CN108564465A - A kind of enterprise credit management method - Google Patents
A kind of enterprise credit management method Download PDFInfo
- Publication number
- CN108564465A CN108564465A CN201810413760.3A CN201810413760A CN108564465A CN 108564465 A CN108564465 A CN 108564465A CN 201810413760 A CN201810413760 A CN 201810413760A CN 108564465 A CN108564465 A CN 108564465A
- Authority
- CN
- China
- Prior art keywords
- enterprise
- index
- credit
- value
- text
- 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
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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
A kind of enterprise credit management method, consider not only numerical indication, it is additionally contemplates that influence of the text index to manufacturing business's credit, and analysis actively is carried out to text and obtains the corresponding index value of text index, the evaluation index of abundant Credit Risk Assessment of Enterprise early warning, test for multi-collinearity processing also is carried out to all evaluation indexes of acquisition, obtain the public evaluation points of the quantity far below evaluation index, while ensure that the information for not losing too many original index, retain the information of its reflection as far as possible, faster and better to carry out early warning to Credit Risk Assessment of Enterprise.The present invention more science, obtain the credit value of enterprise in time, be reflected to enterprise itself and its stakeholder in time before credit risk occurs for enterprise, Informational support provided for its science decision.
Description
Technical field
The present invention relates to a kind of enterprise credit management method more particularly to a kind of manufacturing business's credit management methods.
Background technology
Credit risk identification refer to risk accidents generation before, people with various methods come system, continuously recognize
Know the potential cause that the various risks faced and analysis risk accidents occur.Credit crisis once occurs for enterprise, by band
Carry out huge harm, it is not only possible to cause enterprise itself to be absorbed in a series of predicament mires, can also give investor, the credits of enterprise
People, upstream and downstream supplier, enterprise staff and society etc. bring serious influence.
Pillar industry of the manufacturing industry as Chinese national economy, due to the variation of inside and outside economic situation, in recent years again and again
Credit risk occurs.In the debition of manufacturing industry and China's banking, manufacturing industry about occupies China's Commercial Bank 20%
The amount of the loan, non-performing asset rate is 2.6 times of the whole industry.Therefore, credit risk is carried out to manufacturing business to know
It not and manages, the stabilization and sustainable development to China's economy will play considerable effect.
At present for the research of Credit Risk Assessment of Enterprise primarily focus on the research to the financial data of listed company mostly with
One wealth year is research unit.Current enterprise risk early-warning system is by using initial single financial index gradually to enterprise
Debt paying ability, operation ability, profitability, the changeable figureofmerit of finance based on business growth ability, be re-introduced into cash flow management energy
Power assesses the risk tolerance of enterprise, but it is only merely to finance to assess obtained business risk ability to bear in this way
Judge, it is more unilateral, and the risk of enterprise only not only is from finance, further relates to other aspects.
In addition, usually Logit models or artificial nerve network model etc. is used to comment business risk in the prior art
Estimate, but Logit model calculating process is more complicated, and there are many approximate processings in calculating process, this is inevitably
Influence whether precision of prediction, the accuracy of finally obtained business risk data is not high, and the maximum of artificial nerve network model
The disadvantage is that the randomness of its work is stronger, if to obtain a preferable neural network structure, need artificially to go to debug, it is non-
Normal labor intensive and time, universality is not present, therefore the artificial nerve network model can not be promoted.
Invention content
A kind of enterprise credit management method of present invention offer, more science, the credit value for obtaining enterprise in time, are being looked forward to
Industry occurs to be reflected to enterprise itself and its stakeholder in time before credit risk, and Informational support is provided for its science decision.
In order to achieve the above object, the present invention provides a kind of enterprise credit management method, comprises the steps of:
Step S1, the corresponding finger of numerical indication is obtained from the numerical value report as multiple manufacturing business with reference to enterprise
Scale value;
Step S2, pair analyzed to obtain the corresponding index value of text index with reference to the associated text of enterprise;
Step S3, screening is carried out to all indexs according to multiple index values with reference to enterprise and show that M are suitable for evaluation ginseng
According to the evaluation index of the credit value of enterprise;
Step S4, the index value based on evaluation index carries out all evaluation indexes using factor analysis multiple conllinear
Property inspection processing, obtain N number of public evaluation points associated with evaluation index;
Step S5, based on reference to enterprise the corresponding index of N number of public evaluation points, index value and credit value,
Classification and Identification function is obtained using support vector machines;
Step S6, based on the current corresponding index of N number of public evaluation points of manufacturing business to be evaluated, index value according to
Current credit is calculated in Classification and Identification function;
Wherein, N, M are positive integer more than 1, and M > N.
The enterprise credit management method also comprises the steps of after step S6:
Step S7, the current credit of manufacturing business to be evaluated is exported;
Step S8, credit type corresponding with manufacturing business to be evaluated is judged according to current credit;
Step S9, it based on credit type and presets processing mode corresponding with credit type and is handled;
Step S10, manufacturing business to be evaluated and corresponding processing mode are stored.
Described includes with reference to enterprise:The manufacturing business of credit value termination, and credit value is steady in historical time section
Fixed manufacturing business;
The numerical value report is the financial statement with reference to enterprise;
The numerical indication includes:Financial category index;
The text index includes:Public sentiment class index and law class index;
The step S2 is comprised the steps of:
Step S2.1, text associated with manufacturing business's title is captured from the associated website of text index;
Step S2.2, the keyword in text is identified successively according to pre-stored keyword;
Step S2.3, text where judging keyword based on pre-defined rule is front description or negative description;
Step S2.4, it is counted to obtain predetermined time internal reference enterprise respectively and be referred to according to the judging result of judging unit
Mark relevant number front description or negatively described;
Step S2.5, according to index corresponding with the relevant front description of index or the number setting index negatively described
Value.
In the step S2.3, pre-defined rule is:When occurring the negative vocabulary of odd-times in the sentence where the keyword, sentence
Text where disconnected keyword is negative description;When continuously occurring the negative vocabulary of even-times in the sentence where the keyword, judge to close
Text where keyword is that front describes;When only there is front vocabulary in sentence where keyword, text where keyword is judged
It is described for front.
The step S3 is comprised the steps of:
Step S3.1, it is examined using K-S and individual normal distribution-test is carried out to each index;
If all variables each meet normal distribution, then it is assumed that these all indexs can also meet on the whole
Normal distribution;
Step S3.2, screening terminates to credit value manufacturing business and in historical time section credit value stabilization manufacture
Industry enterprise there are the indexs of significant difference.
The step S5 is comprised the steps of:
Step S5.1, public evaluation is calculated according to the evaluation of estimate of evaluation index corresponding with each public evaluation points
The corresponding public evaluation of estimate of the factor, then obtained public evaluation of estimate substituted into give training sample data TT=(x1, y1),
(x2, y2) ..., (xm, ym), x1 ∈ Rn, yi ∈ { -1 ,+1 } }, the x1 in (x1, y1) includes 7 with reference to enterprise
Public evaluation points, y1 indicate the manufacturing business (- 1) or be the credit in historical time section that the enterprise is credit value termination
The manufacturing business (+1) of value stabilization;
Step S5.2, the kernel function K (x1, x) chosen and penalty factor C is explored, builds and solves following optimize and ask
Topic:
To obtain excellent discriminant function, i.e. Classification and Identification function:
The present invention considers not only numerical indication, it is also contemplated that influence of the text index to manufacturing business's credit, and
Analysis actively is carried out to text and obtains the corresponding index value of text index, enriches the evaluation index of Credit Risk Assessment of Enterprise early warning, also
Test for multi-collinearity processing is carried out to all evaluation indexes of acquisition, obtains the public evaluation of the quantity far below evaluation index
The factor retains the information of its reflection, so as to more preferable as far as possible while ensure that the information for not losing too many original index
Early warning quickly is carried out to Credit Risk Assessment of Enterprise.The present invention more science, obtain the credit value of enterprise in time, sent out in enterprise
It is reflected to enterprise itself and its stakeholder in time before raw credit risk, Informational support is provided for its science decision.
Description of the drawings
Fig. 1 is a kind of flow chart of enterprise credit management method provided by the invention.
Fig. 2 is the schematic diagram of the numerical indication with reference to information storage part storage in the embodiment of the present invention.
Fig. 3 is the schematic diagram of the public sentiment class index with reference to information storage part storage in the embodiment of the present invention.
Fig. 4 is the schematic diagram of the law class index with reference to information storage part storage in the embodiment of the present invention.
Fig. 5 is the selection result schematic diagram of normal distribution screening unit in the embodiment of the present invention.
Fig. 6 is the selection result schematic diagram of conspicuousness distribution screening unit in the embodiment of the present invention.
Fig. 7 is the schematic diagram that test for multi-collinearity is handled in the embodiment of the present invention.
Fig. 8 is a kind of schematic diagram of enterprise credit management system provided by the invention.
Specific implementation mode
Below according to Fig. 1~Fig. 8, presently preferred embodiments of the present invention is illustrated.
As shown in Figure 1, the present invention provides a kind of enterprise credit management method, comprise the steps of:
Step S1, the corresponding finger of numerical indication is obtained from the numerical value report as multiple manufacturing business with reference to enterprise
Scale value;
Step S2, pair analyzed to obtain the corresponding index value of text index with reference to the associated text of enterprise;
Step S3, screening is carried out to all indexs according to multiple index values with reference to enterprise and show that M are suitable for evaluation ginseng
According to the evaluation index of the credit value of enterprise;
Step S4, the index value based on evaluation index carries out all evaluation indexes using factor analysis multiple conllinear
Property inspection processing, obtain N number of public evaluation points associated with evaluation index;
Step S5, based on reference to enterprise the corresponding index of N number of public evaluation points, index value and credit value,
Classification and Identification function is obtained using support vector machines;
Step S6, based on the current corresponding index of N number of public evaluation points of manufacturing business to be evaluated, index value according to
Current credit is calculated in Classification and Identification function;
Step S7, the current credit of manufacturing business to be evaluated is exported, wherein N, M are the positive integer more than 1, and M >
N;
Step S8, it is based on ad hoc rules and credit class corresponding with manufacturing business to be evaluated is judged according to current credit
Type;
Step S9, it based on credit type and presets processing mode corresponding with credit type and is handled;
Step S10, manufacturing business to be evaluated and corresponding processing mode are stored.
In the step S1, described includes with reference to enterprise:The manufacturing business of credit value termination, and in history
Between in section credit value stabilization manufacturing business.In the present embodiment, the manufacturing business of credit value termination has gone bankrupt
Manufacturing business, the manufacturing business of credit value stabilization is the variation of the credit value between 5 years less than predetermined in historical time section
The manufacturing business of value.
The numerical value report is the financial statement with reference to enterprise, and the numerical indication includes:Financial category index is such as schemed
Shown in 2, the financial category index include mainly the debt paying ability of enterprise, operation ability, working capital managerial ability, profitability,
Cost management ability, business growth ability and cash flow management ability, totally 28 indexs.
In the step S2, the text index includes:Public sentiment class index and law class index;Wherein, with carriage
The associated website of feelings class index is financial web site, and text associated with public sentiment class index is financial and economic news, as shown in figure 3,
Public sentiment class index includes mainly public sentiment number, public sentiment classification and public sentiment scoring etc., has 12 indexs altogether;With law class
The associated website of index is law court website, and text associated with law class index is legal documents, as shown in figure 4, law
Class index includes mainly case flow keyword score index, exposure desk keyword score index, law court's bulletin keyword score
Index, announcement of court session keyword score index execute and announce keyword score index, judgement document's keyword score index etc.,
There are 8 indexs altogether.
The step S2 is comprised the steps of:
Step S2.1, text associated with manufacturing business's title is captured from the associated website of text index;
Step S2.2, the keyword in text is identified successively according to pre-stored keyword;
Step S2.3, text where judging keyword based on pre-defined rule is front description or negative description;
Pre-defined rule is:When occurring the negative vocabulary of odd-times in the sentence where the keyword, text where keyword is judged
Negatively to describe;When continuously occurring the negative vocabulary of even-times in the sentence where the keyword, text is just where judging keyword
Face describes;When only there is front vocabulary in sentence where keyword, text describes for front where judging keyword;
Step S2.4, it is counted to obtain predetermined time internal reference enterprise respectively and be referred to according to the judging result of judging unit
Mark relevant number front description or negatively described;
Step S2.5, according to index corresponding with the relevant front description of index or the number setting index negatively described
Value.
The step S3 is comprised the steps of:
Step S3.1, it is examined using K-S and individual normal distribution-test is carried out to each index;
If all variables each meet normal distribution, then it is assumed that these all indexs can also meet on the whole
Normal distribution;
As shown in figure 5, use Z statistics in the present embodiment, using Z values and its corresponding probability P value as reference according to
According to it is 0.05 to give preset value, when followed probability P values>When 0.05, it is believed that the overall distribution of sample is consistent with normal distribution, sample
This totality meets normal distribution;Conversely, working as it<When 0.05, it is believed that there are conspicuousnesses with normal distribution for collection sample overall distribution
Difference, sample do not meet normal distribution on the whole;In the index for 48 Credit Risk Assessment of Enterprise early warning that this experiment is chosen,
There are the followed probability P values of 45 indexs less than given preset value 0.05, only 3 index followed probability P values are higher than the value, i.e.,
3 indexs do not meet normal distribution, therefore screen out 3 indexs;
Step S3.2, the manufacturing business's (credit risk is relatively low) and believe in historical time section that screening terminates to credit value
With manufacturing business's (credit risk is higher) of value stabilization there are the indexs of significant difference;
If significant difference is not present in index, those indexs are rejected, so as to reduce it to predicting mould
The influence of the accuracy rate of type;
As shown in fig. 6, this experiment is using poor with the presence or absence of conspicuousness two groups of variables of Wilcoxon signed rank tests between
It is different to test;In original 48 provided index, in 0.1 given significance, shares 31 indexs and pass through
It examines, shares 31 indexs and pass through inspection, i.e. X1、X2、X3、X4、X5、X6、X8、X9、X11、X14、X15、X17、X18、X20、X23、X24、
X26、X27、X29、X30、X31、X34、X36、X37、X38、X40、X42、X43、X44、X45、X47, therefore, this 31 indexs are to be suitable for evaluation
With reference to the evaluation index of the credit value of enterprise, remaining 17 index does rejecting processing.
As shown in fig. 7, in the step S4, multicollinearity inspection is carried out using 31 evaluation indexes of factor analysis pair
Processing is tested, finally obtains 7 public evaluation points associated with evaluation index;This 7 public evaluation points may be summarized to be
The debt paying ability factor, the working capital managerial ability factor, cost-cash management capability facfor, the profit-business growth ability factor, carriage
Feelings Quantitative factor, the public sentiment scoring factor and the law scoring factor.
The step S5 is comprised the steps of:
Step S5.1, public evaluation is calculated according to the evaluation of estimate of evaluation index corresponding with each public evaluation points
The corresponding public evaluation of estimate of the factor, then obtained public evaluation of estimate substituted into give training sample data TT=(x1, y1),
(x2, y2) ..., (xm, ym), x1 ∈ Rn, yi ∈ { -1 ,+1 } }, the x1 in (x1, y1) includes 7 with reference to enterprise
Public evaluation points, y1 indicate the manufacturing business (- 1) or be the credit in historical time section that the enterprise is credit value termination
The manufacturing business (+1) of value stabilization;
Step S5.2, the kernel function K (x1, x) chosen and penalty factor C is explored, builds and solves following optimize and ask
Topic:
To obtain excellent discriminant function, i.e. Classification and Identification function:
In the step S8, when f (x)=1, then credit type is divided in y=+1 this classification (i.e., corresponding to look forward to
The smaller company of industry credit risk);If f (x)=- 1, credit type is divided in y=-1 this classification (i.e., corresponding to look forward to
The larger company of industry credit risk).
As shown in figure 8, the present invention also provides a kind of enterprise credit management systems, including:
With reference to information storage part 11, wherein being stored with the reference enterprise name as multiple manufacturing business with reference to enterprise
Claim and refers to reference to the corresponding credit value of enterprise name, multiple numerical indications associated with credit value and text with this
Mark;
Numerical indication acquisition unit 12 obtains the corresponding index value of numerical indication from the numerical value report with reference to enterprise;This
In embodiment, numerical indication acquisition unit 12 can only obtain the corresponding index of financial category index from the financial statement with reference to enterprise
Value can be combined with obtaining index corresponding with financial category index with reference in the financial statement of enterprise and the financial statement of bank
Value;
Text index analysis portion 13, pair is analyzed to obtain text index pair with reference to the associated text of enterprise
The index value answered;
Index screening portion 14 carries out screening to all indexs according to multiple index values with reference to enterprise and show that M are applicable in
In evaluation with reference to the evaluation index of the credit value of enterprise;
Test of linearity portion 15, the index value based on evaluation index carry out all evaluation indexes using factor analysis
Test for multi-collinearity processing, obtains N number of public evaluation points associated with evaluation index;
Classification function establishes portion 16, based on reference to enterprise the corresponding index of N number of public evaluation points, index value
And credit value, Classification and Identification function is obtained using support vector machines;
Credit value calculating part 17, based on the current corresponding index of N number of public evaluation points of manufacturing business to be evaluated,
Current credit is calculated according to Classification and Identification function in index value;
As a result output section 18 export the current credit of manufacturing business to be evaluated;
Credit type determination unit 19 is judged and manufacturing business's phase to be evaluated based on ad hoc rules according to current credit
Corresponding credit type;
Processing unit 20 based on credit type and is preset at processing mode corresponding with credit type
Reason;For example, when credit type determination unit 19 judge manufacturing business to be evaluated for Credit Risk Assessment of Enterprise smaller company when, processing
Portion 20 carries out proposing volume processing;When credit type determination unit 19 judges manufacturing business to be evaluated for the larger public affairs of Credit Risk Assessment of Enterprise
When department, processing unit 20 carries out that money is urged to handle;
Record storage portion 21 is handled, manufacturing business to be evaluated and corresponding processing mode are stored;
Temporary storage part 22 is used between each component part of temporarily storage manufacturing business credit management system be exchanged
The data such as data information, such as numerical indication, text index, index value, evaluation index, public evaluation points;
Control unit 23 is used to control with reference to information storage part 11, numerical indication acquisition unit 12, text index analysis portion
13, index screening portion 14, test of linearity portion 15, classification function establish portion 16, credit value calculating part 17, result output section 18, letter
With the operation of type decision portion 19, processing unit 20, processing record storage portion 21 and temporary storage part 22.
Further, the text index analysis portion 13 includes:
Keyword storage unit 131 stores keyword corresponding with text index;Such as it is opposite with public sentiment class index
The keyword answered has the media event that enterprise name, abbreviation, enterprise are likely to occur, such as arrears of wages, tax, fraud, finance
In violation of rules and regulations, service exception, policy pressure, labour dispute, Financing of illegal activities, outside debt promise breaking, great loss, failure in investment, nature
Disaster influence, customer complaint, rectification, single, the bankruptcy recombination of product variety etc.;Keyword corresponding with law class index has enterprise
The full name of industry, the case-involving classification of enterprise, case-involving amount of money etc.;The keyword of different classifications is arranged different keyword weights
Dictionary, according to law court's document points-scoring system to enterprise case flow, exposure desk, law court's bulletin, announcement of court session, execute bulletin,
Related merit occurred in judgement document etc.;
Text placement unit 132, it is associated with manufacturing business's title from the associated website crawl of text index
Text;
Keyword recognition unit 133 identifies the key in text according to the keyword in keyword storage unit successively
Word;
Judging unit 134, text where judging keyword based on pre-defined rule is front description or negative description;
Statistic unit 135 is counted to obtain predetermined time internal reference enterprise according to the judging result of judging unit respectively
Industry and relevant number front description or negatively described of index;
Index value setup unit 136 sets index according to the relevant front description of index or the number negatively described
Corresponding index value.
The index screening portion 14 includes:
Normal distribution screening unit 141 is examined using K-S and carries out individual normal distribution-test to each index;
Conspicuousness is distributed screening unit 142, screens in 48 Raw performances for the manufacturing business of credit value termination
(credit risk is relatively low) and this two class of manufacturing business's (credit risk is higher) of credit value stabilization is looked forward in historical time section
There are the indexs of significant difference between industry.
Experimental result is to when analyzing
This experiment comments manufacturing business's credit using financial category index, public sentiment class index and law class index
Estimate, and is compared from accuracy, rate of false alarm and rate of failing to report.Accuracy strictly describes system to fine or not sample on the whole
Recognition capability.Rate of false alarm describes system by the wrong probability that good pattern representation is bad sample, for bank, if being made
Subsequent post-loan management scheme is carried out for bad client, and the user experience for being likely to result in hospitable family is poor, so as to cause client
It is lost in.Rate of failing to report features the wrong probability that bad sample is judged by accident preferably sample by categorizing system, and for bank, this may lead
It causes that potential bad client is not identified and carries out follow-up post-loan management scheme well, once these bad clients believe
With risk, bank will face serious consequence caused by its promise breaking, will be larger strike for bank, may lead to its bad loan
The rising of money rate.
Experiment stays one every time using the method for 10 foldings crosscheck that is, by the way that training set is divided into 10 parts of equal samples
Part is used as test data, other data to participate in the training of system, different parameters are indicated according to the optimal performance after 10 iteration
The predictive ability of lower system.The quasi- penalty coefficient CC and Gaussian kernel by the way of exponentially-increased in initialization system of experiment
Parameter σ σ in function, i.e.,:CC={ 2-5,2-4,2-3 ..., 25 }, σ σ={ 2-10,2-9,2-8 ..., 25 }, for each net
Lattice node calculates the maximum accuracy of its test set according to the method for traversal.When penalty coefficient is { 2-3,2-2 } and Gaussian kernel
When parameter in function is { 2-7,2-6 }, the operation result of the system of support vector machines is best, and accuracy is 82.04% (table
1)。
Table 1
Credit management system commonly used in the prior art is to rely on BP neural network method, and the system is for the pre- of sample
Performance is surveyed, there are close relationships with parameter setting therein, and historical experience and theory are not merely leaned in the setting of parameter
According to good effect can be reached, also need the multiple training experiment by BP neural network, carry out multiple trial and error,
After the cyclic process of feedback etc., model has reached convergence state, and the parameter inside BP neural network model obtained at this time is only
Best parameter group, i.e. Transfer Parameters, the number of hidden nodes, training function, learning rate, iterations etc..
Sample after treatment, is carried out 10 folding cross validations by the credit management system for relying on BP neural network, to
The training of BP neural network and the search of optimal the number of hidden nodes, and comprehensive accuracy and convergence rate are carried out, is classified
Result (table 2) when best performance.
Table 2
As shown in table 3, manufacturing business's credit management system based on the present embodiment and rely in the prior art BP god
The data of credit management system through network are compared, it can be seen that in accuracy, the manufacturing business of the present embodiment believes
Slightly above rely on the credit management system of BP neural network, the most optimal sorting of the two in the prior art with the accuracy of management system
Class accuracy has respectively reached 82.04% and 80.63%.
The credit management system of BP neural network is relied in the prior art in its multiple classification results, and classification is accurate
Manufacturing business's credit management system of true property variance ratio the present embodiment is big.
Manufacturing business's credit management system of the present embodiment is superior in terms of rate of failing to report and rate of false alarm based on BP nerves
The Credit Risk Assessment of Enterprise identification model of network, wherein the optimal rate of false alarm of manufacturing business's credit management system of the present embodiment
The absolute value of the optimal rate of false alarm low 0.86% of the credit management system of BP neural network, i.e. effect are relied on than in the prior art
Promote 4.7%.On rate of failing to report, manufacturing business's credit management system of the present embodiment compared with the existing technology in rely on BP
For the credit management system of neural network, slightly has advantage.
The comparison result of table 3 SVM and BP neural network
Comparative run | Optimal accuracy | 10 prediction accuracy variances | Optimal rate of false alarm | Optimal rate of failing to report |
The present embodiment | 82.04% | 3.31E-06 | 17.95% | 18.19% |
BP neural network | 80.63% | 1.16E-05 | 27.27% | 19.05% |
To sum up, experiments have shown that manufacturing business's credit management system of the present embodiment compared with the existing technology in rely on BP
For the credit management system of neural network, have better estimated performance, more meets bank and enterprise itself in advance
The correct demand for finding credit risk.
The present invention considers not only numerical indication, it is also contemplated that influence of the text index to manufacturing business's credit, and
Analysis actively is carried out to text and obtains the corresponding index value of text index, enriches the evaluation index of Credit Risk Assessment of Enterprise early warning, also
Test for multi-collinearity processing is carried out to all evaluation indexes of acquisition, obtains the public evaluation of the quantity far below evaluation index
The factor retains the information of its reflection, so as to more preferable as far as possible while ensure that the information for not losing too many original index
Early warning quickly is carried out to Credit Risk Assessment of Enterprise.The present invention more science, obtain the credit value of enterprise in time, sent out in enterprise
It is reflected to enterprise itself and its stakeholder in time before raw credit risk, Informational support is provided for its science decision.
Although present disclosure is discussed in detail by above preferred embodiment, but it should be appreciated that above-mentioned
Description is not considered as limitation of the present invention.After those skilled in the art have read the above, for the present invention's
A variety of modifications and substitutions all will be apparent.Therefore, protection scope of the present invention should be limited to the appended claims.
Claims (10)
1. a kind of enterprise credit management method, which is characterized in that comprise the steps of:
Step S1, the corresponding index of numerical indication is obtained from the numerical value report as multiple manufacturing business with reference to enterprise
Value;
Step S2, pair analyzed to obtain the corresponding index value of text index with reference to the associated text of enterprise;
Step S3, screening is carried out to all indexs according to multiple index values with reference to enterprise and show that M are suitable for evaluation with reference to enterprise
The evaluation index of the credit value of industry;
Step S4, the index value based on evaluation index carries out multicollinearity inspection using factor analysis to all evaluation indexes
Processing is tested, obtains N number of public evaluation points associated with evaluation index;
Step S5, based on reference to enterprise the corresponding index of N number of public evaluation points, index value and credit value, use
Support vector machines obtains Classification and Identification function;
Step S6, based on the current corresponding index of N number of public evaluation points of manufacturing business to be evaluated, index value according to classification
Current credit is calculated in recognition function;
Wherein, N, M are positive integer more than 1, and M > N.
2. enterprise credit management method as described in claim 1, which is characterized in that described includes with reference to enterprise:Credit value
The manufacturing business of termination, and in historical time section credit value stabilization manufacturing business;
The numerical value report is the financial statement with reference to enterprise;
The numerical indication includes:Financial category index;
The text index includes:Public sentiment class index and law class index;
3. enterprise credit management method as claimed in claim 2, which is characterized in that the step S2 is comprised the steps of:
Step S2.1, text associated with manufacturing business's title is captured from the associated website of text index;
Step S2.2, the keyword in text is identified successively according to pre-stored keyword;
Step S2.3, text where judging keyword based on pre-defined rule is front description or negative description;
Step S2.4, according to the judging result of judging unit counted to obtain respectively the predetermined time internal reference enterprise with index phase
The number that the front of pass describes or negatively describes;
Step S2.5, according to index value corresponding with the relevant front description of index or the number setting index negatively described.
4. enterprise credit management method as claimed in claim 3, which is characterized in that in the step S2.3, pre-defined rule
For:When occurring the negative vocabulary of odd-times in the sentence where the keyword, text where judging keyword is negatively to describe;Work as key
When continuously occurring the negative vocabulary of even-times in sentence where word, text where judging keyword is front description;When keyword institute
When only occurring front vocabulary in sentence, text describes for front where judging keyword.
5. enterprise credit management method as claimed in claim 4, which is characterized in that the step S3 is comprised the steps of:
Step S3.1, it is examined using K-S and individual normal distribution-test is carried out to each index;
If all variables each meet normal distribution, then it is assumed that these all indexs can also meet normal state on the whole
Distribution;
Step S3.2, screening terminate to credit value manufacturing business and in historical time section credit value stabilization manufacturing industry enterprise
Industry there are the indexs of significant difference.
6. enterprise credit management method as claimed in claim 5, which is characterized in that the step S5 is comprised the steps of:
Step S5.1, public evaluation points are calculated according to the evaluation of estimate of evaluation index corresponding with each public evaluation points
Corresponding public evaluation of estimate, then obtained public evaluation of estimate substituted into give training sample data TT=(x1, y1), (x2,
Y2) ..., (xm, ym), x1 ∈ Rn, yi ∈ { -1 ,+1 } }, the x1 in (x1, y1) includes that 7 of a reference enterprise are public
Evaluation points, y1 indicate the manufacturing business (- 1) that the enterprise is credit value termination or are that credit value is steady in historical time section
Fixed manufacturing business (+1);
Step S5.2, the kernel function K (x1, x) chosen and penalty factor C is explored, builds and solves following optimization problem:
To obtain excellent discriminant function, i.e. Classification and Identification function:
7. enterprise credit management method as claimed in claim 6, which is characterized in that the enterprise credit management method is in step
It is also comprised the steps of after rapid S6:Step S7, the current credit of manufacturing business to be evaluated is exported.
8. enterprise credit management method as claimed in claim 7, which is characterized in that the enterprise credit management method is in step
It is also comprised the steps of after rapid S7:Step S8, letter corresponding with manufacturing business to be evaluated is judged according to current credit
Use type.
9. enterprise credit management method as claimed in claim 8, which is characterized in that the enterprise credit management method is in step
It is also comprised the steps of after rapid S8:Step S9, based on credit type and processing corresponding with credit type is preset
Mode is handled.
10. enterprise credit management method as claimed in claim 9, which is characterized in that the enterprise credit management method exists
It is also comprised the steps of after step S9:Step S10, manufacturing business to be evaluated and corresponding processing mode are deposited
Storage.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810413760.3A CN108564465A (en) | 2018-05-03 | 2018-05-03 | A kind of enterprise credit management method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810413760.3A CN108564465A (en) | 2018-05-03 | 2018-05-03 | A kind of enterprise credit management method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108564465A true CN108564465A (en) | 2018-09-21 |
Family
ID=63537878
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810413760.3A Pending CN108564465A (en) | 2018-05-03 | 2018-05-03 | A kind of enterprise credit management method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108564465A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110097252A (en) * | 2019-03-29 | 2019-08-06 | 五渡(杭州)科技有限责任公司 | A kind of risk recognition system and its method for supply chain finance |
CN111738581A (en) * | 2020-06-14 | 2020-10-02 | 湖南三湘银行股份有限公司 | Grading system suitable for heavy machinery manufacturing enterprises of bank system |
CN112232377A (en) * | 2020-09-22 | 2021-01-15 | 中财绿指(北京)信息咨询有限公司 | Method and device for constructing ESG (electronic service guide) three-excellence credit model of enterprise |
WO2021088820A1 (en) * | 2019-11-05 | 2021-05-14 | 深圳前海微众银行股份有限公司 | Post-loan information management method and apparatus, device, and storage medium |
CN113988890A (en) * | 2021-11-08 | 2022-01-28 | 大路网络科技有限公司 | Cross-border electric commerce and enterprise assessment method, device and equipment |
CN114066600A (en) * | 2021-11-11 | 2022-02-18 | 中国联合网络通信集团有限公司 | Method and device for analyzing risks after loan of agricultural loan |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105787073A (en) * | 2016-03-02 | 2016-07-20 | 上海源庐加佳信息科技有限公司 | Enterprise credit assessment method based on big data mining technology |
CN107437151A (en) * | 2017-08-08 | 2017-12-05 | 惠国征信服务股份有限公司 | Enterprise credit risk method |
-
2018
- 2018-05-03 CN CN201810413760.3A patent/CN108564465A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105787073A (en) * | 2016-03-02 | 2016-07-20 | 上海源庐加佳信息科技有限公司 | Enterprise credit assessment method based on big data mining technology |
CN107437151A (en) * | 2017-08-08 | 2017-12-05 | 惠国征信服务股份有限公司 | Enterprise credit risk method |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110097252A (en) * | 2019-03-29 | 2019-08-06 | 五渡(杭州)科技有限责任公司 | A kind of risk recognition system and its method for supply chain finance |
WO2021088820A1 (en) * | 2019-11-05 | 2021-05-14 | 深圳前海微众银行股份有限公司 | Post-loan information management method and apparatus, device, and storage medium |
CN111738581A (en) * | 2020-06-14 | 2020-10-02 | 湖南三湘银行股份有限公司 | Grading system suitable for heavy machinery manufacturing enterprises of bank system |
CN112232377A (en) * | 2020-09-22 | 2021-01-15 | 中财绿指(北京)信息咨询有限公司 | Method and device for constructing ESG (electronic service guide) three-excellence credit model of enterprise |
CN113988890A (en) * | 2021-11-08 | 2022-01-28 | 大路网络科技有限公司 | Cross-border electric commerce and enterprise assessment method, device and equipment |
CN114066600A (en) * | 2021-11-11 | 2022-02-18 | 中国联合网络通信集团有限公司 | Method and device for analyzing risks after loan of agricultural loan |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108564465A (en) | A kind of enterprise credit management method | |
Li et al. | DP-LSTM: Differential privacy-inspired LSTM for stock prediction using financial news | |
CN110852856B (en) | Invoice false invoice identification method based on dynamic network representation | |
CN108108352A (en) | A kind of enterprise's complaint risk method for early warning based on machine learning Text Mining Technology | |
Alden et al. | Detection of financial statement fraud using evolutionary algorithms | |
CN111461216A (en) | Case risk identification method based on machine learning | |
CN117455417B (en) | Automatic iterative optimization method and system for intelligent wind control approval strategy | |
CN112767136A (en) | Credit anti-fraud identification method, credit anti-fraud identification device, credit anti-fraud identification equipment and credit anti-fraud identification medium based on big data | |
CN109492097B (en) | Enterprise news data risk classification method | |
Si | Construction and application of enterprise internal audit data analysis model based on decision tree algorithm | |
Wang et al. | Credit fraud risk detection based on XGBoost-LR hybrid model | |
Tong et al. | A novel financial risk early warning strategy based on decision tree algorithm | |
Davalos et al. | A textual analysis of the US Securities and Exchange Commission's accounting and auditing enforcement releases relating to the Sarbanes–Oxley Act | |
Yang | A study on the impact of corporate financial accounting management system on corporate innovation under sustainable development strategy | |
Zhang et al. | Can sentiment analysis help mimic decision-making process of loan granting? A novel credit risk evaluation approach using GMKL model | |
CN108629507A (en) | A kind of enterprise credit management system | |
Chen et al. | Financial distress prediction using data mining techniques | |
CN116611911A (en) | Credit risk prediction method and device based on support vector machine | |
CN112580992B (en) | Illegal fund collecting risk monitoring system for financial-like enterprises | |
Hu | Design and application of a financial distress early warning model based on data reasoning and pattern recognition | |
Yang et al. | An evidential reasoning rule-based ensemble learning approach for evaluating credit risks with customer heterogeneity | |
Su et al. | Detection of tax arrears based on ensemble leaering model | |
Jan et al. | Detection of fraudulent financial statements using decision tree and artificial neural network | |
Zhang et al. | [Retracted] Analysis of Bank Customer Default Risk Based on Embedded Microprocessor Wireless Communication | |
Fan et al. | Unearthing Financial Statement Fraud: Insights from News Coverage Analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180921 |