CN113793212A - Credit assessment method - Google Patents

Credit assessment method Download PDF

Info

Publication number
CN113793212A
CN113793212A CN202111123065.1A CN202111123065A CN113793212A CN 113793212 A CN113793212 A CN 113793212A CN 202111123065 A CN202111123065 A CN 202111123065A CN 113793212 A CN113793212 A CN 113793212A
Authority
CN
China
Prior art keywords
model
variables
final
variable
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111123065.1A
Other languages
Chinese (zh)
Inventor
赵茜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing Fumin Bank Co Ltd
Original Assignee
Chongqing Fumin Bank Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chongqing Fumin Bank Co Ltd filed Critical Chongqing Fumin Bank Co Ltd
Priority to CN202111123065.1A priority Critical patent/CN113793212A/en
Publication of CN113793212A publication Critical patent/CN113793212A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/03—Credit; Loans; Processing thereof
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Abstract

The invention relates to the technical field of data modeling, in particular to a credit assessment method for constructing customer-level consumption and transaction portrait variables and predicting credit risk of a customer by using bank card transaction data acquired after the customer authorizes. The credit evaluation method comprises the steps of variable derivation, wherein variables of transaction preference, transaction activity, consumption level and risk transaction are derived based on the acquired bank card transaction running data; screening variables, namely screening the variables according to the missing values, the missing rates and the IV values of the variables; model fitting, namely performing model training on any group of hyper-parameter combinations by using a training set to obtain a model, calculating an AUC (AUC) value of the model in a test set, taking the hyper-parameter of the model with the highest AUC value in the test set as a hyper-parameter final value, and taking a variable used by the model as a final model entering variable; and combining the training set and the test set, performing model fitting to obtain a final model, and calculating an AUC value on the final model.

Description

Credit assessment method
Technical Field
The invention relates to the technical field of data modeling, in particular to a credit evaluation method based on an XGB algorithm model.
Background
The development of database technology, mathematical statistics technology and computer technology provides a scientific and technological basis for the development and application of scoring models. The credit scoring model taking the prediction model as the core systematically predicts the future credit performance of the customers by mining the behavior characteristics, credit worthiness and the like of the customers and applying a mathematical statistics technology, and queues the customers according to the performance by scores as a decision basis. The prediction model is popularized and used in the seventies and eighties of the 20 th century, and is still the most widely used and well-developed technology in credit management to date. With the increasing improvement of credit investigation information and the expansion of coverage groups, credit scoring models taking credit investigation as a core are widely applied in the fields of automobile credit, personal credit cards, small and micro business credit and the like. However, the core of credit investigation information is mainly the credit card of the client, monthly loan repayment behavior, mainly credit behavior, but does not include other consumption behavior, payment behavior, etc. which may be related to the credit performance of the client.
Disclosure of Invention
The invention aims to provide a credit assessment method for predicting credit risk of a client by constructing client-level consumption and transaction portrait variables by using bank card transaction data acquired after the client authorizes.
The invention relates to a credit evaluation method, which comprises the following steps,
acquiring data, namely acquiring bank card transaction flow data of each customer based on the queried and authorized customer sample;
deriving variables, namely processing a client-level behavior label based on the acquired bank card transaction running data to derive variables of transaction preference, transaction activity, consumption level and risk transaction;
screening variables, namely screening the variables according to the missing values, the missing rates and the IV values of the variables;
model fitting, namely dividing a sample set into a training set, a testing set and a cross-time verification set, selecting an optimal hyper-parameter combination by adopting a grid search method in the model fitting, performing model training on any group of hyper-parameter combinations by using the training set to obtain a model when performing grid search, calculating an AUC value of the model in the testing set, taking a hyper-parameter of the model with the highest AUC value in the testing set as a final hyper-parameter value, and taking a variable used by the model as a final in-mode variable; and combining the training set and the testing set to form a combined set, performing model fitting on the basis of the determined final value of the hyper-parameter and the final in-mold variable to obtain a final model, and calculating an AUC value on the final model.
The invention has the beneficial effects that: after the credit customer authorization is obtained, the transaction flow information of all bank cards under the name of the customer is obtained, the customer-level behavior tags are processed, and variables of the categories such as transaction preference, transaction activity, consumption level and risk transaction are derived. And matching credit performances of corresponding customers, and establishing a customer default probability prediction model by using an XGB algorithm.
Further, in the variable screening, embedded screening by using an XGB algorithm is further included to screen the variables, more screening results can be obtained by using the embedded screening, the embedded screening is performed in multiple rounds, and each round is performed as follows:
randomly dividing a sample set into a training set and a testing set;
selecting an XGB algorithm model, setting the XGB algorithm model hyperparameters, wherein the evaluation index of the XGB algorithm model is AUC;
carrying out model training by adopting a training set;
and evaluating the model effect of the XGB algorithm on the test set, calculating an AUC value, terminating the screening iteration if the AUC value of the round is greatly reduced compared with that of the previous round, taking the variable screened in the previous round as a candidate variable for model fitting, and otherwise, continuing to screen the variable.
Further, the hyper-parameters of the XGB algorithm model are set as follows: the tree depth is set to 3, 4 or 5, the maximum iteration step number is set to 100, 150 or 200, the learning rate is set to 0.01, 0.05 or 0.1, Gamma is set to 0.1, 0.5 or 1, the column random draw ratio is set to 0.5, 0.6 or 0.8, the sample random draw ratio is set to 0.5, 0.6 or 0.8, and the early termination step number is set to 20. To select the optimal number of iterations.
Further, after the effect of the XGB algorithm model is evaluated on the test set, the method further includes: and directly eliminating the variables with the application times of 0, sorting the rest variables according to the importance degree, and selecting the variables with the high importance degree of 80-90% to enter the next round of screening. Therefore, variable screening is carried out according to the importance degree of the variables in the XGB algorithm model.
Further, in model fitting, for each variable to be selected, the PSI of the variable to be selected on the test set and the cross-time verification set is calculated by taking the distribution of the training set as a reference, and the variable to be selected with the PSI larger than 0.1 is removed. Therefore, the stability of the variable is checked, and the unstable variable is removed.
Further, in order to perform various performance evaluations on the constructed final model so as to be beneficial to online application, after model fitting, model evaluation is also included, and samples in a cross-time verification set are adopted to perform evaluation on the final model, wherein the evaluation includes performance evaluation, stability evaluation and explanatory evaluation.
Further, the performance evaluation comprises calculating an AUC value of the final model on the cross-time verification set, comparing the AUC value with the AUC value on the merging set, if the deviation is within a tolerance range, passing the performance evaluation, and otherwise, performing model fitting again.
Further, the stability evaluation comprises calculating the PSI of the final model predicted value on the cross-time verification set by taking the predicted value distribution of the final model on the merging set as a reference, if the PSI is less than 0.1, the stability evaluation is passed, and otherwise, model fitting is carried out again.
And further, the score conversion is also included, and the default probability value output by the final model is converted into a score value to serve as the client credit score. The credit rating system based on the bank card transaction flow can be used as an effective supplement for credit rating assessment of customers, and can more comprehensively evaluate the credit risk of retail credit customers. Meanwhile, when the institution does not have credit investigation qualification, the credit investigation system can be used as a backup for credit investigation credit.
Drawings
FIG. 1 is a flowchart illustrating a credit evaluation method according to an embodiment of the present invention.
Detailed Description
The following is further detailed by way of specific embodiments:
the embodiment is substantially as shown in fig. 1, and the credit evaluation method of the embodiment includes the following steps:
and data acquisition, namely acquiring bank card transaction flow data of each customer based on the queried and authorized customer sample, specifically, backtracking the bank card transaction flow data corresponding to the time according to the first credit application time of the customer.
Deriving variables, processing a client-level behavior label based on the acquired bank card transaction flow data, deriving variables of transaction preference, transaction liveness, consumption level and risk transaction, and certainly deriving other types of variables according to business needs, wherein the specific processing mode is as follows: the method comprises the steps of firstly aggregating bank card transaction running data according to client levels, then carrying out statistics according to variable categories, wherein a statistic time interval can be selected from 1 month, 3 months, 6 months or 12 months, a statistic function can be selected from sum, maximum value, minimum value, mean value, median value and the like, finally obtaining a large number of variables such as 'maximum value of consumption amount of nearly 3 months', 'sum of transaction times of nearly 6 months' and the like, and calculating to obtain a variable quantity value of each client sample.
And (4) variable screening, wherein the variables are screened according to the missing values, the missing rates and the IV values of the variables, and the missing values are incomplete data caused by information lack. If a variable has a large number of missing values on a sample, the actual value of the variable is small and therefore needs to be culled. The missing rate of the variable reflects the percentage of missing values of the variable on the sample, and if the missing rate is more than 85%, the variable should be eliminated. The IV Value (Information Value), i.e. the Information Value index, measures the degree of influence of a variable on the target. Generally, if the IV value of a variable is less than 0.02, the variable should be eliminated.
After the missing value and the IV value are screened, the remaining variables have certain prediction capability on the default probability of the customer and are suitable for constructing a credit model. However, because there is correlation between variables, simple screening may be performed through correlation, but in this embodiment, embedded screening using an XGB algorithm is preferably used to screen variables, and embedded screening may obtain a better result, the embedded screening is performed in multiple rounds, and in this embodiment, 10 rounds are preferred, and each round is performed as follows:
randomly dividing a sample set into a training set and a testing set, wherein the preferred proportion of the implementation is 70% and 30% respectively;
selecting an XGB algorithm model, setting the XGB algorithm model super-parameter as the evaluation index of the XGB algorithm model being AUC, and setting the XGB algorithm model super-parameter as follows: setting the tree depth to be 3, 4 or 5, setting the maximum iteration step number to be 100, 150 or 200, setting the learning rate to be 0.01, 0.05 or 0.1, setting the Gamma to be 0.1, 0.5 or 1, setting the column random extraction proportion to be 0.5, 0.6 or 0.8, setting the sample random extraction proportion to be 0.5, 0.6 or 0.8, and setting the early termination step number to be 20 so as to select the optimal iteration times;
carrying out model training by adopting a training set;
and evaluating the model effect of the XGB algorithm on the test set, calculating an AUC value, terminating the screening iteration if the AUC value of the round is greatly reduced compared with that of the previous round, taking the variable screened in the previous round as a candidate variable for model fitting, and otherwise, continuing to screen the variable.
In this embodiment, after evaluating the effect of the XGB algorithm model on the test set, the method further includes: and (3) performing variable screening according to the importance degree of the variables in the XGB algorithm model, directly removing the variables with the application times of 0, sorting the rest variables according to the importance degree, selecting the variables with the high importance degree of 80-90% to enter the next round of screening, and taking the variables as the variables to be selected if the variables are the last round.
Through embedded screening, the number of variables to be selected can be controlled within a certain range, generally not more than 100, and if the variables to be selected are still large, the number of turns of embedded screening can be increased.
And model fitting, namely dividing the sample set into a training set, a testing set and a cross-time verification set, wherein the cross-time verification set is the part of samples with the closest application time, and is generally 20%, the rest 80% of samples are randomly divided, the training set accounts for 55% -60%, and the testing set accounts for 20% -25%.
In model fitting, for each variable to be selected, calculating the PSI of the variable to be selected on the test set and the cross-time verification set by taking the distribution of the training set as a reference, and removing the variable to be selected with the PSI larger than 0.1.
In model fitting, an optimal hyper-parameter combination is selected by adopting a grid search method, namely, a plurality of selectable values are set for each hyper-parameter (tree depth, maximum iteration step number, learning rate, Gamma, column random extraction proportion and sample random extraction proportion), and specific values refer to an embedded screening link. When grid searching is carried out, model training is carried out on any group of hyper-parameter combinations by using a training set to obtain a model, the AUC value of the model is calculated in a test set, the hyper-parameter of the model with the highest AUC value in the test set is used as the final hyper-parameter value, and a variable used by the model is used as the final model entering variable; and combining the training set and the testing set to form a combined set, performing model fitting on the basis of the determined final value of the hyper-parameter and the final in-mold variable to obtain a final model, and calculating an AUC value on the final model.
After model fitting, model evaluation is further included, and the final model is evaluated by adopting samples in the cross-time verification set, wherein the evaluation comprises performance evaluation, stability evaluation and explanatory evaluation.
And the performance evaluation comprises calculating the AUC value of the final model on the cross-time verification set, comparing the AUC value with the AUC value on the merging set, if the deviation is within a tolerance range, if the attenuation is within 0.05, passing the performance evaluation, and otherwise, carrying out model fitting again.
And the stability evaluation comprises the steps of calculating the PSI of the final model predicted value on the cross-time verification set by taking the predicted value distribution of the final model on the merging set as a reference, if the PSI is less than 0.1, carrying out stability evaluation, and otherwise, carrying out model fitting again.
The model evaluation in this embodiment also includes an explanatory evaluation: and checking the SHAP graph of the model entering variable to ensure that the influence of the variable on the model predicted value accords with business logic or can be explained, and otherwise, carrying out model fitting again.
The credit evaluation method of the embodiment further comprises score conversion, which converts the default probability value output by the final model into a credit score value as the client credit score. For example: in performing the Score conversion, it is necessary to determine the Base Score (Base _ Score), the Base contrast (Base _ Odds), and the PDO. Base _ Score may be selected 600, Base _ Odds 50, PDO 50.
If the model calculates that the default probability of the customer is p, then its Odds is p/(1-p), and Score is a-B i n (odd), where a is Base _ Score + PDO i n (Base _ odd)/i n (2), and B is PDO/i n (2).
Finally, the Score of the customer is limited to the Score interval of 0-999 (i.e. less than 0 is marked as 0, and more than 999 is marked as 999), and the Score is taken as the final Score of the customer credit.
The foregoing is merely an example of the present invention, and common general knowledge in the field of known specific structures and characteristics is not described herein in any greater extent than that known in the art at the filing date or prior to the priority date of the application, so that those skilled in the art can now appreciate that all of the above-described techniques in this field and have the ability to apply routine experimentation before this date can be combined with one or more of the present teachings to complete and implement the present invention, and that certain typical known structures or known methods do not pose any impediments to the implementation of the present invention by those skilled in the art. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the patent. The scope of the claims of the present application shall be determined by the contents of the claims, and the description of the embodiments and the like in the specification shall be used to explain the contents of the claims.

Claims (9)

1. A credit evaluation method, comprising the steps of,
acquiring data, namely acquiring bank card transaction flow data of each customer based on the queried and authorized customer sample;
deriving variables, namely processing a client-level behavior label based on the acquired bank card transaction running data to derive variables of transaction preference, transaction activity, consumption level and risk transaction;
screening variables, namely screening the variables according to the missing values, the missing rates and the IV values of the variables;
model fitting, namely dividing a sample set into a training set, a testing set and a cross-time verification set, selecting an optimal hyper-parameter combination by adopting a grid search method in the model fitting, performing model training on any group of hyper-parameter combinations by using the training set to obtain a model when performing grid search, calculating an AUC value of the model in the testing set, taking a hyper-parameter of the model with the highest AUC value in the testing set as a final hyper-parameter value, and taking a variable used by the model as a final in-mode variable; and combining the training set and the testing set to form a combined set, performing model fitting on the basis of the determined final value of the hyper-parameter and the final in-mold variable to obtain a final model, and calculating an AUC value on the final model.
2. The credit evaluation method of claim 1, further comprising, in the variable screening, screening the variables using embedded screening of the XGB algorithm, the embedded screening being performed in a plurality of rounds, each round performing the following operations:
randomly dividing a sample set into a training set and a testing set;
selecting an XGB algorithm model, setting the XGB algorithm model hyperparameters, wherein the evaluation index of the XGB algorithm model is AUC;
carrying out model training by adopting a training set;
and evaluating the model effect of the XGB algorithm on the test set, calculating an AUC value, terminating the screening iteration if the AUC value of the round is greatly reduced compared with that of the previous round, taking the variable screened in the previous round as a candidate variable for model fitting, and otherwise, continuing to screen the variable.
3. The credit evaluation method of claim 2, wherein the XGB algorithm model hyperparameters are set as: the tree depth is set to 3, 4 or 5, the maximum iteration step number is set to 100, 150 or 200, the learning rate is set to 0.01, 0.05 or 0.1, Gamma is set to 0.1, 0.5 or 1, the column random draw ratio is set to 0.5, 0.6 or 0.8, the sample random draw ratio is set to 0.5, 0.6 or 0.8, and the early termination step number is set to 20.
4. The credit evaluation method of claim 3, further comprising, after evaluating the XGB algorithm model effect on the test set: and directly eliminating the variables with the application times of 0, sorting the rest variables according to the importance degree, and selecting the variables with the high importance degree of 80-90% to enter the next round of screening.
5. The method of claim 1, wherein in the model fitting, for each candidate variable, based on the distribution of the training set, the PSI of the candidate variable in the test set and the cross-time validation set is calculated, and candidate variables with PSI greater than 0.1 are eliminated.
6. The credit evaluation method of claim 1, further comprising model evaluation after model fitting, wherein the final model is evaluated using samples in the cross-time validation set, including performance evaluation, stability evaluation, and explanatory evaluation.
7. The method of claim 6, wherein the performance evaluation comprises calculating an AUC value of the final model over the validation set over time and comparing the AUC value with the AUC value in the combined set, and if the deviation is within a tolerance range, passing the performance evaluation, otherwise performing the model fitting again.
8. The method of claim 6, wherein the stability assessment comprises calculating PSI of the final model prediction values over the time validation set based on the distribution of the final model prediction values over the union set, and if PSI is less than 0.1, passing the stability assessment, otherwise performing the model fitting again.
9. The credit evaluation method of claim 1 or 6, further comprising a score conversion for converting the default probability value outputted from the final model into a score value as the client credit score.
CN202111123065.1A 2021-09-24 2021-09-24 Credit assessment method Pending CN113793212A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111123065.1A CN113793212A (en) 2021-09-24 2021-09-24 Credit assessment method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111123065.1A CN113793212A (en) 2021-09-24 2021-09-24 Credit assessment method

Publications (1)

Publication Number Publication Date
CN113793212A true CN113793212A (en) 2021-12-14

Family

ID=78879248

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111123065.1A Pending CN113793212A (en) 2021-09-24 2021-09-24 Credit assessment method

Country Status (1)

Country Link
CN (1) CN113793212A (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108154430A (en) * 2017-12-28 2018-06-12 上海氪信信息技术有限公司 A kind of credit scoring construction method based on machine learning and big data technology
CN109948668A (en) * 2019-03-01 2019-06-28 成都新希望金融信息有限公司 A kind of multi-model fusion method
CN110163743A (en) * 2019-04-28 2019-08-23 钛镕智能科技(苏州)有限公司 A kind of credit-graded approach based on hyperparameter optimization
CN111652710A (en) * 2020-06-03 2020-09-11 北京化工大学 Personal credit risk assessment method based on ensemble tree feature extraction and Logistic regression
CN111951097A (en) * 2020-08-12 2020-11-17 深圳微众信用科技股份有限公司 Enterprise credit risk assessment method, device, equipment and storage medium
CN112749921A (en) * 2021-02-01 2021-05-04 深圳无域科技技术有限公司 Mathematical modeling method, system, device and computer readable medium

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108154430A (en) * 2017-12-28 2018-06-12 上海氪信信息技术有限公司 A kind of credit scoring construction method based on machine learning and big data technology
CN109948668A (en) * 2019-03-01 2019-06-28 成都新希望金融信息有限公司 A kind of multi-model fusion method
CN110163743A (en) * 2019-04-28 2019-08-23 钛镕智能科技(苏州)有限公司 A kind of credit-graded approach based on hyperparameter optimization
CN111652710A (en) * 2020-06-03 2020-09-11 北京化工大学 Personal credit risk assessment method based on ensemble tree feature extraction and Logistic regression
CN111951097A (en) * 2020-08-12 2020-11-17 深圳微众信用科技股份有限公司 Enterprise credit risk assessment method, device, equipment and storage medium
CN112749921A (en) * 2021-02-01 2021-05-04 深圳无域科技技术有限公司 Mathematical modeling method, system, device and computer readable medium

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
林治洪等: "《SAS数据挖掘与分析项目实战》", 中国铁道出版社有限公司, pages: 182 - 183 *
牛彩芳: "基于XGBoost算法的信用评级系统设计与实现", no. 7, pages 122 - 123 *

Similar Documents

Publication Publication Date Title
CN111291816B (en) Method and device for carrying out feature processing aiming at user classification model
CN110998608B (en) Machine learning system for various computer applications
CN108475393A (en) The system and method that decision tree is predicted are promoted by composite character and gradient
CN112102073A (en) Credit risk control method and system, electronic device and readable storage medium
CN112700324A (en) User loan default prediction method based on combination of Catboost and restricted Boltzmann machine
CN107392217B (en) Computer-implemented information processing method and device
CN112990989B (en) Value prediction model input data generation method, device, equipment and medium
CN117010914A (en) Identification method and device for risk group, computer equipment and storage medium
CN116821688A (en) Method for processing data set in credit card fraud transaction based on clustering downsampling technology
CN113177733B (en) Middle and small micro enterprise data modeling method and system based on convolutional neural network
Tunç Feature selection in credibility study for finance sector
CN113421154B (en) Credit risk assessment method and system based on control chart
CN113793212A (en) Credit assessment method
CN114626940A (en) Data analysis method and device and electronic equipment
CN114693428A (en) Data determination method and device, computer readable storage medium and electronic equipment
CN110570301B (en) Risk identification method, device, equipment and medium
CN114170000A (en) Credit card user risk category identification method, device, computer equipment and medium
CN112823502B (en) Real-time feedback service for resource access rule configuration
CN112529319A (en) Grading method and device based on multi-dimensional features, computer equipment and storage medium
Zeng A comparison study on the era of internet finance China construction of credit scoring system model
Mammadzada et al. Application of bg/nbd and gamma-gamma models to predict customer lifetime value for financial institution
KR102576143B1 (en) Method for performing continual learning on credit scoring without reject inference and recording medium recording computer readable program for executing the method
CN110413967B (en) Account checking chart generation method, device, computer equipment and storage medium
CN117994017A (en) Method for constructing retail credit risk prediction model and online credit service Scoredelta model
Nasution et al. Credit Risk Detection in Peer-to-Peer Lending Using CatBoost

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination