CN115018620A

CN115018620A - Model training method and device, and information processing method and device

Info

Publication number: CN115018620A
Application number: CN202210552573.XA
Authority: CN
Inventors: 王萌; 郑文琛; 张晓军
Original assignee: WeBank Co Ltd
Current assignee: WeBank Co Ltd
Priority date: 2022-05-20
Filing date: 2022-05-20
Publication date: 2022-09-06

Abstract

The invention discloses a model training method and device and an information processing method and device, wherein the method comprises the following steps: at least one group of training samples is obtained, wherein any group of training samples comprises sample object characteristics of sample objects. And training the pre-estimation model according to at least one group of training samples, wherein the pre-estimation model is used for outputting the guiding probability of the sample object according to the characteristics of the sample object, and the guiding probability is used for indicating the probability that the sample object guides other objects to register in the business system. And testing the trained pre-estimation model, and determining the test result parameters of the pre-estimation model. And determining the pre-estimated model after training according to the test result parameters. The technical scheme provided by the invention can effectively improve the effectiveness and pertinence of information pushing.

Description

Model training method and device, and information processing method and device

Technical Field

The present invention relates to computer technologies, and in particular, to a model training method and apparatus, and an information processing method and apparatus.

Background

The user usually applies for the service needed by the service organization, and when the application is successful, the service organization will issue the content related to the service to the user.

In the service handling process, there is a behavior that the user has successfully obtained the issued content and guides other users to apply for the service. In order to effectively increase the number of service transactions, a full-coverage manner is usually adopted to push recommendation information to users who have successfully applied and transacted services, so that the users who have successfully applied guide other users to apply for the services.

However, because the number of users who successfully transact business is huge, the full-coverage approach pushes information to all clients who successfully transact business, which results in lack of pertinence in information pushing.

Disclosure of Invention

The invention mainly aims to provide a model training method and device and an information processing method and device, aiming at improving the pertinence and the effectiveness of information pushing.

In order to achieve the above object, the present invention provides a model training method, including:

obtaining at least one group of training samples, wherein any group of training samples comprise sample object characteristics of sample objects;

training the pre-estimation model according to the at least one group of training samples, wherein the pre-estimation model is used for outputting a guiding probability of the sample object according to the characteristics of the sample object, and the guiding probability is used for indicating the probability that the sample object guides other objects to register in the business system;

testing the trained pre-estimated model, and determining a test result parameter of the pre-estimated model;

and determining the trained pre-estimation model according to the test result parameters.

Also, the present invention provides an information processing method including:

acquiring object characteristics of at least one first object in the preset database, wherein the first object is a registered object in a service system;

processing the object characteristics through a pre-estimation model to obtain a guiding probability corresponding to each first object, wherein the guiding probability is used for indicating the probability that the first object guides a second object to register in the service system;

and determining a target object in the at least one first object according to the guiding probability corresponding to each first object and each first object, and sending recommendation information to equipment corresponding to the target object.

Also, the present invention provides a model training apparatus comprising:

the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring at least one group of training samples, and any group of training samples comprise sample object characteristics of sample objects;

the training module is used for training the pre-estimation model according to the at least one group of training samples, wherein the pre-estimation model is used for outputting the guiding probability of the sample object according to the characteristics of the sample object, and the guiding probability is used for indicating the probability that the sample object guides other objects to register in the business system;

the test module is used for testing the trained pre-estimation model and determining a test result parameter of the pre-estimation model;

and the determining module is used for determining the pre-estimated model after training according to the test result parameters.

And, the present invention provides an information processing apparatus including:

the acquisition module is used for acquiring the object characteristics of at least one first object in the preset database, wherein the first object is an object registered in a service system;

the processing module is used for processing the characteristics of the objects through an estimation model to obtain the guide probability corresponding to each first object, wherein the guide probability is used for indicating the probability that the first object guides the registration of a second object in the service system;

and the determining module is used for determining a target object in the at least one first object according to the guiding probability corresponding to each first object and each first object, and sending recommendation information to the equipment corresponding to the target object.

Also, the present invention provides a model training apparatus including: a memory, a processor, and a model training program stored on the memory and executable on the processor, the model training program when executed by the processor implementing the steps of the model training method as described above.

And, the present invention provides an information processing apparatus including: a memory, a processor and an information processing program stored on the memory and executable on the processor, the information processing program, when executed by the processor, implementing the steps of the information processing method as described above.

Also, the present invention provides a computer-readable storage medium having stored thereon a model training program, which when executed by a processor, implements the steps of the model training method as described above; and (c) a second step of,

the computer-readable storage medium has stored thereon an information processing program which, when executed by a processor, implements the steps of the information processing method as described above.

Also, the invention provides a computer program product comprising a computer program which, when executed by a processor, implements the method as described above.

According to the method, the training sample is obtained, the pre-estimation model is trained according to the training sample, the pre-estimation model is further tested after the training is finished, and the trained pre-estimation model is determined to be obtained when the pre-estimation model passes the test. The pre-estimation model can output the guiding probability of the object according to the object characteristics, and the guiding probability can indicate the probability that the registered object guides other objects to register in the business system. Therefore, the pre-estimation model can be trained to screen out target objects which are most likely to guide other objects to register, and targeted information push is performed on the target objects, so that the effectiveness and the targeting of information push can be effectively improved.

Drawings

FIG. 1 is a schematic diagram of a process of a predictive model according to the present invention;

FIG. 2 is a flowchart of a model training method according to an embodiment of the present invention;

FIG. 3 is a second flowchart of a model training method according to an embodiment of the present invention;

FIG. 4 is a schematic diagram illustrating an implementation of a training prediction model according to an embodiment of the present invention;

fig. 5 is a schematic diagram of an implementation of determining a first target object according to an embodiment of the present invention;

fig. 6 is a schematic diagram of an implementation of determining a second target object according to an embodiment of the present invention;

FIG. 7 is a flowchart of an information processing method according to an embodiment of the present invention;

FIG. 8 is a second flowchart of an information processing method according to an embodiment of the present invention;

FIG. 9 is a schematic diagram illustrating an implementation of determining a target object according to an embodiment of the present invention;

FIG. 10 is a schematic diagram illustrating an implementation of online testing of a predictive model according to an embodiment of the present invention;

FIG. 11 is a schematic structural diagram of a model training apparatus according to an embodiment of the present invention;

FIG. 12 is a schematic structural diagram of an information processing apparatus according to an embodiment of the present invention;

fig. 13 is a schematic diagram of a hardware structure of a model training apparatus according to an embodiment of the present invention;

fig. 14 is a schematic diagram of a hardware structure of an information processing apparatus according to an embodiment of the present invention.

The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

In order to better understand the technical solution of the present invention, the related art related to the present invention will be further described in detail below.

With the continuous development of social processes, various business mechanisms exist at present, users usually apply for the business mechanisms to handle required business, and when the users successfully apply for the business mechanisms, the business mechanisms can issue the content related to the business to the users.

In one specific example, for a banking institution, the applicant may apply for loan transaction to the banking institution, and when the applicant's application passes, the banking institution may issue a credit line to the applicant, and then the applicant may withdraw money.

It can be understood that, in the process of business transaction, there may be an action of the user who has successfully obtained the issued content and guiding other users to apply for the business, and this action may be understood as referral in a colloquial way.

In order to effectively increase the number of service transactions in the service organization, the service organization typically pushes some recommendation information to the user who has successfully obtained the issued content, where the recommendation information may be introduction information of some incentive schemes, so as to encourage the user who has successfully applied to guide the rest of the users to apply for the service. For example, in the loan transaction example described above, the incentive policy may be, for example, interest rate adjustment, or the like.

At present, when a business mechanism pushes recommendation information, a full-coverage mode is adopted to push the recommendation information to all users who have successfully applied and transacted the business.

However, because the number of users who successfully transact business is huge, the full-coverage approach may result in lack of pertinence in information recommendation and may reduce the effectiveness of pushing recommendation information.

Aiming at the problems in the prior art, the invention provides the following technical conception: the object characteristics of the application objects are processed through training the pre-estimation model to obtain the probability that each application object guides other objects to apply for the service, then the target client needing to push the recommendation information is identified according to the probability output by the pre-estimation model, and then the recommendation information is pushed to the target client, so that the pertinence and the effectiveness of the recommendation information pushing can be effectively improved.

Based on the above introduction, it can be understood that the technical solution of the present invention may include two parts, i.e., training of the prediction model and application of the prediction model, for example, the prediction model in the present invention may be understood with reference to fig. 1, where fig. 1 is a schematic processing diagram of the prediction model provided by the present invention.

As shown in fig. 1, the predictive model in the present invention may take an object feature of an application object as an input, process the object feature of the application object, and output a guidance probability corresponding to the application object, where the guidance probability is used to indicate a probability of applying for guiding other objects to perform a service application.

The following describes two parts of the training of the predictive model and the application of the predictive model in the present invention, respectively, with reference to specific embodiments.

Firstly, training of the predictive model is explained, and fig. 2 is a flowchart of a model training method provided in the embodiment of the present invention.

As shown in fig. 2, the method includes:

s201, obtaining at least one group of training samples, wherein any group of training samples comprises sample object characteristics of sample objects.

In this embodiment, in order to train the predictive model, at least one set of training samples, which may include the object features of the sample object, is first obtained. The sample object in this embodiment is an object that has already been registered in the business system. Here, the completion of registration in the service system may be understood as the completion of the service application described above.

For example, assuming that the current business scenario is the scenario of loan application described above, the current sample object may be, for example, a business or a person who has completed the loan application.

Meanwhile, it can be understood that, for the loan service, the behavior related to the withdrawal and repayment of the applicant can reflect the demand degree of the applicant for the loan, and indirectly determine the possibility that the applicant guides other objects to apply for the service. For example, in a general case, the higher the withdrawal amount, the greater the expenses saved if the benefit rate is obtained for the applicant, and the greater the possibility that the applicant will guide other subjects to apply for the service when the recommendation information is pushed to the applicant.

In a possible implementation manner, the object feature in this embodiment may include, for example, a first operation feature and a second operation feature.

In the context of the loan application described above, the first operating characteristic may be understood as a withdrawal characteristic, which may include, for example, at least one of: money drawing amount, money drawing frequency and money drawing time. Further, for example, the total amount currently granted by the sample object, the amount of the withdrawn money, the number of the last x days of the borrowing, the last time of the borrowing, and the like may be specifically mentioned, where x is an integer greater than or equal to 1.

And, the second operating feature may be understood as a payment feature, which may for example comprise at least one of the following: repayment amount, repayment frequency and repayment time. Still further, the payment characteristics may be, for example, the total amount of the current credit granted by the sample object, the amount of the payment that has been paid, the number of the payment on the last y days, the last payment time, and the like, where y is an integer greater than or equal to 1.

Based on the above introduction, it can be understood that in a service scene of a loan application, specific implementation of the withdrawal feature and the repayment feature may also be selected and set according to actual needs, and all features related to the withdrawal behavior may be used as the withdrawal feature in this embodiment, and all features related to the repayment behavior may be used as the repayment feature in this embodiment.

And aiming at service scenes except for loan application, the specific implementation of the first operating characteristic and the second operating characteristic can be selected and set according to actual requirements, as long as the characteristics are related to the operation corresponding to the corresponding service scenes.

In another possible implementation manner, the object characteristics in this embodiment may also include, for example, personalized characteristics of the sample object, that is, some characteristic information of the sample object itself.

For example, the business scenario of loan application described above, the sample object may be a business applying for loan, where the personalized features may include, for example, basic information of the business and historical application information of the business. The basic information of the enterprise may include at least one of the following: business-related basic information of business registered cities, businesses to which the businesses belong, business registered capital, and the like. And the historical application information of the enterprise can be, for example, related information of financial products applied by the enterprise in the past time. In the actual implementation process, the personalized features of the enterprise can be selected and set according to actual requirements, which is not limited in this embodiment.

For the rest of the service scenarios, the personalized features of the sample object can be selected and set according to the actual requirements, and all the information related to the sample object can be used as the personalized features in the embodiment.

It should be noted that, in this embodiment, the sample object is an object that has already completed the registration of the service system, and in a possible implementation manner, all objects that have already completed the registration of the service system are stored in a preset storage space, and at the same time, object information of the objects that have completed the registration of the service system is also stored. At the time of obtaining at least one training sample set, the at least one training sample set may be obtained from a preset storage space, for example.

S202, training a pre-estimation model according to at least one group of training samples, wherein the pre-estimation model is used for outputting the guiding probability of the sample object according to the characteristics of the sample object, and the guiding probability is used for indicating the probability that the sample object guides other objects to register in the business system.

After the at least one set of training samples is obtained, the estimation model can be trained according to the at least one set of training samples.

Based on the above description, it can be determined that the predictive model in this embodiment may process the sample object features, so as to output a guidance probability of the sample object, where the guidance probability is used to indicate a probability that the sample object guides other objects to register in the business system. It should be noted that other objects may or may not be registered in the business system, and this embodiment is not limited to this, as long as the other objects are registered in the sample object guidance.

For example, there is currently an object a, it is assumed that the object a makes a service application, and it is assumed that the object a guides an object B to also make a service application, where the object B may have made a service application before, but does not prevent the object a from making a service application again under the guidance of the object a. Therefore, in this embodiment, it is not concerned whether other objects are registered in the business system, but only whether the sample object guides other objects to register the business system.

Specifically, when the pre-estimation model is trained according to at least one group of training samples, the sample object characteristics of the sample object are actually processed through the pre-estimation model, then the sample guide probability is output, and the pre-estimation model is fed back according to the sample guide probability and the label information corresponding to the sample object, so that the training of the pre-estimation model is realized.

S203, testing the trained pre-estimated model, and determining the test result parameters of the pre-estimated model.

After training of the pre-estimation model according to at least one group of training samples is finished, the trained pre-estimation model can be obtained. However, the training of the pre-estimated model is only finished according to the determined training sample at this time, and the actual effect of the pre-estimated model after training cannot be guaranteed. Therefore, in this embodiment, the trained pre-estimated model is tested to determine the test result parameters of the pre-estimated model.

Wherein, the test result parameter can indicate the correctness and the referential of the guide probability output by the pre-estimation model.

And S204, determining the pre-estimated model after training according to the test result parameters.

After the test result parameters are obtained, the trained pre-estimated model can be determined according to the test result parameters.

In one possible implementation, the test result parameter may be compared to a preset threshold, for example. If the test result parameter is determined to be greater than or equal to the preset threshold, the output correctness of the trained pre-estimated model can be ensured, and therefore the trained pre-estimated model can be determined to be the trained pre-estimated model.

Or, if it is determined that the test result parameter is smaller than the preset threshold, it may be determined that the output correctness of the trained estimation model cannot be guaranteed, so the steps of obtaining the training sample and training the estimation model according to the training sample may be repeatedly performed until the test result parameter of the estimation model is greater than or equal to the preset threshold, and it may be determined that the estimation model after training is obtained.

The model training method provided by the embodiment of the invention comprises the following steps: at least one group of training samples is obtained, wherein any group of training samples comprises sample object characteristics of sample objects. And training the pre-estimation model according to at least one group of training samples, wherein the pre-estimation model is used for outputting the guiding probability of the sample object according to the characteristics of the sample object, and the guiding probability is used for indicating the probability that the sample object guides other objects to register in the business system. And testing the trained pre-estimation model, and determining the test result parameters of the pre-estimation model. And determining the trained pre-estimation model according to the test result parameters. The estimation model is further tested after the training is finished by obtaining the training sample and training the estimation model according to the training sample, and the estimation model after the training is finished is determined to be obtained when the test of the estimation model is determined to pass. The pre-estimation model can output the guiding probability of the object according to the object characteristics, and the guiding probability can indicate the probability that the registered object guides other objects to register in the business system. Therefore, the pre-estimation model can be trained to screen out target objects which are most likely to guide other objects to register, and targeted information push is performed on the target objects, so that the effectiveness and the targeting of information push can be effectively improved.

Based on the above description, the model training method provided by the present invention is further described in detail below with reference to fig. 3 to 6. Fig. 3 is a second flowchart of the model training method provided in the embodiment of the present invention, fig. 4 is a schematic diagram of implementing a training prediction model provided in the embodiment of the present invention, fig. 5 is a schematic diagram of implementing determining a first target object provided in the embodiment of the present invention, and fig. 6 is a schematic diagram of implementing determining a second target object provided in the embodiment of the present invention.

As shown in fig. 3, the method includes:

s301, at least one group of training samples is obtained, wherein any group of training samples comprises sample object characteristics of sample objects.

The implementation manner of S301 is similar to that of S201, and the specific implementation may refer to the above description, which is not described herein again.

In this embodiment, the training samples may further include: a exemplar label for the exemplar object that indicates whether the exemplar object directs the remaining objects to register with the business system.

Based on the above description, it can be determined that the sample object in the present embodiment is an object that has completed registration in the business system, that is, an object that has completed a business application. Meanwhile, in the related art, recommendation information is pushed for the object which completes the service application, so that the part of the objects is encouraged to guide the rest of the objects to carry out the service application.

Therefore, in a possible implementation manner, the sample object in this embodiment may be, for example, an object that is registered in the business system and to which the recommendation information is pushed, that is, in this embodiment, the object that receives the recommendation information is taken as the overall sample set. It can be understood that, after receiving the recommendation information, the sample object that has completed the service application may or may not guide the remaining objects to register in the service system.

Then for each sample object, if it guides the rest of the objects to register in the business system after receiving the push information, its sample label (label) may be 1, for example, to indicate that the sample object guides the rest of the objects to register in the business system.

Or, for each sample object, if it does not guide the rest of the objects to register in the business system after receiving the push information, its sample label (label) may be 0, for example, to indicate that the sample object does not guide the rest of the objects to register in the business system.

And S302, processing the characteristics of the sample object according to the pre-estimation model to obtain the sample guide probability corresponding to the sample object.

After the training samples are determined, the predictive model can be trained according to the training samples. It should be noted that, because there is at least one set of training samples in the present embodiment, the estimation model performs a corresponding training process for any set of training samples, and the implementation manner for each set of training samples is similar, the following description will take any set of training samples as an example.

In a possible implementation manner, for example, the sample object features in the training sample may be processed according to the pre-estimation model, so as to obtain the sample guiding probability corresponding to the sample object.

As can be appreciated with reference to fig. 4, for example, sample object features in a training sample are input into a predictive model such that the predictive model outputs sample leading probabilities for the sample objects.

In a possible implementation manner, the pre-estimation Model in this embodiment may be, for example, a Light Gradient Boosting Model (Light Gradient Boosting Model), or in an actual implementation process, other models may also be selected, and the embodiment does not limit the internal implementation of the pre-estimation Model, as long as it can obtain the guidance probability according to the object feature.

S303, determining a prediction label according to the sample guide probability, wherein the prediction label is used for indicating whether the sample object guides the rest objects to be registered in the business system.

The sample guide probability is used for the probability that the currently calculated sample object guides the rest objects to register in the business system. After determining the sample leading probability of the sample object in this embodiment, referring to fig. 4, a prediction tag may be determined according to the sample leading probability, where the prediction tag may be used to indicate whether the sample object leads the rest of the objects to be registered in the business system.

In one possible implementation, the sample leading probability may be compared with a preset probability, where the preset probability may be understood as a preset probability threshold.

If the sample guidance probability is determined to be greater than or equal to the preset probability, it is determined that the current sample object has a high possibility of guiding the rest of the objects to register in the service system, so that the prediction label may be determined to be 1, for example, to indicate that the sample object will guide the rest of the objects to register in the service system.

Or, if it is determined that the sample guidance probability is smaller than the preset probability, it is considered that the current sample object has a low possibility of guiding the remaining objects to register in the business system, and therefore, for example, the prediction label may be determined to be 0, which is used to indicate that the sample object does not guide the remaining objects to register in the business system.

And S304, updating the model parameters of the pre-estimated model according to the sample label and the prediction label.

After the prediction tag is determined, the model parameters of the pre-estimated model may be updated according to the sample tag and the prediction tag.

The updating of the model parameters of the pre-estimated model may be implemented, for example, as follows: and determining a loss function value according to the sample label and the prediction label. And updating the model parameters of the pre-estimated model according to the loss function values.

It will be appreciated that the exemplar label and the predictive label are similar, except that the exemplar label labels whether the actual exemplar object directs the rest of the objects to be registered with the business system, and the predictive label labels whether the predicted exemplar object directs the rest of the objects to be registered with the business system. That is, one is what actually happens and the other is what the predictive model predicts.

In the process of training the prediction model, in order to improve the accuracy of the output result of the prediction model, it is required to ensure that the prediction label is as close to the sample label as possible. Thus, referring to fig. 4, in this embodiment, the loss function value may be determined based on the sample label and the prediction label.

In one possible implementation, for example, a loss function may be provided, and then the sample label and the prediction label may be input to the loss function, so as to obtain a loss function value. Wherein the loss function value actually characterizes the difference between the prediction tag and the sample tag.

Referring to fig. 4, after obtaining the loss function value, the model parameters of the prediction model may be updated according to the loss function value, so as to optimize the prediction model.

The above description is directed to a complete training process of the pre-estimation model, and it can be understood that the above training process is performed for each training sample. For example, the training process described above may be sequentially performed on each training sample, and after the model parameters of the pre-estimated model are updated according to the completion of training of one training sample each time. The training process of the next training sample can be continued on the basis of the estimation model after the model parameters are updated until the training is completed according to a plurality of training samples, so that the estimation model after the training can be obtained.

S305, obtaining a plurality of test objects in a preset database, wherein the test objects are objects registered in the business system.

After the trained pre-estimation model is obtained, in order to ensure the accuracy and the effectiveness of the output result of the pre-estimation model, the pre-estimation model needs to be further tested, and whether the training of the pre-estimation model is finished can be determined.

When testing the predictive model, for example, a plurality of test objects may be obtained from a preset database, where the test objects may be objects registered in the business system.

It will be appreciated that the test object is similar to the sample object described above, but functions differently. Wherein the sample object is used for training the predictive model, and the test object is used for testing the predictive model.

The test object may be, for example, an object that has been registered in the business system and to which recommendation information is pushed, similar to the sample object described above. Correspondingly, the test object also has a corresponding test tag, for example, when the test tag is 1, the test object may be instructed to guide the rest of the objects to register in the service system, and when the test tag is 0, the test object may be instructed not to guide the rest of the objects to register in the service system.

S306, according to the sample guide probability output by the pre-estimation model, determining K sample objects with the sample guide probability ranked in the front as first target objects, wherein K is an integer greater than or equal to 1.

For testing the predictive model, in this embodiment, for example, a first target object may be determined according to a sample guidance probability output by the predictive model, where the first target object is an object predicted to guide the remaining objects to register in the service system according to the output of the predictive model.

In one possible implementation manner, for example, the sample guidance probabilities may be sorted, and then K sample objects with the sample guidance probabilities being higher than each other are determined as the first target object, where K is an integer greater than or equal to 1.

For example, as can be understood with reference to fig. 5, assuming that there are currently 10 sample objects in total, the sample leading probabilities for the 10 sample objects are currently sorted, for example, the sorting result shown in fig. 5 can be obtained. Meanwhile, assuming that K in the current example is 5, according to the sorting result shown in fig. 5, for example, the sample object 2, the sample object 7, the sample object 5, the sample object 6, and the sample object 4 that are sorted first 5 may be determined as the first target object.

Or in another possible implementation there may also be a recall process, for example, before the sample steering probabilities are sorted. For example, the final capacity may be used as a measure Target, a sample object with a Target Group Index (TGI) greater than or equal to a first threshold is selected, and the application parameter (for example, credit line) greater than or equal to a second threshold is determined as a candidate object.

And then, sequencing according to the sample guide probability of each object to be selected, and determining K sample objects with the sample guide probability being earlier as first target objects, wherein K is an integer greater than or equal to 1. The implementation is similar to the above described examples and will not be illustrated here.

S307, in the test objects, N randomly selected test objects are determined as second target objects, wherein N is an integer greater than or equal to 1.

In order to test the pre-estimation model, in this embodiment, for example, N test objects may be randomly selected from the test objects, and the randomly selected N test objects are determined as second target objects, where the second target objects are objects that are currently aimed at the test objects and that are determined to guide the remaining objects to be registered in the service system.

For example, as can be understood by referring to fig. 6, it is assumed that there are currently 10 test objects in total, and meanwhile, N in the current example is 5, and currently, randomly selecting 5 test objects from the 10 test objects are: test object 1, test object 3, test object 4, test object 7, and test object 10. These 5 test objects selected randomly may be determined as the second target object.

S308, acquiring a first guiding rate corresponding to the first target object, wherein the first guiding rate is a ratio of the number of the first class of objects to the number of the first target object, and the first class of objects are objects which actually guide other objects to be registered in the business system in the first target object.

Based on the above description, it can be determined that the sample object in the present embodiment has the sample label, and therefore, it can be determined whether it actually guides the rest of the objects to be registered in the business system. And the first target object which guides other objects to be registered in the service system under the prediction condition is determined according to the sample guide probability output by the pre-estimation model.

The first target object is, actually, K sample objects with the highest possibility of occurrence of the guiding behavior according to the guiding probability output by the pre-estimation model, and whether the actual guiding behavior occurs or not can be determined for each sample object, for example, the first guiding rate corresponding to the first target object can be determined.

In a possible implementation manner, the first guiding rate is a ratio of the number of the first type objects to the number of the first target objects, where the first type objects are objects that actually guide other objects registered in the business system in the first target objects.

As may be appreciated, for example, with reference to fig. 5, continuing the above example, the currently determined first target object includes: sample object 2, sample object 7, sample object 5, sample object 6, sample object 4, also in connection with the example of fig. 5, wherein the sample label of sample object 2 is 0, for indicating that sample object 2 does not guide the rest of the objects to be registered in the business system; the sample label of the sample object 7 is 1, which is used to instruct the sample object 7 to guide the rest objects to register in the business system; wherein, the sample label of the sample object 5 is 1, which is used to indicate the sample object 5 to guide the rest objects to register in the service system; the sample label of the sample object 6 is 0, which is used to indicate that the sample object 6 does not guide the rest objects to register in the service system; the sample label of the sample object 4 is 1, indicating that the sample object 4 guides the rest of the objects to register with the business system.

Then, based on the current example, it may be determined that the first class of objects that actually guide other objects in the first target object to register with the business system includes: sample object 7, sample object 5 and sample object 1. Then the number of the first type objects is 3, and the total number of the first target objects is 5, from which it can be determined that the first guidance rate corresponding to the first target object is 0.6.

The first guidance rate is thus actually a ratio of the number of objects in the first target object for which guidance behavior exists to the total number of first target objects. The first index ratio here reflects the correctness of the determined first target object.

S309, a second guiding rate corresponding to the second target object is obtained, wherein the second guiding rate is a ratio of the number of the second class of objects to the number of the second target object, and the second class of objects are objects which actually guide other objects to be registered in the service system in the second target object.

And similar to the above description, in this embodiment, for a second target object determined according to the test object, a second index rate corresponding to the second target object may also be determined.

In a possible implementation manner, the second guiding rate is a ratio of a number of second class objects and a number of second target objects, where the second class objects are objects that actually guide other objects to be registered in the business system.

As may be appreciated, for example, with reference to fig. 6, continuing the above example, the currently determined second target object includes: the test objects 1, 3, 4, 7, and 10, in combination with the example of fig. 6, where a sample label of the test object 1 is 0, which is used to indicate that the test object 1 does not guide the rest of the objects to register in the service system; the sample label of the test object 3 is 0, which is used for indicating that the test object 3 does not guide other objects to register in the service system; wherein the sample label of the test object 4 is 0, which is used to indicate that the test object 4 does not guide other objects to register in the service system; the sample label of the test object 7 is 1, which is used for indicating the test object 7 to guide the rest objects to register in the business system; the sample label of the test object 10 is 0, which indicates that the test object 10 does not guide the rest of the objects to register in the business system.

Then, based on the current example, it may be determined that the second class of objects in the second target object that actually guides other objects to register with the business system includes: the sample object 7. Then the number of the second class objects is 1, and the total number of the second target objects is 5, from which it can be determined that the second lead rate corresponding to the second target object is 0.2.

The second guidance rate is thus actually a ratio of the number of objects in the second target object for which guidance behavior exists to the total number of second target objects. The second index rate here may reflect the correctness of the determined second target object.

S310, determining the ratio of the first guiding rate to the second guiding rate as a test result parameter corresponding to the pre-estimated model.

As the first guiding rate of the first target object is determined, it can be understood that the first guiding rate reflects a sample guiding probability output according to the pre-estimated model, and the pre-estimated accuracy of the guiding action of the first target object is obtained.

At the same time, the second guiding rate of the second target object is determined, and it can be understood that the second guiding rate reflects the correctness of the guiding action performed by the randomly selected second target object in the test sample.

Currently, in order to determine, the selection is performed according to the sample guiding probability output by the pre-estimated model, and compared with the implementation of random selection, whether the accuracy can be guaranteed or not is determined. For example, a ratio of the first conductivity ratio and the second conductivity ratio may be made, and the ratio of the first conductivity ratio and the second conductivity ratio may be determined as a test result parameter corresponding to the estimation model. The test result parameter here can also be understood as the lifting ratio of the introduction rate.

For example, in the example of fig. 5 and 6, it is determined that the first conductivity ratio corresponding to the first target object is 0.6, and the second conductivity ratio corresponding to the second target object is 0.2, and then the ratio of the first conductivity ratio to the second conductivity ratio is 3, so that the test result parameter may be determined to be 3.

The test result parameters can indicate that object selection is carried out according to the sample guide probability output by the pre-estimation model, and compared with the gain for randomly carrying out object selection, the accuracy and the effectiveness for carrying out object selection according to the sample guide probability output by the pre-estimation model are reflected to a certain extent.

S311, determining the trained pre-estimation model according to the test result parameters.

After determining the test result parameters, the test result parameters may be compared to a preset threshold, for example, to determine a trained predictive model.

In a possible implementation manner, if it is determined that the test result parameter is greater than or equal to the preset threshold, it may be determined that the accuracy of the current estimation model output may be ensured, and therefore, the estimation model trained according to the at least one sample object may be determined as the estimation model after training.

In another possible implementation manner, if it is determined that the test result parameter is smaller than the preset threshold, it may be determined that the accuracy of the current estimation model output cannot be guaranteed, and therefore, the steps of obtaining the training sample and training the estimation model according to the training sample need to be repeatedly performed until the test result parameter of the estimation model is greater than or equal to the preset threshold.

For example, continuing the above-described example, it is determined that the test result parameter of the predictive model is 3, and assuming that the currently set preset threshold is 1.3 (the corresponding gain reaches 30%), it may be determined that the offline test of the predictive model passes for the above-described example, and then the trained predictive model may be determined as the trained predictive model.

According to the model training method provided by the embodiment of the invention, the sample object characteristics of the sample object are processed through the pre-estimation model to output the sample guide probability of the sample object, then the prediction label of the sample object is determined according to the sample guide probability, and the model parameters of the pre-estimation model are updated according to the prediction label and the sample label of the sample object, so that the updating of the pre-estimation model can be effectively realized, the distance between the prediction label and the sample label is shortened, the correctness of the sample guide probability output by the pre-estimation model is improved, and the training of the pre-estimation model is further effectively realized.

And after training of the pre-estimation model according to the sample object is finished, testing the pre-estimation model according to the test sample, and in the concrete implementation of the test, firstly, according to the sample guide probability output by the pre-estimation model, determining a first target object which has a higher possibility of guiding other objects to perform service system registration, and according to an object which actually guides other objects to perform service system registration in the first target object, determining a first guide rate corresponding to the first target object. And determining a plurality of randomly selected objects in the test objects as second target objects, determining a second guiding rate corresponding to the second target objects according to the objects which actually guide the rest objects to register in the service system in the second target objects, and determining a ratio of the first guiding rate to the second guiding rate as a test result parameter corresponding to the pre-estimation model, so that the gain of the target objects determined according to the pre-estimation model can be effectively quantified. And then comparing the test result parameters with a preset threshold, and determining to obtain the trained pre-estimated model when the test result parameters are determined to be greater than or equal to the preset threshold, so that the correctness and the effectiveness of the guide probability output by the trained pre-estimated model can be effectively ensured.

The above embodiments describe the implementation of training for the predictive model, and the application of the predictive model is described below with reference to specific embodiments.

First, description is made with reference to fig. 7, and fig. 7 is a flowchart of an information processing method according to an embodiment of the present invention.

As shown in fig. 7, the method includes:

s701, acquiring object characteristics of at least one first object in a preset database, wherein the first object is an object registered in a service system.

In this embodiment, the preset database may store related information of a plurality of first objects, where the first objects are objects that have been registered in the business system, and here, the first objects are registered in the business system, for example, it may be understood that a business application is made.

The object characteristics of at least one first object may be obtained in the preset database, where the object characteristics are similar to the sample object characteristics described above, and may include, for example, the first operation characteristics, the second operation characteristics, the personalized characteristics, and so on of the sample object.

In a scenario of an actual loan application, the first operation feature may be, for example, a withdrawal feature, the second operation feature may be a repayment feature, the personalization feature may be, for example, a feature related to the enterprise itself, and the like.

S702, processing the characteristics of each object through the pre-estimation model to obtain the guiding probability corresponding to each first object, wherein the guiding probability is used for indicating the probability that the first object guides the second object to register in the service system.

After determining the object features of the first object, the object features of the first object may be processed through a pre-estimation model, so as to obtain the guidance probability corresponding to the first object, where the pre-estimation model is the pre-estimation model trained in the foregoing embodiment. This is performed for each first object, so that a guiding probability corresponding to each first object can be obtained, wherein the guiding probability is used for indicating the probability that the first object guides the second object to register in the service system.

It should be noted that, the second object may be registered in the business system, or may not be registered in the business system, and this embodiment is not limited to this, as long as the second object is an object different from the first object.

For example, there is currently a first object a, which is assumed to have made a service application, and which is assumed to direct a second object B to make a service application, where the second object B may have made a service application before, but is not prevented from making a service application again currently under the direction of the first object a. Therefore, in this embodiment, it is not concerned whether the second object is registered in the service system, as long as the second object is an object different from the first object.

S703, determining a target object in at least one first object according to the guiding probability corresponding to each first object and each first object, and sending recommendation information to the equipment corresponding to the target object.

It may be determined that the guiding probability may indicate a possibility that the first object guides the second object to register in the service system, and after determining the guiding probabilities corresponding to the first objects, the target object may be determined in the at least one first object according to the guiding probabilities corresponding to the first objects and the first objects.

The target object is determined, and there is a high possibility that the target object will guide the second object to register in the service system, and further, the target object is an object that needs to push recommendation information. Therefore, in this embodiment, after the target object is determined, recommendation information may be sent to a device corresponding to the target object, so as to encourage the target object to guide the second object to register in the service system.

The information processing method provided by the embodiment of the invention comprises the following steps: and acquiring the object characteristics of at least one first object in a preset database, wherein the first object is an object registered in a business system. And processing the characteristics of each object through the pre-estimation model to obtain the guiding probability corresponding to each first object, wherein the guiding probability is used for indicating the probability that the first object guides the second object to register in the service system. And determining a target object in at least one first object according to the guiding probability corresponding to each first object and each first object, and sending recommendation information to equipment corresponding to the target object. The object characteristics of the first object which is registered in the business system are obtained, and the trained pre-estimation model is used for processing the object characteristics of the first object, so that the guiding probability of the first object is obtained, wherein the guiding probability can indicate the probability that the first object guides the second object to be registered in the business system, and then the target object can be determined in the plurality of first objects according to the guiding probability of the first object and the first object, wherein the target object is the object which is determined in the embodiment and needs to be used for sending the recommendation information in a targeted manner, and the target object is the object which is determined at present and most possibly guides other objects to carry out business application, so that the pertinence and the effectiveness of recommendation information pushing can be effectively improved through the method.

Based on the content described in the above embodiments, the information processing method provided by the present invention is further described in detail below with reference to fig. 8 to 9. Fig. 8 is a second flowchart of the information processing method according to the embodiment of the present invention, and fig. 9 is a schematic diagram of implementation of determining a target object according to the embodiment of the present invention.

As shown in fig. 8, the method includes:

s801, obtaining object characteristics of at least one first object in a preset database, wherein the first object is an object registered in a business system.

The implementation manner of S801 is similar to that of S701 described above, and is not described herein again.

S802, processing the characteristics of each object through the pre-estimation model to obtain the guiding probability corresponding to each first object, wherein the guiding probability is used for indicating the probability that the first object guides the second object to register in the service system.

The implementation manner of S802 is similar to that of S702, and is not described herein again.

And S803, acquiring the target group index corresponding to each first object in a preset database, and acquiring the application parameter corresponding to the service application of each first object.

In this embodiment, since the guiding probability may indicate a possibility that the first object guides the second object to register in the service system, after determining the guiding probability, the target object may be determined according to the ranking of the guiding probabilities, for example.

However, in this embodiment, before the ranking according to the guiding probability, there may be a recall process, for example, to perform a certain degree of filtering on the first object.

In a possible implementation manner, for example, the target group index TGI corresponding to each first object may be obtained from a preset database, and the application parameter corresponding to the service application of each first object may be obtained.

For example, the current service scenario is the scenario of loan application described above, and the application parameters corresponding to the number of service applications may be, for example, the credit line granted by the first object.

S804, determining the first object with the target population index being larger than or equal to the first threshold and the application parameter being larger than or equal to the second threshold as the object to be selected.

After determining the target population index and the application parameter for the first object, the first object may be filtered, for example, based on the target population index and the application parameter.

In a possible implementation manner, for example, the target group index is greater than or equal to a first threshold, and the first object whose application parameter is greater than or equal to a second threshold is determined as the object to be selected, and then the objects to be selected are sorted according to the guiding probability of the object to be selected.

That is, the first object with low TGI and low application parameter is screened out with the final capacity as the target, so as to determine the candidate object. Through setting the current screening process, the effectiveness of the final pushed recommendation information can be further improved, and the quality of the first object guiding and registering second object can be improved in a more popular way.

For example, it can be understood in conjunction with fig. 9 that it is assumed that 8 first objects, i.e., the first object 1 to the first object 8 shown in fig. 9, currently exist, and that the first object 1, the first object 3, and the first object 5 are screened out according to the above-described screening conditions. Then obtaining the candidate object includes: first object 2, first object 4, first object 6, first object 7, first object 8.

And S805, sequencing the guiding probabilities corresponding to the objects to be selected.

After determining each object to be selected, the guidance probabilities corresponding to each object to be selected may be sorted.

For example, as can be understood by referring to fig. 9, currently, according to the guiding probability corresponding to the candidate object, the remaining 5 candidate objects are sorted, for example, the order sequentially is: a first object 6, a first object 2, a first object 8, a first object 4, a first object 7.

S806, determining the K objects to be selected with the guide probability ranked in the front as target objects.

After the guidance probabilities corresponding to the respective candidate objects are ranked, for example, K candidate objects ranked in the top may be determined as target objects, where K may be an integer greater than or equal to 1.

Also introduced in conjunction with fig. 9, assuming that K in the present example is 3, in the example of fig. 9, the target objects may be determined to be the first object 6, the first object 2, and the first object 8.

And S807, sending recommendation information to the equipment corresponding to the target object.

After the target object is determined, recommendation information may be sent to a device corresponding to the target object, and specific content of the recommendation information may be selected and set according to an actual requirement, which is not limited in this embodiment.

According to the information processing method provided by the embodiment of the invention, the guidance probability of each first object is determined through the pre-estimation model, the target group index and the application parameter of each first object are obtained, the first objects are screened according to the target group index and the application parameter to obtain the objects to be selected, then the objects to be selected are sequenced according to the guidance probability of the objects to be selected, and a plurality of guidance objects which are sequenced in the front are determined as the target objects, so that the target objects which guide the second object to carry out service application and have the highest possibility can be effectively screened out from the plurality of first objects which carry out service application, and then recommendation information is sent to the target objects, so that the pertinence and the effectiveness of information pushing can be effectively improved, and the maximum return rate can be obtained as far as possible. Meanwhile, when the target object is determined, a screening process is also performed, wherein the screening process can take the maximum capacity as a target, the first object with low TGI and low application parameters is screened out, and only the objects with high TGI and high application parameters are reserved, so that the return rate and the capacity for pushing the recommendation information to the target equipment can be further improved.

Based on the above introduction, it can be determined that, in the training process of the prediction model, the prediction model is subjected to an offline test, wherein the offline test can ensure that the validity and the accuracy of an output result of the prediction model are evaluated before the prediction model is put into use. Further, after the pre-estimation model is put into use, for example, an online test can be performed on the pre-estimation model to determine the effect of determining the target device according to the pre-estimation model.

For example, the implementation of online testing may be understood with reference to fig. 10, where fig. 10 is a schematic diagram illustrating the implementation of online testing of a predictive model according to an embodiment of the present invention.

As shown in fig. 10, it is assumed that a plurality of test subjects currently exist, and here, the test subjects are all registered in the business system.

In one possible implementation, for example, the plurality of objects participating in the test may be divided into a first group a and a second group B, where the first group a may include, for example, at least one first object described above, and the second group B may include, for example, objects other than the first object among the plurality of objects participating in the test, and for convenience of description below, the objects in the second group B are referred to as a third object herein.

Then, for the first object in the first group a, a plurality of target objects are determined according to the processing method based on the prediction model described above.

And randomly selecting a plurality of third objects from the second group B as a plurality of target objects according to a randomly selected implementation mode aiming at the third objects in the second group B.

And then sending recommendation information to the first group of corresponding target objects and the second group of corresponding target objects. Then, after a period of time, the first set of corresponding introduction rates and the second set of corresponding introduction rates may be counted, wherein the introduction rates are implemented in a manner similar to that described in the above embodiments.

Specifically, the introduction rate corresponding to the first group is a ratio of the number of objects guiding the other objects to be registered in the plurality of target objects corresponding to the first group to the number of objects of the plurality of target objects corresponding to the first group. The introduction rate corresponding to the second group is a ratio of the number of objects guiding other objects to be registered in the plurality of target objects corresponding to the second group to the number of objects of the plurality of target objects corresponding to the second group.

And then comparing the introduction rate of the first group A with the introduction rate of the second group B, and if the introduction rate of the first group A is determined to be greater than that of the second group B, determining that the method provided by the invention can effectively improve the introduction rate of the object by a mode of sending recommendation information in a targeted manner.

Fig. 11 is a schematic structural diagram of a model training apparatus according to an embodiment of the present invention. As shown in fig. 11, the apparatus 110 includes: an acquisition module 1101, a training module 1102, a testing module 1103, and a determination module 1104.

An obtaining module 1101, configured to obtain at least one set of training samples, where any set of training samples includes sample object features of a sample object;

a training module 1102, configured to train the pre-estimation model according to the at least one set of training samples, where the pre-estimation model is configured to output a guiding probability of the sample object according to the sample object feature, and the guiding probability is used to indicate a probability that the sample object guides other objects to register in the business system;

the testing module 1103 is configured to test the trained pre-estimated model, and determine a test result parameter of the pre-estimated model;

and the determining module 1104 is configured to determine the trained pre-estimation model according to the test result parameter.

In one possible design, the training samples further include: a sample label of the sample object, the sample label indicating whether the sample object directs the remaining objects to register with the business system;

the training module is specifically configured to:

processing the characteristics of the sample object according to the pre-estimation model to obtain a sample guide probability corresponding to the sample object;

determining a prediction label according to the sample guide probability, wherein the prediction label is used for indicating whether the sample object guides the rest objects to be registered in the business system;

and updating the model parameters of the pre-estimation model according to the sample label and the prediction label.

In one possible design, the training module 1102 is specifically configured to:

if the sample guiding probability is greater than or equal to a preset probability, determining that the prediction tag is used for indicating that the sample object can guide other objects to register in the service system; alternatively, the first and second electrodes may be,

and if the sample guide probability is smaller than the preset probability, determining that the prediction label is used for indicating that the sample object does not guide other objects to be registered in the service system.

In one possible design, the test module 1103 is specifically configured to:

acquiring a plurality of test objects in a preset database, wherein the test objects are objects registered in a business system;

determining K sample objects with sample guide probability sequencing at the front as first target objects according to the sample guide probability output by the pre-estimation model, wherein K is an integer greater than or equal to 1;

determining N randomly selected test objects as second target objects in the test objects, wherein N is an integer greater than or equal to 1;

and determining a test result parameter corresponding to the pre-estimation model according to the first target object and the second target object.

In one possible design, the test module 1103 is specifically configured to:

acquiring a first guiding rate corresponding to the first target object, wherein the first guiding rate is a ratio of the number of first class objects to the number of the first target objects, and the first class objects are objects which actually guide other objects to be registered in the service system in the first target objects;

acquiring a second guiding rate corresponding to the second target object, wherein the second guiding rate is a ratio of the number of second class objects to the number of the second target objects, and the second class objects are objects which actually guide other objects to be registered in the service system in the second target objects;

and determining the ratio of the first guiding rate to the second guiding rate as a test result parameter corresponding to the pre-estimated model.

In one possible design, the determining module 1104 is specifically configured to:

if the test result parameter is greater than or equal to a preset threshold value, determining to obtain a trained pre-estimated model; alternatively, the first and second liquid crystal display panels may be,

if the test result parameter is smaller than the preset threshold value, the steps of obtaining a training sample and training the estimation model according to the training sample are repeatedly executed until the test result parameter of the estimation model is larger than or equal to the preset threshold value.

The apparatus provided in this embodiment may be used to implement the technical solutions of the above method embodiments, and the implementation principles and technical effects are similar, which are not described herein again.

Fig. 12 is a schematic structural diagram of an information processing apparatus according to an embodiment of the present invention. As shown in fig. 12, the apparatus 120 includes: an obtaining module 1201, a processing module 1202, and a determining module 1203.

An obtaining module 1201, configured to obtain an object feature of at least one first object in the preset database, where the first object is an object registered in a service system;

a processing module 1202, configured to process the object features through a pre-estimation model to obtain a guiding probability corresponding to each first object, where the guiding probability is used to indicate a probability that the first object guides a second object to register in the service system;

a determining module 1203, configured to determine, according to the guidance probability corresponding to each first object and each first object, a target object in the at least one first object, and send recommendation information to a device corresponding to the target object.

In one possible design, the determining module 1203 is specifically configured to:

acquiring a target group index corresponding to each first object in the preset database, and acquiring an application parameter corresponding to a service application of each first object;

determining the first object with the target population index being greater than or equal to a first threshold value and the application parameter being greater than or equal to a second threshold value as the object to be selected;

and determining a target object according to the guiding probability corresponding to each object to be selected.

sequencing the guiding probability corresponding to each object to be selected;

and determining the K objects to be selected with the leading probability ranked in the top as the target objects.

Fig. 13 is a schematic diagram of a hardware structure of a model training device according to an embodiment of the present invention, and as shown in fig. 13, a model training device 130 according to the embodiment includes: a processor 1301 and a memory 1302; wherein

A memory 1302 for storing computer-executable instructions;

the processor 1301 is configured to execute the computer executable instructions stored in the memory to implement the steps performed by the model training method in the above embodiments. Reference may be made in particular to the description relating to the method embodiments described above.

Alternatively, the memory 1302 may be separate or integrated with the processor 1301.

When the memory 1302 is separately provided, the model training apparatus further includes a bus 1303 for connecting the memory 1302 and the processor 1301.

Fig. 14 is a schematic diagram of a hardware structure of an information processing apparatus according to an embodiment of the present invention, and as shown in fig. 14, an information processing apparatus 140 according to the embodiment includes: a processor 1401 and a memory 1402; wherein

A memory 1402 for storing computer-executable instructions;

a processor 1401 for executing the computer execution instructions stored in the memory to implement the steps performed by the information processing method in the above embodiments. Reference may be made in particular to the description relating to the method embodiments described above.

Alternatively, the memory 1402 may be separate or integrated with the processor 1401.

When the memory 1402 is provided independently, the information processing apparatus further includes a bus 1403 for connecting the memory 1402 and the processor 1401.

An embodiment of the present invention further provides a computer-readable storage medium, where a computer execution instruction is stored in the computer-readable storage medium, and when a processor executes the computer execution instruction, the model training method executed by the above model training apparatus is implemented, or the information processing method executed by the above information processing apparatus is implemented.

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

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.

The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims

1. A method of model training, comprising:

training a pre-estimation model according to the at least one group of training samples, wherein the pre-estimation model is used for outputting a guiding probability of the sample object according to the characteristics of the sample object, and the guiding probability is used for indicating the probability that the sample object guides other objects to register in a business system;

2. The method of claim 1, wherein the training samples further comprise: a sample label of the sample object, the sample label indicating whether the sample object directs the remaining objects to register with the business system;

the training the pre-estimation model according to the at least one group of training samples comprises:

determining a prediction label according to the sample guiding probability, wherein the prediction label is used for indicating whether the sample object guides other objects to register in the business system;

and updating the model parameters of the pre-estimated model according to the sample label and the prediction label.

3. The method of claim 2, wherein determining a predictive label based on the sample steering probability comprises:

4. The method according to any one of claims 1 to 3, wherein the testing the trained predictive model to determine the test result parameters of the predictive model comprises:

5. The method according to claim 4, wherein the determining, according to the first target object and the second target object, the test result parameters corresponding to the pre-estimation model includes:

6. The method according to any one of claims 1 to 5, wherein the determining the trained predictive model according to the test result parameters comprises:

if the test result parameter is greater than or equal to a preset threshold value, determining to obtain a trained pre-estimated model; alternatively, the first and second electrodes may be,

7. An information processing method characterized by comprising:

acquiring object characteristics of at least one first object in a preset database, wherein the first object is an object registered in a service system;

8. The method according to claim 7, wherein the determining a target object in the at least one first object according to the predicted value corresponding to each first object and each first object comprises:

determining the first object with the target population index being greater than or equal to a first threshold value and the application parameter being greater than or equal to a second threshold value as an object to be selected;

9. The method of claim 8, wherein determining the target object according to the guiding probability corresponding to each candidate object comprises:

sequencing the guiding probability corresponding to each object to be selected;

10. A model training apparatus, comprising:

a training module, configured to train a predictive model according to the at least one set of training samples, where the predictive model is configured to output a guiding probability of the sample object according to the sample object feature, and the guiding probability is used to indicate a probability that the sample object guides other objects to register in a business system;

the testing module is used for testing the trained pre-estimated model and determining a testing result parameter of the pre-estimated model;

and the determining module is used for determining the trained pre-estimation model according to the test result parameters.

11. An information processing apparatus characterized by comprising:

the system comprises an acquisition module, a storage module and a processing module, wherein the acquisition module is used for acquiring the object characteristics of at least one first object in a preset database, and the first object is an object registered in a service system;

the processing module is used for processing the characteristics of each object through a pre-estimation model to obtain a guiding probability corresponding to each first object, wherein the guiding probability is used for indicating the probability that the first object guides a second object to register in the service system;

and the determining module is used for determining a target object in the at least one first object according to the guiding probability corresponding to each first object and each first object, and sending recommendation information to equipment corresponding to the target object.

12. A model training apparatus, characterized in that the model training apparatus comprises: memory, a processor and a model training program stored on the memory and executable on the processor, the model training program when executed by the processor implementing the steps of the model training method of any one of claims 1 to 6.

13. An information processing apparatus characterized by comprising: memory, processor and information processing program stored on the memory and executable on the processor, which when executed by the processor implements the steps of the information processing method according to any one of claims 7 to 9.

14. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a model training program which, when executed by a processor, implements the steps of the model training method according to any one of claims 1 to 6; and (c) a second step of,

the computer-readable storage medium has stored thereon an information processing program which, when executed by a processor, realizes the steps of the information processing method according to any one of claims 7 to 9.

15. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 6 or 7 to 9.