Target object determination method and apparatus, and electronic device

Abstract

The invention discloses a target object determination method and device and electronic equipment. The method comprises the steps of obtaining target management and control operation information corresponding to a target financial product and target data of a plurality of first objects, wherein the target data is used for representing object features of the first objects; calculating a target responsivity value corresponding to the first object in the plurality of first objects according to a target prediction model and the target data of the plurality of first objects, wherein the target prediction model is obtained by carrying out anti-fact inference on the sample data set according to the initial model and carrying out cross training on the initial model according to an anti-fact inference result; and determining a target object from the plurality of first objects according to the target responsivity value so as to execute a target control operation on the target object. According to the method and the device, the technical problem that the accuracy of the determined management and control object is low due to the fact that the assessment accuracy is low when the financial product is managed and the overall risk assessment is carried out on the customer group in the related technology is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and in particular, to a method, apparatus, and electronic device for determining a target object. Background Art

[0002] When managing financial products, some control operations may bring negative effects. Taking the credit scenario as an example, when an institution's credit granting to a customer is revolving credit, the customer can initiate a new draw at any time after the credit approval. Therefore, it is necessary to monitor and control the customer during the loan period. For example, control operations such as reducing the credit limit or freezing are performed on customers who are predicted to be poor during the loan period. However, in the related art, the overall risk of the customer group is evaluated, focusing on the overall effect and gain of the customer group, and the evaluation accuracy is relatively low. The accuracy of the determined control object is relatively low, and the negative effects brought by the control operation to the customer are not considered. For example, the original intention of the freeze operation during the loan period is to reduce the overdue rate, but it may cause customers who were not overdue to become overdue due to the freeze operation, resulting in negative effects.

[0003] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of this application provide a method, apparatus, and electronic device for determining a target object, so as to at least solve the technical problem in the related art that when managing financial products, the overall risk of the customer group is evaluated, resulting in relatively low evaluation accuracy and relatively low accuracy of the determined control object.

[0005] According to one aspect of the embodiments of this application, a method for determining a target object is provided, including: obtaining target control operation information corresponding to a target financial product and target data of multiple first objects, where the target control operation information is used to indicate a target control operation on the target financial product, the first object is an object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; calculating a target response degree value corresponding to the first object among the multiple first objects according to the target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on a sample data set according to an initial model and performing cross-training on the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity of the first object to the target control operation; determining a target object from the multiple first objects according to the target response degree value, so as to perform the target control operation on the target object.

[0006] Furthermore, the target prediction model includes a first target prediction model and a second target prediction model, and the training samples corresponding to the first target prediction model and the second target prediction model are different. Among them, calculating the target response degree value corresponding to the first object in multiple first objects according to the target prediction model and the target data of multiple first objects includes: calculating the target data of multiple first objects according to the first target prediction model to obtain the first target response degree value corresponding to the first object in multiple first objects; calculating the target data of multiple first objects according to the second target prediction model to obtain the second target response degree value corresponding to the first object in multiple first objects; performing weighted calculation on the first target response degree value and the second target response degree value to obtain the target response degree value corresponding to the first object in multiple first objects.

[0007] Furthermore, determining the target object from multiple first objects according to the target response degree value includes: respectively comparing the target response degree value corresponding to the first object in multiple first objects with a preset threshold, and taking the first object corresponding to the target response degree value greater than the preset threshold as the target object.

[0008] Furthermore, the target prediction model is generated through the following steps: determining the control time period and the target control operation according to the task requirements; obtaining the sample data set and randomly dividing the sample data set into a first sample data set and a second sample data set. Among them, the sample data set is composed of the sample data of multiple sample objects, and the sample data is used to characterize the object characteristics of the sample objects. The first sample data set is used to implement the target control operation, and the second sample data set is used not to implement the target control operation; performing the target control operation on the sample objects in the first sample data set during the control time period, and after the end of the control time period, obtaining the true response degree value corresponding to the sample objects in the first sample data set and the true response degree value corresponding to the sample objects in the second sample data set; training the initial model according to the first sample data set, the true response degree value corresponding to the sample objects in the first sample data set, the second sample data set, and the true response degree value corresponding to the sample objects in the second sample data set to obtain the target prediction model.

[0009] Further, the initial model includes a first initial model and a second initial model, and the training samples corresponding to the first initial model and the second initial model are different. Among them, the initial model is trained based on the first sample dataset, the true response degree values corresponding to the sample objects in the first sample dataset, the second sample dataset, and the true response degree values corresponding to the sample objects in the second sample dataset to obtain a target prediction model, including: forming a first training sample set by combining the sample data in the first sample dataset and the true response degree values corresponding to the sample objects in the first sample dataset, and forming a second training sample set by combining the sample data in the second sample dataset and the true response degree values corresponding to the sample objects in the second sample dataset; training the first initial model based on the first training sample set to obtain a trained first initial model, and training the second initial model based on the second training sample set to obtain a trained second initial model; performing counterfactual inference on the second training sample set based on the trained first initial model to obtain a first response degree value, and performing counterfactual inference on the first training sample set based on the trained second initial model to obtain a second response degree value; training the trained first initial model and the trained second initial model based on the sample data in the first sample dataset, the sample data in the second sample dataset, the first response degree value, and the second response degree value to obtain a target prediction model.

[0010] Further, training the trained first initial model and the trained second initial model based on the sample data in the first sample dataset, the sample data in the second sample dataset, the first response degree value, and the second response degree value to obtain a target prediction model, including: calculating the difference between the second response degree value and the true response degree value corresponding to the sample object in the first sample dataset to obtain a third response degree value, and calculating the difference between the first response degree value and the true response degree value corresponding to the sample object in the second sample dataset to obtain a fourth response degree value; forming a third training sample set by combining the sample data in the first sample dataset and the third response degree value, and forming a fourth training sample set by combining the sample data in the second sample dataset and the fourth response degree value; training the trained first initial model based on the third training sample set to obtain a first target prediction model, and training the trained second initial model based on the fourth training sample set to obtain a second target prediction model; combining the first target prediction model and the second target prediction model to form a target prediction model.

[0011] Further, training the trained first initial model based on a third training sample set to obtain a first target prediction model, including: inputting the training samples in the third training sample set into the trained first initial model to obtain the predicted response degree values corresponding to the training samples in the third training sample set; training and optimizing the trained first initial model based on the third response degree values included in the training samples in the third training sample set, the predicted response degree values corresponding to the training samples in the third training sample set, and the third training sample set until the loss function of the first target prediction model meets a preset condition, thereby obtaining the first target prediction model.

[0012] Further, training the trained second initial model based on a fourth training sample set to obtain a second target prediction model, including: inputting the training samples in the fourth training sample set into the trained second initial model to obtain the predicted response degree values corresponding to the training samples in the fourth training sample set; training and optimizing the trained second initial model based on the fourth response degree values included in the training samples in the fourth training sample set, the predicted response degree values corresponding to the training samples in the fourth training sample set, and the fourth training sample set until the loss function of the second target prediction model meets a preset condition, thereby obtaining the second target prediction model.

[0013] According to another aspect of the embodiments of the present application, a method for determining a target object is further provided, including: obtaining the target control operation information corresponding to the target financial product uploaded by the client and the target data of multiple first objects, where the target control operation information is used to indicate the target control operation on the target financial product, the first object is the object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; calculating the target response degree value corresponding to the first object among the multiple first objects in the cloud server based on the target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on the sample data set based on the initial model and cross-training the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity of the first object to the target control operation; determining the target object from the multiple first objects according to the target response degree value; and feeding back the target object to the client to perform the target control operation on the target object.

[0014] According to another aspect of the embodiments of the present application, there is also provided a determining device for a target object, including: a first acquisition unit, configured to acquire target control operation information corresponding to a target financial product and target data of a plurality of first objects, wherein the target control operation information is used to indicate a target control operation on the target financial product, the first object is an object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; a first processing unit, configured to calculate a target response degree value corresponding to a first object among the plurality of first objects according to a target prediction model and the target data of the plurality of first objects, wherein the target prediction model is obtained by performing counterfactual inference on a sample data set according to an initial model and performing cross-training on the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity degree of the first object to the target control operation; a first determination unit, configured to determine a target object from the plurality of first objects according to the target response degree value, so as to perform a target control operation on the target object.

[0015] Further, the target prediction model includes a first target prediction model and a second target prediction model, and the training samples corresponding to the first target prediction model and the second target prediction model are different. Among them, the first processing unit includes: a first calculation subunit, configured to calculate according to the first target prediction model the target data of the plurality of first objects to obtain a first target response degree value corresponding to a first object among the plurality of first objects; a second calculation subunit, configured to calculate according to the second target prediction model the target data of the plurality of first objects to obtain a second target response degree value corresponding to a first object among the plurality of first objects; a third calculation subunit, configured to perform weighted calculation on the first target response degree value and the second target response degree value to obtain a target response degree value corresponding to a first object among the plurality of first objects.

[0016] Further, the first determination unit includes: a comparison subunit, configured to respectively compare the target response degree value corresponding to a first object among the plurality of first objects with a preset threshold, and use the first object corresponding to the target response degree value greater than the preset threshold as the target object.

[0017] Further, the determining device for the target object further includes the following units for generating a target prediction model through the following steps: a second determining unit for determining a control time period and a target control operation according to the task requirements; a second obtaining unit for obtaining a sample data set and randomly dividing the sample data set into a first sample data set and a second sample data set. The sample data set is composed of sample data of multiple sample objects, and the sample data is used to characterize the object features of the sample objects. The first sample data set is used to perform the target control operation, and the second sample data set is used to not perform the target control operation; a second processing unit for performing the target control operation on the sample objects in the first sample data set during the control time period, and after the control time period ends, obtaining the true response degree values corresponding to the sample objects in the first sample data set and the true response degree values corresponding to the sample objects in the second sample data set; a third processing unit for training an initial model based on the first sample data set, the true response degree values corresponding to the sample objects in the first sample data set, the second sample data set, and the true response degree values corresponding to the sample objects in the second sample data set to obtain the target prediction model.

[0018] Further, the initial model includes a first initial model and a second initial model, and the training samples corresponding to the first initial model and the second initial model are different. The third processing unit includes: a generating subunit for forming a first training sample set by combining the sample data in the first sample data set and the true response degree values corresponding to the sample objects in the first sample data set, and forming a second training sample set by combining the sample data in the second sample data set and the true response degree values corresponding to the sample objects in the second sample data set; a first processing subunit for training the first initial model based on the first training sample set to obtain a trained first initial model, and training the second initial model based on the second training sample set to obtain a trained second initial model; a second processing subunit for performing counterfactual inference on the second training sample set based on the trained first initial model to obtain a first response degree value, and performing counterfactual inference on the first training sample set based on the trained second initial model to obtain a second response degree value; a third processing subunit for training the trained first initial model and the trained second initial model based on the sample data in the first sample data set, the sample data in the second sample data set, the first response degree value, and the second response degree value to obtain the target prediction model.

[0019] Further, the third processing sub-unit includes: a calculation module, configured to calculate a difference between the second responsivity value and the true responsivity value corresponding to the sample object in the first sample dataset to obtain a third responsivity value, and calculate a difference between the first responsivity value and the true responsivity value corresponding to the sample object in the second sample dataset to obtain a fourth responsivity value; a first generation module, configured to form a third training sample set by combining the sample data in the first sample dataset and the third responsivity value, and form a fourth training sample set by combining the sample data in the second sample dataset and the fourth responsivity value; a processing module, configured to train the trained first initial model based on the third training sample set to obtain a first target prediction model, and train the trained second initial model based on the fourth training sample set to obtain a second target prediction model; a second generation module, configured to form a target prediction model by combining the first target prediction model and the second target prediction model.

[0020] Further, the processing module includes: a first processing sub-module, configured to input the training samples in the third training sample set into the trained first initial model to obtain the predicted responsivity values corresponding to the training samples in the third training sample set; a second processing sub-module, configured to train and optimize the trained first initial model based on the third responsivity value included in the training samples in the third training sample set, the predicted responsivity values corresponding to the training samples in the third training sample set, and the third training sample set until the loss function of the first target prediction model meets a preset condition, so as to obtain the first target prediction model.

[0021] Further, the processing module further includes: a third processing sub-module, configured to input the training samples in the fourth training sample set into the trained second initial model to obtain the predicted responsivity values corresponding to the training samples in the fourth training sample set; a fourth processing sub-module, configured to train and optimize the trained second initial model based on the fourth responsivity value included in the training samples in the fourth training sample set, the predicted responsivity values corresponding to the training samples in the fourth training sample set, and the fourth training sample set until the loss function of the second target prediction model meets a preset condition, so as to obtain the second target prediction model.

[0022] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium storing a program, wherein when the program runs, it controls the device where the storage medium is located to execute the method for determining a target object in any one of the above.

[0023] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including: a memory storing an executable program; a processor configured to run the program, wherein when the program runs, it executes the method for determining a target object in any one of the above.

[0024] In the embodiments of the present application, by obtaining the target control operation information corresponding to the target financial product and the target data of multiple first objects, where the target control operation information is used to indicate the target control operation of the target financial product, the first object is the object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; calculating the target response degree value corresponding to the first object in the multiple first objects according to the target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on the sample data set based on the initial model and cross-training the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity of the first object to the target control operation; determining the target object from the multiple first objects according to the target response degree value, and performing counterfactual inference on the sample data set and the initial model in the manner of performing the target control operation on the target object, comprehensively considering the impacts brought by the control operation and the non-control operation. Through the cross-training method, the error brought by the samples of the small sample group can be improved by using the samples of the diverse group, that is, the sample imbalance problem can be solved, and the accuracy of the model prediction response degree value is effectively improved, enabling the model to better evaluate the response degree of customers to the control operation, so as to clarify whether the impact brought by the control operation itself is a positive promotion effect or a negative containment effect, thereby effectively avoiding selecting customers with negative effects, achieving the purpose of better realizing the control of the target financial product, thus achieving the technical effects of improving the accuracy of risk assessment and the accuracy of the determined control object, and further solving the technical problem in the related art that when managing financial products, overall risk assessment of the customer group has low assessment accuracy, resulting in low accuracy of the determined control object. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0026] Figure 1 is a schematic diagram of a computer terminal provided in Embodiment 1 of the present application;

[0027] Figure 2 is a flowchart of a method for determining a target object provided in Embodiment 1 of the present application;

[0028] Figure 3 is a schematic flowchart of an optional process for determining a control object provided in Embodiment 1 of the present application;

[0029] Figure 4 is a schematic flowchart of an optional process for X-learner modeling provided in Embodiment 1 of the present application;

[0030] Figure 5 It is a flowchart of the method for determining the target object provided in Embodiment 2 of the present application;

[0031] Figure 6 It is a schematic diagram of the device for determining the target object provided in Embodiment 3 of the present application;

[0032] Figure 7 It is a schematic diagram of the computing terminal provided in Embodiment 4 of the present application. Detailed implementation manners

[0033] In order to enable those skilled in the art of the present technology to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0034] It should be noted that the terms "first", "second", etc. in the description and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily need to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0035] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards in the relevant regions, and corresponding operation entrances are provided for the user to choose to authorize or refuse.

[0036] First, some nouns or terms that appear during the description of the embodiments of the present application are applicable to the following explanations:

[0037] During the loan: In the credit field, the entire credit life cycle is generally divided into before the loan, during the loan, and after the loan. During the loan refers to the life stage where the customer has passed the credit application and has not become a bad debt customer, and is normally borrowing and repaying.

[0038] A / B Testing: A scientific method that randomly divides a sample group into two samples, namely A and B, and implements different actions for each. Generally, one group conducts a certain type of experiment, called the experimental group, and the other group does not conduct the experiment, called the control group. The experimental effect is evaluated by comparing the experimental results of the experimental group and the control group.

[0039] X-learner: A model framework for inferring the causal effect degree caused by a certain behavior.

[0040] Example 1

[0041] According to the embodiments of the present application, a method for determining a target object is further provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0042] The method embodiments provided by the first embodiment of the present application can be executed on a mobile terminal, a computer terminal or a similar computing device. Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing the method for determining a target object is shown. As Figure 1 shown, the computer terminal (or mobile device) 10 may include a set of processors 102 (the set of processors 102 may include, but is not limited to, processing devices such as a microprocessor MCU (Microcontroller Unit) or a field programmable gate array FPGA (Field Programmable Gate Array), and the set of processors 102 may include a set of processors, Figure 1 which are shown as 102a, 102b,..., 102n in ), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB, Universal Serial Bus) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 may further include more or fewer components than Figure 1 shown in , or have a different configuration from Figure 1 shown in .

[0043] It should be noted that one or more of the above-mentioned processors 102 and / or other data processing circuits can generally be referred to as "data processing circuits" herein. The data processing circuit can be embodied in software, hardware, firmware, or any combination thereof, in whole or in part. In addition, the data processing circuit can be a single independent processing module, or be incorporated, in whole or in part, into any one of other elements in the computer terminal 10 (or mobile device).

[0044] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage devices corresponding to the method for determining the target object in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the method for determining the target object described above. The memory 104 can include high-speed random access memory, and can also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 can further include a memory remotely set relative to the processor 102, and these remote memories can be connected to the computer terminal 10 through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, intranet, local area network, mobile communication network, and their combinations.

[0045] The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network can include the wireless network provided by the communication provider of the computer terminal 10. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet wirelessly.

[0046] The display can be, for example, a touch-screen liquid crystal display, which enables the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0047] When managing financial products, some control operations may bring negative effects. Taking the credit scenario as an example, when an institution's credit granting to a customer is revolving credit, the customer can initiate a new draw at any time after the credit approval. Therefore, it is necessary to conduct in - loan monitoring and control of the customer. For example, control operations such as reducing the credit limit and freezing are carried out on customers judged to be poor during the in - loan period. However, in the related technologies, an overall risk assessment is performed on the customer group, focusing on the overall effect and gain of the customer group, with relatively low assessment accuracy. The accuracy of the determined control objects is relatively low, and the negative effects brought by the control operations to the customers are not considered. For example, the original intention of the in - loan freezing operation is to reduce the default rate, but it may cause customers who would not default to default due to the freezing operation, resulting in negative effects.

[0048] In the above technical background, this application provides a method for determining a target object as shown in Figure 2 the following. Figure 2 FIG. is a flowchart of the method for determining a target object provided in Embodiment 1 of this application. The method includes:

[0049] Step S201: Obtain the target control operation information corresponding to the target financial product and the target data of multiple first objects. Among them, the target control operation information is used to indicate the target control operation on the target financial product, the first object is the object holding the target financial product, and the target data is used to characterize the object characteristics of the first object.

[0050] Taking the target financial product as a credit product as an example, the target control operation information is used to indicate the target control operation on the credit product. The target control operation can be an operation to increase the loan amount, an operation to reduce the loan amount, an operation to freeze the loan amount, etc. The first object can be a loan customer whose repayment risk is to be evaluated, and the target data can be characteristic data of the loan customer in multiple dimensions, such as occupation, education level, credit data, etc.

[0051] For example, during the in - loan management of a credit product, the target control operation information corresponding to the credit product (such as information on the operation to increase the loan amount, etc.) and the characteristic data of multiple dimensions corresponding to multiple loan customers whose repayment risks are to be evaluated are regularly obtained for repayment risk assessment.

[0052] Step S202: Calculate the target response degree value corresponding to the first object among the multiple first objects according to the target prediction model and the target data of the multiple first objects. Among them, the target prediction model is obtained by performing counterfactual inference on the sample data set based on the initial model and cross - training the initial model according to the counterfactual inference result. The target response degree value is used to characterize the sensitivity of the first object to the target control operation.

[0053] Based on the target control operation information, a target prediction model can be determined. For example, if the target control operation information indicates a loan amount increase operation, the target prediction model is determined as the prediction model corresponding to the loan amount increase operation; if the target control operation information indicates a loan amount decrease operation, the target prediction model is determined as the prediction model corresponding to the loan amount decrease operation.

[0054] Based on the target prediction model and the target data of multiple first objects, the target response degree value corresponding to the first object in the multiple first objects can be calculated. For example, based on the prediction model corresponding to the loan amount decrease operation and the feature data of multiple dimensions corresponding to multiple loan customers to be evaluated for repayment risk, the sensitivity of the loan customers to be evaluated for repayment risk to the loan amount decrease operation (i.e., the target response degree value) can be calculated.

[0055] Among them, the target prediction model is obtained by performing counterfactual inference on the sample data set based on the initial model and cross-training the initial model based on the counterfactual inference result. The initial model includes a control group (also called an experimental group) model and a control group model. In this solution, the target prediction model is obtained based on X-learner and A / B testing.

[0056] Step S2*3*, based on the target response degree value, determine the target object from the multiple first objects to perform the target control operation on the target object.

[0057] Based on the target response degree value, the target object can be determined from the multiple first objects to perform the target control operation on the target object. For example, if the target control operation is a loan amount decrease operation, the sensitivity of the loan customers to be evaluated for repayment risk to the loan amount decrease operation calculated based on the prediction model corresponding to the loan amount decrease operation can be used to determine the customers to be controlled (i.e., the target objects) from the multiple loan customers, and the loan amount decrease operation is performed on the customers to be controlled.

[0058] In this solution, by performing counterfactual inference based on the sample data set and the initial model, the comprehensive consideration of the impact of the control operation and the impact of not performing the control operation is realized. Through the cross-training method, the error brought by the samples of the small sample group can be improved by the samples of the diverse group, that is, the sample imbalance problem can be solved, and the accuracy of the model prediction response degree value is effectively improved. The model can better evaluate the response degree of customers to the control operation to clarify whether the impact brought by the control operation itself is a positive promotion effect or a negative containment effect, so as to effectively avoid selecting customers with negative effects, better realize the control of the target financial product, improve the accuracy of risk assessment, and improve the accuracy of determining the control object.

[0059] How to determine the sensitivity of loan customers whose repayment risks are to be evaluated to the target control operation is crucial. Therefore, in the method for determining the target object provided in the first embodiment of this application, the target prediction model includes a first target prediction model and a second target prediction model, and the training samples corresponding to the first target prediction model and the second target prediction model are different. Among them, calculating the target response degree value corresponding to the first object among the multiple first objects based on the target prediction model and the target data of the multiple first objects includes: calculating the target data of the multiple first objects according to the first target prediction model to obtain the first target response degree value corresponding to the first object among the multiple first objects; calculating the target data of the multiple first objects according to the second target prediction model to obtain the second target response degree value corresponding to the first object among the multiple first objects; performing weighted calculation on the first target response degree value and the second target response degree value to obtain the target response degree value corresponding to the first object among the multiple first objects.

[0060] Optionally, the first target prediction model may be a control group response degree model, denoted as τ1, and the second target prediction model may be a control group response degree model, denoted as τ0. The first target response degree value is the response degree value calculated based on the control group response degree model τ1 for the customer feature data, denoted as τ1(x), and the second target response degree value is the response degree value calculated based on the control group response degree model τ0 for the customer feature data, denoted as τ0(x). For example, calculating the response degree value τ1(x) corresponding to the loan customer whose repayment risk is to be evaluated based on the control group response degree model τ1 for the feature data of the loan customer whose repayment risk is to be evaluated, and calculating the response degree value τ0(x) corresponding to the loan customer whose repayment risk is to be evaluated based on the control group response degree model τ0 for the feature data of the loan customer whose repayment risk is to be evaluated.

[0061] Performing weighted calculation on the first target response degree value and the second target response degree value can obtain the target response degree value corresponding to the first object among the multiple first objects. Among them, the response degree (Average Treatment Effect, abbreviated as ATE) is also called the average treatment effect, and the target response degree value is denoted as ATE. For example, the target response degree value ATE is calculated through the following formula:

[0062] ATE = ζτ0(x) + (1 - ζ)τ1(x)

[0063] Among them, ζ represents the propensity score. When the samples are unbalanced, it can be set as the proportion of the control samples, that is, the proportion of the number of control samples to the total number of samples (that is, the sum of the number of control samples and the number of control group samples); when the samples are balanced, it can be set as 0.5.

[0064] It should be noted that by evaluating the responsiveness based on customer characteristic data, the value improvement brought by the control operation itself can be obtained, achieving a better model prediction effect.

[0065] In order to accurately determine the control object, in the method for determining the target object provided in the first embodiment of this application, based on the target responsiveness value, the target object is determined from multiple first objects, including: respectively comparing the target responsiveness value corresponding to the first object in the multiple first objects with a preset threshold, and taking the first object corresponding to the target responsiveness value greater than the preset threshold as the target object.

[0066] For example, the loan customers to be evaluated for repayment risk are scored according to the target prediction model, and the score obtained is the customer responsiveness estimated by the model. Then, the target responsiveness value corresponding to the loan customers to be evaluated for repayment risk is respectively compared with the preset threshold, and the comparison result corresponding to the loan customers to be evaluated for repayment risk can be obtained. If the comparison result indicates that the target responsiveness value is greater than the preset threshold, the loan customers corresponding to the target responsiveness value greater than the preset threshold are taken as the control objects, that is, the loan customers with a responsiveness value higher than the preset threshold are circled for in - loan control.

[0067] It should be noted that the target prediction model obtained based on counterfactual inference and cross - training can better evaluate customer responsiveness, so as to accurately determine the control object based on the customer responsiveness estimated by the model.

[0068] In order to accurately determine the control object, in the method for determining the target object provided in the first embodiment of this application, the target prediction model is generated through the following steps: determining the control time period and the target control operation according to the task requirements; obtaining the sample data set and randomly dividing the sample data set into a first sample data set and a second sample data set. Among them, the sample data set is composed of the sample data of multiple sample objects, and the sample data is used to represent the object characteristics of the sample objects. The first sample data set is used to perform the target control operation, and the second sample data set is used not to perform the target control operation; performing the target control operation on the sample objects in the first sample data set within the control time period, and after the end of the control time period, obtaining the true responsiveness value corresponding to the sample objects in the first sample data set and the true responsiveness value corresponding to the sample objects in the second sample data set; training the initial model according to the first sample data set, the true responsiveness value corresponding to the sample objects in the first sample data set, the second sample data set, and the true responsiveness value corresponding to the sample objects in the second sample data set to obtain the target prediction model.

[0069] In the process of generating the target prediction model, the control time period and the target control operation can be determined according to the task requirements. For example, the task requirements can be to increase the loan amount, freeze the loan amount, etc., then the determined target control operations can be loan amount increase operations, loan amount freeze operations, etc. Here, the control time period represents the observation window period, that is, the observation time period for implementing the control operation on the control group samples.

[0070] In this solution, the target prediction model is built based on X-learner and A / B testing. Obtain the sample data set. For example, select loan customers who want to evaluate the repayment risk from loan customers as multiple sample objects, and obtain the feature data (i.e., sample data) of multiple dimensions corresponding to the multiple sample objects, such as occupation, education level, credit data, etc. These data form the sample data set. Randomly divide the sample data set into the first sample data set (such as the control group sample data set) and the second sample data set (such as the control group sample data set).

[0071] Apply in-loan control to the control group sample data set. For example, perform the target control operation (such as loan amount increase operation) on the control sample objects in the control group sample data set during the control time period, and after the control time period ends, obtain the true response value corresponding to the control sample object and the true response value corresponding to the control sample object in the control group sample data set. Taking the loan amount increase operation as an example, the task indicator (i.e., the response) that is concerned about can be the daily average amount usage increase rate in the future period of time. According to the relevant data during the control time period, the daily average amount usage increase rate of the control sample customers can be calculated as the true response value corresponding to the control sample object, and the daily average amount usage increase rate of the control sample customers can be calculated as the true response value corresponding to the control sample object. Then, train the initial model based on the control group sample data set, the true response value corresponding to the control sample object, the control group sample data set, and the true response value corresponding to the control sample object, and the target prediction model can be obtained.

[0072] In order to obtain the target prediction model, in the method for determining the target object provided in the first embodiment of this application, the initial model includes a first initial model and a second initial model, and the training samples corresponding to the first initial model and the second initial model are different. Among them, the initial model is trained based on the first sample dataset, the true response degree values corresponding to the sample objects in the first sample dataset, the second sample dataset, and the true response degree values corresponding to the sample objects in the second sample dataset to obtain the target prediction model, including: forming a first training sample set by combining the sample data in the first sample dataset and the true response degree values corresponding to the sample objects in the first sample dataset, and forming a second training sample set by combining the sample data in the second sample dataset and the true response degree values corresponding to the sample objects in the second sample dataset; training the first initial model based on the first training sample set to obtain a trained first initial model, and training the second initial model based on the second training sample set to obtain a trained second initial model; performing counterfactual inference on the second training sample set based on the trained first initial model to obtain a first response degree value, and performing counterfactual inference on the first training sample set based on the trained second initial model to obtain a second response degree value; training the trained first initial model and the trained second initial model based on the sample data in the first sample dataset, the sample data in the second sample dataset, the first response degree value, and the second response degree value to obtain the target prediction model.

[0073] The initial model includes a first initial model (such as the control group model to be trained, denoted as μ1) and a second initial model (such as the control group model to be trained, denoted as μ0). For example, the sample data in the control group sample dataset (denoted as X i (1), representing the feature vector) and the true response degree value corresponding to the control sample object (denoted as Y i (1), used as the true label) form the first training sample set (i.e., the control group sample set, denoted as X i (1), Y i (1)), and the sample data in the control group sample dataset (denoted as X i (0), representing the feature vector) and the true response degree value corresponding to the control sample object (denoted as Y i (0), used as the true label) form the second training sample set (i.e., the control group sample set, denoted as X i (0), Y i (0)).

[0074] Based on the control group sample set X i (1), Y i (1) train the control group model μ1 to be trained to obtain a trained control group model μ1 (i.e., the trained first initial model), and based on the control group sample set X i(0), Y i (0) The control group model μ0 to be trained is trained to obtain the trained control group model μ0 (i.e., the trained second initial model).

[0075] Since the two models (i.e., μ1 and μ0) are independent of each other and no connection is established after modeling the control group and the control group respectively, the sample size and dimension will be quite different, resulting in a large deviation in the prediction results. Therefore, in this scheme, the target prediction model is obtained based on X-learner modeling, that is, the two models (i.e., μ1 and μ0) are used to perform counterfactual inference and cross-training on each other. For example, based on the trained control group model μ1, the control group sample set X i (0), Y i (0) Perform counterfactual inference and obtain the first responsiveness value (denoted as μ1(X i (0)), used to calculate the new label), that is, assuming that the control group samples are controlled, the response value is obtained; according to the trained control group model μ0, the control group sample set X i (1), Y i (1) Perform counterfactual inference to obtain the second responsiveness value (denoted as μ0(X i (1)) is used to calculate the new label), that is, the response value obtained by assuming that the control group samples have not been controlled. At this point, the connection between unrelated samples is established.

[0076] Based on the sample data X in the control group sample data set i (1) Sample data X in the control group sample data set i (0), the first response value μ1(X i (0)) and the second responsivity value μ0(X i (1)) The trained control group model μ1 and the trained control group model μ0 are trained, that is, cross-training is achieved, and the target prediction model can be obtained.

[0077] In order to achieve cross-training to obtain a target prediction model, in the method for determining a target object provided in Embodiment 1 of this application, the trained first initial model and the trained second initial model are trained based on the sample data in the first sample dataset, the sample data in the second sample dataset, the first response degree value, and the second response degree value to obtain a target prediction model, including: calculating the difference between the second response degree value and the true response degree value corresponding to the sample object in the first sample dataset to obtain a third response degree value, and calculating the difference between the first response degree value and the true response degree value corresponding to the sample object in the second sample dataset to obtain a fourth response degree value; forming a third training sample set with the sample data in the first sample dataset and the third response degree value, and forming a fourth training sample set with the sample data in the second sample dataset and the fourth response degree value; training the trained first initial model based on the third training sample set to obtain a first target prediction model, and training the trained second initial model based on the fourth training sample set to obtain a second target prediction model; and forming a target prediction model with the first target prediction model and the second target prediction model.

[0078] Optionally, calculating the difference between the second response degree value μ0(X i (1)) and the true response degree value Y i (1) of the controlled sample object can obtain a third response degree value (denoted as D i (1), used as a new label), that is, the third response degree value D i (1) = Y i (1) - μ0(X i (1)); calculating the difference between the first response degree value μ1(X i (0)) and the true response degree value Y i (0) of the control sample object can obtain a fourth response degree value (denoted as D i (0), used as a new label), that is, the fourth response degree value D i (0) = μ1(X i (0)) - Y i (0).

[0079] It should be noted that since the trained control group model μ0 is the model corresponding to the situation without control operations, therefore, based on the trained control group model μ0, counterfactual inference is performed on the controlled group sample set X i (1), Y i (1), and it can be inferred what the corresponding response degree value would be if the samples in the controlled group did not undergo control operations, that is, the second response degree value μ0(X i(1)), so as to analyze the corresponding difference in response values between the control group samples with and without control operations, that is, based on the second response value μ0(X i (1)) and the true response value Y i (1) of the control sample object, calculate the difference to obtain the third response value D i (1), that is, the third response value D i (1) = Y i (1) - μ0(X i (1)). Furthermore, this difference can be used as a new label to train the trained control group model μ1, realizing the construction of the connection between unrelated samples to effectively improve the accuracy of determining the control object.

[0080] Combine the sample data X i (1) and the third response value D i (1) in the control group sample data set to form the third training sample set (denoted as X i (1), D i (1)), and combine the sample data X i (0) and the fourth response value D i (0) in the control group sample data set to form the fourth training sample set (denoted as X i (0), D i (0)).

[0081] Train the trained control group model μ1 based on the third training sample set X i (1), D i (1) to obtain the first target prediction model (i.e., the control group response model τ1), and train the trained control group model μ0 based on the fourth training sample set X i (0), D i (0) to obtain the second target prediction model (i.e., the control group response model τ0), and combine the control group response model τ1 and the control group response model τ0 to form the target prediction model.

[0082] In order to obtain the control group response model τ1, in the method for determining a target object provided in the first embodiment of this application, the trained first initial model is trained based on the third training sample set to obtain a first target prediction model, including: inputting the training samples in the third training sample set into the trained first initial model to obtain the predicted response values corresponding to the training samples in the third training sample set; training and optimizing the trained first initial model based on the third response values included in the training samples in the third training sample set, the predicted response values corresponding to the training samples in the third training sample set, and the third training sample set until the loss function of the first target prediction model meets the preset conditions, and obtaining the first target prediction model.

[0083] Optionally, input the training samples in the third training sample set X i (1), D i (1) into the trained control group model μ1, and output the predicted response value of the model (i.e., the predicted response value, denoted as H i (1)), and based on the sample label (here referring to the third response value D i (1)), the predicted response value H i (1), and the third training sample set X i z(1), D i (1) train and optimize the trained control group model μ1 until the loss function of the control group response model τ1 meets the preset conditions, and obtain the control group response model τ1. For example, compare the difference between the sample label and the predicted response value, construct a loss function with the goal of minimizing the difference, and perform iterative training to obtain the control group response model τ1, which will not be elaborated here.

[0084] In order to obtain the control group response model τ0, in the method for determining a target object provided in the first embodiment of this application, the trained second initial model is trained based on the fourth training sample set to obtain a second target prediction model, including: inputting the training samples in the fourth training sample set into the trained second initial model to obtain the predicted response values corresponding to the training samples in the fourth training sample set; training and optimizing the trained second initial model based on the fourth response values included in the training samples in the fourth training sample set, the predicted response values corresponding to the training samples in the fourth training sample set, and the fourth training sample set until the loss function of the second target prediction model meets the preset conditions, and obtaining the second target prediction model.

[0085] Optionally, input the training samples in the fourth training sample set X i (0), D i (0) into the trained control group model μ0, and output the predicted response value of the model (i.e., the predicted response value, denoted as Hi (0)), based on the sample label (here refers to the fourth response value D i (0)), the predicted response value H i (0) and the fourth training sample set X i (0), D i (0), train and optimize the trained control group model μ0 according to D(0), until the loss function of the control group response model τ0 meets the preset conditions, and obtain the control group response model τ0. For example, compare the difference between the sample label and the predicted response value, construct a loss function with the goal of minimizing the difference and perform iterative training to obtain the control group response model τ0, which will not be elaborated here.

[0086] In an optional embodiment, the Figure 3 schematic diagram shown can be used to determine the controlled object. Figure 3 is a schematic diagram of an optional process for determining the controlled object provided in Embodiment 1 of the present application. As Figure 3 shown, it includes target definition, customer group selection (random), application of in - loan control, task index calculation, X - learner modeling, and intelligent customer acquisition.

[0087] Target definition means that it is necessary to clarify the type of in - loan control (that is, the control operation to be performed), and define the task index (that is, the response) to be concerned according to the type of in - loan control. For example, if it is an operation to increase the loan amount, the task index can be the daily average amount usage increase rate in a future period; if it is an operation to freeze the loan amount, the task index can be whether there is an overdue situation or the overdue amount in a future period, etc.

[0088] Customer group selection means obtaining a sample data set. For example, select loan customers who want to evaluate the repayment risk from loan customers as multiple sample objects, and obtain the feature data of multiple dimensions corresponding to the multiple sample objects, such as occupation, education level, credit data, etc. These data form a sample data set. Randomly divide the sample data set into two sample data sets, one for implementing control (experimental group), and the other without implementing control (control group).

[0089] It should be noted that in order to minimize the intervention in financial products (such as credit products), the samples in the control group (that is, the group performing the control operation) will be relatively few, and the samples in the control group (that is, the group not performing the control operation) will be relatively many. In this way, there will be an insufficient sample set and a sufficient sample set. Based on X - learner through the cross - method, the samples of the multi - sample group can be used to improve the error brought by the model of the few - sample group, and the sample imbalance problem is better solved.

[0090] Loan mid-term control refers to performing operations such as increasing the loan amount or freezing the loan amount for customers in the experimental group (i.e., the controlled group) during the observation window period (i.e., the control time period); task indicator calculation refers to being able to obtain the true response degree value corresponding to the controlled sample object and the true response degree value corresponding to the control sample object after the end of the observation window period. Among them, the setting of the observation window period needs to be determined according to the product form. When the window period is too short, it is not sufficient to observe clearly, while when the window period is too long, it will increase the implementation time of the entire solution. For example, if the loan mid-term control operation is to increase the loan amount, a short window period (such as 1 month) can be set to observe whether customers will use more loans within 1 month.

[0091] In an optional embodiment, the Figure 4 shown schematic diagram can be used to implement X-learner modeling. Figure 4 is an optional process schematic diagram based on X-learner modeling provided in Embodiment 1 of the present application. As Figure 4 shown, a controlled group model μ1 is trained based on the controlled group samples (i.e., X i (1)) and the true labels corresponding to the controlled group samples (i.e., Y i (1)), and a control group model μ0 is trained based on the control group samples (i.e., X i (0)) and the true labels corresponding to the control group samples (i.e., Y i (0)). Among them, X i ∈R d represents a d-dimensional feature vector, 1 represents the controlled group, and 0 represents the control group.

[0092] Since after separately modeling the controlled group and the control group, the two models (i.e., μ1 and μ0) are independent of each other and no connection is established, there will be large differences in the sample size and dimension, resulting in a large deviation in the prediction result. Therefore, in this solution, the two models (i.e., μ1 and μ0) are respectively used to perform counterfactual inference and cross-training with each other. As Figure 4 shown, counterfactual inference is performed on the control group sample X i (0) according to the trained controlled group model μ1 to obtain the response degree value μ1(X i (0)), that is, the control result corresponding to the control group sample X i (0), which is also the response degree value situation obtained assuming that the control group sample has undergone the control operation; counterfactual inference is performed on the controlled group sample X i (1) according to the trained control group model μ0 to obtain the response degree value μ0(X i (1)), that is, the response degree value corresponding to the controlled group sample X i(1) The corresponding unregulated results, that is, assuming that the samples in the control group have not undergone control operations, the obtained response value situation. Thus, the connection between unrelated samples is constructed, which can effectively improve the evaluation accuracy of the model, thereby improving the accuracy of determining the control object, and further improving the accuracy of control.

[0093] As Figure 4 shown, subtract the original sample label of the samples in the control group (i.e., Y i (1), representing the control result) from the model result (herein referring to μ0(X i (1)), representing the unregulated result) to obtain the response value of the samples in the control group (i.e., D i (1)). Use this response value as the new label value to train the control group response model τ1; subtract the original sample label of the samples in the control group (i.e., Y i (0), representing the unregulated result) from the model result (herein referring to μ1(X i (0), representing the control result) to obtain the response value of the samples in the control group (i.e., D i (0)). Use this response value as the new label value to train the control group response model τ0. Thus, in the Figure 3 shown intelligent customer acquisition stage, based on the target prediction models (i.e., the control group response model τ1 and the control group response model τ0), evaluate the loan customers whose repayment risks are to be evaluated, predict the customer response value (i.e., the response value ATE = ζτ0(x)+(1 - ζ)τ1(x)), and select the customers with response values higher than the preset threshold for corresponding mid-loan control operations. Moreover, it can be iterated continuously to optimize the model.

[0094] It should be noted that in the model construction, by using X-learner, the connection between unrelated samples is cleverly constructed. Moreover, for the samples in the control group and the control group, two models are respectively constructed, that is, for the samples in the control group, the control group model μ1 and the control group response model τ1 are constructed; for the samples in the control group, the control group model μ0 and the control group response model τ0 are constructed, which improves the sample utilization rate and fully excavates the sample characteristics.

[0095] It should be noted that the types of the four models (i.e., μ1, τ1, μ0, τ0) involved in this solution are determined according to the task requirements and task scenarios. For example, they can be binary classification models or multi-classification models or regression models, etc.

[0096] It should be noted that in this solution, a modeling solution based on X-learner is proposed to evaluate the customer response (i.e., the value improvement degree brought by the control operation itself) according to the customer characteristic data. After obtaining the customer response, customers with high response can be selected, effectively avoiding selecting customers with negative effects.

[0097] In the embodiments of the present application, by obtaining the target control operation information corresponding to the target financial product and the target data of multiple first objects, where the target control operation information is used to indicate the target control operation of the target financial product, the first object is the object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; calculating the target response degree value corresponding to the first object in the multiple first objects according to the target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on the sample data set based on the initial model and performing cross-training on the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity degree of the first object to the target control operation; determining the target object from the multiple first objects according to the target response degree value, and performing counterfactual inference based on the sample data set and the initial model in the manner of performing the target control operation on the target object, comprehensively considering the impacts brought by the control operation and the non-control operation. Through the cross-training method, the error brought by the samples of the small sample group can be improved by using the samples of the diverse group, that is, the sample imbalance problem can be solved, and the accuracy of the model prediction response degree value can be effectively improved, so that the model can better evaluate the response degree of customers to the control operation, so as to clarify whether the impact brought by the control operation itself is a positive promotion effect or a negative containment effect, thereby effectively avoiding selecting customers with negative effects, achieving the purpose of better realizing the control of the target financial product, thus achieving the technical effects of improving the accuracy of risk assessment and improving the accuracy of the determined control object, and further solving the technical problem that when managing financial products in the related art, overall risk assessment is performed on the customer group, resulting in lower assessment accuracy and lower accuracy of the determined control object.

[0098] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0099] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present application.

[0100] Embodiment 2

[0101] According to an embodiment of the present application, a method for determining a target object is further provided, as Figure 5 shown, the method includes:

[0102] Step S501, obtaining target control operation information corresponding to a target financial product and target data of multiple first objects uploaded by a client. Among them, the target control operation information is used to indicate a target control operation on the target financial product, the first object is an object holding the target financial product, and the target data is used to characterize the object characteristics of the first object;

[0103] Step S502, calculating a target response degree value corresponding to a first object among the multiple first objects in a cloud server according to a target prediction model and the target data of the multiple first objects. The target prediction model is obtained by performing counterfactual inference on a sample data set according to an initial model and performing cross-training on the initial model according to the counterfactual inference result. The target response degree value is used to characterize the sensitivity of the first object to the target control operation; determining a target object from the multiple first objects according to the target response degree value;

[0104] Step S503, feeding back the target object to the client to perform a target control operation on the target object.

[0105] Through the above solution, counterfactual inference is carried out based on the sample data set and the initial model, comprehensively considering the impacts brought by control operations and non-control operations. Through the cross-training method, the errors brought by the samples of the few-sample group can be improved by the samples of the multi-sample group, that is, the sample imbalance problem can be solved, effectively improving the accuracy of the model prediction response value, enabling the model to better evaluate the response degree of customers to control operations, so as to clarify whether the impact brought by the control operation itself is a positive promotion effect or a negative containment effect, thereby effectively avoiding selecting customers with negative effects, achieving the purpose of better realizing the control of the target financial product, thus achieving the technical effects of improving the accuracy of risk assessment and the accuracy of the determined control objects, and further solving the technical problem in the related art that when managing financial products, the overall risk assessment of the customer group has low assessment accuracy, resulting in low accuracy of the determined control objects.

[0106] In the cloud server, the specific method for determining the target object is the same as the method in Embodiment 1, and will not be elaborated here.

[0107] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0108] Through the description of the above implementation manners, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such an understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of the various embodiments of the present application.

[0109] Embodiment 3

[0110] According to an embodiment of the present application, there is also provided a target object determination device for implementing the above target object determination method, as Figure 6 shown. The device includes: a first acquisition unit 601, a first processing unit 602, and a first determination unit 603.

[0111] A first acquisition unit 601, configured to acquire target control operation information corresponding to a target financial product and target data of multiple first objects, where the target control operation information is used to indicate a target control operation on the target financial product, the first object is an object holding the target financial product, and the target data is used to characterize the object characteristics of the first object;

[0112] A first processing unit 602, configured to calculate a target response degree value corresponding to a first object among the multiple first objects according to a target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on a sample data set according to an initial model and performing cross-training on the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity of the first object to the target control operation;

[0113] A first determination unit 603, configured to determine a target object from the multiple first objects according to the target response degree value, so as to perform a target control operation on the target object.

[0114] In the determining device for a target object provided in Embodiment 3 of the present application, a target control operation information corresponding to a target financial product and target data of multiple first objects are obtained by a first obtaining unit 601, where the target control operation information is used to indicate a target control operation on the target financial product, the first object is an object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; a first processing unit 602 calculates a target response degree value corresponding to a first object among the multiple first objects according to a target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on a sample data set according to an initial model and performing cross-training on the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity degree of the first object to the target control operation; a first determining unit 603 determines a target object from the multiple first objects according to the target response degree value, so as to perform a target control operation on the target object. In this solution, by performing counterfactual inference according to the sample data set and the initial model, the comprehensive consideration of the influence brought by the control operation and the influence brought by not performing the control operation is realized. Through the cross-training method, the error brought by the samples of the small sample group can be improved by using the samples of the diverse group, that is, the sample imbalance problem can be solved, and the accuracy of the model prediction response degree value is effectively improved, so that the model can better evaluate the response degree of customers to the control operation, so as to clarify whether the influence brought by the control operation itself is a positive promotion effect or a negative containment effect, thereby effectively avoiding selecting customers with negative effects, achieving the purpose of better realizing the control of the target financial product, thus achieving the technical effects of improving the accuracy of risk assessment and the accuracy of the determined control object, and further solving the technical problem that when managing financial products in the related art, the overall risk assessment of the customer group has low assessment accuracy, resulting in low accuracy of the determined control object.

[0115] Optionally, in the determining device for a target object provided in Embodiment 3 of the present application, the target prediction model includes a first target prediction model and a second target prediction model, and the training samples corresponding to the first target prediction model and the second target prediction model are different. Among them, the first processing unit includes: a first calculation subunit, configured to calculate according to the first target prediction model the target data of the multiple first objects to obtain a first target response degree value corresponding to a first object among the multiple first objects; a second calculation subunit, configured to calculate according to the second target prediction model the target data of the multiple first objects to obtain a second target response degree value corresponding to a first object among the multiple first objects; a third calculation subunit, configured to perform weighted calculation on the first target response degree value and the second target response degree value to obtain a target response degree value corresponding to a first object among the multiple first objects.

[0116] Optionally, in the determining device for a target object provided in Embodiment 3 of the present application, the first determining unit includes: a comparison subunit, configured to respectively compare the target response degree value corresponding to the first object among multiple first objects with a preset threshold, and use the first object corresponding to the target response degree value greater than the preset threshold as the target object.

[0117] Optionally, in the determining device for a target object provided in Embodiment 3 of the present application, the determining device for the target object further includes the following units for generating a target prediction model through the following steps: a second determining unit, configured to determine a control time period and a target control operation according to task requirements; a second obtaining unit, configured to obtain a sample data set and randomly divide the sample data set into a first sample data set and a second sample data set, where the sample data set is composed of sample data of multiple sample objects, the sample data is used to characterize the object features of the sample objects, the first sample data set is used to perform the target control operation, and the second sample data set is used to not perform the target control operation; a second processing unit, configured to perform the target control operation on the sample objects in the first sample data set within the control time period, and after the control time period ends, obtain the true response degree values corresponding to the sample objects in the first sample data set and the true response degree values corresponding to the sample objects in the second sample data set; a third processing unit, configured to train an initial model according to the first sample data set, the true response degree values corresponding to the sample objects in the first sample data set, the second sample data set, and the true response degree values corresponding to the sample objects in the second sample data set to obtain the target prediction model.

[0118] Optionally, in the determining device for the target object provided in Embodiment 3 of the present application, the initial model includes a first initial model and a second initial model, and the training samples corresponding to the first initial model and the second initial model are different. Among them, the third processing unit includes: a generating subunit, configured to form a first training sample set by combining the sample data in the first sample dataset and the true response degree values corresponding to the sample objects in the first sample dataset, and form a second training sample set by combining the sample data in the second sample dataset and the true response degree values corresponding to the sample objects in the second sample dataset; a first processing subunit, configured to train the first initial model based on the first training sample set to obtain a trained first initial model, and train the second initial model based on the second training sample set to obtain a trained second initial model; a second processing subunit, configured to perform counterfactual inference on the second training sample set based on the trained first initial model to obtain a first response degree value, and perform counterfactual inference on the first training sample set based on the trained second initial model to obtain a second response degree value; a third processing subunit, configured to train the trained first initial model and the trained second initial model based on the sample data in the first sample dataset, the sample data in the second sample dataset, the first response degree value, and the second response degree value to obtain a target prediction model.

[0119] Optionally, in the determining device for the target object provided in Embodiment 3 of the present application, the third processing subunit includes: a calculation module, configured to perform a difference calculation based on the second response degree value and the true response degree value corresponding to the sample object in the first sample dataset to obtain a third response degree value, and perform a difference calculation based on the first response degree value and the true response degree value corresponding to the sample object in the second sample dataset to obtain a fourth response degree value; a first generating module, configured to form a third training sample set by combining the sample data in the first sample dataset and the third response degree value, and form a fourth training sample set by combining the sample data in the second sample dataset and the fourth response degree value; a processing module, configured to train the trained first initial model based on the third training sample set to obtain a first target prediction model, and train the trained second initial model based on the fourth training sample set to obtain a second target prediction model; a second generating module, configured to form a target prediction model by combining the first target prediction model and the second target prediction model.

[0120] Optionally, in the target object determination device provided in the third embodiment of this application, the processing module includes: a first processing sub-module, configured to input the training samples in the third training sample set into the trained first initial model to obtain the predicted response degree values corresponding to the training samples in the third training sample set; a second processing sub-module, configured to train and optimize the trained first initial model according to the third response degree values included in the training samples in the third training sample set, the predicted response degree values corresponding to the training samples in the third training sample set, and the third training sample set until the loss function of the first target prediction model meets the preset conditions, and obtain the first target prediction model.

[0121] Optionally, in the target object determination device provided in the third embodiment of this application, the processing module further includes: a third processing sub-module, configured to input the training samples in the fourth training sample set into the trained second initial model to obtain the predicted response degree values corresponding to the training samples in the fourth training sample set; a fourth processing sub-module, configured to train and optimize the trained second initial model according to the fourth response degree values included in the training samples in the fourth training sample set, the predicted response degree values corresponding to the training samples in the fourth training sample set, and the fourth training sample set until the loss function of the second target prediction model meets the preset conditions, and obtain the second target prediction model.

[0122] It should be noted here that the above-mentioned first acquisition unit 601, first processing unit 602, and first determination unit 603 correspond to steps S201 to S203 in Embodiment 1. The above units have the same examples and application scenarios as the corresponding steps, but are not limited to the content disclosed in the above-mentioned Embodiment 1. It should be noted that the above modules, as part of the device, can run in the computer terminal 10 provided in Embodiment 1.

[0123] It should be noted that the preferred implementation schemes involved in the above embodiments of this application are the same as the schemes, application scenarios, and implementation processes provided in Embodiment 1, but are not limited to the schemes provided in Embodiment 1.

[0124] Embodiment 4

[0125] An embodiment of this application can provide a computer terminal, and this computer terminal can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the above computer terminal can also be replaced with a terminal device such as a mobile terminal.

[0126] Optionally, in this embodiment, the above computer terminal can be located in at least one of multiple network devices in a computer network.

[0127] In this embodiment, the above computer terminal may execute the program code of the following steps in the method for determining a target object: obtaining target control operation information corresponding to a target financial product and target data of multiple first objects, where the target control operation information is used to indicate a target control operation on the target financial product, the first object is an object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; calculating a target response degree value corresponding to the first object among the multiple first objects according to the target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on a sample data set according to an initial model and performing cross-training on the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity of the first object to the target control operation; determining a target object from the multiple first objects according to the target response degree value so as to perform a target control operation on the target object.

[0128] The above computer terminal may also execute the program code of the following steps in the method for determining a target object: calculating, according to a first target prediction model, the target data of the multiple first objects to obtain a first target response degree value corresponding to the first object among the multiple first objects; calculating, according to a second target prediction model, the target data of the multiple first objects to obtain a second target response degree value corresponding to the first object among the multiple first objects; performing weighted calculation on the first target response degree value and the second target response degree value to obtain a target response degree value corresponding to the first object among the multiple first objects.

[0129] The above computer terminal may also execute the program code of the following steps in the method for determining a target object: respectively comparing the target response degree value corresponding to the first object among the multiple first objects with a preset threshold value, and taking the first object corresponding to the target response degree value greater than the preset threshold value as the target object.

[0130] The above computer terminal can also execute the program code of the following steps in the method for determining a target object: determining a control time period and a target control operation according to task requirements; obtaining a sample data set and randomly dividing the sample data set into a first sample data set and a second sample data set, wherein the sample data set is composed of sample data of multiple sample objects, the sample data is used to characterize the object characteristics of the sample objects, the first sample data set is used to perform the target control operation, and the second sample data set is used to not perform the target control operation; performing the target control operation on the sample objects in the first sample data set during the control time period, and after the control time period ends, obtaining the true response degree values corresponding to the sample objects in the first sample data set and the true response degree values corresponding to the sample objects in the second sample data set; training an initial model according to the first sample data set, the true response degree values corresponding to the sample objects in the first sample data set, the second sample data set, and the true response degree values corresponding to the sample objects in the second sample data set to obtain a target prediction model.

[0131] The above computer terminal can also execute the program code of the following steps in the method for determining a target object: combining the sample data in the first sample data set and the true response degree values corresponding to the sample objects in the first sample data set to form a first training sample set, and combining the sample data in the second sample data set and the true response degree values corresponding to the sample objects in the second sample data set to form a second training sample set; training a first initial model according to the first training sample set to obtain a trained first initial model, and training a second initial model according to the second training sample set to obtain a trained second initial model; performing counterfactual inference on the second training sample set according to the trained first initial model to obtain a first response degree value, and performing counterfactual inference on the first training sample set according to the trained second initial model to obtain a second response degree value; training the trained first initial model and the trained second initial model according to the sample data in the first sample data set, the sample data in the second sample data set, the first response degree value, and the second response degree value to obtain a target prediction model.

[0132] The above computer terminal can also execute the program code of the following steps in the method for determining a target object: perform a difference calculation based on the second response degree value and the true response degree value corresponding to the sample object in the first sample dataset to obtain a third response degree value, and perform a difference calculation based on the first response degree value and the true response degree value corresponding to the sample object in the second sample dataset to obtain a fourth response degree value; form a third training sample set by combining the sample data in the first sample dataset and the third response degree value, and form a fourth training sample set by combining the sample data in the second sample dataset and the fourth response degree value; train the trained first initial model based on the third training sample set to obtain a first target prediction model, and train the trained second initial model based on the fourth training sample set to obtain a second target prediction model; and form a target prediction model by combining the first target prediction model and the second target prediction model.

[0133] The above computer terminal can also execute the program code of the following steps in the method for determining a target object: input the training samples in the third training sample set into the trained first initial model to obtain the predicted response degree values corresponding to the training samples in the third training sample set; train and optimize the trained first initial model based on the third response degree value included in the training samples in the third training sample set, the predicted response degree values corresponding to the training samples in the third training sample set, and the third training sample set until the loss function of the first target prediction model meets the preset condition to obtain the first target prediction model.

[0134] The above computer terminal can also execute the program code of the following steps in the method for determining a target object: input the training samples in the fourth training sample set into the trained second initial model to obtain the predicted response degree values corresponding to the training samples in the fourth training sample set; train and optimize the trained second initial model based on the fourth response degree value included in the training samples in the fourth training sample set, the predicted response degree values corresponding to the training samples in the fourth training sample set, and the fourth training sample set until the loss function of the second target prediction model meets the preset condition to obtain the second target prediction model.

[0135] Optionally, Figure 7 is a structural block diagram of a computer terminal according to an embodiment of the present application. As Figure 7 shown, the computer terminal 10 may include: one or more ( Figure 7 only one is shown in the figure) processors 102, a memory 104. The computer terminal 10 may also include a storage controller for controlling and managing the memory 104 through the storage controller; the computer terminal 10 may also include a peripheral interface for connecting a radio frequency module, an audio module, and a display screen, etc. through the peripheral interface.

[0136] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the method and device for determining the target object in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implements the above-mentioned method for determining the target object. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory may further include a memory remotely set relative to the processor, and these remote memories can be connected to the terminal 10 through a network. Examples of the above network include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and combinations thereof.

[0137] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: obtaining the target control operation information corresponding to the target financial product and the target data of multiple first objects, where the target control operation information is used to indicate the target control operation on the target financial product, the first object is the object holding the target financial product, and the target data is used to characterize the object characteristics of the first object; calculating the target response degree value corresponding to the first object among the multiple first objects according to the target prediction model and the target data of the multiple first objects, where the target prediction model is obtained by performing counterfactual inference on the sample data set according to the initial model and performing cross-training on the initial model according to the counterfactual inference result, and the target response degree value is used to characterize the sensitivity of the first object to the target control operation; determining the target object from the multiple first objects according to the target response degree value, so as to perform the target control operation on the target object.

[0138] Optionally, the above processor may further execute the program code of the following steps: calculating the first target response degree value corresponding to the first object among the multiple first objects according to the first target prediction model; calculating the second target response degree value corresponding to the first object among the multiple first objects according to the second target prediction model; performing weighted calculation on the first target response degree value and the second target response degree value to obtain the target response degree value corresponding to the first object among the multiple first objects.

[0139] Optionally, the above processor may further execute the program code of the following steps: respectively comparing the target response degree value corresponding to the first object among the multiple first objects with a preset threshold, and taking the first object corresponding to the target response degree value greater than the preset threshold as the target object.

[0140] Optionally, the above-mentioned processor may also execute the program code of the following steps: Determine the control time period and the target control operation according to the task requirements; Obtain the sample data set and randomly divide the sample data set into a first sample data set and a second sample data set. Among them, the sample data set is composed of the sample data of multiple sample objects, and the sample data is used to characterize the object characteristics of the sample object. The first sample data set is used to implement the target control operation, and the second sample data set is used not to implement the target control operation; Perform the target control operation on the sample objects in the first sample data set during the control time period, and after the control time period ends, obtain the true response degree value corresponding to the sample objects in the first sample data set and the true response degree value corresponding to the sample objects in the second sample data set; Train the initial model according to the first sample data set, the true response degree value corresponding to the sample objects in the first sample data set, the second sample data set, and the true response degree value corresponding to the sample objects in the second sample data set to obtain the target prediction model.

[0141] Optionally, the above-mentioned processor may also execute the program code of the following steps: Combine the sample data in the first sample data set and the true response degree value corresponding to the sample objects in the first sample data set to form a first training sample set, and combine the sample data in the second sample data set and the true response degree value corresponding to the sample objects in the second sample data set to form a second training sample set; Train the first initial model according to the first training sample set to obtain the trained first initial model, and train the second initial model according to the second training sample set to obtain the trained second initial model; Perform counterfactual inference on the second training sample set according to the trained first initial model to obtain a first response degree value, and perform counterfactual inference on the first training sample set according to the trained second initial model to obtain a second response degree value; Train the trained first initial model and the trained second initial model according to the sample data in the first sample data set, the sample data in the second sample data set, the first response degree value, and the second response degree value to obtain the target prediction model.

[0142] Optionally, the above-mentioned processor may also execute the program code of the following steps: calculate the difference between the second response value and the true response value corresponding to the sample object in the first sample dataset to obtain a third response value, and calculate the difference between the first response value and the true response value corresponding to the sample object in the second sample dataset to obtain a fourth response value; form a third training sample set by combining the sample data in the first sample dataset and the third response value, and form a fourth training sample set by combining the sample data in the second sample dataset and the fourth response value; train the trained first initial model based on the third training sample set to obtain a first target prediction model, and train the trained second initial model based on the fourth training sample set to obtain a second target prediction model; combine the first target prediction model and the second target prediction model to form a target prediction model.

[0143] Optionally, the above-mentioned processor may also execute the program code of the following steps: input the training samples in the third training sample set into the trained first initial model to obtain the predicted response values corresponding to the training samples in the third training sample set; train and optimize the trained first initial model based on the third response value included in the training samples in the third training sample set, the predicted response values corresponding to the training samples in the third training sample set, and the third training sample set until the loss function of the first target prediction model meets the preset conditions to obtain the first target prediction model.

[0144] Optionally, the above-mentioned processor may also execute the program code of the following steps: input the training samples in the fourth training sample set into the trained second initial model to obtain the predicted response values corresponding to the training samples in the fourth training sample set; train and optimize the trained second initial model based on the fourth response value included in the training samples in the fourth training sample set, the predicted response values corresponding to the training samples in the fourth training sample set, and the fourth training sample set until the loss function of the second target prediction model meets the preset conditions to obtain the second target prediction model.

[0145] Those of ordinary skill in the art can understand that Figure 7 the structure shown is only illustrative, and the computer terminal may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a handheld computer, and terminal devices such as Mobile Internet Devices (MID), PAD, etc. Figure 7 It does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 may also include more or fewer components (such as a network interface, a display device, etc.) than those shown Figure 7 in the figure, or have a different configuration from that shown Figure 7 in the figure.

[0146] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, and the storage medium can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disk, etc.

[0147] Embodiment 5

[0148] An embodiment of the present application also provides a computer-readable storage medium. Optionally, in this embodiment, the above storage medium can be used to store the program code executed by the method for determining the target object provided in the first embodiment above.

[0149] Optionally, in this embodiment, the above storage medium can be located in any one of the computer terminals in the computer terminal group in the computer network, or in any one of the mobile terminals in the mobile terminal group.

[0150] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0151] In the above embodiments of the present application, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0152] In the several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in an electrical or other form.

[0153] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0154] In addition, in each embodiment of the present application, each functional unit may be integrated into one processing unit, may exist separately as individual physical units, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0155] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, may be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present application. The aforementioned storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0156] The above is only the preferred embodiment of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for determining a target object, characterized in that: include: Obtaining target management and control operation information corresponding to a target financial product and target data of a plurality of first objects, wherein the target management and control operation information indicates a target management and control operation on the target financial product, the first objects are objects holding the target financial product, and the target data characterizes object characteristics of the first objects; Calculating a target responsiveness value corresponding to a first object among the multiple first objects based on a target prediction model and target data of the multiple first objects, wherein the target prediction model is obtained by performing counterfactual inference on a sample data set based on an initial model and cross-training the initial model based on the counterfactual inference result, and the target responsiveness value is used to characterize the sensitivity of the first object to the target control operation; A target object is determined from the plurality of first objects according to the target responsiveness value, so as to perform the target control operation on the target object.

2. The method according to claim 1, characterized in that The target prediction model includes a first target prediction model and a second target prediction model, the first target prediction model and the second target prediction model correspond to different training samples, wherein calculating a target responsiveness value corresponding to a first object among the multiple first objects based on the target prediction model and the target data of the multiple first objects includes: Calculating target data of the plurality of first objects according to the first target prediction model to obtain a first target responsiveness value corresponding to a first object among the plurality of first objects; Calculating the target data of the plurality of first objects according to the second target prediction model to obtain a second target responsivity value corresponding to a first object among the plurality of first objects; A weighted calculation is performed on the first target responsiveness value and the second target responsiveness value to obtain a target responsiveness value corresponding to a first object among the plurality of first objects.

3. The method according to claim 1, characterized in that Determining a target object from the plurality of first objects according to the target responsivity value includes: The target responsiveness values corresponding to the first objects among the plurality of first objects are respectively compared with a preset threshold, and the first object corresponding to the target responsiveness value greater than the preset threshold is taken as the target object.

4. The method according to any one of claims 1 to 3, characterized in that The target prediction model is generated by the following steps: Determine the control time period and the target control operations according to task requirements; Acquire the sample data set, and randomly divide the sample data set into a first sample data set and a second sample data set, wherein the sample data set is composed of sample data of a plurality of sample objects, the sample data is used to characterize object features of the sample objects, the first sample data set is used to implement the target control operation, and the second sample data set is used not to implement the target control operation; performing the target control operation on the sample objects in the first sample data set within the control time period, and obtaining, after the control time period ends, true responsiveness values corresponding to the sample objects in the first sample data set and true responsiveness values corresponding to the sample objects in the second sample data set; The initial model is trained based on the first sample data set, the true response values corresponding to the sample objects in the first sample data set, the second sample data set, and the true response values corresponding to the sample objects in the second sample data set to obtain the target prediction model.

5. The method according to claim 4, characterized in that The initial model includes a first initial model and a second initial model, the first initial model and the second initial model corresponding to different training samples, wherein the initial model is trained based on the first sample data set, the true response value corresponding to the sample objects in the first sample data set, the second sample data set, and the true response value corresponding to the sample objects in the second sample data set to obtain the target prediction model, including: The sample data in the first sample data set and the true response values corresponding to the sample objects in the first sample data set form a first training sample set, and the sample data in the second sample data set and the true response values corresponding to the sample objects in the second sample data set form a second training sample set; Training the first initial model based on the first training sample set to obtain a trained first initial model, and training the second initial model based on the second training sample set to obtain a trained second initial model; Performing counterfactual inference on the second training sample set based on the trained first initial model to obtain a first responsiveness value, and performing counterfactual inference on the first training sample set based on the trained second initial model to obtain a second responsiveness value; The trained first initial model and the trained second initial model are trained according to the sample data in the first sample data set, the sample data in the second sample data set, the first response value, and the second response value to obtain the target prediction model.

6. The method according to claim 5, characterized in that The trained first initial model and the trained second initial model are trained based on the sample data in the first sample data set, the sample data in the second sample data set, the first response value, and the second response value to obtain the target prediction model, including: performing a difference calculation based on the second responsivity value and a true responsivity value corresponding to the sample object in the first sample data set to obtain a third responsivity value, and performing a difference calculation based on the first responsivity value and the true responsivity value corresponding to the sample object in the second sample data set to obtain a fourth responsivity value; The sample data in the first sample data set and the third responsiveness value form a third training sample set, and the sample data in the second sample data set and the fourth responsiveness value form a fourth training sample set; Training the trained first initial model based on the third training sample set to obtain a first target prediction model, and training the trained second initial model based on the fourth training sample set to obtain a second target prediction model; The first target prediction model and the second target prediction model are combined into the target prediction model.

7. The method according to claim 6, characterized in that Training the trained first initial model based on the third training sample set to obtain a first target prediction model includes: Inputting the training samples in the third training sample set into the trained first initial model to obtain predicted responsiveness values corresponding to the training samples in the third training sample set; The trained first initial model is trained and optimized based on the third responsiveness value contained in the training samples in the third training sample set, the predicted responsiveness value corresponding to the training samples in the third training sample set, and the third training sample set until the loss function of the first target prediction model meets the preset conditions, thereby obtaining the first target prediction model.

8. The method according to claim 6, characterized in that Training the trained second initial model according to the fourth training sample set to obtain a second target prediction model, including: Inputting the training samples in the fourth training sample set into the trained second initial model to obtain predicted responsiveness values corresponding to the training samples in the fourth training sample set; The trained second initial model is trained and optimized based on the fourth responsiveness value contained in the training samples in the fourth training sample set, the predicted responsiveness value corresponding to the training samples in the fourth training sample set, and the fourth training sample set until the loss function of the second target prediction model meets the preset conditions, thereby obtaining the second target prediction model.

9. A method for determining a target object, characterized in that: include: Obtaining target management and control operation information corresponding to a target financial product uploaded by a client and target data of a plurality of first objects, wherein the target management and control operation information indicates a target management and control operation on the target financial product, the first objects are objects holding the target financial product, and the target data characterizes object characteristics of the first objects; Calculating, in the cloud server, a target responsiveness value corresponding to a first object among the multiple first objects based on a target prediction model and target data of the multiple first objects, wherein the target prediction model is obtained by performing counterfactual inference on a sample data set based on an initial model and cross-training the initial model based on the counterfactual inference result, and the target responsiveness value is used to characterize the sensitivity of the first object to the target control operation; determining a target object from the multiple first objects based on the target responsiveness value; Feedback the target object to the client to perform the target management and control operation on the target object.

10. A device for determining a target object, characterized in that: include: a first acquiring unit, configured to acquire target management and control operation information corresponding to a target financial product and target data of a plurality of first objects, wherein the target management and control operation information indicates a target management and control operation on the target financial product, the first objects are objects holding the target financial product, and the target data characterizes object characteristics of the first objects; a first processing unit, configured to calculate a target responsiveness value corresponding to a first object among the multiple first objects based on a target prediction model and target data of the multiple first objects, wherein the target prediction model is obtained by performing counterfactual inference on a sample data set based on an initial model and cross-training the initial model based on the counterfactual inference result, and the target responsiveness value is used to characterize the sensitivity of the first object to the target control operation; The first determining unit is configured to determine a target object from the plurality of first objects according to the target responsiveness value, so as to perform the target control operation on the target object.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the storage medium is located is controlled to execute the target object determination method according to any one of claims 1 to 9.

12. An electronic device, characterized in that: include: a memory storing an executable program; A processor, configured to run a program, wherein the method for determining a target object according to any one of claims 1 to 9 is executed when the program is run.

Images