CN113450202A

CN113450202A - Credit rating system and credit rating method

Info

Publication number: CN113450202A
Application number: CN202010231385.8A
Authority: CN
Inventors: 黄麒修
Original assignee: Taiwan United Financial Technology Co ltd
Current assignee: Taiwan United Financial Technology Co ltd
Priority date: 2020-03-27
Filing date: 2020-03-27
Publication date: 2021-09-28

Abstract

The invention provides a credit rating system and a credit rating method. The method comprises the steps of determining the scores of each customer in each variable according to the relative conditions of a plurality of customers in a plurality of variables, and calculating the credit scores of each customer according to the weights of each variable preset or adjustable along with the score distribution. In addition, the credit rating of each client can be adjusted by adjusting the weight of each variable according to the score distribution of the client with poor credit performance relative to all clients in each variable.

Description

Credit rating system and credit rating method

Technical Field

The present invention relates to a credit rating system and a credit rating method, and more particularly, to a credit rating system and a credit rating method for calculating credit ratings of clients according to the distribution status of scores of the clients in a plurality of variables.

Background

Along with the development of society, fund lending and investment financing are more and more common behaviors. Whether it is a capital provider or a capital demander, a good credit rating mechanism is required to properly evaluate each capital loan and/or each investment to manage money, to judge whether it is worth to engage and to decide how to maximize benefits and minimize risks, and particularly for corporate institutions that are going to engage in commercial activities such as loan or investment, if the credit rating of each client can be accurately and uniquely calculated, the competitiveness of the corporate institutions can be significantly enhanced.

Generally, to obtain a credit rating or the like of a specific client, the credit rating or the like may be obtained from a private company providing a credit rating such as standard popularization (standard popularization), moody (moods), and reputation (fiducings), or may be obtained by a mechanism of collecting public information or even non-public information of the specific client and performing a credit rating of its own. However, the former method is limited by the fact that the companies often only provide credit ratings of certain specific clients and all can pay for the credit ratings, and the latter method is limited by the scope of data collected and the mechanisms of credit rating performed, and the obtained credit ratings of the specific clients are not unique or accurate enough, or even have difficulty in failing to obtain the credit ratings of the specific clients.

In particular, in the prior art, the credit rating mechanism used in any of the above manners is basically that, at a specific time, data covering certain predetermined variables, which can be obtained up to the specific time, related to specific clients are collected, and then the data are processed by using a predetermined formula to calculate the credit rating of each specific client at the specific time. However, the formulas and variables used cannot be properly adjusted after this particular time as the customer evolves over time, but instead new formulas must be developed and tested in addition. In addition, in the prior art, professional evaluation can be added to adjust data of one or more variables, but the influence of human subjective evaluation on credit evaluation and the like obtained by calculation cannot be properly and systematically adjusted dynamically after the specific time.

In view of the above, it can be seen that there exists a long-standing need in the prior art for providing credit scores that are unique and can be systematically and dynamically adjusted, and there is a need to improve the prior art to solve the problem.

Disclosure of Invention

The credit evaluation mechanism used in the invention is that any variable used for calculating the credit evaluation of a plurality of clients uses the relative distribution of the original data of the clients to determine the score of each client at the variable or even determine the weight of the variable, thereby integrating the scores and the weights of a plurality of variables to obtain the credit evaluation of each client calculated according to the relative performance of the clients.

The credit rating mechanism used in the invention can be adjusted according to the credit performance of some clients in the real world after the credit rating is given to the clients. According to the point distribution of the credit-poor-performing customers relative to all the customers in the variables, the variables with obviously lower points of the credit-poor-performing customers are found, and then the weights corresponding to the variables are increased.

The invention provides a credit rating system, which at least comprises a data storage unit, a conversion score unit and a weight score unit, and can further comprise a weight adjusting unit and a rating updating unit. The data storage unit is used for storing data, and the data are respectively stored according to the corresponding variables and the corresponding clients. The conversion score unit is used for converting the relative distribution of the data of the clients in each variable into the respective scores of the clients in the variables. The weight score unit is used for calculating the credit score and the like of each customer according to the weight of each variable and the score of each customer in each variable, wherein the weight of any variable can be preset or adjusted according to the relative distribution of the scores of the customers in the variables. In addition, the weight adjusting unit is used for finding out one or more variables with low credit performance of the customers according to the score distribution difference of the credit performance poor part and the integral part of the customers in all variables in a period of time and correspondingly adjusting the weight of the variables. Meanwhile, the evaluation updating unit is used for regularly and irregularly updating the data stored in the data storage unit and enabling the conversion score unit and the weight score unit to calculate again to obtain a new credit evaluation result. The credit evaluation system provided by the present invention may be configured on a single computer device or configured on a plurality of computer devices connected to each other through a wired network and/or a wireless network, and any unit may be composed of any hardware with storage function and any hardware with calculation function and may execute software and/or firmware.

The invention provides a credit rating method, which at least comprises the following steps: first, public data and non-public data about any variable for evaluating credit rating and any client expected to make credit rating are stored separately. Then, the score of each customer in the variable is calculated according to the relative distribution of the data of the customers in each variable. Then, the credit score of each client is calculated according to the weight of each variable and the score of each client in each variable, and the weight of any variable can be either predetermined or adjusted according to the score distribution condition of the clients among the variables. In addition, the method can also comprise at least one step of finding out partial variables with low scores of the customers with poor credit performance according to the score distribution difference of the partial and the whole part of the customers with poor credit performance in all variables in the real world after calculating the credit rating of each customer and the like, and correspondingly adjusting the weight of the partial variables. Second, the data stored in the storage is periodically and aperiodically updated, and the respective updated credit ratings of the clients are recalculated based on the updated stored data.

Drawings

FIG. 1 is a schematic diagram of the operation of the present invention.

Fig. 2A to 2C are some structural diagrams of the credit rating system according to the present invention.

Fig. 3A to 3C are some flowcharts of the credit rating method according to the present invention.

Fig. 4A shows a data format obtained by using a crawler for business registration information to retrieve enterprise information to be queried through a network or an external hard disk in the data storage unit and the data storage step according to the present invention.

Fig. 4B shows the data format of the litigation data crawler for retrieving the enterprise information to be queried through the network or external hard disk used in the data storage unit and data storage step according to the present invention.

Fig. 4C shows the data format of the government standard form crawler for searching the enterprise information to be searched through the network or external hard disk used in the data storage unit and data storage step according to the present invention. Fig. 5A shows that in the score conversion unit and score conversion step, the previously obtained information is classified into three categories, i.e. original information, analyzable variables and numerical values, and each piece of original information of each client has corresponding analyzable data and numerical values.

FIG. 5B shows that in the score conversion unit and score conversion step, the values of each variable of each customer are integrated to obtain a plurality of values of the variables of the customers.

Fig. 6A illustrates that in the weight fraction unit and weight fraction step, the variables and their corresponding values can be de-scaled to eliminate the misinterpretation that may be caused by the size difference between different values of different variables (such as the company capital amount or the company operating time), and to emphasize the relative distribution of different values of different customers in each variable.

FIG. 6B is a result of the present invention in the weight score unit and weight score step, which can be scaled using the values accumulated by the clients in the variables.

FIG. 6C shows that the weight fraction unit and the weight fraction step of the present invention can divide and establish the clustering interval.

Fig. 6D shows that in the weight-score unit and the weight-score step of the present invention, the scores of the clients at the variables can be calculated by using the corresponding intervals of the clients at the variables.

FIG. 6E shows the credit scores of all clients that can be presented using the ranking type in the weight score unit and weight score step according to the present invention.

Fig. 7A shows an embodiment of the present invention, during the operation of the weight adjustment unit and the weight adjustment step, the respective weights of the variables can be adjusted according to the differences between the mean and the standard deviation of the variables.

Fig. 7B is a diagram illustrating the detailed operation of the weight adjustment unit and the weight adjustment process according to the present invention, the data can be integrated into the credit rating information of the customers, and the average and standard deviation of the scores of all the customers in each variable can be calculated.

Fig. 7C shows a specific operation of the weight adjusting unit and the weight adjusting step, according to the present invention, only a portion of the specific clients marked to have been rejected or refunded during the operation time can be calculated as the average value and the standard deviation of the scores of the specific clients in each variable.

Fig. 7D illustrates the specific operation of the weight adjustment unit and the weight adjustment process according to the present invention, the average score and the normalized falling point of each variable of the specific clients can be calculated according to the values of each variable in fig. 7B and 7C.

Fig. 7E shows the specific operation of the weight adjustment unit and the weight adjustment step of the present invention, in order to adjust the weights corresponding to each variable, the lower the normalized falling point is, the higher the weight score is given, such that if the normalized falling point of a certain variable falls in the interval 1, the weight score is 5, and if the normalized falling point falls in the interval 2, the weight score is 4, and the weight score is decreased to 0.

Fig. 7F illustrates the detailed operation of the weight adjustment unit and the weight adjustment process of the invention, which can adjust the weights of the variables according to the different weight scores between the variables, the basic concept is that if the weight score of an item is higher, the adjusted weight is heavier, and this is the variable with lower normalized drop point represents that the difference between the specific clients and all the clients in the variable is less obvious.

[ description of main element symbols ]

Credit rating system 100

Data Source 101

Credit rating result 102

Credit rating system 200

Data storage unit 201

Conversion score Unit 202

Weight fraction unit 203

Weight adjustment unit 204

Rating updating unit 205

Credit rating method 300

Data storage step 301

Score conversion step 302

Weight fraction step 303

Weight adjustment step 304

Evaluation update step 305

Detailed Description

The operation of the credit rating system and the credit rating method provided by the invention can be summarized as shown in fig. 1. A credit rating system 100, which may perform a credit rating method, receives data from one or more data sources 101 to be used in calculating credit ratings of a plurality of clients, and outputs credit rating results 102 of the clients. Here, different data sources 101 may respectively provide data covering one or more different variables and corresponding to one or more clients, and the credit evaluation result 102 includes a plurality of clients and their corresponding credit evaluations.

The invention is not limited or restricted by the details of how the credit system 100, data source 102 and credit rating results 102 interact. For example, different data provided by different data sources 101 may be either public data or non-public data. For example, some data sources 101 may be their credit ratings for certain customers, such as those provided by folk companies that provide credit ratings such as standard poors, mucky and boomer. For example, some data sources 101 may be open web sites or pay-for-demand databases that provide information such as government publications, corporate and industrial business registration information, property mortgage information, real estate mortgage information, legal action information, ticket information, import and export performance information, government filing information, stock and bond market information, and the like. For example, some data sources may be mass media, credit unions, expert students' opinions, or credit rating system 100 that provide information about the customers whose credit is to be assessed from previous records of their interactions with one or more customers. In addition, there are many ways to transfer data from these data sources 101 to the credit evaluation system 100, such as using a crawler to find the needed data from the relevant website and transfer the data to the credit evaluation system 100, using different word processors and different logic algorithms to convert data from different sources with different formats into variables that can be analyzed, using a semantic parser to capture data from a text file for one or more variables of a customer. It is clear that the present invention is independent of how the required data is acquired and of how the generated credit scores etc. are applied, and that the system and method proposed by the present invention is focused on how to calculate the respective credit scores etc. of the individual clients from the data relating to these clients. The present invention is also not limited to the kind and number of variables used to calculate the credit rating, and the more the number of data sources and the richer the contents are, the more the kinds of variables used in the present invention are. For example, the variables used in the present invention may be credit evaluation given to the target customer by a company such as credit evaluation, annual revenue, capital amount, pre-tax credit, number of ticket jumps, interest rate of debt of the issuing company, credit record of the business class of the target customer or pre-criminal department, number of change of the director list of the target customer, number of times reported by a newspaper magazine, and employee number.

The basic architecture of the credit rating system proposed by the present invention can be summarized as shown in fig. 2A. The credit rating system 200 comprises at least a data storage unit 201, a conversion score unit 202 and a weight score unit 203, wherein one or more data from at least one data source external to the credit rating system 200 is transmitted to the data storage unit 201, and the calculated credit rating of each customer is output from the weight score unit 203 to the outside of the credit rating system 200. Further, as shown in fig. 2B and fig. 2C, the credit rating system 200 may further include a weight adjustment unit 204 and a rating update unit 205, so that not only the credit rating of some clients may be calculated, but also the process and the result of calculating the credit rating may be updated. Briefly, the data storage unit 201 is used for storing data according to corresponding variables and corresponding clients, the score conversion unit 202 is used for converting the relative distribution of the data of the clients in each variable into respective scores of the clients in each variable, and the weight score unit 203 is used for calculating credit scores of the clients according to the respective weights of the variables and the respective scores of the clients in the variables. And, the weight adjusting unit 204 is used for finding out one or more variables with low credit performance of the customers according to the score distribution difference of the credit performance part and the whole part of the customers in all variables in a period of time after calculating the credit rating of the customers and adjusting the weight of the variables correspondingly. Meanwhile, the evaluation updating unit 205 is used for periodically and aperiodically updating the data stored in the data storage unit after calculating the credit evaluation of the clients, and letting the conversion score unit and the weight score unit calculate again to obtain a new credit evaluation result.

The credit evaluation system 200 may be configured on a single computer device, or may be configured on a plurality of computer devices connected to each other via a wired network and/or a wireless network, and any one of the units is a part of the computer device as long as it can provide the functions required by the unit. For example, the computer device can be a server, a host computer, a desktop computer, a notebook computer, a tablet computer, a smart phone, and combinations thereof, and the network can be a wired network, a wireless network, and combinations thereof. For example, any unit is composed of at least one piece of hardware with storage function and/or at least one piece of hardware with calculation function, and can execute software and/or firmware, the usable hardware with storage function can be dynamic random access memory, read only memory, flash memory, hard disk, solid state disk, flash disk, network hard disk, optical disk and their combination, and the usable hardware with calculation function at least includes analog circuit, digital circuit, central processing unit, programmable processor, digital signal processor, programmable controller, graphics processor, integrated circuit, special purpose integrated circuit and their combination. Indeed, the present invention may use any existing, developing, or future hardware to construct the credit rating system 200.

The basic flow chart of the credit rating method provided by the invention can be summarized as shown in fig. 3A. The credit rating method 300 comprises a data storage step 301, a conversion score step 302, and a weight score step 303, wherein one or more data from one or more data sources is transmitted to the data storage step 301, and the calculated credit rating for each customer is output after the weight score step 303 is completed. Further, as shown in fig. 3B and fig. 3C, the credit rating method 300 may further include a weight adjustment step 304, and may further include a rating updating step 305, so that not only the credit ratings of some clients may be calculated, but also the process and the result of calculating the credit rating may be updated. Briefly, the data storage step 301 is used to store public data and non-public data related to any variable for evaluating credit rating and any client expected to make credit rating, the score conversion step 302 is used to calculate the score of each client in the variable according to the relative distribution of the data of the clients in each variable, and the weight score step 303 is used to calculate the credit rating of each client according to the respective weight of each variable and the score of each client in each variable. And, the weight adjusting step 304 is used for finding out one or more variables with poor credit performance of the customers with low scores according to the score distribution difference of the poor credit performance part and the whole part of the customers in all variables in a period of time after calculating the credit scores of the customers and adjusting the weights of the variables correspondingly. Meanwhile, the evaluation updating step 305 is used for periodically and aperiodically updating the data stored in the data storage unit after calculating the credit evaluation of the clients, and letting the conversion score unit and the weight score unit calculate again to obtain a new credit evaluation result.

Reasonably, the credit rating system 200 of the present invention is corresponding to the credit rating method of the present invention, and the data storage unit 201, the conversion score unit 202, the weight score unit 203, the weight adjustment unit 204 and the rating update unit 205 are respectively corresponding to the data storage step 301, the conversion score step 302, the weight score step 303, the weight adjustment step 304 and the rating update step 305.

The data storage unit 201 and the data storage step 301 do not limit how the data are found and transferred from various public or private databases and the like to be stored, nor do they limit how the data are stored, only the raw data of each client that can be presented in each variable. For example, when the variables used are the company capital amount, the company stock price and the company sales amount, the company customers who need to make credit assessment can be queried about the respective capital amount, the stock price and the sales amount. That is, the present invention is not limited to and does not limit how these data sources are found and transmitted for storage, and any existing, developing, and future technologies and products may be used to find and transmit these data.

For example, the crawler used in the data storage unit 201 and the data storage step 301 to retrieve the enterprise information to be queried through a network or an external hard disk may be a web-based information registration crawler and obtain the data format shown in fig. 4A, a litigation data crawler and obtain the data format shown in fig. 4B, a government standard form crawler and obtain the data format shown in fig. 4C, or a ticketing crawler, an online data registration crawler for a mobile asset guarantee transaction, or other types of crawlers.

The conversion score unit 202 and the conversion score step 302 are a big feature of the present invention, and the calculation of the score of each customer in any variable is determined according to the relative distribution of the customers in the raw data of the variable, wherein the closer the raw data of some customers are to each other, the closer the score of the variable is, and the farther the average of the raw data of any customer and the raw data of the customers is, the larger (or smaller) the score of the variable is. In other words, compared to the prior art that the original data of a customer at a variable is calculated by using a predetermined formula to obtain the score of the customer at the variable, so that a plurality of customers with similar original data all have similar scores, the present invention can distinguish the differences of the customers at the variable because the present invention can still give different scores according to the relative distribution of the customers with each other (or the difference of the original data).

The score conversion unit 202 and score conversion step 302 give each variable its own score range upper and lower limits, i.e., different variables may have different score ranges. This is because the more important variables have different importance levels when calculating the credit rating, the more important variables need a larger score range to make the score difference between the variables of different customers larger, and the weighting unit 203 and the weighting score step 303 can obtain different credit ratings when calculating the credit rating using the weights of the variables and the scores of the variables of the customers. For example, variables corresponding to sales, ticket skip, and capital amounts of individual customers tend to have a large range of scores, but variables corresponding to winning times and reporting times of newspaper magazines of individual customers tend to have a small range of scores.

The score conversion unit 202 and the score conversion step 302 calculate the score of each customer in each variable according to the respective raw data of each customer in the data stored in the data storage unit 201 and the data storage step 301 and the average value and the standard deviation of the raw data of the customers when processing each variable. That is, for each variable, the average and standard deviation of the distribution of the variable corresponding to the clients are calculated using all the raw data of all the clients, and then the score of the variable corresponding to each client is given to each client one by one according to the condition of the raw data of each client relative to the average and standard deviation.

The score conversion unit 202 and score conversion step 302 may vary how each individual customer is given their respective scores in some variable. For example, the score of a customer at a variable may be set to the middle of the range of scores for the variable when the raw data for the customer differs from the average of the raw data for the customers by no more than X standard deviations, where X is a number greater than zero. For example, when the absolute value of the difference between the original data of a certain client and the average value of the original data of the clients is greater than NX standard deviations but not greater than (N +1) X standard deviations, the absolute value of the difference between the score of the client at the variable and the median of the score range of the variable may be set to N, but if N is not less than half of the score range of the variable, the absolute value of the difference between the score of the client at the variable and the median of the score range of the variable may be set to half of the score range, where N is a positive integer and X is a number greater than zero. For example, at least one variable may be set such that the score of the variable is closer to the upper limit of the score range of the variable when the raw data of any client is larger than the average value of the raw data of the clients, and the score of the variable is closer to the lower limit of the score range of the variable when the raw data of any client is smaller than the average value of the raw data of the clients. For example, at least one variable may be set to have a score closer to a lower limit of the score range of the variable when the raw data of any client is larger than the average value of the raw data of the clients, and may be set to have a score closer to an upper limit of the score range of the variable when the raw data of any client is smaller than the average value of the raw data of the clients. That is, in at least one variable, the middle value of the score range of the variable can be made to correspond to the average value of the raw data of the clients, and then the score of the variable in the score range is increased or decreased as the difference between the raw data of each client and the average value increases until the difference is so large that the score can not be increased or decreased any more due to the fact that the difference between the raw data of a client and the average value is measured by using the standard deviation as a unit, because the standard deviation is generally used in the statistical concept to reflect the dispersion degree among a plurality of individuals in a group. The score of the customer at the variable is increased or decreased as the difference between the customer raw data and the average of the customer raw data increases, depending on the nature of the variables. For example, if a variable is obtained corresponding to a company capital amount, a company sales amount, or a company tax, it can be that the more standard deviation a customer's raw data is above the average, the higher the customer's score for the variable. For example, if a variable is the interest rate corresponding to the number of times a company jumps out tickets or debt issued by a company, it may be that the raw data of a customer has a higher score for the variable when the raw data is a standard deviation greater than the average. Of course, for variables with more important or sensitive credit ratings, the corresponding value of X is smaller (as measured by half the standard deviation when X equals 0.5), thereby highlighting differences among different customers.

The score conversion unit 202 and score conversion step 302 may vary how each individual customer is given their respective scores in some variable. For example, the score of the client at the variable may be set as the lower limit of the score range of the variable when at least one variable is that the raw data of the client is higher than the average value of the raw data of the clients by at least MX standard deviations, and the score of the client at the variable may be set as the lower limit of the score range of the variable plus N when the raw data of the client is higher than the average value of the raw data of the clients by between (M-N) standard deviations and (M- (N-1)) standard deviations, but the score of the client at the variable may be set as the upper limit of the score range of the variable when N is so large that the lower limit of the score range of the variable plus N is not less than the upper limit of the score range, where M and N are positive integers respectively, and X is a number greater than zero. For example, the score of the client at the variable may be set as the lower limit of the score range of the variable when the raw data of the client is lower than the average value of the raw data of the clients by at least MX standard deviations, and the score of the client at the variable may be set as the lower limit of the score range of the variable plus N when the raw data of the client is lower than the average value of the raw data of the clients by between (M-N) standard deviations and (M- (N-1)) standard deviations, but the score of the client at the variable may be set as the upper limit of the score range of the variable when N is so large that the lower limit of the score range of the variable plus N is not less than the upper limit of the score range, where M and N are positive integers, respectively, and X is a number greater than zero. That is, in at least one variable, if the nature of the variable is as low as possible, the original data is as low as possible (e.g., the pre-crime department of the company's business hierarchy or the number of times the company is penalized by the government authority), it may be set such that if the original data of a particular company is lower than the average of the original data of the customers by a certain number of standard deviations, the score of the particular company at the variable is set to the upper limit of the score range of the variable (i.e., the highest score that can be obtained at the variable), and then, at other customers, the score at the variable is set to the upper limit of the score range of the variable minus the number of standard deviations of the original data of the particular company, depending on how many standard deviations the original data is higher than the original data of the particular company. Of course, if the nature of a variable is such that the higher the raw data is (e.g., the company receives a credit rating on a standard pul or the like), it can be compared in a similar manner, except that it is changed to set the company to set the upper limit of the range of the score of the variable when the raw data of the company is higher than the average of the raw data of the customers by a certain number of standard deviations.

For example, in the score conversion unit 202 and the score conversion step 302, the previously obtained information may be listed into three categories of original information, analyzable variables and values, and each original information of each customer has corresponding analyzable data and values, as shown in fig. 5A, and then each value of each variable of each customer is input and integrated to obtain a plurality of values of the variables of the customers, as shown in fig. 5B.

The weight-score unit 203 and the weight-score step 303 are a big feature of the present invention, and the weight of any variable herein can be determined according to the relative distribution of the customers in the original data of the variable or the relative distribution of the scores of the customers in the variable, besides being predetermined. In other words, in addition to giving fixed weights to each of some variables to weight and add the scores of the variables of a certain client to calculate the credit rating of the client, the weights of different variables basically depend only on the nature of the variable, and the invention can adjust the weights of the variables according to the relative distribution (or difference) of the raw data of the client in one or more variables, so that the variables which are sensitive to the performance of different clients can have larger weights, and the invention can further distinguish the differences of the clients in the calculated credit rating and the like.

How a variable is given a weight in a weight scoring unit 203 and weight scoring step 303 may vary widely. For example, the weights used for at least one variable may be predetermined, i.e., the predetermined weights for the variables may be received separately when receiving data needed to measure credit ratings and the like for the clients. For example, the predetermined weights used for at least two variables may be different, that is, different weights may be given according to different properties of different variables, and the variables having larger influence on credit rating and the like may be given larger weights. For example, the weights used in at least one variable may be adjusted according to the relative distribution of the clients in the variable, i.e., not only the score of each client in the variable depends on the mutual distribution of the raw data of the clients. Generally, if the relative distribution of the raw data of the clients in a variable is more divergent, the weight of the variable can be increased, so that the credit score obtained by the final calculation can further highlight the difference of the clients. For example, the relative proportions of the weights of the variables between at least two variables may be adjusted according to the relative proportions of the standard deviations of the customer's fractional distributions between the variables, where the greater the standard deviation of the customer's fractional distributions for any one variable, the greater the weight of that variable. For example, after the weights of at least one variable are adjusted, the weights of all the variables are multiplied by an adjustment multiple, so that the total weight of the adjusted variables is equal to the total weight of the variables before adjustment, thereby keeping the obtained credit scores within a certain quantity range, and conveniently comparing the results of credit scores before and after one or more times of weight adjustment.

For example, the detailed operations of the weight fraction unit 203 and the weight fraction step 303 can be illustrated as follows. First, as shown in fig. 6A, the variables and their corresponding values are descaled to eliminate the misinterpretation that may be caused by the size difference between different values of different variables (such as the company capital amount or the company operating time), and to emphasize the relative distribution of different values of different customers in each variable. The scale value of the variable related to the company scale is determined by using various variables with different scales established by accumulating information of each company, then the scale values are multiplied to calculate a single scale factor, and then the variable influenced by the company scale is divided by the previously calculated scale factor, thereby carrying out the descale. For example, if variables 1 and 2 are the accumulated capital amount and operating time information, the average and standard deviation calculated in the two variables are used to divide each variable into nine intervals and classify the scale variable into 1-9 values, wherein a lower value indicates a smaller company capital amount or a longer company operating time. If the variables 3 and 4 are the number of civil litigation and the number of criminal litigation accumulated by the company, business disputes are more likely to occur and litigation is more likely to occur as the capital amount of the company increases and the operating time increases, but the balance can be achieved by dividing the scale factor calculated previously. Then, the results after the scale processing of each company are substituted into the calculation of credit evaluation and the like. Here, as shown in fig. 6B, the values of the variables accumulated by the clients are used to scale up, and the clustering interval is divided and established as shown in fig. 6C. For example, interval 1 used to estimate variable 1 from variable 1 at the parameter shown in fig. 6B may be between the mean 1-2 standard deviation 1 and the mean 1-1.5 standard deviation, and interval 2 used to estimate variable 1 from variable 1 at the parameter shown in fig. 6B may be between the mean 1-1.5 standard deviation 1 and the mean 1-1 standard deviation 1. By analogy, the drop point of each client in each variable interval can be judged. Next, the score of each client for each variable is calculated using the section to which each client corresponds for each variable, as shown in fig. 6D. For example, if a client falls in the interval 2 of the variable 1, it indicates that the client has a score of 2 in the variable 1, and if a client falls in the interval 3 of the variable 2, it indicates that the client has a score of 3 in the variable 1. Finally, for any customer, the scores of all the variables are added to obtain the credit rating of the customer (presented in the form of score). Of course, the average (m) and standard deviation(s) of the credit ratings of the customers may be calculated by accumulating the credit ratings (score type) of the customers, and then a credit rating presented by the rating type may be given to any customer according to how many standard deviations the credit rating (score type) is higher or lower than the average. For example, if the credit rating (score type) of a client is between m +2 s and m +2.5 s, the credit rating in the form of a rating is considered as a1 rating, and if the credit rating (score type) of another client is between m +1.5 s and m +2 s, the credit rating in the form of a rating is considered as a B1 rating. By analogy, credit scores of all customers can be obtained, as in the example shown in fig. 6E.

The weight adjustment unit 204 and the weight adjustment step 304 are a major feature of the present invention, and the respective weights of the variables can be adjusted according to the relative distribution of the scores of some special customers, which are inferior in performance in the real world, in relation to all customers in the variables among the customers to be calculated credit scores. In other words, at a certain starting time, there is often not enough credit performance data (such as the refund record or the number of times of refused transactions) of the clients, so that credit evaluation can be performed only according to the data available at that time, but some special clients among the clients may have poor credit performance (such as refunded or refused transactions) in the real world within a period of time after calculating the credit evaluation of the clients, so that the difference between the distribution of the variables of the special clients and the distribution of the variables of all clients can be utilized to find one or more special variables that can distinguish the special clients from all clients, and the weights of the special variables are increased to highlight the special variables that are more likely to predict poor credit performance of the clients in the future, the impact of the calculated credit rating.

The weight adjustment unit 204 and the weight adjustment step 304 can be summarized as how to find out one or more variables from the variables that can identify the customers with poor credit performance and adjust the weights of the variables accordingly, first, calculate the average value and standard deviation of each variable of the customers with pending credit rating and the like at a starting time according to the data available at that time. Secondly, some special clients whose records are less well-documented in the real world are received and stored for a period of time after the starting time. Then, the specific mean and standard deviation of each variable of the specific clients are calculated, and the way how to calculate the mean and standard deviation of each corresponding variable from some data is the same except that only the data of the specific clients are used instead of the data of all clients, and is not burdensome. Then, each variable is calculated with its corresponding quality factor, and the calculation process of the quality factor corresponding to any variable is to calculate the difference between the special average value of the special customers and the average value of all customers in the variable, and then divide the difference by the standard deviation of all customers. Then, the maximum quality factor value and the minimum quality factor value in the quality factors are found, and the value between the maximum quality factor value and the minimum quality factor value is divided into equal parts of K, wherein K is a positive integer. Finally, in each variable, when the corresponding quality factor is located in the J-th part, the weight corresponding to the variable is adjusted to J, wherein J is a positive integer. For example, if the calculated quality factors for each variable are-1.5, -2.0, -5, 1.5, 2, …, -2, -1.7, respectively, i.e., the minimum quality factor is-2.0 and the maximum quality factor is 2.0, then the range of-2 to 2 can be divided into five equal parts, such that variables having quality factors closer to-2.0 have adjusted weights closer to 5 and variables having quality factors closer to 2.0 have adjusted weights closer to 1. Obviously, if some clients in the real world already have records of bad credit performance such as refunds or refusals, etc. some clients may not be able to receive any credit, especially if the number of such bad records is so large that statistically sufficient data stability is obtained, the weight adjustment feature of the present invention can increase the weight of one or more variables that are more likely to identify poor credit performance in a future period of time at the start of the process, such that after adjusting the respective weights of the variables, customers who are more likely to have poor credit performance in the future may be given lower credit ratings because one or more variables have lower adjusted weights, and may allow those who are less likely to have poor credit performance in the future to have a higher credit rating because of the higher adjusted weight of one or more variables.

The weight adjustment unit 204 and the weight adjustment step 304 can adjust the weights of these variables in various ways. For example, after the weights of the variables are adjusted, the weights of all the variables may be multiplied by an adjustment multiple, so that the total weight of the variables after adjustment is equal to the total weight of the variables before adjustment, thereby making the credit rating of each client obtained before adjustment and the credit rating of each client obtained after adjustment fall within the same value range.

For example, the detailed operations of the weight adjusting unit 204 and the weight adjusting step 304 can be illustrated as follows. First, when the credit evaluation mechanism is just established and operated, there is no way to set the weights corresponding to the different variables respectively, because there is not enough information to verify the credit evaluation given by the credit evaluation mechanism. However, as shown in fig. 7A, the weights of the variables may be adjusted according to the differences between the mean and the standard deviation of the variables. The basic concept is that a variable with a larger standard deviation represents a different client with a larger difference in the variable, i.e. the larger the effect of the variable on the difference between clients is, the more weight it is worth to give.

For example, the detailed operations of the weight adjusting unit 204 and the weight adjusting step 304 can be illustrated as follows. After the credit rating mechanism has been established and operated for a period of time, if the number of times the customers have been rejected and refunded is large enough (e.g., reaching a statistically significant amount), the data can be integrated into the credit rating information of the customers, and the average and standard deviation of the scores of all customers in each variable can be calculated, as shown in fig. 7B. Then, only for the part of the specific customers marked as having been rejected or refunded during the operation time, the average value and the standard deviation of the scores of the part of the specific customers in each variable are calculated, as shown in fig. 7C. Of course, the same calculation is used, except that all of the customers are calculated or only the differences for that particular customer are calculated. Next, the average score and the normalized drop point of each variable of the specific clients are calculated based on the values of each variable in fig. 7B and 7C, as shown in fig. 7D. Next, using the values of FIG. 7D for each variable, the mean m 'and standard deviation s' are calculated based on the normalized landing points of the variables, and the interval is divided by this estimate. For example, if a division into 5 intervals is set, interval 1 is between m '-2.5 s' and m '-1.5 s' and interval 2 is between m '-1.5 s' and m '-0.5 s', and so on. Then, in order to adjust the weights corresponding to the variables, all variables are given higher weight scores as the normalized falling points are lower, such as if the normalized falling point of a variable falls in the interval 1, the weight score is 5, and if the normalized falling point falls in the interval 2, the weight score is 4, and all variables are decremented until the score reaches 0, as shown in fig. 7E. Finally, the weights of the variables are adjusted according to the different weight scores of the variables, and the basic concept is that the higher the weight score of a certain item is, the more the weight of the item is adjusted, and this is that the lower the normalized drop point is, the less obvious the difference between the specific clients and all the clients in the variable is, as shown in fig. 7F.

The evaluation updating unit 205 and the evaluation updating step 305 are a big feature of the present invention, and the credit evaluation of each client is not fixed after being calculated and generated at a certain time, or the weights of one or more variables are not adjusted only after one or more clients are refunded or rejected in the real world, but the credit evaluation of the clients is updated regularly or irregularly. In other words, the present invention can also continuously obtain updated credit scores of each client as the obtained data is updated.

How the credit ratings of the clients are updated regularly or irregularly by the rating update unit 205 and the rating update step 305 can be summarized as follows, first, the average value and the standard deviation of each variable of the clients are obtained at a start time. Then, the stored public data and non-public data of at least one client are updated at a time after the starting time, wherein the time may be a predetermined time length and the updating is performed at intervals of a predetermined time length after the starting time, or the time may be an unspecified time length and the updating is performed at intervals of a predetermined time length, such as receiving an updating instruction from the outside, and the like. Then, an updated score for each customer at each variable is calculated based on the relative distribution of the customers in the updated data. And finally, calculating the updated credit score of each client according to the weight of each variable and the updated score of each client in each variable.

How the rating updating unit 205 and the rating updating step 305 update the respective credit ratings of the customers may be varied. For example, the credit rating of each client may be updated only after at least R clients have data updated, where R is a positive integer. For example, the credit rating of each client may be updated only after at least S variables have data updated, where S is a positive integer. Therefore, the efficiency of the credit evaluation system and the credit evaluation method provided by the invention can be further improved, and the burden that the credit evaluation of all clients is completely updated only by updating a little data is reduced.

In summary of the above discussion, the credit rating system and the credit rating method provided by the present invention can obviously provide credit rating with uniqueness and systematization and dynamic adjustment. First, in the present invention, when calculating the credit rating of each client, the credit rating of the client is not calculated according to the data of all variables of the client and the predetermined formula as in the prior art, but is calculated according to the relative condition of each variable of the clients. I.e. the process is repeated. The present invention may particularly highlight the portion of customers where the credit is particularly good or particularly bad based on their mutual performance, without being limited by the degree of discrimination of the predetermined formula used and/or the integrity and accuracy of the data received and used. In addition, after calculating the credit rating of some customers, the invention can find out one or more variables with better discrimination of the credit performance and correspondingly lower and increase the weight of the one or more variables with better discrimination of the credit performance on the calculated credit rating according to the distribution of the part of special customers with poor performance in the real world relative to all the customers in the variables, thereby highlighting the influence of the one or more variables with better discrimination of the credit performance on the calculated credit rating. That is, the present invention can systematically and dynamically adjust the mechanism for calculating the credit rating according to the actual credit performance of the clients, without adjusting the credit rating mechanism by modifying the formula used as part of the prior art, and can avoid the human subjective influence when referring to the professional rating as part of the prior art. Of course, since the credit rating mechanism is fixed, which is how to calculate the credit rating of each client according to the received and stored data, the present invention may also periodically and/or aperiodically update the received and stored data, and further periodically and/or aperiodically update the credit rating of each client.

Obviously, many modifications and variations of the present invention are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims, the invention may be practiced otherwise than as specifically described. The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention; it is intended that all such equivalent changes and modifications be included within the scope of the present invention as defined by the appended claims.

Claims

1. A credit rating system, comprising:

the data storage unit is used for respectively storing data according to the corresponding variables and the corresponding clients;

a conversion score unit for converting the relative distribution of the data of the clients in each variable into respective scores of the clients in the variables; and

and the weight score unit is used for calculating the credit rating and the like of each client according to the respective weight of the variables and the respective scores of the clients in the variables.

2. The credit rating system of claim 1, further comprising at least one of:

the weight adjusting unit is used for finding out one or more variables with poor credit performance of the customers with low scores and correspondingly adjusting the weights of the variables according to the score distribution differences of the poor credit performance part and the whole part of the customers in all the variables after the credit performance of the customers is calculated for a period of time; and

and the evaluation updating unit is used for regularly and irregularly updating the data stored in the data storage unit after calculating the credit evaluation of the clients, and enabling the conversion score unit and the weight score unit to calculate again to obtain a new credit evaluation result.

3. The credit rating system of claim 1 or 2, further comprising at least one of:

the credit rating system is constructed on a single calculator device, wherein the calculator device is selected from a server, a computer host, a desktop computer, a notebook computer, a tablet computer, a smart phone and a combination thereof;

the credit evaluation system is a plurality of calculator devices which are mutually connected through a network, wherein the calculator devices are selected from a server, a computer host, a desktop computer, a notebook computer, a tablet computer, an intelligent mobile phone and a combination thereof; and

any unit is composed of at least one piece of hardware with storage function and/or at least one piece of hardware with calculation function and can execute software and/or firmware, wherein the hardware with storage function is selected from dynamic random access memory, read only memory, flash memory, hard disk, solid state disk, flash disk, network hard disk, optical disk and combination thereof, and the hardware with calculation function is selected from analog circuit, digital circuit, central processing unit, programmable processor, digital signal processor, programmable controller, graphic processor, integrated circuit, special purpose integrated circuit and combination thereof.

4. The credit rating system of claim 1, wherein the conversion scoring unit is configured to give upper and lower limits to respective score ranges of different variables, and to calculate a score of each customer at each variable based on respective raw data of each customer in the data stored in the data storage unit and an average and standard deviation of the raw data of the customers.

5. The credit rating system of claim 4, further comprising at least one of:

the conversion score unit sets the score of a client to the middle value of the score range of the variable when at least one variable is different from the average value of the raw data of the clients by no more than X standard deviations, wherein X is a number larger than zero; and

the conversion score unit sets, in at least one variable, an absolute value of a difference between a score of a certain client and a median of the raw data of the clients, when the absolute value of the difference between the raw data of the client and the average of the raw data of the clients is greater than NX standard deviations but not greater than (N +1) X standard deviations, to N, and sets, if N is not less than a half of the score range of the variable, the absolute value of a difference between the score of the client in the score range of the variable and the median of the score range of the variable, to a half of the score range, where N is a positive integer and X is a number greater than zero.

6. The credit rating system of claim 5, further comprising at least one of:

the conversion score unit sets the upper limit of the score range of the variable when the original data of any client is larger than the average value of the original data of the clients, and sets the lower limit of the score range of the variable when the original data of any client is smaller than the average value of the original data of the clients; and

the conversion score unit sets the lower limit of the score range of the variable when the original data of any client is larger than the average value of the original data of the clients, and sets the upper limit of the score range of the variable when the original data of any client is smaller than the average value of the original data of the clients.

7. The credit rating system of claim 4, further comprising at least one of:

the conversion fraction unit sets the fraction of a client to be the lower limit of the fraction range of the variable when the original data of the client is higher than the average value of the original data of the clients by at least MX standard deviations, sets the fraction of the client to be the lower limit of the fraction range of the variable plus N when the original data of the client is higher than the average value of the original data of the clients by (M-N) standard deviations to (M- (N-1)) standard deviations, but sets the fraction of the client to be the upper limit of the fraction range of the variable when N is so large that the lower limit of the fraction range of the variable plus N is not smaller than the upper limit of the fraction range, wherein M and N are positive integers respectively, and X is a number larger than zero; and

the conversion score unit sets, in at least one variable, a score of a client to a lower limit of a score range of the variable when raw data of the client is lower than an average of raw data of the clients by at least MX standard deviations, sets the score of the client to the lower limit of the score range of the variable plus N when the raw data of the client is lower than the average of raw data of the clients by between (M-N) standard deviations and (M- (N-1)) standard deviations, but sets the score of the client to the upper limit of the score range of the variable when N is so large that the lower limit of the score range of the variable plus N is not less than the upper limit of the score range, where M and N are positive integers and X is a number greater than zero, respectively.

8. The credit rating system of claim 1, further comprising at least one of:

the weight score block is predetermined in the weight used in the calculation of the credit score for the at least one variable;

the weight score block is configured to calculate a credit score based on the at least two variables using different predetermined weights;

the weight score block adjusts the weight used in calculating the credit score in at least one variable according to the relative distribution of the customers in the variable;

a weight score block that adjusts the relative proportions of the weights of the variables between at least two variables according to the relative proportions of the standard deviations of the customer's score distributions among the variables, where the greater the standard deviation of the customer's score distributions for any one variable, the greater the weight of that variable; and

the weight fraction block multiplies the weights of all the variables by an adjustment factor after adjusting the weights of the variables, so that the sum of the weights of the variables after adjustment is equal to the sum of the weights of the variables before adjustment.

9. The credit rating system of claim 2, further comprising:

the data storage unit, the conversion score unit and the weight score unit obtain the average value and the standard deviation of each variable of the clients at the starting moment;

the data storage unit receives and stores some special clients with relatively poor financial records of all the clients in the real world in a period of time after the starting time;

after the time, the conversion score unit and the weight score unit obtain the special average value and the special standard deviation of each variable of the special clients;

the weight score unit calculates the quality factor corresponding to each variable, wherein the quality factor corresponding to any variable is the difference between the special average value of the special customers and the average value of all the customers in the variable, and the difference is divided by the standard deviation of all the customers;

the weight fraction unit finds out the maximum quality factor value and the minimum quality factor value in the quality factors, and divides the value between the maximum quality factor value and the minimum quality factor value into K equal parts, wherein K is a positive integer; and

the weight fraction unit adjusts the weight of each variable to J when the quality factor corresponding to the variable is located in the J-th equal part, wherein J is a positive integer.

10. The system of claim 9, further comprising a weight score unit that sums the adjusted weights of the variables to obtain a sum, and divides the weight associated with each variable by the sum to obtain a re-adjusted weight.

11. The credit rating system of claim 2, further comprising:

the data storage unit and the conversion score unit obtain the average value and the standard deviation of the clients at each variable at the starting moment;

the data storage unit updates the stored public data and non-public data of at least one client at a period of time after the starting time;

the conversion score unit calculates the updated score of each customer in each variable according to the relative distribution of the customers in the updated data; and

the weight score unit calculates an update credit score and the like of each client according to the weight of each variable and the updated score of each client in each variable.

12. The credit rating system of claim 11, further comprising at least one of:

the conversion score unit updates the credit scores of all the clients and the like after at least R clients have data to be updated, wherein R is a positive integer; and

the weight score unit updates the credit score of each client and the like after at least S variables have data to be updated, wherein S is a positive integer.

13. A credit rating method, comprising:

storing public data and non-public data respectively related to any variable for evaluating credit rating and any client expected to make credit rating;

calculating the score of each customer in each variable according to the relative distribution of the data of the customers; and

and calculating the credit rating and the like of each client according to the weight of each variable and the score of each client in each variable.

14. The method of claim 13, further comprising at least one of:

after calculating the credit scores of the clients for a period of time, finding one or more variables with low credit performance client scores according to the score distribution difference of the credit performance part and the integral part of the clients in all variables during the period of time, and correspondingly adjusting the weights of the variables; and

after calculating the credit rating of the clients, the data stored in the data storage unit is periodically and irregularly updated, and the conversion score unit and the weight score unit are calculated again to obtain a new credit rating result.

15. The method of claim 13, wherein the different variables are related to the upper and lower limits of their respective score ranges, and the score of each customer at each variable is calculated based on the respective raw data of the customers in the stored data and the mean and standard deviation of the raw data of the customers.

16. The method of claim 15, further comprising at least one of:

in at least one variable, when the original data of a certain client and the average value of the original data of the clients are different from each other by no more than X standard deviations, the score of the certain client is the middle value of the score range of the variable, wherein X is a number larger than zero; and

in at least one variable, when the absolute value of the difference between the original data of a certain customer and the average value of the original data of the customers is greater than NX standard deviations but not greater than (N +1) X standard deviations, the absolute value of the difference between the score of the certain customer and the middle value of the score range of the variable is different from N, but when N is not less than half of the score range of the variable, the absolute value of the difference between the score of the certain customer and the middle value of the score range of the variable is set to be half of the score range, wherein N is a positive integer and X is a number greater than zero.

17. The method of claim 16, further comprising at least one of:

in at least one variable, the larger the raw data of any customer is, the closer the score of the customer is to the upper limit of the score range of the variable, and the smaller the raw data of the customer is, the closer the score of the customer is to the lower limit of the score range of the variable; and

in at least one variable, the larger the raw data of any customer is than the average value of the raw data of the customers, the closer the score of the customer is to the lower limit of the score range of the variable, and the smaller the raw data of the customers is, the closer the score of the customer is to the upper limit of the score range of the variable.

18. The method of claim 15, further comprising at least one of:

at least one variable, when the original data of a client is higher than the average value of the original data of the clients by at least MX standard deviations, the score of the client is the lower limit of the score range of the variable, when the original data of the client is higher than the average value of the original data of the clients by between (M-N) standard deviations and (M- (N-1)) standard deviations, the score of the client is the lower limit of the score range of the variable plus N, but when N is so large that the lower limit of the score range of the variable plus N is not less than the upper limit of the score range, the score of the client is the upper limit of the score range of the variable, wherein M and N are both positive integers respectively, and X is a number greater than zero; and

in at least one variable, when the raw data of a client is lower than the average value of the raw data of the clients by at least MX standard deviations, the score of the client is the lower limit of the score range of the variable, when the raw data of the client is lower than the average value of the raw data of the clients by between (M-N) standard deviations and (M- (N-1)) standard deviations, the score of the client is the lower limit of the score range of the variable plus N, but when N is so large that the lower limit of the score range of the variable plus N is not less than the upper limit of the score range, the score of the client is the upper limit of the score range of the variable, wherein M and N are both positive integers respectively, and X is a number greater than zero.

19. The method of claim 13, further comprising at least one of:

the weight used in calculating the credit rating or the like in at least one variable is predetermined;

the predetermined weights used in the calculation of the credit rating and the like for the at least two variables are not the same; and

the weight used in calculating the credit rating in at least one variable is adjusted according to the relative distribution of the customers in that variable.

20. The method of claim 19, further comprising at least one of:

adjusting the relative proportions of the weights of the variables between at least two variables according to the relative proportions of the standard deviations of the customer's fractional distributions among the variables, wherein the greater the standard deviation of the customer's fractional distributions for any one variable, the greater the weight of that variable; and

after the weights of the variables are adjusted, the weights of all the variables are multiplied by an adjusting multiple, so that the sum of the weights of the variables after adjustment is equal to the sum of the weights of the variables before adjustment.

21. The method of credit rating according to claim 14, further comprising:

obtaining the average value and the standard deviation of each variable of the clients at a starting moment;

selecting special clients with poor credit records according to the performance of all the clients in the real world in a period of time after the starting time;

after this time, the particular average and the particular standard deviation of the particular customers at each variable are obtained;

calculating a quality factor corresponding to each variable, wherein the quality factor corresponding to any variable is the difference between the special average value of the special customers and the average value of all customers in the variable, and the standard deviation of all customers is divided;

finding out the maximum quality factor value and the minimum quality factor value in the quality factors, and dividing the value between the maximum quality factor value and the minimum quality factor value into K equal parts, wherein K is a positive integer; and

in each variable, when the corresponding quality factor is located in the J-th equal part, the weight corresponding to the variable is adjusted to J, wherein J is a positive integer.

22. The method of claim 21, further comprising summing the adjusted weights of the variables to obtain a sum, and dividing the weight associated with each variable by the sum to obtain a re-adjusted weight.

23. The method of credit rating according to claim 14, further comprising:

updating the stored public data and non-public data of at least one client at a time after the starting time;

calculating the updated score of each customer in each variable according to the relative distribution of the customers in the updated data; and

and calculating the update credit score and the like of each client according to the weight of each variable and the updated score of each client in each variable.

24. The method of claim 23, further comprising at least one of:

updating the credit rating of each client after at least R clients have data to be updated, wherein R is a positive integer; and

the credit rating of each customer is updated only after at least S variables, where S is a positive integer, have data updated.