CN113298394A - Automobile financial post-loan risk early warning method - Google Patents

Abstract

The application provides a risk early warning method after automobile financial loan, which comprises the following steps: constructing a machine learning model for early warning; formulating an early warning strategy rule for early warning; inputting relevant data of the automobile financial customer into the machine learning model; outputting a risk early warning grade according to data output by the machine learning model and an early warning strategy rule; and matching corresponding early warning measures according to the risk early warning grade. The method has the beneficial effects that the automobile financial post-loan risk early warning method is provided, which integrates the vehicle consumption staging scene so as to find the post-loan risk in time.

Description

Automobile financial post-loan risk early warning method

Technical Field

The application relates to a risk early warning method after automobile financial loan.

Background

With the increase of the marketing level and the open degree of the financial industry, the competition among financial institutions is increased, and the franchise value of the bank is continuously reduced. Meanwhile, along with the fluctuation of domestic economic situation and the deterioration of the operating condition of enterprises, the balance of bad loans of commercial banks is gradually increased, and the reject ratio gradually approaches the supervision requirement. The traditional bank is uniformly managed by a head office after loan, and branches can not control risks after loan to a certain extent, so that only risks before loan are controlled in many cases, and customers who deteriorate in the middle of economy after loan can not be distinguished, and can know the risks after the loan actually occurs and enter a receiving stage. And the terminal of the collection is late, and most results can only form a bad account and a bad account. Under the condition, how to take risk prevention after loan and identify the loan in transit about to generate risk in advance is the key importance of commercial banks to improve the core competitiveness and reduce the overall reject ratio.

Disclosure of Invention

In order to overcome the defects in the prior art, the application provides an automobile financial post-loan risk early warning method, which comprises the following steps: constructing a machine learning model for early warning; formulating an early warning strategy rule for early warning; inputting relevant data of an automobile financial customer to the machine learning model; outputting a risk early warning grade according to the data output by the machine learning model and the early warning strategy rule; and matching corresponding early warning measures according to the risk early warning grade.

Further, the constructing of the machine learning model for early warning includes the following steps: and selecting a model template for the machine learning model.

Further, the constructing of the machine learning model for early warning includes the following steps: sample data is prepared for the machine learning model.

Further, preparing training data for the machine learning model comprises the steps of: and dividing the sample data into good samples and bad samples according to the repayment condition of the automobile financial customer.

Further, preparing training data for the machine learning model further comprises the steps of: and preparing the sample data, processing indexes and cleaning the data.

Further, in preparing the sample data, the data dimensions of the sample data include: application data, overdue data after loan, vehicle loan related data, customer basic data, income and debt data and customer historical deduction data.

Further, in preparing the sample data, the data dimension of the sample data further includes: three-party external data, mortgage rate data, post-credit collection data and post-credit investigation data.

Further, the index processing of the sample data includes: and counting the number data of the times of the success of the non-deduction within one year and the average expected days.

Further, the data cleaning of the sample data comprises missing value filling and abnormal value processing.

Further, the making of an early warning strategy rule for early warning comprises strategy analysis and strategy making.

The application has the advantages that: the automobile financial post-loan risk early warning method is provided, which integrates vehicle consumption staging scenes to facilitate timely finding of post-loan risks.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this application, serve to provide a further understanding of the application and to enable other features, objects, and advantages of the application to be more apparent. The drawings and their description illustrate the embodiments of the invention and do not limit it. In the drawings:

fig. 1 is a block diagram illustrating steps of a risk pre-warning method after automobile financial loan according to an embodiment of the application.

Detailed Description

In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only partial embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It should be understood that the data so used may be interchanged under appropriate circumstances such that embodiments of the application described herein may be used. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

In this application, the terms "upper", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings. These terms are used primarily to better describe the present application and its embodiments, and are not used to limit the indicated devices, elements or components to a particular orientation or to be constructed and operated in a particular orientation.

Moreover, some of the above terms may be used to indicate other meanings besides the orientation or positional relationship, for example, the term "on" may also be used to indicate some kind of attachment or connection relationship in some cases. The specific meaning of these terms in this application will be understood by those of ordinary skill in the art as appropriate.

Furthermore, the terms "mounted," "disposed," "provided," "connected," and "sleeved" are to be construed broadly. For example, it may be a fixed connection, a removable connection, or a unitary construction; can be a mechanical connection, or an electrical connection; may be directly connected, or indirectly connected through intervening media, or may be in internal communication between two devices, elements or components. The specific meaning of the above terms in the present application can be understood by those of ordinary skill in the art as appropriate.

It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.

Referring to fig. 1, the application provides a method for early warning risk after financial loan of an automobile, which includes the following steps: constructing a machine learning model for early warning; formulating an early warning strategy rule for early warning; inputting relevant data of the automobile financial customer into the machine learning model; outputting a risk early warning grade according to data output by the machine learning model and an early warning strategy rule; and matching corresponding early warning measures according to the risk early warning grade.

Specifically, the construction of a machine learning model for early warning comprises the following steps: and selecting a model template for the machine learning model.

Specifically, the construction of a machine learning model for early warning comprises the following steps: sample data is prepared for the machine learning model.

Specifically, preparing training data for a machine learning model includes the steps of: and dividing the sample data into good samples and bad samples according to the repayment condition of the automobile financial customer.

Specifically, preparing training data for the machine learning model further comprises the steps of: and preparing sample data, processing indexes and cleaning the data.

Specifically, when sample data is prepared, the data dimensions of the sample data include: application data, overdue data after loan, vehicle loan related data, customer basic data, income and debt data and customer historical deduction data.

Specifically, when preparing sample data, the data dimension of the sample data further includes: three-party external data, mortgage rate data, post-credit collection data and post-credit investigation data.

Specifically, the index processing of the sample data includes: and counting the number data of the times of the success of the non-deduction within one year and the average expected days.

Specifically, data cleansing of sample data includes missing value population and outlier processing.

Specifically, the early warning strategy rule for early warning is formulated and comprises strategy analysis and strategy formulation.

As a specific embodiment, the method of the present application comprises the following two parts:

design of model

Model design scheme

And dividing the customers according to new vehicles/used vehicles, first-stage vehicles/non-first-stage vehicles, and then modeling each customer group respectively.

(II) model sample selection

(III) determining target variables and defining good and bad samples

And defining that the client does not pay for the bill 23 days after the bill is issued as a bad sample, and otherwise, the client does not pay for the bill as a good sample.

(IV) data preparation, index processing, and data cleaning

The data dimensions mainly include:

(a) application data: applying for grading, and classifying customers into groups;

(b) post-loan overdue data: accumulating the number of overdue times, the maximum number of overdue days, the current number of expired times, etc.;

(c) vehicle loan related data: first payment proportion, staging period number, monthly supply and the like;

(d) client base data: including age at the time point, gender;

(e) income liability data: monthly liability, income;

(f) client historical deduction data: including accumulating the number of outstanding deductions, etc.;

(g) three-party external data sources: including FICO score, medium integrity score, etc.;

(h) mortgage class data;

(i) post-loan collection of data;

(j) credit investigation data for post-credit management;

index processing: including statistical indicators such as the number of times of unsubscribed money in the last 1 year, average number of overdue days, etc.;

data cleaning: including missing value padding, outlier processing, etc.

(V) feature engineering

The method comprises the steps of characteristic correlation test, VIF test, characteristic screening, characteristic derivation, univariate analysis and the like;

and (3) correlation test: correlation > =0.7, culling variables with small iv values;

and (4) VIF (visual inspection): characteristic VIF < = 5;

and (3) feature screening:

(a) IV screening, IV > = 0.02;

(b) and (4) random forest feature importance screening, wherein the first 25 features are high in importance.

(VI) model development and evaluation

And (3) algorithm selection: and integrating a learning algorithm. Such as XGBoost, LightGBM, etc.;

evaluation indexes are as follows: ks > =0.35, AUC > =0.75, mode-entering characteristic PSI < 0.1;

and (3) effect measurement: and (4) performing swap in/swap out analysis.

(VII) model review, deployment and acceptance

And (3) evaluating the model: outputting a model document and a model effect measuring and calculating report;

model deployment and acceptance:

(a) inputting and outputting a model characteristic processing requirement document;

(b) output model files, such as PMML files, score card documents;

(c) after the development is completed, performing field analysis acceptance and model result acceptance.

(eighth) model monitoring

The monitoring indexes are as follows: and if the variable PSI exceeds 0.1 or the KS value is lower than 0.3, eliminating the corresponding variable from the model to iterate the model.

Second, strategy design

Strategy analysis:

(a) pre-credit risk analysis;

(b) analyzing repayment capacity after loan;

income: recent social security/public deposit payment record;

negative debt: newly adding loan in X months, using rate of the credit card in X months, paying amount required in X months in the future, and the like.

Fund tension: credit card, loan application times, network loan application times, loan APP activity, whether a large amount of loans payed in one time or not in the near X months/day, and the like.

Behavior: whether to gamble consumption records such as lotteries and the like recently, whether to be complained in the court, and the like;

recording the performance: recent overdue amount, recent repayment amount, etc.

(c) Post-loan payment willingness analysis

And (4) repayment recording: automatic deduction record, active repayment record, other people compensation record, delay record and the like;

and (4) collection and recording: record of connection, such as an offer of payment record;

and (3) accelerating the collection of the label: phone status tags, communication content tags (e.g., deny loan, malicious repudiation, etc.), contact/co-Return feedback tags (e.g., loss of contact, legal case, accident, etc.);

slight risk: forget to pay, and the habitual overdue.

(d) Process risk analysis

Mortgage recording: car purchase mortgage transaction records, car mortgage and pledge loan records after car purchase and the like;

channel monitoring: risks after channel loan, cases of channel complaints, channel management states and the like;

vehicle information: the track of the vehicle is inconsistent with the actual active address of the client, whether the vehicle has related consumption after the vehicle is purchased or not, and the like;

relation network: and the repayment records of the network clients where the users are located, and the like.

(II) strategy making:

(a) and making an early warning disposal strategy based on the early warning model classification and strategy analysis results. And determining the risk level of the user by the risk level + rule, and determining an early warning handling mode.

(b) The early warning level is divided into four types, namely risk A level, AI intelligent voice reminding and risk B level, short message risk C level is reminded in a pushing mode, slow processing is carried out, if overdue occurs, AI reminding is carried out, risk D level is carried out, and processing is not carried out.

As can be seen from the foregoing, the present application aims to identify an in-transit loan that is about to generate a risk in advance by using a post-loan model early warning inspection mechanism and applying a method of combining quantitative analysis and qualitative analysis, to allow a local bank to participate in asset disposition of post-loan risk customers, to take targeted processing measures for loans of different risk levels, to take timely precaution, control and preposition risk processing, and to reduce post-loan loss.

The technical scheme is as follows: firstly, a risk early warning model is established, and early warning scores are scored for each user. The early warning model passes through an ensemble learning algorithm. For example, XGboost and LightGBM develop a set of models for identifying risk occurrence probability of paid and unfinished clients, and comprehensively judge whether the clients are about to generate overdue risks or not by means of application scoring, post-loan recording, basic information of the users and the like and dispose under the condition that the risks are not completely outbreak, so that the probability of overdue occurrence is reduced. The risk that the early warning model is mainly preventing is as follows:

pre-loan risk (misplacement): pre-loan model scoring, pre-loan income/liability, etc

② ability risk of repayment: high liability, sudden events, etc., resulting in insufficient payment sources

③ risk of repayment willingness: multiple overdue, refusal to urge to accept, malicious repudiation, etc

Fourthly, process risk: the participation of channels, intermediaries, etc. may result in a client not actually purchasing cars, not transacting mortgages on time, etc

And secondly, the existing data of the bank is fully mined, and a rule strategy which is possibly missed by a decision tree multi-dimensional cross formulation model is adopted. For example, the credit investigation message data is used for cleaning and characteristic engineering processing, and data which may influence the client to generate repayment risk, such as user's unionpay data, accumulation fund/social security data, and data of court, post-loan behavior, collection log, three-party multi-head life, etc., are collected.

Thirdly, according to different risks possibly generated under the analysis condition, strategy situations are customized, such as short message reminding, ivr voice reminding, manual intervention and the like, a coping strategy is output, and loss is timely recovered.

Generally, the machine learning model outputs a credit score of each client based on input variable characteristics of each client, and the early warning policy rules add rule items on the basis of the score output by the machine learning model to warn the client, for example, the model outputs a client credit score of 600, and the policy setting model score < =600 warns the client. And the policy can also output rules independently of the model, such as rule: if the age is greater than 50 and the M1 times are more than 1 after 3 months, the client who hit the rule will be warned.

By adopting the scheme, a set of powerful risk early warning system can be provided for the bank vehicle loan business, in addition, the scheme of the application can timely predict the risk points which may occur in the credit business, high-risk customers can be found in advance, so that the high-risk customers can be intervened and disposed as soon as possible, corresponding coping strategies are formulated according to different risks, and the overdue rate of M1 is reduced.

The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims (10)

1. A risk early warning method after automobile financial loan is characterized by comprising the following steps:

the automobile financial post-loan risk early warning method comprises the following steps:

constructing a machine learning model for early warning;

formulating an early warning strategy rule for early warning;

inputting relevant data of an automobile financial customer to the machine learning model;

outputting a risk early warning grade according to the data output by the machine learning model and the early warning strategy rule;

and matching corresponding early warning measures according to the risk early warning grade.

2. The method of claim 1, wherein the method comprises:

the construction of the machine learning model for early warning comprises the following steps:

and selecting a model template for the machine learning model.

3. The method of claim 2, wherein the method comprises:

the construction of the machine learning model for early warning comprises the following steps:

sample data is prepared for the machine learning model.

4. The method of claim 3, wherein the method comprises:

preparing training data for the machine learning model comprises the steps of:

and dividing the sample data into good samples and bad samples according to the repayment condition of the automobile financial customer.

5. The method of claim 4, wherein the method comprises:

preparing training data for the machine learning model further comprises the steps of:

and preparing the sample data, processing indexes and cleaning the data.

6. The method of claim 5, wherein the method comprises:

in preparing the sample data, the data dimensions of the sample data include:

application data, overdue data after loan, vehicle loan related data, customer basic data, income and debt data and customer historical deduction data.

7. The method of claim 6, wherein the method comprises:

in preparing the sample data, the data dimensions of the sample data further include:

three-party external data, mortgage rate data, post-credit collection data and post-credit investigation data.

8. The method of claim 7, wherein the method comprises:

the index processing of the sample data comprises the following steps: and counting the number data of the times of the success of the non-deduction within one year and the average expected days.

9. The method of claim 8, wherein the method comprises:

and the data cleaning of the sample data comprises missing value filling and abnormal value processing.

10. The method of claim 9, wherein the method comprises:

the early warning strategy rule for early warning is formulated and comprises strategy analysis and strategy formulation.

Images