KR102249028B1 - System for Debt Repayment Capability Evaluation Of Corporation - Google Patents

Abstract

The debt repayment ability evaluation system of a company, which can evaluate the debt repayment ability using financial data generated based on the financial statements of the target company, as well as financial information, public procurement information, etc. A data acquisition unit that acquires financial data obtained as a basis and non-financial data other than the financial data; The main forecasting period from the evaluation point to the predetermined forecasting point after the evaluation point is divided into a plurality of sub-prediction periods, and the financial data and the non-financial data are input to a plurality of debt repayment ability evaluation models corresponding to each sub-prediction period. A first calculation unit for calculating the debt repayment capability evaluation result of the target company for each of the debt repayment capability evaluation models; And a second calculation unit that calculates a final debt repayment ability evaluation result of the target company by using at least one evaluation result among the debt repayment ability evaluation results calculated for each debt repayment ability evaluation model.

Description

System for Debt Repayment Capability Evaluation Of Corporation}

The present invention relates to credit ratings, and more specifically, to credit ratings of companies.

In inter-company transactions, in order to prevent damage due to arrears or default of debts of a customer company or an investment partner company (hereinafter referred to as'company'), evaluation of the debt repayment ability of the relevant company is required in advance.

Therefore, a credit rating system has been proposed to objectively evaluate a company's debt repayment ability by objectively quantifying the possibility of a company's default or the size of its creditworthiness.

A general credit rating system evaluates a company's debt repayment ability using financial data generated based on financial statements. However, the credit rating system can only use financial data up to before the financial statements are prepared, and the financial data generated after the financial statements are prepared cannot be reflected in the evaluation. There is a problem that it cannot be accurately evaluated.

In addition, since the credit rating system evaluates the debt repayment ability at the time of evaluation, there is a problem in that it cannot be known even if there is a possibility that a problem may occur in the debt repayment ability at some point after the evaluation point.

The present invention is to solve the above-described problems, as well as financial data generated based on the financial statements of a target company, as well as non-financial data including financial information, public procurement information, etc. of a company capable of evaluating its debt repayment ability. Its technical task is to provide a system for evaluating debt repayment ability.

In addition, the present invention is a company capable of calculating the probability of at least one of arrears, insolvent, and default of a target company for each forecast period by using a plurality of debt repayment capability evaluation models that evaluate debt repayment capability for each predetermined forecast period. Its technical task is to provide a system for evaluating its debt repayment ability.

In addition, in the present invention, each debt repayment capability evaluation model is based on the financial data, non-financial data of the insolvent company generated in the expected insolvent period corresponding to each forecast period among the financial data, non-financial data, and derivative data of the insolvent company in which bankruptcy or delinquency occurred. The technical task is to provide a system for evaluating the debt repayment ability of a company in which data and derivative data are learned.

A system for evaluating the debt repayment ability of a company according to an aspect of the present invention for achieving the above object comprises: a data acquisition unit for acquiring financial data obtained based on the financial statements of a target company and non-financial data other than the financial data; The main forecasting period from the evaluation point to the predetermined forecasting point after the evaluation point is divided into a plurality of sub-prediction periods, and the financial data and the non-financial data are input to a plurality of debt repayment ability evaluation models corresponding to each sub-prediction period. A first calculation unit for calculating the debt repayment capability evaluation result of the target company for each of the debt repayment capability evaluation models; And a second calculation unit that calculates a final debt repayment ability evaluation result of the target company by using at least one evaluation result among the debt repayment ability evaluation results calculated for each debt repayment ability evaluation model.

According to the present invention, it is possible to evaluate the debt repayment ability using non-financial data as well as the financial data of the target company, so even if changes occur in the target company after the financial statements are prepared, this can be reflected, so that the debt repayment ability can be accurately calculated. There is an effect.

According to the present invention, the probability of occurrence of at least one of delinquency, insolvent, and default of a target company can be calculated for each forecast period by using a plurality of debt repayment capacity evaluation models that evaluate the debt repayment capacity for each predetermined forecast period. In addition, it has the effect of being able to evaluate not only the debt repayment ability of the target company at the time of evaluation, but also the debt repayment ability of the target company at the time after the time of evaluation.

In addition, according to the present invention, since the probability of occurrence of at least one of delinquency, insolvent, and default of a target company can be calculated for each forecast period, the debt repayment ability is abnormal in any one of the plurality of forecast periods. However, if the debt repayment ability is normal in other forecast periods, it is effective in preventing the determination of the debt repayment ability in the first forecast period as normal.

According to the present invention, since the probability of occurrence of at least one of delinquency, insolvent, and default of the target company can be calculated for each predetermined forecast period, the target company can seek a response to the target company before the delinquency or default occurs. It has the effect of minimizing economic losses by getting opportunities.

In addition, according to the present invention, each debt repayment capability evaluation model is the financial data of the insolvent company generated in the expected insolvent period corresponding to each forecast period among the financial data, non-financial data, and derivative data of the insolvent company in which bankruptcy or delinquency occurred, Since non-financial data and derivative data are learned, each debt repayment capability evaluation model learns the financial data, non-financial data, and derivative data of insolvent companies generated in different periods of insolvency. It has the effect of being able to evaluate.

1 is a diagram schematically showing the configuration of a system for evaluating a company's debt repayment ability according to an embodiment of the present invention.
2 is a diagram showing the configuration of a debt repayment ability evaluation system 100 according to an embodiment of the present invention
3A is a diagram illustrating that a first calculator according to an embodiment of the present invention uses a probability distribution to calculate a grade for an insolvency probability.
3B is a diagram showing a composition ratio for each grade according to an embodiment of the present invention.
4 is a diagram showing a configuration of a second calculation unit according to an embodiment of the present invention.
5A is a diagram illustrating an example of a table generated by a table generation unit.
5B is a diagram showing an example of a debt repayment capability level.
6 is a diagram illustrating that a modeling unit generates a debt repayment capability evaluation model according to an embodiment of the present invention.
7 is a diagram showing an example of a modeling unit verifying a debt repayment capability evaluation model.
FIG. 8A is a diagram illustrating that first to fourth models according to an embodiment of the present invention calculate an insolvency probability for each prediction period.
8B is a diagram illustrating that a first model according to an embodiment of the present invention calculates an insolvency probability for each prediction period.

The meaning of the terms described in this specification should be understood as follows.

Singular expressions should be understood as including plural expressions unless clearly defined differently in context, and terms such as “first” and “second” are used to distinguish one element from other elements, The scope of rights should not be limited by these terms.

It is to be understood that terms such as "comprises" or "have" do not preclude the presence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

The term “at least one” is to be understood as including all possible combinations from one or more related items. For example, the meaning of “at least one of the first item, the second item, and the third item” means 2 among the first item, the second item, and the third item as well as each of the first item, the second item, and the third item. It means a combination of all items that can be presented from more than one.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram schematically showing the configuration of a system for evaluating the debt repayment ability of a company (hereinafter referred to as'debt repayment ability evaluation system') according to an embodiment of the present invention.

The user terminal 10 receives identification information on a company (hereinafter referred to as a “target company”) to be evaluated for debt repayment capability from the user and transmits it to the debt repayment capability evaluation system 100. In this case, the identification information may include a business operator number, a business name, and the like.

The user terminal 10 may include a mobile terminal, a tablet PC, a computer, and the like.

The debt repayment capability evaluation system 100 evaluates the debt repayment capability of the target company according to the request of the user terminal 10. Specifically, the debt repayment capability evaluation system 100 uses a plurality of debt repayment capability evaluation models to determine non-financial data including financial data, financial information, public procurement information, etc., generated based on the financial statements of the target company. Each forecast period can be analyzed to evaluate the debt repayment ability of the target company.

Hereinafter, the debt repayment ability evaluation system 100 will be described in more detail with reference to FIG. 2.

2 is a diagram showing the configuration of a debt repayment ability evaluation system 100 according to an embodiment of the present invention. 2, the debt repayment ability evaluation system 100 includes a data acquisition unit 210, a derivative data generation unit 220, a first calculation unit 230, a second calculation unit 240, and a database ( 245), a modeling unit 250, and a debt repayment ability evaluation model 260.

The data acquisition unit 210 acquires financial data generated based on the financial statements of the target company and non-financial data other than the financial data of the target company. In one embodiment, when the identification information of the target company is received from the user terminal 10, the data acquisition unit 210 transmits the received identification information of the target company to the external server 20 from the external server 20. Financial and non-financial data of the target company can be obtained.

In an embodiment, the financial data acquired by the data acquisition unit 210 may include at least one of financial statement information, financial ratio information, financial model information, and borrowing information. At this time, the financial statement information includes at least one of financial position information, income statement information, and cash flow information, and the financial ratio information includes general financial ratio information. The financial model information includes at least one of financial model input variable information and financial issue information, and the borrowing information includes information on calculating borrowing capacity.

The non-financial data acquired by the data acquisition unit 210 may include financial information, public procurement information, human resource information, and penalty information. In this case, the company information may include a company overview, and the financial information may include credit grant information. The public procurement information may include at least one of bidding information and contract information. Human resource information may include at least one of national pension information and job posting information. The penalty information may include at least one of information on disposition of fraudulent disposition and information on disposition of construction license.

The derivative data generation unit 220 generates derived data that is correlated with the debt repayment ability from the financial data and non-financial data of the target company obtained by the data acquisition unit 210.

In one embodiment, the derivative data is the ratio of the latest month's debt to sales, the amount of the guarantee of the corporate debt to the equity capital, the amount of the guarantee of the corporate debt to the real capital, the ratio of the 3-month credit contribution to the operating profit, the ratio of the end of the previous month to the equity capital, and monthly. Borrowing capacity analysis, monthly debt dependence, average number of credit loans in the last 3 months, average deposit amount in the last 3 months, rate of exhaustion of 1 financial sector loan limit for the last 3 months, monthly interest expense, proportion of policy funds, bill discount proportion, checking loan availability, Overdraft transaction amount, payment guarantee ratio, share of collateral, current month's entry rate, current month's resignation rate, current month's number of employees increase/decrease rate, previous month's notice amount increase/decrease rate, average notified amount, 3 month resignation rate, 6 month resignation rate, 3 month employment rate , 6-month entry rate, 3-month employee increase/decrease rate, 6-month employee increase/decrease rate, 3-month-per-person average notification amount increase/decrease rate, and 6-month-per-person average notification amount increase/decrease rate may be included.

The derivative data generation unit 220 generates derived data from the financial data and non-financial data of the target company and transmits the generated data to the first calculation unit 230.

The first calculation unit 230 calculates a result of evaluating the debt repayment ability of the target company. Specifically, the first calculation unit 230 divides the main prediction period from the evaluation time to a predetermined prediction time after the evaluation time into a plurality of sub prediction periods. The first calculation unit 230 inputs the financial data and non-financial data of the target company into a plurality of debt repayment capability evaluation models 260 corresponding to each sub-prediction period, and the debt repayment capability of the target company for each debt repayment capability evaluation model. Calculate the evaluation result. At this time, the time point of evaluation means the time point at which the debt repayment ability is evaluated.

In the above-described embodiment, it has been described that the first calculation unit 230 uses only financial data and non-financial data, but the first calculation unit 230 additionally adds the derived data generated by the derived data generation unit 220. It can also be used to calculate the result of evaluating the debt repayment ability of the target company.

In one embodiment, the first calculation unit 230 includes financial data and non-financial data of the target company in a plurality of debt repayment capability evaluation models 260 that evaluate the debt repayment capability for each sub-prediction period in which a predetermined main forecast period is divided. By inputting data, a target grade indicating the probability of occurrence of insolvency events of the target company for each sub-prediction period is calculated.

Specifically, the first calculation unit 230 calculates the probability of insolvency of the target company for each sub-prediction period by inputting the financial data and non-financial data of the target company into the plurality of debt repayment capability evaluation models 260, and each sub-prediction The insolvency probability calculated for each period is calculated as a target grade according to a predetermined criterion.

In one embodiment, each sub-prediction period may divide the main prediction period into quartiles. According to this embodiment, the plurality of debt repayment capability evaluation models 260 may evaluate the debt repayment capability for each quartile.

In an embodiment, the first calculator 230 may calculate the insolvency probability calculated for each sub-prediction period as a target grade according to a section located on the reference insolvency probability distribution. In this case, the reference insolvency probability distribution means a probability distribution of a plurality of test insolvency probabilities calculated by inputting a plurality of test data into each debt repayment ability evaluation model 260. On the reference insolvency probability distribution, sections of each grade may be set according to the composition ratio of each grade.

For example, as shown in FIGS. 3A and 3B, a grade may be set for each section according to the composition ratio of each grade on the reference insolvency probability distribution. Accordingly, the first calculation unit 230 may calculate a target grade according to a section in which the insolvency probability calculated for each sub-prediction period is located on the reference insolvency probability distribution. In FIG. 3B, it is shown that there are seven grades, but this is only an example, and there may be six or less grades, and eight or more grades may exist.

The reason why the first calculation unit 230 according to the present invention calculates the insolvency probability calculated for each debt repayment capability evaluation model 260 as a target grade, is that each debt repayment capability evaluation model 260 has different expected insolvent periods. Since it was created by learning the financial data, non-financial data, and derivative data of the insolvent company generated in the insolvency, it is impossible to determine the superiority between each insolvency probability, so the insolvency probability calculated in each debt repayment ability evaluation model (260) is converted into a target grade. Therefore, it is to calculate as one result value by determining the superiority between target grades.

In one embodiment, the first calculation unit 230 may determine the enterprise type of the target company according to the classification criteria, and select a plurality of debt repayment capability evaluation models 260 according to the determined enterprise type of the target company. In this case, the classification criteria may include the type of business, the size of the company, whether or not a company has external audit. In the case of this embodiment, sub-prediction by inputting the financial data, non-financial data, and derivative data of the target company into a plurality of debt repayment capability evaluation models 260 selected according to the first calculation unit 230, the company type of the target company. It is possible to calculate the target grade for each period.

For example, when the target company is determined to be an externally audited company, the first calculation unit 230 inputs financial data, non-financial data, and derivative data of the target company into a plurality of debt repayment ability evaluation models 260 for the externally audited company. You can calculate the target grade for each prediction period.

The second calculation unit 240 evaluates the final debt repayment capability of the target company using at least one evaluation result of the debt repayment capability evaluation results calculated for each debt repayment capability evaluation model 260 by the first calculation unit 230. Yields results.

In one embodiment, the second calculation unit 240 is the median value of the debt repayment capabilities of the target company calculated for each debt repayment capability evaluation model 260, the lowest evaluation result among the debt repayment capability evaluation results of the target company. , And one of the highest evaluation results among the evaluation results of the target company's debt repayment ability may be selected. The second calculation unit 240 may calculate a final debt repayment capability evaluation result of the target company by using the selected evaluation result.

For example, the second calculation unit 240 may select the lowest evaluation result among the debt repayment capabilities of the target company, and calculate the selected lowest evaluation result as the final debt repayment capability evaluation result of the target company.

In one embodiment, the second calculation unit 240 calculates a target debt repayment capability level of the target company using a plurality of target grades calculated by the first calculation unit 230. Specifically, the second calculation unit 230 may calculate a target debt repayment capability level of the target company by using a final grade that is one of a plurality of target grades calculated for each sub-prediction period.

In one embodiment, the second calculation unit 240 is for each debt repayment ability evaluation model 260 according to the distance on the time axis from the evaluation point to a reference point included in each sub-prediction period of the plurality of debt repayment ability evaluation models. The target debt repayment capability level of the target company can be calculated using a table in which the priorities of a plurality of grades are recorded. At this time, the time point of evaluation means the time point at which the debt repayment ability is evaluated. The reference point may include the start point or end point of each prediction period.

To this end, the second calculation unit 240 includes a table generation unit 410, a table storage 420, a rating selection unit 430, a level determination unit 440, and a score calculation unit 450 as shown in FIG. 4. ) Can be included.

The table generation unit 410 generates a table in which the priorities of a plurality of grades for each debt repayment ability evaluation model 260 are recorded.

For example, when there are n debt repayment capability evaluation models 260 and m grades, the table generator 410 may generate a table having n*m fields. At this time, the priority of the grade for each debt repayment capability evaluation model may be recorded in each field.

For example, as illustrated in FIG. 5A, when four first to fourth models exist and seven grades exist, the table generator 410 may generate a table having 28 fields.

In one embodiment, the table generation unit 410 is in the field of each table according to the distance on the time axis from the evaluation point to the reference point of the prediction period of each debt repayment ability evaluation model 260 ( 260) Priority of a plurality of stars can be recorded.

In one embodiment, the table generator 410 has a higher priority as the distance on the time axis from the evaluation point to the reference point of each prediction period is shorter for the same grades of the different debt repayment ability evaluation models 260 on the table. The different grades of the same debt repayment ability evaluation model 260 may be recorded, and the higher the grade, the higher the priority. In this case, the higher the grade, the worse the debt repayment ability may mean. In contrast to the above-described embodiment, on the contrary, when a lower grade means that the debt repayment ability is poor, the table generation unit 410 prioritizes different grades of the same debt repayment ability evaluation model 260 as the lower grade. Can be recorded high.

For example, as shown in FIG. 5A, when the distance on the time axis from the evaluation point to the reference point of the prediction period increases sequentially in the first to fourth models, the table generator 410 has priority among the same grades. It can be recorded sequentially lower in the 1st model to the 4th model. In addition, when the higher the grade means that the debt repayment ability is poor, the table generator 410 may record the higher priority as the grade increases between different grades. In FIG. 5A, the smaller the priority value is, the higher the priority is, but this is only an example, and the higher the priority value, the higher the priority.

In an embodiment, the table generator 410 may group a plurality of grades for each debt repayment capability evaluation model on the table into at least one group to set the debt repayment capability level for each group. As an example, the table generation unit 410 groups some of the plurality of grades of the debt repayment ability evaluation model 260 on the table between the same grades of the different debt repayment ability evaluation models 260, and the other grades are the same. It is possible to group between different successive grades of the debt repayment ability evaluation model 260. Accordingly, the table generator 410 may set the debt repayment capability level for the grouped groups.

For example, as shown in FIG. 5A, the table generator 410 may group the first to fourth models into a first group and set the first group to the first level. The table generator 410 may group the first to fourth models of the second grade into a second group and set the second group to a second level. The table generator 410 may group the first to fourth models into a third group between the third grades and set the third group to a third level. The table generator 410 may group the 4th grades of the first to fourth models into a fourth group and set the fourth group to a 4th level. The table generator 410 may group the seventh grades of the first to fourth models into a seventh group and set the seventh group to a seventh level.

In addition, the table generator 410 may group the 5th to 6th grades of the first model into the 6B group and set the 6B group to the 6B level. The table generator 410 may group the 5th to 6th grades of the second model into the 6A group and set the 6A group to the 6A level. The table generator 410 may group the 5th to 6th grades of the third model into the 5B group and set the 5B group to the 5B level. The table generator 410 may group the 5th to 6th grades of the fourth model into the 5A group and set the 5A group to the 5A level.

As shown in FIG. 5B, the level set by the table generation unit 410 may mean that the debt repayment ability of the target company decreases in ascending order. Levels 1 and 2 can mean that the target company's debt repayment ability is normal, levels 3 and 4 mean that the target company's debt repayment ability needs to be observed, and levels 5 to 7 have problems with the debt repayment ability. It can mean that there is.

At this time, the 5A level may mean that the target company has a high probability of occurrence of insolvent events within 9 to 12 months from the time of evaluation, and the 5B level indicates that the target company has a high probability of occurrence of insolvency events within 6 to 9 months from the time of evaluation. It can mean that. In addition, level 6A may mean that the target company has a high probability of occurrence of insolvent events within 3 to 6 months from the time of evaluation, and level 6B means that the target company has a high probability of occurrence of insolvency events within 3 months from the time of evaluation. I can. Level 7 can mean that a bad event has already occurred in the target company.

The table generation unit 410 stores the generated table in the table storage 420.

In the above-described embodiment, it has been described that the table generation unit 410 is included in the second calculation unit 240, but the table generation unit 410 may be included in the debt repayment ability evaluation system 100 in a separate configuration.

The table storage 420 stores tables created by the table generation unit. The table storage 420 may be implemented in the form of a database.

The grade selection unit 430 may select a final grade with the highest priority matching each target grade among a plurality of target grades using a table stored in the table storage 420.

For example, as shown in FIG. 5A, the first calculation unit 230 uses 4 grades of the first model (priority: 13), 3 grades of the second model (priority: 18), and 5 of the third model. When the grade (priority: 10) and the sixth grade (priority: 11) of the fourth model are calculated, the grade selection unit 430 selects the fifth grade (priority: 10) of the third model with the highest priority. It can be selected as the final grade.

The level determination unit 440 determines the final grade selected by the grade selection unit 430 as the target debt repayment capability level of the target company. In one embodiment, the level determination unit 440 may determine the debt repayment capability level set in the group including the final grade as the target debt repayment capability level of the target company.

For example, as shown in FIG. 5A, when the grade 5 is selected as the final grade by the grade selection unit 430, the level determination unit 440 targets the 5B level set in the group including the field where the grade 5 is located. It can be determined by the target debt repayment capability level of the company.

The level determination unit 440 transmits the determined debt repayment capability level to the user terminal 10.

The score calculation unit 450 may convert the target debt repayment capability level into a debt repayment capability score. In one embodiment, the score calculation unit 450 inputs the priority recorded in the field where the final grade selected by the grade selection unit 430 is located in a predetermined equation to determine the target debt repayment capability level of the target company. It can be converted into debt repayment ability points. As an example, the predetermined equation may be Equation 1 below.

In Equation 1, Y denotes a debt repayment ability score, and X denotes priority.

For example, as shown in FIG. 5A, when the 5th grade is selected by the grade selection unit 430, the score calculation unit 450 inputs 10, which is the priority recorded in the field where the 5th grade is located, into Equation 1 Thus, the debt repayment ability score can be calculated. Accordingly, the first calculation unit 450 may calculate a debt repayment ability score of 37 points.

The reason why the score calculation unit 450 according to the present invention calculates the debt repayment capability score is to provide users with accurate debt repayment capability by further subdividing the debt repayment capability level of the target company.

The score calculation unit 450 transmits the calculated debt repayment capability score to the user terminal 10.

Meanwhile, in the above-described embodiment, the second calculation unit 240 calculates a plurality of target grades as the target debt repayment capability level using a table in which the priorities are recorded, but in a modified embodiment, the second calculation unit 240 A plurality of target grades can be calculated as target debt repayment capability levels using weights for each grade.

Referring back to FIG. 2, when the second calculation unit 240 receives information on occurrence of a non-donation event from the data acquisition unit 210, the second calculation unit 240 may calculate the level of the debt repayment capability of the target company as a level representing the actual occurrence of the bad event. have. For example, as shown in FIG. 5B, the second calculation unit 240 may calculate a 7 level when a non-donation event occurs in the target company.

In the database 245, the modeling unit 240 stores financial and non-financial data of companies (hereinafter referred to as'comparative companies') in order to generate the debt repayment capability evaluation model 260, and the derivative data generation unit 220 The derivative data of the comparative company generated by is stored. At this time, the target company may be included in the comparative company.

In one embodiment, the database 245 may be stored by acquiring financial data and non-financial data of comparative companies from the external server 20 by the data acquisition unit 210.

The modeling unit 250 generates a debt repayment ability evaluation model 260. In an embodiment, the modeling unit 250 may generate an AI-based debt repayment ability evaluation model 260 using a gradient boosting machine (GBM) algorithm. In this case, the GBM algorithm is an algorithm that generates a decision tree sequentially and corrects an error in the previous tree to derive a result.

Unlike the above-described embodiment, the modeling unit 250 may generate an AI-based debt repayment capability evaluation model 260 using a random forest algorithm, a deep neural network (DNN) algorithm, or the like.

In one embodiment, the modeling unit 250 generates a plurality of debt repayment capability evaluation models 260 by learning financial data, non-financial data, and derivative data of a comparative company stored in the database 245.

In particular, the modeling unit 250 according to the present invention learns the financial data, non-financial data, and derivative data of the insolvent company in which the insolvent event occurred among the comparative companies, and evaluates the debt repayment ability of the target company for each sub-prediction period. A repayment ability evaluation model 260 is created. In this case, the insolvency event may include at least one of short-term overdue, public overdue, financial overdue, card overdue, commerce overdue, and closed business.

In one embodiment, the modeling unit 250 includes financial data, non-financial data, and financial data, non-financial data, and By learning the derived data, each debt repayment ability evaluation model can be created.

Hereinafter, with reference to FIG. 6, it will be described in more detail that the modeling unit 250 generates a plurality of debt repayment ability evaluation models 260.

6 is a view showing that the modeling unit 250 generates a plurality of debt repayment ability evaluation models 260 according to an embodiment of the present invention. As shown in FIG. 6, the modeling unit 250 includes a first learning unit 610, a second learning unit 620, a third learning unit 630, and a fourth learning unit 640. In addition, the debt repayment ability evaluation model 260 generated by the modeling unit 250 includes a first model 660, a second model 670, a third model 680, and a fourth model 690. do.

In FIG. 6, the modeling unit 250 is shown to include four learning units 610 to 640, but there may be three or less learning units or five or more learning units. Accordingly, the number of four models 660 to 690 generated by the modeling unit 260 may be three or less, or five or more. Hereinafter, for convenience of description, it will be described that there are four learning units 610-640 and four models 660-690.

The first learning unit 610 generates a first model 660 by learning financial data, non-financial data, and derivative data of a non-performing company in which a non-performing event has occurred among the comparison companies stored in the database 245. Specifically, the first learning unit 610 includes the financial data, non-financial data, and derivative data of the insolvent company in which the insolvency event has occurred, from the insolvency point where the insolvency event occurs to the first insolvency point before the insolvency point. The first model 660 is generated by learning the financial data, non-financial data, and derivative data generated during the first insolvency period.

For example, the first scheduled insolvency may mean three months before the insolvency. When following this example, the first learning unit 610 learns the financial data, non-financial data, and derivative data of the insolvent company that occurred 3 months ago from the insolvency point in which the insolvent event occurred to create the first model 660. Can be generated.

The second learning unit 620 generates a second model 670 by learning financial data, non-financial data, and derivative data of a non-performing company in which a non-performing event has occurred among the comparison companies stored in the database 245. Specifically, the second learning unit 620 is for the second expected insolvent period from the first expected insolvent time to the second expected insolvent time before the first expected insolvency among financial data, non-financial data, and derivative data of the insolvent company. The generated financial data, non-financial data, and derived data are learned to generate a second model 670.

For example, the second scheduled insolvency may mean three months before the first insolvency scheduled time. When following this example, the second learning unit 620 learns the financial data, non-financial data, and derivative data of the insolvent company generated between 3 months to 6 months before the occurrence of the insolvency event, and the second model 670 ) Can be created.

The third learning unit 630 generates a third model 680 by learning financial data, non-financial data, and derivative data of a non-performing company in which a non-performing event has occurred among the comparison companies stored in the database 245. Specifically, the third learning unit 630 occurs during the third expected insolvency period from the second learning point to the third expected insolvency point before the second expected insolvency point among the financial data, non-financial data, and derivative data of the insolvent company. The third model 680 is generated by learning the generated financial data, non-financial data, and derived data.

For example, the third expected insolvency may mean three months before the second expected insolvency. When following this example, the third learning unit 630 learns the financial data, non-financial data, and derivative data of the insolvent company generated between 3 months to 6 months before the occurrence of the insolvency event, and the third model 680 ) Can be created.

The fourth learning unit 640 generates a fourth model 690 by learning financial data, non-financial data, and derivative data of a non-performing company in which a non-performing event has occurred among the comparison companies stored in the database 245. Specifically, the third learning unit 640 is for the fourth expected insolvency period from the third expected insolvency point to the fourth expected insolvency point before the third expected insolvency point among the financial data, non-financial data, and derivative data of the insolvent company. The generated financial data, non-financial data, and derived data are learned to generate a fourth model 690.

For example, the fourth expected insolvency may mean three months before the third expected insolvency. When following this example, the fourth learning unit 640 learns the financial data, non-financial data, and derivative data of the insolvent company generated between 9 months to 12 months before the occurrence of the insolvent event, and the fourth model 690 ) Can be created.

Referring back to FIG. 2, in an embodiment, the modeling unit 250 may generate a plurality of debt repayment capability evaluation models 260 for each type of company of the insolvent company in which the insolvent event has occurred. In this case, the enterprise type may be set as a classification criterion for classifying the enterprise. In one embodiment, the classification criterion may include a type of business, a company size, whether an externally audited company, and the like. Here, the externally audited company means a company that is obligated to undergo an accounting audit.

For example, when the classification criterion is the size of the company and whether it is an externally audited company, the modeling unit 250 may generate a plurality of debt repayment capability evaluation models 260 for each type of enterprise classified according to the size of the comparative company and whether or not it is an externally audited company. have. In this case, the type of enterprise may include externally audited enterprises, non-exclusively audited enterprises, small and medium-sized enterprises (hereinafter referred to as'small enterprises'), and individual entrepreneurs.

When following this example, the modeling unit 250 generates a plurality of debt repayment capability evaluation models 260 using financial data, non-financial data, and derivative data of a non-performing company corresponding to an externally audited company among a plurality of comparative companies. I can. In addition, the modeling unit 250 may generate a plurality of debt repayment capability evaluation models 260 using financial data, non-financial data, and derivative data of a non-performing company corresponding to a non-extraordinary company among a plurality of insolvent companies. In addition, the modeling unit 250 may generate a plurality of debt repayment capability evaluation models 260 by using financial data, non-financial data, and derivative data of a bad company corresponding to a small company among a plurality of bad companies. In addition, the modeling unit 250 may generate a plurality of debt repayment capability evaluation models 260 using financial data, non-financial data, and derivative data of a bad company corresponding to an individual business among a plurality of bad companies.

According to this embodiment, the modeling unit 250 may generate k*l debt repayment capability evaluation models 260 according to k enterprise types and l prediction periods.

For example, the first to fourth learning units 610 to 640 included in the modeling unit 250 learn from the financial data, non-financial data, and derivative data of the external audit enterprise in which the insolvency event has occurred, and You can create a model.

In addition, the first to fourth learning units 610 to 640 may generate the first to fourth models of non-instrumented companies by learning from financial data, non-financial data, and derivative data of non-instrumented companies in which the insolvency event has occurred. .

In addition, the first to fourth learning units 610 to 640 may generate first to fourth small business models by learning from financial data, non-financial data, and derivative data of a small company in which an insolvent event has occurred.

In addition, the first to fourth learning units 610 to 640 may generate first to fourth private business models learned from financial data, non-financial data, and derivative data of an individual business operator in which an insolvent event has occurred.

The modeling unit 250 may input test data into the generated plurality of debt repayment capability evaluation models 260 to verify whether the comparison company's actual insolvency probability and the match. In this case, the test data may include financial data, non-financial data, and derivative data of the comparative company. In an embodiment, the modeling unit 250 may verify a plurality of debt repayment capability evaluation models 250 using Area Under the ROC Curve (AUC) or Kolmogorov Smirnov Statistics. For example, as shown in FIG. 7, the modeling unit 250 may verify a plurality of debt repayment capability evaluation models.

According to this embodiment, the first to fourth learning units 660 to 690 included in the modeling unit 250 respectively verify the first to fourth models, or the modeling unit 250 A separate verification unit (not shown) for verifying the repayment capability evaluation model 260 may be further included.

The plurality of debt repayment capability evaluation models 260 receive financial data and non-financial data of a target company and calculate a probability of insolvency at which insolvency events will occur in the target company for each predetermined sub-prediction period. Specifically, when the financial data and non-financial data of the target company are input in the plurality of debt repayment capability evaluation models 250, each debt repayment capability evaluation model is targeted for each predetermined sub-prediction period. Calculate the probability of insolvency at which insolvent events will occur in a company. In one embodiment, the plurality of debt repayment capability evaluation models 250 additionally receive derivative data by the first calculation unit 230, and calculate the insolvency probability that a non-performing event will occur in the target company for each sub-prediction period. Can do it

In an embodiment, the sub-prediction period may be determined as at least one of daily, monthly, and quarterly. In the case of this embodiment, the plurality of debt repayment capability evaluation models 260 may evaluate the debt repayment capability of the target company according to at least one forecast period of daily, monthly, and quarterly.

In an embodiment, the sub-prediction period may divide the main prediction period into quartiles. When following this embodiment, the plurality of debt repayment ability evaluation models 260 may evaluate the debt repayment ability for each quintile.

In one embodiment, each debt repayment capability evaluation model 260 divides the main prediction period into a plurality of sub prediction periods, and may be learned for each insolvency expected period of a non-performing company corresponding to each divided sub prediction period. Specifically, each debt repayment capability evaluation model 260 is the financial data and non-financial data of the insolvent company generated during the expected insolvent period corresponding to each sub-prediction period among the financial data, non-financial data, and derivative data of the insolvent company in which the insolvent event occurred. Data, and derived data can be learned.

Hereinafter, a plurality of debt repayment ability evaluation models 260 will be described in more detail with reference to FIG. 6.

As shown in FIG. 6, the plurality of debt repayment capability evaluation models 260 may include a first model 660, a second model 670, a third model 680, and a fourth model 690. have. As described above, the number of four models 660 to 690 may be three or less, and may be five or more.

When the financial data, non-financial data and derivative data of the target company are input by the first calculation unit 230, the first model 660 is applied to the target company during the first sub-prediction period from the time of evaluation to the time of the first sub-prediction. Calculate the probability of insolvency that an insolvency event will occur. In this case, the first prediction point refers to a point in time later than the evaluation point.

For example, the first sub-prediction time point may mean three months after the evaluation time point. Following this example, the first model 660 may calculate a probability of insolvency at which insolvency events will occur in the target company from the time of evaluation to three months.

In one embodiment, the first model 660 is a first insolvency period from the insolvency point where the insolvency event is generated in the insolvent company by the first learning unit 610 to the first insolvency schedule before the insolvency point. The generated financial data, non-financial data, and derivative data of the insolvent company may be learned.

Following this embodiment, the first expected insolvent period may correspond to the first sub-prediction period. For example, if the first sub-prediction period is 3 months from the time of evaluation, the first expected insolvency period is 3 months from the time of non-conduct.

When the financial data, non-financial data, and derivative data of the target company are input by the first calculation unit 230, the second model 670 is applied during the second sub-prediction period from the first sub-prediction time to the second sub-prediction time. Calculate the probability of insolvency that an insolvent event will occur in the target company. In this case, the second sub-prediction time point means a time point later than the first sub-prediction time point.

For example, the second sub-prediction time may mean three months after the first sub-prediction time. Following this example, the second model 670 may calculate a probability that an insolvency event will occur in the target company from 3 months to 6 months from the time point of evaluation.

In one embodiment, the second model 670 is generated by the second learning unit 620 during the second expected insolvency period from the first expected insolvency to the second expected insolvency before the first expected insolvency. Financial data, non-financial data, and derivative data of the insolvent company may be learned.

Following this embodiment, the second insolvency expected period may correspond to the second sub-prediction period. For example, if the second sub-prediction period is from 3 months to 6 months from the point of evaluation, the second expected period of insolvency is from 3 months to 6 months from the point of insolvency.

When the financial data, non-financial data, and derivative data of the target company are input by the first calculation unit 230, the third model 680 is performed during the third sub-prediction period from the second sub-prediction time to the third sub-prediction time. Calculate the probability of insolvency at which insolvency events will occur in the target company. In this case, the third sub-prediction time point means a time point later than the second sub-prediction time point.

For example, the third sub-prediction time may mean three months after the second sub-prediction time. Following this example, the third model 680 may calculate the probability that a bad event will occur in the target company from 6 months to 9 months from the time of evaluation.

In one embodiment, the third model 680 is generated by the third learning unit 630 during the third expected insolvency period from the second expected insolvency to the third expected insolvency before the second expected insolvency. Financial data, non-financial data, and derivative data of the insolvent company may be learned.

Following this embodiment, the third expected insolvency period may correspond to the third sub-prediction period. For example, if the third sub-prediction period is from 6 months to 9 months from the time of evaluation, the third expected insolvency period is from 6 months to 9 months from the time of non-implementation.

In the fourth model 690, when financial data, non-financial data, and derivative data of the target company are input by the first calculation unit 230, during the fourth sub-prediction period from the third sub-prediction time to the fourth sub-prediction time. Calculate the probability of insolvency that an insolvent event will occur in the target company. In this case, the fourth sub-prediction time point means a time point later than the third sub-prediction time point.

For example, the fourth sub-prediction time may mean three months after the third sub-prediction time. Following this example, the fourth model 690 may calculate a probability that an insolvent event will occur in the target company from 9 months to 12 months from the time point of evaluation.

In one embodiment, the fourth model 690 is generated by the fourth learning unit 640 during the fourth expected insolvency period from the third expected insolvency to the fourth expected insolvency before the third expected insolvency. Financial data, non-financial data, and derivative data of the insolvent company may be learned.

Following this embodiment, the fourth insolvency expected period may correspond to the fourth sub-prediction period. For example, if the fourth sub-prediction period is from 9 months to 12 months from the time of evaluation, the fourth expected period of insolvency is from 9 months to 12 months from the time of non-implementation.

The first to fourth models 660 to 690 according to the present invention can calculate the insolvency probability of the target company in the first to fourth sub-prediction periods as shown in FIG. 8A. In a modified embodiment different from the above-described embodiment, one first model 660 may calculate the insolvency probability of the target company in the first to fourth sub-prediction periods, respectively, as shown in FIG. 8B. In addition, in another modified embodiment, one first model 660 may calculate the insolvency probability during the total prediction period of the first to fourth sub-prediction periods.

In an embodiment, the first to fourth models 660-690 may be generated for each company type by the first to fourth learning units 610-640. In this case, the enterprise type includes at least one of an externally audited enterprise, a non externally audited enterprise, a small enterprise, and an individual business entity.

When following this embodiment, the first model to the fourth model (660-690) may be a plurality of each according to the type of company.

For example, when the type of enterprise is an externally audited enterprise, a non-exclusively audited enterprise, a small enterprise, and an individual entrepreneur, the first to fourth models 660-690 may be the first to fourth externally audited enterprise models, and the first to fourth models (660-690) may be the 1st to 4th non-exempted enterprise model. Further, the first to fourth models 660-690 may be first to fourth small business models, and the first to fourth models 660-690 may be first to fourth private business models. Accordingly, four first to fourth models 660 may be generated by the learning units 610-640.

Meanwhile, the debt repayment capability evaluation system 100 may further include a debt repayment capability evaluation correction model 280 and a correction unit 290 in order to improve the accuracy of the debt repayment capability as shown in FIG. 2.

The first calculation unit 230 inputs the financial data, non-financial data, and derivative data of the target company into the plurality of debt repayment capability evaluation correction models 280, and additionally inputs the commercial data of the target company to target each forecast period. The company's corrected failure probability is calculated, and the corrected failure probability calculated for each forecast period is calculated as a target correction grade according to a predetermined criterion. In one embodiment, the commercial transaction data is internal transaction history data, and may include unrecovered information on the receivables.

The second calculation unit 240 calculates a target debt repayment ability correction level of the target company using a plurality of target correction grades calculated by the first calculation unit 230. Specifically, the second calculation unit 230 may calculate a target debt repayment ability correction level of the target company using a final correction grade, which is one of a plurality of target correction grades calculated for each prediction period.

The modeling unit 250 additionally learns not only the financial data, non-financial data, and derivative data of the insolvent company where the insolvent event occurred, but also the commercial data of the insolvent company, and the debt repayment ability of the target company within a predetermined sub-prediction period. A plurality of debt repayment ability evaluation correction models 280 are generated for evaluating.

The reason that the modeling unit 250 according to the present invention additionally learns commercial transaction data to generate the debt repayment ability evaluation correction model 280 is, in general, the evidence that appears first in the progress of corporate insolvency is delinquent in transaction For this reason, by learning the commercial transaction data including this information to generate a plurality of debt repayment ability evaluation correction models 280, it is possible to more accurately evaluate the debt repayment ability of the target company.

The plurality of debt repayment capability evaluation correction models 280 calculates the corrected insolvency probability at which insolvency events will occur in the target company for each predetermined sub-prediction period. Specifically, when the plurality of debt repayment ability evaluation correction models 280 are inputted by the first calculation unit 230 of financial data, non-financial data, derivative data, and commercial data of the target company, each debt repayment ability evaluation correction model 280 calculates a corrected insolvency probability that a bad event will occur in the target company for each predetermined sub-prediction period.

The correction unit 290 compares the target debt repayment ability evaluation level and correction level calculated by the second calculation unit 240, and when the debt repayment ability evaluation level and the debt repayment ability evaluation correction level differ by more than a predetermined reference value, The target debt repayment ability evaluation level is corrected to the target debt repayment ability evaluation correction level. The correction unit 290 transmits the corrected target debt repayment ability evaluation correction level to the user terminal 10.

On the other hand, the debt repayment ability evaluation system 100 may further include a payment capacity evaluation unit 295.

The payment capacity evaluation unit (295) is a marginal borrowing inference model (not shown) generated by learning through correlation analysis between the borrowing information among the financial data of a company with financial statements and the credit contribution information among the non-financial data of the company, and appropriate borrowings. Using an inference model (not shown), the payment capacity of the target company can be calculated. At this time, marginal borrowing means a borrowing that the target company can bear up to the limit in consideration of operating activities, assets, and industrial average, and the appropriate borrowing means the borrowing that the target company can bear in consideration of business activities. The payment capacity may refer to the range of marginal borrowings and the amount of appropriate borrowings.

Referring back to FIG. 1, the external server 20 collects financial data and non-financial data for the target company based on the financial statements for the target company and transmits it to the debt repayment capability evaluation system 100. In an embodiment, when a non-donation event occurs in the target company, the external server 20 may transmit information on the occurrence of the non-donation event to the second calculation unit 240 through the data acquisition unit 210.

Those skilled in the art to which the present invention pertains will be able to understand that the above-described present invention can be implemented in other specific forms without changing the technical spirit or essential features thereof.

Therefore, it should be understood that the embodiments described above are illustrative in all respects and not limiting. The scope of the present invention is indicated by the claims to be described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

100: Enterprise's debt repayment ability evaluation system
210: data acquisition unit 220: derived data generation unit
230: first calculation unit 240: second calculation unit
250: modeling unit 260: debt repayment ability evaluation model
280: debt repayment ability evaluation correction model 290: correction unit
295: Payment power evaluation unit

Claims (17)

A data acquisition unit that acquires financial data obtained based on the financial statements of the target company and non-financial data other than the financial data;
The main prediction period from the evaluation point to the predetermined prediction point after the evaluation point is divided into n sub-prediction periods (n is a natural number of 2 or more), and the debt repayment ability evaluation model corresponding to each sub-prediction period is described above. A first calculation unit for inputting financial data and the non-financial data to calculate a debt repayment capability evaluation result of the target company for each of the debt repayment capability evaluation models; And
And a second calculation unit that calculates a final debt repayment ability evaluation result of the target company by using at least one evaluation result of the n debt repayment ability evaluation results calculated for each of the n debt repayment ability evaluation models,
The n debt repayment ability evaluation model calculates the debt repayment ability evaluation result of the target company by determining any one of m predetermined grades (m is a natural number of 2 or more) as a target grade,
The second calculation unit determines the priorities of target grades calculated by the n debt repayment ability evaluation models based on the n debt repayment ability evaluation models and the priorities assigned according to the m grades, and the priority A system for evaluating debt repayment ability of a company, characterized in that by calculating the highest target grade as the final grade and calculating the final debt repayment ability evaluation result of the target company. The method of claim 1,
The n debt repayment capability evaluation models are learned for each of the insolvency expected periods of the insolvent companies corresponding to each of the sub-prediction periods. The method of claim 2,
The n debt repayment ability evaluation models are formed into different evaluation models by learning the financial data and non-financial data of the insolvent company according to the expected insolvency period. The method of claim 1,
The n debt repayment ability evaluation models,
A first model in which the financial data and non-financial data of the insolvent company generated during the first insolvency period from the insolvent event at which the insolvent event occurs to the first insolvency schedule before the insolvency event is learned; And
And a second model in which the financial and non-financial data of the insolvent company generated during the second insolvent period from the first expected insolvent time to the second insolvent expected time before the first insolvency schedule are learned. A system for evaluating the debt repayment ability of a company. The method of claim 4,
Each of the sub-prediction periods divides the entire prediction period into quartiles, and the n debt repayment capacity evaluation models evaluate the debt repayment capacity for each quintile,
The n debt repayment ability evaluation models,
A third model in which financial and non-financial data of the insolvent company generated during a third insolvency period from the second expected insolvency to a third insolvency before the second expected insolvency are learned; And
And a fourth model in which financial and non-financial data of the insolvent company generated during the fourth insolvency period from the third expected insolvency point to the fourth insolvency expected point before the third insolvency point. A system for evaluating the company's debt repayment ability. The method of claim 1,
The n debt repayment ability evaluation model is an enterprise's debt repayment ability evaluation system, characterized in that the additional learning of derivative data generated from the financial data and non-financial data of the insolvent enterprise. delete delete The method of claim 1,
The second calculation unit,
And a score calculation unit for calculating a debt repayment capability score by inputting a priority matching the final grade into a predetermined equation. The method of claim 1,
The second calculation unit,
Debt repayment ability level is set for each group in which a plurality of grades per debt repayment ability evaluation model are grouped into at least one group, and a debt set in the group containing the final grade among a plurality of target grades calculated for each debt repayment ability evaluation model A debt repayment ability evaluation system of a company, characterized in that calculating a repayment capability level as a target debt repayment capability level of the target company. The method of claim 1,
The second calculation unit,
Debt by group, in which some of the plurality of ratings per debt repayment ability evaluation model are grouped between the same grades of different debt repayment ability evaluation models, and other grades are grouped between different successive grades of the same debt repayment ability evaluation model A debt repayment ability evaluation system of a company, characterized in that: setting a repayment capability level, and calculating a debt repayment capability level set in a group including the final grade among a plurality of target grades as a target debt repayment capability level of the target company. The method of claim 1,
The non-financial data includes at least one of the target company's corporate information, credit grant information, bidding information, contract information, national pension information, job posting information, fraudulent disposition information, and construction license disposition information. Repayment ability evaluation system. The method of claim 1,
The financial data includes at least one of financial statement information, financial ratio information, financial model information, and borrowing information. The method of claim 1,
Further comprising a derivative data generating unit for generating derived data correlated with the debt repayment capability from the financial data and non-financial data,
The first calculation unit calculates the debt repayment ability of the target company by additionally using the derivative data. The method of claim 1,
Further comprising a modeling unit for generating the n debt repayment capability evaluation models by learning the financial data, non-financial data, and derivative data of the insolvent company where the insolvent event has occurred for each predetermined insolvency period corresponding to each of the sub-prediction periods. A system for evaluating the debt repayment ability of a company, characterized in that. The method of claim 15,
The insolvency event includes at least one of short-term overdue, public overdue, financial overdue, card overdue, commerce overdue, and closed business. The method of claim 1,
Further comprising a modeling unit for generating the n debt repayment ability evaluation model,
The modeling unit generates each of the n debt repayment capability evaluation models for each company type,
The first calculation unit selects n debt repayment capability evaluation models that match the type of company to which the target company belongs, and calculates the debt repayment capability evaluation result of the target company for each of the n debt repayment capability evaluation models. A system for evaluating the debt repayment ability of a company.

Images