CN109636244A - Enterprise's Rating Model construction method, enterprise's methods of marking and device - Google Patents
Enterprise's Rating Model construction method, enterprise's methods of marking and device Download PDFInfo
- Publication number
- CN109636244A CN109636244A CN201910012442.0A CN201910012442A CN109636244A CN 109636244 A CN109636244 A CN 109636244A CN 201910012442 A CN201910012442 A CN 201910012442A CN 109636244 A CN109636244 A CN 109636244A
- Authority
- CN
- China
- Prior art keywords
- variance
- weight
- judgment matrix
- modular character
- enter
- 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 49
- 238000010276 construction Methods 0.000 title claims abstract description 30
- 239000011159 matrix material Substances 0.000 claims abstract description 220
- 238000013480 data collection Methods 0.000 claims abstract description 72
- 230000009467 reduction Effects 0.000 claims abstract description 12
- 238000004364 calculation method Methods 0.000 claims description 60
- 238000013507 mapping Methods 0.000 claims description 46
- 238000012512 characterization method Methods 0.000 claims description 27
- 239000012141 concentrate Substances 0.000 claims description 27
- 238000012545 processing Methods 0.000 claims description 20
- 238000012790 confirmation Methods 0.000 claims description 19
- 230000008859 change Effects 0.000 claims description 16
- 238000004140 cleaning Methods 0.000 claims description 14
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 claims description 9
- 230000029305 taxis Effects 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 5
- 239000004744 fabric Substances 0.000 claims description 3
- 230000006870 function Effects 0.000 description 24
- 230000008569 process Effects 0.000 description 13
- 230000008901 benefit Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
Abstract
The present invention provides a kind of enterprise's Rating Model construction method, enterprise's methods of marking and devices, enterprise's Rating Model construction method can include: obtain the representation data and management data of type of industry, dimension-reduction treatment is carried out to representation data and management data, from the result of dimensionality reduction, select into modular character;Classify to modular character is entered, constructs class index;For class index, execute: the important relationship entered between modular character for including according to class index determines judgment matrix;Judgment matrix is normalized, determines the corresponding judgment matrix weighted data collection of judgment matrix;Using variance method, the corresponding variance weighted data collection of class index is calculated;According to judgment matrix weighted data collection and variance weighted data collection, calculate into modular character weight.Scheme provided by the invention, which greatly reduces human interference or subjective factor, to be influenced, to effectively improve scoring accuracy.
Description
Technical field
The present invention relates to big data analysis technical field, in particular to a kind of enterprise's Rating Model construction method, enterprise are commented
Divide method and apparatus.
Background technique
Enterprise is that objective channel is obtained in the very important front end of such as bank, financial institution, is carried out using big data technology to enterprise comprehensive
Scoring is closed, financial institution can be helped to pick out outstanding enterprise, while financial institution can also be helped to enterprise's credit mistake
Good air control system is established in journey.
Currently, carrying out the mode of comprehensive score to enterprise is mainly to acquire several indexs of enterprise, it is that these indexs are artificial
Ground distributes corresponding weight and score rule, calculates enterprise's score according to the weight and score rule artificially distributed.It is existing this
It is lower to often result in scoring accuracy since human interference or subjective factor influence for the mode of kind comprehensive score.
Summary of the invention
The embodiment of the invention provides a kind of enterprise's Rating Model construction method, enterprise's methods of marking and devices, subtract significantly
Having lacked human interference or subjective factor influences, to effectively improve scoring accuracy.
A kind of enterprise's Rating Model construction method obtains the representation data and management data of type of industry, further includes:
Dimension-reduction treatment is carried out to the representation data and the management data, from the result of dimensionality reduction, selects at least two
Enter modular character;
Enter modular character to described at least two to classify, constructs at least one class index;
For each class index, execute:
According to the class index include described in enter important relationship between modular character, determine judgment matrix;
The judgment matrix is normalized, determines the corresponding judgment matrix weighted data of the judgment matrix
Collection, the judgment matrix weighted data collection, comprising: the class index include it is each enter modular character corresponding judgment matrix power
Weight;
Using variance method, the corresponding variance weighted data collection of the class index is calculated, the corresponding variance power of the class index
Weight data set, comprising: the class index include it is each enter the corresponding variance weight of modular character;
According to the judgment matrix weighted data collection and the variance weighted data collection, enter modular character weight described in calculating.
Preferably, above-mentioned enterprise's Rating Model construction method further comprises: drawing hierarchical chart, wherein the layer
Hierarchical structure chart includes: general objective layer, sub-goal layer and solution layer;
It is described to enter modular character to described at least two and classify, construct at least one class index, comprising:
Receive the general objective layer and sub-goal layer of external definition;
According to it is described enter modular character and the sub-goal layer correlation, enter modular character for described at least two and be filled into institute
State solution layer;
The sub-goal layer for determining definition is the class index.
Preferably, above-mentioned enterprise's Rating Model construction method further comprises:
Using power function mapping, the representation data and the management data are pre-processed;
Using linear mapping function, by the pretreated representation data and the pretreated management data points
The stable data section mapping of cloth determines the score of the stable data segment of distribution between 0~100;
Using sigmoid mapping function, by the pretreated representation data and the pretreated management data
Middle there are the data section mappings of wave crest between 0~100, determines that described there are the scores of the data segment of wave crest;
It is described to utilize variance method, calculate the corresponding variance weighted data collection of the class index, comprising:
Enter the score of the corresponding stable data segment of distribution of modular character according to each of described class index
With described there are the score of the data segment of wave crest, calculate and enter the corresponding average mark of modular character described in each;
The average mark for entering modular character according to each in the class index calculates and enters modular character pair described in each
The first intra-class variance and the first between-group variance answered;
According to it is described enter corresponding first intra-class variance of modular character and the first between-group variance, it is corresponding that modular character is entered described in calculating
The first intra-class variance weight and the first between-group variance weight, it is described enter the corresponding variance weight of modular character be it is described enter modular character
The sum of the corresponding first intra-class variance weight and the first between-group variance weight;
The variance weight combination for entering modular character described in will be all in the class index, becomes the corresponding variance of the class index
Weighted data collection.
Preferably,
It is described according to the class index include described in enter important relationship between modular character, determine judgment matrix, comprising:
Enter point of the corresponding stable data segment of distribution of modular character described in each for including according to the class index
It is several and described there are the score of the data segment of wave crest, it calculates and enters the corresponding average mark of modular character described in each;
According to the average mark, the significance level that every two in the class index enters modular character is compared;
According to comparing result, the score value of modular character distribution 1~9 is entered for the every two, determines judgment matrix.
Preferably, described that the judgment matrix is normalized, determine the corresponding judgement square of the judgment matrix
Battle array weighted data collection, comprising:
By belong in the judgment matrix it is same it is described enter modular character score value form a group;
Calculate each described group mean value, the second intra-class variance and the second between-group variance;
According to the mean value, second intra-class variance and second between-group variance, intra-class variance weight calculation is utilized
Formula and between-group variance weight calculation formula calculate the second intra-class variance weight and the second between-group variance weight;
Intra-class variance weight calculation formula:
Wherein, the wkCharacterize the corresponding intra-class variance weight of kth group in the judgment matrix;σkCharacterize the judgement square
The corresponding intra-class variance of kth group in battle array;σiCharacterize i-th group of corresponding intra-class variance in the judgment matrix;N characterizes the judgement
The total number organized in matrix;
Between-group variance weight calculation formula:
Wherein, the skCharacterize the corresponding between-group variance weight of kth group in the judgment matrix;μiCharacterize the judgement square
I-th group of corresponding mean value in battle array;μ characterizes the mean value of the judgment matrix;pkiCharacterize in the judgment matrix kth group and i-th group
Between between-group variance;N characterizes the total number organized in the judgment matrix;
Using the second intra-class variance weight and the second between-group variance weight, described group of population variance power is determined
Weight;
By corresponding all described groups of the population variance weight combination of the judgment matrix, it is corresponding to become the judgment matrix
Weighted data collection.
Preferably, described according to the judgment matrix weighted data collection and the variance weighted data collection, enter described in calculating
Modular character weight, comprising:
Using it is following enter modular character weight calculation formula, enter modular character weight described in calculating;
Enter modular character weight calculation formula:
Wherein, the QkIt characterizes and enters modular character weight into modular character k;The WkCharacterize the judgment matrix weighted data
It concentrates into the corresponding judgment matrix weight of modular character k;PkThe variance weighted data is characterized to concentrate into the corresponding variance of modular character k
Weight;WiIt characterizes the judgment matrix weighted data and concentrates i-th of judgment matrix weight;PiCharacterize the variance weighted data collection
In i-th of variance weight;N characterizes total number/variance power that the judgment matrix weighted data concentrates judgment matrix weight
Tuple is according to the total number for concentrating variance weight.
Preferably, above-mentioned enterprise's Rating Model construction method is applied to electric business enterprise.
Preferably, it is described enter modular character: include: run a shop duration, information integrity, to manage duration, change number, commodity full
Meaning degree, asset item risk number, total assets, risk case number, judicial risk number, case-involving amount of funds, is owed at service satisfaction
Tax number, sales volume ranking list, business number industry ranking, industry weighting ranking, moon sales volume, investment number, output investment ratio
Any two or multiple in example, intangible asset quantity and grading of paying taxes;
The class index, comprising: any one or more in stability, credit worthiness and loan repayment capacity.
A kind of enterprise's methods of marking realized based on any of the above-described enterprise's Rating Model, comprising:
Clean Target Enterprise data, and by the data after cleaning distribute to it is each enter modular character;
For the data after the corresponding cleaning of modular character are entered described in each, execute:
Data using power function mapping, after pre-processing the cleaning;
Using linear mapping function, exist stable data section mapping is distributed in the data after the pretreated cleaning
Between 0~100, the score of the stable data segment of distribution is determined;
Using sigmoid mapping function, by there are the data segments of wave crest to reflect in the data after the pretreated cleaning
It penetrates between 0~100, determines that described there are the scores of the data segment of wave crest;
Using the score of the stable data segment of the distribution and described there are the score of the data segment of wave crest, enter described in calculating
The average mark of modular character;
According to following enterprise's score calculation formula, the score of the Target Enterprise is calculated;
Enterprise's score calculation formula:
Wherein, the score of the F characterization Target Enterprise;QiEnter modular character in characterization enterprise's Rating Model for i-th enters mould
Index weights;fiCharacterize i-th of average mark for entering modular character;Include in n characterization enterprise's Rating Model enters the total of modular character
Number.
A kind of enterprise's Rating Model construction device, comprising: basic data processing unit, index confirmation unit and index power
Re-computation unit, wherein
The basic data processing unit, for obtaining the representation data and management data of type of industry, to the picture
As data and the management data carry out dimension-reduction treatment;
The index confirmation unit, for from the result for the dimensionality reduction that the basic data processing unit obtains, selection to be extremely
Few two enter modular character, enter modular character to described at least two and classify, construct at least one class index;
The index weights computing unit is executed for each class index for index confirmation unit building:
According to the class index include described in enter important relationship between modular character, determine judgment matrix;To the judgment matrix into
Row normalized, determines the corresponding judgment matrix weighted data collection of the judgment matrix, the judgment matrix weighted data collection,
Include: the class index include it is each enter the corresponding judgment matrix weight of modular character;Using variance method, the class index is calculated
Corresponding variance weighted data collection, the corresponding variance weighted data collection of the class index, comprising: the class index includes each
Enter the corresponding variance weight of modular character;According to the judgment matrix weighted data collection and the variance weighted data collection, institute is calculated
It states into modular character weight.
Preferably, the index weights computing unit, comprising: average mark computation subunit, first variance computation subunit
And variance weight sets constructs subelement, wherein
The basic data processing unit is further used for pre-processing the representation data and institute using power function mapping
State management data;Using linear mapping function, by the pretreated representation data and the pretreated management data
The middle stable data section mapping of distribution determines the score of the stable data segment of distribution between 0~100;Using sigmoid
Mapping function, by there are the data segments of wave crest to reflect in the pretreated representation data and the pretreated management data
It penetrates between 0~100, determines that described there are the scores of the data segment of wave crest;
The average mark computation subunit, it is each in the class index for being constructed according to the index confirmation unit
It is a it is described enter the score of the stable data segment of the distribution determined of the corresponding basic data processing unit of modular character and
It is described there are the score of the data segment of wave crest, calculate and enter the corresponding average mark of modular character described in each;
The first variance computation subunit, for according to the calculated class index of the average mark computation subunit
In enter the average mark of modular character described in each, calculate and enter corresponding first intra-class variance of modular character and first described in each
Between-group variance;
The variance weight sets constructs subelement, for according to the first variance computation subunit it is calculated it is described enter
Corresponding first intra-class variance of modular character and the first between-group variance enter the corresponding first intra-class variance weight of modular character described in calculating
With the first between-group variance weight, it is described enter the corresponding variance weight of modular character be it is described enter modular character it is described first group corresponding in
The sum of variance weight and the first between-group variance weight;Enter the variance weight group of modular character described in will be all in the class index
It closes, becomes the corresponding variance weighted data collection of the class index.
Preferably,
The index weights computing unit further comprises: judgment matrix confirms subelement,
The judgment matrix confirms subelement, the class index for determining according to the basic data processing unit
Enter the score of the stable data segment of the corresponding distribution of modular character described in each for including and described there are the data of wave crest
The score of section calculates and enters the corresponding average mark of modular character described in each;From high to low according to the average mark, institute is determined
State into modular character significance level from high to low;According to it is described enter modular character significance level from high to low, for it is described enter mould refer to
The score value of mark distribution 1~9;Using it is described enter the corresponding score value of modular character, determine judgment matrix.
Preferably,
The index weights computing unit further comprises: judge that weight sets constructs subelement,
The judgement weight sets constructs subelement, for will belong in the judgment matrix it is same it is described enter modular character
Score value forms a group;Calculate each described group mean value, the second intra-class variance and the second between-group variance;According to described equal
Value, second intra-class variance and second between-group variance, utilize intra-class variance weight calculation formula and between-group variance weight
Calculation formula calculates the second intra-class variance weight and the second between-group variance weight;
Intra-class variance weight calculation formula:
Wherein, the wkCharacterize the corresponding intra-class variance weight of kth group in the judgment matrix;σkCharacterize the judgement square
The corresponding intra-class variance of kth group in battle array;σiCharacterize i-th group of corresponding intra-class variance in the judgment matrix;N characterizes the judgement
The total number organized in matrix;
Between-group variance weight calculation formula:
Wherein, the skCharacterize the corresponding between-group variance weight of kth group in the judgment matrix;μiCharacterize the judgement square
I-th group of corresponding mean value in battle array;μ characterizes the mean value of the judgment matrix;pkiCharacterize in the judgment matrix kth group and i-th group
Between between-group variance;N characterizes the total number organized in the judgment matrix;
Using the second intra-class variance weight and the second between-group variance weight, described group of population variance power is determined
Weight;By corresponding all described groups of the population variance weight combination of the judgment matrix, become the corresponding weight of the judgment matrix
Data set.
Preferably,
The index weights computing unit further comprises: weight calculation subelement,
The weight calculation subelement, for using it is following enter modular character weight calculation formula, enter modular character described in calculating
Weight;
Enter modular character weight calculation formula:
Wherein, the QkIt characterizes and enters modular character weight into modular character k;The WkCharacterize the judgment matrix weighted data
It concentrates into the corresponding judgment matrix weight of modular character k;PkThe variance weighted data is characterized to concentrate into the corresponding variance of modular character k
Weight;WiIt characterizes the judgment matrix weighted data and concentrates i-th of judgment matrix weight;PiCharacterize the variance weighted data collection
In i-th of variance weight;N characterizes total number/variance power that the judgment matrix weighted data concentrates judgment matrix weight
Tuple is according to the total number for concentrating variance weight.
Preferably, above-mentioned apparatus is applied to electric business enterprise.
Preferably, it is described enter modular character: include: run a shop duration, information integrity, to manage duration, change number, commodity full
Meaning degree, asset item risk number, total assets, risk case number, judicial risk number, case-involving amount of funds, is owed at service satisfaction
Tax number, sales volume ranking list, business number industry ranking, industry weighting ranking, moon sales volume, investment number, output investment ratio
Any two or multiple in example, intangible asset quantity and grading of paying taxes;
The class index, comprising: any one or more in stability, credit worthiness and loan repayment capacity.
The embodiment of the invention provides a kind of enterprise's Rating Model construction method, enterprise's methods of marking and device, the enterprises
Rating Model construction method includes: the representation data and management data for obtaining type of industry, to representation data and management data
Dimension-reduction treatment is carried out, from the result of dimensionality reduction, selects at least two to enter modular character;Enter modular character at least two to classify,
Construct at least one class index;For each class index, execute: the important pass entered between modular character for including according to class index
System, determines judgment matrix;Judgment matrix is normalized, determines the corresponding judgment matrix weighted data of judgment matrix
Collection, judgment matrix weighted data collection, comprising: class index include it is each enter the corresponding judgment matrix weight of modular character;Utilize side
Poor method calculates the corresponding variance weighted data collection of class index, the corresponding variance weighted data collection of class index, comprising: class index packet
Contain it is each enter the corresponding variance weight of modular character;According to judgment matrix weighted data collection and variance weighted data collection, calculate into
Modular character weight, the building of enterprise's Rating Model are the confirmation into modular character weight, and in the process of weight confirmation, various variances are complete
It is determined entirely by data distribution, without any subjective factor with time change, data structure changes, and weight also can be with change
Change, greatly reducing human interference or subjective factor influences, to effectively improve scoring accuracy.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is a kind of flow chart of enterprise's Rating Model construction method provided by one embodiment of the present invention;
Fig. 2 is a kind of flow chart for enterprise's Rating Model construction method that another embodiment of the present invention provides;
Fig. 3 is that enterprise provided by one embodiment of the present invention enters modular character hierarchical chart;
Fig. 4 is a kind of flow chart of enterprise's methods of marking provided by one embodiment of the present invention;
Fig. 5 is enterprise's Rating Model appraisal result distribution map provided by one embodiment of the present invention;
Fig. 6 is a kind of enterprise's Rating Model inspection figure provided by one embodiment of the present invention;
Fig. 7 is the structural schematic diagram of enterprise's Rating Model construction device provided by one embodiment of the present invention;
Fig. 8 is the structural schematic diagram for enterprise's Rating Model construction device that another embodiment of the present invention provides.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments, based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
As shown in Figure 1, the embodiment of the invention provides a kind of enterprise's Rating Model construction method, this method may include with
Lower step:
Step 101: obtaining the representation data and management data of type of industry;
Step 102: dimension-reduction treatment being carried out to representation data and management data, from the result of dimensionality reduction, selects at least two
Enter modular character;
Step 103: entering modular character at least two and classify, construct at least one class index;
Step 104: it is directed to each class index, executes the important relationship entered between modular character for including according to class index,
Determine judgment matrix;
Step 105: judgment matrix being normalized, determines the corresponding judgment matrix weighted data of judgment matrix
Collection, judgment matrix weighted data collection, comprising: class index include it is each enter the corresponding judgment matrix weight of modular character;
Step 106: utilizing variance method, calculate the corresponding variance weighted data collection of class index, the corresponding variance power of class index
Weight data set, comprising: class index include it is each enter the corresponding variance weight of modular character;
Step 107: according to judgment matrix weighted data collection and variance weighted data collection, calculating into modular character weight.
Wherein, representation data is directly obtained from other systems such as Tianyuan big data platform etc., for electric business enterprise
For, management data is directly acquired from electric business website, can be straight from enterprise official website for other non-electrical commercial businesses industry
It obtains.
In addition, being reduced to processing is mainly merged into a dimension for a variety of data, existing dimensionality reduction side can be directly selected
Formula obtains.
For electric business enterprise, enter modular character can include: duration of running a shop, information integrity manage duration, change time
Number, commodity satisfaction, service satisfaction, asset item risk number, total assets, risk case number, judicial risk number, case-involving money
Golden number amount, tax arrear number, sales volume ranking list, business number industry ranking, industry weighting ranking, moon sales volume, investment time
Number, ratio between investments, intangible asset quantity and grading of paying taxes.
For electric business enterprise, class index can include: stability, credit worthiness and loan repayment capacity.
In the embodiment shown in fig. 1, the representation data and management data for obtaining type of industry, to representation data and warp
It seeks data and carries out dimension-reduction treatment, from the result of dimensionality reduction, select at least two to enter modular character;Enter modular character at least two to carry out
Classification, constructs at least one class index;For each class index, executes: being entered between modular character according to what class index included
Important relationship determines judgment matrix;Judgment matrix is normalized, determines the corresponding judgment matrix weight of judgment matrix
Data set, judgment matrix weighted data collection, comprising: class index include it is each enter the corresponding judgment matrix weight of modular character;Benefit
With variance method, the corresponding variance weighted data collection of class index, the corresponding variance weighted data collection of class index, comprising: class refers to are calculated
Mark include it is each enter the corresponding variance weight of modular character;According to judgment matrix weighted data collection and variance weighted data collection, meter
Modular character weight is counted, the building of enterprise's Rating Model is the confirmation into modular character weight, in the process of weight confirmation, various sides
Difference is determined that, without any subjective factor with time change, data structure changes by data distribution completely, and weight also can be with
Variation, greatly reducing human interference or subjective factor influences, to effectively improve scoring accuracy.
In an alternative embodiment of the invention, above-mentioned enterprise's Rating Model construction method further comprises: drawing hierarchical structure
Figure, wherein hierarchical chart includes: general objective layer, sub-goal layer and solution layer;Enter modular character at least two to classify,
Construct at least one class index, comprising: receive the general objective layer and sub-goal layer of external definition;According to entering modular character and sub-goal
The correlation of layer, enters modular character at least two and is filled into solution layer;The sub-goal layer for determining definition is class index.Pass through level
Structure chart can be more convenient the building of class index, while be more clear the division of class index.
In an alternative embodiment of the invention, above-mentioned enterprise's Rating Model construction method, is further comprised: being reflected using power function
It penetrates, pre-processes representation data and management data;Using linear mapping function, by pretreated representation data and pretreated
It is distributed stable data section mapping in management data between 0~100, determines the score for being distributed stable data segment;Using
Sigmoid mapping function, by there are the data segments of wave crest to reflect in pretreated representation data and pretreated management data
It penetrates between 0~100, determines that there are the scores of the data segment of wave crest;Using variance method, the corresponding variance weight of class index is calculated
The specific embodiment of data set includes: to enter the corresponding stable data segment of distribution of modular character according to each of class index
Score and there are the score of the data segment of wave crest, calculates each and enters the corresponding average mark of modular character;According to every in class index
One enters the average mark of modular character, calculates each and enters corresponding first intra-class variance of modular character and the first between-group variance;Root
According to corresponding first intra-class variance of modular character and the first between-group variance is entered, calculate into the corresponding first intra-class variance weight of modular character
With the first between-group variance weight, entering the corresponding variance weight of modular character is into the corresponding first intra-class variance weight of modular character and
The sum of one between-group variance weight;By in class index it is all enter modular character variance weight combination, become the corresponding variance of class index
Weighted data collection.
In an alternative embodiment of the invention, the important relationship entered between modular character for including according to class index determines judgement
The specific embodiment of matrix can include: the corresponding stable data segment of distribution of modular character is entered according to each that class index includes
Score and there are the score of the data segment of wave crest, calculate each and enter the corresponding average mark of modular character;According to average mark,
Every two enters the significance level of modular character in comparison class index;According to comparing result, enter modular character distribution 1~9 for every two
Score value determines judgment matrix.
In an alternative embodiment of the invention, judgment matrix is normalized, determines the corresponding judgement of judgment matrix
Matrix weight data set specific embodiment can include: by belong in judgment matrix it is same enter modular character score value form one
Group;Calculate mean value, the second intra-class variance and the second between-group variance of each group;According to mean value, the second intra-class variance and second
Between-group variance calculates the second intra-class variance weight using intra-class variance weight calculation formula and between-group variance weight calculation formula
With the second between-group variance weight;
Intra-class variance weight calculation formula:
Wherein, wkCharacterize the corresponding intra-class variance weight of kth group in judgment matrix;σkCharacterize kth group in the judgment matrix
Corresponding intra-class variance;σiCharacterize i-th group of corresponding intra-class variance in judgment matrix;The total number organized in n characterization judgment matrix;
Between-group variance weight calculation formula:
Wherein, skCharacterize the corresponding between-group variance weight of kth group in judgment matrix;μiCharacterize i-th group of correspondence in judgment matrix
Mean value;The mean value of μ characterization judgment matrix;pkiCharacterize the between-group variance in judgment matrix between kth group and i-th group;N characterization is sentenced
The total number organized in disconnected matrix;
Using the second intra-class variance weight and the second between-group variance weight, the population variance weight of group is determined;By judgment matrix
Corresponding all groups of population variance weight combination, becomes the corresponding weighted data collection of judgment matrix.
In an alternative embodiment of the invention, it according to judgment matrix weighted data collection and variance weighted data collection, calculates into mould
Index weights specific embodiment can include: using it is following enter modular character weight calculation formula, calculate into modular character weight;
Enter modular character weight calculation formula:
Wherein, QkIt characterizes and enters modular character weight into modular character k;WkCharacterization judgment matrix weighted data is concentrated into modular character k
Corresponding judgment matrix weight;PkCharacterization variance weighted data is concentrated into the corresponding variance weight of modular character k;WiCharacterization judges square
Battle array weighted data concentrates i-th of judgment matrix weight;PiIt characterizes variance weighted data and concentrates i-th of variance weight;N characterization judgement
The total number of judgment matrix weight/variance weighted data concentrates the total number of variance weight in matrix weight data set.
In an embodiment of the invention, above-mentioned enterprise's Rating Model construction method can be applied to electric business enterprise.
In an embodiment of the invention, it is above-mentioned enter modular character can include: duration of running a shop, information integrity, manage duration,
Change number, commodity satisfaction, service satisfaction, asset item risk number, total assets, risk case number, judicial risk number,
Case-involving amount of funds, tax arrear number, sales volume ranking list, business number industry ranking, industry weighting ranking, moon sales volume,
Any two or multiple in investment number, ratio between investments, intangible asset quantity and grading of paying taxes;
In an embodiment of the invention, above-mentioned class index can include: any in stability, credit worthiness and loan repayment capacity
It is one or more.
For thinking electric business enterprise building enterprise's Rating Model, as shown in Fig. 2, enterprise's Rating Model construction method can wrap
Include following steps:
Step 201: drawing hierarchical chart;
Step 202: obtaining the representation data and management data of type of industry;
The representation data can be directly acquired from other systems or platform, such as enterprise marketing product type, enterprise
Industry manages grade etc.;
Step 203: using power function mapping, pre-process representation data and management data;
Step 204: linear mapping function is used, by pretreated representation data and pretreated management data points
The stable data section mapping of cloth determines the score for being distributed stable data segment between 0~100;
Step 205: sigmoid mapping function is used, by pretreated representation data and pretreated management data
Middle there are the data section mappings of wave crest between 0~100, determines that there are the scores of the data segment of wave crest;
There is no strict sequence between above-mentioned steps 204 and step 205.Through the above steps 204 and step 205 energy
Enough make the score distribution of data relatively uniform, reduces the appearance of extreme value to the greatest extent.
Step 206: to representation data, management data carries out dimension-reduction treatment with treated, from the result of dimensionality reduction, selection
At least two enter modular character;
The step and above-mentioned steps 203 are not to having strict sequence between step 205.The dimension-reduction treatment, being will be several
The data of seed type are merged into the same dimension.For example, this dimension of information integrity is by electric business business entity information, electric business
A dimension obtained from Corporate finance information, electric business enterprise business address etc. merge, for another example, commodity satisfaction are sale
Dimension obtained from the evaluation situation of each commodity merges etc., there may be intersections for the data between each dimension.For
For electric business enterprise, enter modular character may include run a shop duration, information integrity, manage duration, change number, commodity satisfaction,
Service satisfaction, asset item risk number, total assets, risk case number, judicial risk number, case-involving amount of funds, tax arrear time
Number, sales volume ranking list, business number industry ranking, industry weighting ranking, moon sales volume, investment number, ratio between investments, nothing
Shape amount of assets and grading of paying taxes.
Step 207: receiving the general objective layer and sub-goal layer of external definition;
As shown in figure 3, include: general objective layer, sub-goal layer and solution layer for the hierarchical chart that electric business enterprise draws,
In, general objective layer be enterprise scoring, sub-goal layer be stability, credit worthiness and loan repayment capacity, solution layer be it is each enter modular character.
Step 208: according to the correlation for entering modular character with sub-goal layer, entering modular character at least two and be filled into scheme
Layer determines that the sub-goal layer of definition is class index;
As shown in figure 3, the enter modular character bigger with stability correlation is when running a shop duration, information integrity, operation
Long and change number;The bigger modular character that enters is commodity satisfaction, service satisfaction, asset item wind with credit worthiness correlation
Dangerous number, total assets, risk case number, judicial risk number, case-involving amount of funds, tax arrear number;With loan repayment capacity correlation ratio
It is biggish enter modular character be sales volume ranking list, business number industry ranking, industry weight ranking, moon sales volume, investment time
Number, ratio between investments, intangible asset quantity and grading of paying taxes.Then class index is stability, credit worthiness and loan repayment capacity.
Step 209: according to the score for entering the corresponding stable data segment of distribution of modular character and there are the data segments of wave crest
Score is calculated into the corresponding average mark of modular character;
Step 210: being directed to each class index, compare the significance level that every two in class index enters modular character;
Step 211: according to comparing result, the score value of modular character distribution 1~9 is entered for every two, determines judgment matrix;
Judgment matrix is A=aij, which meets aij=1/aji, aii=1, wherein i and j characterizes a class index
In any two enter modular character.
Such as: for class index " stability ", to enter modular character since it includes four, the judgment matrix formed is
4 × 4 matrixes.
The corresponding judgment matrix of class index " stability ":Wherein, it 1 characterizes and refers into mould
It marks " duration of running a shop ", 2 characterize into modular character " information integrity ", and 3 characterize and characterize into modular character " manage duration " 4 into modular character
" change number ", and a11=1, a22=1, a33=1, a44=1;a12=1/a21, a13=1/a31, a14=1/a41, a23=1/a32,
a24=1/a42, a34=1/a43。
Step 212: by belong in judgment matrix it is same enter modular character score value form a group;
Such as the corresponding judgment matrix of class index " stability ", by a11, a12, a13, a14, a21, a31, a41It is divided into one
Group, a21, a22, a23, a24, a12, a32, a42It is divided into one group,
a31, a32, a33, a34, a13, a23, a43It is divided into one group, a41, a42, a43, a44, a14, a24, a34It is divided into one group.
Step 213: calculating mean value, the second intra-class variance and the second between-group variance of each group;
The calculating of the mean value, intra-class variance and between-group variance can be calculated using existing calculation formula, herein not
It repeats again.
Step 214: according to mean value, the second intra-class variance and the second between-group variance, calculating the second intra-class variance weight and the
Two between-group variance weights;
The step is mainly, using intra-class variance weight calculation formula and between-group variance weight calculation formula, to calculate second
Intra-class variance weight and the second between-group variance weight;
Intra-class variance weight calculation formula:
Wherein, wkCharacterize the corresponding intra-class variance weight of kth group in judgment matrix;σkIt is corresponding to characterize kth group in judgment matrix
Intra-class variance;σiCharacterize i-th group of corresponding intra-class variance in judgment matrix;The total number organized in n characterization judgment matrix;
Between-group variance weight calculation formula:
Wherein, skCharacterize the corresponding between-group variance weight of kth group in the judgment matrix;μiI-th group is characterized in judgment matrix
Corresponding mean value;The mean value of μ characterization judgment matrix;pkiCharacterize the between-group variance in judgment matrix between kth group and i-th group;N table
The total number organized in sign judgment matrix;
Step 215: utilizing the second intra-class variance weight and the second between-group variance weight, determine the population variance weight of group;
Population variance=second the+the second between-group variance of intra-class variance.
Step 216: combining corresponding all groups of the population variance weight of judgment matrix, become the corresponding weight of judgment matrix
Data set;
Above-mentioned steps 209 to step 216 is to construct the process of weighted data collection, with following variance weighted data collection structures
The process of building be it is relatively independent, without strict sequence.
Step 217: the score of the corresponding stable data segment of distribution of modular character being entered according to each of class index and is deposited
In the score of the data segment of wave crest, calculates each and enter the corresponding average mark of modular character;
Step 218: entering the average mark of modular character according to each in class index, calculating each, to enter modular character corresponding
First intra-class variance and the first between-group variance;
Intra-class variance is utilized to be calculated into the corresponding each score of modular character, and between-group variance is utilized into modular character
Average mark be calculated.Existing variance calculation formula is directlyed adopt to obtain.
Step 219: according to corresponding first intra-class variance of modular character and the first between-group variance is entered, calculating corresponding into modular character
The first intra-class variance weight and the first between-group variance weight;
It is public that the first intra-class variance weight and the first between-group variance weight are utilized respectively above-mentioned between-group variance weight calculation
Formula and between-group variance weight calculation formula are calculated.
Step 220: by class index it is all enter modular character variance weight combine, become the corresponding variance weight of class index
Data set;
The variance weight for entering modular character is into the corresponding first intra-class variance weight of modular character and the first between-group variance weight
The sum of.
Step 221: according to judgment matrix weighted data collection and variance weighted data collection, calculating into modular character weight.
Using it is following enter modular character weight calculation formula, calculate into modular character weight;
Enter modular character weight calculation formula:
Wherein, QkIt characterizes and enters modular character weight into modular character k;WkThe judgment matrix weighted data is characterized to concentrate into mould
The corresponding judgment matrix weight of index k;PkCharacterization variance weighted data is concentrated into the corresponding variance weight of modular character k;WiCharacterization is sentenced
I-th of judgment matrix weight in disconnected matrix weight data set;PiIt characterizes variance weighted data and concentrates i-th of variance weight;N characterization
Judgment matrix weighted data concentrates total number/variance weighted data concentration variance weight total number of judgment matrix weight.
It is commented as shown in figure 4, the embodiment of the present invention provides the enterprise that any enterprise's Rating Model based on above-mentioned building is realized
Divide method, it may include following steps:
Step 401: clean the data of Target Enterprise, and by the data after cleaning distribute to it is each enter modular character;
Step 402: enter the data after the corresponding cleaning of modular character for each, execute: using power function mapping, it is pre- to locate
Clear the data after washing;
Step 403: using linear mapping function, reflected stable data segment is distributed in the data after pretreated cleaning
It penetrates between 0~100, determines the score for being distributed stable data segment;
Step 404: sigmoid mapping function is used, by the data in the data after pretreated cleaning there are wave crest
Section is mapped between 0~100, determines that there are the scores of the data segment of wave crest;
Step 405: using the score for being distributed stable data segment and there are the score of the data segment of wave crest, calculating and refer into mould
Target average mark;
Step 406: based on enterprise's Rating Model and the average mark for entering modular character, calculating the score of Target Enterprise.
The step is mainly, according to following enterprise's score calculation formula, to calculate the score of Target Enterprise;
Enterprise's score calculation formula:
Wherein, F characterizes the score of Target Enterprise;QiEnter modular character in characterization enterprise's Rating Model for i-th enters modular character
Weight;fiCharacterize i-th of average mark for entering modular character;The total number for entering modular character for including in n characterization enterprise's Rating Model.
The scoring obtained using above-mentioned Rating Model, most of score are distributed within 3 standard deviations such as Fig. 5 i.e.
Between (39.08,91.88).According to scoring inspection result, as shown in Figure 6, it can be seen that score is substantially distributed in 45 degree of diagonal lines
On, normal distribution is substantially conformed to, central-limit theorem is met.
As shown in fig. 7, the embodiment of the present invention provides a kind of enterprise's Rating Model construction device, comprising: basic data processing
Unit 701, index confirmation unit 702 and index weights computing unit 703, wherein basic data processing unit 701 is used for
The representation data and management data for obtaining type of industry carry out dimension-reduction treatment to representation data and management data;
Index confirmation unit 702, for from the result for the dimensionality reduction that basic data processing unit 701 obtains, selection to be at least
Two enter modular character, enter modular character at least two and classify, and construct at least one class index;
Index weights computing unit 703, each class index for constructing for index confirmation unit 702, executes: root
The important relationship entered between modular character for including according to class index, determines judgment matrix;Judgment matrix is normalized, really
Determine the corresponding judgment matrix weighted data collection of judgment matrix, judgment matrix weighted data collection, comprising: class index include it is each enter
The corresponding judgment matrix weight of modular character;Using variance method, the corresponding variance weighted data collection of class index is calculated, class index is corresponding
Variance weighted data collection, comprising: class index include it is each enter the corresponding variance weight of modular character;According to judgment matrix weight
Data set and variance weighted data collection, calculate into modular character weight.
In an alternative embodiment of the invention, as shown in figure 8, These parameters weight calculation unit 703, comprising: average mark meter
Operator unit 7031, first variance computation subunit 7032 and variance weight sets construct subelement 7033, wherein
Basic data processing unit 701 is further used for pre-processing representation data and the operation using power function mapping
Data;Using linear mapping function, smoothly number will be distributed in pretreated representation data and pretreated management data
It is mapped between 0~100 according to section, determines the score for being distributed stable data segment;Using sigmoid mapping function, will pre-process
There are the data section mappings of wave crest between 0~100 in representation data and pretreated management data afterwards, determines that there are waves
The score of the data segment at peak;
The average mark computation subunit 7031, it is each in the class index for being constructed according to index confirmation unit 702
It is a enter the score of the stable data segment of distribution determined of the corresponding basic data processing unit 701 of modular character and there are wave crests
The score of data segment calculates each and enters the corresponding average mark of modular character;
First variance computation subunit 7032, for according to every in the calculated class index of average mark computation subunit 7031
One enters the average mark of modular character, calculates each and enters corresponding first intra-class variance of modular character and the first between-group variance;
Variance weight sets construct subelement 7033, for according to first variance computation subunit 7032 it is calculated enter mould refer to
Corresponding first intra-class variance and the first between-group variance are marked, is calculated into the corresponding first intra-class variance weight of modular character and first group
Between variance weight, enter the corresponding variance weight of modular character be between the corresponding first intra-class variance weight of modular character and first group side
The sum of poor weight;By in class index it is all enter modular character variance weight combination, become the corresponding variance weighted data of class index
Collection.
In an alternative embodiment of the invention, These parameters weight calculation unit 703 further comprises: judgment matrix confirmation
Subelement (not shown),
Judgment matrix confirms subelement, and the class index for being determined according to basic data processing unit 701 includes every
One enters the score of the corresponding stable data segment of distribution of modular character and there are the score of the data segment of wave crest, calculates each and enter
The corresponding average mark of modular character;From high to low according to average mark, determine the significance level of modular character from high to low;According to
Enter the significance level of modular character from high to low, for the score value for entering modular character distribution 1~9;Using the corresponding score value of modular character is entered, really
Determine judgment matrix.
In still another embodiment of the process, index weights computing unit further comprises: judging that weight sets constructs subelement
(not shown),
Judge weight sets construct subelement, for by belong in judgment matrix it is same enter modular character score value form one
Group;Calculate mean value, the second intra-class variance and the second between-group variance of each group;According to mean value, the second intra-class variance and second
Between-group variance calculates the second intra-class variance weight using intra-class variance weight calculation formula and between-group variance weight calculation formula
With the second between-group variance weight;
Intra-class variance weight calculation formula:
Wherein, wkCharacterize the corresponding intra-class variance weight of kth group in judgment matrix;σkIt is corresponding to characterize kth group in judgment matrix
Intra-class variance;σiCharacterize i-th group of corresponding intra-class variance in judgment matrix;The total number organized in n characterization judgment matrix;
Between-group variance weight calculation formula:
Wherein, skCharacterize the corresponding between-group variance weight of kth group in judgment matrix;μiCharacterize i-th group of correspondence in judgment matrix
Mean value;The mean value of μ characterization judgment matrix;pkiCharacterize the between-group variance in judgment matrix between kth group and i-th group;N characterization is sentenced
The total number organized in disconnected matrix;
Using the second intra-class variance weight and the second between-group variance weight, the population variance weight of group is determined;By judgment matrix
Corresponding all groups of population variance weight combination, becomes the corresponding weighted data collection of judgment matrix.
In still another embodiment of the process, index weights computing unit further comprises: weight calculation subelement is (in figure
It is not shown),
Weight calculation subelement, for using it is following enter modular character weight calculation formula, calculate into modular character weight;
Enter modular character weight calculation formula:
Wherein, QkIt characterizes and enters modular character weight into modular character k;WkCharacterization judgment matrix weighted data is concentrated into modular character k
Corresponding judgment matrix weight;PkCharacterization variance weighted data is concentrated into the corresponding variance weight of modular character k;WiCharacterization judges square
Battle array weighted data concentrates i-th of judgment matrix weight;PiIt characterizes variance weighted data and concentrates i-th of variance weight;N characterization judgement
The total number of judgment matrix weight/variance weighted data concentrates the total number of variance weight in matrix weight data set.
In an embodiment of the invention, above-mentioned enterprise's Rating Model construction device is applied to electric business enterprise.
In an embodiment of the invention, it is above-mentioned enter modular character: include: run a shop duration, information integrity, manage duration,
Change number, commodity satisfaction, service satisfaction, asset item risk number, total assets, risk case number, judicial risk number,
Case-involving amount of funds, tax arrear number, sales volume ranking list, business number industry ranking, industry weighting ranking, moon sales volume,
Any two or multiple in investment number, ratio between investments, intangible asset quantity and grading of paying taxes;Above-mentioned class index, comprising:
It is any one or more in stability, credit worthiness and loan repayment capacity.
The contents such as the information exchange between each unit, implementation procedure in above-mentioned apparatus, due to implementing with the method for the present invention
Example is based on same design, and for details, please refer to the description in the embodiment of the method for the present invention, and details are not described herein again.
The embodiment of the invention provides a kind of readable mediums, including execute instruction, when the processor of storage control executes
Described when executing instruction, the storage control executes the method that any of the above-described embodiment of the present invention provides.
The embodiment of the invention provides a kind of storage controls, comprising: processor, memory and bus;The memory
It is executed instruction for storing, the processor is connect with the memory by the bus, when the storage control is run
When, the processor executes the described of memory storage and executes instruction, so that the storage control executes in the present invention
The method that any embodiment offer is provided.
In conclusion more than the present invention each embodiment at least has the following beneficial effects:
1, the representation data and management data in embodiments of the present invention, obtaining type of industry, to representation data and warp
It seeks data and carries out dimension-reduction treatment, from the result of dimensionality reduction, select at least two to enter modular character;Enter modular character at least two to carry out
Classification, constructs at least one class index;For each class index, executes: being entered between modular character according to what class index included
Important relationship determines judgment matrix;Judgment matrix is normalized, determines the corresponding judgment matrix weight of judgment matrix
Data set, judgment matrix weighted data collection, comprising: class index include it is each enter the corresponding judgment matrix weight of modular character;Benefit
With variance method, the corresponding variance weighted data collection of class index, the corresponding variance weighted data collection of class index, comprising: class refers to are calculated
Mark include it is each enter the corresponding variance weight of modular character;According to judgment matrix weighted data collection and variance weighted data collection, meter
Modular character weight is counted, the building of enterprise's Rating Model is the confirmation into modular character weight, in the process of weight confirmation, various sides
Difference is determined that, without any subjective factor with time change, data structure changes by data distribution completely, and weight also can be with
Variation, greatly reducing human interference or subjective factor influences, to effectively improve scoring accuracy.
2, in embodiments of the present invention, by drawing hierarchical chart, wherein hierarchical chart includes: general objective layer, son
Destination layer and solution layer;Receive the general objective layer and sub-goal layer of external definition;It is related to sub-goal layer according to modular character is entered
Property, enter modular character at least two and is filled into solution layer;The sub-goal layer for determining definition is class index.It can by hierarchical chart
To be more convenient the building of class index, while it is more clear the division of class index.
3, in embodiments of the present invention, using power function mapping, representation data and management data are pre-processed;Using linearly reflecting
Function is penetrated, stable data section mapping will be distributed 0~100 in pretreated representation data and pretreated management data
Between, determine the score for being distributed stable data segment;Using sigmoid mapping function, by pretreated representation data and in advance
There are the data section mappings of wave crest between 0~100 in treated management data, determines point there are the data segment of wave crest
Number;The score distribution of data can be made relatively uniform, reduce the appearance of extreme value to the greatest extent, to further increase the standard of scoring
True property.
It should be noted that, in this document, such as first and second etc relational terms are used merely to an entity
Or operation is distinguished with another entity or operation, is existed without necessarily requiring or implying between these entities or operation
Any actual relationship or order.Moreover, the terms "include", "comprise" or its any other variant be intended to it is non-
It is exclusive to include, so that the process, method, article or equipment for including a series of elements not only includes those elements,
It but also including other elements that are not explicitly listed, or further include solid by this process, method, article or equipment
Some elements.In the absence of more restrictions, the element limited by sentence " including one ", is not arranged
Except there is also other identical factors in the process, method, article or apparatus that includes the element.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can store in computer-readable storage medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed;And storage medium above-mentioned includes: ROM, RAM, magnetic disk or light
In the various media that can store program code such as disk.
Finally, it should be noted that the foregoing is merely presently preferred embodiments of the present invention, it is merely to illustrate skill of the invention
Art scheme, is not intended to limit the scope of the present invention.Any modification for being made all within the spirits and principles of the present invention,
Equivalent replacement, improvement etc., are included within the scope of protection of the present invention.
Claims (10)
1. a kind of enterprise's Rating Model construction method, which is characterized in that the representation data and management data of type of industry are obtained,
Further include:
Dimension-reduction treatment is carried out to the representation data and the management data, from the result of dimensionality reduction, selects at least two to enter mould
Index;
Enter modular character to described at least two to classify, constructs at least one class index;
For each class index, execute:
According to the class index include described in enter important relationship between modular character, determine judgment matrix;
The judgment matrix is normalized, determines the corresponding judgment matrix weighted data collection of the judgment matrix, institute
State judgment matrix weighted data collection, comprising: the class index include it is each enter the corresponding judgment matrix weight of modular character;
Using variance method, the corresponding variance weighted data collection of the class index is calculated, the corresponding variance weight number of the class index
According to collection, comprising: the class index include it is each enter the corresponding variance weight of modular character;
According to the judgment matrix weighted data collection and the variance weighted data collection, enter modular character weight described in calculating.
2. enterprise's Rating Model construction method according to claim 1, which is characterized in that further comprise: drawing level
Structure chart, wherein the hierarchical chart includes: general objective layer, sub-goal layer and solution layer;
It is described to enter modular character to described at least two and classify, construct at least one class index, comprising:
Receive the general objective layer and sub-goal layer of external definition;
According to it is described enter modular character and the sub-goal layer correlation, enter modular character for described at least two and be filled into the side
Pattern layer;
The sub-goal layer for determining definition is the class index.
3. enterprise's Rating Model construction method according to claim 1, which is characterized in that further comprise:
Using power function mapping, the representation data and the management data are pre-processed;
It is flat by being distributed in the pretreated representation data and the pretreated management data using linear mapping function
Steady data section mapping determines the score of the stable data segment of distribution between 0~100;
Using sigmoid mapping function, will be deposited in the pretreated representation data and the pretreated management data
Wave crest data section mapping between 0~100, determine that described there are the scores of the data segment of wave crest;
It is described to utilize variance method, calculate the corresponding variance weighted data collection of the class index, comprising:
Enter score and the institute of the corresponding stable data segment of distribution of modular character according to each of described class index
The score for stating the data segment there are wave crest calculates and enters the corresponding average mark of modular character described in each;
The average mark for entering modular character according to each in the class index, calculates that enter modular character described in each corresponding
First intra-class variance and the first between-group variance;
According to it is described enter corresponding first intra-class variance of modular character and the first between-group variance, enter modular character corresponding described in calculating
One intra-class variance weight and the first between-group variance weight, it is described enter the corresponding variance weight of modular character be it is described enter modular character it is corresponding
The first intra-class variance weight and the sum of the first between-group variance weight;
The variance weight combination for entering modular character described in will be all in the class index, becomes the corresponding variance weight of the class index
Data set.
4. enterprise's Rating Model construction method according to claim 3, which is characterized in that
It is described according to the class index include described in enter important relationship between modular character, determine judgment matrix, comprising:
Enter described in each for including according to the class index the corresponding stable data segment of distribution of modular character score and
It is described there are the score of the data segment of wave crest, calculate and enter the corresponding average mark of modular character described in each;
According to the average mark, the significance level that every two in the class index enters modular character is compared;
According to comparing result, the score value of modular character distribution 1~9 is entered for the every two, determines judgment matrix.
5. enterprise's Rating Model construction method according to claim 3, which is characterized in that it is described to the judgment matrix into
Row normalized determines the corresponding judgment matrix weighted data collection of the judgment matrix, comprising:
By belong in the judgment matrix it is same it is described enter modular character score value form a group;
Calculate each described group mean value, the second intra-class variance and the second between-group variance;
According to the mean value, second intra-class variance and second between-group variance, intra-class variance weight calculation formula is utilized
With between-group variance weight calculation formula, the second intra-class variance weight and the second between-group variance weight are calculated;
Intra-class variance weight calculation formula:
Wherein, the wkCharacterize the corresponding intra-class variance weight of kth group in the judgment matrix;σkIt characterizes in the judgment matrix
The corresponding intra-class variance of kth group;σiCharacterize i-th group of corresponding intra-class variance in the judgment matrix;N characterizes the judgment matrix
Middle group of total number;
Between-group variance weight calculation formula:
Wherein, the skCharacterize the corresponding between-group variance weight of kth group in the judgment matrix;μiIt characterizes in the judgment matrix
I-th group of corresponding mean value;μ characterizes the mean value of the judgment matrix;pkiIt characterizes in the judgment matrix between kth group and i-th group
Between-group variance;N characterizes the total number organized in the judgment matrix;
Using the second intra-class variance weight and the second between-group variance weight, described group of population variance weight is determined;
By corresponding all described groups of the population variance weight combination of the judgment matrix, become the corresponding weight of the judgment matrix
Data set.
6. method according to any one of claims 1 to 5, which is characterized in that described according to the judgment matrix weighted data
Collection and the variance weighted data collection enter modular character weight described in calculating, comprising:
Using it is following enter modular character weight calculation formula, enter modular character weight described in calculating;
Enter modular character weight calculation formula:
Wherein, the QkIt characterizes and enters modular character weight into modular character k;The WkThe judgment matrix weighted data is characterized to concentrate
Enter the corresponding judgment matrix weight of modular character k;PkThe variance weighted data is characterized to concentrate into the corresponding variance power of modular character k
Weight;WiIt characterizes the judgment matrix weighted data and concentrates i-th of judgment matrix weight;PiThe variance weighted data is characterized to concentrate
I-th of variance weight;N characterizes total number/variance weight that the judgment matrix weighted data concentrates judgment matrix weight
The total number of variance weight in data set;
And/or
Applied to electric business enterprise;
And/or
It is described enter modular character: include: run a shop duration, information integrity, to manage duration, change number, commodity satisfaction, service full
Meaning degree, asset item risk number, total assets, risk case number, judicial risk number, case-involving amount of funds, tax arrear number, sale
Quantity ranking list, business number industry ranking, industry weighting ranking, moon sales volume, investment number, ratio between investments, intangible asset
Any two or multiple in quantity and grading of paying taxes;
The class index, comprising: any one or more in stability, credit worthiness and loan repayment capacity.
7. a kind of enterprise's methods of marking that any enterprise's Rating Model constructed based on claim 1 to 6 is realized, feature exist
In, comprising:
Clean Target Enterprise data, and by the data after cleaning distribute to it is each enter modular character;
For the data after the corresponding cleaning of modular character are entered described in each, execute:
Data using power function mapping, after pre-processing the cleaning;
Using linear mapping function, will be distributed in the data after the pretreated cleaning stable data section mapping 0~
Between 100, the score of the stable data segment of distribution is determined;
Using sigmoid mapping function, by there are the data section mappings of wave crest 0 in the data after the pretreated cleaning
Between~100, determine that described there are the scores of the data segment of wave crest;
Using the score of the stable data segment of the distribution and described there are the score of the data segment of wave crest, enters mould described in calculating and refer to
Target average mark;
According to following enterprise's score calculation formula, the score of the Target Enterprise is calculated;
Enterprise's score calculation formula:
Wherein, the score of the F characterization Target Enterprise;QiI-th of modular character that enters for entering modular character is weighed in characterization enterprise's Rating Model
Weight;fiCharacterize i-th of average mark for entering modular character;The total number for entering modular character for including in n characterization enterprise's Rating Model.
8. a kind of enterprise's Rating Model construction device characterized by comprising basic data processing unit, index confirmation unit
And index weights computing unit, wherein
The basic data processing unit, for obtaining the representation data and management data of type of industry, to the portrait number
Dimension-reduction treatment is carried out according to the management data;
The index confirmation unit, for selecting at least two from the result for the dimensionality reduction that the basic data processing unit obtains
It is a enter modular character, enter modular character to described at least two and classify, construct at least one class index;
The index weights computing unit is executed for each class index for index confirmation unit building: according to
The class index include it is described enter modular character between important relationship, determine judgment matrix;The judgment matrix is returned
One change processing determines the corresponding judgment matrix weighted data collection of the judgment matrix, the judgment matrix weighted data collection, packet
Include: the class index include it is each enter the corresponding judgment matrix weight of modular character;Using variance method, the class index pair is calculated
The variance weighted data collection answered, the corresponding variance weighted data collection of the class index, comprising: the class index include it is each enter
The corresponding variance weight of modular character;According to the judgment matrix weighted data collection and the variance weighted data collection, described in calculating
Enter modular character weight.
9. device according to claim 8, which is characterized in that the index weights computing unit, comprising: average mark calculates
Subelement, first variance computation subunit and variance weight sets construct subelement, wherein
The basic data processing unit is further used for pre-processing the representation data and the warp using power function mapping
Seek data;Using linear mapping function, by the pretreated representation data and the pretreated management data points
The stable data section mapping of cloth determines the score of the stable data segment of distribution between 0~100;It is mapped using sigmoid
Function, by there are the data section mappings of wave crest to exist in the pretreated representation data and the pretreated management data
Between 0~100, determine that described there are the scores of the data segment of wave crest;
The average mark computation subunit, each of described class index for being constructed according to index confirmation unit institute
State the score of the stable data segment of the distribution determined into the corresponding basic data processing unit of modular character and described
There are the score of the data segment of wave crest, calculates and enter the corresponding average mark of modular character described in each;
The first variance computation subunit, for according to every in the calculated class index of the average mark computation subunit
The average mark for entering modular character described in one calculates to enter between corresponding first intra-class variance of modular character and first group described in each
Variance;
The variance weight sets constructs subelement, for according to the first variance computation subunit it is calculated it is described enter mould refer to
Corresponding first intra-class variance and the first between-group variance are marked, the corresponding first intra-class variance weight of modular character and are entered described in calculating
One between-group variance weight, it is described enter the corresponding variance weight of modular character be it is described enter corresponding first intra-class variance of modular character
The sum of weight and the first between-group variance weight;Enter the variance weight combination of modular character described in will be all in the class index,
As the corresponding variance weighted data collection of the class index.
10. device according to claim 9, which is characterized in that
The index weights computing unit further comprises: judgment matrix confirms subelement,
The judgment matrix confirms subelement, and the class index for being determined according to the basic data processing unit includes
Each described in enter the score of the stable data segment of the corresponding distribution of modular character and described there are the data segment of wave crest
Score calculates and enters the corresponding average mark of modular character described in each;From high to low according to the average mark, enter described in determining
The significance level of modular character is from high to low;According to it is described enter modular character significance level from high to low, for it is described enter modular character point
With 1~9 score value;Using it is described enter the corresponding score value of modular character, determine judgment matrix;
And/or
The index weights computing unit further comprises: judge that weight sets constructs subelement,
The judgement weight sets constructs subelement, for by belong in the judgment matrix it is same it is described enter modular character score value
Form a group;Calculate each described group mean value, the second intra-class variance and the second between-group variance;According to the mean value, institute
The second intra-class variance and second between-group variance are stated, it is public using intra-class variance weight calculation formula and between-group variance weight calculation
Formula calculates the second intra-class variance weight and the second between-group variance weight;
Intra-class variance weight calculation formula:
Wherein, the wkCharacterize the corresponding intra-class variance weight of kth group in the judgment matrix;σkIt characterizes in the judgment matrix
The corresponding intra-class variance of kth group;σiCharacterize i-th group of corresponding intra-class variance in the judgment matrix;N characterizes the judgment matrix
Middle group of total number;
Between-group variance weight calculation formula:
Wherein, the skCharacterize the corresponding between-group variance weight of kth group in the judgment matrix;μiIt characterizes in the judgment matrix
I-th group of corresponding mean value;μ characterizes the mean value of the judgment matrix;pkiIt characterizes in the judgment matrix between kth group and i-th group
Between-group variance;N characterizes the total number organized in the judgment matrix;
Using the second intra-class variance weight and the second between-group variance weight, described group of population variance weight is determined;It will
Corresponding all described groups of the population variance weight combination of the judgment matrix, becomes the corresponding weighted data of the judgment matrix
Collection;
And/or
The index weights computing unit further comprises: weight calculation subelement,
The weight calculation subelement, for using it is following enter modular character weight calculation formula, enter modular character weight described in calculating;
Enter modular character weight calculation formula:
Wherein, the QkIt characterizes and enters modular character weight into modular character k;The WkThe judgment matrix weighted data is characterized to concentrate
Enter the corresponding judgment matrix weight of modular character k;PkThe variance weighted data is characterized to concentrate into the corresponding variance power of modular character k
Weight;WiIt characterizes the judgment matrix weighted data and concentrates i-th of judgment matrix weight;PiThe variance weighted data is characterized to concentrate
I-th of variance weight;N characterizes total number/variance weight that the judgment matrix weighted data concentrates judgment matrix weight
The total number of variance weight in data set;
And/or
Applied to electric business enterprise;
And/or
It is described enter modular character: include: run a shop duration, information integrity, to manage duration, change number, commodity satisfaction, service full
Meaning degree, asset item risk number, total assets, risk case number, judicial risk number, case-involving amount of funds, tax arrear number, sale
Quantity ranking list, business number industry ranking, industry weighting ranking, moon sales volume, investment number, ratio between investments, intangible asset
Any two or multiple in quantity and grading of paying taxes;
The class index, comprising: any one or more in stability, credit worthiness and loan repayment capacity.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910012442.0A CN109636244A (en) | 2019-01-07 | 2019-01-07 | Enterprise's Rating Model construction method, enterprise's methods of marking and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910012442.0A CN109636244A (en) | 2019-01-07 | 2019-01-07 | Enterprise's Rating Model construction method, enterprise's methods of marking and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109636244A true CN109636244A (en) | 2019-04-16 |
Family
ID=66058087
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910012442.0A Pending CN109636244A (en) | 2019-01-07 | 2019-01-07 | Enterprise's Rating Model construction method, enterprise's methods of marking and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109636244A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111930815A (en) * | 2020-06-22 | 2020-11-13 | 航天信息股份有限公司 | Method and system for constructing enterprise portrait based on industry attribute and business attribute |
CN113435746A (en) * | 2021-06-28 | 2021-09-24 | 平安银行股份有限公司 | User workload scoring method and device, electronic equipment and storage medium |
-
2019
- 2019-01-07 CN CN201910012442.0A patent/CN109636244A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111930815A (en) * | 2020-06-22 | 2020-11-13 | 航天信息股份有限公司 | Method and system for constructing enterprise portrait based on industry attribute and business attribute |
CN113435746A (en) * | 2021-06-28 | 2021-09-24 | 平安银行股份有限公司 | User workload scoring method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Galagedera et al. | Australian mutual fund performance appraisal using data envelopment analysis | |
Bulgurcu | Application of TOPSIS technique for financial performance evaluation of technology firms in Istanbul stock exchange market | |
Hull et al. | CVA and wrong-way risk | |
Usui | Policy adjustments to the oil boom and their evaluation: the Dutch disease in Indonesia | |
CN108090652A (en) | The method for building the electricity transaction index system based on big data technology | |
CN108399453A (en) | A kind of Electric Power Customer Credit Rank Appraisal method and apparatus | |
CN109191189A (en) | Power sales decontrol lower power customer value assessment method | |
CN107705207A (en) | Method, apparatus, equipment and the computer-readable storage medium that customer value is assessed | |
CN107833137A (en) | Quantization trading strategies generation method and device, equipment and storage medium based on multiple-objection optimization | |
Ngoc Nguyen et al. | Concentration and efficiency in the Vietnamese banking system between 1999 and 2009: A structural model approach | |
CN107808337A (en) | Factor Clustering and device, equipment and storage medium | |
Shukla | Analysing financial strength of public and private sector banks: a CAMEL approach | |
CN109636244A (en) | Enterprise's Rating Model construction method, enterprise's methods of marking and device | |
CN110415125A (en) | A kind of method, apparatus, equipment and computer storage medium carrying out credit rating to issuing subject online | |
Visalakshmi et al. | An integrated fuzzy DEMATEL-TOPSIS approach for financial performance evaluation of GREENEX industries | |
Siegmann et al. | Hedge fund innovation | |
Bluhm | Persistent liquidity shocks and interbank funding | |
CN108416673A (en) | Quantify hedging portfolio method of commerce, equipment and storage medium | |
Majumder et al. | Limitation of ARIMA in extremely collapsed market: A proposed method | |
Carriero et al. | Forecasting government bond yields with large Bayesian VARs | |
Rajan et al. | Misalignment of the Baht, trade imbalances and the crisis in Thailand | |
CN110298759A (en) | A kind of fund diagnostic method, device and computer readable storage medium | |
CN108446844A (en) | A kind of new energy power station assets value comprehensive estimation method and system | |
Cady | The equilibrium real exchange rate of the Malagasy franc: Estimation and assessment | |
Luo et al. | Downside Risk-Parity Portfolio. |
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 | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20190710 Address after: 214029 No. 999 Gaolang East Road, Binhu District, Wuxi City, Jiangsu Province (Software Development Building) 707 Applicant after: Chaozhou Zhuoshu Big Data Industry Development Co.,Ltd. Address before: 250100 S06 tower, 1036, Chao Lu Road, hi tech Zone, Ji'nan, Shandong. Applicant before: SHANDONG INSPUR CLOUD INFORMATION TECHNOLOGY Co.,Ltd. |
|
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190416 |