TWI805880B - Internal system of bank for creidt risk evluation and methohd thereof - Google Patents

Abstract

An internal system of bank for credit risk evaluation includes an information managing device and an evaluation device. The information managing device which includes a database of internal credit automatically accesses an internal credit information of a business requester stored in the database of internal credit and calculates a number of the internal credit information to determine whether to hand in a request for providing at least one external rating toward at least one external rating organization according to the number of the information and a threshold number of the information. When receiving the approval of the request, the information managing device accesses at least one external rating. The evaluation device automatically calculates the internal credit information based on at least one default formula or automatically transforms at least one external rating based on a rating-transformation table in order to reach out at least one probability of default before weighted.

Description

Internal credit risk assessment system and method of the bank

The invention relates to a credit risk assessment system, in particular to a credit risk assessment system internally used by a bank to assess its customers.

Those with financial business needs (such as individuals) apply to financial institutions (such as banks) for related financial services. If the financial services required by individuals are more related to financing or borrowing, the bank must first carry out the above-mentioned individuals in accordance with internal regulations and procedures. Evaluation and measurement of financial credit. If the individual's credit history is poor, or the assessment results show that the probability of default is high, that is, the credit risk is high, the bank may need to further adjust and increase the debt interest rate for repayment, and may even be unable to provide the required financial services business.

However, at present, the calculation and evaluation of the credit risk of the above-mentioned financial business needs is still done manually. Therefore, it will take a long time for the evaluation procedure and inefficiency, which will indirectly prolong the review time of the evaluation case. In addition, there is still the problem that the financial institutions themselves have insufficient credit evaluation information, making it impossible to conduct evaluations. Based on the above, how to reduce the working time of the evaluation process, improve work efficiency, and provide the necessary credit risk evaluation even in the case of insufficient information of the financial institution itself has become a problem to be solved in this field.

In view of the above-mentioned problems to be solved and their reasons, specifically, the present invention provides a credit risk assessment system within a bank. When a business demander applies to a bank for business, the above-mentioned system is used to evaluate a breach of contract of the above-mentioned business demander. Probability, the above system includes: a data management device and an evaluation device.

The above-mentioned data management device includes an internal credit database, and the above-mentioned data management device automatically reads one of the internal credit data of the above-mentioned business needs in the above-mentioned internal credit database, and calculates the number of one data of the above-mentioned internal credit data, based on The above-mentioned number of data and one data threshold number automatically determine whether to request at least one external rating unit to provide at least one external rating of the above-mentioned business needs. When the request is approved, the above-mentioned data management device receives the above-mentioned at least one external rating .

The aforementioned evaluation device automatically calculates the aforementioned internal credit data according to at least one default formula, or automatically converts the aforementioned at least one external rating according to a rating comparison table, so as to obtain at least one unweighted probability of default.

According to an embodiment, wherein the evaluation device further calculates a discrete value of the at least one unweighted probability of default according to a discrete formula, compares the discrete value with a discrete threshold, and corrects the initial weight according to an initial weight or automatic judgment as A modified weight, which is calculated according to the above-mentioned initial weight or the above-mentioned modified weight, and obtains a default probability and an internal rating corresponding to the above-mentioned default probability.

According to another embodiment, a credit evaluation data is further modified through the above data management device, the above credit evaluation data includes the above data threshold number, the above default formula, the above Commentary comparison table, the above-mentioned discrete formula, the above-mentioned discrete threshold value, the above-mentioned initial weight, and a threshold ratio of missing value.

According to yet another embodiment, the assessment device further calculates an undervalue ratio of the at least one unweighted probability of default, compares the undervalue ratio with an undervalue threshold ratio, and automatically determines an initial weight or modifies the above initial weight to be a The modified weight is calculated according to the above-mentioned initial weight or the above-mentioned modified weight, and a default probability and an internal rating corresponding to the above-mentioned default probability are obtained.

According to yet another embodiment, wherein the evaluation device automatically outputs the at least one unweighted probability of default based on the discrete value and the discrete threshold, and automatically converts the at least one unweighted probability of default according to the rating comparison table. internal rating.

In addition to providing the credit risk assessment system inside the bank, the present invention further provides a credit risk assessment method inside the bank. When a business demander applies to a bank for business, the above-mentioned system is used to evaluate a business demander. Probability of default, the above-mentioned method includes: submitting a credit evaluation request to the above-mentioned system; automatically reading and calculating the data number of one of the internal credit data of the above-mentioned business needs; when the number of the above-mentioned data is not less than a data threshold number, Automatically calculate the above-mentioned internal credit data according to at least one default formula to obtain at least one unweighted probability of default; when the number of the above-mentioned data is less than the above-mentioned threshold number of data, automatically request at least one external rating unit to provide at least one of the above-mentioned business needs an external rating, when the request is granted, receiving said at least one external rating and automatically converting said at least one external rating according to a ratings comparison table to obtain said at least one unweighted probability of default; automatically calculated according to a discrete formula A discrete value of the above at least one unweighted probability of default; when the above discrete value is less than a discrete threshold value, the above at least one unweighted probability of default is automatically calculated according to an initial weight, and the above default contract probability; when the above-mentioned discrete value is not less than the above-mentioned discrete threshold value, the above-mentioned initial weight is automatically corrected to a revised weight, and the above-mentioned at least one unweighted default probability is calculated according to the above-mentioned revised weight, and the above-mentioned default probability is obtained; and according to the above-mentioned default probability The above-mentioned rating comparison table automatically converts the above-mentioned default probability to an internal rating, and outputs the above-mentioned default probability and the above-mentioned internal rating.

According to yet another embodiment, after the request for credit evaluation is made, a credit evaluation data is modified, and the credit evaluation data includes the number of data thresholds, the above-mentioned default formula, the above-mentioned rating comparison table, the above-mentioned discrete formula, and the above-mentioned discrete threshold value, the above initial weights, and an undervalue threshold ratio.

According to yet another embodiment, wherein one of the at least one unweighted probability of default default ratio is further calculated, and when the above-mentioned short-value ratio is not less than a short-value threshold ratio, the above-mentioned initial weight is automatically adjusted to the above-mentioned revised weight, and according to the above-mentioned revised Calculate the at least one unweighted default probability by weight, and calculate the at least one unweighted default probability according to the above initial weight when the above-mentioned undervalue ratio is less than the above-mentioned undervalue threshold ratio.

According to yet another embodiment, when the above-mentioned discrete value is not less than the above-mentioned discrete threshold value, the above-mentioned at least one unweighted default probability can be selected or automatically output, and the above-mentioned at least one unweighted default probability can be automatically converted according to the above-mentioned rating comparison table At least one of the above internal ratings.

100: Bank

110: Credit Evaluation Request

120:Credit evaluation information

121: Number of data threshold transactions

122: Default formula

123: Rating Comparison Table

124: Discrete formula

125: discrete threshold

126: initial weight

127:Deficiency threshold ratio

200: Bank internal credit risk assessment system

210: data management device

211: Internal credit database

212: External Rating Request

220: Evaluation device

221:Internal Rating

222: Default probability

300: External Rating Unit

310: External Rating

400: The process of credit risk assessment method within the bank

410-450: Steps

421, 431-434, 441-442: Steps

In order to make the above and other objects, features, advantages and embodiments of the present invention more comprehensible, the accompanying drawings are described as follows:

FIG. 1 is a structural diagram of the bank's internal credit risk assessment system of the present invention.

Fig. 2 is a content structure diagram of credit assessment data inside a bank according to the present invention.

FIG. 3 is a flow chart of the bank internal credit risk assessment method of the present invention.

In view of the above-mentioned problems to be overcome, the present invention provides a credit risk assessment system for internal use by banks. The above-mentioned system can automatically determine whether the bank itself has sufficient internal credit risk data for the assessment of individuals with business needs . Even further, when the bank itself does not have sufficient internal credit risk data, it can request at least one external rating unit to provide the required external credit risk data for assessing the credit risk of individuals. After the above-mentioned system calculates the above-mentioned internal credit data, or converts the above-mentioned external credit data, it outputs the internal rating results within the bank to continue the subsequent business undertaking content.

For a more clear description of the implementation of the present invention, please refer to FIG. 1 , which is a structural diagram of the bank's internal credit risk assessment system of the present invention. The present invention provides a credit risk assessment system 200 within a bank. When a business needer applies to a bank 100 for business, the system 200 is used to evaluate a default probability 222 of the business needer. The system 200 includes: a data management device 210 and an evaluation device 220 .

The above-mentioned data management device 210 is further described as follows. When the above-mentioned bank 100 receives the business needs of the above-mentioned business needs, the above-mentioned bank 100 can automatically read the above-mentioned business needs in the above-mentioned internal credit database 211 through the above-mentioned data management device 210 (including an internal credit database 211) An internal credit information of the above-mentioned internal credit information, and calculate the number of a data of the above-mentioned internal credit information. Among them, the above-mentioned internal credit information includes the credit information of various risk exposure positions, and the above-mentioned exposure positions can be loan information, guarantee ratio, collateral appraisal rate, exposure of group companies, proportion of liabilities, enterprise default loss data, mortgage default exposure data, personal Credit index type/ladder type/other types of mortgage information, personal guarantees/no guarantees, equivalent credit lines and other information.

The data management device 210 automatically compares the number of data obtained through the above calculation with a threshold number of data 121 , and automatically determines whether to submit at least one external rating request 212 to at least one external rating unit 300 .

Wherein, the above-mentioned data management device 210 may perform a numerical comparison between the number of data items obtained by the above calculation and the above-mentioned threshold number of data items 121 . That is, the above-mentioned data management device 210 retrieves the internal credit data in the above-mentioned internal credit database 211 that can be used for evaluating the required exposure position. For example, if you have not conducted relevant financial business at the above-mentioned bank and need to request credit risk information of relevant financial business from other banks, or if the bank only holds, for example, 5 relevant internal credit data, which is less than the above-mentioned data threshold of 121 , for example, if 7 transactions are required, it will be automatically judged and determined as "insufficient internal credit information". Conversely, those who have conducted relevant financial business in the above-mentioned banks and hold, for example, more than or equal to 7 related internal credit data (that is, the above-mentioned data threshold number of 121) will be automatically judged and determined as "sufficient internal credit data". The data management device 210 can automatically judge and decide whether to submit the at least one external rating request 212 to the at least one external rating unit 300 according to whether the internal credit data is "sufficient" or "insufficient".

Wherein, the above-mentioned at least one external rating unit 300 includes credit assessments in the capital market (such as China Credit Ratings [Taiwan Ratings], Standard & Poor's [Standard & Poor's, referred to as S&P], Moody's Corporation, referred to as Moody's] , and Fitch International [Fitch Group]), credit assessment in the commercial market (such as The Dun & Bradstreet Corporation [The Dun & Bradstreet Corporation]), consumer credit assessment (such as Equifax [Equifax], Link Information [TransUnion], and Experian [Experian]), China National Personal credit evaluation (such as Lianhe Credit, China Chengxin International, and Dagong International), Chinese personal credit evaluation (such as Zhima Credit [Ant Financial], and Tencent WeChat [Tencent]), credit evaluation of other banks, or other credit evaluations that can provide and other non-bank internal units, etc.

Further, if the above-mentioned data management device 210 automatically determines that the above-mentioned external rating request 212 needs to be submitted, then continue to select and determine the above-mentioned at least one external rating unit 300, and then generate the above-mentioned at least one external rating request 212, and submit the above-mentioned The external rating request 212 is to the selected at least one external rating unit 300 . Among them, the above-mentioned at least one external rating request 212 can clearly include the request serial number, date and time of request generation, query unit code, data provider code, query reason, query object’s unified compilation or ID number, query items, and Query the message content such as the number of transactions.

When the external rating request 212 is approved, the data management device 210 receives the at least one external rating 310 from the at least one external rating unit 300 . Among them, the above-mentioned at least one external rating 310 can be returned in the form of a rating result. For example, the ratings of S&P, Moody's and China Credit Ratings in Table 1 below, Table 1 shows different external rating units and Rating indicates the corresponding rating definition comparison table. In Table 1, under the same rating definition, different sources may provide the at least one external rating 310, that is, the at least one external rating unit 300 may provide a variety of ratings according to different ratings. and the at least one external rating 310 indicated by the representation method. For example, in the case where the rating is also defined as the first column "very good reputation, almost no risk", the S&P rating is expressed as "AAA", the Moody's rating is expressed as "Aaa", and the China Credit Rating is Expressed as "twAAA", that is, three same rating definitions are generated, but the above-mentioned at least one external rating 310 is returned in different rating representation modes.

The evaluation device 220 described above is further described as follows. When the above-mentioned data management device 210 automatically judges and determines that "internal credit data is sufficient", the above-mentioned evaluation device 220 automatically calculates the above-mentioned internal credit data whose number of data is greater than or equal to the data threshold 121 according to at least one default formula 122 material. Among them, at least one default formula 122 includes any financial calculation method that can calculate and convert credit risk ratings, such as logistic regression (Logistic Regression), linear regression (Linear Regression), discriminant analysis (Discrimination Analysis), and artificial neural network analysis (Artificial Neural Network) and other calculation formulas.

According to an embodiment, the evaluation device 220 further calculates a discrete value of the at least one unweighted probability of default according to a discrete formula 124, so as to compare the discrete value with a discrete threshold 125, and automatically determine that the at least one unweighted probability of default The degree of dispersion of weighted default probabilities. Wherein, the above-mentioned discrete formula 124 can be any statistical formula that can calculate scattered data or multiple data materials without intermediate values, and present the degree of dispersion of the above-mentioned multiple data data with a scattered reference value (i.e. the above-mentioned discrete value) .

According to another embodiment, when the data management device 210 automatically determines that "the internal credit data is sufficient", the evaluation device 220 uses the discrete formula 124 to calculate the discrete value of the internal credit data. Among them, the above-mentioned discrete value and above-mentioned discrete formula 124 are further illustrated by the following Table 2 and the following formula 1-2. Table 2 is a discrete table of unweighted default probability of internal credit data.

Please refer to Table 2 below. If the above-mentioned unweighted default probabilities of internal credit data are calculated using the four above-mentioned default formulas 122 (N=4), after calculating the above-mentioned unweighted default probabilities of each of the above-mentioned default formulas 122, the above formula 1 Calculate the average of the above unweighted probabilities of default (11.5%). Then, according to the above-mentioned average value (11.5%) and the above-mentioned discrete formula 124 (such as the above-mentioned formula 2), the above-mentioned discrete value (2.7%) of the above-mentioned unweighted probability of default can be calculated and obtained. If the above-mentioned discrete threshold value 125 is set to 5.0%, compare the above-mentioned discrete threshold value 125 (5.0%) with the above-mentioned calculated discrete value (2.7%), it can be known that the above-mentioned discrete value (2.7%) is less than the above-mentioned discrete threshold value 125 (5.0%), so the above-mentioned unweighted probability of default is judged to be "discrete and not obvious".

And when the above-mentioned data management device 210 automatically judges and determines that "internal credit information is insufficient", the above-mentioned evaluation device 220 automatically converts the above-mentioned at least one external rating from the above-mentioned at least one external rating unit 300 according to a rating comparison table 123 310 to obtain at least one unweighted probability of default. Wherein, the above-mentioned rating comparison table 123 can be any rating comparison conversion method that can meet the external rating defined internally by the bank 100 . Here, the following Table 3 is used as an example of the above-mentioned rating comparison table 123. Table 3 is a comparison table for conversion of external ratings to corresponding internal ratings and unweighted probability of default. For the same exposure position, S&P may return "A", please refer to Table 3 below, which corresponds to the internal unweighted default probability of "1.05-1.36%"; and Moody's returns "Baa1", which corresponds to the internal The unweighted probability of default is "1.55-2.23%". That is to say, the lower the grade of the above-mentioned at least one external rating 310 returned, the higher the probability of possible default, and thus the higher the unweighted probability of default converted into internal.

According to another embodiment, when the above-mentioned data management device 210 automatically judges and determines that "internal credit information is insufficient", the above-mentioned evaluation device 220 uses the above-mentioned rating comparison table 123 to convert the above-mentioned at least one external rating 310, and the above-mentioned at least one external rating The rating 310 is converted to an internal corresponding unweighted probability of default, and the above-mentioned discrete value of the above-mentioned unweighted probability of default is calculated. Among them, the above-mentioned discrete value and the above-mentioned discrete formula 124 are further illustrated by the following Table 4 and the above-mentioned Formula 1-2. Table 4 is a discrete table of the unweighted probability of default of the external rating.

Please refer to Table 4 below. If the above-mentioned unweighted default probabilities are obtained by converting the four sources of the above-mentioned external ratings 310 (N=4), after converting the above-mentioned unweighted default probabilities of each of the above-mentioned external ratings 310 , the average (26.0%) of the above-mentioned unweighted probability of default needs to be obtained from the above formula 1. Then, according to the above-mentioned average value (26.0%) and the above-mentioned discrete formula 124 (such as the above-mentioned formula 2), the above-mentioned discrete value (34.6%) of the above-mentioned unweighted probability of default can be calculated and obtained. If the above discrete threshold value 125 is set to 0.05, compare the above discrete threshold value 125 (5.0%) with the above calculated discrete value (34.6%), you can know that the above discrete value (34.6%) is greater than the above discrete threshold value 125 (5.0%), so it is judged that the above-mentioned unweighted probability of default is "significantly discrete".

Regardless of whether the above-mentioned internal credit data is calculated through the above-mentioned evaluation device 220 or the above-mentioned at least one external rating 310 is converted, the internally corresponding unweighted default probability and the above-mentioned discrete value can be obtained, and can be further obtained by combining the above-mentioned discrete value with the above-mentioned discrete threshold The value 125 is compared to judge the degree of dispersion of the above-mentioned unweighted probability of default. According to the degree of dispersion of the above-mentioned unweighted default probability, the evaluation device 220 can further automatically determine whether to maintain an initial weight 126 or automatically judge and modify the above-mentioned initial weight 126 to be a modified weight. The evaluation device 220 calculates according to the initial weight 126 or the modified weight to obtain a default probability 222 and the default probability 222 corresponds to an internal rating 221 of the rating comparison table 123 .

Furthermore, according to yet another embodiment, the following table 5 is used for illustration, and table 5 is a calculation table of default probability with insignificant dispersion. Please refer to Table 5 below. For example, the above-mentioned initial weight 126 of each of the above-mentioned unweighted default probabilities is initially set to 1.0. When the evaluation device 220 judges that the unweighted default probabilities are "discrete and not obvious", the initial weight 126 of each of the unweighted default probabilities is still maintained at 1.0. The evaluation device 220 multiplies the above-mentioned unweighted default probability by the corresponding initial weights 126 according to the following formula 3, and obtains the above-mentioned default probability 222 (11.5%). The evaluation device 220 can convert the default probability 222 (11.5%) into the internal rating 221 (grade 13) according to the rating comparison table 123 (such as Table 3 above).

Furthermore, according to yet another embodiment, the following table 6 is used for illustration, and table 6 is a table for calculating the probability of default with obvious discreteness. Please refer to Table 6 below. For example, the above-mentioned initial weight 126 of each of the above-mentioned unweighted default probabilities is initially set to 1.0. When the evaluation device 220 judges that the unweighted default probabilities are "discrete and obvious", the initial weights 126 of each of the unweighted default probabilities are automatically adjusted to the modified weights (2.6, 1.2, 0.9 and 0.2). The evaluation device 220 multiplies the above-mentioned unweighted probability of default by the corresponding modified weights according to the following formula 4, and obtains the above-mentioned probability of default 222 (10.8%). The evaluation device 220 can convert the default probability 222 (10.8%) into the internal rating 221 (level 12) according to the rating comparison table 123 (such as Table 3 above).

Wherein, according to yet another embodiment, the above-mentioned initial weight 126 is adjusted to a calculation method of the above-mentioned modified weight, which is briefly described as follows. As mentioned above, there are four sources of external ratings 310 (N=4), according to the above table 4 and the following formula 5 (wherein, the above-mentioned unweighted probability of default R _i must first be arranged in a numerical sequence [Number sequence] , and not limited to the order from large to small or from small to large, N is the number of the above-mentioned default formula 122 or the number of sources of the above-mentioned external rating 310, the number of the above-mentioned unweighted probability of default R _i , the N in the example here is the above-mentioned The number of sources of external rating 310, N=4), and the median (Median, referred to as MED, MED=13.0) of the above-mentioned unweighted probability of default series of the above-mentioned external rating 310 is calculated. Then according to the above table 4 and the following formula 6, the above-mentioned correction weights (0.18, 2.60, 1.18 and 0.87) corresponding to the above-mentioned initial weights 126 can be calculated respectively, as shown in the following table 6.

According to another embodiment, please refer to FIG. 2 . FIG. 2 is a content structure diagram of credit evaluation data inside a bank according to the present invention. In the above-mentioned system 200, a credit evaluation data 120 is built in itself. The above-mentioned credit evaluation data 120 includes the above-mentioned data threshold number 121, the above-mentioned default formula 122, the above-mentioned rating comparison table 123, the above-mentioned discrete formula 124, the above-mentioned discrete threshold 125, the above-mentioned initial weight 126, and a short-value threshold ratio 127. If the above-mentioned bank 100 recognizes the need to change or modify any of the above-mentioned credit evaluation data 120 , the above-mentioned credit evaluation data 120 can be further changed or modified through the above-mentioned data management device 210 . For example, the above-mentioned bank 100 can adjust and set the value of the above-mentioned initial weight 126 through the above-mentioned data management device 210 , and the above-mentioned evaluation device 220 calculates the above-mentioned initial weight 126 when calculating the above-mentioned default probability 222 .

According to yet another embodiment, the evaluation device 220 further calculates an undervalue ratio of the at least one unweighted probability of default, so as to compare the above-mentioned undervalue ratio with an undervalue threshold ratio 127, and automatically determine the above-mentioned at least one unweighted probability of default Undervaluation of the probability of default. Among them, on The missing value ratio mentioned above can be any reference value that can be calculated and can present the missing or missing value of multiple data data.

According to yet another embodiment, the evaluation device 220 calculates the above-mentioned shortfall ratio of the at least one unweighted default probability, which is further illustrated by the following Table 7 and the following formula 7. Table 7 is a table of the shortfall ratio of the unweighted default probability.

Please refer to Table 7 below. The evaluation device 220 respectively calculates the missing value ratio of each data source of the above-mentioned unweighted probability of default. The above-mentioned missing ratio (0.5) is calculated from the "actual number of missing data" and "theoretical data" of the data source of China Credit Rating. If the above-mentioned undervalue threshold ratio 127 is set to 0.25, compare the above-mentioned undervalue threshold ratio 127 (0.25) with the above-mentioned calculated undervalue ratio (0.5), it can be known that the above-mentioned undervalue ratio (0.5) is greater than the above-mentioned undervalue threshold The proportion is 127 (0.25), so it is automatically judged that the above data source is "the proportion of missing value is too high".

According to yet another embodiment, the evaluation device 220 can further adjust the initial weight 126 to the modified weight according to the initial weight 126 or automatically through the situation of the missing value ratio and the comparison result. The evaluation means 220 is calculated according to the initial weight 126 or the modified weight. Wherein, according to yet another embodiment, the above-mentioned initial weight 126 is automatically adjusted to the above-mentioned calculation method of the correction weight, which is briefly described as follows. As mentioned above, there are the same four sources of external rating 310 (N=4), according to the above table 7 and the following formula 8 (wherein, the above-mentioned missing value ratios must first be arranged in a numerical sequence in order of value, and are not limited to the largest From small to large, N is the number of sources of the above-mentioned default formula 122 or the number of sources of the above-mentioned external rating 310, the number of the above-mentioned unweighted probability of default R _i , and N in this example is the number of sources of the above-mentioned external rating 310 , N=4), calculate the median (MED=0.2) of the above-mentioned missing value ratio series of the above-mentioned external rating 310. Then, according to the above table 7 and the following formula 9, the above-mentioned correction weights (0.63, 1.25, 1.00 and 1.00) corresponding to the above-mentioned initial weights 126 can be calculated respectively, as shown in the above table 7. The evaluation device 220 adjusts the correction weight in the above-mentioned manner, and supplements the above formula 4 to obtain the above-mentioned default probability 222 (20.1%) and the above-mentioned internal rating 221 (level 14) corresponding to the above-mentioned default probability 222 (20.1%) .

According to yet another embodiment, the above-mentioned evaluation device 220 can automatically output the above-mentioned at least one unweighted default probability according to the above-mentioned discrete value and the above-mentioned discrete threshold value 125, and automatically convert the above-mentioned at least one unweighted default probability according to the above-mentioned rating comparison table 123 The aforementioned at least one internal rating 221 of probability. Take Table 4 above as an example. The above-mentioned discrete value calculated is 34.6%, and the above-mentioned discrete threshold value 125 is set at 5.0%. Therefore, it is judged that the above-mentioned unweighted probability of default is "discretely obvious", that is, directly output at least one of the above-mentioned values in Table 8 below. Unweighted probability of default, and at least one internal rating 221 corresponding to it.

In addition, please refer to FIG. 3 , which is a flow chart of the bank internal credit risk assessment method of the present invention. Corresponding to the credit risk assessment system 200 inside the bank mentioned above, the present invention discloses another process 400 of the credit risk assessment method inside the bank. When a business When a demander applies to a bank 100 for business, use the above-mentioned system 200 to evaluate a default probability 222 of the above-mentioned business demander, which includes the following steps:

First of all, referring to FIG. 3 , as in step 410 , the bank 100 can submit a credit evaluation request 110 to the system 200 when receiving a service request from a person in need of the above-mentioned business. Automatically read one of the internal credit data of the above-mentioned business needs, so as to calculate and obtain a data number of the above-mentioned internal data according to the above-mentioned internal credit data.

Next, as in step 430 , after the above-mentioned number of data items is obtained, the relationship between the above-mentioned number of data items and a threshold number of data items 121 is further compared. When the above-mentioned number of data items is not less than (that is, greater than or equal to) the above-mentioned number of data threshold items, as in step 431, the above-mentioned internal credit data is automatically calculated according to at least one default formula 122, and at least one unweighted default probability is obtained. Wherein, the above-mentioned at least one default formula 122 includes any financial calculation method that can calculate and convert credit risk ratings, such as Logistic Regression, Linear Regression, Discrimination Analysis, and Artificial Neural Network and other calculation formulas.

When the above-mentioned number of data items is less than the above-mentioned number of data threshold items, as in step 432, at least one external rating request 212 is automatically submitted to at least one external rating unit 300 to request at least one external rating 310 of the above-mentioned business needs . When the at least one external rating unit 300 further approves the at least one external rating request 212 , in step 433 , the at least one external rating 310 is received. In step 434 , the at least one external rating 310 is automatically converted according to a rating comparison table 123 to obtain the at least one unweighted default probability within the bank 100 corresponding to the at least one external rating 310 . Among them, the above-mentioned rating comparison table 123 can be any In line with the rating comparison conversion method of the external rating defined internally by the bank 100 above, for example, Table 3 above, the conversion example has been shown above, and will not be repeated here.

Then, regardless of the at least one unweighted probability of default calculated through the above at least one default formula 122 or the above-mentioned at least one unweighted default probability obtained through the conversion of the above-mentioned rating comparison table 123, the above-mentioned at least one unweighted default probability can be automatically calculated according to a discrete formula 124. A discrete value for an unweighted probability of default. Wherein, the above-mentioned discrete formula 124 can be any statistical formula that can calculate scattered data or multiple data materials without intermediate values, and present the degree of dispersion of the above-mentioned multiple data data with a scattered reference value (i.e. the above-mentioned discrete value) . In addition, the calculation example of the above-mentioned discrete formula 124 is as shown above, and will not be repeated here.

Next, as in step 440 , the calculated discrete value is compared with a discrete threshold value 125 . Wherein, an example of the way of comparing the above discrete value with a discrete threshold value 125 is as shown above, and will not be repeated here. When the discrete value is smaller than the discrete threshold value 125, in step 442, the at least one unweighted default probability is automatically calculated according to an initial weight 126, and the default probability 222 is obtained. Wherein, when the above discrete value is less than the above discrete threshold value 125 and it is judged as "discrete is not obvious" and the example of the calculation method is also as described above, it will not be repeated here.

When the above discrete value is not less than (ie greater than or equal to) the above discrete threshold value 125 , in step 441 , the above initial weight 126 is automatically corrected as a modified weight. Calculate the at least one unweighted default probability according to the modified weight, and obtain the default probability 222 . Wherein, when the above-mentioned discrete value is not less than the above-mentioned discrete threshold value of 125 and it is judged to be "obviously discrete" and the example of the calculation method is as shown above, it will not be repeated here.

Finally, automatically convert the default probability 222 to an internal rating 221 according to the default probability 222 and the rating comparison table 123 , and output the default probability 222 and the internal rating 221 as in step 450 . Among them, examples of the output format of the above-mentioned probability of default 222 and the above-mentioned internal rating 221 can be shown in Table 5-7 above, or other output formats that can provide data content including the above-mentioned probability of default 222 and the above-mentioned internal rating 221.

According to yet another embodiment, after submitting the above-mentioned credit evaluation request 110, as in step 420, if the above-mentioned bank 100 assesses that there is a need to change or modify any of the above-mentioned credit evaluation data 120, as in step 421, it can be modified 1. Credit evaluation data 120. Among them, the above-mentioned credit evaluation data 120 is the data that has been built in the above-mentioned system 200 and has initial settings, including the above-mentioned data threshold number 121, the above-mentioned default formula 122, the above-mentioned rating comparison table 123, the above-mentioned discrete formula 124, and the above-mentioned discrete threshold value 125 , the above-mentioned initial weight 126 , and a threshold ratio 127 of shortfall. For example, the value of the above-mentioned initial weight 126 can be adjusted and set in the aforementioned manner, which will not be repeated here.

According to yet another embodiment, an undervalue ratio of the at least one unweighted probability of default is further calculated, and the undervalue ratio is compared with an undervalue threshold ratio 127 . When the aforementioned missing value ratio is not less than (that is, greater than or equal to) the aforementioned missing value threshold ratio 127, the aforementioned initial weight 126 is automatically adjusted to the aforementioned corrected weight. Next, the at least one unweighted probability of default is calculated according to the modified weight. And when the above-mentioned undervalue ratio is smaller than the above-mentioned undervalue threshold ratio 127 , the above-mentioned at least one unweighted default probability is calculated according to the above-mentioned initial weight 126 . Among them, when the above-mentioned missing value ratio is not less than the above-mentioned missing value threshold ratio 127 and it is judged as "the missing value ratio is too high" and the example of the calculation method is as shown above, it will not be repeated here.

According to yet another embodiment, when the above-mentioned discrete value is not less than (that is, greater than or equal to) the above-mentioned discrete threshold value 125, the above-mentioned at least one unweighted probability of default can be manually selected or automatically output, and automatically output according to the above-mentioned rating comparison table 123 Converting said at least one internal rating 222 of said at least one unweighted probability of default. Among them, when the above-mentioned discrete value is not less than the above-mentioned discrete threshold value 125 and it is judged to be "obviously discrete" and the calculation method, the above-mentioned at least one unweighted default probability and the corresponding above-mentioned at least one internal rating 222 are directly and automatically output , as shown above, and will not be repeated here.

Combining the above-mentioned processes of the bank's internal credit risk assessment system and bank's internal credit risk assessment method, when a business requester applies for a loan or other related business content at the bank, the bank can automatically and quickly use the system of the present invention Credit risk data for credit risk assessment. And through multiple default risk calculation methods, discrete formulas for credit risk data, and technical means such as credit risk data gap ratios, multiple checks are made on the credit risk assessment results. Therefore, while providing real-time credit risk assessment results, it can achieve a balance between shortening the waiting time for business demanders and accurately calculating the bank's credit risk assessment.

Even when the bank's internal credit information is insufficient, the present invention can better integrate the rating results made by multiple external credit rating units, and further convert it into a commonly used rating method inside the bank, and can provide credible and real-time credit assessment results. Therefore, the present invention really solves the problems of long working hours and low work efficiency of the assessment procedures in the field, and the inability to conduct credit risk assessment of business demanders under the condition of insufficient internal bank information.

The present invention is only disclosed in preferred embodiments herein, but anyone skilled in the art should understand that the above embodiments are only used to describe the present invention, and are not intended to limit the scope of patent rights claimed by the present invention. All changes or replacements that are equal or equivalent to the above-mentioned embodiments are all It should be construed as being included within the spirit or scope of the present invention. Therefore, the scope of protection of the present invention should be defined by the scope of the following patent application.

100: Bank

110: Credit Evaluation Request

120:Credit evaluation information

200: Bank internal credit risk assessment system

210: data management device

211: Internal credit database

212: External Rating Request

220: Evaluation device

221:Internal Rating

222: Default probability

300: External Rating Unit

310: External Rating

Claims (2)

An internal credit risk assessment system of a bank. When a business demander applies to a bank for business, the system is used to assess a default probability of the business demander. The system includes: a data management device, including an internal credit database , the data management device automatically reads the internal credit data of the business demander in the internal credit database, and calculates the data number of the internal credit data, when the data number is less than a data threshold number, Automatically request at least one external rating of the business demander from at least one external rating unit, and when the request is approved, the data management device receives the at least one external rating; and an evaluation device, automatically according to a rating comparison table converting the at least one external rating to an internal rating to obtain at least one unweighted probability of default, and then calculating a discrete value of the at least one unweighted probability of default according to a discrete formula, and calculating the at least one unweighted probability of default A missing value ratio, when the discrete value is greater than or equal to a discrete threshold value and when the missing value ratio is greater than or equal to a missing value threshold ratio, an initial weight is automatically corrected to be a modified weight, so as to calculate and obtain a value according to the modified weight probability of default. The internal credit risk assessment system of the bank as described in claim item 1, wherein a credit assessment data is modified through the data management device, and the credit assessment data includes the data threshold number, the default formula, the rating comparison table, the A discrete formula, the discrete threshold value, the initial weight, and a missing value threshold ratio.

Images