KR102336462B1 - Apparatus and method of credit rating - Google Patents

Abstract

A credit rating information providing apparatus of an embodiment of the present invention includes: a data collection unit which collects CB rating and non-financial data; a data processing unit which selects and classifies a change value of the non-financial data according to a change in the CB rating; a credit rating prediction evaluation model unit which generates an evaluation model for predicting future CB rating changes by using the change in the CB rating according to an evaluation period and a change value of the non-financial data; and a computing unit which receives personal data and compares the personal data and the evaluation model to predict a change in the individual's CB rating. The embodiment has the effect of more effectively measuring an individual's credit rating by providing an evaluation model capable of predicting credit rating changes by using non-financial data.

Description

Apparatus and method for providing credit rating information {APPARATUS AND METHOD OF CREDIT RATING}

The embodiment relates to an apparatus for providing credit rating information for evaluating a credit rating using non-financial data.

In general, personal credit evaluation is mainly made based on credit information and financial information data. In particular, recently, analysis and measurement using non-financial data have been conducted to increase the accuracy of credit evaluation.

In addition, various methods using data are being attempted, such as the recent revision of the Data 3 Act, including the Credit Information Act, and the creation of new data/credit evaluation-related businesses such as My Data Up and Non-Financial CB Business.

However, there are still not many prediction methods or evaluation models for credit information/credit rating management for individuals using various data.

Until now, it is common to slightly increase credit score by submitting communication fee payment details. Other than that, most of the credit management guides are simple methods such as credit card usage and loan management. Most of the follow-up management is follow-up, such as changes in credit card payments, increase in card usage, and information on loan maturity dates.

In the modern credit society, credit rating is very important and plays an absolute role in individual financial activities. In particular, if the credit rating has been lowered, it takes at least several months to recover, even if an activity to raise the credit rating is performed, and damage to the use of finance occurs such as an increase in interest rates and the inability to use some financial products.

Therefore, some financial defaults are detected before credit rating changes due to lack of financial knowledge, negligence, mistakes, and negative factors in simple life patterns/habits, except for individuals, and provide alarm and information to individuals in various ways As such, a credit rating prediction evaluation model that can provide consulting is required.

In order to solve the above-mentioned problems, the embodiment aims to provide a credit rating information providing apparatus and method for predicting credit rating changes by using personal non-financial data and providing appropriate information.

In addition, the embodiment predicts credit rating changes by using personal non-financial data, and provides a credit rating information providing device and Another purpose is to provide a method.

The credit rating information providing device of the embodiment includes a data collection unit that collects CB rating and non-financial data, a data processing unit that selects and classifies a change value of the non-financial data according to the change in the CB rating, and the CB rating change according to the evaluation period and a credit rating prediction evaluation model unit for generating an evaluation model for predicting future CB rating changes by using the change value of the non-financial data; It may include a calculation unit for predicting a change.

The data processing unit may align the non-financial data according to the CB grade, find and sample the degree of change of the non-financial data according to the change of the CB grade, and quantify the tabled non-financial data using a sigmoid function have.

The data processing unit may classify the quantified non-financial data according to the CB grade according to a period.

The evaluation model may include at least one or more of the CB grade change value and the average value of increase or decrease in the number of delinquent communication charges according to the CB grade change value, the average value of increase or decrease in annual income, and the average value of the change in subscription service usage amount.

It may include an evaluation model verification unit that verifies the evaluation model.

The non-financial data collected by the data collection unit may include at least one of communication data information, periodic payment information, tax information, savings information, and management fee information.

Communication data information includes at least one or more of negative indicators including SIM opening information, suspended line information, new subscription information, and content payment information, non-payment information, payment information, terminal usage days information, counseling center contact information, and data usage information. may include

The periodic payment information may include at least one of a subscription service type, the number of subscription services, the number of subscription service payments, the subscription service payment amount, the subscription service non-payment number, and the subscription service non-payment amount.

The tax information may include at least one of a type of tax paid, arrears of tax, the number of non-payment of tax, and a payment period of tax.

The savings information may include at least one of periodic savings information, the number of deposits and duration information.

The management fee information may include at least one or more of the number of times of payment of the management fee and the number of overdue times.

In addition, the embodiment provides a method for providing credit rating information performed in a credit rating information providing device, the method comprising: collecting CB rating and non-financial data; measuring a change value of the non-financial data according to the change in the CB rating; generating an evaluation model for predicting a future CB rating change using the change in the CB rating and the non-financial data according to the evaluation period; and comparing the personal data with the evaluation model by receiving personal data. predicting a change in the individual's CB rating.

The embodiment has the effect of more effectively measuring an individual's credit rating by providing an evaluation model capable of predicting changes in credit rating using non-financial data.

In addition, the embodiment has an effect of rapidly recovering an individual's credit rating by predicting a credit rating change using non-financial data.

In addition, the embodiment has the effect of reducing the damage caused by the personal credit rating decline by predicting the credit rating change using non-financial data.

1 is a block diagram illustrating an apparatus for providing credit rating information according to an embodiment.
2 is a diagram illustrating types of data collected by a data collection unit of an apparatus for providing credit rating information according to an embodiment.
3 is a flowchart illustrating a data processing process of a data collection unit of an apparatus for providing credit rating information according to an embodiment.
4 is a flowchart illustrating a data processing process of a data processing unit of an apparatus for providing credit rating information according to an embodiment.
5 is a diagram illustrating an evaluation model generated by a credit rating prediction evaluation model unit of the apparatus for providing credit rating information according to an embodiment.
6 is a view showing an operation performed by the evaluation model verification unit of the apparatus for providing credit rating information according to the embodiment.
7 is a flowchart illustrating a method of providing credit rating information according to an embodiment.
8 is a diagram illustrating an operation of an apparatus for providing credit rating information according to an embodiment.

Hereinafter, the embodiment will be described in detail with reference to the drawings.

1 is a block diagram illustrating a credit rating information providing apparatus according to an embodiment, FIG. 2 is a diagram illustrating types of data collected by a data collection unit of a credit rating information providing apparatus according to an embodiment, and FIG. 3 is an embodiment is a flow chart showing a data processing process of the data collection unit of the credit rating information providing apparatus according to the embodiment It is a view showing the evaluation model generated by the credit rating prediction evaluation model unit of the credit rating information providing device, and FIG. 6 is a view showing the operation performed by the evaluation model verification unit of the credit rating information providing device according to the embodiment.

Referring to FIG. 1 , an apparatus 1000 for providing credit rating information according to an embodiment includes a data collection unit 100 , a data processing unit 200 , a credit rating prediction evaluation model unit 300 , an evaluation model verification unit 400 and It may include an operation unit 500 .

The data collection unit 100 may collect non-financial data. Non-financial data may include a Credit Bureau (CB) rating. The CB grade may consist of grades 1 to 10, but is not limited thereto.

As shown in FIG. 2 , the non-financial data may include communication data. The communication data may include negative indicators and positive indicators.

Negative indicators of communication data are: Usim alone opened more than 2 lines within the last 2 months, Usim changed more than 3 times recently, illegal / expedient stopped, spam suspended lines, 3 or more new subscriptions within the last 3 months, after opening It can include information data such as whether or not the amount used for payment of a small amount + content payment is more than 200,000 won per month within 3 months. Positive indicators of communication data include non-payment amount, number of self-payments, number of non-payment months, non-payment/full payment cycle, average usage amount for the previous three months, average payment days, elapsed subscription date, number of days of terminal use, contact information of the counseling center, data usage, etc. It may contain data.

In telecommunication data, weights are differentially added between variables that significantly affect creditworthiness and variables that do not affect creditworthiness and can be used as data for credit evaluation. For example, a weight of 1 may be assigned to information on the number of times the SIM card was opened alone, within 3 months after opening, the number of self-payments, and the elapsed date of subscription, and a weight lower than 1 may be assigned to the rest of the information. Based on the weight 1, the lowest value of the weight may be 0.35. A ratio of the highest value to the lowest value of the weight may be 1:0.35. The weight has the effect of more accurately measuring credit information by adding a high weight to information with high reliability.

Non-financial data may include periodic payment information. The periodic payment information may include information data such as the type, number, number of payments, payment amount, non-payment number, and non-payment amount for the periodic payment service.

Non-financial data may include tax information. The tax information may include information data such as a type of tax to be paid, a tax overdue, the number of non-payment, a tax payment period, and the like. Here, a weight of 1 may be assigned to the tax payment period, and a weight lower than 1 may be assigned to the remaining information.

Tax information may not take into account the amount of tax. The reason is that the large or small amount of tax is directly related to assets and income, as it may be outside the scope of non-financial data. Therefore, only the fidelity information of tax payment can be included in the non-financial data.

Non-financial data may include savings information. Savings information may include information data such as periodic savings, the number and duration of deposits.

The non-financial data may include utility information such as administrative expenses. The management fee information may include the number of times of payment and delinquency information data of management fees, city gas, district heating, and the like.

Non-financial data may include, but is not limited to, annual income information and increase/decrease information.

As shown in FIG. 3 , looking at the order of operations of the data collection unit 100 , first, external non-financial data may be collected using an API, screen scraping, or a dedicated line (S110). External non-financial data may include CB rating, periodic payment information, tax information, savings information, and management fee information.

The data collection unit 100 may collect internal non-financial data (S120). Internal non-financial data may be collected using a database (DB) provided therein. The internal non-financial data may include communication data previously collected from a communication company.

The data collection unit 100 may perform encryption processing on external non-financial data and internal non-financial data (S130). The data collection unit 100 may store and store the encrypted non-financial data (S140).

Returning to FIG. 1 , the data processing unit 200 may load and analyze the big data collected by the data collection unit 100 .

The data processing unit 200 may select and classify the change value of the non-financial data according to the CB grade change. The data processing unit 200 may sort the non-financial data according to the CB grade by using the non-financial data collected from the data collection unit 100 .

As shown in FIG. 4 , looking at the order of operations of the data processing unit 200 , first, the data processing unit 200 may sort non-financial data. The data processing unit 200 may sort each non-financial data corresponding thereto in turn based on the CB grade.

Subsequently, the data processing unit 200 may find and sample the degree of change in the non-financial data according to the change in the CB grade. For sampling, KS statistic, chi-square test, and t test can be used.

The KS statistic can represent the maximum difference % of the cumulative distribution of the good group and the bad group of a specific variable. For example, more than 3 times of change of heart -> 10% of group with 3 or more times, 90% of group with less than that, 80% of KS statistic, KS statistic of each variable is 40~50 or more in financial-based credit evaluation, 20 or more in corporate credit rating Above, the credit rating of a small business may be 5 or more. In the example, since the difference between the good and bad groups in the variables is not large, the KS statistic 5 is used as the standard, and instead, the number of variables is secured as much as possible. If it is 5 or more, it can be used as a variable.

The chi-square test is a method of verifying whether the observed frequency in categorical data is different from the expected frequency. Categorical data is data having two or more categories, and can be calculated by Equation 1 below.

[Equation 1]

The t-test compares the mean between two groups for continuous data, and can be used especially when measuring the mean of a normal distribution. In this embodiment, it can be utilized for the presence or absence of suspension according to the CB grade, the presence or absence of contact with the counseling center, and the like. The method to obtain the t-test number is as follows.

1) t score = mean difference of two groups / data deviation of two groups

2) t-score follows standard normal distribution

3) The standard normal distribution can be formulated as a probability density function that can calculate the area of the distribution

4) The value obtained by subtracting the area corresponding to the t-score on the x-axis from 1 in the probability density function is the p-value (p-value)

In the embodiment, as a variable, a KS statistic of 0.05 or more and a p-value of 0.05 or less of chi-square or t-test may be selected as variables.

Subsequently, the data processing unit 200 may quantify the data selected as variables according to grades 1 to 10 ( S230 ). Data quantification may use a sigmoid function as shown in Equation 2 so that all indicators have values of 0 to 1.

[Equation 2]

Here, x is the value of the variable, and a is the weight. The sigmoid function can converge the output value to 0~1 for all input values. The weight may be determined as a constant verified in the final prediction model.

Then, the data processing unit 200 may classify the non-financial data (S240). The data processing unit may classify the quantified non-financial data according to the CB grade according to a period. For example, it is possible to check the amount of change in non-financial data in units of 3 months, 6 months, 9 months, and 12 months, but is not limited thereto.

Returning to FIG. 1 , the credit rating prediction evaluation model unit 300 may generate an evaluation model for predicting a future CB rating change by using the change in the CB rating according to the evaluation period and the change value of the non-financial data.

For example, the credit rating prediction evaluation model unit 300 based on the data sorted based on the CB rating, based on the data change of 9 months out of 12 months of historical data, the evaluation model according to the CB rating change of 3 months thereafter can create

As shown in Fig. 5, the evaluation model includes the CB rating change value for a specific period, the average value of increase or decrease in the number of delinquent communication charges calculated from non-financial data according to the CB rating, the average value of change in annual income, and the average value of change in subscription service usage. can do. Here, the evaluation period may be 6 months, but is not limited thereto.

The number of CB grade transfer cases can be a total of 100 such as Grade 1 -> Grade 1, Grade 1 -> Grade 2, Grade 1 -> Grade 3, etc. Data change for the past 24 months in units of 3 months from the present time can be measured. In this case, the past 24 months may be extended in some cases.

In the evaluation model, the concept of time for each of the unspecified people corresponding to non-financial data can be considered. For the personal data of many unspecified people, at least 3 to 24 months of data can be applied to the evaluation model.

A logistic regression model can be used for sample data change according to CB grade change. The concept of time can be applied by Equation (3). In Equation 3, the prediction probability may have a tendency to decrease as the time decreases.

[Equation 3]

Here, 24 is the number of months, and x is the time of accumulated data of a specific individual.

Returning to FIG. 1 , the evaluation model verification unit 400 may verify the evaluation model generated by the credit rating prediction evaluation model unit 300 . Here, the evaluation model validation can be performed using K-fold Cross Validation to validate and evaluate the model. In K-fold cross-validation, the existing evaluation model can be updated frequently by obtaining the average of the evaluation model through multiple cross-validation or finding the optimal model in the current state.

As shown in FIG. 6 , after each non-financial data is extracted into a 24-month period, it can be randomly classified into 10 folds without allowing duplicates. Then, in each of the 10 classified folds, 10% becomes the validation set, and the remaining 90% can be determined as the training set.

Each of the ten folds can then be evaluated as a training set and compared to the validation set. Then, the average value of these can be taken as the latest result value of the corresponding predictive model. The evaluation model can be verified by adjusting the weights of the existing data with the latest results.

Returning to FIG. 1 , the calculating unit 500 may predict a change in an individual's CB grade by comparing the personal data with the evaluation model.

The calculator 500 may receive personal data from the individual 10 who has requested credit rating information and compare it with an evaluation model. The calculating unit 500 may calculate a future CB grade change with respect to an index change by using the data of the individual 10 for a specific period.

The calculation unit 500 may send the individual's CB rating change information to a financial institution or credit rating agency 20, and the financial institution or credit rating institution 20 uses the individual's CB rating for credit, such as a rate cut. Additional points may be awarded.

Here, personal data may be collected after consent to use of personal information, and the collected personal data may be compared with an evaluation model in the operation unit 500 . Personal data may be collected by means of APIs, leased lines, or screen scraping.

The embodiment has the effect of predicting credit ratings more effectively by predicting credit rating changes using non-financial data.

In addition, the embodiment has an effect of inducing recovery of an individual's credit rating by predicting a change in credit rating using non-financial data.

In addition, the embodiment has the effect of reducing the damage caused by the personal credit rating decline by predicting the credit rating change using non-financial data.

7 is a flowchart illustrating a method of providing credit rating information according to an embodiment.

Referring to FIG. 7 , the credit rating information providing method according to the embodiment includes the steps of collecting CB rating and non-financial data (S310), and selecting and classifying the change value of the non-financial data according to the change in the CB rating (S320). ) and generating an evaluation model for predicting future CB rating changes using the change in the CB rating and the non-financial data according to the evaluation period (S330); It may include a step (S340) of predicting a change in the individual's CB grade by comparing the evaluation model. Here, the credit rating information providing method according to the embodiment may be performed by the credit rating information providing apparatus.

Collecting the CB rating and non-financial data (S310) may be performed in the data collection unit. In the step (S310) of collecting CB grade and non-financial data, the non-financial data may include communication data, periodic payment information, tax information, savings information, and management fee information.

The step (S320) of selecting and classifying the change value of the non-financial data according to the CB grade change may be performed by the data processing unit. In the step (S320) of selecting and classifying the change value of the non-financial data according to the change in the CB rating, each non-financial data corresponding thereto may be sequentially arranged based on the CB rating. Then, it is possible to find and sample the degree of change in non-financial data as the CB rating changes. Then, data selected as variables according to grades 1 to 10 can be quantified. Then, the quantified non-financial data according to the CB rating can be classified according to period.

The step (S330) of generating an evaluation model for predicting a future CB rating change using the change in the CB rating and the non-financial data according to the evaluation period (S330) may be performed in the credit rating prediction evaluation model unit.

In the step (S330) of generating an evaluation model for predicting future CB rating change by using the change value of the CB rating change and the non-financial data according to the evaluation period (S330), the evaluation model uses the CB rating change value and the CB rating for a specific period It may include an average value of increase and decrease in the number of delinquent communication charges calculated from non-financial data according to , an average value of increase or decrease in annual income, and an average value of change in subscription service usage amount.

The step (S340) of receiving personal data and comparing the personal data with the evaluation model to predict a change in the individual's CB grade (S340) may be performed by the calculating unit.

In the step (S340) of receiving personal data and comparing the personal data and the evaluation model to predict the change in the CB rating of the individual, the future CB rating change for the index change is calculated using the personal data for a specific period. can

8 is a diagram illustrating an operation of an apparatus for providing credit rating information according to an embodiment.

As shown in FIG. 8 , the credit rating information providing apparatus 1000 may periodically collect de-identified statistical data from a data holding institution ( S401 ).

The apparatus for providing credit rating information may establish an evaluation model for predicting a credit rating (S402). Here, the credit rating information providing apparatus may generate an evaluation model using non-financial data, and may be generated based on a change value of the non-financial data according to a change in the CB rating.

When a prediction of the credit rating is requested from the user (S403), the credit rating information providing device performs the identity authentication and information provision consent procedure (S404). Users can perform authentication and consent processes according to the provided procedures.

The credit rating information providing apparatus 1000 may request or transmit data from the data retention institution 30 ( S405 ).

The credit rating information providing apparatus 1000 may predict a change in rating using an evaluation model generated using past and present data (S406). The credit rating information providing apparatus 1000 may deliver the predicted rating change value to the user (S407).

At the same time, the credit rating information providing apparatus 1000 may transmit the user's predicted change in rating to the financial company 20 (S408).

The financial company 20 may provide financial information to the user 10 according to the changed credit rating (S409). The credit rating information providing apparatus 1000 may receive updated personal credit rating information from the user (S410).

The embodiment may help the user to receive a financial product according to the changed credit rating by providing a future change rate of the user's credit rating to the financial company using non-financial data.

Although the above has been described with reference to the drawings and embodiments, it will be understood by those skilled in the art that various modifications and changes can be made to the embodiments without departing from the spirit of the embodiments described in the claims below. will be able

100: data collection unit
200: data processing unit
300: credit rating prediction evaluation model unit
400: evaluation model verification unit
500: arithmetic unit

Claims (12)

a data collection unit that collects CB ratings and non-financial data;
a data processing unit for selecting and classifying a change value of the non-financial data according to the change in the CB rating;
a credit rating prediction evaluation model unit for generating an evaluation model for predicting future CB rating changes by using the CB rating change according to the evaluation period and the change value of the non-financial data;
an evaluation model verification unit that verifies the evaluation model; and
Comprising a computing unit that receives personal data and compares the personal data with the evaluation model to predict a change in the individual's CB grade,
The CB rating change and the change value of the non-financial data are values collected from an unspecified number for a certain period of time,
The credit rating prediction evaluation model unit generates an evaluation model for predicting the CB rating change after the evaluation period by using the change value of the CB rating change and the non-financial data corresponding to the evaluation period during the predetermined period,
The evaluation model verification unit verifies the evaluation model using k-fold cross-validation,
The evaluation model verification unit generates a plurality of folds consisting of a training set and a verification set by extracting non-financial data of a certain period, calculates each verification error for the plurality of folds, and based on the average value of the verification errors, the Validate the evaluation model;
The evaluation model verification unit determines 90% of the total non-financial data of each fold as a training set, and determines the remaining 10% as a verification set,
The data processing unit aligns the non-financial data according to the CB grade, samples the degree of change of non-financial data according to the change of the CB grade using any one of KS statistics, chi-square test, and t test, and sampled non-financial data Credit rating information providing device that quantifies using a sigmoid function. delete According to claim 1,
The data processing unit,
Credit rating information providing device for classifying the quantified non-financial data according to the CB rating according to period. According to claim 1,
The evaluation model includes at least one of the CB rating change value and the average value of increase or decrease in the number of delinquent communication charges according to the CB rating change value, the average change in annual income, and the average change in the subscription service usage amount. delete 5. The method of any one of claims 1, 3 and 4,
The non-financial data collected by the data collection unit includes at least one of communication data information, periodic payment information, tax information, savings information, and management fee information. 7. The method of claim 6,
The communication data information includes at least one or more of a negative indicator including SIM opening information, stop line information, new subscription information, and content payment information, non-payment information, payment information, terminal usage days information, counseling center contact information, and data usage information. Credit rating information providing device including a positive indicator including a. 7. The method of claim 6,
The periodic payment information may include at least one of a subscription service type, the number of subscription services, the number of subscription service payments, the subscription service payment amount, the subscription service non-payment number, and the subscription service non-payment amount. 7. The method of claim 6,
The tax information is a credit evaluation information providing device including at least one of a type of tax paid, arrears of tax, number of non-payment of tax, and payment period of tax. 7. The method of claim 6,
The savings information is a credit rating information providing device including at least one of periodic savings information, the number of deposits and duration information. 7. The method of claim 6,
The management fee information is a credit rating information providing device including at least one of the number of times of payment of the management fee and the number of delinquency. In the credit rating information providing method performed by the credit rating information providing device,
collecting CB rating and non-financial data;
measuring a change value of the non-financial data according to the change in the CB rating;
generating an evaluation model for predicting a future CB rating change by using the change in the CB rating according to the evaluation period and the change value of the non-financial data;
verifying the evaluation model; and
Comprising the step of receiving personal data and comparing the personal data and the evaluation model to predict a change in the CB grade of the individual,
The CB rating change and the change value of the non-financial data are values collected from an unspecified number for a certain period of time,
The generating of the evaluation model comprises predicting the change in the CB rating after the evaluation period using the change in the CB rating corresponding to the evaluation period and the change value of the non-financial data during the predetermined period to generate an evaluation model,
The step of verifying the evaluation model is to verify the evaluation model using k-fold cross-validation,
The step of verifying the evaluation model extracts non-financial data of a certain period to generate a plurality of folds consisting of a training set and a verification set, calculates each verification error for the plurality of folds, and calculates the average value of the verification errors. Validating the evaluation model based on
In the step of verifying the evaluation model, 90% of the total non-financial data of each fold is determined as a training set, and the remaining 10% is determined as a validation set,
The step of measuring the change value of the non-financial data according to the CB grade change aligns the non-financial data according to the CB grade, and the degree of change of the non-financial data according to the change of the CB grade is KS statistic, chi-square test and t test A method of providing credit rating information that samples using any one of the methods and quantifies the sampled non-financial data using a sigmoid function.

Images