CN110110981A - A kind of credit rating Default Probability estimates and method for prewarning risk - Google Patents
A kind of credit rating Default Probability estimates and method for prewarning risk Download PDFInfo
- Publication number
- CN110110981A CN110110981A CN201910344162.XA CN201910344162A CN110110981A CN 110110981 A CN110110981 A CN 110110981A CN 201910344162 A CN201910344162 A CN 201910344162A CN 110110981 A CN110110981 A CN 110110981A
- Authority
- CN
- China
- Prior art keywords
- credit
- probability
- default
- default probability
- data
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
-
- 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
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Technology Law (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to financial information data administrative skill field, discloses a kind of credit rating Default Probability and estimate and method for prewarning risk;Include: excavation valid data, calculates Default Probability;Loss given default prediction model is constructed, the consistency of rating result is examined;It constructs binary and responds risk warning model, realize " early warning " strategy by data mining technology.By the research and application to credit rating mathematical model, a set of credit system that can be strong with the feasibility of data real example is ultimately formed using data mining technology, and obtain effective practical experience.Research achievement can also provide credit product and technological service with credit business cooperation for it, provide data reference for the trade decision of main market players with financial industry cooperation simultaneously, to make contributions to credit demand and credit industries development.
Description
Technical field
The invention belongs to financial information data administrative skill field more particularly to a kind of credit rating Default Probability estimate with
Method for prewarning risk.
Background technique
Currently, the immediate prior art:
The key problem of credit rating is the research of Default Probability, is divided into the assessment of Default Probability estimated with Default Probability
Two aspects.The former is to be solved the problem of being how to calculate Default Probability;The research purpose of the latter is that analysis is determined, influenced
The correlative factor and its importance of promise breaking and Default Probability.
The measurement research of Default Probability can be classified as several classes: the credit grade promise breaking based on credit rating historical summary is general
Rate, " radix " rate of violation based on option pricing theory, estimating based on Insurance Actuarial Science Default Probability and be based on risk neutral
(RN) rate of violation in market is estimated.Representing research includes: the Creditmetrics mould that Default Probability is showed in the form of transfer matrix
Type;It is included in the Creditportfolio View model of Macroeconomic Factors design conditions Default Probability;It is managed using option valuation
The violation correction model of opinion --- credit monitor model (KMV);Mortality model based on life insurance thought and based on protecting
The Creditrisk+ model of dangerous actuarial Development of Framework.Default Probability evaluation studies mainly have following a few alanysis methods: single argument
Analysis, multi-variables analysis, modern Evaluation Method.It is typical for representing the research that research includes: " predicting failure in operation with financial ratios "
The Default Probability of representative assesses univariate analysis method;The ZETA Rating Model established with multivariate discriminant analysis thinking;With cluster
Analyze the credit evaluation model to consumption loan application person.Currently, assuming that event of default using more in positive research
Occur to obey Logit regression model and neural network model that Logistic is distributed.Using nerual network technique as the promise breaking of core
Probabilistic model do not require to make probability distribution it is assumed that and can it is deep enough excavate predictive variable between hide correlativity,
This just can effectively handle abnormal, nonlinear credit risk analysis problem.But neural network model is faced with again
Network structure is difficult to the problems such as determining and overfitting, largely influences its predictive ability and mould to new samples promise breaking
Type application.
With the continuous improvement that economic market requires credit system, the subdivision that deepens continuously of credit rating research, logarithm
Seem more urgent with the needs of efficient process according to effective collect.The update progress of data mining technology is dug according to data, analysis number
According to ability credit rating research in it is more and more important.Data mining refers to mentioning from a large amount of random real data
Take the information of potential using value and the process of knowledge.Its target is from mass data, and discovery is hidden in rule thereafter
Or the relationship between data, to serve decision.Data mining technology has converged database, artificial intelligence, statistics, visual
The difference subject such as change, parallel computation and crossing domain.Its classical processes divide are as follows: definition, data collection and the pre- place of problem
Reason, the execution of data mining algorithm and the explanation of result and assessment.Hypothesis item is needed not rely on since data mining has
Part is capable of handling the advantages that large-scale data and is widely applied in financial analysis, risk profile, is particularly concentrated on credit card
Audit, sequences, stock market analysis, the fields such as investment decision.Since finance data is mostly large sample, high-dimensional data, and data are dug
The big data of processing multivariable is good in pick, and it is scientific and reasonable, data that financial market risks are portrayed using data mining algorithm
Digging technology provides the technological means of numerous feasibilities for risk early-warning system.Therefore letter is examined by data mining technology
It is theoretical with grading, Risk Management System is improved, realizes Risk-warning strategy, mutually merged between theoretical research and science and technology,
It mutually promotes, is the development trend of credit rating research.
Financial circles occupy very important status in national economy, and financial system is improved and development degree direct relation
National economy can continue, healthily develop.And risk is as the inherent substantive characteristics of financial circles, also along with finance
Deepening development and it is more noticeable.Due to the stochastic volatility in financial market, the objective reality of financial risks has been expedited the emergence of pair
The demand of financial risk management.Important participant of the credit rating as financial risk management, plays specialized risk and takes off
It is shown as using.Credit rating is the assessment of ability and trusted degree that business such as from about repays capital with interest on time of being in debt debt main body,
It is the evaluation to debt repayment risk, basic goal is to disclose the credit risk of loan floatation people, reduces transaction cost and investment
Risk.
Credit rating industry in China's is established with market economic system, is gradually grown up from late 1980s.By
Nearly development in 30 years, although the credit rating industry in China achieves significant progress, with the international same trade compared to still in
Step section, due to rating model is not mature enough and rating technique falls behind relatively etc., faced in credit rating development process with
The problem of lower three aspects:
1. history rate of violation and same period Default Probability are fitted undesirable due to China's reform of the financing system and bond market
Development, the credit activity of economic entity is increased, produces huge database size, excavates valid data statistical history and disobeys
About rate difficulty increases.Again because rating system, method or the practical operation of rating organization are not sufficiently stable, additional macroeconomic environment
Etc. factors influence, the data of acquisition have very strong randomness, this proposes new challenge to measuring and calculating Default Probability.Fact hair
Existing, the true Default Probability of the rate of violation and the same period that are counted according to history credit data has differences, truthful data with it is original
Default Probability measure model fitting it is undesirable.And Default Probability is to calculate the key problem of credit risk.Rating organization is directed to
The debt subject definition of different attribute goes out corresponding credit rating grade, these definition are all based on Default Probability.Default Probability
Value makes credit rating result have more intuitive explanation strengths, while providing more accurately about the measurement of risk level,
It provides the foundation to carry out risk monitoring and control with management, a large amount of Credit Risk Model, which is all based on, calculates Default Probability work.Cause
This, Default Probability estimates the problem of primarily studying, and improves original measure model, reduces history rate of violation and same period Default Probability
Difference is very necessary.
2. rating result is lack of consistency inspection from 1841 first hand commercial credit rating organization Louis in the world
Tappan is set up so far in New York, and grading person, statistician, mathematician and software programmers create innumerable credit and comment
Grade model, develops various softwares of efficiently grading, and rating organization, China has also introduced many advanced grading softwares, but
It is in actual credit rating work, people's more attention is rating result, but lacks the consistency inspection to rating result
It tests and corrects.Its reason is that macroeconomy and financial environment variation are very fast, can inevitably go out between rating result and actual conditions
Existing deviation, and such deviation is objective reality.The difficult point of rating result consistency check be the credit quality of bond with
Consistency check between its credit grade.The key index of bond credit quality is the expected loss rate (expected loss of bond
Rate=Default Probability × loss given default), wherein influencing loss given default factor complexity, there are correlations between factor, therefore
Only using related Mathematical Method, science, which filters out, influences loss given default Main Factors, establishes prediction model, Cai Nengzhun
Really obtain expected loss rate.The consistency check between credit quality and credit grade is captured, rating result and grading mould are improved
The amendment of type makes science, correct credit rating judgement, is the essential link of credit system.
3. tracking grading operational difficulties, Risk-warning lag, tracking grading refer to that credit rating organization is complete in grading work
Cheng Hou still pays close attention to the associated change information of ratee, and its purpose is to ensure that the credit risk of ratee can be in item
Mesh obtains lasting tracking and announcement in the survival phase, so that rating result be avoided to fail.But at present at home, credit rating result
The range relatively narrower of application, and the variation of itself credit standing, credit rating organization are not concerned with after main body to be appraised acquisition rating result
The credit data of acquisition is confined to index number before grading, and later data missing is serious, leads to tracking grading operational difficulties, grading
The timeliness of information is low, Risk-warning lag.The building of risk early-warning system and realizing needs numerous subjects theories and technology
It participates in.Financial Engineering, mathematical model, financial management, computer and information technology etc. be all support risk early-warning system operation and
The important ambit of development.With the raising that financial investment market requires Risk-warning, credit risk early-warning index is improved
System, by corresponding data mining technology implementation " early warning " strategy be under big data era there is an urgent need to.
The participation of data mining technology for credit rating research provide more diversification data processing technique means and
Theory, rating system certainly will develop to multi-angle, multi-level direction.Mutually melt between credit rating and data mining technology
The trend close, mutually promoted will be further strengthened, and the seamless connection of the two is the desirability and scientific theory of reality
To the necessary process of practice operation conversion.
In conclusion existing credit rating technology is:
(1) enterprise self-determining develops Credit rating system, carries out Default Probability calculating to data using existing mathematical model;
(2) credit rating Factor system carries out quantitative analysis with mathematical model mostly based on quantitative data;
(3) most of Credit rating systems use principal component analytical method, carry out data drop to credit rating multiple factors
Credit grade sequence is done in dimension processing, reuse factor analysis;
(4) enterprise judges according to credit grade result.
In conclusion problem of the existing technology is:
(1) neural network model network structure is difficult to determining and overfitting, largely influences it to new samples
The predictive ability and model application of promise breaking.
(2) the true Default Probability of the rate of violation and the same period that are counted according to history credit data has differences, true number
It is undesirable according to being fitted with original Default Probability measure model.
(3) in actual credit rating work, people's more attention is rating result, is but lacked to rating result
Consistency check and amendment.
(4) later data missing is serious, leads to tracking grading operational difficulties, the timeliness of rating information is low, Risk-warning
Lag.
(5) original mathematical model seldom considers variable in the randomness in financial market, the rating system of rating organization, side
Method or practical operation are not sufficiently stable;
(6) some in actual credit rating work is qualitative index factor, if ignoring qualitative factor to grading
Influence, the accuracy that Default Probability is estimated can be reduced;
(7) Principal Component Analysis Algorithm can lose the information of a part of initial data during dimensionality reduction, by principal component
The data processing of parser, if contribution rate accounts for 85% or more, then system will be considered that subsequent data result has Statistical Value,
But the data information lost directly affects the accuracy and the degree of reliability of grading;
Solve the difficulty of above-mentioned technical problem:
(1) method of Default Probability p is estimated in research by rate of violation;
(2) principal element for influencing loss given default is determined;
(3) data mining technology of " early warning " strategy is realized.
Solve the meaning of above-mentioned technical problem:
(1) parameter lambda is added on the basis of original Moody's model, considers the random fluctuation of non-Credit Factors to separated in model
The about influence of probability;
(2) qualitative index factor is added, qualitative index area of feasible solutions is set, default loss comprehensive index system is established;
(3) credit rating mathematical model is improved, accuracy and the degree of reliability that Default Probability is estimated are improved;
(4) consistency check for improving evaluation result answers model foundation-model by rating result correction model algorithm
Become a closed loop with-Modifying model;
(5) grading early warning system is established using the classificating thought of the supervised learning in machine learning, uses data mining technology
Realize prediction policy.
Summary of the invention
In view of the problems of the existing technology, estimate the present invention provides a kind of credit rating Default Probability and Risk-warning
Method.
The invention is realized in this way a kind of credit rating Default Probability estimates and method for prewarning risk.
The Credit rating system is by Default Probability Likelihood Computation module, credit rating index system module, supervised learning
Mathematical model visualization model, data query and grading report generation module composition.
The Default Probability Likelihood Computation module can be realized data acquisition by data mining technology and Default Probability is surveyed
Degree calculates, load, parameter setting, Default Probability Likelihood Computation including data.
The credit rating index system module can be realized the preservation of qualitative index quantitative target source data, data column altogether
Linear analysis, the pretreatment of initial data, including data cleansing, data integration, data transformation, hough transformation and credit matter
The consistency check of amount and credit grade.
The supervised learning mathematical model visualization model can be realized the prediction of credit rating result, including logic is returned
Return, the comparison of the data result of support vector machines, neural network model, optimal rating result compared with Default Probability is estimated,
And the further revision of Moody's model.
The data query and grading report generation module can be realized data backup, data query, Print Preview function,
And later period promise breaking warning function.
The data query and grading report generation module further comprise:
Step 1: three factors (m (R), n (R), l (R)) for influencing rate of violation, setting are filtered out using data mining technology
Parameter lambda, λ ∈ [0,1] calculate rate of violation p (R) by improved Moody's model, utilize the χ of family of distributions2Fitness Test method is examined
P (R) probability distribution, estimates Default Probability by numerical characteristic, is that foundation does credit grade division, grade with Default Probability value
It is divided into M class, M is 1 to 5 integer.M=1 indicates that credit level AAA, M=2 indicate that credit level AA, M=3 indicate credit level A,
M=4 indicates that credit level BBB, M=5 indicate credit level BB.Default Probability value is bigger, and M value is bigger, and default risk is bigger.
Step 2: extracting the qualitative index data and quantitative target data of sample using data mining technology, save source number
According to the Pearson correlation coefficient for calculating data column analyzes synteny, sets correlation coefficient r, r ∈ (0,1), if gained Pearson came phase
Relationship number is greater than r, then is change of variable xij=xi/xj, eliminate synteny, then do data cleansing, data integration, data transformation,
The data predictions such as hough transformation.
Step 3: first by Default Probability value be foundation credit grade be set as same period credit grade, then with history credit
Grade does consistency check, if consistency check does not pass through, parameter lambda is revised suddenly back to step 1, if consistency check is logical
It crosses, is then the same period credit of foundation by Default Probability value using the factor for influencing LGD in Credit Appraisal Index System as explanatory variable
Grade is explained variable, distinguishes test sample using logistic regression, support vector machines, neural network model, relatively more accurate
Rate and recall rate visualize ROC curve and AUC value, and Systematic selection optimal models are as later period credit rating Early-warning Model.
Step 4: setting credit grade M=2, A class are devoid of risk credit level, and assignment 0, B class is risk credit level, is assigned
Value 1, building binary respond risk warning model, realize " early warning " strategy by data mining technology.Data backup generates credit
Grading visualization report, saves as WORD or PDF document, prints credit appraisal result.
The credit rating Default Probability is estimated includes: with method for prewarning risk
Step 1 excavates valid data, calculates Default Probability;
Step 2 constructs loss given default prediction model, examines the consistency of rating result;
Step 3, building binary respond risk warning model, realize " early warning " strategy by data mining technology.
Further, the step 1 excavates valid data, calculates the method for Default Probability are as follows: it is general to estimate promise breaking by rate of violation
Rate p.Due to being difficult to that the corresponding Default Probability of credit grade is accurately calculated in advance, estimate that promise breaking is general with rate of violation
Rate.Rate of violation refers to that debtor fails to provide to fulfil its finance obligation, that is, the actual history frequency broken a contract according to contract.And
Default Probability refers to debt human hair raw a possibility that breaking a contract within given period in future.By to rating organization's credit rating history
The tracking and analysis of data obtain rate of violation, estimate Default Probability from rate of violation probability distribution.
Further, the measuring and calculating Default Probability method specifically includes:
The first step is filtered out three factors (m (R), n (R), l (R)) for influencing rate of violation using data mining, calculated
Rate of violation p (R);
Second step utilizes the χ of family of distributions2Fitness Test method is examined p (R) probability distribution, is estimated by numerical characteristic separated
About probability.
Further, the measuring and calculating Default Probability first step specifically:
Moody's model dynamic group is improved, year rate of violation p (R) formula for obtaining the publisher that grade is R is as follows:
Wherein m (R): the number broken a contract in the publisher that grade is R;N (R): grade is original of the publisher of R
Number;L (R): grade is the number being revoked in the publisher of R due to non-credit related causes;λ: proportionality coefficient determines ratio system
The feasible section of number λ.
If 1) Default Probability does not change over time, for large sample, the rate of violation p (R) of ttIt is weighted and averaged with T
The approximate Normal Distribution of rate of violation p (R), have
Wherein p: Default Probability;mt: publisher's number of t;M:T adds up to publisher's number,
2) if Default Probability fluctuates at any time, but each annual oscillations is mutually indepedent, then
Wherein σ: fluctuating factor
3) if Default Probability fluctuates at any time, and fluctuates there are continuity, then the average rate of violation DR in the entire survival phase
Approximation obeys following normal distribution:
Wherein,
θ: different continuity degree.
Further, the measuring and calculating Default Probability second step specifically includes:
Using the 2 Fitness Test method of χ of family of distributions, p (R) probability distribution is examined, Default Probability is estimated by numerical characteristic.
IfDefault Probability p is then its mathematic expectaion, estimates p with maximum likelihood estimate.
Detailed process is as follows:
Null hypothesis H first0: p (R)tPossible probability density function is
μ, σ are obtained by maximum likelihood estimate again, estimated valueThen divide Ω, obtain
To event A1, A2..., Ak, calculate frequency fiWithTo calculateLevel of significance α is taken, ifThen refuse H0.If receiving H0, then Default Probability
Further, the step 2 constructs loss given default prediction model, and the consistency of rating result is examined to specifically include:
The consistency check of credit rating result refers to be existed between the main body Default Probability of floater and its credit grade
Consistency, there are consistency between the credit quality of bond and its credit grade.For floater, main body promise breaking is general
Consistency between rate and credit grade is not related to specific default loss, thus the main historical data broken a contract using main body come
It estimates rate of violation, and carries out consistency check.It collects, arrange, screening main body promise breaking historical data, establishing database is to examine to disobey
The about premise of probability and credit grade consistency.Quality and ranking consistence test stage, since the credit quality of bond includes
Factor is more, and internal relation is complicated, and the simple Default Probability for investigating bond is not sufficient to describe its actual credit performance, because
This, should be main investigation factor with expected loss rate (that is: Default Probability × loss given default) to the consistency check of bond.Its
The size of middle measuring and calculating loss given default LGD is not only influenced by debt main body oneself factor, but also with the tool of debt project
Body design is closely related, and the factor for influencing LGD includes project factor (such as liquidation priority, guarantee), company factor (as always
Assets and total liability etc.), industry choice (such as rate of recovery), macroeconomy periodic factors (such as economic indicator).Examine grading
As a result consistency, it is important to determine loss given default impact factor, establish the prediction model of loss given default, from bond
The angle inspection quality of credit quality and the consistency of grade.For the foundation of loss given default prediction model, can use because
Sub- analytic approach determines the principal element for influencing loss given default LGD, and between multiple regression data fitting and loss given default
Quantitative relation expected loss rate is predicted to predict loss given default with this, and examine the consistency of rating result.
Further, examining the consistency of rating result, there are two aspects: first is that Default Probability and credit grade consistency, two
It is credit quality and credit grade consistency.Difficult point is the consistency check between the credit quality of bond and its credit grade,
Reason is that credit quality contains Default Probability, severity of loss and grade and the factors such as shifts risk, internal relation
It is complicated.It therefore, is main investigation bond consistency with expected loss rate (expected loss rate=Default Probability × loss given default)
The factor of inspection, wherein research, the determining factor for influencing loss given default LGD just become the critical issue of research.
Further, the step 3 building binary responds risk warning model, realizes " early warning " by data mining technology
Strategy specifically includes:
The purpose for tracking main body to be appraised is in order to ensure the credit risk of ratee can be held within the project survival phase
Continuous announcement.If main body to be appraised has been tracked record a period of time, main economic activity data in this period are collected, if
Duration is t, then the probability that the main body is broken a contract before 1+t has much, this is the problem of risk probability is often answered.It is this kind of
Problem is it can be appreciated that the risk probability of duration t is exactly the risk broken a contract between t and 1+t.And survival analysis, then disclose by
Comment main body when may promise breaking.Risk and survivorship curve provide the fast of debt main body life cycle (meeting one's engagements the phase) to be appraised
According to.It is renewed after rating result announcement and establishes early stage risk warning model, detection promise breaking caution signal.
For the building and realization of binary response Early-warning Model, using a kind of special shape --- the needle of linear regression model (LRM)
To the logistic regression analysis method of qualitative variable.But linear regression problem is converted into logistic regression, can be with similar to simple shellfish
The thinking of this model of leaf carrys out estimated probability multiplied by a string of likelihoods, is then converted into probability.It constructs binary and responds Early-warning Model
Key point is that the probability of its promise breaking is estimated according to the duration of main body to be appraised.Since the value range of probability is 0~1, and probability
Range be 0~+∞, therefore utilizeProbability is converted into probability.Then the logarithm of probability is taken to generate one from negative
It is infinite to arrive just infinite function, regression equation is established as target variable using the logarithm of probability:
Default Probability of the obtained logical function for the duration calculates, the subsequent economic activity data of debt main body to be appraised
It can be fitted and be predicted by logistic regression analysis model.
In conclusion advantages of the present invention and good effect are as follows:
Calculating Default Probability is problem most crucial in Credit rating system, it is to quantify credit risk to input change
Amount, is of great significance to the management of credit risk.It is dug under big data background relevant to rate of violation effective according to going out
Data calculate history rate of violation;The dynamic group of Moody's model is improved, the relationship between rate of violation and Default Probability is studied,
Its reasonability is examined with real data, while determining the feasible section of proportionality coefficient;Utilize rate of violation probability distribution research
Default Probability Measurement Method estimates a possibility that main body to be appraised is broken a contract by Default Probability, reduces the credit risk for being exposed to promise breaking.
Constructing binary response risk warning model can be using credit evaluation system progress credit risk pre-control, when still having time
When taking measures to reduce exposure, the caution signal of promise breaking being detected, and issuing warning, effective dissolving debts main body to be appraised is non-
The generation voluntarily broken a contract makes loss caused by credit risk minimize degree.Construct binary respond risk warning model, be
The snapshot collection of all main bodys to be appraised of given point in time, detects whether t time of the main body to be appraised after the snapshot date breaks a contract, and utilizes
Main body to be appraised or lower assessment main body'choice " early warning " strategy that data mining technology closes on contract period, remind debt main body to be appraised to answer
The contract fulfiled reduces the risk of involuntary promise breaking with this.
By the research and application to credit rating mathematical model, being ultimately formed using data mining technology a set of can use number
The strong credit system of the feasibility factually demonstrate,proved, and obtain effective practical experience.Research achievement can be with financial row simultaneously
Industry cooperation can also provide credit product and technological service with credit business cooperation for it, mention for the trade decision of main market players
For data reference, to make contributions to credit demand and credit industries development.
Innovative point of the present invention is to create the effective way that Default Probability is estimated by rate of violation probability distribution, and uses data
It digs and realizes algorithm according to technology.Data mining technology and credit rating are theoretical to combine closely, make practice technology and scientific theory it
Between form the developing state for mutually merging, mutually promoting.The infiltration of data mining technology is the application and reality of credit Rating Model
Card becomes possibility, enhances the feasibility of model testing.On the basis of existing credit rating algorithm, improve, innovation credit
Risk warning model and related research and development, the test job for matching set of model, enable the corresponding credit quality of credit grade more section
It learns, accurately quantified.
(1) Yin Mudi model is added to parameter lambda, it is contemplated that the shadow that the randomness of non-Credit Factors estimates Default Probability
It rings, keeps model application more extensive;
(2) qualitative index is combined with quantitative target, is established credit appraisal factor index system, is examined credit appraisal
The factor examined is more fully;
(3) data are collected using data mining technology, under big data background, credit appraisal result is more true and reliable;
(4) keep the grading accuracy predicted higher using the supervised learning algorithm in machine learning, credit rating early warning
Strategy is more accurate.
Detailed description of the invention
Fig. 1 is Credit rating system structural schematic diagram provided in an embodiment of the present invention.
In figure: 1, Default Probability Likelihood Computation module;2, credit rating index system module;3, supervised learning mathematical model
Visualization model;4, data query and grading report generation module.
Fig. 2 is that credit rating Default Probability provided in an embodiment of the present invention is estimated and method for prewarning risk flow chart.
Fig. 3 is measuring and calculating Default Probability method flow diagram provided in an embodiment of the present invention.
Fig. 4 is that SVM confusion matrix training set provided in an embodiment of the present invention is schemed as the result is shown.
Fig. 5 is that SVM confusion matrix test set provided in an embodiment of the present invention is schemed as the result is shown.
Fig. 6 is that neural network confusion matrix training set provided in an embodiment of the present invention is schemed as the result is shown.
Fig. 7 is that neural network confusion matrix test set provided in an embodiment of the present invention is schemed as the result is shown.
Fig. 8 is that logistic regression confusion matrix test set provided in an embodiment of the present invention is schemed as the result is shown.
Fig. 9 is that neural network confusion matrix training set provided in an embodiment of the present invention is schemed as the result is shown.
Figure 10 is provided in an embodiment of the present invention
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
Application principle of the invention is described in detail with reference to the accompanying drawing.
As shown in Figure 1, Credit rating system provided in an embodiment of the present invention is by Default Probability Likelihood Computation module 1, credit
Standard diagrams system module 2, supervised learning mathematical model visualization model 3, data query and grading 4 structure of report generation module
At.
The Default Probability Likelihood Computation module 1 can be realized data acquisition by data mining technology and Default Probability is surveyed
Degree calculates, load, parameter setting, Default Probability Likelihood Computation including data.
The credit rating index system module 2 can be realized the preservation of qualitative index quantitative target source data, data column
The pretreatment of Collinearity Diagnosis Analysis, initial data, including data cleansing, data integration, data transformation, hough transformation and credit
The consistency check of quality and credit grade.
The supervised learning mathematical model visualization model 3 can be realized the prediction of credit rating result, including logic is returned
Return, the comparison of the data result of support vector machines, neural network model, optimal rating result compared with Default Probability is estimated,
And the further revision of Moody's model.
The data query and grading report generation module 4 can be realized data backup, data query, Print Preview function
Energy and later period promise breaking warning function.
The data query and grading report generation module 4 further comprise:
Step 1: three factors (m (R), n (R), l (R)) for influencing rate of violation, setting are filtered out using data mining technology
Parameter lambda, λ ∈ [0,1] calculate rate of violation p (R) by improved Moody's model, utilize the χ of family of distributions2Fitness Test method is examined
P (R) probability distribution, estimates Default Probability by numerical characteristic, is that foundation does credit grade division, grade with Default Probability value
It is divided into M class, M is 1 to 5 integer.M=1 indicates that credit level AAA, M=2 indicate that credit level AA, M=3 indicate credit level A,
M=4 indicates that credit level BBB, M=5 indicate credit level BB.Default Probability value is bigger, and M value is bigger, and default risk is bigger.
Step 2: extracting the qualitative index data and quantitative target data of sample using data mining technology, save source number
According to the Pearson correlation coefficient for calculating data column analyzes synteny, sets correlation coefficient r, r ∈ (0,1), if gained Pearson came phase
Relationship number is greater than r, then is change of variable xij=xi/xj, eliminate synteny, then do data cleansing, data integration, data transformation,
The data predictions such as hough transformation.
Step 3: first by Default Probability value be foundation credit grade be set as same period credit grade, then with history credit
Grade does consistency check, if consistency check does not pass through, parameter lambda is revised suddenly back to step 1, if consistency check is logical
It crosses, is then the same period credit of foundation by Default Probability value using the factor for influencing LGD in Credit Appraisal Index System as explanatory variable
Grade is explained variable, distinguishes test sample using logistic regression, support vector machines, neural network model, relatively more accurate
Rate and recall rate visualize ROC curve and AUC value, and Systematic selection optimal models are as later period credit rating Early-warning Model.
Step 4: setting credit grade M=2, A class are devoid of risk credit level, and assignment 0, B class is risk credit level, is assigned
Value 1, building binary respond risk warning model, realize " early warning " strategy by data mining technology.Data backup generates credit
Grading visualization report, saves as WORD or PDF document, prints credit appraisal result.
Such as Fig. 2-3, credit rating Default Probability provided in an embodiment of the present invention is estimated includes: with method for prewarning risk
S101: excavating valid data, calculates Default Probability;
S102: building loss given default prediction model examines the consistency of rating result;
S103: building binary responds risk warning model, realizes " early warning " strategy by data mining technology.
Further, the step 1 excavates valid data, calculates the method for Default Probability are as follows: it is general to estimate promise breaking by rate of violation
Rate p.Due to being difficult to that the corresponding Default Probability of credit grade is accurately calculated in advance, estimate that promise breaking is general with rate of violation
Rate.Rate of violation refers to that debtor fails to provide to fulfil its finance obligation, that is, the actual history frequency broken a contract according to contract.And
Default Probability refers to debt human hair raw a possibility that breaking a contract within given period in future.By to rating organization's credit rating history
The tracking and analysis of data obtain rate of violation, estimate Default Probability from rate of violation probability distribution.
Further, the measuring and calculating Default Probability method specifically includes:
S201: three factors (m (R), n (R), l (R)) for influencing rate of violation are filtered out using data mining, are calculated separated
About rate p (R);
S202: the χ of family of distributions is utilized2Fitness Test method examines p (R) probability distribution, estimates promise breaking by numerical characteristic
Probability.
Further, the measuring and calculating Default Probability first step specifically:
Moody's model dynamic group is improved, year rate of violation p (R) formula for obtaining the publisher that grade is R is as follows:
Wherein m (R): the number broken a contract in the publisher that grade is R;N (R): grade is original of the publisher of R
Number;L (R): grade is the number being revoked in the publisher of R due to non-credit related causes;λ: proportionality coefficient, determine λ can
Row section;
If 1) Default Probability does not change over time, for large sample, the rate of violation p (R) of ttIt is weighted and averaged with T
The approximate Normal Distribution of rate of violation p (R), have
Wherein p: Default Probability;mt: publisher's number of t;M:T adds up to publisher's number,
2) if Default Probability fluctuates at any time, but each annual oscillations is mutually indepedent, then
Wherein σ: fluctuating factor
3) if Default Probability fluctuates at any time, and fluctuates there are continuity, then the average rate of violation DR in the entire survival phase
Approximation obeys following normal distribution:
Wherein,
θ: different continuity degree.
Further, the measuring and calculating Default Probability second step specifically includes:
Utilize the χ of family of distributions2Fitness Test method examines p (R) probability distribution, estimates Default Probability by numerical characteristic.
IfDefault Probability p is then its mathematic expectaion, can estimate p with maximum likelihood estimate.
Detailed process is as follows:
Null hypothesis H first0: p (R)tPossible probability density function is
μ, the estimated value of σ are obtained by maximum likelihood estimate againThen divide Ω, obtain
To event A1, A2..., Ak, calculate frequency fiWithTo calculateLevel of significance α is taken, ifThen refuse H0.If receiving H0, then Default Probability
Further, the step 2 constructs loss given default prediction model, and the consistency of rating result is examined to specifically include:
The consistency check of credit rating result refers to be existed between the main body Default Probability of floater and its credit grade
Consistency, there are consistency between the credit quality of bond and its credit grade.For floater, main body promise breaking is general
Consistency between rate and credit grade is not related to specific default loss, thus the main historical data broken a contract using main body come
It estimates rate of violation, and carries out consistency check.It collects, arrange, screening main body promise breaking historical data, establishing database is to examine to disobey
The about premise of probability and credit grade consistency.Quality and ranking consistence test stage, since the credit quality of bond includes
Factor is more, and internal relation is complicated, and the simple Default Probability for investigating bond is not sufficient to describe its actual credit performance, because
This, should be main investigation factor with expected loss rate (that is: Default Probability × loss given default) to the consistency check of bond.Its
The size of middle measuring and calculating loss given default LGD is not only influenced by debt main body oneself factor, but also with the tool of debt project
Body design is closely related, and the factor for influencing LGD includes project factor (such as liquidation priority, guarantee), company factor (as always
Assets and total liability etc.), industry choice (such as rate of recovery), macroeconomy periodic factors (such as economic indicator).Examine grading
As a result consistency, it is important to determine loss given default impact factor, establish the prediction model of loss given default, from bond
The angle inspection quality of credit quality and the consistency of grade.For the foundation of loss given default prediction model, can use because
Sub- analytic approach determines the principal element for influencing loss given default LGD, and between multiple regression data fitting and loss given default
Quantitative relation expected loss rate is predicted to predict loss given default with this, and examine the consistency of rating result.
Further, examining the consistency of rating result, there are two aspects: first is that Default Probability and credit grade consistency, two
It is credit quality and credit grade consistency.Difficult point is the consistency check between the credit quality of bond and its credit grade,
Reason is that credit quality contains Default Probability, severity of loss and grade and the factors such as shifts risk, internal relation
It is complicated.It therefore, is main investigation bond consistency with expected loss rate (expected loss rate=Default Probability × loss given default)
The factor of inspection, wherein research, the determining factor for influencing loss given default LGD just become the critical issue of research.
Further, the step 3 building binary responds risk warning model, realizes " early warning " by data mining technology
Strategy specifically includes:
The purpose for tracking main body to be appraised is in order to ensure the credit risk of ratee can be held within the project survival phase
Continuous announcement.If main body to be appraised has been tracked record a period of time, main economic activity data in this period are collected, if
Duration is t, then the probability that the main body is broken a contract before 1+t has much, this is the problem of risk probability is often answered.It is this kind of
Problem is it can be appreciated that the risk probability of duration t is exactly the risk broken a contract between t and 1+t.And survival analysis, then disclose by
Comment main body when may promise breaking.Risk and survivorship curve provide the fast of debt main body life cycle (meeting one's engagements the phase) to be appraised
According to.It is renewed after rating result announcement and establishes early stage risk warning model, detection promise breaking caution signal.
For the building and realization of binary response Early-warning Model, using a kind of special shape --- the needle of linear regression model (LRM)
To the logistic regression analysis method of qualitative variable.But linear regression problem is converted into logistic regression, can be with similar to simple shellfish
The thinking of this model of leaf carrys out estimated probability multiplied by a string of likelihoods, is then converted into probability.It constructs binary and responds Early-warning Model
Key point is that the probability of its promise breaking is estimated according to the duration of main body to be appraised.Since the value range of probability is 0~1, and probability
Range be 0~+∞, therefore utilizeProbability is converted into probability.Then the logarithm of probability is taken to generate one from negative
It is infinite to arrive just infinite function, regression equation is established as target variable using the logarithm of probability:
Default Probability of the obtained logical function for the duration calculates, the subsequent economic activity data of debt main body to be appraised
It can be fitted and be predicted by logistic regression analysis model.
In conclusion credit rating Default Probability is estimated and method for prewarning risk principle are as follows: firstly, collected using R language,
It arranges, storage debt principal sector of the economy activity initial data, establishes database.Secondly, being counted using list sample S-K checking computation Z
Amount and accordingly together probability P value, according to p value method principle come judge data from overall distribution situation, according to stochastic variable
The numerical characteristic and maximum likelihood estimate of rate of violation accordingly obtain the estimated value of parameter μ, and Default Probability p can be obtained.Again,
The software realization of loss given default prediction model follows factorial analysis principle and thinking, molecule score coefficient is obtained, in EViews
Regression model setting is carried out in statistical software and is examined, and predicts loss given default.And the software of binary response risk warning model
It realizes, condition early warning can be set and named first to qualitative variable again assignment 0 and 1, Test database promise breaking caution signal, and
To the main body to be appraised or lower assessment main body automatic prompt for reaching critical point of meeting one's engagements.
Main contents of the invention are all linked with one another, and regulations are clear, and scheme is reasonable.Research contents has certain depth and wide
Degree, the demonstration of research approach can reach the needs of current credit rating real work, to become compatible with internationally accepted practices.Data mining
Technical hardware and software are able to satisfy the needs of research and development with set of model, test job.The statistical software that this project is related to can be straight
The data for importing Excel format are connect, convenient for collecting data, establish database, and R language provides many higher management function
Can, it is capable of handling mass data.Flexibility, safety and ease for use of the invention provides good item for rating model programming
Part.
The present invention formulated reasonable project implementation schedule, selected the veteran research team of research and teaching,
Good cooperation relation is established, guarantees the smooth execution of project.This project associate professor group membership three, lecturer two, assiatant
One, constitute reasonable research echelon.Six members are college statistics practice instruction teacher, are engaged in random process, application
The theoretical research of statistics and real training practical work for many years, have very strong data-handling capacity and software editing ability.
Chongqing library, the second college of education has had subscribed the digital resources such as Chinese journals and foreign journals database, this is conducive to project
Group membership consults some relevant documents, provides for the implementation of this project and more complete looks into New Terms.Chongqing second is pedagogical to be learned
Institute's mathematics and information engineering establish statistical laboratory, are equipped with statisticians software, and develop gold cooperatively with Ruan Ku company
Melt Mathematical Experiments practical training project, provides higher platform for computer, mathematics, financial multi-crossed disciplines and innovation research.This is
The implementation of this project provides good soft and hardware support.Research achievement can further with credit rating organization, financial industry
Cooperation, provides credit product and service, provides data reference and suggestion for the trade decision of main market players.
The invention will be further described combined with specific embodiments below.
Embodiment:
1020 effective sample data of data mining, index for selection system partial data do binary classification prediction if M=2
It examines, obtains following result with support vector machines, neural network, Logic Regression Models test sample respectively:
(1) SVM operation result is analyzed, as shown in Fig. 4, Fig. 5 and table one, table two.
All kinds of evaluation criterion indexs of 1 SVM test set of table
All kinds of evaluation criterion indexs of 2 SVM training set of table
(2) neural network operation result is analyzed, as shown in Fig. 6, Fig. 7 and table three, table four.
All kinds of evaluation criterion indexs of 3 neural network test set of table
All kinds of evaluation criterion indexs of 4 neural metwork training collection of table
(3) logistic regression operation result is analyzed, as shown in Fig. 8, Fig. 9 and table five, table six.
All kinds of evaluation criterion indexs of 5 logistic regression test set of table
All kinds of evaluation criterion indexs of 6 logistic regression training set of table
SVM model, neural network model, Logic Regression Models algorithm are crossed it is recognised that in three kinds of algorithms, from data knot
Fruit sees, illustrates to establish the fitting of Logic Regression Models algorithm preferably, has statistical significance.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (8)
1. a kind of credit rating Default Probability estimates and method for prewarning risk, which is characterized in that the credit rating Default Probability
Estimate and includes: with method for prewarning risk
Step 1 excavates valid data, calculates Default Probability;
Step 2 constructs loss given default prediction model, examines the consistency of rating result;
Step 3, building binary respond risk warning model, realize " early warning " strategy by data mining technology.
2. credit rating Default Probability as described in claim 1 estimates and method for prewarning risk, which is characterized in that the step
One excavates valid data, calculates the method for Default Probability are as follows: estimate Default Probability p by rate of violation;Rate of violation refers to debtor not
It can provide to fulfil its finance obligation, that is, the actual history frequency broken a contract according to contract;Default Probability refers to given in future
Debt human hair raw a possibility that breaking a contract in period;It is disobeyed by the tracking to rating organization's credit rating historical data with analysis
About rate estimates Default Probability from rate of violation probability distribution.
3. credit rating Default Probability as described in claim 1 estimates and method for prewarning risk, which is characterized in that the measuring and calculating
Default Probability method specifically includes:
The first step filters out three factor m (R) for influencing rate of violation, n (R), l (R) using data mining, calculates rate of violation p
(R);
Second step examines p (R) probability distribution using the 2 Fitness Test method of χ of family of distributions, and it is general to estimate promise breaking by numerical characteristic
Rate.
4. credit rating Default Probability as claimed in claim 3 estimates and method for prewarning risk, which is characterized in that the measuring and calculating
The Default Probability first step specifically: improve Moody's model dynamic group, year rate of violation p (R) for obtaining the publisher that grade is R is public
Formula is as follows:
Wherein m (R): the number broken a contract in the publisher that grade is R;N (R): grade is the original number of the publisher of R;l
(R): grade is the number being revoked in the publisher of R due to non-credit related causes;λ: proportionality coefficient determines proportionality coefficient λ
Feasible section;
If 1) Default Probability does not change over time, for large sample, the rate of violation p (R) of ttIt is average weighted separated with T
About rate p (R) approximation Normal Distribution, has;
Wherein p: Default Probability;mt: publisher's number of t;M:T adds up to publisher's number,
2) if Default Probability fluctuates at any time, but each annual oscillations is mutually indepedent, then;
Wherein σ: fluctuating factor;
3) if Default Probability fluctuates at any time, and there are continuitys for fluctuation, then the average rate of violation DR in the entire survival phase is approximate
Obey following normal distribution:
Wherein;
θ: different continuity degree.
5. credit rating Default Probability as claimed in claim 3 estimates and method for prewarning risk, which is characterized in that the measuring and calculating
Default Probability second step specifically includes:
Using the 2 Fitness Test method of χ of family of distributions, p (R) probability distribution is examined, Default Probability is estimated by numerical characteristic, ifDefault Probability p is then its mathematic expectaion, estimates p with maximum likelihood estimate;
Detailed process is as follows:
Null hypothesis H first0: p (R)tPossible probability density function is;
μ, σ are obtained by maximum likelihood estimate again, estimated valueThen divide Ω, obtain thing
Part A1, A2..., Ak, calculate frequency fiWithTo calculateLevel of significance α is taken, ifThen refuse H0;If receiving H0, then Default Probability
6. credit rating Default Probability as described in claim 1 estimates and method for prewarning risk, which is characterized in that the step
Two building loss given default prediction models, examine the consistency of rating result to specifically include:
The consistency check of credit rating result, which refers between the main body Default Probability of floater and its credit grade, to be existed unanimously
Property, there are consistency between the credit quality of bond and its credit grade;For floater, main body Default Probability with
Consistency between credit grade is not related to specific default loss;It collects, arrange, screening main body promise breaking historical data, establishing
Database is the premise for examining Default Probability and credit grade consistency;Quality and ranking consistence test stage, to bond
Consistency check should be main investigation factor with expected loss rate;The size of loss given default LGD is wherein calculated not only by debt
The influence of business main body oneself factor, but also it is closely related with the specifically design of debt project, the factor for influencing LGD includes project
Factor, company factor, industry choice, macroeconomy periodic factors;
Examine the consistency of rating result, it is important to determine loss given default impact factor, establish the pre- of loss given default
Model is surveyed, from the angle inspection quality of bond credit quality and the consistency of grade;For building for loss given default prediction model
It is vertical, the principal element for influencing loss given default LGD is determined using factor analysis, and damaged with multiple regression data fitting and promise breaking
Quantitative relation between mistake rate predicts loss given default, predicts expected loss rate with this, and examine the consistency of rating result.
7. credit rating Default Probability as described in claim 1 estimates and method for prewarning risk, which is characterized in that the inspection
There are two aspects for the consistency of rating result: first is that Default Probability and credit grade consistency, second is that credit quality and credit etc.
Level consistency.
8. credit rating Default Probability as described in claim 1 estimates and method for prewarning risk, which is characterized in that the step
Three building binary respond risk warning model, realize that " early warning " strategy specifically includes by data mining technology: according to master to be appraised
The duration of body estimates the probability of its promise breaking;Since the value range of probability is 0~1, and the range of probability is 0~+∞, is utilizedProbability is converted into probability;Then it is infinite to just infinite function from bearing to generate one to take the logarithm of probability, with several
The logarithm of rate establishes regression equation as target variable:
Default Probability of the obtained logical function for the duration calculates, the subsequent economic activity data of debt main body to be appraised
To be fitted and be predicted by logistic regression analysis model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910344162.XA CN110110981B (en) | 2019-04-26 | 2019-04-26 | Credit rating default probability measure and risk early warning method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910344162.XA CN110110981B (en) | 2019-04-26 | 2019-04-26 | Credit rating default probability measure and risk early warning method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110110981A true CN110110981A (en) | 2019-08-09 |
CN110110981B CN110110981B (en) | 2021-07-20 |
Family
ID=67486931
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910344162.XA Active CN110110981B (en) | 2019-04-26 | 2019-04-26 | Credit rating default probability measure and risk early warning method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110110981B (en) |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110659979A (en) * | 2019-09-03 | 2020-01-07 | 深圳中兴飞贷金融科技有限公司 | Method and apparatus for predicting loss rate of default, storage medium, and electronic device |
CN110689427A (en) * | 2019-10-12 | 2020-01-14 | 杭州绿度信息技术有限公司 | Consumption stage default probability model based on survival analysis |
CN110782339A (en) * | 2019-10-22 | 2020-02-11 | 黑龙江工业学院 | Default probability prediction method, system and readable storage medium |
CN110969370A (en) * | 2020-01-14 | 2020-04-07 | 深圳市建筑科学研究院股份有限公司 | Quality risk analysis method for building structural member |
CN111179068A (en) * | 2020-01-02 | 2020-05-19 | 百望股份有限公司 | Method, device and storage medium for determining enterprise performance credit level |
CN111340236A (en) * | 2020-03-03 | 2020-06-26 | 中债金融估值中心有限公司 | Bond default prediction method based on bond valuation data and integrated machine learning |
CN111582751A (en) * | 2020-05-19 | 2020-08-25 | 国网吉林省电力有限公司 | Time-weighted electricity purchasing risk early warning method |
CN111652430A (en) * | 2020-05-29 | 2020-09-11 | 蚌埠学院 | Internet financial platform default rate prediction method and system |
CN112073147A (en) * | 2020-04-08 | 2020-12-11 | 成都理工大学 | Feature selection method based on network structure |
CN112102094A (en) * | 2020-09-16 | 2020-12-18 | 中国建设银行股份有限公司 | Method and apparatus for securitization and stratification of assets |
CN112116402A (en) * | 2020-09-28 | 2020-12-22 | 中国建设银行股份有限公司 | Credit risk customer-level default rate determination method and device |
CN112712424A (en) * | 2020-12-07 | 2021-04-27 | 苏宁消费金融有限公司 | Consumption financial risk assessment method for measuring unexpected loss |
CN112862214A (en) * | 2021-03-10 | 2021-05-28 | 重庆第二师范学院 | Parking service recommendation method, device, medium and server based on big data |
CN113222255A (en) * | 2021-05-17 | 2021-08-06 | 上海生腾数据科技有限公司 | Method and device for contract performance quantification and short-term default prediction |
CN113282886A (en) * | 2021-05-26 | 2021-08-20 | 北京大唐神州科技有限公司 | Bank loan default judgment method based on logistic regression |
CN113283861A (en) * | 2021-05-18 | 2021-08-20 | 上海示右智能科技有限公司 | Method for constructing intelligent enterprise compliance |
CN115760365A (en) * | 2022-11-23 | 2023-03-07 | 中电金信软件有限公司 | Method and device for constructing default loss rate prediction model |
US11914719B1 (en) | 2020-04-15 | 2024-02-27 | Wells Fargo Bank, N.A. | Systems and methods for cyberthreat-risk education and awareness |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800016A (en) * | 2012-06-18 | 2012-11-28 | 大连理工大学 | Credit rating adjusting method based on credit rating and loss given default matching |
CN107527288A (en) * | 2017-06-16 | 2017-12-29 | 兴业数字金融服务(上海)股份有限公司 | A kind of bond survival phase Warning System |
CN109670676A (en) * | 2018-11-26 | 2019-04-23 | 安徽继远软件有限公司 | Distributing net platform region method for prewarning risk and system based on Support Vector data description |
-
2019
- 2019-04-26 CN CN201910344162.XA patent/CN110110981B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800016A (en) * | 2012-06-18 | 2012-11-28 | 大连理工大学 | Credit rating adjusting method based on credit rating and loss given default matching |
CN107527288A (en) * | 2017-06-16 | 2017-12-29 | 兴业数字金融服务(上海)股份有限公司 | A kind of bond survival phase Warning System |
CN109670676A (en) * | 2018-11-26 | 2019-04-23 | 安徽继远软件有限公司 | Distributing net platform region method for prewarning risk and system based on Support Vector data description |
Non-Patent Citations (4)
Title |
---|
D. V. BOREIKO: "Numerical Modeling of Dependent Credit Rating Transitions with Asynchronously Moving", 《COMPUTATIONAL ECONOMICS》 * |
HAOMIN WANG: "Cost-sensitive classifiers in credit rating: A comparative study on P2P lending", 《2018 7TH INTERNATIONAL CONFERENCE ON COMPUTERS COMMUNICATIONS AND CONTROL (ICCCC)》 * |
HSIN-MIN LU: "Credit Rating Change Modeling Using News and Financial Ratios", 《ACM TRANSACTIONS ON MANAGEMENT INFORMATION SYSTEMS》 * |
MAURIZIOPOMPELLA: "Ratings based Inference and Credit Risk: Detecting likely-to-fail Banks with the PC-Mahalanobis Method", 《ECONOMIC MODELLING》 * |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110659979A (en) * | 2019-09-03 | 2020-01-07 | 深圳中兴飞贷金融科技有限公司 | Method and apparatus for predicting loss rate of default, storage medium, and electronic device |
CN110689427A (en) * | 2019-10-12 | 2020-01-14 | 杭州绿度信息技术有限公司 | Consumption stage default probability model based on survival analysis |
CN110782339A (en) * | 2019-10-22 | 2020-02-11 | 黑龙江工业学院 | Default probability prediction method, system and readable storage medium |
CN111179068A (en) * | 2020-01-02 | 2020-05-19 | 百望股份有限公司 | Method, device and storage medium for determining enterprise performance credit level |
CN110969370A (en) * | 2020-01-14 | 2020-04-07 | 深圳市建筑科学研究院股份有限公司 | Quality risk analysis method for building structural member |
CN110969370B (en) * | 2020-01-14 | 2023-05-02 | 深圳市建筑科学研究院股份有限公司 | Quality risk analysis method for building structural member |
CN111340236A (en) * | 2020-03-03 | 2020-06-26 | 中债金融估值中心有限公司 | Bond default prediction method based on bond valuation data and integrated machine learning |
CN111340236B (en) * | 2020-03-03 | 2023-09-22 | 中债金融估值中心有限公司 | Bond breach prediction method based on bond estimation data and integrated machine learning |
CN112073147B (en) * | 2020-04-08 | 2022-08-19 | 成都理工大学 | Credit card default prediction method based on network structure feature selection |
CN112073147A (en) * | 2020-04-08 | 2020-12-11 | 成都理工大学 | Feature selection method based on network structure |
US11914719B1 (en) | 2020-04-15 | 2024-02-27 | Wells Fargo Bank, N.A. | Systems and methods for cyberthreat-risk education and awareness |
CN111582751A (en) * | 2020-05-19 | 2020-08-25 | 国网吉林省电力有限公司 | Time-weighted electricity purchasing risk early warning method |
CN111582751B (en) * | 2020-05-19 | 2022-06-14 | 国网吉林省电力有限公司 | Time-weighted electricity purchasing risk early warning method |
CN111652430A (en) * | 2020-05-29 | 2020-09-11 | 蚌埠学院 | Internet financial platform default rate prediction method and system |
CN112102094A (en) * | 2020-09-16 | 2020-12-18 | 中国建设银行股份有限公司 | Method and apparatus for securitization and stratification of assets |
CN112116402A (en) * | 2020-09-28 | 2020-12-22 | 中国建设银行股份有限公司 | Credit risk customer-level default rate determination method and device |
CN112712424A (en) * | 2020-12-07 | 2021-04-27 | 苏宁消费金融有限公司 | Consumption financial risk assessment method for measuring unexpected loss |
CN112712424B (en) * | 2020-12-07 | 2022-08-02 | 苏宁消费金融有限公司 | Consumption financial risk assessment method for measuring unexpected loss |
CN112862214A (en) * | 2021-03-10 | 2021-05-28 | 重庆第二师范学院 | Parking service recommendation method, device, medium and server based on big data |
CN113222255A (en) * | 2021-05-17 | 2021-08-06 | 上海生腾数据科技有限公司 | Method and device for contract performance quantification and short-term default prediction |
CN113222255B (en) * | 2021-05-17 | 2024-03-05 | 上海生腾数据科技有限公司 | Method and device for quantifying contract performance and predicting short-term violations |
CN113283861A (en) * | 2021-05-18 | 2021-08-20 | 上海示右智能科技有限公司 | Method for constructing intelligent enterprise compliance |
CN113283861B (en) * | 2021-05-18 | 2024-04-16 | 上海示右智能科技有限公司 | Method for constructing compliance of intelligent enterprise |
CN113282886A (en) * | 2021-05-26 | 2021-08-20 | 北京大唐神州科技有限公司 | Bank loan default judgment method based on logistic regression |
CN115760365A (en) * | 2022-11-23 | 2023-03-07 | 中电金信软件有限公司 | Method and device for constructing default loss rate prediction model |
CN115760365B (en) * | 2022-11-23 | 2024-05-07 | 中电金信软件有限公司 | Method and device for constructing loss-in-default prediction model |
Also Published As
Publication number | Publication date |
---|---|
CN110110981B (en) | 2021-07-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110110981A (en) | A kind of credit rating Default Probability estimates and method for prewarning risk | |
Valle-Cruz et al. | From E-budgeting to smart budgeting: Exploring the potential of artificial intelligence in government decision-making for resource allocation | |
Sun et al. | Financial distress early warning based on group decision making | |
Jiang et al. | Bubble diagnosis and prediction of the 2005–2007 and 2008–2009 Chinese stock market bubbles | |
Sahajwala et al. | Supervisory risk assessment and early warning systems | |
CN107633265A (en) | For optimizing the data processing method and device of credit evaluation model | |
Bateni et al. | Bankruptcy prediction using logit and genetic algorithm models: A comparative analysis | |
Acheampong et al. | Does soft information determine credit risk? Text-based evidence from European banks | |
CN107590735A (en) | Data digging method and device for credit evaluation | |
Yang et al. | Evaluate the sustainable reuse strategy of the corporate financial management based on the big data model | |
Zhang et al. | Enterprise financial risk early warning using bp neural network under internet of things and rough set theory | |
Chen et al. | Forecasting PGR of the financial industry using a rough sets classifier based on attribute-granularity | |
CN117575328A (en) | Industrial chain risk assessment method and device based on graph attention neural network | |
Xie et al. | An interpretable machine-learned model for international oil trade network | |
Mahmoud et al. | Expert system for banking credit decision | |
Tomlinson et al. | Some systems aspects of strategic planning | |
Hu et al. | Application of Data Mining Combined with Fuzzy Multicriteria Decision‐Making in Credit Risk Assessment from Legal Service Companies | |
Jiang | Research on Influencing Factors and Early Warning of University Finance Based on Decision Tree Model | |
López et al. | Multicriteria Decision Model to Support the Evaluation of Common Jurisdiction Violence in the Capital Cities of the States of Mexico | |
Lapishko et al. | Immatious-automatic simulation of processes in insurance activities | |
Li et al. | Construction of Credit Evaluation Index System for Two-Stage Bayesian Discrimination: An Empirical Analysis of Small Chinese Enterprises. | |
Zhang et al. | Construction and analysis of the financing risk network of Chinese fisheries enterprises | |
Janková et al. | Analysis of Financial Market Using Soft Computing Techniques | |
CN113610638B (en) | Rating system and method for matching credit rating with default loss rate based on SMAA-DS | |
Zhou et al. | An Improved Data Mining Model for Predicting the Impact of Economic Fluctuations |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |