CN113592622A

CN113592622A - Credit data optimization method, device, equipment and medium

Info

Publication number: CN113592622A
Application number: CN202110790628.6A
Authority: CN
Inventors: 李萌; 丁楠; 苏绥绥
Original assignee: Beijing Qiyu Information Technology Co Ltd
Current assignee: Beijing Qiyu Information Technology Co Ltd
Priority date: 2021-07-13
Filing date: 2021-07-13
Publication date: 2021-11-02

Abstract

The invention relates to the field of internet information processing, and provides a credit granting data optimization method, a device, equipment and a medium for solving the defect of risk runaway caused by mismatching of data and resources of actual service requirements, which cannot be accurately evaluated and predicted by the existing data processing. Therefore, the optimized change or the promotion amplitude can be accurately, reasonably and effectively obtained to obtain new credit granting data, the existing credit granting data promotion mode is optimized through multi-angle multi-model calculation, the safe and credible data is guaranteed, meanwhile, the accurate, reasonable and optimal credit granting data enables the normal and safe operation of the business, and the overdue risk is effectively controlled.

Description

Credit data optimization method, device, equipment and medium

Technical Field

The invention relates to the field of internet information processing, in particular to a user credit granting data optimization method and device, electronic equipment and a computer readable storage medium.

Background

With the rapid development of internet finance and the change of people's consumption concept, advanced consumption in a credit mode is more and more accepted by people. Due to the credit type financial data processed by network transmission, the requirements on information security and credibility of related data are higher, for example, existing credit products set data threshold limits such as use limit, namely the maximum limit that a credit user can use circularly under normal conditions. For financial data related to credit classes in the internet, the amount of credit data to be processed is huge, the risk and profitability changes of credit users are relatively more complicated, and the data security is guaranteed while the data security is adapted to the actual business change requirements, for example, in a credit business scene, when the data security and the credibility are guaranteed, the business is continuously realized, including that a credit customer obtains income and a credit party improves the profitability, and the like, the development changes of information related to numerous users are often evaluated through effective and accurate data processing. Taking a credit business scenario as an example, in an actual internet financial data processing process, for a credit-term user with a credited middle link of a business, a qualified customer base is usually selected, a Behavior scoring Card (behavor score Card, commonly referred to as B Card) model is used for evaluating data of changes related to credit customers to predict future overdue risks of the users in the customer base, whether safety threshold limits such as credit line of the user are proper or not is judged according to the grading of the risks, and then whether the magnitude of the amount of the credit is proper or not, for example, the amount of the credit is proper is not proper. However, when such data processing is used for evaluation, due to different quota providing ranges for the client, the future risks of the client are different, that is, the quota providing ranges adversely affect the risks, so that it cannot be determined how much quota is provided for the client to be the most appropriate, that is, the risk is out of control due to inaccurate credit data (specifically, credit limit), and further, the situation that the benefit cannot be maximized and the resources cannot be accurately matched due to large data processing errors in actual services exists.

Therefore, in the data processing process of the financial business, the credit granting data is further optimized while the data security is ensured, and the change of the credit granting data is more accurately and reasonably evaluated so as to control the risk in the business.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides a credit granting data optimization method, a credit granting data optimization device, electronic equipment and a computer readable medium, which can solve the technical problem of how to provide optimized credit granting data while ensuring data security in the process of processing mass network financial data; furthermore, the technical problem of how to accurately and reasonably evaluate and predict the change gradient, particularly the data growth gradient, so as to obtain the optimal user credit data can be solved, so that safe, reliable and accurate credit data change is provided for the user credit data, particularly the credit data is improved, and the overdue risk can be effectively controlled in financial services.

In order to solve the above technical problem, a first aspect of the present invention provides a method for optimizing credit data, including: based on the acquired risk model and response model, respectively evaluating each user of the user group to acquire a risk score and a response score, and calculating the gradient of credit promotion data of each user to judge whether the promotion level of the credit promotion data is changed; determining a sum of the unused information of the corresponding user group based on the unused information of each user predicted by the obtained prediction model; if the level of the credit granting data is determined to be continuously changed according to the sum, repeating the processing procedures of the two previous steps; otherwise, taking the lifting amplitude corresponding to the current lifting level as the final lifting amplitude to optimize the credit granting data of the user; wherein, the information for use is the information that the user uses the relevant service data in the service; wherein, the unused information is the information of the relevant service data reserved by the user in the service.

According to a preferred embodiment of the present invention, the obtaining the risk model specifically includes: obtaining at least one of said risk models; the risk model is used for evaluating the risk of a user when credit granting data is improved and outputting a risk score for improving the credit granting data; the obtaining of the response model specifically includes: obtaining at least one of the response models; the response model is used for evaluating the user information when the user promotes the credit data and outputting a response score for promoting the credit data; the obtaining of the prediction model specifically includes: obtaining at least one predictive model; the prediction model is used for predicting the user unused information in a preset period after the user promotes the credit data and outputting the user unused information after the credit data is promoted.

According to a preferred embodiment of the present invention, obtaining at least one risk model specifically includes: acquiring a user credit granting data label; promoting credit granting data to the user according to different promotion amplitudes based on the credit granting data label of the user; acquiring risk data of the user after credit granting data are promoted for the user according to different promotion amplitudes, and establishing promotion labels and risk labels; establishing at least one risk model according to the credit granting data label, the lifting label and the risk label of the user; and/or, acquiring at least one response model, specifically comprising: acquiring a user credit granting data label; promoting credit granting data to the user according to different promotion amplitudes based on the credit granting data label of the user; acquiring user information appearing after credit granting data is promoted for a user according to different promotion amplitudes, and establishing a promotion label and a response label; establishing at least one response model according to the credit granting data label, the promotion label and the response label of the user; and/or, obtaining at least one prediction model, specifically comprising: acquiring a user credit granting data label; promoting credit granting data to the user according to different promotion amplitudes based on the credit granting data label of the user; acquiring unused information of a user after credit data is promoted for the user according to different promotion amplitudes, and establishing a promotion tag and an unused tag for promoting the credit data; and establishing at least one prediction model according to the credit data label, the lifting label and the unused label of the user.

According to a preferred embodiment of the present invention, based on the obtained risk model and response model, the risk score and response score obtained by evaluating each user of the user group are respectively calculated, and the gradient of the trust promotion data of each user is calculated, specifically including: dividing the user quota raising amplitude into n levels; for each user in the user group_jObtaining a user_jCurrent credit data of_j(ii) a Initializing i-n, the current stage of said initializationThe lifting amplitude corresponding to the other n is the highest; wherein i, j and n are positive integers, and i is less than or equal to n; obtaining users from risk model and response model_jCorresponding to the current credit data_jAnd the lifting amplitude_ijRisk of improving risk score of credit data_ijResponse of response score_ijAnd corresponding to current credit data_jAnd the lifting amplitude_(i-1)jRisk of improving risk score of credit data_(i-1)jResponse of response score_(i-1)j(ii) a For each user in the user group_jCalculating current credit data_jGradient of underlying enhanced trust data_ij。

According to a preferred embodiment of the present invention, determining whether to change the level of trust data promotion specifically includes: will correspond to each user_jCurrent credit data_jThe gradient of_ijComparing with a first threshold if the user is_jSaid gradient of_ijAnd if the value is larger than or equal to the first threshold value, setting i-1, otherwise, setting i-i.

According to a preferred embodiment of the present invention, determining a sum of the unused information of the corresponding user group based on the unused information of each user predicted by the obtained prediction model specifically includes: for each user in the user group_jObtaining the each user from the prediction model_jCorresponding to the current credit data_jAnd the lifting amplitude_ijFor information not used after promotion of credit data_ij(ii) a Wherein the unused information_ijTo set up corresponding to the user_jStep i of_ijThe unused information after the credit granting data is promoted; the total sum of the unused information is M, and is equal to the total sum of the unused information of all users in the current user group after credit data promotion: unused information of M ═ sigma_ij。

According to a preferred embodiment of the present invention, if it is determined to continue to change the level of trust data promotion according to the sum, the processing procedure of the previous two steps is repeated; otherwise, taking the lifting amplitude corresponding to the current lifting level as the final lifting amplitudeSo as to optimize the credit granting data of the user, specifically comprising: comparing the sum with a preset second threshold: if the sum is greater than the second threshold value, setting i-1, repeating the steps of evaluating and calculating a gradient to determine and determining the sum until the sum is equal to or less than the second threshold value, and increasing the amplitude of lift at that time_ijAs the final determined optimum lifting amplitude_ijTo obtain the optimal credit data; if the sum is equal to or less than the second threshold, the lifting amplitude at that time is determined_ijAs the final determined optimum lifting amplitude_ij。

According to a preferred embodiment of the present invention, the dividing into n stages includes: dividing the promotion amplitude of the credit data of the user from 100% to 0% into n levels; the gradient_ijNot (risk)_ijRisk of_(i-1)j) /(response)_ij-response to_(i-1)j)。

According to a preferred embodiment of the present invention, further comprising: the first threshold and the second threshold are determined by the promotion credit data of the historical user and a machine learning model; or the first threshold and the second threshold are given by manual data analysis.

According to a preferred embodiment of the present invention, the risk model is a rate-increasing risk model, the response model is a rate-increasing response model, and the prediction model is a rate-increasing balance model; the risk score is an improvement risk score, the response score is an improvement response score, and the gradient is an improvement response gradient; the credit granting data is credit granting amount, the promotion amplitude is promotion amplitude, the information for use is dynamic support information of the user, and the information for use is balance information of the user; the sum is the total amount of money put, which is equal to the balance of all users in the current user group.

According to a preferred embodiment of the invention, based on the acquired risk model and response model, the risk score and response score obtained by evaluating each user of the user group are respectively calculated, and the gradient of credit promotion data of each user is calculated to judge whether the credit promotion level is changed or not, so that the method has the advantages of improving the credit promotion level of the credit promotion data, and improving the credit promotion level of the credit promotion dataThe body includes: dividing the user quota amplitude into n levels, aiming at each user in the user group_jObtaining a user_jCurrent credit line_jInitializing i ═ n, wherein i, j and n are positive integers, and i is less than or equal to n; obtaining users from an offer risk model and an offer response model_jCorresponding to the current credit line_jAnd amplitude of the promotion_ijRisk of the increased risk score of_ijScored responses to rate of offer_ijAnd corresponding to the current credit line_jAnd the next level of lifting amount_(i-1)jRisk of the increased risk score of_(i-1)jScored responses to rate of offer_(i-1)j(ii) a For each user in the user group_jCalculating the current credit line_jGradient of lift of lower_ij(ii) a Will correspond to each user_jCurrent credit line_jGradient of_ijComparing with a preset first threshold value if the user_jGradient of (2)_ijGreater than or equal to the first threshold, setting i-1, otherwise setting i-i; and/or, determining a sum of the unused information of the corresponding user group based on the unused information of each user predicted by the obtained prediction model, specifically comprising: for each user in the user group_jObtaining the credit limit of the user from the credit balance model_jAnd amplitude of the promotion_ijThe balance information of_ij(ii) a Calculating a total deposit amount M which is equal to the balance of all users in the current user group_ijThe sum of (a); and/or if the level of the credit data promotion is determined to be continuously changed according to the sum, repeating the processing procedures of the two previous steps; otherwise, taking the lifting amplitude corresponding to the current lifting level as the final lifting amplitude to optimize the credit granting data of the user, specifically comprising: comparing the total deposit amount M with a preset second threshold value; if the total deposit amount M is greater than the second threshold value, setting i to i-1, repeating the steps of calculating a gradient to judge and determining the sum until M is equal to or less than the second threshold value, and determining the amount of the withdrawal at that time_ijAs the final determined magnitude of the boost_ij(ii) a If the total deposit amount M is equal to or less than the second threshold value, the amount increase amplitude determined at this time_ijAs the final determined magnitude of the boost_ij。

The second aspect of the present invention provides a trust data optimization apparatus, including: the gradient module is used for respectively evaluating each user of the user group to obtain a risk score and a response score based on the risk model and the response model, and calculating the gradient of credit promotion data of each user so as to judge whether the credit promotion level is changed; a prediction module for determining a sum of unused information of the corresponding user group based on the unused information predicted for each user by the prediction model; the condition module is used for repeating the processing processes of the gradient module and the prediction module if the level of credit granting data promotion is determined to be continuously changed according to the sum, otherwise, the promotion amplitude corresponding to the current promotion level is used as the final promotion amplitude to optimize the credit granting data of the user; the use information is information that the user uses the relevant service data in the service; the unused information is information of related service data reserved in the service by the user.

A third aspect of the present invention provides an electronic device, comprising: a processor; and a memory storing computer executable instructions that, when executed, cause the processor to perform the method of the aforementioned first aspect.

A fourth aspect of the present invention proposes a computer readable storage medium, wherein the computer readable storage medium stores one or more programs which, when executed by a processor, implement the method of the aforementioned first aspect.

In one embodiment of the invention, through a plurality of evaluation models (risk models and response models) and prediction models, the change gradient, especially the growing gradient, of the credit data is dynamically determined along with the change of the actual business data (namely, the introduced gradient is used for measuring the direction of one optimizing tone), and the actual increasable amplitude of the optimal credit data corresponding to the current user is determined. Therefore, the multiple models dynamically evaluate and predict according to the multi-angle data of the user to determine the gradient/grade so as to determine the optimal promotion amplitude of the credit data which can be provided for the user, namely, to obtain the optimized changed or promoted new credit data. And furthermore, reasonable and optimal credit granting data can ensure that overdue risks can be effectively controlled in the process of normally realizing profits by the business.

Further, the credit granting data is changed according to different promotion amplitudes (amplification for short) by using the current credit granting data label of the user, the corresponding overdue condition is obtained, a promotion label and a risk label are established, so that the label of the credit granting data is combined, at least one risk model for evaluating risks is established, and the risk of the current user for changing the credit granting data is evaluated; similarly, acquiring information for use, so as to establish a lifting tag and a lifting response tag, and in combination with the tag of the credit granting data, establishing at least one response model for evaluating the use condition, and evaluating the lifting response of the current user for changing the credit granting data; similarly, the unused information is obtained, a lifting tag and an unused tag are established, at least one prediction model is established by combining the tags of the credit data, and the unused situation after the credit data of the current user is changed is predicted. And according to the evaluation and prediction results of the three models, calculating the amplification gradient of the credit granting data of the current user, and further determining the optimal amplification to provide the optimized credit granting data corresponding to the current user. By the evaluation and prediction of specific multi-model multi-angle data, risk and response evaluation and prediction of amplification unused information can be more accurately and reasonably realized, so that the safety and credibility of data processing are ensured, optimized credit granting data can be obtained, further, the optimal amplification can be determined for a corresponding user by introducing dynamic gradient change, the optimal credit granting data of the user in actual business is optimized, and further, overdue risks can be effectively controlled in financial business according to the accurate and reasonable credit granting data.

And further, determining a lifting gradient based on the combination of the amplification grading with a risk model, the risk score of a response model and the response score, predicting the unused information of corresponding amplification according to a prediction model, determining the final amplification, then circulating the processes until a preset condition is met, and taking the final amplification meeting the preset condition (the grading is finished or a threshold is reached) as the optimal amplification to further determine corresponding optimal credit granting data. The optimal amplification can be further accurately and reasonably determined by the cyclic processing of the amplification grading and the condition limitation, and the optimal promoted credit granting data can be obtained.

Furthermore, the method and the device are applied to a credit business scene, and by adopting the technical scheme of the invention, the change of the quota can be optimized to realize dynamic change, and the risk can be stabilized while the total sum of the loan of the user group is kept unchanged.

Drawings

In order to make the technical problems solved by the present invention, the technical means adopted and the technical effects obtained more clear, the following will describe in detail the embodiments of the present invention with reference to the accompanying drawings. It should be noted, however, that the drawings described below are only illustrations of exemplary embodiments of the invention, from which other embodiments can be derived by those skilled in the art without inventive step.

Fig. 1 is a main flowchart of an embodiment of the trust data optimization method according to the present invention.

Fig. 2 is a functional module architecture block diagram of an embodiment of the trust data optimization apparatus according to the present invention.

Fig. 3 is a block diagram of an exemplary embodiment of an electronic device according to the present invention.

FIG. 4 is a schematic diagram of one embodiment of a computer-readable medium of the present invention.

Detailed Description

Exemplary embodiments of the present invention will now be described more fully hereinafter with reference to the accompanying drawings, in which exemplary embodiments of the invention may be embodied in many specific forms, and should not be construed as limited to the embodiments set forth herein. Rather, these exemplary embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the invention to those skilled in the art.

The structures, properties, effects or other characteristics described in a certain embodiment may be combined in any suitable manner in one or more other embodiments, while still complying with the technical idea of the invention.

In describing particular embodiments, specific details of structures, properties, effects, or other features are set forth in order to provide a thorough understanding of the embodiments by one skilled in the art. However, it is not excluded that a person skilled in the art may implement the invention in a specific case without the above-described structures, performances, effects or other features.

The flow chart in the drawings is only an exemplary flow demonstration, and does not represent that all the contents, operations and steps in the flow chart are necessarily included in the scheme of the invention, nor does it represent that the execution is necessarily performed in the order shown in the drawings. For example, some operations/steps in the flowcharts may be divided, some operations/steps may be combined or partially combined, and the like, and the execution order shown in the flowcharts may be changed according to actual situations without departing from the gist of the present invention.

The block diagrams in the figures generally represent functional entities and do not necessarily correspond to physically separate entities. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.

The same reference numerals denote the same or similar elements, components, or parts throughout the drawings, and thus, a repetitive description thereof may be omitted hereinafter. It will be further understood that, although the terms first, second, third, etc. may be used herein to describe various elements, components, or sections, these elements, components, or sections should not be limited by these terms. That is, these phrases are used only to distinguish one from another. For example, a first device may also be referred to as a second device without departing from the spirit of the present invention. Furthermore, the term "and/or", "and/or" is intended to include all combinations of any one or more of the listed items.

In an embodiment of the invention, through a plurality of evaluation models (risk models, response models) and prediction models, the change gradient, especially the growth gradient, of credit granting data is dynamically determined along with the change of actual business data, and the optimal promotable amplitude corresponding to the current user is determined, so that the optimal credit granting data, namely the resource matching is most suitable, is determined.

[ example 1 ]

Fig. 1 is a main flow diagram of an embodiment of a method according to the invention. As shown in fig. 1, the method of the present invention at least comprises the following steps:

step S1, based on the risk model and the response model, respectively evaluating each user of the user group to obtain a risk score and a response score, and calculating the gradient of the promotion authorization data of each user to judge whether to change the promotion level of the credit data; specifically, the authorization data includes trust data, and all of the trust data will be taken as an example below.

Step S2, determining the sum of the information for the user groups based on the information for the user groups predicted by the prediction model;

step S3, if the level of credit granting data promotion is determined to be changed continuously according to the sum, repeating the process from S1 to S2, otherwise, taking the promotion amplitude corresponding to the current promotion level as the final promotion amplitude to optimize the credit granting data of the user;

wherein, the information for use is the information that the user uses the relevant service data in the service;

wherein, the unused information is the information of the relevant service data reserved by the user in the service.

At least one risk model in S1 can be obtained by the following example:

acquiring a user credit granting data label;

promoting the credit granting data to the user according to different promotion amplitudes based on the credit granting data label data of the user;

acquiring overdue data of the user after the credit granting data is promoted for the user according to different promotion amplitudes, and establishing a promotion amplitude label (promotion label for short) and an insurance label;

and according to the credit granting data label of the user, the promotion label and the risk label, and by combining behavior change in user service and qualification change in service, establishing at least one risk model for promoting the credit granting data.

The acquired at least one risk model may be used to evaluate a risk (e.g., an overdue risk) when credit data is promoted for the user, and output a risk score for promoting the credit data.

At least one response model in S1 can be obtained by the following example:

acquiring a user credit granting data label;

promoting credit granting data to the user according to different promotion amplitudes based on the credit granting data label of the user;

acquiring information (such as user dynamic information in credit) for use of the user after credit data is promoted for the user according to different promotion amplitudes, and establishing promotion tags and response tags;

and establishing at least one response model according to the credit data label, the promotion label and the response label of the user and by combining behavior change in user service and qualification change in service.

The obtained at least one response model can be used for evaluating user dynamic information (user dynamic support information) when credit data is improved for a user, and outputting a response score for improving the credit data.

At least one prediction model in S2 can be obtained by the following example:

acquiring a user credit granting data label;

acquiring user unused information (such as user balance information in credit) of the user after credit data is promoted to the user according to different promotion amplitudes, and establishing a promotion label and an unused label;

and establishing at least one prediction model according to the credit granting data label, the lifting label and the unused label of the user and by combining behavior change in user service and qualification change in service.

The obtained at least one prediction model can be used for predicting the user unused information in a preset period after the credit data is promoted to the user, and outputting the user unused information corresponding to the promoted credit data.

In S1, based on the risk model and the response model, a risk score and a response score obtained by evaluating each user of the user group are respectively calculated, and a gradient of credit promotion data of each user is calculated to determine whether to change a level of credit promotion of the credit promotion data. The user's optimum magnitude of lift, or level of lift, can thus be calculated from the gradient. The method specifically comprises the following steps:

first, a risk model, a response model, and a prediction model are used in a real-time business process of a user group including j users.

Calculating the gradient of the promotion credit data of the user, such as:

for each user in the user group_jDividing the user lifting amplitude from 100% to 0% into n levels to obtain the user_jCurrent credit data of_jInitializing i-n (i.e., initializing the current level to have a boost magnitude of 100%).

Through the evaluation of each risk model and response model, the user is obtained_jRespectively corresponding to the current credit data_jAnd the lifting amplitude_ijAnd the lifting amplitude of the next stage_(i-1)jThe risk score and the response score of the credit data are improved.

Therein, risk of_ijTo correspond to the user_jIn response to the risk score of the increased trust data of the increased amplitude of the ith level_ijTo correspond to the user_jThe response score of the promotion credit data of the ith level, i, j and n are positive integers, and i is less than or equal to n.

The current credit granting data of the user is obtained by obtaining the user credit granting data tag. Namely, each user has a corresponding credit data tag for credit.

For each user in the user group_jCalculating the gradient of credit promotion data under the current credit promotion data_ij；

Gradient of gradient_ijNot (risk)_ijRisk of_(i-1)j) /(response)_ij-response to_(i-1)j)

The gradient is used for measuring the direction of optimizing the lifting amplitude, and the higher the gradient value is, the higher the risk increase of the user under the current credit data lifting is represented, or the lower the user use (user dynamic support) will be, therefore, if from the optimization perspective, the lower the gradient of the user is, the higher the lifting amplitude should be preferentially increased to the part of users, and if the gradient of the user is higher, the lower the lifting amplitude should be decreased to the part of users.

The judgment is carried out according to the gradient, and specifically, for example:

will correspond to each user_jGradient under current credit data_ijComparing with a first threshold if the user is_jGradient of (2)_ijAnd if the value is larger than or equal to the first threshold value, setting i-1, namely lowering the level, otherwise, setting i-i.

The first threshold may be determined by the machine learning model and the historical user's contribution risk data.

The first threshold value may also be given by manual data analysis.

The first threshold may rank the decision for the gradient of each user in the user group. After the gradient of the elevation of each user in the user group is calculated, sorting the users in the user group according to the gradient of the elevation of the users, and determining the first threshold value according to the sorting.

For example, the user ranking the top 10% of the gradients in the user group is selected to decrease the boost amplitude, and the first threshold is preset so that the gradient of the user in the user group greater than the first threshold belongs to the top 10%.

In S2, determining the sum of the non-used information of the corresponding user group based on the non-used information predicted by the prediction model for each user may specifically include:

the sum of all unused information that may be provided to the user population (e.g., the total amount of the deposit in the credit) M is calculated.

For each user in the user group_jObtaining the each user from the prediction model_jCorresponding to the current credit data_jAnd the lifting amplitude_ijFor information not used after promotion of credit data_ij。

Wherein the unused information_ijTo set up corresponding to the user_jStep i of_ijThe unused information after the credit granting data is improved.

The sum of all the unused information provided to the user group (e.g., the total amount paid in credit) M is equal to the sum of the unused information after the credit data is promoted for all users in the current user group. I.e. M ═ sigma unused information_ij。

In S3, if it is determined according to the sum that the level of trust data elevation continues to be changed, repeating the process from S1 to S2, otherwise, using the elevation amplitude corresponding to the current elevation level as the final elevation amplitude to optimize the trust data of the user, which may specifically include:

comparing said M with a preset second threshold:

if M is equal to or less than the preset second threshold value, the determined lifting amplitude at this time_ijAs a final determined lifting amplitude_ij；

If M is greater than the preset second threshold value, setting i-1, repeating the above steps S1 to S2 until M is equal to or less than the second threshold value, and determining the lifting amplitude at that time_ijAs a final determined lifting amplitude_ijNamely the optimal amplification of the current user, and then the credit granting data can be updated to obtain the optimal credit granting data.

The second threshold may be determined by the machine learning model and the historical user's contribution risk data. The second threshold value may also be given by manual data analysis.

According to the embodiment of the invention, the existing trust data promotion mode is optimized when the data security and credibility are guaranteed, and the risk can be further stabilized while the total provided data of the user group is kept unchanged.

[ example 2 ]

Similarly, an embodiment of the corresponding trust data optimization apparatus corresponds to the method. As shown in fig. 2, the apparatus according to an embodiment of the present invention may specifically include:

the gradient module is used for respectively evaluating each user of the user group to obtain a risk score and a response score based on the risk model and the response model, and calculating the gradient of credit promotion data of each user so as to judge whether the credit promotion level is changed; for specific functions, see specific steps and contents of S1, which are not described herein again.

A prediction module for determining a sum of the unused information of the corresponding user group based on the unused information predicted by the prediction model for each user; for specific functions, see specific steps and contents of S2, which are not described herein again.

The condition module is used for repeating the processing processes of the two modules if the level of credit granting data promotion is determined to be continuously changed according to the sum, otherwise, the promotion amplitude corresponding to the current promotion level is used as the final promotion amplitude to optimize the credit granting data of the user; for specific functions, see specific steps and contents of S3, which are not described herein again.

Because such trust data optimization schemes are more applied to the internet credit scene of fussy and complex mass data processing, the details are as follows: the service may correspond to a credit service of the credit scenario; the credit data of the user can be, for example, credit limit of the user; the label of the credit data can be corresponding credit limit label data; risks such as overdue risks in this scenario, and risk data is often overdue risk data; the boost amplitude, i.e. the amplification, may be a quantum boost amplitude; the promotion or change of the credit data can be promotion or change of credit limit; the lifting tag can lift amplitude tag data; the risk label may be overdue risk label data; the call-up information may be information of the call-up of the user; the promotion response tag may be promotion response tag data; the unused information may be user balance information of the user; the inactive tag may be balance tag data; the risk model may be an offer risk model, the response model may be an offer response model, and the prediction model may be an offer balance model; and so on.

The implementation of the present invention will be described in detail below in the context of an internet credit scenario.

[ example 3 ]

In one embodiment of the method applied to a credit scenario, at least the following steps are included:

s101: and obtaining at least one quota risk model, wherein the quota risk model is used for evaluating the risk when the credit line is promoted for the user and outputting a quota risk score.

Wherein, the step S101 of obtaining at least one quoting risk model further includes:

s1011: acquiring the credit line label data of a user;

s1012: based on the credit line label data of the user, the credit line is promoted to the user according to different promotion amplitudes;

s1013: acquiring overdue data of the user after the credit line is increased for the user according to different quota increasing ranges, and establishing quota increasing range label data and overdue risk label data;

s1014: and establishing at least one quota risk model according to the credit line label data, quota amplitude label data and overdue risk label data of the user and by combining behavior change in loan and qualification change in loan of the user.

Wherein the establishing of overdue risk label data in the step S1013 further includes:

establishing whether at least one of overdue, number of overdue days, and/or amount of overdue risk label data.

As an example, in the process of establishing the quota increasing risk model, firstly, part of users need to be extracted as training samples, the credit line of the users serving as the training samples is obtained, and credit line label data is established. Then, the training sample users are proportionally promoted based on the credit line of the users, and the training sample users are promoted by 5%, 10%, … … 100% and the like as examples.

And thirdly, after the training sample user is subjected to the quota extraction, observing for a period of time, acquiring overdue data of the training sample user after the quota extraction is 5%, 10% and … … 100% and establishing quota amplitude label data and overdue risk label data.

And establishing an amount-increasing risk model according to the credit line, the amount-increasing amplitude and overdue risk data of the training sample user and by combining behavior change in the user loan and qualification change in the user loan.

The change of behavior in the user loan has a non-negligible influence on the overdue risk of the user, so before the risk model is established, the behavior change in the user loan needs to be acquired as a reference. The behavior change in credits comprises: a change in the installment repayment, a change in the repayment amount, a change in the loan frequency, and/or a change in the loan amount.

The change of qualification in user loan has a non-negligible influence on the overdue risk of the user, and the qualification change in user loan is required to be obtained as a reference, and the qualification change in user loan includes: revenue changes, and/or liability changes

When establishing the rate-increasing risk scoring model, it is necessary to acquire training sample user attribute label data, where the user attribute label data includes: gender, age, location, academic history, income, and/or liabilities. The sources of the user attribute tag data include: user registration data, external credit agency data, and/or associated user data, etc.

In the embodiment of the invention, at least one quota risk model is established according to different credit lines and quota ranges. After the quota-increasing risk model is established, the model is trained and used in a real-time user loan process.

S102: and acquiring at least one quota response model, wherein the quota response model is used for evaluating the user dynamic support information when the credit line is promoted for the user, and outputting a quota response score.

Wherein, the step S102 of obtaining at least one referral response model further comprises:

s1021: acquiring the credit line label data of a user;

s1022: based on the credit line label data of the user, the credit line is promoted to the user according to different promotion amplitudes;

s1023: acquiring dynamic support information of a user after the credit line is increased for the user according to different quota increasing amplitudes, and establishing quota increasing amplitude tag data and quota responding tag data;

the dynamic support information of the user comprises whether the dynamic support is performed or not in a preset period after the user increases the quota, wherein the preset period is any one of the following periods: 1 day, 3 days, 7 days, 10 days, 1 month, 2 months, 3 months, half a year, one year, etc.

S1024: and establishing at least one quota response model according to the credit line label data, the quota amplitude label data and the quota response label data of the user and by combining the behavior change in the user loan and the qualification change in the user loan.

By way of example, in the process of establishing the quota response model, a part of users need to be extracted as training samples. And obtaining the credit line of the user as a training sample, and establishing credit line label data. Then, the training sample users are proportionally promoted based on the credit line of the users, and the training sample users are promoted by 5%, 10%, … … 100% and the like as examples. Optionally, based on the credit line of the user, the user is increased according to the credit line value, namely, the user is increased by 500, 1000 … … and the like.

And thirdly, after the user of the training sample is subjected to the quota increasing, observing for a period of time, acquiring the dynamic support data of the user of the training sample after the quota increasing is 5%, 10% and … … 100%, and establishing quota amplitude label data and quota response label data.

And establishing a rate-increasing response model according to the credit line, the rate-increasing amplitude and the rate-increasing response data of the training sample user and by combining the behavior change in the user credit and the qualification change in the user credit.

S103: and acquiring at least one quota balance model, wherein the quota balance model is used for predicting the user borrowing balance in a preset period after the credit line is upgraded for the user, and outputting the user balance information.

Wherein, the step S103 of obtaining at least one balance model further includes:

s1031: acquiring the credit line label data of a user;

s1032: based on the credit line label data of the user, the credit line is promoted to the user according to different promotion amplitudes;

s1033: acquiring user balance information of a user after the credit line is increased for the user according to different increase amplitudes, and establishing increase amplitude tag data and increase balance tag data;

the user balance information of the user comprises the user balance which is uncompensated in a preset period after the user promotes the credit line.

The preset period is any one of the following periods: 1 day, 3 days, 7 days, 10 days, 1 month, 2 months, 3 months, half a year, one year, etc.

S1034: and establishing at least one quota balance model according to the credit line label data, quota amplitude label data and quota balance label data of the user and by combining action change in credit and qualification change in credit of the user.

By way of example, in the process of establishing the quota balance model, part of users are required to be extracted as training samples. And obtaining the credit line of the user as a training sample, and establishing credit line label data. Then, the training sample users are proportionally promoted based on the credit line of the users, and the training sample users are promoted by 5%, 10%, … … 100% and the like as examples. Optionally, based on the credit line of the user, the user is increased according to the credit line value, namely, the user is increased by 500, 1000 … … and the like.

And thirdly, observing for a period of time after the amount of the training sample user is increased, acquiring balance data of the training sample user after the amount of the training sample user is increased by 5%, 10% and … … 100%, and establishing the amount increase label data and the amount balance label data.

And establishing a credit balance model according to the credit line, the credit promotion amplitude and the credit balance data of the training sample user and by combining the behavior change in the credit and the qualification change in the credit of the user.

S104: and calculating the user rate gradient by using the rate-increasing risk model, the rate-increasing response model and the rate-increasing balance model, and calculating the optimal rate-increasing amplitude of the user according to the gradient.

As an example, the offer risk model, the offer response model, and the offer balance model are used in a real-time loan process for a user group including j users.

Wherein, the step S104 calculates a user rate gradient, and calculates an optimal rate amplitude of the user according to the gradient, further comprising:

s1041: for each user in the user group_jDividing the user quota amplitude from 100% to 0% into n levels to obtain the user_jCurrent credit line_jInitializing i-n (i.e., initializing the current quota to 100%).

S1042: obtaining users from each of the rate-of-offer risk models and rate-of-offer response models_jRespectively corresponding to the current credit line_jAnd amplitude of the promotion_ijAnd the amplitude of the increase_(i-1)jThe rate-of-offer risk score and the rate-of-offer response score.

Therein, risk of_ijTo correspond to the user_jIn response to the rating risk score for the rating amplitude of the ith level_ijTo correspond to the user_jI, j, n are positive integers and i is less than or equal to n.

The current credit line of the user is obtained by obtaining the data of the credit line label of the user.

S1043: for each user in the user group_jCalculating the rate-increasing gradient under the current credit line_ij；

The gradient is used for measuring the direction of optimizing the amplitude of the quota, and the higher the gradient value is, the higher the quota risk of the user under the premise is increased, or the lower the willingness of the user is, therefore, if the gradient of the user is lower from the optimization perspective, the increase amplitude of the quota should be preferentially increased for the part of the users, and if the gradient of the user is higher, the decrease amplitude of the quota should be decreased for the part of the users.

S1044: will correspond to each user_jGradient under current credit line_ijComparing with a first threshold if the user is_jGradient of (2)_ijAnd if the value is larger than or equal to the first threshold value, setting i-1, otherwise, setting i-i.

The first threshold value may also be given by manual data analysis.

The first threshold may rank the decision for the gradient of each user in the user group. After the rate-of-rise gradient of each user in the user group is calculated, the users in the user group are sorted according to the rate-of-rise gradient of the users, and the first threshold value is determined according to the sorting.

For example, the user ranking the top 10% of the gradients in the user group is selected to reduce the magnitude of the increase, and a first threshold is preset so that the gradients of users in the user group greater than the first threshold belong to the top 10%.

S1045: and calculating the total deposit amount M of the user group.

For each user in the user group_jObtaining the credit limit of the user from the credit balance model_jAnd amplitude of the promotion_ijThe balance information of the quota.

Wherein the balance_ijTo be arranged asCorresponding to the user_jStep i of_ijThe balance information of the quota.

And the total amount M of the deposit is equal to the sum of the balance information of all the users in the current user group. I.e. M ═ sigma balance_ij。

S1046: comparing the total deposit amount M with a preset second threshold value, and if the total deposit amount M is equal to or less than the preset second threshold value, determining the amount-increasing range at the moment_ijAs the final determined magnitude of the boost_ijIf the total deposit amount M is greater than the preset second threshold, setting i to i-1, repeating the above steps S1042 to S1045 until M is equal to or less than the second threshold, and determining the amount increase range at this time_ijAs the final determined magnitude of the boost_ij。

The method optimizes the existing quota improving mode, and can stabilize the risk while keeping the total sum of the group payment of the users unchanged.

The method of the invention uses the trained model in the real-time user loan process, based on the users with different credit line grades, outputs the quotation model scores corresponding to each user according to each line, comprehensively considers the risk indexes, selects the good user quotation with the line requirements and gradually explores the space of the quotation, namely, the user after the judgment of the quotation model observes the subsequent behavior again, continuously iterates the model and explores the upper limit of the quotation of each type of user.

As an example, 100% of the amount is increased for each user in the user group, the amount increasing range of the user is divided into 20 levels, the current credit line and the current amount increasing range are input into an amount increasing risk and amount increasing response model by 100%, and the gradient of each user is calculated according to the output of the model. If the user gradient is in the top 10% of the user group, the rate-increasing amplitude of the user is set as the rate-increasing amplitude of the next level, namely, the rate-increasing amplitude of the part of the users with the top 10% of the gradient rank in the user group is adjusted to 95%, and the rate-increasing amplitude of the part of the users with the bottom 90% of the gradient rank in the user group is still 100%. The above process is repeated until the total amount of the deposit across the user population matches the deposit plan, resulting in an optimized magnitude of the lift for each customer.

Those skilled in the art will appreciate that all or part of the steps for implementing the above-described embodiments are implemented as programs executed by data processing apparatuses (including computers), i.e., computer programs. When the computer program is executed, the method provided by the invention can be realized. Furthermore, the computer program may be stored in a computer readable storage medium, which may be a readable storage medium such as a magnetic disk, an optical disk, a ROM, a RAM, or a storage array composed of a plurality of storage media, such as a magnetic disk or a magnetic tape storage array. The storage medium is not limited to centralized storage, but may be distributed storage, such as cloud storage based on cloud computing.

[ example 4 ]

Embodiments of the apparatus of the present invention are described below, which may be used to perform method embodiments of the present invention. The details described in the device embodiments of the invention should be regarded as complementary to the above-described method embodiments; reference is made to the above-described method embodiments for details not disclosed in the apparatus embodiments of the invention. The device is applied to an embodiment of an internet credit scene and specifically comprises the following steps:

a module 201 for obtaining at least one rate-increasing risk model, a module 202 for obtaining at least one rate-increasing response model, a module 203 for obtaining at least one rate-increasing balance model, and a module 204 for calculating a rate-increasing gradient of the user using the rate-increasing risk model, the rate-increasing response model, and the rate-increasing balance model, and calculating an optimal rate-increasing amplitude of the user according to the gradient.

The module 201 for obtaining at least one rate-increasing risk model, the module 202 for obtaining at least one rate-increasing response model, the module 203 for obtaining at least one rate-increasing balance model, and the module 204 for calculating a rate-increasing gradient of the user by using the rate-increasing risk model, the rate-increasing response model, and the rate-increasing balance model, and calculating an optimal rate-increasing amplitude of the user according to the gradient correspond to the method steps S101 to S104 in embodiment 3, respectively, which are not described herein again.

Those skilled in the art will appreciate that the modules in the above-described embodiments of the apparatus may be distributed as described in the apparatus, and may be correspondingly modified and distributed in one or more apparatuses other than the above-described embodiments. The modules of the above embodiments may be combined into one module, or further split into multiple sub-modules.

[ example 5 ]

In the following, embodiments of the electronic device of the present invention are described, which may be regarded as an implementation in physical form for the above-described embodiments of the method and apparatus of the present invention. Details described in the embodiments of the electronic device of the invention should be considered supplementary to the embodiments of the method or apparatus described above; for details which are not disclosed in embodiments of the electronic device of the invention, reference may be made to the above-described embodiments of the method or the apparatus.

Fig. 3 is a block diagram of an exemplary embodiment of an electronic device according to the present invention. The electronic device shown in fig. 3 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.

As shown in fig. 3, the electronic apparatus 200 of the exemplary embodiment is represented in the form of a general-purpose data processing apparatus. The components of the electronic device 200 may include, but are not limited to: at least one processing unit 210, at least one memory unit 220, a bus 230 connecting different system components (including the memory unit 220 and the processing unit 210), a display unit 240, and the like.

The storage unit 220 stores a computer readable program, which may be a code of a source program or a read-only program. The program may be executed by the processing unit 210 such that the processing unit 210 performs the steps of various embodiments of the present invention. For example, the processing unit 210 may perform the steps as shown in fig. 1.

The memory unit 220 may include readable media in the form of volatile memory units, such as a random access memory unit (RAM)2201 and/or a cache memory unit 2202, and may further include a read only memory unit (ROM) 2203. The storage unit 220 may also include a program/utility 2204 having a set (at least one) of program modules 2205, such program modules 2205 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

Bus 230 may be one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

The electronic device 200 may also communicate with one or more external devices 300 (e.g., keyboard, display, network device, bluetooth device, etc.), enable a user to interact with the electronic device 200 via the external devices 300, and/or enable the electronic device 200 to communicate with one or more other data processing devices (e.g., router, modem, etc.). Such communication may occur via input/output (I/O) interfaces 250, and may also occur via network adapter 260 with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network such as the Internet). The network adapter 260 may communicate with other modules of the electronic device 200 via the bus 230. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in the electronic device 200, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

[ example 6 ]

FIG. 4 is a schematic diagram of one computer-readable medium embodiment of the present invention. As shown in fig. 4, the computer program may be stored on one or more computer readable media. The computer readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. The computer program, when executed by one or more data processing devices, enables the computer-readable medium to implement the above-described methods of the present invention.

Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments of the present invention described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiment of the present invention can be embodied in the form of a software product, which can be stored in a computer-readable storage medium (which can be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to make a data processing device (which can be a personal computer, a server, or a network device, etc.) execute the above-mentioned method according to the present invention.

The computer readable storage medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable storage medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).

In summary, the present invention can be implemented as a method, an apparatus, an electronic device, or a computer-readable medium executing a computer program. Some or all of the functions of the present invention may be implemented in practice using a general purpose data processing device such as a microprocessor or a Digital Signal Processor (DSP).

While the foregoing embodiments have described the objects, aspects and advantages of the present invention in further detail, it should be understood that the present invention is not inherently related to any particular computer, virtual machine or electronic device, and various general-purpose machines may be used to implement the present invention. The invention is not to be considered as limited to the specific embodiments thereof, but is to be understood as being modified in all respects, all changes and equivalents that come within the spirit and scope of the invention.

Claims

1. A trust data optimization method is characterized by comprising the following steps:

based on the acquired risk model and response model, respectively evaluating each user of the user group to acquire a risk score and a response score, and calculating the gradient of credit promotion data of each user to judge whether the promotion level of the credit promotion data is changed;

determining a sum of the unused information of the corresponding user group based on the unused information of each user predicted by the obtained prediction model;

if the level of the credit granting data is determined to be continuously changed according to the sum, repeating the processing procedures of the two previous steps; otherwise, taking the lifting amplitude corresponding to the current lifting level as the final lifting amplitude to optimize the credit granting data of the user;

2. The method of claim 1,

the obtaining of the risk model specifically includes: obtaining at least one of said risk models;

the risk model is used for evaluating the risk of a user when credit granting data is improved and outputting a risk score for improving the credit granting data;

the obtaining of the response model specifically includes: obtaining at least one of the response models;

the response model is used for evaluating the user information when the user promotes the credit data and outputting a response score for promoting the credit data;

the obtaining of the prediction model specifically includes: obtaining at least one predictive model;

the prediction model is used for predicting the user unused information in a preset period after the user promotes the credit data and outputting the user unused information after the credit data is promoted.

3. The method according to claim 2, wherein obtaining at least one of the risk models specifically comprises:

acquiring a user credit granting data label;

acquiring risk data of the user after credit granting data are promoted for the user according to different promotion amplitudes, and establishing promotion labels and risk labels;

establishing at least one risk model according to the credit granting data label, the lifting label and the risk label of the user;

and/or the presence of a gas in the gas,

obtaining at least one response model specifically includes:

acquiring a user credit granting data label;

acquiring user information appearing after credit granting data is promoted for a user according to different promotion amplitudes, and establishing a promotion label and a response label;

establishing at least one response model according to the credit granting data label, the promotion label and the response label of the user;

and/or the presence of a gas in the gas,

obtaining at least one prediction model, specifically comprising:

acquiring a user credit granting data label;

acquiring unused information of a user after credit data is promoted for the user according to different promotion amplitudes, and establishing a promotion tag and an unused tag for promoting the credit data;

and establishing at least one prediction model according to the credit data label, the lifting label and the unused label of the user.

4. The method according to any one of claims 1 to 3, wherein the step of calculating a gradient of the promotion credit data of each user based on a risk score and a response score obtained by evaluating each user of the user group based on the obtained risk model and response model comprises:

dividing the user quota raising amplitude into n levels;

for each user in the user group_jObtaining a user_jCurrent credit data of_j；

Initializing i to n, wherein the lifting amplitude corresponding to the initialized current level n is the highest;

wherein i, j and n are positive integers, and i is less than or equal to n;

obtaining users from risk model and response model_jCorresponding to the current credit data_jAnd the lifting amplitude_ijRisk of improving risk score of credit data_ijResponse of response score_ijAnd corresponding to current credit data_jAnd the lifting amplitude_(i-1)jRisk of improving risk score of credit data_(i-1)jResponse of response score_(i-1)j；

For each user in the user group_jCalculating current credit data_jGradient of underlying enhanced trust data_ij。

5. The method according to claim 4, wherein determining whether to change the level of trust data promotion specifically comprises:

will correspond to each user_jCurrent credit data_jThe gradient of_ijComparing with a first threshold if the user is_jSaid gradient of_ijAnd if the value is larger than or equal to the first threshold value, setting i-1, otherwise, setting i-i.

6. The method according to claim 4 or 5, wherein determining a sum of the unpopulated information of the corresponding user group based on the unpopulated information of each user predicted by the obtained prediction model comprises:

for each user in the user group_jObtaining the each user from the prediction model_jCorresponding to the current credit data_jAnd the lifting amplitude_ijFor information not used after promotion of credit data_ij；

Wherein the unused information_ijTo set up corresponding to the user_jStep i of_ijThe unused information after the credit granting data is promoted;

the sum of the information for non-use is M, which is equal to all the information in the current user groupThe sum of unused information of the user after the credit data is promoted: unused information of M ═ sigma_ij。

7. The method according to any one of claims 4 to 6, characterized in that if the level of credit data promotion is determined to be changed continuously according to the sum, the processing procedures of the previous two steps are repeated; otherwise, taking the lifting amplitude corresponding to the current lifting level as the final lifting amplitude to optimize the credit granting data of the user, specifically comprising:

comparing the sum with a preset second threshold:

if the sum is greater than the second threshold value, setting i-1, repeating the steps of evaluating and calculating a gradient to determine and determining the sum until the sum is equal to or less than the second threshold value, and increasing the amplitude of lift at that time_ijAs the final determined optimum lifting amplitude_ijTo obtain the optimal credit data;

if the sum is equal to or less than the second threshold, the lifting amplitude at that time is determined_ijAs the final determined optimum lifting amplitude_ij。

8. The method according to any one of claims 4 to 7,

the dividing into n stages includes: dividing the promotion amplitude of the credit data of the user from 100% to 0% into n levels;

the gradient_ijNot (risk)_ijRisk of_(i-1)j) /(response)_ij-response to_(i-1)j)。

9. The method of any of claims 5 to 7, further comprising:

the first threshold and the second threshold are determined by the promotion credit data of the historical user and a machine learning model; or,

the first threshold and the second threshold are given by manual data analysis.

10. The method according to any one of claims 1 to 9,

the risk model is a quota risk model, the response model is a quota response model, and the prediction model is a quota balance model;

the risk score is an improvement risk score, the response score is an improvement response score, and the gradient is an improvement response gradient;

the credit granting data is credit granting amount, the promotion amplitude is promotion amplitude, the information for use is dynamic support information of the user, and the information for use is balance information of the user;

the sum is the total amount of money put, which is equal to the balance of all users in the current user group.

11. The method of claim 10,

based on the obtained risk model and response model, respectively evaluating each user of the user group to obtain a risk score and a response score, and calculating the gradient of credit promotion data of each user to judge whether to change the promotion level of the credit promotion data, specifically comprising:

dividing the user quota amplitude into n levels, aiming at each user in the user group_jObtaining a user_jCurrent credit line_jInitializing i ═ n, wherein i, j and n are positive integers, and i is less than or equal to n;

obtaining users from an offer risk model and an offer response model_jCorresponding to the current credit line_jAnd amplitude of the promotion_ijRisk of the increased risk score of_ijScored responses to rate of offer_ijAnd corresponding to the current credit line_jAnd the next level of lifting amount_(i-1)jRisk of the increased risk score of_(i-1)jScored responses to rate of offer_(i-1)j；

For each user in the user group_jCalculating the current credit line_jGradient of lift of lowerⁱ _j；

Will be paired withFor each user_jCurrent credit line_jGradient of_ijComparing with a preset first threshold value if the user_jGradient of (2)_ijGreater than or equal to the first threshold, setting i-1, otherwise setting i-i;

and/or the presence of a gas in the gas,

determining the sum of the unused information of the corresponding user group based on the unused information of each user predicted by the obtained prediction model, which specifically comprises the following steps:

for each user in the user group_jObtaining the credit limit of the user from the credit balance model_jAnd amplitude of the promotion_ijThe balance information of_ij(ii) a Calculating a total deposit amount M which is equal to the balance of all users in the current user group_ijThe sum of (a);

and/or the presence of a gas in the gas,

if the level of the credit granting data is determined to be continuously changed according to the sum, repeating the processing procedures of the two previous steps; otherwise, taking the lifting amplitude corresponding to the current lifting level as the final lifting amplitude to optimize the credit granting data of the user, specifically comprising:

comparing the total deposit amount M with a preset second threshold value;

if the total deposit amount M is greater than the second threshold value, setting i to i-1, repeating the steps of calculating a gradient to judge and determining the sum until M is equal to or less than the second threshold value, and determining the amount of the withdrawal at that time_ijAs the final determined magnitude of the boost_ij；

If the total deposit amount M is equal to or less than the second threshold value, the amount increase amplitude determined at this time_ijAs the final determined magnitude of the boost_ij。

12. An apparatus for optimizing trusted data, comprising:

the gradient module is used for respectively evaluating each user of the user group to obtain a risk score and a response score based on the risk model and the response model, and calculating the gradient of credit promotion data of each user so as to judge whether the credit promotion level is changed;

a prediction module for determining a sum of unused information of the corresponding user group based on the unused information predicted for each user by the prediction model;

the condition module is used for repeating the processing processes of the gradient module and the prediction module if the level of credit granting data promotion is determined to be continuously changed according to the sum, otherwise, the promotion amplitude corresponding to the current promotion level is used as the final promotion amplitude to optimize the credit granting data of the user;

the use information is information that the user uses the relevant service data in the service;

the unused information is information of related service data reserved in the service by the user.

13. An electronic device, comprising:

a processor; and

a memory storing computer-executable instructions that, when executed, cause the processor to perform the steps of the method of any of claims 1 to 11.

14. A computer readable storage medium, wherein the computer readable storage medium stores one or more programs which, when executed by a processor, implement the method of any of claims 1-11.