CN104463673A - P2P network credit risk assessment model based on support vector machine - Google Patents
P2P network credit risk assessment model based on support vector machine Download PDFInfo
- Publication number
- CN104463673A CN104463673A CN201410801984.3A CN201410801984A CN104463673A CN 104463673 A CN104463673 A CN 104463673A CN 201410801984 A CN201410801984 A CN 201410801984A CN 104463673 A CN104463673 A CN 104463673A
- Authority
- CN
- China
- Prior art keywords
- result
- length
- reference point
- support vector
- function
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a P2P network credit risk assessment model based on a support vector machine. The P2P network credit risk assessment model comprises the steps of (1) data preprocessing, (2) model building and (3) risk assessment according to a result provided by the model. In this way, the P2P network credit risk assessment model based on the support vector machine has the advantages of reducing the influence of human factors, being objective, scientific, better in prediction effect and higher in accuracy, reducing unrelated attributes, greatly reducing the data processing quantity, being shorter in leaning time needed by prediction and low in requirement for the sample data quantity and the like, and the P2P network credit risk assessment model based on the support vector machine has the wide market prospect in popularization.
Description
Technical field
The present invention relates to network loan valuation model field, particularly relate to a kind of P2P network loan risk evaluation model based on support vector machine.
Background technology
P2P(Peer-to-Peer) network loan refers to the unsecured debt-credit of interindividual small amount, it is not intermediary with bank and other financial mechanism, directly establish debtor-creditor relationship by internet platform and complete relationship trading formality, achieve " financial disintermediation ", there is the features such as loan threshold is low, broad covered area, information flow is fast, transaction procedure is convenient, amount of money involved is little, the life of loan is shorter, there is very large development potentiality.
And P2P network loan exists certain risk, how to ensure that lended out money is all (creditor have the ability repay) that can regain, namely " risk control " seem very important.So according to the essential information that above creditor provides, assessed the risk class of current loan by the model method of objective science, thus provide valuable decision references information and then seem abnormal important.
But the problems such as the risk control measure that is asymmetric, platform that is virtual, information because of network is unsound, add the possibility of borrower's promise breaking, cause the loaning bill rate of violation in P2P network loan market higher, credit risk is difficult to control.
The domestic research to assessing credit risks at present takes mode qualitatively mostly, seldom can carry out quantitative test and research to it.This model chooses algorithm of support vector machine, its credit risk of qualitative assessment targetedly according to features such as P2P network loan data high latitude, non-linear and small samples just.
Suppose that the Shen that a network Shen loan people provides is borrowed shown in the following form of essential information:
People's Basic Information Table is borrowed in table 1 Shen
How to go out this single degree of risk of providing a loan according to the loan people essential information science objective assessment of above Shen is exactly the work that the present invention will do.
Summary of the invention
The technical matters that the present invention mainly solves is to provide a kind of P2P network loan risk evaluation model based on support vector machine, by the analysis to past annual data, choose and have the attribute of considerable influence as primary attribute to result, then risk evaluation model is set up according to support vector cassification algorithm, when there being unknown record to input, this model can provide final Risk Evaluation result, prediction effect is better, accuracy rate is higher, has market outlook widely popularizing of the P2P network loan risk evaluation model based on support vector machine.
For solving the problems of the technologies described above, the invention provides a kind of P2P network loan risk evaluation model based on support vector machine, comprising the following steps:
(1) data prediction: obtain P2P loan documentation in former years from database, carry out data prediction, main employing principal component analysis (PCA) technology, namely provides every bar attribute to the influence degree of net result, chooses percentage contribution sum and is greater than former Column Properties of certain weight proportion as primary attribute;
(2) Modling model: former years P2P loan documentation primary attribute basis on Modling model,
A () feature space maps: in order to strengthen linear separability, the original input space is mapped to a higher-dimension point product space, i.e. feature space,
If non-linear phasor function g (x)=[g1 (x) ..., gl (x)] m is tieed up input vector x to be mapped to l dimensional feature space, then the linear decision function of feature space is:
（1）
According to Hilaert-Schmiat theorem, if a symmetric function H (x, x) meets
(2)
In formula (2), M is natural number, h
i, h
jfor real number,
Then there is mapping function g (x), x can be mapped to dot product feature space, this mapping function meets
（3）
If formula (2) is set up, then
（4）
Formula (2) or formula (4) are called Mercer condition, and the function of any one met in above-mentioned two formulas is called positive semidefinite kernel function or Mercer kernel function;
B () utilizes trial method to obtain optimized parameter:
Setting C initial value, reference point, change direction, step-length, train first and second SVM, i=2,
Calculate the ASVR of i-th time,
Judge whether changing value exceeded thresholding compared with last time,
(b.1) if result is "Yes", then judge whether SVR declines compared with reference point,
(b.1.1) if result is "No", then judge whether the step-length of reference point increased,
(b.1.1.1) if result is "Yes", then judge whether the step-length of reference point reduced,
(b.1.1.1.1) if result is "Yes", then make corresponding modify according to the scope of step-length: if current step is greater than 1, reduce step-length; If be less than 1, increase step-length, obtain the value of next C,
(b.1.1.1.2) if result is "No", then reduce step-length, note current reference point step size reduced, and obtained the value of next C,
(b.1.1.2) if result is "No", then increase step-length, note current reference point step size increased, and to upgrade reference point be currency, obtained the value of next C,
(b.1.2) if result is "Yes", then step-length is constant, and change direction is constant, and to upgrade reference point be currency, obtains the value of next C,
(b.2) if result is "No", then C, step-length, change direction are constant, and to upgrade reference point be currency,
Make i+1 assignment to i, and return the ASVR of calculating i-th time, finally obtain optimized parameter;
C () is based on the support vector cassification of kernel function: utilize the advantage of kernel function to be no longer to need directly to process high-dimensional feature space,
Employing kernel function H (x, x ') replace g (x), then original optimization problem is converted into:
（5）
Constraint condition is:
,
,
Because H (x, x ') is a positive semidefinite kernel function, therefore, in formula (5), optimization problem is a quadratic convex programming problem, has globally optimal solution,
According to KKT complementarity condition, can be in the hope of now categorised decision function
（6）
Wherein bias term a determines (getting non-border support vector mean value) by following formula
（7）
Unknown data x classification results is:
If class 1 A (x) is >0
If class 2 A (x) is <0
If A (x)=0, then x is inseparable,
Bring the optimized parameter obtained in step (b) into above-mentioned model and obtain result;
(3) result provided according to model makes risk assessment.
In a preferred embodiment of the present invention, the numerical value of the described certain weight proportion in step (1) is 90%.
The invention has the beneficial effects as follows: the P2P network loan risk evaluation model that the present invention is based on support vector machine have reduce human factor impact, objective science, prediction effect is better, accuracy rate is higher, reduce irrelevant attribute, greatly reduce data processing amount, make the learning time needed for prediction shorter, to advantages such as sample data amount are less demanding, the P2P network loan risk evaluation model based on support vector machine universal on have market outlook widely.
Accompanying drawing explanation
In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing used required in describing embodiment is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings, wherein:
Fig. 1 is the algorithm flow chart of the optimized parameter of a preferred embodiment of the P2P network loan risk evaluation model that the present invention is based on support vector machine;
Fig. 2 is the process flow diagram of a preferred embodiment of the P2P network loan risk evaluation model that the present invention is based on support vector machine.
Embodiment
Be clearly and completely described to the technical scheme in the embodiment of the present invention below, obviously, described embodiment is only a part of embodiment of the present invention, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making other embodiments all obtained under creative work prerequisite, belong to the scope of protection of the invention.
The embodiment of the present invention comprises:
Based on a P2P network loan risk evaluation model for support vector machine, comprise the following steps:
(1) data prediction: obtain P2P loan documentation in former years from database, carry out data prediction, main employing principal component analysis (PCA) technology, namely provides every bar attribute to the influence degree of net result, chooses percentage contribution sum and is greater than former Column Properties of certain weight proportion as primary attribute;
(2) Modling model: former years P2P loan documentation primary attribute basis on Modling model,
A () feature space maps: in order to strengthen linear separability, the original input space is mapped to a higher-dimension point product space, i.e. feature space,
If non-linear phasor function g (x)=[g1 (x) ..., gl (x)] m is tieed up input vector x to be mapped to l dimensional feature space, then the linear decision function of feature space is:
（1）
According to Hilaert-Schmiat theorem, if a symmetric function H (x, x) meets
(2)
In formula (2), M is natural number, h
i, h
jfor real number,
Then there is mapping function g (x), x can be mapped to dot product feature space, this mapping function meets
（3）
If formula (2) is set up, then
（4）
Formula (2) or formula (4) are called Mercer condition, and the function of any one met in above-mentioned two formulas is called positive semidefinite kernel function or Mercer kernel function;
B () utilizes trial method to obtain optimized parameter:
Setting C initial value, reference point, change direction, step-length, train first and second SVM, i=2,
Calculate the ASVR of i-th time,
Judge whether changing value exceeded thresholding compared with last time,
(b.1) if result is "Yes", then judge whether SVR declines compared with reference point,
(b.1.1) if result is "No", then judge whether the step-length of reference point increased,
(b.1.1.1) if result is "Yes", then judge whether the step-length of reference point reduced,
(b.1.1.1.1) if result is "Yes", then make corresponding modify according to the scope of step-length: if current step is greater than 1, reduce step-length; If be less than 1, increase step-length, obtain the value of next C,
(b.1.1.1.2) if result is "No", then reduce step-length, note current reference point step size reduced, and obtained the value of next C,
(b.1.1.2) if result is "No", then increase step-length, note current reference point step size increased, and to upgrade reference point be currency, obtained the value of next C,
(b.1.2) if result is "Yes", then step-length is constant, and change direction is constant, and to upgrade reference point be currency, obtains the value of next C,
(b.2) if result is "No", then C, step-length, change direction are constant, and to upgrade reference point be currency,
Make i+1 assignment to i, and return the ASVR of calculating i-th time, finally obtain optimized parameter;
C () is based on the support vector cassification of kernel function: utilize the advantage of kernel function to be no longer to need directly to process high-dimensional feature space,
Employing kernel function H (x, x ') replace g (x), then original optimization problem is converted into:
（5）
Constraint condition is:
,
,
Because H (x, x ') is a positive semidefinite kernel function, therefore, in formula (5), optimization problem is a quadratic convex programming problem, has globally optimal solution,
According to KKT complementarity condition, can be in the hope of now categorised decision function
（6）
Wherein bias term a determines (getting non-border support vector mean value) by following formula
（7）
Unknown data x classification results is:
If class 1 A (x) is >0
If class 2 A (x) is <0
If A (x)=0, then x is inseparable,
Bring the optimized parameter obtained in step (b) into above-mentioned model and obtain result;
(3) result provided according to model makes risk assessment.
Preferably, the numerical value of the described certain weight proportion in step (1) is 90%.
According to the value of loan status attribute lobn_stbtus, choose the record that expires of all loans and form real example data set, record of on time refunding in overdue record is called record of normally refunding, if the refund that exceeds the time limit is called abnormal record of refunding.Therefrom randomly draw training set and test set that record forms this system respectively, it is as shown in the table and as example for test set and training set data distribution situation:
| Data set scale | Normal refund record | Extremely to refund record | |
| Training dataset | 6669 | 4793 | 1876 |
| Test data set | 3800 | 3000 | 800 |
| Conceptual data collection | 10649 | 7793 | 2676 |
Lenaing clua loan documentation has more than 100 attribute, therefrom chooses the input attributes of attribute as model refund state lobn_stbtus being had to considerable influence.
The model that Modling model just can allow Evaluation accuracy converge on all properties set to set up is carried out with the attribute of 7 wherein.Experimentally result selects following 7 attributes of record: the lobn_bmnt(apply for loan amount of money), the term(length of maturity), int_rbte(loan interest rate), grbae(credit grade), emp_length(is engaged in time of this work), home_ownership(house has situation), bnnubl_inc(annual income), and add the state that lobn_stbtus(provides a loan) this categorical attribute.
7 attributes are converted into numeric type.Length of maturity term has 36 months and 60 months two classes, represents respectively with numerical value 3 and 5.Credit grade grbae has B, A, C, A, E Pyatyi from high in the end, is set to 1,2,3,4,5 accordingly.Being engaged in the time emp_length of work, is a grade every year, and being greater than 10 years assignment is 11.Have situation home_ownership for house, rent a house and give weights 1, having house weights is 5.Loan status lobn_stbtus except fully pbia state of on time refunding be except 0, other state lbte, chbrgea off, aefbult, in grbce perioa are 1 entirely, represent delay refund.Wherein the apply for loan amount of money, loan interest rate, annual income attribute are originally as numeric type data, need not change.
In mbtlba, function Y=zscore (X) is adopted to carry out data normalization, to eliminate the dimension difference between data.
Below by trial method determination support vector machine optimized parameter.
(1) get g=0.5, trial method determines C, and wherein accuracy of detection refers to be recorded in ratio shared in total test set by the data of correctly classifying.
| C | Accuracy of detection |
| 0.01 | 84.48% |
| 0.1 | 85% |
| 1 | 86.66% |
| 10 | 81.18% |
| 20 | 79.54% |
| 50 | 77.56% |
| 100 | 75.96% |
(2) as C=1, the value of g is determined.
| G | Accuracy of detection |
| 0.001 | 79% |
| 0.01 | 86% |
| 0.1 | 85.6% |
| 1 | 84.52% |
| 10 | 83.32% |
| 50 | 80.02% |
From 1 attribute, increase attribute gradually, accuracy of detection also increases thereupon, and after being increased to 7 attributes, accuracy of detection is tending towards convergence, then increases attribute and can not significantly improve precision.Along with attribute increases, the travelling speed of supporting vector machine model also and then reduces, and considers precision and speed factor, and 7 attributes are best input attributes combinations.The accuracy of detection that the parameter determined by above-mentioned experiment is C=1, g=0.01, have the supporting vector machine model of 7 input attributes finally to obtain is 85.6%.
Detailed test result and contrasting with the assessment models of another kind based on AP neural network, result is as follows:
| Model | Accuracy rate | Error type I rate | Error type II rate |
| Support vector machine | 85.6% | 1.02% | 13.4% |
| AP neural network | 78.6% | 4.8% | 16.6% |
Upper form shows, and " the P2P network loan risk evaluation model based on support vector machine " has higher credit risk discrimination and accuracy of identification than " model based on AP neural network ".And it is very good to control Error type I, can control risk to the full extent, ensures the interests by borrowing people.Above experimental result illustrates that this model is effective.
The beneficial effect that the present invention is based on the P2P network loan risk evaluation model of support vector machine is:
One, by the analysis to past annual data, choose and have the attribute of considerable influence as primary attribute to result, then risk evaluation model is set up according to support vector cassification algorithm, when there being unknown record to input, this model can provide final Risk Evaluation result, reduces the impact of human factor, for science decision provides technical support while liberation manpower, compared to the neural network learning forecasting mechanism extensively adopted at present, the prediction effect based on support vector machine is better, and accuracy rate is higher;
Two, by with the addition of principal component analysis (PCA) technology on the basis of support vector machine, irrelevant attribute can be reduced, greatly reduce data processing amount, make the learning time needed for prediction shorter;
Three, algorithm of support vector machine mainly carries out a kind of method learning, classify and predict for Small Sample Database, can solve the indeterminable problem concerning study excessively of other sorting algorithms, less demanding to sample data amount.
The foregoing is only embodiments of the invention; not thereby the scope of the claims of the present invention is limited; every utilize description of the present invention to do equivalent structure or equivalent flow process conversion; or be directly or indirectly used in other relevant technical field, be all in like manner included in scope of patent protection of the present invention.
Claims (2)
1., based on a P2P network loan risk evaluation model for support vector machine, it is characterized in that, comprise the following steps:
(1) data prediction: obtain P2P loan documentation in former years from database, carry out data prediction, main employing principal component analysis (PCA) technology, namely provides every bar attribute to the influence degree of net result, chooses percentage contribution sum and is greater than former Column Properties of certain weight proportion as primary attribute;
(2) Modling model: former years P2P loan documentation primary attribute basis on Modling model,
A () feature space maps: in order to strengthen linear separability, the original input space is mapped to a higher-dimension point product space, i.e. feature space,
If non-linear phasor function g (x)=[g1 (x) ..., gl (x)] m is tieed up input vector x to be mapped to l dimensional feature space, then the linear decision function of feature space is:
（1）
According to Hilaert-Schmiat theorem, if a symmetric function H (x, x) meets
(2)
In formula (2), M is natural number, h
i, h
jfor real number,
Then there is mapping function g (x), x can be mapped to dot product feature space, this mapping function meets
（3）
If formula (2) is set up, then
（4）
Formula (2) or formula (4) are called Mercer condition, and the function of any one met in above-mentioned two formulas is called positive semidefinite kernel function or Mercer kernel function;
B () utilizes trial method to obtain optimized parameter:
Setting C initial value, reference point, change direction, step-length, train first and second SVM, i=2,
Calculate the ASVR of i-th time,
Judge whether changing value exceeded thresholding compared with last time,
(b.1) if result is "Yes", then judge whether SVR declines compared with reference point,
(b.1.1) if result is "No", then judge whether the step-length of reference point increased,
(b.1.1.1) if result is "Yes", then judge whether the step-length of reference point reduced,
(b.1.1.1.1) if result is "Yes", then make corresponding modify according to the scope of step-length: if current step is greater than 1, reduce step-length; If be less than 1, increase step-length, obtain the value of next C,
(b.1.1.1.2) if result is "No", then reduce step-length, note current reference point step size reduced, and obtained the value of next C,
(b.1.1.2) if result is "No", then increase step-length, note current reference point step size increased, and to upgrade reference point be currency, obtained the value of next C,
(b.1.2) if result is "Yes", then step-length is constant, and change direction is constant, and to upgrade reference point be currency, obtains the value of next C,
(b.2) if result is "No", then C, step-length, change direction are constant, and to upgrade reference point be currency,
Make i+1 assignment to i, and return the ASVR of calculating i-th time, finally obtain optimized parameter;
C () is based on the support vector cassification of kernel function: utilize the advantage of kernel function to be no longer to need directly to process high-dimensional feature space,
Employing kernel function H (x, x ') replace g (x), then original optimization problem is converted into:
（5）
Constraint condition is:
,
,
Because H (x, x ') is a positive semidefinite kernel function, therefore, in formula (5), optimization problem is a quadratic convex programming problem, has globally optimal solution,
According to KKT complementarity condition, can be in the hope of now categorised decision function
（6）
Wherein bias term a determines (getting non-border support vector mean value) by following formula
（7）
Unknown data x classification results is:
If class 1 A (x) is >0
If class 2 A (x) is <0
If A (x)=0, then x is inseparable,
Bring the optimized parameter obtained in step (b) into above-mentioned model and obtain result;
(3) result provided according to model makes risk assessment.
2. the P2P network loan risk evaluation model based on support vector machine according to claim 1, it is characterized in that, the numerical value of the described certain weight proportion in step (1) is 90%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410801984.3A CN104463673A (en) | 2014-12-22 | 2014-12-22 | P2P network credit risk assessment model based on support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410801984.3A CN104463673A (en) | 2014-12-22 | 2014-12-22 | P2P network credit risk assessment model based on support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104463673A true CN104463673A (en) | 2015-03-25 |
Family
ID=52909664
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410801984.3A Pending CN104463673A (en) | 2014-12-22 | 2014-12-22 | P2P network credit risk assessment model based on support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104463673A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107025596A (en) * | 2016-02-01 | 2017-08-08 | 腾讯科技（深圳）有限公司 | A kind of methods of risk assessment and system |
| CN107038165A (en) * | 2016-02-03 | 2017-08-11 | 腾讯科技（深圳）有限公司 | A kind of service parameter acquisition methods and device |
| CN107133867A (en) * | 2017-04-16 | 2017-09-05 | 信阳师范学院 | Credit method for anti-counterfeit based on SVMs |
| CN107767259A (en) * | 2017-09-30 | 2018-03-06 | 平安科技（深圳）有限公司 | Loan risk control method, electronic installation and readable storage medium storing program for executing |
| CN110135970A (en) * | 2019-04-15 | 2019-08-16 | 深圳壹账通智能科技有限公司 | Loan valuation method, apparatus, computer equipment and storage medium |
| CN110619024A (en) * | 2019-09-20 | 2019-12-27 | 苏州大学 | Credit evaluation method, system and related device |
| CN110909805A (en) * | 2019-11-26 | 2020-03-24 | 西安交通大学城市学院 | Financial wind control system based on big data and increment V3 deep network model |
| CN111553800A (en) * | 2020-04-30 | 2020-08-18 | 上海商汤智能科技有限公司 | Data processing method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101996381A (en) * | 2009-08-14 | 2011-03-30 | 中国工商银行股份有限公司 | Method and system for calculating retail asset risk |
| CN102376067A (en) * | 2010-08-20 | 2012-03-14 | 许威 | Scorecard system based on financial credit loan and realization method for scorecard system |
-
2014
- 2014-12-22 CN CN201410801984.3A patent/CN104463673A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101996381A (en) * | 2009-08-14 | 2011-03-30 | 中国工商银行股份有限公司 | Method and system for calculating retail asset risk |
| CN102376067A (en) * | 2010-08-20 | 2012-03-14 | 许威 | Scorecard system based on financial credit loan and realization method for scorecard system |
Non-Patent Citations (2)
| Title |
|---|
| 傅彦铭等: "P2P网络贷款信用的风险评估", 《统计与决策》 * |
| 周玖玖: "入侵检测中支持向量机参数选择方法研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107025596A (en) * | 2016-02-01 | 2017-08-08 | 腾讯科技（深圳）有限公司 | A kind of methods of risk assessment and system |
| CN107025596B (en) * | 2016-02-01 | 2021-07-16 | 腾讯科技（深圳）有限公司 | Risk assessment method and system |
| CN107038165A (en) * | 2016-02-03 | 2017-08-11 | 腾讯科技（深圳）有限公司 | A kind of service parameter acquisition methods and device |
| CN107038165B (en) * | 2016-02-03 | 2021-02-02 | 腾讯科技（深圳）有限公司 | Service parameter acquisition method and device |
| CN107133867A (en) * | 2017-04-16 | 2017-09-05 | 信阳师范学院 | Credit method for anti-counterfeit based on SVMs |
| CN107767259A (en) * | 2017-09-30 | 2018-03-06 | 平安科技（深圳）有限公司 | Loan risk control method, electronic installation and readable storage medium storing program for executing |
| CN110135970A (en) * | 2019-04-15 | 2019-08-16 | 深圳壹账通智能科技有限公司 | Loan valuation method, apparatus, computer equipment and storage medium |
| CN110619024A (en) * | 2019-09-20 | 2019-12-27 | 苏州大学 | Credit evaluation method, system and related device |
| CN110909805A (en) * | 2019-11-26 | 2020-03-24 | 西安交通大学城市学院 | Financial wind control system based on big data and increment V3 deep network model |
| CN110909805B (en) * | 2019-11-26 | 2022-02-18 | 西安交通大学城市学院 | Financial wind control system based on big data and increment V3 deep network model |
| CN111553800A (en) * | 2020-04-30 | 2020-08-18 | 上海商汤智能科技有限公司 | Data processing method and device, electronic equipment and storage medium |
| CN111553800B (en) * | 2020-04-30 | 2023-08-25 | 上海商汤智能科技有限公司 | Data processing method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104463673A (en) | P2P network credit risk assessment model based on support vector machine | |
| Ye et al. | Loan evaluation in P2P lending based on random forest optimized by genetic algorithm with profit score | |
| Ye et al. | A novel forecasting method based on multi-order fuzzy time series and technical analysis | |
| Ala’raj et al. | Modelling customers credit card behaviour using bidirectional LSTM neural networks | |
| Harris | Credit scoring using the clustered support vector machine | |
| Crook et al. | Recent developments in consumer credit risk assessment | |
| Shen et al. | A cost-sensitive logistic regression credit scoring model based on multi-objective optimization approach | |
| Babenko et al. | Classical machine learning methods in economics research: Macro and micro level example | |
| CN110751557B (en) | Abnormal fund transaction behavior analysis method and system based on sequence model | |
| Babaei et al. | A multi-objective instance-based decision support system for investment recommendation in peer-to-peer lending | |
| Höglund | Tax payment default prediction using genetic algorithm-based variable selection | |
| Chen et al. | The predictive power of Japanese candlestick charting in Chinese stock market | |
| JP2003535387A (en) | Rapid evaluation of asset portfolios such as financial products | |
| CN106570631B (en) | P2P platform-oriented operation risk assessment method and system | |
| Fan et al. | Improved ML‐based technique for credit card scoring in Internet financial risk control | |
| Wei | [Retracted] A Method of Enterprise Financial Risk Analysis and Early Warning Based on Decision Tree Model | |
| Lu et al. | A clustering-based portfolio strategy incorporating momentum effect and market trend prediction | |
| Deng et al. | A hybrid method for crude oil price direction forecasting using multiple timeframes dynamic time wrapping and genetic algorithm | |
| Chen | [Retracted] Risk Assessment of Government Debt Based on Machine Learning Algorithm | |
| Li et al. | Application of XGBoost in P2P default prediction | |
| Kurihara et al. | AR model or machine learning for forecasting GDP and consumer price for G7 countries | |
| Glennon et al. | Development and validation of credit scoring models | |
| CN113344692B (en) | Method for establishing network loan credit risk assessment model with multi-information-source fusion | |
| Zhou et al. | The bubble and anti-bubble risk resistance analysis on the metal futures in China | |
| Zeng | [Retracted] Research on Risk Measurement and Early Warning of Electronic Banking Business Based on GMDH Algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| 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: 20150325 |