CN107133867A - Credit method for anti-counterfeit based on SVMs - Google Patents

Credit method for anti-counterfeit based on SVMs Download PDF

Info

Publication number
CN107133867A
CN107133867A CN201710275338.1A CN201710275338A CN107133867A CN 107133867 A CN107133867 A CN 107133867A CN 201710275338 A CN201710275338 A CN 201710275338A CN 107133867 A CN107133867 A CN 107133867A
Authority
CN
China
Prior art keywords
mrow
msup
msub
sample
credit
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
Application number
CN201710275338.1A
Other languages
Chinese (zh)
Inventor
郭华平
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xinyang Normal University
Original Assignee
Xinyang Normal University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xinyang Normal University filed Critical Xinyang Normal University
Priority to CN201710275338.1A priority Critical patent/CN107133867A/en
Publication of CN107133867A publication Critical patent/CN107133867A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q40/00Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/03Credit; Loans; Processing thereof
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • G06F18/241Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411Classification 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Other Investigation Or Analysis Of Materials By Electrical Means (AREA)

Abstract

The present invention provides a kind of credit method for anti-counterfeit based on SVMs.The credit method for anti-counterfeit based on SVMs is in the training stage, the sorting technique and homing method of SVMs are organically combined using new object function, and then ensure the more doubtful credits swindle samples of segregation section inside covering of study, and segregation section both sides are respectively normal sample and credit swindle sample;In forecast period, if sample to be predicted falls inside segregation section, predict that the example is swindled for doubtful credit;If on segregation section both sides, according to its fall particular location, it is the swindle of normal or credit to predict it.Compared with correlation technique, the credit method for anti-counterfeit based on SVMs that the present invention is provided can effectively analyze whether sample belongs to doubtful credit swindle, and can keep high-accuracy on and credit swindle sample normal in prediction.

Description

Credit method for anti-counterfeit based on SVMs
Technical field
The present invention relates to data analysis technique field, more particularly to a kind of credit method for anti-counterfeit based on SVMs.
Background technology
In practice, loan fraud case occurs repeatedly, and huge loss is caused to bank, and loan swindle comes from bank mostly Bad lending, occur loan swindle be bank's non-performing loan ultimate performance, and it is more the problem of be commercial banking Embody a concentrated reflection of poor for loan quality, bad-loan ratio remains high.At present, this ratio mean height of Ji great nationalized banks is up to two Digit, has had a strong impact on the normal operation of business bank, and heavy losses are caused to bank and people's property.
As continuing to develop for China market economy and deepening continuously for market economic system, especially economic structure turn The formation of rail, the expansion of socio-economic activity, and modern finance system, required specification is inevitable with a large amount of during setting up Phenomenon of being out of normal activity, loan fraudulent act a large amount of appearance be exactly a more typical example, loan swindle crime very disruptive Normal financial order, while also destroy the social credit mechanism on one of socialist market economy basis.
Credit swindle classification problem can regard the classification problem of three classes as, i.e., normal, credit swindle and doubtful letter With swindle.Compared to normal and credit fraud, dubiety fraud is often more worth research, because, this part of credit Problem is likely to become real credit fraud.
For such issues that, traditional multi-class support vector machine is come pair using one-to-one, a pair other and DAGSVM methods Multi-class problem is classified.Wherein, one-to-one and a pair of other methods extensive errors are unbounded.Also, it is one-to-one and one The quantity of sub-classifier to being constructed needed for other methods increases on classification number k into superlinearity, and common k (k-1) is individual, in test rank Section, it is necessary to calculate all sub- discriminant functions and predict the outcome, then obtains final classification result using voting method.This method It is not directly perceived, do not utilize understanding of the expert to problem.In addition, each sub-classifier must carry out specification to data in One-against-one Change, this causes also one most obvious shortcoming to be exactly that each sub-classifier will must be adjusted carefully very much, if some Sub-classifier lack of standardizationization, then whole categorizing system will tend to study.DAGSVM methods solve inseparable regional issue, and And be not necessarily intended to calculate all subclassification decision functions, but position of each sub-classifier in directed acyclic graph also can be to dividing Class system produces large effect.Because credit swindles the particularity of data, the model that these methods are built is not easy to reason Solve and performance is not ideal enough, be not suitable for the classification that credit swindles sample.Therefore design is needed to be directed to credit swindle data set Sorting technique.
The content of the invention
The present invention builds new supporting vector machine model to solve prior art problem, design it is a kind of based on support to The credit method for anti-counterfeit of amount machine, while this method ensures effectively to predict doubtful credit swindle sample, predict exactly it is normal and Credit swindles sample.
The present invention provides a kind of credit method for anti-counterfeit based on SVMs, including:
Step 1, build one's credit loan variable information table is described;Training sample is obtained, table pair is described according to described information Training sample is handled, and sets up training samples information table;
Step 2, object function is set up, and utilizes described object function by support vector cassification strategy and supporting vector Machine returns strategy and merged, and described object function is:
Wherein:
yi(w·xj)≥1-ξj, j=l+1, L, l+m+n,
(w·xi)-yi≤1+ξi, i=1, L, l,
ξj>=0, j=l+1, L, l+m+n,
Wherein, w is the parameter of hyperplane, and l is the number of doubtful credit swindle sample in training sample table, and m and n are respectively The number of credit swindle sample and normal sample in training sample table, power of punishment when C1 is doubtful credit swindle mistake classification Weight, punishment weight when C2 is credit swindle sample or the classification of normal sample mistake;
Step 3, parameter C1 and C2 value is determined using 10 folding cross validations:It is 10 that stochastic averagina, which divides training dataset, Folding, other to roll over for training pattern for each folding, the folding is used for the performance for testing the model trained;Average each folding As a result the final excellent Generalization Capability of model is obtained;It is respectively [0.1,0.2,0.3 ..., 1.0] to set C1 and C2 spans;It is right In with the corresponding C1 and C2 values of optimal Generalization Capability model,
Step 4, setOn overall training set, final mask, profit are trained using iterative method The parameter w solved with object function;Iterative process with for:
Wherein w (k) is kth time iteration w value,It is gradient of the object function at w (k) places, Hk represents kth time The Hesse matrixes of iterative target function, its initial value H0 is unit matrix, and Hk is calculated using following alternative manner:
Wherein p(k)=w(k+1)-w(k),
Step 5, list processing forecast sample x ' is described using the information of step 1, calculates the output function y=of SVMs Wx '+b, so the class that is associated with forecast sample x ' of model prediction marked as:
It is preferred that, in step 1, described information describes table and fiduciary loan type of variables is divided into classification, numerical value, two It is first or discrete.
It is preferred that, in steps of 5, described forecast sample x ' is available:
Substituted into after non-linearization the output function of SVMs.
It is preferred that, the missing attribute of the sample in forecast sample x ' in the training sample and step 5 that are obtained in step 1 Value, is filled after being averaged using other samples are known with property value.
Compared with correlation technique, the credit method for anti-counterfeit based on SVMs that the present invention is provided can be analyzed effectively Whether sample belongs to doubtful credit swindle, and can keep high-accuracy on and credit swindle sample normal in prediction.
Brief description of the drawings
Fig. 1 swindles sample linear classification schematic diagram for the credit method for anti-counterfeit based on SVMs that the present invention is provided;
The swindle sample Nonlinear Classification signal of the credit method for anti-counterfeit based on SVMs that Fig. 2 provides for the present invention Figure.
Embodiment
Describe the present invention in detail below with reference to accompanying drawing and in conjunction with the embodiments.
A kind of credit method for anti-counterfeit based on SVMs, including:
Step 1, build one's credit loan variable information table is described;Training sample is obtained, table pair is described according to described information Training sample is handled, and sets up training samples information table.
Information describes table and is shown in Table 1.Training samples information table is shown in Table 2.
Table 1:Information describes table
In table 1, build one's credit loan variable information table is described, by fiduciary loan type of variables be divided into classification, numerical value, It is binary, discrete.
Table 2:Training samples information table
In table 2, fetch according to first 4 of concentration as the training sample x obtained, table is described to training according to described information Sample x processing.The missing attribute values of sample in training sample, average using known to other samples with property value After fill, the 7th row Salary value in such as table 2, data item the 2nd be its 1st, 3, the average value of the Salary of 4 value.
Step 2, object function is set up, and is returned support vector cassification and SVMs using described object function Return merging, described object function is:
Wherein:
yi(w·xj)≥1-ξj, j=l+1, L, l+m+n,
(w·xi)-yi≤1+ξi, i=1, L, l,
ξj>=0, j=l+1, L, l+m+n,
Wherein, w is the parameter of hyperplane, and l is the number of doubtful credit swindle sample in training sample table, and m and n are respectively The number of credit swindle sample and normal sample in training sample table, power of punishment when C1 is doubtful credit swindle mistake classification Weight, punishment weight when C2 is credit swindle sample or the classification of normal sample mistake.
Step 3, parameter C1 and C2 value is determined using 10 folding cross validations:It is 10 that stochastic averagina, which divides training dataset, Folding, other to roll over for training pattern for each folding, the folding is used for the performance for testing the model trained;Average each folding As a result the final excellent Generalization Capability of model is obtained;It is respectively [0.1,0.2,0.3 ..., 1.0] to set C1 and C2 spans;It is right In with the corresponding C1 and C2 values of optimal Generalization Capability model,
Step 4, setOn overall training set, final mask, profit are trained using iterative method The parameter w solved with object function;Iterative process with for:
Wherein w (k) is kth time iteration w value,It is gradient of the object function at w (k) places, Hk represents kth time The Hesse matrixes of iterative target function, its initial value H0 is unit matrix, and Hk is calculated using following alternative manner:
Wherein p(k)=w(k+1)-w(k),
Step 5, list processing forecast sample x ' is described using the information of step 1, calculates the output function y=of SVMs Wx '+b, so the class that is associated with forecast sample x ' of model prediction marked as:
Train a supporting vector machine model so that normal sample falls in the side of segregation section, as shown in Fig. 1 "+";Credit Swindle sample falls another survey in segregation section, as shown in Fig. 1 "○";And doubtful credit swindle sample falls in the inside of segregation section, Such as Fig. 1It is shown.
The class of normal, doubtful and credit swindle sample is mapped respectively marked as+1,0 and -1.Operate, standardize for convenience Supporting vector to hyperplane (dotted line in Fig. 1) distance be 1.
In addition, the forecast sample x ' in step 5 is available:
Substituted into after non-linearization The output function of SVMs.Obtain result as shown in Figure 2:Normally, the differentiation degree of doubtful fraud, fraud is than linear point Class is more accurate.
The missing attribute values of the sample in forecast sample x ' in step 5, can utilize same property value known to other samples Filled after averaging.
Experimental result:Commented, used a pair as evaluation index using accuracy rate (accuracy) and recall rate (recall) One SVMs is used as comparison other.Accuracy rate is defined as the ratio correctly classified, and recall rate is defined as doubtful credit swindle The ratio correctly classified in sample.It the results are shown in Table shown in 3.
Table 3:Experimental result table
In table 3, the output function SVM-3C-Linear of linear SVM and the output letter of Nonlinear Support Vector Machines Number SVM-3C-NonLinear are to support linear and non-linear method, SVM-Linear and SVM- in the inventive method respectively NonLinear represents man-to-man linear SVM and Nonlinear Support Vector Machines respectively.
Embodiments of the invention are the foregoing is only, are not intended to limit the scope of the invention, it is every to utilize this hair Equivalent structure or equivalent flow conversion that bright specification and accompanying drawing content are made, or directly or indirectly it is used in other related skills Art field, is included within the scope of the present invention.

Claims (4)

1. a kind of credit method for anti-counterfeit based on SVMs, it is characterised in that including:
Step 1, build one's credit loan variable information table is described;Training sample is obtained, table is described to training according to described information Sample is handled, and sets up training samples information table;
Step 2, object function is set up, and is returned support vector cassification strategy and SVMs using described object function Return tactful merging, described object function is:
<mrow> <munder> <mi>min</mi> <mrow> <mi>w</mi> <mo>,</mo> <mi>b</mi> </mrow> </munder> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <msup> <mi>w</mi> <mn>2</mn> </msup> <mo>+</mo> <msub> <mi>C</mi> <mn>1</mn> </msub> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>l</mi> </munderover> <mrow> <mo>(</mo> <msubsup> <mi>&amp;xi;</mi> <mi>i</mi> <mn>2</mn> </msubsup> <mo>+</mo> <msubsup> <mover> <mi>&amp;xi;</mi> <mo>^</mo> </mover> <mi>i</mi> <mn>2</mn> </msubsup> <mo>)</mo> </mrow> <mo>+</mo> <msub> <mi>C</mi> <mn>2</mn> </msub> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mi>l</mi> <mo>+</mo> <mn>1</mn> </mrow> <mrow> <mi>l</mi> <mo>+</mo> <mi>m</mi> <mo>+</mo> <mi>n</mi> </mrow> </munderover> <msubsup> <mi>&amp;xi;</mi> <mi>j</mi> <mn>2</mn> </msubsup> </mrow>
Wherein:
yi(w·xj)≥1-ξj, j=l+1, L, l+m+n,
(w·xi)-yi≤1+ξi, i=1, L, l,
<mrow> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>-</mo> <mrow> <mo>(</mo> <mi>w</mi> <mo>&amp;CenterDot;</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>&amp;le;</mo> <mn>1</mn> <mo>+</mo> <msub> <mover> <mi>&amp;xi;</mi> <mo>^</mo> </mover> <mi>i</mi> </msub> <mo>,</mo> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mi>L</mi> <mo>,</mo> <mi>l</mi> <mo>,</mo> </mrow>
ξj>=0, j=l+1, L, l+m+n,
<mrow> <msub> <mi>&amp;xi;</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mover> <mi>&amp;xi;</mi> <mo>^</mo> </mover> <mi>i</mi> </msub> <mo>&amp;GreaterEqual;</mo> <mn>0</mn> <mo>,</mo> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mi>L</mi> <mo>,</mo> <mi>l</mi> </mrow>
Wherein, w is the parameter of hyperplane, and l is the number of doubtful credit swindle sample in training sample table, and m and n are training respectively The number of credit swindle sample and normal sample in sample table, punishment weight when C1 is doubtful credit swindle mistake classification, C2 Punishment weight when being credit swindle sample or the classification of normal sample mistake;
Step 3, parameter C1 and C2 value is determined using 10 folding cross validations:It is 10 foldings that stochastic averagina, which divides training dataset, Other to roll over for training pattern for each folding, the folding is used for the performance for testing the model trained;The result of average each folding Obtain the final excellent Generalization Capability of model;It is respectively [0.1,0.2,0.3 ..., 1.0] to set C1 and C2 spans;For tool There are the corresponding C1 and C2 values of optimal Generalization Capability model,
Step 4, setOn overall training set, final mask is trained using iterative method, mesh is utilized The parameter w that scalar functions are solved;Iterative process with for:
<mrow> <msup> <mi>w</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </msup> <mo>=</mo> <msup> <mi>w</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> <mo>-</mo> <msub> <mi>H</mi> <mi>k</mi> </msub> <mo>&amp;dtri;</mo> <mi>L</mi> <mrow> <mo>(</mo> <msup> <mi>w</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> <mo>)</mo> </mrow> </mrow>
Wherein w (k) is kth time iteration w value,It is gradient of the object function at w (k) places, Hk represents kth time iteration The Hesse matrixes of object function, its initial value H0 is unit matrix, and Hk is calculated using following alternative manner:
<mrow> <msub> <mi>H</mi> <mrow> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> <mo>=</mo> <msub> <mi>H</mi> <mi>k</mi> </msub> <mo>+</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mfrac> <mrow> <msup> <mi>q</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mi>T</mi> </mrow> </msup> <msub> <mi>H</mi> <mi>k</mi> </msub> <msup> <mi>q</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> </mrow> <mrow> <msup> <mi>p</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mi>T</mi> </mrow> </msup> <msup> <mi>q</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> </mrow> </mfrac> <mo>)</mo> </mrow> <mfrac> <mrow> <msup> <mi>p</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> <msup> <mi>p</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mi>T</mi> </mrow> </msup> </mrow> <mrow> <msup> <mi>p</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mi>T</mi> </mrow> </msup> <msup> <mi>q</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> </mrow> </mfrac> <mo>-</mo> <mfrac> <mrow> <msup> <mi>p</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> <msup> <mi>q</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mi>T</mi> </mrow> </msup> <msub> <mi>H</mi> <mi>k</mi> </msub> <mo>+</mo> <msub> <mi>H</mi> <mi>k</mi> </msub> <msup> <mi>q</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> <msup> <mi>p</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mi>T</mi> </mrow> </msup> </mrow> <mrow> <msup> <mi>p</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mi>T</mi> </mrow> </msup> <msup> <mi>q</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </msup> </mrow> </mfrac> </mrow>
Wherein
Step 5, list processing forecast sample x ' described using the information of step 1, calculate the output function y=wx ' of SVMs+ B, so the class that is associated with forecast sample x ' of model prediction marked as:
2. the credit method for anti-counterfeit according to claim 1 based on SVMs, it is characterised in that in step 1, institute The information stated describes table and fiduciary loan type of variables is divided into classification, numerical value, binary or discrete.
3. the credit method for anti-counterfeit according to claim 1 based on SVMs, it is characterised in that in steps of 5, institute The forecast sample x ' stated is available:
Substituted into after non-linearization the output function of SVMs.
4. the credit method for anti-counterfeit according to claim 1 based on SVMs, it is characterised in that obtained in step 1 Training sample and step 5 in forecast sample x ' in sample missing attribute values, can be belonged to together using known to other samples Property value average after fill.
CN201710275338.1A 2017-04-16 2017-04-16 Credit method for anti-counterfeit based on SVMs Pending CN107133867A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710275338.1A CN107133867A (en) 2017-04-16 2017-04-16 Credit method for anti-counterfeit based on SVMs

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710275338.1A CN107133867A (en) 2017-04-16 2017-04-16 Credit method for anti-counterfeit based on SVMs

Publications (1)

Publication Number Publication Date
CN107133867A true CN107133867A (en) 2017-09-05

Family

ID=59716622

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710275338.1A Pending CN107133867A (en) 2017-04-16 2017-04-16 Credit method for anti-counterfeit based on SVMs

Country Status (1)

Country Link
CN (1) CN107133867A (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070005538A1 (en) * 2001-06-18 2007-01-04 Wisconsin Alumni Research Foundation Lagrangian support vector machine
US20090125466A1 (en) * 2007-10-12 2009-05-14 Kyle Randolph Hinsz Rapidly Maturing Expanded Data Transaction Profiles
CN102663643A (en) * 2012-03-16 2012-09-12 暨南大学 Method of farmers' multi-layer one-way network piecewise linear credit rating
CN104463673A (en) * 2014-12-22 2015-03-25 中国科学技术大学苏州研究院 P2P network credit risk assessment model based on support vector machine
CN105512938A (en) * 2016-02-03 2016-04-20 宜人恒业科技发展(北京)有限公司 Online credit risk assessment method based on long-term using behavior of user
CN106408184A (en) * 2016-09-12 2017-02-15 中山大学 User credit evaluation model based on multi-source heterogeneous data

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070005538A1 (en) * 2001-06-18 2007-01-04 Wisconsin Alumni Research Foundation Lagrangian support vector machine
US20090125466A1 (en) * 2007-10-12 2009-05-14 Kyle Randolph Hinsz Rapidly Maturing Expanded Data Transaction Profiles
CN102663643A (en) * 2012-03-16 2012-09-12 暨南大学 Method of farmers' multi-layer one-way network piecewise linear credit rating
CN104463673A (en) * 2014-12-22 2015-03-25 中国科学技术大学苏州研究院 P2P network credit risk assessment model based on support vector machine
CN105512938A (en) * 2016-02-03 2016-04-20 宜人恒业科技发展(北京)有限公司 Online credit risk assessment method based on long-term using behavior of user
CN106408184A (en) * 2016-09-12 2017-02-15 中山大学 User credit evaluation model based on multi-source heterogeneous data

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
周绮凤 等: "商业银行信用风险评估中"拒真纳伪"", 《厦门大学学报》 *
成洪静: "基于SVM的银行信贷风险评估模型研究", 《中国优秀硕士学位论文全文数据库 经济与管理科学辑》 *

Similar Documents

Publication Publication Date Title
Huang et al. Enterprise credit risk evaluation based on neural network algorithm
Harris Credit scoring using the clustered support vector machine
Karaivanov et al. Dynamic financial constraints: Distinguishing mechanism design from exogenously incomplete regimes
Pacelli et al. An artificial neural network approach for credit risk management
CN103258069B (en) A kind of Forecasting Methodology of steel industry electricity needs
Zhang et al. An improved SMO algorithm for financial credit risk assessment–evidence from China’s banking
Tang et al. Neural networks analysis in business failure prediction of Chinese importers: A between-countries approach
CN105279691A (en) Financial transaction detection method and equipment based on random forest model
Chung et al. Bankruptcy prediction using cerebellar model neural networks
CN114048436A (en) Construction method and construction device for forecasting enterprise financial data model
Zou et al. Geographic deregulation and commercial bank performance in US state banking markets
CN104463673A (en) P2P network credit risk assessment model based on support vector machine
Rasiah et al. Financing technological upgrading in East Asia
CN108805331A (en) A kind of electricity demand forecasting method
Sun et al. Does financial development hamper or improve the resource curse? Analysis based on the panel threshold effect model
CN107403391A (en) A kind of optimal screening method for oil field development pre-arranged project
Alonso de Armiño et al. Analysing the intermeshed patterns of road transportation and macroeconomic indicators through neural and clustering techniques
CN107292433A (en) Stock exchange behavior Forecasting Methodology based on double cluster results and naive Bayesian and AdaBoost
Simutis et al. A flexible neural network for ATM cash demand forecasting
Zeng [Retracted] Research on Risk Measurement and Early Warning of Electronic Banking Business Based on GMDH Algorithm
CN107133867A (en) Credit method for anti-counterfeit based on SVMs
Tang [Retracted] A Study of the Spatial Effects of Financial Support and Economic Growth under Optimal Control of Nonlinear Generalized Complex Systems
Sun et al. Dynamic financial distress prediction based on class-imbalanced data batches
Chen et al. A stable credit rating model based on learning vector quantization
Guo et al. Research and verification analysis on early warning model through support vector machine algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20170905