KR101323206B1 - Improved simulation distribution data managing method of credit var calculation system for reduction of data storage - Google Patents

Abstract

PURPOSE: A simulation distribution data management improving method for reducing the data storage capacity of a credit VaR calculation system is provided to easily calculate and offer portfolio credit VaR and VaR contribution, thereby shortening reading/writing time for a file or a database. CONSTITUTION: A parameter estimation unit (120) generates a reference table for a class transition threshold for each transition probability after constant periods of classes. A random number generation unit (130) generates asset profit rate random numbers for a transaction component for each scenario. An asset calculation unit (140) derives a predicted credit class for the transaction opponent for the random numbers and stores the information for the predicted credit class in a database (111). The asset calculation unit generates a reference prediction value table for the classes for each account of the transaction opponent and stores the same in the database. [Reference numerals] (110) Control unit; (111) Database; (120) Parameter estimation unit; (130) Random number generation unit; (140) Asset calculation unit; (150) VaR calculation unit (data restoring/inquiry is available)

Description

Improved Simulation Distribution Data Management Method for Credit VaR Calculation System for Reduction of Data Storage

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Credit VaR (Loss Amount) calculation system, which can reduce storage capacity and significantly reduce read / write time for a file or database, and a data processing method thereof. It is about.

The key management indicators of financial institutions, such as banks and securities firms, consist of size, profitability, and soundness.In general, soundness is assessed as the ratio of risks to financial institutions' holding capital. Possible loss amount (risk) ”. Risks are classified into credit risks, market risks, and operational risks. Among them, credit risks are assets such as loans and securities held by financial institutions due to bankruptcy (failure of default) or declining credit rating. Means the likelihood of a loss occurring.

A variety of statistical techniques are used to predict risk as the potential for indeterminate “future” losses, with VaR (Value at Risk) methodology being representative. VaR is an extreme lossable amount that can occur under a certain level of confidence. A 99% VaR is a 1% chance (100%-) in the light of past experience when a current asset is held for a period of time (eg 1 year). 99%) means the possible loss amount.

Credit VaR (Loss Amount) means the amount of loss that can be incurred by a counterparty's default or declining credit rating under a certain level of confidence, from the definition of credit risk and VaR as above. In Monte Carlo Simulation, we use the Monte Carlo Simulation technique to generate a large number of virtual yield variables (random numbers) to estimate the future creditworthiness of counterparties and calculate the amount of change in asset value by the number of times of simulation. From this, we calculate the lossable amount of a certain level of confidence, ie the extreme lossable amount of credit VaR.

Financial institutions use the Credit VaR calculation system to calculate Credit VaR and use it to manage credit risk. An expensive large-capacity data storage device is used to store a large number of random numbers and value distributions for the purpose of querying or reporting risks. I use it.

For example, to calculate the VaR Contribution by account and to view the calculation process, it is necessary to store and manage the value distribution for each scenario / account in a storage device. A commercial bank usually has about hundreds of thousands of accounts. For example, it is common to perform about 1,000 simulations. For example, considering 300,000 accounts, the storage capacity required to store the value distribution for each scenario / account is 300,000 accounts × 1,000 simulations × 8 Bytes ( Storage capacity of real data) = 2,400,000,000 Bytes = 2.4GB. In addition, the computer execution time for reading and writing data such as 2.4 GB or more into a database or a file varies depending on the specification, but usually reaches tens of minutes, causing problems in timeliness of business processing.

In addition, to increase the accuracy of simulations, the number of simulations must be increased. However, there are many cases in which sufficient simulation cannot be performed because the storage capacity increases in proportion to the number of simulations. Thus, to increase the timeliness and utilization of the Credit VaR calculation system, There is a need to improve the account-by-scenario-specific value distribution storage and retrieval capabilities.

Accordingly, an object of the present invention is to provide a data processing technique related to portfolio credit VaR, and to significantly reduce storage capacity and significantly reduce read / write time for a file or database. In order to be able to do so, only simple basic data can be stored and managed and then a simple relationship can be used to restore the large value distribution of accounts for multiple trading partners of a financial institution. The present invention provides a credit VaR calculation system that can calculate and provide VaR Contribution, and a data processing method thereof.

In addition, the present invention provides a Credit VaR calculation system and its data processing method that can easily derive a large value distribution for each scenario and portfolio account even if the number of simulations increases, thereby increasing timeliness and usability.

First, to summarize the features of the present invention, in accordance with an aspect of the present invention for achieving the object of the present invention, the Credit VaR data processing method, processing the data related to the calculation of Credit VaR in the Credit VaR calculation system A credit VaR data processing method comprising generating N * M asset return random numbers for M (natural number) counterparts per N (natural number) scenarios, and for each generated random number, each transition probability after a plurality of predetermined periods. N * M expected credit ratings are derived based on the table of the thresholds of the previously created rating transfer thresholds, and the information about the estimated credit ratings is stored in the database. Generate a reference expected value table for the account and store it in the database; for all accounts of the counterparty, the example The portfolio credit VaR is provided by estimating the portfolio expected value for each scenario based on credit rating information and the estimated values for each random number based on the standard expected value table. The portfolio credit VaR or the account is restored by applying the stored information about the expected credit rating and the reference expected value table to a relational expression to restore the expected values and portfolio expected values for each random number for all accounts of the counterparty. Characterized to calculate and provide a star Credit VaR Contribution.

The expected values of each random number and portfolio expected values for all the restored accounts are sorted in the order of the portfolio expected value, and the N for the input confidence level X% of the inquiry request among the sorted expected values. It is possible to output the X% portfolio credit VaR by subtracting Y = N (100-X) / 100th from the average of the portfolio's expected values.

The contribution of each account may be calculated as the Credit VaR Contribution for each account, based on the proportion of the expected value of all the accounts in a predetermined section before and after the Y-th portfolio expected value and the Y-th expected value.

And, according to another aspect of the present invention, the Credit VaR calculation system for processing data associated with the calculation of Credit VaR, the parameter for generating a reference table for the grade transition threshold for each transition probability after a certain period of a plurality of grades Estimator; A random number generator for generating N * M yield random numbers by the number of counterparts (M) for each simulation time (N); For each random number generated by the random number generation unit, information about the expected credit rating derived by deriving an expected credit rating for each of the counterparties of M (natural numbers) is stored in a database, and the plurality of accounts for each account held by the corresponding counterparty. An asset value calculator configured to generate a reference predicted value table for a grade and store it in the database; And providing portfolio credit VaR for all accounts of the counterparty by estimating expected value and portfolio expected value for each account for each scenario, based on the information on the estimated credit rating and the reference expected value table. And a VaR calculation unit, wherein the VaR calculation unit applies the information on the expected credit rating and the reference expected value table stored in the database to a relational expression according to the inquiry request, and applies the scenario to all accounts of the counterparty. Restoration of the discriminated value and the portfolio expected value can be used to calculate and provide portfolio credit var or account specific credit var contributions.

According to the Credit VaR calculation system and its data processing method according to the present invention, it is possible to store and manage only simple basic data and then to restore a large value distribution of accounts for a plurality of counterparties of a financial institution by using a simple relationship. Portfolio Credit VaR, VaR Contribution, etc. can be easily calculated and provided, thereby reducing storage capacity and drastically reducing the read and write time for a file or database.

In addition, even if the number of simulations increases, time value and usability can be improved by easily deriving a large value distribution for each scenario and portfolio account.

1 is a block diagram illustrating a credit var calculation system according to an embodiment of the present invention.
2 is a flowchart illustrating an operation of a credit VaR calculation system according to an embodiment of the present invention.
3 is an example of a reference table for the probability of transition to each credit rating after one year for each credit rating.
4 is an example of a reference table for a grade transition threshold for a BBB grade.
FIG. 5 is a diagram for describing a concept of a grade transition threshold of FIG. 4.
6 is an example of a reference expected value table for account credit rating.
7 is an example of the expected credit rating for each counterparty generated with reference to the table of FIG.
8 is an example of a holding account for each counterparty.
9 is an example of a portfolio expected value generated for a large number of counterparties.
10 is a flowchart illustrating the restoration of value distribution data.
11 is a diagram for explaining the portfolio VaR.
12 is a diagram for explaining VaR Contribution.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout.

Credit VaR calculation system 100 according to an embodiment of the present invention to be described below, in the financial institutions, such as banks, securities companies, loans held by financial institutions in accordance with the default (debt default) or credit rating of the counterparty, It is a system to manage financial soundness by predicting the possibility of loss (credit risk) on assets such as securities.It uses Monte Carlo Simulation technique to generate a large number of virtual yield variables (random numbers). It is a system to calculate the lossable amount of a certain confidence level, that is, the extreme lossable amount of credit VaR, from the value distribution derived by estimating the future creditworthiness of the counterparties and calculating the amount of change in asset value according to the yield. .

1 is a block diagram illustrating a credit var calculation system 100 according to an embodiment of the present invention.

Referring to Figure 1, Credit VaR calculation system 100 according to an embodiment of the present invention, the control unit 110, database 111, parameter estimation unit 120, random number generation unit 130, asset value calculation The unit 140 includes a VaR calculator 150. The controller 110 corresponds to a processor that is in charge of the overall control of all the components of the credit VaR calculation system 100 as described above, and may be implemented to include some of the functions of the other components as described above. All components of the Credit VaR calculation system 100 may be implemented in hardware, software, or a combination thereof.

First, the operation of each part of the credit VaR calculation system 100 will be briefly described, and the detailed description thereof will be described with reference to the flowcharts of FIGS. 2 and 10.

The parameter estimator 120 generates a reference table (see FIG. 4) for the grade transition threshold for each transition probability after a predetermined period of the plurality of grades.

The random number generator 130 generates a plurality of yield random numbers in each scenario round according to a Monte Carlo simulation technique. The random number generation unit 130 generates N * M asset yield random numbers as many as M (natural numbers) of the counterparty for each N (natural number) scenarios.

The asset value calculation unit 140 derives an expected credit rating for each of the counterparties of M (natural numbers) for each random number generated by the random number generator 130 (see FIG. 7). Is stored in the database 111, and generates a reference expected value table (see FIG. 6) for the plurality of grades for each account held by the counterparties and stores it in the database 111.

VaR calculation unit 150 for all the accounts of the counterparty (see Fig. 8), the scenario as described above based on the above information on the expected credit rating (see Fig. 7) and the reference expected value table (see Fig. 6) We estimate portfolio value and portfolio estimate (see portfolio estimate in Figure 9) for each account to provide portfolio credit VaR or account-specific credit vaR contribution.

In the present invention, as described above, simple information such as information on the estimated credit rating (see FIG. 7), a reference expected value table (see FIG. 6), and information on all accounts of the counterparty (see FIG. 8) are provided in the database 111. Only the basic data are stored, and according to the inquiry request, the VaR calculation unit 150 applies the predetermined relation using the stored information of the database 111, so that each random number of all the accounts of the trading counterpart (see FIG. 8) is applied. It was possible to calculate and provide the Portfolio Credit VaR or the Account-specific Credit VaR Contribution by restoring the expected values and the portfolio expected value (see FIG. 9).

Accordingly, according to the data processing technique of storing and managing only simple basic data in the Credit VaR calculation system 100 of the present invention, and then restoring a large value distribution of accounts for a plurality of trading partners of a financial institution using a simple relationship. From this, portfolio credit VaR, VaR Contribution, etc. can be easily calculated and provided, thereby reducing storage capacity and drastically reducing the read / write time for a file or database. Even with the increase, the large value distribution by scenario and portfolio account can be easily derived to increase timeliness and usability.

Hereinafter, the operation of the Credit VaR calculation system 100 according to an embodiment of the present invention will be described in more detail with reference to the flowchart of FIG. 2.

First, the parameter estimator 120 estimates simulation parameters required for Monte Carlo simulation (S110). The parameter estimator 120 obtains (collects) a table for the probability of transition to a credit rating after a predetermined period of time (for example, one year later) for a plurality of trading partners' current credit ratings as shown in FIG. 3. can do. Here, the credit rating is described by taking the ratings in the order of AAA, AA, A, BBB, BB, B, CCC, Default (D) from the top to the bottom, but is not limited to this, and financial institutions as needed A table of transition probabilities can be prepared by separating more or fewer credit ratings.

3, the parameter estimator 120 may generate a reference table (see FIG. 4) for a rating transition threshold for each transition probability after a predetermined period of credit ratings as described above. have. 4 is an example of a reference table for the grade transition threshold for the current BBB grade, and the reference table for the grade transition threshold may be generated in a similar manner for other grades AAA, AA, A, BB, B, CCC, and the like. . The reference table for the grade transition threshold is a section grade random transition (−∞ to + ∞) when generating a random number (in step S140) to have a standard normal distribution of N (0,1) as shown in FIG. 5. For example, in FIG. 4, when the yield random number is generated as a value between −1.1601 and +1.4632, the table refers to maintaining an expected credit rating for the current rating BBB. Each threshold of the reference table for the class transition threshold is a standard normal distribution N (0 (average), 1 (standard deviation)) to satisfy each transition probability centered on the BBB class having the highest probability as shown in FIG. It can be obtained by properly distinguishing the left and right sections of the random number (-∞ ~ + ∞) of.

In addition, in the above parameter estimating step, the asset value calculation unit 140 generates a reference predicted value table (see FIG. 6 for the account b1 of the counterparty B) of the credit ratings for each account as described above in advance. ) To save. This may be done at the time of account value calculation in step S160. The baseline predicted value table (see FIG. 6) represents the expected value of the account for the transition grade after a predetermined period (eg, after one year), and thus the predetermined time period for each transition grade included in the estimated value table. The later expected value may be calculated and stored in the database 111 based on a predetermined function relationship to the transition grade.

On the other hand, after the simulation parameter is estimated, the calculation of each step from S140 to S170 is repeated N times by the random number of N (natural number), and the counterpart of M (natural number) in each scenario round. The calculation of each step from S140 to S160 is repeatedly performed M times with respect to the different random numbers by. The same random number applies to all accounts for the same counterparty. Here, N may be freely input, for example, 1000 may be input as described below. In addition, the number M of trading partners can also be freely input. For example, as shown in FIG. 8, trading partners A, B, C, and D are trading partners A, accounts a1, a2, a3, and trading partners B, respectively. b1, the trading partner C may input information indicating that the accounts c1 and c2, and the trading partner D have the accounts d1 and d2.

In operation S140, the random number generator 130 generates a random number of yields (see FIG. 5) according to a Monte Carlo Simulation technique. In this case, as described above, a yield random number (see FIG. 5) may be randomly generated in left and right intervals of random numbers (−∞ to + ∞) of the standard normal distribution N (0, 1). In this manner, the random number generation unit 130 generates asset yield random numbers in each scenario round as M (natural numbers) of counterparts for each N (natural number) scenarios, and total assets for each scenario (N) scenarios. Generate N * M yield random numbers.

According to the random number generation, the asset value calculation unit 140 derives an estimated credit rating for each counterparty for each random number generated by the random number generation unit 130 (see FIG. 7). ) Is stored in the database 111 (S150). The counterparty's expected credit rating may be derived based on the rating transition threshold threshold table (see reference to Figure 4), which is based on the rating threshold section in which random numbers are generated.

In addition, the asset value calculation unit 140 generates a reference predicted value table (see FIG. 6) that defines the estimated values for each of the plurality of credit ratings as described above, and stores the reference value table in the database 111. For all accounts of the counterparty (see Fig. 8), the estimated values for each of the scenario rounds above (based on the above information on the expected credit rating (see Fig. 7) and the baseline expected value table (see Fig. 6) ( Reference value estimates for each account of FIG. 9) may be calculated (S160). For example, for each of the accounts a1, a2, and a3 of the counterparty A, each scenario turn as shown in FIG. 9 with reference to the anticipated credit rating of the counterparty A of FIG. 7 and the corresponding estimated credit rating of FIG. 6. The expected value of can be calculated.

Each of the steps S140 to S160 is repeated by the number M of counterparties to calculate the expected value for the corresponding random number as shown in FIG. 9, and for all accounts of the counterpart counterpart (see FIG. 8) in M iterations. Based on the same expected credit rating information (see Figure 7) and the baseline expected value table (see Figure 6), the estimated values for that random number (see account-specific estimates in Figure 9) are computed. The calculator 150 may calculate a portfolio expected value (see portfolio expected value of FIG. 9) for each scenario, which is the sum of the expected values of all the accounts for the random number (S170).

Each of the steps S140 to S170 is repeated by the number of occurrences of the random number N to calculate the expected value and the portfolio expected value for each account for each scenario as shown in FIG. 9, and the VaR calculator 150 generates the scenario as shown in FIG. 9. Portfolio Credit VaR or account-specific Credit VaR Contribution may be provided based on the expected value of each account and the portfolio expected value of each account (S180, S190).

In particular, in the present invention, in the process of generating data to provide portfolio credit VaR or account-specific Credit VaR Contribution according to each process as described above, to reduce the storage capacity and to read and write the file or database In order to be greatly shortened, a large number of financial institutions can be stored using simple relational expressions by storing and managing only basic data (Figs. 6, 7, and 8) without storing information on the large value distribution for each account as shown in FIG. It is possible to restore the large value distribution of accounts to counterparties in. Even if the number of simulations (N), the number of counterparties (M), or the number of accounts increases, the large value distribution by account as shown in FIG.

The value distribution data restoration in the Credit VaR calculation system 100 according to an embodiment of the present invention will be described with reference to the flowchart of FIG. 10.

The user or operator can execute the Credit VaR calculation system 100 to restore the expected values and portfolio expected values for each random number (refer to FIG. 9), or to query the portfolio Credit VaR or Credit VaR Contribution by account. The inquiry can be requested by inputting necessary inquiry request information through a user interface to a device such as a computer.

For example, the user or the operator may input the confidence level X (%) of the portfolio credit VaR as the inquiry request information (S210), and the VaR calculator 150 may calculate Y = N ( 100-X) / 100 may be calculated in advance (S220).

On the other hand, if the inquiry request as described above, in order to restore the expected value and portfolio expected value for each account for each scenario, as shown in Figure 9, the information on the estimated credit rating stored in the database 111 in the process of FIG. Using only simple basic data, such as the reference predicted value table (see FIG. 6) and the reference predicted value table (see FIG. 6), the VaR calculator 150 applies this information to a given relation, thereby all accounts of the counterparty (FIG. 8). (Refer to FIG. 9) to calculate and provide a portfolio Credit VaR (S240) or an account-specific Credit VaR Contribution (contribution) (S250) for each random number. Can be.

For example, as shown in [Relationship 1], the index i of the simulation N and information (trade partner ID) I3 (a) (a is an account ID) for all accounts of the trading partner (see FIG. 8) are variables. Using the function I1, information (see FIG. 7) about the anticipated credit ratings of the counterparts of the trading partners may be calculated, and the function I2 (a, r) (r = I1) can be used to restore the expected values and portfolio expected values for each random number (see FIG. 9) O = I2 for all accounts of the trading counterpart (see FIG. 8). That is, I3 (a function that receives an account ID and outputs a counterparty ID of the corresponding account with reference to the relationship of FIG. 8) receives I1 (a simulation turn (i) and a counterpart ID) as shown in FIG. 7. Information on the expected credit rating for each scenario is used. In addition, information on the estimated credit rating for each scenario by I1 (see FIG. 7) is used for I2 (a function that receives the account ID and the expected credit rating and outputs the estimated value by referring to the table of FIG. 6). Discrimination For each account, the corresponding estimated value for each expected credit rating (see Figure 9) is extracted, and the portfolio expected value for each scenario is computed (see Figure 9), the estimated values for all random numbers and the portfolio estimated value. Is restored.

[Relation 1]

O (i, a) = I2 (a, I1 (i, I3 (a)))

As such, when the VaR calculator 150 restores the expected values and portfolio expected values for each random number for all the accounts of the trading counterpart (see FIG. 8), the VaR calculator 150 returns the expected values for each random number. 9, the expected values and portfolio estimates of each random number restored as shown in FIG. 9 can be sorted in descending order or ascending order of the portfolio expected values as shown in FIG. 11. We can output X (%) Portfolio Credit VaR by subtracting Y = N (100-X) / 100th from the average of N portfolio expected values for the input confidence level of X (%). For example, in FIG. 11, the 99% portfolio Credit VaR corresponds to KRW 27,149 corresponding to the 165,882-Y (= 10) th value (138,773).

In addition, the VaR calculation unit 150, as shown in FIG. 12, based on the ratio (average loss amount ratio) (percentage) of the estimated values of all the accounts to the portfolio estimated value of the Yth (10) above. Contributions can be calculated as Credit VaR Contribution by account.

According to other inquiry request conditions, the VaR calculation unit 150, based on the value distribution data restored as shown in FIG. 9, as shown in FIG. 12, the total average value of each account, the section average value, the average loss amount, and the like. Various data can be calculated and provided.

In the Credit VaR calculation system 100 and its data processing method according to an embodiment of the present invention, the functions used may be embodied as computer readable codes on a computer readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the computer-readable recording medium include a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, a hard disk, a removable storage device, And the like. In addition, it is understood that the computer readable recording medium can be distributed over network coupled computer systems and stored and executed in the form of computer readable code in a distributed fashion.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the equivalents of the claims, as well as the claims.

Control unit 110
Database (111)
Parameter Estimator 120
Random number generator 130
Asset Value Calculator (140)
VaR calculator 150

Claims (4)

In the Credit VaR data processing method for processing data associated with the calculation of the Credit VaR in the Credit VaR calculation system,
N (Natural Number) Scenarios Generate N * M asset return random numbers for M (natural number) counterparts per turn, and generate a pre-generated rating for each transition probability after a certain period of multiple grades for each random number generated Deriving an expected credit rating for each counterparty based on a scenario table based on a reference table for a transition threshold, and storing information about the derived expected credit rating in a database;
Generating a reference expected value table for the plurality of grades for each account held by the corresponding trading partner and storing the table in the database; And
In accordance with the inquiry request, for all accounts of the counterparty, the expected values for each random number are calculated based on the information on the expected credit rating stored in the database and the reference expected value table, and for each scenario. Estimating a portfolio expected value for each scenario by adding the expected values of all accounts of the counterparty, and calculating and providing a portfolio credit var or a credit var contribution for each account.
Credit VaR data processing method comprising a. The method of claim 1,
The expected values of each random number and the portfolio expected values for all of the restored accounts are sorted in the order of the size of the portfolio expected value, and the input confidence level X% for the inquiry request of the sorted expected values is A method of processing Credit VaR data, comprising: outputting Y% of the portfolio's expected values by subtracting Y = N (100-X) / 100th as X% Portfolio Credit VaR. 3. The method of claim 2,
Credit VaR data processing method characterized in that for calculating the contribution of each account as the Credit VaR Contribution based on the proportion of the expected value of all the accounts to the Y-th portfolio expected value. In the Credit VaR calculation system for processing data related to the calculation of Credit VaR,
A parameter estimator configured to generate a reference table for a grade transition threshold for each transition probability after a plurality of grades;
A random number generator for generating N * M asset return random numbers for M counterparts for each N (natural number) scenarios;
For each random number generated by the random number generator, information about the estimated credit rating derived by deriving an expected credit rating for each of the trading partners is stored in a database, and the criteria for the plurality of ratings for each account held by the corresponding trading partner. An asset value calculator configured to generate an expected value table and store it in the database; And
In accordance with the inquiry request, for all accounts of the counterparty, the expected values for each random number are calculated based on the information on the expected credit rating stored in the database and the reference expected value table, and for each scenario. VaR calculation unit for estimating the portfolio expected value of each scenario by adding the expected values of all the accounts of the trading partner, calculates the portfolio Credit VaR or Credit VaR Contribution by account
Credit VaR calculation system comprising a.

Images