CN115601159A - Method for rating credit of main body of civil enterprise - Google Patents
Method for rating credit of main body of civil enterprise Download PDFInfo
- Publication number
- CN115601159A CN115601159A CN202211495732.3A CN202211495732A CN115601159A CN 115601159 A CN115601159 A CN 115601159A CN 202211495732 A CN202211495732 A CN 202211495732A CN 115601159 A CN115601159 A CN 115601159A
- Authority
- CN
- China
- Prior art keywords
- data
- credit
- enterprise
- rating
- civil
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000012216 screening Methods 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 8
- 239000003795 chemical substances by application Substances 0.000 claims description 63
- 239000013598 vector Substances 0.000 claims description 60
- 238000006243 chemical reaction Methods 0.000 claims description 39
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000011176 pooling Methods 0.000 claims description 16
- 238000013507 mapping Methods 0.000 claims description 15
- 238000013519 translation Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 206010037660 Pyrexia Diseases 0.000 claims 1
- 230000008569 process Effects 0.000 description 3
- 230000002349 favourable effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for grading the credit of a main body of a civil-camp enterprise, belonging to the technical field of credit grading, and the method comprises the steps of constructing a credit data set of the civil-camp enterprise; screening a data set; establishing an agent rating model; training an agent rating model; evaluating the quality of the agent rating model; and collecting data items of the civil enterprise to be evaluated, screening key data items, and inputting the key data items into the agent rating model to obtain a credit rating result. According to the method, a credit data set of the civil-operated enterprise is constructed, the data set is screened, an agent rating model is established by adopting the screened data set, the agent rating model is trained and evaluated in quality, the agent rating model with high accuracy is obtained, meanwhile, data items of the civil-operated enterprise to be evaluated are collected, key data items are screened out according to the importance of each data item, the key data items are input into the obtained agent rating model, a credit rating result is output, and credit rating based on partial observable data of the civil-operated enterprise to be evaluated can be achieved.
Description
Technical Field
The invention relates to credit the field of rating techniques and the like, in particular to a method for grading the credit of a main body of a civil enterprise.
Background
The credit rating, also called credit rating, is a social intermediary service, has positive effects on investors, financial institutions, financial markets, enterprises and supervisory institutions, provides a simple and objective indication for credit risks, reduces market information cost, is favorable for pricing of financial products, can be used as a basis for signing contracts and developing services in daily operation of enterprises, and is finally favorable for the sound and perfect completion of the whole market supervision mechanism.
The lack of credit rating of the civil-operating enterprise leads to difficulty in helping the enterprise gain more market opportunities, widening financing channels and improving the management level through the credit rating, and also cannot assist the financial market to comprehensively grasp the credit situation of the civil-operating enterprise, optimize the ecological rating and promote the high-quality development of the rating business.
Disclosure of Invention
The invention aims to provide a method for rating credit of a main body of a civil enterprise to solve the defects in the prior art, and the technical problem to be solved by the invention is realized by the following technical scheme.
The method disclosed by the invention comprises the following steps:
extracting credit rating results of the civil-service enterprises which have entered the credit rating market and publicly disclosed various types of data to construct a credit data set S of the civil-service enterprises;
screening data item sets corresponding to various types of publicly disclosed data through related domain knowledge rules or domain experts, and acquiring screened data sets according to the screened data item sets and mapping function sets corresponding to the screened data item sets
；
The screened data set
The related data in (1) is subjected to correlation operation to obtain an embedded layer defined by a formula, and the embedded layer is fusedAcquiring a pooling layer defined by a formula according to the maximum value of each column vector in a vector matrix obtained by combining related data in an embedding layer, and acquiring a conversion layer defined by the formula by carrying out polynomial operation on the vector output by the pooling layer;
sequentially overlapping an embedded layer defined by a formula, a pooling layer defined by the formula and a conversion layer defined by the formula to establish a proxy rating model;
measuring the difference between the enterprise credit level in the screened data set and the rating level output by the agent rating model according to a cross entropy loss function, and training the agent rating model by continuously adjusting the parameters of the agent rating model by adopting an Adam algorithm so as to minimize the difference;
building and screening a post-data set
Including different data sets of civil and commercial enterprises
Obtaining a data set via a proxy rating model
The credit rating set obtained by the agent rating model and the data set
Comparing the credit rating sets contained in the agent rating model, judging whether the quality of the agent rating model meets the requirements, and when the quality of the agent rating model does not meet the requirements, reestablishing and training the agent rating model until the obtained agent rating model meets the quality requirements;
the method comprises the steps of collecting data items of the to-be-evaluated civil enterprises, screening out key data items according to the importance of the data items, inputting the key data items into an agent rating model which is trained and meets quality requirements, and taking enterprise credit levels corresponding to the maximum values of elements in output vectors of the agent rating model as credit rating results of the to-be-evaluated civil enterprises.
In the above-described aspect of the present invention,the credit data set of the civil and commercial enterprises is expressed as S =<O，A，V，L>Wherein O = { O = 1 , o 2 , …，o N Denotes a set of N private enterprises that have entered the credit rating market, o i A civil enterprise indicating that the ith household has entered the credit rating market; a = { a = 1 , a 2 ,…，a D D represents a set of data items corresponding to each type of publicly disclosed data;
representing a set of functions, mapping an enterprise to data values corresponding to data items
Assigning a private enterprise-specific data value that has entered the credit rating market; l = { L 1， l 2， …，l M Denotes a set of credit levels corresponding to the M credit rating results, by a rating function
A private business that has entered the credit rating market is given a particular credit rating.
In the above scheme, the filtered data set
Expressed as:
wherein O = { O = 1 , o 2 , …，o N Denotes a set of N private enterprises that have entered the credit rating market, o i A civil enterprise indicating that the ith household has entered the credit rating market;
a set of data items representing a filter, wherein,
for the selected data item orderSet of symbols, J = {1,2, …, D is a set of D data item sequence numbers corresponding to a set A of data items corresponding to each type of publicly disclosed data, wherein A = { a = { (a) } 1 , a 2 ,…，a D }；
Set of data items to be filtered
A corresponding set of mapping functions; l = { L 1， l 2， …，l M Expressing the set of credit levels corresponding to the M credit rating results through a rating function
A private business that has entered the credit rating market is given a particular credit rating.
In the above scheme, the filtered data set
Performing a correlation operation on the correlation data to obtain an embedding layer defined by a formula comprises:
acquiring a data set
Business in o i Corresponding j-th data item corresponds to a data value v j (o i ) The data type of (2);
when the data value v j (o i ) Is classified data, and the data value v is expressed by a first formula j (o i ) Conversion into vectors
Wherein the first formula is:
wherein, in the step (A),
is a mapping matrix corresponding to the jth data item, d j The number of the type values contained in the j-th dimension data, k is the dimension of the embedding layer vector space,
to map the categorical data as d j A function of the Wikstro representation, reLU (-) being an activation function;
when the data value v j (o i ) Is numerical data, and the data value v is obtained by a second formula j (o i ) Conversion into vectors
Wherein the second formula is:
wherein, in the process,
is a mapping vector corresponding to the jth data item, k being the dimension of the embedding layer vector space;
to enterprise o i The vector converted by the data value corresponding to each data item of (1) forms a vector matrix X (i) Wherein X is (i) The expression formula of (a) is:
，X (i) is enterprise o i Corresponding j-th data item corresponding data value v j (o i ) The vector of the translation is then converted into a vector,
representing a data set
The number of data items in (a);
by vector matrix X (i) Obtaining an embedding layer defined by a formula
And H is a vector set of the category type data conversion, and m is a vector set of the numerical type data conversion.
In the above solution, the pooling layer defined by the formula is:
wherein, in the step (A),
representing a vector matrix X (i) Maximum value in j-th column.
In the above solution, the formula-defined translation layer is
Wherein, in the step (A),
for converting the vector x output by the pooling layer as a conversion function of the ith layer of the conversion layer, A i A conversion matrix for converting the i-th layer, b i And z represents the number of layers of the conversion layer, A is a conversion matrix set of the conversion layer, and b is a bias vector set of the conversion layer. In the above scheme, the agent rating model is
Wherein, in the step (A),θthe parameter set of the proxy rating model is a parameter set of the proxy rating model, which includes a, b, H, and m.
In the above solution, the data set obtained by the agent rating model
The credit rating set obtained by the agent rating model and the data set
The credit rating contained inComparing, and judging whether the quality of the agent rating model meets the requirement or not comprises the following steps:
calculating to obtain data set by using agent rating model
The credit levels of each enterprise form an enterprise credit level set L q ；
Item-by-item comparison enterprise credit level set L q And data set
The enterprise credit level set L contained in the system is calculated to obtain the enterprise credit level set L q The number of enterprises occupying the data set and having the same credit level as the credit level of the enterprise in the credit level set L
The proportion of the total number of middle enterprises;
and comparing the calculated proportion with the set expected proportion, wherein if the calculated proportion exceeds the set expected proportion, the quality of the proxy rating model meets the requirement, and if the calculated proportion is less than or equal to the set expected proportion, the quality of the proxy rating model does not meet the requirement.
In the above scheme, the collecting data items of the civil enterprise to be evaluated, and screening out key data items according to the importance of each data item includes:
inputting all data items of the civil enterprises to be evaluated into an agent rating model which is trained and meets the quality requirement for credit rating;
after the values of the first data items in all the civil enterprises to be evaluated are set to be missing, inputting the missing values into an agent rating model which is trained and meets quality requirements to carry out credit rating again, wherein k =1,2, …, Y and Y are the total number of the data items of the civil enterprises to be evaluated;
calculating the importance of the kth data item in the data items of all the civil enterprises to be evaluated through an importance calculation formula;
constructing an importance set of each data item according to the importance of each data item in the data items of all the civil enterprises to be evaluated;
and calculating the mean value of the importance set and the variance of the importance set, calculating the difference value of the importance of each data item and the mean value of the importance set, and taking the data item corresponding to the variance of which the difference value is less than three times the importance set as a key data item.
In the above scheme, the importance calculation formula is
N is the total number of the civil enterprises to be evaluated,
in order to input the data item of the civil enterprise to be evaluated into the agent rating model to obtain an output value and set the value of the kth data item of the civil enterprise to be evaluated as the square of the output value difference obtained by inputting the missing data item into the agent rating model, the specific formula is as follows:
The embodiment of the invention has the following advantages:
the method for grading the credit of the main body of the civil-camp enterprise solves the problem of grading the credit of most of the civil-camp enterprises which do not enter the credit grading market, and comprises the steps of constructing a credit data set of the civil-camp enterprise,
Screening a data set, establishing an agent rating model through the screened data set, training and evaluating the quality of the agent rating model, acquiring the agent rating model with high accuracy, acquiring data items of the to-be-evaluated civil enterprises, screening out key data items according to the importance of each data item, inputting the key data items into the acquired agent rating model, and outputting a credit rating result, so that credit rating can be performed on the basis of partial observable data of the to-be-evaluated civil enterprises.
Drawings
FIG. 1 is a flow chart of steps of an embodiment of a method for rating credit of a subject of a private enterprise of the present invention;
FIG. 2 is a flow chart of steps of an embodiment of the present invention to obtain an embedding layer defined by a formula.
FIG. 3 is a flow chart of the steps of the present invention for determining whether the quality of the agent rating model meets the requirements.
FIG. 4 is a flowchart of the steps of the present invention for screening key data items.
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present invention will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
As shown in fig. 1, the method for rating credit of a main body of a civil enterprise provided by the present invention includes:
step S1: extracting credit rating results of the civil-service enterprises which have entered the credit rating market and publicly disclosed various types of data to construct a credit data set S of the civil-service enterprises;
step S2: screening the data item set corresponding to each type of publicly disclosed data through related domain knowledge rules or domain experts, and acquiring the screened data set according to the screened data item set and the mapping function set corresponding to the screened data item set
；
And step S3: the screened data set
The relevant data in the data processing system is subjected to relevant operation to obtain an embedding layer defined by a formula, the maximum value of each column vector in a vector matrix obtained by fusing the relevant data in the embedding layer is fused to obtain a pooling layer defined by the formula, and a vector output by the pooling layer is subjected to multiple operations to obtain a conversion layer defined by the formula;
and step S4: sequentially overlapping an embedded layer defined by a formula, a pooling layer defined by the formula and a conversion layer defined by the formula to establish a proxy rating model;
step S5: measuring the screened data set according to the cross entropy loss function
Training the agent rating model by continuously adjusting the parameters of the agent rating model by adopting an Adam algorithm according to the difference between the credit level of the enterprise and the rating level output by the agent rating model, so that the difference is minimized;
step S6: building and screening a post-data set
Including different data sets of civil and commercial enterprises
Obtaining a data set via a proxy rating model
The credit rating set obtained by the agent rating model and the data set
Comparing the credit rating sets contained in the agent rating model, judging whether the quality of the agent rating model meets the requirements, and when the quality of the agent rating model does not meet the requirements, reestablishing and training the agent rating model until the obtained agent rating model meets the quality requirements;
step S7: collecting data items of the to-be-evaluated civil enterprise, screening out key data items according to the importance of each data item, inputting the key data items into an agent rating model which is trained and meets quality requirements, and taking an enterprise credit grade corresponding to the maximum value of elements in an output vector of the agent rating model as a credit rating result of the to-be-evaluated civil enterprise.
In this embodiment, in step S2, different domain knowledge rules may be selected according to different emphasis dimensions of credit rating in a specific business, or different domain experts may be requested to manually perform screening, and when a data item set corresponding to each type of publicly disclosed data is screened by using related domain knowledge rules or domain experts, a data item set closely related to enterprise credit is screened by using the introduced related domain knowledge rules or domain experts, where the related domain knowledge rules include a credit index system and the like.
In the embodiment, the credit data set S of the civil and commercial enterprises is represented as S =<O，A，V，L>Wherein O = { O = 1 , o 2 , …，o N Denotes a set of N private enterprises that have entered the credit rating market, o i A civil enterprise indicating that the ith household has entered the credit rating market; a = { a = 1 , a 2 ,…，a D D represents a set of data items corresponding to each type of publicly disclosed data;
representing a set of functions, mapping an enterprise to data values corresponding to data items
Assigning a private enterprise-specific data value that has entered the credit rating market; l = { L 1， l 2， …，l M Expressing the set of credit levels corresponding to the M credit rating results through a rating function
A private business that has entered the credit rating market is given a particular credit rating.
In this embodiment, the filtered data set
Expressed as:
wherein O = { O = 1 , o 2 , …，o N Denotes a set of N private enterprises that have entered the credit rating market, o i A civil enterprise indicating that the ith household has entered the credit rating market;
a set of data items representing a filter, wherein,
for the set of screened data item sequence numbers, J = {1,2, …, D } is a set of data item sequence numbers corresponding to D sets A of publicly disclosed data items corresponding to each type of data, where A = { a = 1 , a 2 ,…，a D }；
For a set of filtered data items
A corresponding set of mapping functions; l = { L 1， l 2， …，l M Denotes a set of credit levels corresponding to the M credit rating results, by a rating function
A private business that has entered the credit rating market is given a particular credit rating.
In the present embodiment, the collection is performed by the data items
The screening of the method realizes knowledge driving,
this allows the number of data items to be reduced as much as possible while meeting the credit rating information requirements.
As shown in FIG. 2, the filtered data set
InThe related data is subjected to related operation to obtain an embedded layer defined by a formula, and the method comprises the following steps:
step S311: acquiring a data set
Business in o i Corresponding j-th data item corresponding data value v j (o i ) The data type of (d);
step S312: when the data value v j (o i ) Is classified data, and the data value v is expressed by a first formula j (o i ) Conversion into vectors
Wherein the first formula is:
wherein, in the step (A),
is a mapping matrix corresponding to the jth data item, d j The number of the type values contained in the j-th dimension data, k is the dimension of the embedding layer vector space,
to map the categorical data as d j A function of the Wikstro representation, reLU (-) being an activation function; when the data value v j (o i ) Is numerical data, and the data value v is obtained by a second formula j (o i ) Conversion into vectors
Wherein the second formula is:
wherein, in the step (A),
is the mapping vector corresponding to the jth data item, k is the dimension of the embedding layer vector space;
step S313: to enterprise o i The vector converted by the data value corresponding to each data item of (1) forms a vector matrix X (i) Wherein X is (i) The expression formula of (a) is:
step S314: by vector matrix X (i) Obtaining an embedding layer defined by a formula
H is a vector set of category type data conversion, m is a vector set of numerical type data conversion, and H and m are parameters to be learned of the agent rating model.
In this embodiment, reLU (-) converts vector values smaller than 0 in the vector into 0, and outputs vector values larger than 0 as they are.
In this embodiment, the pooling layer defined by the formula is:
In the present embodimentThe formula-defined translation layer is
Wherein, in the step (A),
for converting the vector x output by the pooling layer as a conversion function of the ith layer of the conversion layer, A i A conversion matrix for converting the i-th layer, b i And z represents the number of layers of the conversion layer, A is a conversion matrix set of the conversion layer, and b is a bias vector set of the conversion layer.
In the embodiment, the translation layer is formed by adopting a plurality of layers of fully-connected neural networks, and each layer of the network adopts ReLU (-) as an activation function of the neuron.
In this embodiment, the agent rating model is
And theta is a parameter set of the proxy rating model, and the parameter set of the proxy rating model comprises A, b, H and m.
As shown in FIG. 3, the data set obtained by the proxy rating model
The credit rating set obtained by the agent rating model and the data set
The step of comparing the credit rating sets contained in the data base and judging whether the quality of the agent rating model meets the requirement comprises the following steps:
step S61: calculating to obtain data set by using agent rating model
The credit levels of each enterprise form an enterprise credit level set L q ；
Step S62: item-by-item comparison enterprise credit level set L q And data set
The enterprise credit level set L contained in the system is calculated to obtain the enterprise credit level set L q The number of enterprises occupying the data set and having the same credit level as the credit level of the enterprise in the credit level set L
The proportion of the total number of middle enterprises;
step S63: and comparing the calculated proportion with the set expected proportion, wherein if the calculated proportion exceeds the set expected proportion, the quality of the proxy rating model meets the requirement, and if the calculated proportion is less than or equal to the set expected proportion, the quality of the proxy rating model does not meet the requirement.
As shown in fig. 4, the collecting data items of the civil enterprise to be evaluated, and screening out key data items according to the importance of each data item includes:
step S711: inputting all data items of the civil enterprises to be evaluated into an agent rating model which is trained and meets the quality requirement for credit rating;
step S712: after the value of the kth data item in all the to-be-evaluated civil enterprises is set to be missing, inputting the missing value into an agent rating model which is trained and meets the quality requirement to perform credit rating again, wherein k =1,2, …, and Y are the total number of the data items of the to-be-evaluated civil enterprises;
step S713: calculating the importance of the first data item in the data items of all the civil enterprises to be evaluated through an importance calculation formula;
step S714: constructing an importance set of each data item according to the importance of each data item in the data items of all the civil enterprises to be evaluated;
step S715: and calculating the mean value of the importance set and the variance of the importance set, calculating the difference value between the importance of each data item and the mean value of the importance set, and taking the data item corresponding to the variance of which the difference value is less than three times the importance set as a key data item.
In the embodiment, data driving is realized by screening key data items, the credit rating of an enterprise can be acquired based on partial observable data, and the credit rating problem of a civil enterprise under the condition of missing data items is solved.
In this embodiment, the importance calculation formula is
N is the total number of the civil enterprises to be evaluated,
in order to set the value of the data item of the civil service enterprise to be evaluated into the square of the output value difference obtained by inputting the data item of the civil service enterprise to be evaluated into the proxy rating model and the value of the first data item of the civil service enterprise to be evaluated into the proxy rating model after being deleted, the specific formula is as follows:
The types of data disclosed by the publication include, but are not limited to, financial data, basic business data, capital market data.
It should be noted that the above detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular is intended to include the plural unless the context clearly dictates otherwise. Furthermore, it will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in other sequences than described of illustrated herein.
Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements explicitly listed, but may include other steps or elements not explicitly listed or inherent to such process, method, article, or apparatus.
For ease of description, spatially relative terms such as "over … …", "over … …", "over … …", "over", etc. may be used herein to describe the spatial positional relationship of one device or feature to another device or feature as shown in the figures. It will be understood that the spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if a device in the figures is turned over, devices described as "above" or "on" other devices or configurations would then be oriented "below" or "under" the other devices or configurations. Thus, the exemplary term "above … …" may include both orientations of "above … …" and "below … …". The device may also be oriented in other different ways, such as by rotating it 90 degrees or at other orientations, and the spatially relative descriptors used herein interpreted accordingly.
In the foregoing detailed description, reference is made to the accompanying drawings, which form a part hereof. In the drawings, like numerals typically identify like components, unless context dictates otherwise. The illustrated embodiments described in the detailed description, drawings, and claims are not meant to be limiting. Other embodiments may be utilized, and other changes may be made, without departing from the spirit or scope of the subject matter presented here.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for rating credit of a subject of a civil enterprise, the method comprising:
extracting credit rating results of the civil-service enterprises which have entered the credit rating market and publicly disclosed various types of data to construct a credit data set S of the civil-service enterprises;
screening data item sets corresponding to various types of publicly disclosed data through related domain knowledge rules or domain experts, and acquiring screened data sets according to the screened data item sets and mapping function sets corresponding to the screened data item sets
；
The screened data set
The relevant data in the data processing system is subjected to relevant operation to obtain an embedded layer defined by a formula, the maximum value of each column vector in a vector matrix obtained by fusing the relevant data in the embedded layer is obtained to obtain a pooling layer defined by the formula, and a vector output by the pooling layer is subjected to multiple operations to obtain a conversion layer defined by the formula;
sequentially overlapping an embedded layer defined by a formula, a pooling layer defined by the formula and a conversion layer defined by the formula to establish a proxy rating model;
measuring the screened data set according to the cross entropy loss function
According to the difference between the enterprise credit level and the rating level output by the agent rating model, the Adam algorithm is adopted to continuously adjust the parameters of the agent rating model to train the agent rating model, so that the difference is minimized;
building and screening a post-data set
Including different data sets of civil and commercial enterprises
Obtaining a data set via a proxy rating model
The credit rating set obtained by the agent rating model and the data set
Comparing the credit rating sets contained in the agent rating model, judging whether the quality of the agent rating model meets the requirements, and when the quality of the agent rating model does not meet the requirements, reestablishing and training the agent rating model until the obtained agent rating model meets the quality requirements;
collecting data items of the to-be-evaluated civil enterprise, screening out key data items according to the importance of each data item, inputting the key data items into an agent rating model which is trained and meets quality requirements, and taking an enterprise credit grade corresponding to the maximum value of elements in an output vector of the agent rating model as a credit rating result of the to-be-evaluated civil enterprise.
2. The method of claim 1, wherein the credit data set S of the civil-camp enterprise is S =<O，A，V，L>Wherein O = { O = 1 , o 2 , …，o N Denotes a set of N private enterprises that have entered the credit rating market, o i A civil enterprise indicating that the ith household has entered the credit rating market; a = { a = 1 , a 2 ,…，a D D represents a set of data items corresponding to each type of publicly disclosed data;
representing a set of functions, mapping an enterprise to data values corresponding to data items
Assigning a private enterprise-specific data value that has entered the credit rating market; l = { L 1， l 2， …，l M Denotes a set of credit levels corresponding to the M credit rating results, by a rating function
A private business that has entered the credit rating market is given a particular credit rating.
3. The method of claim 1, wherein the screened data set is used as a credit rating measure for the subject of a private enterprise
Expressed as:
wherein O = { O = 1 , o 2 , …，o N Denotes a set of N private enterprises that have entered the credit rating market, o i A civil enterprise indicating that the ith household has entered the credit rating market;
a set of data items representing a filter, wherein,
for the set of screened data item sequence numbers, J = {1,2, …, D } is a set of data item sequence numbers corresponding to a set a of D publicly disclosed data items of each type, where a = { a = 1 , a 2 ,…，a D }；
Set of data items to be filtered
A corresponding set of mapping functions; l = { L 1， l 2， …，l M Denotes a set of credit levels corresponding to the M credit rating results, by a rating function
A private business that has entered the credit rating market is given a particular credit rating.
4. The method of claim 1, wherein the filtered data set is used to rate credit for the subject of a private enterprise
Performing a correlation operation on the correlation data to obtain an embedding layer defined by a formula comprises:
acquiring a data set
Business in o i Corresponding j-th data item corresponds to a data value v j (o i ) The data type of (2);
when the data value v j (o i ) Is classified data, and the data value v is expressed by a first formula j (o i ) Conversion into vectors
Wherein the first formula is:
wherein, in the step (A),
is a mapping matrix corresponding to the jth data item, d j The number of the type values contained in the j-th dimension data, k is the dimension of the embedding layer vector space,
to map the categorical data as d j A function represented by Uygur fever, reLU (-) being an activation function;
when the data value v j (o i ) Is numerical data, and the data value v is obtained by a second formula j (o i ) Conversion into vectors
Wherein the second formula is:
wherein, in the step (A),
is the mapping vector corresponding to the jth data item, k is the dimension of the embedding layer vector space;
to enterprise o i The vector converted by the data value corresponding to each data item of (1) forms a vector matrix X (i) Wherein X is (i) The expression formula of (a) is:
，X (i) is enterprise o i Corresponding j-th data item corresponds to a data value v j (o i ) The vector of the translation is then converted into a vector,
representing a data set
The number of data items in (a);
by vector matrix X (i) Obtaining an embedding layer defined by a formula
And H is a vector set of the category type data conversion, and m is a vector set of the numerical type data conversion.
5. The method of claim 1, wherein the formula-defined pooling level is
Wherein, in the step (A),
representing a vector matrix X (i) Maximum value in j-th column.
6. The method of claim 1, wherein the formula-defined translation layer is
Wherein, in the step (A),
for converting the vector x output by the pooling layer as a conversion function of the ith layer of the conversion layer, A i A conversion matrix for converting the ith layer, b i And z represents the number of layers of the conversion layer, A is a conversion matrix set of the conversion layer, and b is a bias vector set of the conversion layer.
7. The method of claim 1, wherein the agent rating model is a credit rating model
Wherein, in the step (A),θthe parameter set of the proxy rating model is a parameter set of the proxy rating model comprising a, b, H and m.
8. The method of claim 1, wherein the data set is obtained via an agent rating model
The credit rating set obtained by the agent rating model and the data set
Comparing the credit level sets contained in the data base, and judging whether the quality of the agent rating model meets the requirement or not comprises the following steps:
calculating to obtain data set by using agent rating model
The credit levels of each enterprise form an enterprise credit level set L q ；
Item-by-item comparison enterprise credit level set L q And data set
The enterprise credit level set L contained in the system is calculated to obtain the enterprise credit level set L q The number of enterprises occupying the data set and having the same credit level as the credit level of the enterprise in the credit level set L
The proportion of the total number of middle enterprises;
and comparing the calculated proportion with the set expected proportion, wherein if the calculated proportion exceeds the set expected proportion, the quality of the proxy rating model meets the requirement, and if the calculated proportion is less than or equal to the set expected proportion, the quality of the proxy rating model does not meet the requirement.
9. The method of claim 1, wherein the credit rating of the subject of the civil enterprise,
the method for acquiring the data items of the civil enterprise to be evaluated and screening out the key data items according to the importance of each data item comprises the following steps:
inputting all data items of the civil and private enterprises to be evaluated into an agent rating model which completes training and meets quality requirements for credit rating;
setting the value of the kth data item in all the to-be-evaluated civil enterprises to be deleted, and inputting the kth data item to an agent rating model which is trained and meets the quality requirement to perform credit rating again, wherein k =1,2, …, Y and Y are the total number of the data items of the to-be-evaluated civil enterprises;
calculating the importance of the kth data item in the data items of all the civil enterprises to be evaluated through an importance calculation formula;
constructing an importance set of each data item according to the importance of each data item in the data items of all the civil enterprises to be evaluated;
and calculating the mean value of the importance set and the variance of the importance set, calculating the difference value of the importance of each data item and the mean value of the importance set, and taking the data item corresponding to the variance of which the difference value is less than three times the importance set as a key data item.
10. The method of claim 9, wherein the importance calculation formula is
N is the total number of the civil enterprises to be evaluated,
in order to input the data item of the civil enterprise to be evaluated into the agent rating model to obtain an output value and set the value of the kth data item of the civil enterprise to be evaluated as the square of the output value difference obtained by inputting the missing data item into the agent rating model, the specific formula is as follows:
and setting the value of the kth data item of the civil enterprise to be evaluated as the missing enterprise sample data, wherein the kth data item is the total number of the credit levels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211495732.3A CN115601159A (en) | 2022-11-28 | 2022-11-28 | Method for rating credit of main body of civil enterprise |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211495732.3A CN115601159A (en) | 2022-11-28 | 2022-11-28 | Method for rating credit of main body of civil enterprise |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115601159A true CN115601159A (en) | 2023-01-13 |
Family
ID=84852746
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211495732.3A Pending CN115601159A (en) | 2022-11-28 | 2022-11-28 | Method for rating credit of main body of civil enterprise |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115601159A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115511506A (en) * | 2022-09-30 | 2022-12-23 | 中国电子科技集团公司第十五研究所 | Enterprise credit rating method, device, terminal equipment and storage medium |
-
2022
- 2022-11-28 CN CN202211495732.3A patent/CN115601159A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115511506A (en) * | 2022-09-30 | 2022-12-23 | 中国电子科技集团公司第十五研究所 | Enterprise credit rating method, device, terminal equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kozeny | Genetic algorithms for credit scoring: Alternative fitness function performance comparison | |
| Pal et al. | Development of a hybrid methodology for dimensionality reduction in Mahalanobis–Taguchi system using Mahalanobis distance and binary particle swarm optimization | |
| CN113256636B (en) | Bottom-up parasite species development stage and image pixel classification method | |
| CN103065042B (en) | Multiple target comprehensive decision evaluation method based on sight | |
| CN112085252B (en) | Anti-fact prediction method for set type decision effect | |
| CN108874959A (en) | A kind of user's dynamic interest model method for building up based on big data technology | |
| CN113407864B (en) | Group recommendation method based on mixed attention network | |
| CN109523021A (en) | A kind of dynamic network Structure Prediction Methods based on long memory network in short-term | |
| CN110110372B (en) | Automatic segmentation prediction method for user time sequence behavior | |
| CN115601159A (en) | Method for rating credit of main body of civil enterprise | |
| CN115270004B (en) | Educational resource recommendation method based on field factor decomposition | |
| CN110930038A (en) | Loan demand identification method, loan demand identification device, loan demand identification terminal and loan demand identification storage medium | |
| CN104077493A (en) | Method for constructing state evaluation index system of electric relaying protection system | |
| CN114757432A (en) | Future execution activity and time prediction method and system based on flow log and multi-task learning | |
| Hakimpoor et al. | Management information systems, conceptual dimensions of information quality and quality of managerial decisions: modelling artificial neural networks | |
| CN110032551A (en) | Model generates system, method and forecasting system | |
| CN112633556A (en) | Short-term power load prediction method based on hybrid model | |
| CN116861924A (en) | Project risk early warning method and system based on artificial intelligence | |
| CN117036760A (en) | Multi-view clustering model implementation method based on graph comparison learning | |
| CN115660293A (en) | Comprehensive evaluation method for complete life cycle of complex electromechanical product based on digital twinning | |
| CN108846489A (en) | A kind of large scale equipment health Evaluation method based on SVM | |
| CN117575328A (en) | Industrial chain risk assessment method and device based on graph attention neural network | |
| Shankar et al. | Analyzing attrition and performance of an employee using machine learning techniques | |
| CN111428861A (en) | Manufacturing enterprise overseas investment site selection decision method based on multi-scale multi-dimensional proximity | |
| CN117058882A (en) | Traffic data compensation method based on multi-feature double-discriminant |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230113 |