CN114240595A - Anti-fraud model taking recognition repayment capability as core - Google Patents

Abstract

The invention discloses an anti-fraud model taking recognition repayment capability as a core, which solves the technical problems of the prior model that the model is not strict enough, the risk recognition is not perfect enough and the accuracy rate is low; reading the identity card number of a client to be borrowed and credited, and initializing the score R =0 of the client to be checked; extracting corresponding data of a related database by taking the identity card number as an index; if the identification number hits fields related to social stability and the like, directly returning a negative result, and if the identification number does not hit fields, entering the process; if the identity card number hits and relates to feature 4, every hit is R + 20; if a hit involves feature 5, every hit is one R + 15; obtaining the total value of R accumulated in the step; if R > =60, returning a negative result; if 0< R <60, returning a judicious result and recommending the financial institution to carry out further scrutiny; if R =0, returning a continued result. The invention combs a logic backbone from a plurality of influence factors to identify a real risk source, and the logic backbone can be used as a base stone and a preposed input of the existing anti-fraud model system.

Description

Anti-fraud model taking recognition repayment capability as core

Technical Field

The invention relates to the technical field of financial anti-fraud, in particular to an anti-fraud model taking recognition repayment capability as a core.

Background

The relatively common loan auditing of financial institutions is to evaluate credit based on authorization information of users applying for loan services, establish a static or dynamic rule base constructed based on artificial experience according to a mode of combining experience and credit evaluation data, establish a credit risk model based on rules by using data mainly from people's bank credit investigation, asset information, a known blacklist base and the like, and refuse loan application when the output result of the model is poor or a service risk code is hit. The other machine learning model represented by a scoring card model is mainly used, a plurality of characteristics and a multi-party data source are constructed to serve as a training set, a verification set and a test set, and the model has methods of random forests, XGboost, graph neural networks, transfer learning, unsupervised clustering and the like.

The existing rule model does not take rapid loss of repayment capacity as a core, lacks deep cognition on financial fraud risk control, does not establish a set of strict and effective risk identification system, mainly piles up rules refined by scattered experience, and detects the established rules when hit, but most models cannot effectively deal with continuously changing fraud means.

The model based on the machine learning method depends on a large amount of user data and third-party data, so that the legality of the third-party data is problematic, the continuity of the business cannot be guaranteed, the fraud samples are insufficient in sufficient quantity, the sample is unbalanced in the model training process particularly when a novel fraud problem is faced, the accuracy and recall rate in prediction cannot meet the actual requirements, and the generalization capability is poor when different financial business scenes are dealt with. In addition, machine learning models are currently mainly used to identify credit risk issues.

Disclosure of Invention

The invention aims to provide an anti-fraud model taking recognition of rapid loss of repayment capacity as a core, the key of the model lies in deep cognition and high abstraction of financial risk control based on years of industry and actual working experience, a logic backbone is combed from a plurality of influence factors to identify a real risk source, and the model can be used as a base stone and a preposed input of the existing anti-fraud model system.

In order to achieve the above purpose, the anti-fraud model with the payment identification capability as the core provided by the invention comprises the following specific steps;

s1, reading the identity card number of the client to be borrowed and initializing the score R =0 of the client to be checked;

s2, extracting corresponding data of the relevant database by taking the ID card number as an index;

s3, if the identity card number hits the related feature 1, directly returning a rejection result, reporting the related condition to related departments, avoiding the credit and debit political risks, and if not, entering the next step;

s4, if the identity card number hits and relates to the feature 2, directly returning a rejection result, and reporting the relevant condition to the relevant department, if not, entering the next step;

s5, if the identity card number hits and relates to the feature 3, directly returning a rejection result, and if not, entering the next step;

s6, if the identity number hits and relates to the feature 4, every hit is R + 20; if a hit involves feature 5, every hit is one R + 15; if hit and involve the other fields which threaten the repayment ability, hit every R +10, get R total score accumulated in this step;

s7, if R > =60, returning a negative result; if 0< R <60, returning a judicious result and recommending the financial institution to carry out further scrutiny; if R =0, returning a continued result, and determining whether further examination is needed by the financial institution.

The core logic of the invention is as follows: (1) credit risk is relatively stable, fraud risk fluctuation is larger, influence on bank badness rate is more important, loan audit is critical to identify fraud risk, but most anti-fraud models focus on credit risk level at present; (2) the key to identifying the risk of fraud is to identify the determinant of the rapid loss of repayment ability, and ignoring this factor is also an important reason for the poor effect of other anti-fraud models; (3) the ability to identify repayment includes not only conventional evidence of assets, but also non-asset factors that are often ignored, and parties involved in criminal crime cases often have a higher risk of such kind; (4) in view of the processing rule that criminals are superior to civilian, the model considers that the rapid loss of repayment capacity through criminal factor recognition can be used as the high priority of an anti-fraud model; (5) the lender which loses repayment ability quickly can be identified according to the data of public security criminal crime, drug-related, gambling, foretell and the like.

The invention provides an anti-fraud model taking the recognition of the fast change probability of repayment capacity as the core, and focuses on the key problem of predicting the loss of repayment capacity of loan customers, so that a real fraud risk source is grasped, a 'one-ticket denial' result can be provided without manual examination, the anti-fraud model system can be used as a basic stone and a preposed input of the existing anti-fraud model system, the anti-fraud model system is integrated with other effective anti-fraud models, and other payment model interfaces are not required to be called when the 'one-ticket denial' result occurs.

Drawings

In order to more clearly illustrate the embodiments of the present application or technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present invention, and other drawings can be obtained by those skilled in the art according to the drawings.

FIG. 1 is a flow chart of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, those skilled in the art will now describe the present invention in further detail with reference to the accompanying drawings.

The invention provides an anti-fraud model taking the ability of identifying repayment as a core, which is shown in figure 1 and comprises the following specific steps;

s1, reading the identity card number of the client to be borrowed and initializing the score R =0 of the client to be checked;

s2, extracting corresponding data of the relevant database by taking the ID card number as an index;

s3, if the identity card number hits the related feature 1, directly returning a rejection result, reporting the related condition to related departments, avoiding the credit and debit political risks, and if not, entering the next step;

s4, if the identity card number hits and relates to the feature 2, directly returning a rejection result, and reporting the relevant condition to the relevant department, if not, entering the next step;

s5, if the identity card number hits and relates to the feature 3, directly returning a rejection result, and if not, entering the next step;

s6, if the identity number hits and relates to the feature 4, every hit is R + 20; if a hit involves feature 5, every hit is one R + 15; if hit and involve the other fields which threaten the repayment ability, hit every R +10, get R total score accumulated in this step;

s7, if R > =60, returning a negative result; if 0< R <60, returning a judicious result and recommending the financial institution to carry out further scrutiny; if R =0, returning a continued result, and determining whether further examination is needed by the financial institution.

The key point of the anti-fraud model disclosed by the invention is that an anti-fraud logic system is constructed by identifying the probability of losing repayment capacity rapidly, and the anti-fraud logic system comprises the following steps: (1) the credit risk is relatively stable, the fraud risk is more important, and the loan audit key is to identify the fraud risk rather than the credit risk; (2) the key to identifying the risk of fraud is the determining factor of the rapid loss of ability to identify repayment; (3) the ability to identify repayment includes not only conventional evidence of assets, but also non-asset factors that are often ignored, and parties involved in criminal crime cases often have a higher risk of such kind; (4) given the criminal's rules over the civilian, such risk of quickly losing repayment ability may be a high priority as an anti-fraud model; (5) the lender which loses repayment ability quickly can be identified according to the data of public security criminal crime, drug-related, gambling, foretell and the like.

While certain exemplary embodiments of the present invention have been described above by way of illustration only, it will be apparent to those of ordinary skill in the art that the described embodiments may be modified in various different ways without departing from the spirit and scope of the invention. Accordingly, the drawings and description are illustrative in nature and should not be construed as limiting the scope of the invention.

Claims (1)

1. An anti-fraud model with payment recognition capability as a core, characterized in that: the method comprises the following specific steps;

s1, reading the identity card number of the client to be borrowed and initializing the score R =0 of the client to be checked;

s2, extracting corresponding data of the relevant database by taking the ID card number as an index;

s3, if the identity card number hits the related feature 1, directly returning a rejection result, reporting the related condition to related departments, avoiding the credit and debit political risks, and if not, entering the next step;

s4, if the identity card number hits and relates to the feature 2, directly returning a rejection result, and reporting the relevant condition to the relevant department, if not, entering the next step;

s5, if the identity card number hits and relates to the feature 3, directly returning a rejection result, and if not, entering the next step;

s6, if the identity number hits and relates to the feature 4, every hit is R + 20; if a hit involves feature 5, every hit is one R + 15; if hit and involve the other fields which threaten the repayment ability, hit every R +10, get R total score accumulated in this step;

s7, if R > =60, returning a negative result; if 0< R <60, returning a judicious result and recommending the financial institution to carry out further scrutiny; if R =0, returning a continued result, and determining whether further examination is needed by the financial institution.

Images