Disclosure of Invention
Aiming at the technical problems that most of the existing e-commerce platforms fail to present results and do not give a detailed push to merchants, and how to solve the cause of the presentation failure, if the presentation failure is caused by credit problems, the platform does not give a specific solution, the invention provides a platform merchant presentation credit management method and a platform merchant presentation credit management system.
The technical scheme adopted by the invention is as follows: a platform merchant presenting limit management method specifically comprises the following steps:
The merchant sends out a request for increasing the cash register on the platform, inputs cash register information related to the increase of the cash register, and outputs an application result of the increase of the cash register;
the application result of the increase of the cash line comprises: the withdrawal limit application fails and the withdrawal limit application succeeds;
and establishing a risk prediction model for predicting whether the platform merchant presentation has risk.
If the request for the credit is failed, the following steps are carried out:
acquiring merchant information stored in a platform, and judging whether the presentation information input by the user is accurate or not based on the merchant information stored in the platform;
acquiring all historical rendering data of a merchant, and extracting features of all the historical rendering data to acquire a plurality of historical rendering feature data;
Acquiring a cash register result of each piece of history cash register data, and recording the cash register result as history cash register result data;
Based on a plurality of historical rendering feature data, a judging and predicting model is established, the input of the judging and predicting model is rendering information of the application rendering, and the output of the judging and predicting model is the failure result reason of the rendering.
Further, based on a plurality of historical demonstration feature data, establishing a judgment prediction model specifically comprises:
establishing a one-to-one correspondence between each piece of historical rendering amount result data and corresponding historical rendering feature data;
Based on the credit-offering results, the historical credit-offering result data are classified, and a plurality of historical credit-offering result data corresponding to each cause of the credit-offering application failure are obtained.
Further, based on a one-to-one correspondence between the historical rendering credit result data and the corresponding historical rendering feature data, acquiring the historical rendering feature data corresponding to each cause of the failure of the rendering credit application;
Carrying out statistical analysis on a plurality of historical rendering feature data corresponding to each cause of the failure of the rendering credit application, and respectively acquiring a plurality of rendering features related to each cause of the failure of the rendering credit application and related weights between each cause of the failure of the rendering credit application and the rendering features;
establishing a feature matching slave model, wherein the feature matching slave model is used for calculating the matching degree of the input withdrawal information of the application withdrawal and the reasons of each withdrawal failure;
Establishing a screening subordinate model, screening one or more reasons of presentation failure which are most matched with presentation information of the application quota based on a calculation result of the feature matching subordinate model by the screening subordinate model, and outputting the reasons as a result;
further, the calculation formula of the correlation weight between the reason of the failure of the withdrawal limit application and the withdrawal feature is as follows:
In the formula, yi is the related weight between the reason of the failure of the withdrawal line application and the ith withdrawal feature, b Total (S) is the total number of the historical withdrawal feature data corresponding to the reason of the failure of the withdrawal line application, and bi is the total number of the historical withdrawal feature data with the ith withdrawal feature in the total number of the historical withdrawal feature data corresponding to the reason of the failure of the withdrawal line application.
Further, the specific calculation formula for screening the subordinate model is as follows:
Wherein, Tj is the matching degree of the input withdrawal information of the application withdrawal and the reason of the failure of the j-th withdrawal request, and Nj is the total number of corresponding historical withdrawal feature data of the reason of the failure of the j-th withdrawal request;
N0 is the total number of features of the input application quota and N As same as is the same total number of features in the history quota extracting feature data corresponding to the reason of the j-th quota extracting application failure;
Alpha'l is the correlation weight between the feature data of the j-th cause of the withdrawal request failure and the withdrawal information of the first input request withdrawal and the same feature in the corresponding history withdrawal feature data of the j-th cause of the withdrawal request failure, and max () is the maximum function.
Further, the method includes inputting the presenting information of the request to the judgment prediction model, and obtaining the cause of the failure of the presenting amount specifically includes:
extracting features of the input data of the application quota extracting information to obtain a plurality of rendering features corresponding to the data of the application quota extracting information;
Invoking a feature matching subordinate model, and respectively calculating the matching degree of the data of the input application quota prompting information and the reasons of failed prompting quota application based on a plurality of prompting features;
Invoking a screening subordinate model, screening out one or more reasons of application failure of the withdrawal amount with the data matching degree of the withdrawal information of the input application withdrawal amount being larger than a preset value, and taking the reasons as reasons of prediction application failure;
the reasons of the failure of the predicted application and the matching degree corresponding to the reasons of the failure of the predicted application are output after a one-to-one correspondence is established, so that the reasons of the failure of the request of the present amount are obtained;
based on the reason of the result of the failure of the withdrawal limit application, generating the withdrawal recommendation scheme through the withdrawal limit recommendation logic comprises the following steps:
judging whether a high risk exists in reasons of the application failure of the withdrawal amount, if not, outputting a recommended solution according to the reasons of the application failure of the withdrawal amount with the largest matching degree with the withdrawal information of the input application withdrawal amount, and if so, outputting the recommended solution according to the high risk type.
Further, the solution for outputting recommendations according to high risk categories specifically includes:
establishing a high-risk rating failure library, and setting a preliminary solution and a rechecking solution for each cause of high-risk rating failure in the high-risk rating failure library based on historical rating experience;
Based on the screened reasons of the rate increase failure, judging whether the matching degree is larger than a first matching preset value, if so, outputting a preliminary solution corresponding to the reasons of the high-risk rate increase failure as a recommended solution, and if not, outputting the preliminary solution corresponding to the reasons of the high-risk rate increase failure and a rechecking solution as recommended solutions.
The preliminary solution may preliminarily address the cause of the forehead lifting failure with high risk;
The reinspection scheme can accurately solve the reason of the forehead lifting failure with high risk.
Further, the risk prediction model is:
Wherein q=1 represents that the platform merchant presents an existing risk, q=0 represents that the platform merchant presents no risk;
h is the prediction probability of the risk prediction model;
aP is an index of risk information existing in the presentation information related to the application of the increase of the presentation amount;
aT is an index of error information existing in the presentation information related to the application of the increase of the presentation amount;
alpha, beta1 and beta2 are all coefficients of the risk prediction model.
Further, the calculation method of alpha, beta1 and beta2 is as follows:
Classifying the historical presentation data according to whether the presentation amount increasing request is successful or not to obtain historical presentation data of failure in presentation amount application in a plurality of sets of presentation amount increasing requests and historical presentation data of success in presentation amount application in a plurality of sets of presentation amount increasing requests;
The historical presentation data of the failed presentation credit application and the historical presentation data of the successful presentation credit application are calculated by a maximum likelihood method according to alpha, beta1 and beta2;
And (3) checking the significance of alpha, beta1 and beta2 on the parameters of the risk prediction model, and judging whether alpha, beta1 and beta2 meet the significance requirement.
Further, a platform merchant presenting credit management system includes:
the application module is used for the platform merchant to apply for a request for increasing the cash amount;
the output module is used for outputting an application result of increasing the cash register amount by the platform;
the prediction module is used for predicting whether the platform merchant presentation has risks or not;
the analysis module is used for solving the cause of the result of the credit withdrawal failure when the credit withdrawal application fails;
The recommending module recommends a solution based on reasons of the credit-on-line failure result;
and the control module is used for controlling all the modules.
The beneficial effects of the invention are as follows: compared with the prior art, in the invention, whether the reasons for the failure of the application of the credit limit of the prediction model are high risk or not is judged, if the reasons for the failure of the application of the credit limit of the prediction model are not high risk, the reasons for the failure of the application of the credit limit are output, and a specific solution is provided to assist the platform in carrying out the management of the credit limit, if the reasons for the failure of the application of the credit limit are high risk, the recommended solution can be output based on the screened reasons for the failure of the credit limit, and if the reasons for the failure of the application of the credit limit are not high risk, another solution is output to assist the user in carrying out the perfect request of the credit limit management of the merchant.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, in order to solve the problems existing in the background art, the present application proposes the following technical scheme: a platform merchant presents a credit management method.
Example 1
The method specifically comprises the following steps:
step S101: the merchant sends out a request for increasing the cash register on the platform, inputs cash register information related to the increase of the cash register, and outputs an application result of the increase of the cash register;
step S102: the application result of the increase of the cash line comprises: the withdrawal limit application fails and the withdrawal limit application succeeds;
step S103: and establishing a risk prediction model for predicting whether the platform merchant presentation has risk.
The method mainly aims at failure in proposition, and performs reason analysis and gives a solution, so that an auxiliary platform helps merchants to manage the amount;
Specific:
step S104: if the request for the credit is failed, the following steps are carried out:
Step S105: acquiring merchant information stored in a platform, and judging whether the presentation information input by a user is accurate or not based on the merchant information stored in the platform (the merchant information comprises the time of entering the platform, the presented bank card number, the name of the presented user, the identification card number and other relevant information for presentation);
step S106: acquiring all historical rendering data of the merchants (namely, data of successful rendering or failure in the histories of all or part of the merchants or the one merchant, wherein the data of failed rendering is mainly acquired), and performing feature extraction on all the historical rendering data to acquire a plurality of historical rendering feature data;
step S107: acquiring a cash register result of each piece of history cash register data, and recording the cash register result as history cash register result data;
Step S108: based on a plurality of historical rendering feature data, a judging and predicting model is established, the input of the judging and predicting model is rendering information of the application rendering, and the output of the judging and predicting model is the failure result reason of the rendering.
In a further design, based on a plurality of historical demonstration feature data, establishing the judgment prediction model specifically comprises:
Step 1, establishing a one-to-one correspondence between each piece of history presenting limit result data and corresponding history presenting feature data, specifically, managing mutual mapping, such as a piece of history presenting feature data with failed data corresponding to failed data;
Step 2, classifying historical withdrawal result data based on withdrawal result (mainly acquired results of withdrawal failure or successful results are acquired and used for recommending auxiliary solutions), and acquiring a plurality of historical withdrawal result data (acquired reasons of withdrawal failure and data of withdrawal application) corresponding to each reason of withdrawal request failure;
Step 3, based on the one-to-one correspondence between the historical rendering amount result data and the corresponding historical rendering feature data, acquiring the historical rendering feature data corresponding to each reason of the failure of the rendering amount application;
step 4, carrying out statistical analysis on a plurality of historical rendering feature data corresponding to each reason for failure of rendering the credit application, and respectively acquiring a plurality of rendering features related to each reason for failure of rendering the credit application and related weights between each reason for failure of rendering the credit application and the rendering features (the reason for failure of rendering the credit application corresponds to the rendering features);
step 5, establishing a feature matching subordinate model, wherein the feature matching subordinate model is used for calculating the matching degree of the input withdrawal information of the application withdrawal and the reasons of each withdrawal failure;
Step 6, establishing a screening subordinate model, screening one or more reasons of presentation failure which are most matched with presentation information of the application quota based on a calculation result of the feature matching subordinate model by the screening subordinate model, and outputting the reasons as a result;
The specific results are as follows:
the calculation formula of the correlation weight between the reason of the failure of the withdrawal limit application and the withdrawal feature is as follows:
In the formula, yi is the related weight between the reason of the failure of the withdrawal line application and the ith withdrawal feature, b Total (S) is the total number of the historical withdrawal feature data corresponding to the reason of the failure of the withdrawal line application, and bi is the total number of the historical withdrawal feature data with the ith withdrawal feature in the total number of the historical withdrawal feature data corresponding to the reason of the failure of the withdrawal line application.
The whole explanation of the above technical scheme is as follows:
In some embodiments, the total number of all the historical presentation feature data corresponding to the reason a of the application failure of the presentation credit (e.g. a is insufficient in credibility and the time of the merchant entering the platform does not reach the set time) is 1000, wherein 200 pieces of the historical inquiry feature data each include the presentation feature b, and the correlation weight between the reason a of the application failure and the presentation feature b is 200/1000=0.2
In a further design, the specific calculation formula for screening the subordinate model is as follows:
Wherein, Tj is the matching degree of the input withdrawal information of the application withdrawal and the reason of the failure of the j-th withdrawal request, and Nj is the total number of corresponding historical withdrawal feature data of the reason of the failure of the j-th withdrawal request;
N0 is the total number of features of the input application quota and N As same as is the same total number of features in the history quota extracting feature data corresponding to the reason of the j-th quota extracting application failure;
Alpha'l is the correlation weight between the feature data of the j-th cause of the withdrawal request failure and the withdrawal information of the first input request withdrawal and the same feature in the corresponding history withdrawal feature data of the j-th cause of the withdrawal request failure, and max () is the maximum function.
The following explanation is made for the above technical scheme: according to the scheme, the matching degree calculation is performed based on the two indexes, on one hand, the matching degree of the feature types is the matching degree of the characteristics, on the other hand, the correlation weight of the overlapping characteristics and the reasons of application failure, and the matching degree of the application upgrading information and the reasons of application failure is calculated through the combination of the two indexes, so that the types of the application failure reasons possibly existing in the application upgrading information can be accurately identified.
In a further design, the method for obtaining the failure result of the withdrawal amount specifically includes the steps of:
Step one, extracting features of input data of the application quota raising information to obtain a plurality of quota raising features corresponding to the data of the application quota raising information;
Step two, calling a feature matching subordinate model, and respectively calculating the matching degree of the data of the input application quota withdrawal information and the reasons of each withdrawal quota application failure (the most specific reasons of withdrawal failure can be conveniently obtained) based on a plurality of withdrawal features;
Step three, a screening subordinate model is called, and one or more reasons for the failure of the application of the withdrawal amount, of which the data matching degree with the withdrawal information of the input application withdrawal amount is larger than a preset value, are screened out and used as the reasons for the failure of the prediction application; (because the merchant has one or more reasons for failure when the platform fails to present, for example, failure caused by incorrect information filling, failure caused by insufficient credit, failure caused by not self-embodiment, failure caused by risk of presenting environment, failure caused by non-adult operation embodiment, and the like);
Fourth, the reasons of the failure of the predicted application and the matching degree corresponding to the reasons of the failure of the predicted application are output after a one-to-one correspondence is established, and the reasons of the result of the failure of the application of the withdrawal limit are obtained;
step five, generating a presenting recommendation scheme through presenting quota recommendation logic based on the reason of the result of presenting quota application failure comprises the following steps:
Firstly, judging whether a high risk exists in reasons of the application failure of the withdrawal amount (the high risk refers to whether the withdrawal environment has risks or not, whether the withdrawal environment has suspicion of being stolen by hacking operation or not), if not, outputting a recommended solution according to the reasons of the application failure of the withdrawal amount with the maximum matching degree with the withdrawal information of the input application withdrawal amount (for example, the withdrawal failure caused by the fact that the merchant is not long in residence time but insufficient in reliability, the outputted recommended solution is to wait for the merchant, increase residence time and then perform other operations to improve reliability, thereby completing the withdrawal), if yes, outputting the recommended solution according to the high risk type (for example, if the withdrawal environment has risks, suspicion of being controlled by hacking, the merchant user is required to complete operations such as various personal verification, login password modification and the like, thereby maintaining a good withdrawal environment).
In another design;
The solution for outputting recommendations according to high risk categories specifically includes:
Step a, a high-risk rating failure library is established (the high-risk rating failure library stores various reasons for high-risk rating failures), and a preliminary solution and a rechecking scheme are set for each reason for high-risk rating failure in the high-risk rating failure library based on historical rating experience;
And b, judging whether the matching degree is larger than a first matching preset value based on the screened reasons of the rate increasing failure, if so, outputting a primary solution corresponding to the reasons of the high-risk rate increasing failure as a recommended solution, and if not, outputting the primary solution corresponding to the reasons of the high-risk rate increasing failure and a reinspection solution as the recommended solution.
Step c, the primary solution can primarily solve the reason of the forehead raising failure with high risk;
step d, the reinspection scheme can accurately solve the reason of the forehead lifting failure with high risk.
The following explanation is made for the above technical scheme:
In the scheme, if high risk exists in reasons of the withdrawal request failure, outputting reasons of the withdrawal failure if the high risk does not exist, and providing a specific solution to assist a platform to assist a merchant in carrying out the withdrawal management, if the reasons of the withdrawal request failure exist, judging whether the matching degree of the reasons of the withdrawal request failure is larger than a first matching preset value or not based on the screened reasons of the withdrawal request failure, if the reasons of the withdrawal request failure are larger than the first matching preset value, outputting a preliminary solution corresponding to the reasons of the withdrawal request failure of the high risk as a recommended solution, and if the reasons of the withdrawal request failure of the high risk are not larger than the first matching preset value, outputting a preliminary solution corresponding to the reasons of the withdrawal request failure of the high risk as a recommended solution; the primary solution can primarily solve the reason of the high-risk forehead lifting failure, and the reinspection solution can accurately solve the reason of the high-risk forehead lifting failure.
Example two
In this embodiment, step S103 is performed in the embodiment;
In order to embody whether the predicted merchant has risk in the platform, a specific risk prediction model is:
Wherein q=1 represents that the platform merchant presents an existing risk, q=0 represents that the platform merchant presents no risk;
h is the prediction probability of the risk prediction model;
AP is an index of risk information existing in the presentation information related to the application of the increase of the presentation amount (for example, high risk exists, and the high risk refers to whether the presentation environment is at risk or not, and whether the presentation environment is suspected of being stolen by hacking or not);
aT is an index of error information (such as filled information is in error) existing in the presentation information related to the application of the increase of the presentation amount;
alpha, beta1 and beta2 are all coefficients of the risk prediction model.
Specific:
The calculation method of alpha, beta1 and beta2 comprises the following steps:
Firstly, classifying historical withdrawal data according to whether withdrawal request is successful or not, and obtaining historical withdrawal data of withdrawal request failure in a plurality of groups of withdrawal request and historical withdrawal data of withdrawal request success in a plurality of groups of withdrawal request;
Secondly, carrying out calculation of alpha, beta1 and beta2 by using a maximum likelihood method according to historical presentation data of failed presentation credit application and a plurality of groups of historical presentation data of successful presentation credit application;
And (3) checking the significance of alpha, beta1 and beta2 on the parameters of the risk prediction model, and judging whether alpha, beta1 and beta2 meet the significance requirement.
The risk prediction model is established based on a Logistic regression model principle, and a fault risk prediction model is established, wherein the Logistic regression model is a generalized linear regression analysis model and is commonly used in the fields of data mining, result prediction and the like;
According to the risk prediction model provided by the scheme, whether the merchant presents risks or whether the risk can be presented to the platform or not is calculated, and the merchant can be assisted in timely and fast self-checking according to the presentation failure of prediction calculation and the probability of the risks, particularly when the presentation environment has high risks, the risk prediction can be performed through the model.
Example III
A platform merchant credit-offering management system, comprising: the system comprises an application module, an output module, a prediction module, an analysis module, a recommendation module and a control module.
Specific:
the application module is used for a platform merchant to apply for a request for increasing the credit limit;
The output module is used for outputting an application result of increasing the cash register amount by the platform;
The prediction module is used for predicting whether the platform merchant presentation has risks or not;
when the analysis module fails to apply for the credit, the cause of the credit failure result is checked;
the recommending module recommends a solution based on reasons of the result of the withdrawal limit failure;
The control module is used for controlling all modules.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.