Executive is company "INO_TEL" Ltd. Executive provides service on estimation and analysis of credit risk of borrower.

Customer is juridical person, Bank or Credit Bureau (Loan Office), which is provided by service "estimation and analysis of credit risk of borrower".

Service is construction of logical and probabilstic model (LP-model) of credit risk under statistics of bank and estimation of credit risk of borrower. The Software of Executive for training of LP-model of credit risk with groups of incompatible events, estimation and analysis of risk of concrete borrower, is used.

Order 1. Bank (credit Bureau) gives statistics of issued earlier credits according to Form 1. File with statistics in impersonal form is created by bank, then zipped and sent by e-mail.

Filenames must have the next form:
       name of zipped file: STNNN_OUT.zip, where:
           ST - sign of file with statistics,
           NNN - conditional number of bank (3 last number of BIK),
           OUT - outcoming file.
       name of internal file:  NNNN_XXXXXX.txt, where:
           NNNN - registration number of bank,
           XXXXXX - sequence number of message.  

Updating of statistics and construction of new LP-model of credit risk are performed periodically every 1-4 quarter according to agreement with Customer.

N
n
N1 ... Nj ... Nn
Y1      Z11      Z21     ...     Zn1
Y2      Z12      Z22     ...   Zn2
.......................
YN      Z1 N1     Z2 N2     ...    Zn Nn

where N - the number of credits;
       n - the number of signs;
       N1 ... Nj ... Nn - the number of grades in every sign;
       Y1, Y2, ..., YN - credit's success (1 - "good"; 0 - "bad") ;
       Z1 1, Z1 2 ..., Zn Nn - the values grades of sign.
Values of grades of signs are placed in squares of table.

Example:
1000
20
4  10   5  11  10   5   5   4   4   3   4   4   5   3   3   4   4   2   2   2

1   1    3   5    3    3   1   2   4   2   1   4   2   1   3   1   1   3   1   1   1
1   1    2   5    1    5   1   3   2   3   1   2   1   2   3   1   2   3   2   1   1
1   2    2   3   10   2   2   4   2   2   1   4   1   1   3   1   1   2   1   1   1
1   1    2   5    1    4   1   3   3   3   1   2   1   2   3   1   2   2   2   1   2
..........................
..........................
..........................
1   1    1   5    1    3   1   2   2   3   1   4   1   3   3   2   1   2   2   1   2
1   4    3   5    4    3   1   1   4   2   1   4   3   5   3   2   2   1   1   1   1
0   2    4   3    4    5   3   1   1   2   1   4   4   1   3   1   1   1   1   1   1

Customer's data is used for training of LP-model of risk. Numbers of good and bad credits are calculated on file automatically. Also, the calculation of number of similarly described credits is performed and it's identificated what grades of signs are not used for description of credits. It allows to control data of Order 1 and results of training of LP-model of risk.

For credit risk of juridical persons categories of clients are presented as grades 1, 2, 3, ... of sign "client's category". This parameter is placed in last column of file and additional risk attributes are calculated for it.

Order 2. Bank gives order for estimation and analysis of credit risk for one or several borrowers in impersonal from according to Form 2. File with order is zipped and sent by e-mail. Filenames must have the following form:

    name of zipped file: RENNN_OUT.zip, where:
      RE - sign of file for risk estimation (Risk Estimation),
      NNN - conditional number of bank (3 last number of BIK),
      OUT - outcoming.
    name of internal file:  NNNN_XXXXXX.txt,  where:
      NNNN - registration number of bank,
      XXXXXX - sequence number of message.

Form 2. Impersonal file NNNN_XXXXXX.txt of order for estimation and analysis of risk of credits

{yy.mm.dd hh:mm:ss}
UserID  Z1, Z2, ..., Zn
where:
     yy.mm.dd - year, month and day of order,
     hh.mm.ss - hour, minutes and seconds when order was created,
     UserID - user's identificator (bank creates),
     Z1, Z2, ..., Zn - numbers of grades in every sign.
Example:
{2006.08.17 01:03:24}
10345678  1  3  5  3  3  1  2  4  2  1  4  2  1  3  1  1  3  1  1  1
10234673  1  2  5  1  5  1  3  2  3  1  2  1  2  3  1  2  3  2  1  1
10543572  2  6  5  4  4  1  5  4  3  1  4  1  2  3  2  1  3  1  1  1
10862346  2  6  3  6  4  1  2  4  3  1  1  4  2  3  1  1  2  1  1  1
..........................
10642686  1  3  3  1  3  1  2  4  2  1  3  3  1  3  1  1  3  1  2  1
10135467  4  3  5  7  4  2  3  4  2  1  2  1  2  3  2  2  3  1  1  1

Results for order 1. Construction of LP-model of credit risk of bank takes a time up to 12 hours.

Results are sent to Customer by E_mail. Filenames have the following form:

     name of zipped file: STNNN_IN.zip, where:
         ST - sign of file with results of training of risk model by statistics,
         NNN - conditional number of bank (3 last numbers of BIK),
         IN - incoming.
     name of internal file:  NNNN_XXXXXX.txt,  where:
         NNNN - registration number of bank,
         XXXXXX - sequence number of message.

Customer is informed when LP-model of credit risk was being constructed and basic attributes and quality indicators (accuracy, robustness, recognition asymmetry) of model are passed to Customer in according to Form 3.

Let us make some explanations (fig. 1): 

Ng20 - the number of very good credits with least risk, containing 20% of common calculated number Ngc of good credits;

Ngc - Ng20   - good credits;

Nb20   - the number of very bad credits with largest risk, containing 20% of common calculated number Nbc of bad credits;

Nbc - Nb20   - bad credits.

For juridical persons for "categories of clients" (last column in Form 1) following attributes are calculated additionally: the frequencies of categories in all credits, in "bad" and in "good" credits, average value of risk of credits for categories. It allows to estimate the adequacy of dividing of clients into categories.

Form 3. File NNNN_XXXXXX.txt with parameters of LP-model of credit risk of bank

Identificator	Parameter's name
N	Number of credits in bank's statistics
Ng	Number of "good" credits in statistics
Nb	Number of "bad" credits in statistics
Pm	Average risk of bank's credits
Pad	Admissible credit risk
Pmin	Minimal risk of credit in statistical data
Pmax	Maximal risk of credit in statistical data
Ng20	Number of very "good" credits
Ngc - Ng20	Number of "good" credits
Nb20	Number of very "bad" credits
Nbc - Nb20	Number of "bad" credits
Em	Average mistake in recognition of credits
Eb	Mistake in recognition of "bad" credits
Egb	Asymmetry ratio in recognition of credits
Fmax	Number of correctly recognized credits in bank's statistics
Feff	Reduction of number of wrongly recognized credits in LP-technique in comparison with technique, used by bank
F1	Reduction of number of correctly recognized credits under exception of sign 1
F2	Reduction of number of correctly recognized credits under exception of sign 2
...	...
Fn	Reduction of number of correctly recognized credits under exception of sign n

Fig.1. Diagram of classification of credits

Results of order 2. Results of estimation and analysis of credit risk of one or several credits are sent to Customer by E_mail as a file in Form 4 during 36 hours. Filenames have the following form:

   name of zipped file: RENNN_IN.zip, where:
       RE - sign of file with results of estimation of risks of credits (Risk Estimation),
       NNN - conditional number of bank (3 last numbers of BIK),
       IN - incoming.
   name of internal file:  NNNN_XXXXXX.txt,  where:
       NNNN - registration number of bank,
       XXXXXX - sequence number of message.

Information of Forms 3 and 4 allow to construct formula for price (percent) for credit risk. For example, the simplest formula is following:

C_i = C_ad + k (P_risk - P_ad ),

Example:

10345678   1   0.199218
 ....
10234673   0   0.203452

Up