediction model with the logarithmic probability of the prediction result of the current risk prediction model; Based on the comparison between the mapped risk probability and the preset risk control threshold corresponding to the risk control business, a risk control result is generated; the risk control result represents whether the business service requested by the business processing request is provided to the target object.
2. The method according to claim 1, characterized in that The method further includes generating the prediction result mapping function, wherein generating the prediction result mapping function includes: Inputting the risk performance sample data into the historical risk prediction model for risk prediction processing to obtain a first sample prediction result of the risk performance sample data; Taking the first sample prediction result as input data, constructing a first logarithmic probability regression function based on the first sample prediction result and label information of the risk performance sample data corresponding to the first sample prediction result; the label information indicates whether it is the preset risk category; Inputting the risk performance sample data into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of the risk performance sample data; Taking the second sample prediction result as input data, and constructing a second logarithmic probability regression function based on the second sample prediction result and label information of the risk performance sample data corresponding to the second sample prediction result; The first logarithmic probability regression function and the second logarithmic probability regression function are logarithmically aligned to obtain the prediction result mapping function.
3. The method according to claim 2, characterized in that The constructing a first logarithmic probability regression function based on the first sample prediction result and the label information of the risk performance sample data corresponding to the first sample prediction result includes: Constructing a first initial logistic regression model corresponding to the first logarithmic probability regression function; inputting each of the first sample prediction results into the first initial logistic regression model respectively to obtain a first logistic regression result corresponding to each of the first sample prediction results; Based on the difference between the first logistic regression result corresponding to each of the first sample prediction results and the label information corresponding to the corresponding first sample prediction result, adjusting the parameters of the first initial logistic regression model until a preset end condition is reached, thereby obtaining the first logarithmic probability regression function; The method of taking the second sample prediction result as input data and constructing a second logarithmic probability regression function based on the second sample prediction result and label information of the risk performance sample data corresponding to the second sample prediction result includes: Constructing a second initial logistic regression model corresponding to the second logarithmic probability regression function; inputting each of the second sample prediction results into the second initial logistic regression model respectively to obtain a second logistic regression result corresponding to each of the second sample prediction results; Based on the difference between the second logistic regression results corresponding to each of the second sample prediction results and the label information corresponding to the corresponding second sample prediction results, the parameters of the second initial logistic regression model are adjusted until the preset end condition is reached to obtain the second logarithmic probability regression function.
4. The method according to claim 2, characterized in that: Before inputting the risk performance sample data into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of the risk performance sample data, the method further includes: Based on the first sample prediction results of the risk performance sample data, target risk performance sample data are screened from the risk performance sample data; the first sample prediction results of each target risk performance sample data cover a preset numerical range, and the first sample prediction results of the target risk performance sample data are used to construct the first logarithmic probability regression function; The step of inputting the risk performance sample data into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of the risk performance sample data includes: The target risk performance sample data is input into the current risk prediction model for risk prediction processing to obtain a second sample prediction result of the target risk performance sample data; the second sample prediction result of the target risk performance sample data is used to construct the second logarithmic probability regression function.
5. The method according to any one of claims 1 to 4, characterized in that: Before inputting the risk control feature data of the target object into the current risk prediction model corresponding to the current model application cycle for risk prediction processing, the method further includes: Acquire risk performance sample data within the application period of the historical model; each of the risk performance sample data includes sample risk control feature data and label information indicating whether it is the preset risk category; Inputting the sample risk control feature data in each of the risk performance sample data into the risk prediction model to be trained for risk prediction processing, and obtaining the training prediction results of each of the risk performance sample data; Determining a loss value based on a difference between a training prediction result of each of the risk performance sample data and label information of the corresponding risk performance sample data; Adjusting the model parameters of the risk prediction model to be trained based on the loss value until a preset training end condition is met, thereby obtaining the current risk prediction model; The model used for risk prediction processing in the risk control business is updated to the current risk prediction model, and enters the current model application cycle.
6. The method according to claim 5, characterized in that The method further comprises: When the current model application cycle reaches a preset duration, the current model application cycle is used as the historical model application cycle, and the current risk prediction model is used as the historical risk prediction model.
7. The method according to claim 1, characterized in that The generating of the risk control result based on the comparison between the mapped risk probability and the preset risk control threshold corresponding to the risk control business includes: When the mapped risk probability exceeds a preset risk control threshold corresponding to the risk control business, a first risk control result is generated; the first risk control result indicates that the business service requested by the business processing request is refused to be provided to the object; When the mapped risk probability does not exceed the preset risk control threshold corresponding to the risk control business, a second risk control result is generated; the second risk control result indicates that the business service requested by the business processing request is provided to the object.
8. A business risk control device, characterized in that: The device comprises: A feature data acquisition module, configured to acquire risk control feature data of a target object in response to a business processing request of the target object for a risk control business; A risk prediction processing model is used to input the risk control feature data of the target object into the current risk prediction model corresponding to the current model application cycle for risk prediction processing to obtain a predicted risk probability; the predicted risk probability represents the possibility that the target object belongs to a preset risk category, the current risk prediction model is trained based on risk performance sample data within a historical model application cycle, and the historical model application cycle performs risk prediction processing based on a historical risk prediction model; A prediction result mapping module, used for calling a prediction result mapping function to perform mapping processing on the predicted risk probability to obtain a mapped risk probability; the prediction result mapping function represents the mapping relationship between the prediction result of the historical risk prediction model and the prediction result of the current risk prediction model, and the prediction result mapping function is obtained based on aligning the logarithmic probability of the prediction result of the historical risk prediction model with the logarithmic probability of the prediction result of the current risk prediction model; A risk control result generation module is used to generate a risk control result based on the comparison between the mapped risk probability and the preset risk control threshold corresponding to the risk control business; the risk control result represents whether the business service requested by the business processing request is provided to the target object.
9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the business risk control method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by a processor to implement the business risk control method according to any one of claims 1 to 7.
11. A computer program, characterized in that When the computer program is executed by a processor, the business risk control method according to any one of claims 1 to 7 is implemented.
