CN110009479A - Credit assessment method and device, storage medium, computer equipment - Google Patents
Credit assessment method and device, storage medium, computer equipment Download PDFInfo
- Publication number
- CN110009479A CN110009479A CN201910156799.6A CN201910156799A CN110009479A CN 110009479 A CN110009479 A CN 110009479A CN 201910156799 A CN201910156799 A CN 201910156799A CN 110009479 A CN110009479 A CN 110009479A
- Authority
- CN
- China
- Prior art keywords
- user
- credit
- fraud
- data
- model
- 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
- 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
- G06Q20/00—Payment architectures, schemes or protocols
- G06Q20/38—Payment protocols; Details thereof
- G06Q20/40—Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
- G06Q20/401—Transaction verification
- G06Q20/4016—Transaction verification involving fraud or risk level assessment in transaction processing
-
- 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
- G06Q20/00—Payment architectures, schemes or protocols
- G06Q20/38—Payment protocols; Details thereof
- G06Q20/40—Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
- G06Q20/403—Solvency checks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Technology Law (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
This application discloses a kind of credit assessment method and device, storage medium, computer equipments, this method comprises: obtaining the credit data of user and the application business of user；According to credit data and anti-fraud rule, anti-fraud investigation is carried out to user；If user is checked by fraud, model is cheated according to credit data and clique, clique's fraud investigation is carried out to user；If user is checked by clique's fraud, according to credit data and preceding credit scoring model is borrowed, calculates credit scoring before the loan of user；If credit scoring is greater than or equal to the corresponding default threshold value that scores of making loans of application business before the loan of user, made loans according to application business.The application is by checking scale and highly professional clique's fraud, and then the credit scoring according to user is made loans, and safety of making loans is improved, and helps to reduce bad credit rate.
Description
Technical field
This application involves credit appraisal technical fields, are situated between particularly with regard to a kind of credit assessment method and device, storage
Matter, computer equipment.
Background technique
As explosive growth is presented in the behavioral data that the fast development and internet of internet finance are precipitated,
Mass data is relied on, the personal credit file technology for serving internet financial business is just come into being.This is interconnected to increase
The examination & approval efficiency of finance is netted, accelerates that the general favour finance construction of country is promoted to produce positive power-assisted.
However at present in credit evaluation link, anti-fraud is mainly for identity fraud, information falseness, history fraud, agency
The fraud type such as people's fraud passes through identity verification, information comparison, blacklist comparison and agent's signature analysis (monitoring) etc.
Means carry out prevention and control.Scale and highly professional clique's fraud are lacked and do not checked, financial company is once
It encounters and will cause extreme loss.
Summary of the invention
In view of this, being helped this application provides a kind of credit assessment method and device, storage medium, computer equipment
It makes loans safety in raising.
According to the one aspect of the application, a kind of credit assessment method is provided characterized by comprising
Obtain the credit data of user and the application business of user；
According to the credit data and anti-fraud rule, anti-fraud investigation is carried out to the user；
If the user is checked by the fraud, model is cheated according to the credit data and clique, it is right
The user carries out clique's fraud investigation；
If the user is checked by clique's fraud, according to the credit data and preceding credit scoring is borrowed
Model calculates credit scoring before the loan of the user；
If credit scoring is greater than or equal to the corresponding default threshold value that scores of making loans of the application business before the loan of the user,
Then made loans according to the application business.
According to the another aspect of the application, a kind of credit appraisal device is provided characterized by comprising
First obtains module, for obtaining the credit data of user and the application business of user；
Anti- fraud investigation module is used for according to the credit data and anti-fraud rule, counter to the user to be taken advantage of
Swindleness behavior investigation；
Clique's fraud investigation module, if being checked for the user by the fraud, according to the credit number
Accordingly and model is cheated by clique, carries out clique's fraud investigation to the user；
Credit scoring computing module before borrowing, if being checked for the user by clique's fraud, according to institute
Credit scoring model before stating credit data and borrowing, calculates credit scoring before the loan of the user；
It makes loans module, if it is corresponding default to be greater than or equal to the application business for credit scoring before the loan of the user
It makes loans the threshold value that scores, is then made loans according to the application business.
According to the application another aspect, a kind of storage medium is provided, computer program, described program are stored thereon with
Above-mentioned credit assessment method is realized when being executed by processor.
According to the application another aspect, a kind of computer equipment is provided, including storage medium, processor and be stored in
On storage medium and the computer program that can run on a processor, the processor realize above-mentioned credit when executing described program
Evaluation method.
By above-mentioned technical proposal, a kind of credit assessment method and device, storage medium, computer provided by the present application are set
It is standby, by carrying out the investigation of anti-fraud and clique's fraud to the user of submission business application, thus to row is passed through
The user looked into carries out borrowing preceding credit scoring, and the business for borrowing preceding credit scoring and user's application is combined to make loans.Compared to
The identity fraud of identification user is relied primarily in the prior art, the modes services of screening such as deceptive information, history fraud are provided
User, the application is also by checking scale and highly professional clique's fraud, and then the letter according to user
It is made loans with scoring, improves safety of making loans, help to reduce bad credit rate.
Above description is only the general introduction of technical scheme, in order to better understand the technological means of the application,
And it can be implemented in accordance with the contents of the specification, and in order to allow above and other objects, features and advantages of the application can
It is clearer and more comprehensible, below the special specific embodiment for lifting the application.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present application, constitutes part of this application, this Shen
Illustrative embodiments and their description please are not constituted an undue limitation on the present application for explaining the application.In the accompanying drawings:
Fig. 1 shows a kind of flow diagram of credit assessment method provided by the embodiments of the present application；
Fig. 2 shows the flow diagrams of another credit assessment method provided by the embodiments of the present application；
Fig. 3 shows a kind of flow diagram of Rating Model method for building up provided by the embodiments of the present application；
Fig. 4 shows a kind of structural schematic diagram of credit appraisal device provided by the embodiments of the present application；
Fig. 5 shows the structural schematic diagram of another credit appraisal device provided by the embodiments of the present application.
Specific embodiment
The application is described in detail below with reference to attached drawing and in conjunction with the embodiments.It should be noted that not conflicting
In the case of, the features in the embodiments and the embodiments of the present application can be combined with each other.
A kind of credit assessment method is provided in the present embodiment, as shown in Figure 1, this method comprises:
Step 101, the credit data of user and the application business of user are obtained.
When user has business application demand, the credit data of user and the application business of user, the letter of user are obtained
With data may include the identity information of user, income proves, assets prove, consumption information, reference information etc., apply for business
It may include the loan limit that user submits.
Step 102, according to credit data and anti-fraud rule, anti-fraud investigation is carried out to user.
Using acquisition credit data and pre-establish anti-fraud rule, judge one by one user credit data whether
Meet the requirement of anti-fraud rule, if meeting the regulation of anti-fraud rule, illustrates that personal fraud is not present in user, if not
The regulation for meeting anti-fraud rule, then illustrate that there may be frauds by user, if make loans user will be present it is higher bad
Account possibility, therefore the business of user's application should not be accepted.
Step 103, if user is checked by fraud, model is cheated according to credit data and clique, to user
Carry out clique's fraud investigation.
If the regulation for meeting anti-fraud rule is checked by anti-fraud, according further to the credit data of user,
Clique's fraud of user is checked using clique's fraud model.Since clique's fraud has scale and profession
Property, financial company, which once encounters, will cause extreme loss, therefore carry out clique's fraud investigation to user, facilitate enterprise's rule
Keep away the loss such as bad credit.
Step 104, if user is checked by clique's fraud, according to credit data and preceding credit scoring mould is borrowed
Type calculates credit scoring before the loan of user.
If user has passed through anti-fraud investigation and clique's fraud investigation, illustrate user there is no personal and
The fraud of clique, then should further data according to the user's credit, using credit scoring model before borrowing, before the loan of user
Credit scoring is assessed.
Step 105, if credit scoring is greater than or equal to the corresponding default threshold value that scores of making loans of application business before the loan of user,
Then made loans according to application business.
If higher by the credit scoring for calculating user, that is, it is greater than or equal to corresponding default make loans of its applied business and comments
Divide threshold value, illustrates that user meets loan application condition, the subsequent risk for generating bad credit in gathering is lower, then can be according to user
Applied business realizing is made loans.
Technical solution by applying this embodiment, by submit the user of business application carry out anti-fraud and
The investigation of clique's fraud, to borrow preceding credit scoring to the user by checking, and combine borrow before credit scoring with
And the business of user's application is made loans.The identity fraud of identification user is relied primarily in compared with the prior art, falseness is provided
The modes such as information, history fraud screen service user, and the application is also by taking advantage of scale and highly professional clique
Swindleness behavior is checked, and then the credit scoring according to user is made loans, and safety of making loans is improved, and helps to reduce bad credit
Rate.
Further, as the refinement and extension of above-described embodiment specific embodiment, in order to completely illustrate the present embodiment
Specific implementation process, provide another credit assessment method, as shown in Fig. 2, this method comprises:
Step 201, the credit data of user and the application business of user are obtained.
Step 202, according to credit data and anti-fraud rule, anti-fraud investigation is carried out to user.
Obtain user credit data and user application business after, first with user credit data and in advance
The anti-fraud rule of setting, carries out the investigation of anti-fraud.
It should be noted that the anti-fraud rule in the application establishes mode and anti-fraud investigation method is mainly wrapped
It includes:
Step 2021, the credit data of sample of users is obtained, wherein credit data includes a variety of.
For example, the credit data of user may include the identity information of user, history loan defaults information, jurisdictional information etc.
Deng.
Step 2022, according to default maximum case number, default minimum leaf segment points and default minimum leaf node accounting, to sample
The information data of this user carries out the processing of decision tree branch mailbox, obtains multiple branch mailbox.
The embodiment of the present application carries out optimal characteristics Variable Selection (i.e. the screening of credit data) using branch mailbox mode, presses first
According to default maximum case number, default minimum leaf segment points and default minimum leaf node accounting, to the information data of sample of users into
The processing of row decision tree branch mailbox.It should be noted that minimum leaf segment points can be chosen according to sample size in the embodiment of the present application, one
As the sample size of every case be all larger than minimum leaf segment points, such as sample size minimum leaf segment points at 50,000 or more are set as 100,
Or it is bigger, minimum leaf segment points are set as 50 when sample size is below 50,000.In addition, the sample size in each branch mailbox accounts for sample
The specific gravity of total amount should be greater than the ratio that the corresponding sample size of minimum leaf node accounts for non-empty sample size.
In addition, rule formulation generally require consideration interpretation, participate in Rulemaking variable much contain it is directional
Information, such as in debt-credit intention, debt-credit number is more, bad sample (bad sample refer mainly to there are the samples of bad credit behavior) it is general
Rate is bigger, in order to make the rule finally filtered out meet business interpretation, can assign inceptive direction 1 or -1 to variable, and right
Branch mailbox data carry out respective handling, to enhance the interpretation of variable.
Step 2023, it calculates the negative of each branch mailbox and comments rate, filter out the negative branch mailbox for commenting rate to comment rate greater than preset negative.
It calculates separately the negative of each branch mailbox and comments rate, bear rate is commented to refer to that bad sample accounts for the ratio of whole samples in the present embodiment
Example filters out the negative branch mailbox for commenting rate to be greater than preset value, to tentatively establish using the branch mailbox filtered out monotropic from whole branch mailbox
Gauge is then.In addition, in addition to commenting rate to be screened using negative other branch mailbox quality can also can be embodied using minimum lift degree etc.
Content carry out branch mailbox screening, do not do other citings herein.Rate size is commented to can be set it should be noted that bearing in the application
It is 30%, minimum lift degree size can be set to 3.
Step 2024, according to the branch mailbox after screening, the corresponding anti-fraud rule of every kind of credit data is determined.
After screening to branch mailbox, according to the branch mailbox after screening, list corresponding to credit data corresponding with branch mailbox is determined
The anti-fraud rule of variable.Each branch mailbox can specifically be first passed through and determine corresponding single argument rule, then judge each sample with
The hit relationship of every single argument rule, is adjusted single argument rule.
Step 2025, calculate separately corresponding anti-the negative of fraud rule of every kind of credit data and comment rate, and according to it is negative comment rate from
Small sequence anti-fraud rule corresponding to every kind of credit data is arrived greatly to be ranked up.
It calculates separately and instead cheats rule according to single argument obtained above and negative comment rate, this reality to what sample of users was checked
Apply born in example comment rate value instead cheat in rule judgement sample user according to single argument there are the numbers of users of fraud to account for whole
The total quantity of sample of users comments the sequence of rate from big to small rule of cheating counter to single argument to be ranked up according to negative.
Step 2026, successively the anti-fraud rule after sequence is added in anti-fraud rule set, and calculates and is newly put into counter take advantage of
Cheat the related coefficient between the anti-fraud rule of others in the anti-fraud rule and anti-fraud rule set in rule set.
The anti-fraud investigation time is wasted in order to avoid establishing excessive rule using the higher variable of correlation, is increased
Workload is checked, the embodiment of the present application needs rule of cheating counter to above-mentioned single argument to carry out eliminating correlation processing.It establishes anti-
Rule set is cheated, anti-fraud rule set here is for storing the lower rule of correlation from each other, specifically, successively will
In the anti-fraud rule set of anti-fraud rule addition after sequence, and in one rule of new addition every time, it is new all to calculate this
Related coefficient in the rule and rule set of addition between other anti-fraud rules, to establish anti-fraud rule set.
Step 2027, if related coefficient is greater than preset correlation coefficient number, the anti-fraud in anti-fraud rule set will be newly put into
Redundant rule elimination.
The case where preset correlation coefficient number (such as 0.8) is greater than for related coefficient, the redundant rule elimination that will should be newly added, such as
The rule being newly added is a, and the related coefficient between regular b and a in rule set is 0.9, then deletes regular a.
Step 2028, the anti-fraud rule for including in anti-fraud rule set is intersected two-by-two, and will be anti-after intersection
Fraud rule is put into anti-fraud rule set.
In rule is established, the effect of rule will be established jointly using two variables than using only two variables sometimes
The effect for establishing two rules is good, and therefore, the embodiment of the present application intersects any two rule in anti-fraud rule set.
Step 2029, it calculates separately the negative of each anti-fraud rule in anti-fraud rule set and comments rate, and comment rate according to negative
Determine final anti-fraud rule.
Recycle the negative of every rule in anti-fraud rule set that rate is commented to screen final anti-fraud rule, specifically
For, it comments rate descending to arrange according to negative rule in rule set, and be sequentially placed into effective rule set, is put into effective rule set
In rule need to meet, be greater than the bad sample size of effective rule set hit and preset bad sample size, wherein preset bad sample
Quantity is related with the regular quantity in effective rule set, such as it is 10 that bad sample smallest incremental, which is arranged,.Finally, in effective rule set
The anti-fraud that as finally determines of rule it is regular.
Step 2020, anti-fraud investigation is carried out to user according to final anti-fraud rule.
It is right according to the anti-fraud rule finally determined after establishing anti-fraud rule using above-mentioned steps 2021 to step 2029
User carries out anti-fraud investigation.
Step 203, if user is checked by fraud, model is cheated according to credit data and clique, to user
Carry out clique's fraud investigation.
In the embodiment of the present application, the step of clique's fraud investigation includes:
Step 2031, believed according to the identity information of user, social information, geographical location information, consumption information, lend-borrow action
Breath obtains the credit data and credit attribute tags with the other users of user-association in credit database；
Step 2032, according to the credit data of user and other users, user and each other users are calculated separately
The degree of association；
Step 2033, according to the degree of association and the credit attribute tags of other users, to clique's fraud of user into
Row investigation.
Traditional anti-fraud cheats type mainly for identity fraud, information falseness, history fraud, agent's fraud etc.,
It is compared by identity verification, information comparison, blacklist and the means such as agent's signature analysis (monitoring) carries out effective prevention and control,
But for recessive clique's fraud, difficulty has effective means to be taken precautions against.And on the other hand, clique is cheated due to scale
Property and it is professional, financing corporation once meets, and loss is very big.Therefore, the embodiment of the present application checks clique's fraud.
The present embodiment is by including identity information, social information, geographical location information, consumption information, lend-borrow action information
Etc. magnanimity, multidimensional data, find being associated between user and the other users in database, and further calculate user with it is above-mentioned
Association user between the degree of association, i.e., association between user is strong and weak, thus according to the degree of association and the credit of association user
Attribute tags determine business application user with the presence or absence of clique's fraud.
Step 204, if user is checked by clique's fraud, according to credit data and preceding credit scoring mould is borrowed
Type calculates credit scoring before the loan of user.
It should be noted that credit scoring model and the self-healing hereinafter referred to scoring before loan in the embodiment of the present application
Model and collection Rating Model are all made of following methods foundation, as shown in figure 3, this method comprises:
Step 301, any model before borrowing in credit scoring model, self-healing Rating Model and collection Rating Model is obtained
Corresponding training sample, wherein before any model is to borrow when credit scoring model, training sample includes the credit of sample of users
Data, when any model is self-healing Rating Model or collection Rating Model, training sample includes the behavioral data of sample of users
And debt data.
The corresponding training sample of each model is obtained, for example, the corresponding training sample of credit scoring model includes sample before borrowing
The multinomial credit data of this user, using credit data as characteristic variable needed for model training.The purpose of model foundation is root
Distribution situation of the credit scoring on user's sample, which determines, before borrowing according to user optimal by refusal threshold point and corresponding passes through
Rate and pass through bad credit rate.
Step 302, the multiple characteristic variables for including according to training sample, carry out the derivative of characteristic variable, obtain new spy
Levy variable.
The quality of data used in model training will directly influence the training effect of model, and therefore, the embodiment of the present application needs
The multiple characteristic variables for including in training sample are centainly handled, so that the characteristic variable generated is able to ascend model instruction
Practice effect.
Firstly, derivation process is carried out to feature, except the feature deriving method on basis, as the frequency summarizes, numerical value sums it up, average
Value, consistency, standard deviation, coefficient of variation etc., and done outside the derivative variable of intersection by the methods of decision tree, linear fit, also
It can integrate using xgboost, neural network and AutoEncoder (autocoder) scheduling algorithm, construction more various dimensions are stronger
The derivative variable of effect.Characteristic variable deriving method is exemplified below:
1) decision tree: when deriving using the intersection that the non-root node of decision tree carries out variable, generally to consider that business is explained
Property, be conducive to generate and distinguish that effect is strong, the more variables of branch mailbox dimension；
2) Xgboost: it can be directed to a certain class variable (or the single unconspicuous variable of variable effect of a few classes), used
Xgboost generates the stronger generalized variable of predictive ability (utilizing the intermediate result of Xgboost)；
3) neural network: neural network algorithm passes through essence other than it can be efficiently used for the model training of classification problem
The heart designs its input variable and hiding layer parameter, and the node of hidden layer can be made to become effective compound characteristics.
4) Autoencoder: autocoder by by depth network settings be identical mapping, i.e., so that output valve with it is defeated
Enter to be worth equal, then by backpropagation scheduling algorithm, calculates the mapping relations between intermediate each hiding node layer.Due to hiding
The quantity of node layer is traditionally arranged to be less compared with input variable, thus may finally realize and excavate while variable dimensionality reduction
Imitate the target of feature.
Step 303, new characteristic variable corresponding to training sample carries out branch mailbox processing.
The most important purpose of variable branch mailbox is so that the stability of variable enhances, and it is excessively quasi- from single argument dimension to reduce model
The risk of conjunction, while can be that model introduces non-linear, lift scheme ability to express by characteristic crossover after sliding-model control.It removes
Except common contour, wide branch mailbox method, can also using there is the optimal branch mailbox method of supervision, such as:
1) card side's branch mailbox: it is preferential to merge two small casees of chi-square value by calculating the chi-square value between adjacent branch mailbox, and so on
Until termination condition meets；
2) decision tree branch mailbox, random forest branch mailbox, Xgboost branch mailbox: since tree-model can be referring to certain in growth
Standard (comentropy, Gini impurity level or loss function etc.), has contained some information relevant to simulated target, therefore such side
Method mainly utilizes the intermediate result of tree-model, finds the reasonable branch mailbox threshold value of variable.Compared to the decision for being usually used in single argument branch mailbox
Branch mailbox is set, it is both rear that usually can a large amount of variables be carried out with branch mailbox simultaneously.
3) WOE/IV branch mailbox: each step is found so that the maximum merging of variable IV value or division critical point, so circulation are straight
To meeting termination condition.
Above-mentioned branch mailbox method, which is suitable for continuous variable, generally may map to sequence of natural numbers for Ordinal variable
It is handled afterwards using the branch mailbox method of similar continuous variable.For Nominal variable, then it is contemplated that being encoded using WOE
Branch mailbox is carried out afterwards or the worry of branch mailbox is saved using one-hot coding.
Step 304, each branch mailbox is calculated separately treated the corresponding separating capacity of characteristic variable, according to separating capacity pair
Characteristic variable is screened.
Feature Selection i.e. realize characteristic variable dimensionality reduction, reduce model over-fitting risk while, improve model stability and
Model training efficiency.
The method of Feature Selection is very more, mainly screens according to each characteristic variable to the influence power of scoring,
That is the separating capacity of characteristic variable, be integrated with here common methods such as single argument feature selecting (such as Pearson correlation coefficient, IV,
Geordie variance etc.), variable importance sequence (such as result of XGBoost, Random Forest algorithm), (such as L1 is just for regularization
Then/Lasso), Variable cluster etc..Live the combination innovated to these methods, cross-reference.Feature Selection mode can be as
Under:
1) IV (Information Value): variable information value is used to measure separating capacity of the variable on fine or not sample:
Wherein, DistributionBad indicates that the bad sample number in the i of some section accounts for the ratio of all bad sample numbers,
DistributionGood indicates that the good sample number in the i of some section accounts for the ratio of all good sample numbers.The comprehensive reflection of IV value
Quality is distributed in the difference on the variable out, and IV value more High Defferential is bigger, and the variable is stronger for the separating capacity of quality；
2) Feature Importance: variable importance (feature importance) sequence, XGBoost with
Random Forest algorithm can output variable feature importance, can be used as the foundation of selection variables；
3) Lasso (least absolute shrinkage and selection operator): briefly,
LASSO is to joined L1 regular terms in generalized linear model, and due to regular terms non-zero, the feature institute that this just forces those weak is right
The coefficient answered becomes 0.The presence of regular terms makes generalized linear model it is possible to prevente effectively from over-fitting, and L1 norm is in regular terms
In use LASSO allowed to play the role of Variable Selection.
4) Variable cluster: the variable after standardization is considered as " individual ", between the related coefficient description " individual " between variable
Similarity degree.After Variable cluster, the thinking for representing feature from all kinds of middle selections can refer to preset expert opinion library, this feature
Representativeness to class where it or with the correlation of explained variable etc..
5) using BIC or RMSE as optimization aim, variable is rejected from variable pond one by one and is modeled (in view of calculating effect
Rate uses Weak Classifier), until optimization aim is got and is most worth.
Step 305, according to the characteristic variable training Rating Model after screening.
In model training link, the embodiment of the present application can use naive Bayesian, support vector machines, and logistic regression (is commented
Point card), random forest, XGBoost, LightGBM and neural network etc. tradition and advanced algorithm.Also, in order to further
The effect and stability of lift scheme, can also introduce Stacking technology in training process, i.e., to being exported by above-mentioned classifier
As a result fusion again is carried out to obtain final prediction label, and amalgamation mode can apply the realization of simple classification device.
Join link (for the model containing hyper parameter) in the tune of model training, uses GridSearch (grid search) etc.
The function of adjusting ginseng automatically may be implemented in technology.Model needs just dispose after verifying and passing through online for production after generating
It calls.Pass through K-S (model discrimination index) value, GINI (Gini coefficient), AUC statistic (area under ROC curve), PSI
Equal model testings index verifies model overall predictivity, stability.Wherein, common K-S index has been Customer Score
It is distributed the maximum value of the difference of cumulative percentage and bad Customer Score distribution cumulative percentage.K-S index is higher, show hospitable family and
The distance between bad client is bigger, and the separating capacity of model is stronger.PSI (Population Stability Index) is used to weigh
The stability that amount model shows on different samples.It is generally acknowledged that model stability is very high when PSI is less than 0.1；If PSI between
Although then showing that model is available but stability needs give more sustained attention when between 0.1-0.25；Illustrate model if PSI is greater than 0.25
Stability is poor, needs to select a good opportunity and rebuild to model.
In addition, can also be optimized to model after model foundation, such as can by semi-supervised learning model optimization method
With for correcting to the optimization of misalignment (or unstability) model, the model after optimization can be applied to refuse that part precisely fish out back that grade is smart to model
More demanding scene is spent, or can be used for business initial stage sample performance phase deficiency not established supervision credit evaluation mould
The scene of type (y label is insufficient).By excavating the value of existing credit scoring and data product, fusion Unsupervised clustering algorithm and
The Supervised classification algorithm on a variety of bases is first realized to semi-supervised learning mechanism repetitive exercise is used after small sample mark, in turn
Originally classify to bulk sample.Precisely effective credit evaluation model is established based on classification results.
1) data prediction: Missing Data Filling, category feature numerical value conversion, feature normalization；
2) dimensionality reduction: by principal component analysis (PCA), take the main variables of explained variance accounting 80% or so as subsequent
The feature of model training, realizing under the premise of retaining information content effectively will dimension；
3) the accurate mark of small sample: hundred, which melt credit scoring, has certain fine or not separating capacity.Take scoring it is lower (risk compared with
It is high) sub-fraction sample (k=2) iteration is clustered by multiple KMeans, each iteration takes scoring water as bad sample set
Putting down lower sample set is mark bad sample set, similarly to part sample mark at good sample set, as beating for semi-supervised learning
Mark sample set；
4) iteration marks: the sample set based on mark, using the learning machine of logistic regression and decision-tree model dual model
System, continuous iteration, be iterating through every time logistic regression and decision tree jointly determination can mark sample set, until all samples beat
Mark.
5) Boosting: when containing sample with true tag on a small quantity, pass through full dose mark sample after iteration label
This exploitation overall model.Boosting method can be introduced later, and it is above-mentioned to improve weight progress to the sample of artificial mark mistake
The iteration optimization of model, every step optimization generate new Weak Classifier, and the model of final output is the weighting of all Weak Classifiers, power
Weight is linked up with each Weak Classifier with the error rate on true tag sample.
It should be noted that the above-mentioned Rating Model method for building up in the application be not limited to borrow before credit scoring model,
Self-healing Rating Model and collection Rating Model establish work, can be applicable to the Risk-warning and two in stage in borrowing
Other purposes such as secondary marketing intention determination.
Step 205, if credit scoring is greater than or equal to the corresponding default threshold value that scores of making loans of application business before the loan of user,
Then made loans according to application business.
In addition, before making loans, the maximum for needing to assess user makes loans amount and gathering interest rate, specific steps include:
Step 2051, it according to amount calculation formula of making loans, calculates and makes loans amount to the maximum of user, amount of making loans calculates public
Formula is
Wherein, x indicate user loan before credit scoring credit scoring before the loan of sample of users sort from low to high in it is right
The quantile answered, y (x) indicate the corresponding amount of making loans of user, AmeanIndicate amount of averagely making loans, [Amin,Amax] it is volume of making loans
Section is spent,
Historical sample analysis based on a large amount of financial clients, under same credit risk level, amount in certain section
Height there is no apparent risks to distinguish, that is to say, that whether applicant can break a contract, mainly related with its risk level, and
Obvious relation is had no with its amount.Based on this theory, enterprise thinks the client low for risk, gives high amount as far as possible；
For the high client of risk, low amount is given as far as possible, can effectively promote the earning rate for the amount of money of having made loans.It therefore, is realization wind
Dangerous differentiation, the amount of money bad credit rate target constant lower than population bad credit rate, the equal amount put on of part, give the equal amount A of partmean, volume
Spend section [Amin,Amax].In addition, the above-mentioned amount calculation formula of the application is to establish based on linear function, it is also based on finger
The corresponding amount calculation formula of foundation, the other modes such as number function, power function do not illustrate herein.
In addition, on the basis of above-mentioned formula, it can considering from other dimensions of superposition.Such as: the income level of client, from individual
The angle of debt ratio, the client relatively low to personal liability improve amount, reduce amount on the contrary；Preference is horizontal by stages, from profit
Angle, it is opposite to the client that preference by stages is high to improve amount, reduce amount on the contrary；The level of consumption, from the angle of customer demand,
It is opposite to the client that consuming capacity is vigorous to improve amount, it is opposite to reduce amount etc..
Step 2052, according to interest rate calculation formula, the gathering interest rate of user is calculated, interest rate calculation formula is
Wherein, r indicates the corresponding gathering interest rate of user, the corresponding bad credit rate of credit scoring before the loan of p expression user, ro=
rmean(1-pmean)-pmean, rmeanIndicate interest rate of averagely collecting money, pmeanIndicate average bad credit rate.
Specifically, for top-tier customer (low-risk client), more preferential interest rate is given, promotes top-tier customer experience；
For secondary client (high risk client), higher interest rate level is provided, to cover the bad debt risk being likely to occur.
Furthermore, it is contemplated that the high client of credit scoring is the objective group that market is liked, selectable product is more, may be for valence
The sensibility of lattice is high, conversely, credit scoring it is high client it is low for the sensibility of price.Accordingly, adjustable above-mentioned formula are as follows:β (p) can be the corresponding regulation coefficient of client.
Step 2053, according to the maximum of user make loans amount, gathering interest rate and application business, make loans.
It calculates maximum and makes loans amount with after gathering interest rate, in conjunction with the application business of user, make loans within the scope of amount in maximum
It makes loans to user, in addition, generating corresponding collection amount and payment collection time.
Step 206, within the survival phase of making loans, the behavioral data of user is obtained.
After making loans to user, the stage in loan can also carry out secondary battalion to high-quality user according to the behavioral data of user
Pin.
Step 207, according to behavioral data, the risk class of user is evaluated.
According to the behavioral data of user, the risk class of user is evaluated, such as from low to high by the risk of user
It is divided into tetra- grades of A, B, C, D, excavation is further worth for the user of A grades and B grades, in addition, for C grades and D grades of use
Family can carry out Risk-warning prompt in advance.
Step 208, if the risk class of user is greater than or equal to secondary marketing risk grade, according to credit data, divide
Analyse the secondary sales service being adapted to user.
Value is excavated mainly judges its economic strength class according to data such as income, the assets of user, and further
According to the analysis of the other information of the economic strength class of user and user to the secondary marketing product of user, to user in net
After state is verified, networked users are determined as to the target user of secondary marketing, are used to be combined for secondary marketing product
The information such as family credit data calculate make loans amount and the gathering interest rate of secondary marketing product, and then touch and reach client, if user receives
Product can then make loans to user is secondary.In addition, can specifically understand user's if user does not receive secondary marketing product
Reason For Denial, so that the formulation for subsequent secondary marketing product is adjusted.
Step 209, within the refund self-healing phase, the debt data of user are obtained.
Step 210, according to credit data, behavioral data and self-healing Rating Model, the self-healing scoring of user is calculated.
Step 211, according to the self-healing scoring of user and debt data, determine that refund is mentioned from the overdue interior refund to user
Show mode.
In step 209 to step 211, within shorter refund self-healing phase overdue time of refunding, the debt of user is obtained
The debts data such as time and amount owed, and self-healing Rating Model is utilized, the self-healing scoring of user is calculated, thus according to self-healing
Scoring and debt data determine corresponding refund prompting mode, carry out collection.
Step 212, within the refund collection phase, according to credit data, behavioral data and collection Rating Model, user is calculated
Collection scoring；
Step 213, it according to the collection scoring of user and debt data, determines and the refund of user is mentioned in the refund collection phase
Show mode.
In addition, in step 212 and step 213, it is similar with the refund self-healing phase, in overdue time longer refund of refunding
In the collection phase, determine also that corresponding refund prompting mode carries out collection.
For example, table 1 shows a kind of collection prompt table of the embodiment of the present application, by determining that self-healing scoring and collection are scored
Grade, in conjunction with debt aging, to determine refund prompting mode using the table.
Table 1
As explosive growth is presented in the behavioral data that the fast development and internet of internet finance are precipitated,
Mass data is relied on, the personal credit file technology for serving internet financial business is just come into being.This is interconnected to increase
The examination & approval efficiency of finance is netted, accelerates that the general favour finance construction of country is promoted to produce positive power-assisted.
In this tide, equipment is counter to be cheated, advanced machine learning, intelligent algorithm etc. fresh-core technique quilt one after another
It is introduced into the field of personal big data air control.
However it is on the one hand limited by the thin of air control consciousness, it is on the other hand limited to the air control idea of traditional financial industry,
It is many that financial company is when carrying out internet financial business or air control mode is excessively rough or air control means are too cautious
It is inefficient.It is happy to embrace the financial company of new technology and financial technology company even for most of, the air control technology used
Show the features such as covering surface is single, and accuracy is not high, and development efficiency is low.And work as the own service not yet amount of rising, or modeling
When the ratio of risky performance is too low in sample, many schemes are even more helpless.Meanwhile in terms of application scenarios, these sides
Case is often only conceived to some part in air control process, can not effectively support Life cycle risk management.In mating application
Tactful aspect, lacks the quantization logic combined with business.
Technical solution by applying this embodiment establishes the development approach of anti-fraud rule set, greatly improves rule
Then collect the efficiency of exploitation and ultimately generates the promotion effect of rule set；Clique's fraud investigation compensates in previous anti-fraud scheme
For the blank of clique's crime, cooperate personal anti-fraud rule set, can effectively before loan, borrow in link will be into part or storage
Fraud molecule catch all in one draft；The method for building up of Rating Model from feature derivative, branch mailbox, Feature Selection, model training & verifying etc.
Several aspects have carried out innovation and experience is integrated, and while so that model development efficiency greatly promotes, there has also been more for model accuracy
Height ensures.And the application scenarios of this programme run through from borrow before into loan to the Life cycle after loan, be related to business and cover loan
Customer portrait building after preceding assessing credit risks, loan risk investigation and early warning and loan when collection.It is matched from this system
From the point of view of strategy, examination & approval strategy facilitate financial institution quantization formulation match with financial objectives business objective (percent of pass and
Pass through bad credit rate etc.), and the amount and pricing strategy of risk differentiation can also further help financial institution to promote profitability,
Realize the risk management of fining.Facilitating financial company reduces financial risks, realizes maximum revenue.
Further, the specific implementation as Fig. 1 method, the embodiment of the present application provide a kind of credit appraisal device, such as
Shown in Fig. 4, which includes: the first acquisition module 41, anti-letter before module 43, loan is checked in fraud investigation module 42, clique's fraud
With scoring computing module 44, module of making loans 45.
First obtains module 41, for obtaining the credit data of user and the application business of user；
Anti- fraud investigation module 42, for carrying out anti-fraud to user according to credit data and anti-fraud rule
Investigation；
Clique's fraud investigation module 43, if being checked for user by fraud, according to credit data and clique
Model is cheated, clique's fraud investigation is carried out to user；
Credit scoring computing module 44 before borrowing, if being checked for user by clique's fraud, according to credit data
And credit scoring model before borrowing, calculate credit scoring before the loan of user；
Module of making loans 45 is commented if being greater than or equal to corresponding default make loans of application business for credit scoring before the loan of user
Divide threshold value, is then made loans according to application business.
In specific application scenarios, as shown in figure 5, the device further include: second obtains module 46, risk class determines
Module 47, secondary marketing module 48.
Second obtains module 46, for obtaining the behavioral data of user within the survival phase of making loans；
Risk class determining module 47, for evaluating the risk class of user according to behavioral data；
Secondary marketing module 48, if the risk class for user is greater than or equal to secondary marketing risk grade, basis
Credit data analyzes the secondary sales service being adapted to user.
In specific application scenarios, as shown in figure 5, the device further include: third obtains module 49, self-healing scoring calculates
Module 410, the first cue module 411, collection scoring computing module 412, the second cue module 413.
Third obtains module 49, for obtaining the debt data of user within the refund self-healing phase；
Self-healing scoring computing module 410, for calculating and using according to credit data, behavioral data and self-healing Rating Model
The self-healing at family is scored；
First cue module 411 is determined and is refunded from overdue interior right for the self-healing scoring and debt data according to user
The refund prompting mode of user；
Collection scoring computing module 412, is used within the refund collection phase, according to credit data, behavioral data and collection
Rating Model calculates the collection scoring of user；
Second cue module 413, for the collection scoring and debt data according to user, it is right in the refund collection phase to determine
The refund prompting mode of user.
In specific application scenarios, as shown in figure 5, module 45 of making loans, specifically includes: amount of making loans computing unit 451,
Gathering interest rate computing unit 452, unit 453 of making loans.
Amount of making loans computing unit 451 makes loans volume to the maximum of user for calculating according to amount calculation formula of making loans
Degree, amount of making loans calculation formula are
Wherein, x indicate user loan before credit scoring credit scoring before the loan of sample of users sort from low to high in it is right
The quantile answered, y (x) indicate the corresponding amount of making loans of user, AmeanIndicate amount of averagely making loans, [Amin,Amax] it is volume of making loans
Section is spent,
Gathering interest rate computing unit 452, for calculating the gathering interest rate of user according to interest rate calculation formula, interest rate is calculated
Formula is
Wherein, r indicates the corresponding gathering interest rate of user, the corresponding bad credit rate of credit scoring before the loan of p expression user, ro=
rmean(1-pmean)-pmean, rmeanIndicate interest rate of averagely collecting money, pmeanIndicate average bad credit rate；
It makes loans unit 453, makes loans amount, gathering interest rate and application business, make loans for the maximum according to user.
In specific application scenarios, as shown in figure 5, the device further include: the 4th obtains the derivative mould of module 414, feature
Block 415, branch mailbox module 416, Variable Selection module 417, training module 418.
4th obtains module 414, borrows preceding credit scoring model, self-healing Rating Model and collection Rating Model for obtaining
In the corresponding training sample of any model, wherein when any model be borrow before credit scoring model when, training sample includes sample
The credit data of this user, when any model is self-healing Rating Model or collection Rating Model, training sample includes that sample is used
The behavioral data and debt data at family；
Feature derives module 415, and multiple characteristic variables for including according to training sample carry out spreading out for characteristic variable
It is raw, obtain new characteristic variable；
Branch mailbox module 416, for carrying out branch mailbox processing to the corresponding new characteristic variable of training sample；
Variable Selection module 417, treated for the calculating separately each branch mailbox corresponding separating capacity of characteristic variable, root
Characteristic variable is screened according to separating capacity；
Training module 418, for according to the characteristic variable training Rating Model after screening.
In specific application scenarios, as shown in figure 5, clique's fraud investigation module 43, specifically includes: data capture unit
431, calculation of relationship degree unit 432, clique's fraud investigation unit 433.
First acquisition unit 431, for being believed according to the identity information of user, social information, geographical location information, consumption
Breath, lend-borrow action information obtain the credit data and credit attribute with the other users of user-association in credit database
Label；
Calculation of relationship degree unit 432, for the credit data according to user and other users, calculate separately user with it is each
The degree of association of a other users；
Clique's fraud investigation unit 433, for the credit attribute tags according to the degree of association and other users, to user's
Clique's fraud is checked.
In specific application scenarios, as shown in figure 5, anti-fraud investigation module 42, specifically includes: second acquisition unit
421, branch mailbox unit 422, branch mailbox screening unit 423, the first rule determination unit 424, rule compositor unit 425, related coefficient
Computing unit 426, rule set establish unit 427, regular cross unit 428, rule determination unit 429, anti-fraud investigation unit
4210。
Second acquisition unit 421, for obtaining the credit data of sample of users, wherein credit data includes a variety of；
Branch mailbox unit 422, for being accounted for according to default maximum case number, default minimum leaf segment points and default minimum leaf node
Than carrying out the processing of decision tree branch mailbox to the information data of sample of users, obtaining multiple branch mailbox；
Branch mailbox screening unit 423 comments rate for calculating the negative of each branch mailbox, filters out to bear and rate is commented to comment rate greater than preset negative
Branch mailbox；
First rule determination unit 424, for determining the corresponding anti-fraud of every kind of credit data according to the branch mailbox after screening
Rule；
Rule compositor unit 425 is commented rate for calculating separately corresponding anti-the negative of fraud rule of every kind of credit data, and is pressed
The sequence of rate from big to small anti-fraud rule corresponding to every kind of credit data is commented to be ranked up according to negative；
Related coefficient computing unit 426, for successively the anti-fraud rule after sequence to be added in anti-fraud rule set, and
It calculates between the anti-fraud rule of others in the anti-fraud rule and anti-fraud rule set being newly put into anti-fraud rule set
Related coefficient；
Rule set establishes unit 427, if being greater than preset correlation coefficient number for related coefficient, will newly be put into anti-fraud rule
The anti-fraud redundant rule elimination concentrated；
Regular cross unit 428, for the anti-fraud rule for including in anti-fraud rule set to be intersected two-by-two, and will
Anti- fraud rule after intersection is put into anti-fraud rule set；
Rule determination unit 429, for calculate separately each anti-fraud rule in anti-fraud rule set it is negative comment rate, and
Rate is commented to determine final anti-fraud rule according to negative；
Anti- fraud investigation unit 4210, for carrying out anti-fraud investigation to user according to final anti-fraud rule.
It should be noted that other of each functional unit involved by a kind of credit appraisal device provided by the embodiments of the present application
Corresponding description, can be with reference to the corresponding description in Fig. 1 and Fig. 2, and details are not described herein.
Based on above-mentioned method as depicted in figs. 1 and 2, correspondingly, the embodiment of the present application also provides a kind of storage medium,
On be stored with computer program, which realizes above-mentioned credit assessment method as depicted in figs. 1 and 2 when being executed by processor.
Based on this understanding, the technical solution of the application can be embodied in the form of software products, which produces
Product can store in a non-volatile memory medium (can be CD-ROM, USB flash disk, mobile hard disk etc.), including some instructions
With so that computer equipment (can be personal computer, server or the network equipment an etc.) execution the application is each
Method described in implement scene.
Based on above-mentioned method as shown in Figure 1 and Figure 2 and Fig. 3, virtual bench embodiment shown in Fig. 4, in order to realize
Above-mentioned purpose, the embodiment of the present application also provides a kind of computer equipments, are specifically as follows personal computer, server, network
Equipment etc., the computer equipment include storage medium and processor；Storage medium, for storing computer program；Processor is used
In execution computer program to realize above-mentioned credit assessment method as depicted in figs. 1 and 2.
Optionally, which can also include user interface, network interface, camera, radio frequency (Radio
Frequency, RF) circuit, sensor, voicefrequency circuit, WI-FI module etc..User interface may include display screen
(Display), input unit such as keyboard (Keyboard) etc., optional user interface can also connect including USB interface, card reader
Mouthful etc..Network interface optionally may include standard wireline interface and wireless interface (such as blue tooth interface, WI-FI interface).
It will be understood by those skilled in the art that a kind of computer equipment structure provided in this embodiment is not constituted to the meter
The restriction for calculating machine equipment, may include more or fewer components, perhaps combine certain components or different component layouts.
It can also include operating system, network communication module in storage medium.Operating system is management and preservation computer
The program of device hardware and software resource supports the operation of message handling program and other softwares and/or program.Network communication
Module is for realizing the communication between each component in storage medium inside, and between other hardware and softwares in the entity device
Communication.
Through the above description of the embodiments, those skilled in the art can be understood that the application can borrow
It helps software that the mode of necessary general hardware platform is added to realize, can also be passed through by hardware realization to submission business application
User carries out the investigation of anti-fraud and clique's fraud, so that credit is commented before being borrowed to the user by checking
Point, and the business for borrowing preceding credit scoring and user's application is combined to make loans.Identification is relied primarily in compared with the prior art
The modes such as identity fraud, offer deceptive information, the history fraud of user screen service user, and the application is also by scale
Property and highly professional clique's fraud are checked, and then the credit scoring according to user is made loans, and is improved and is put
Money safety helps to reduce bad credit rate.
It will be appreciated by those skilled in the art that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, module in attached drawing or
Process is not necessarily implemented necessary to the application.It will be appreciated by those skilled in the art that the mould in device in implement scene
Block can according to implement scene describe be distributed in the device of implement scene, can also carry out corresponding change be located at be different from
In one or more devices of this implement scene.The module of above-mentioned implement scene can be merged into a module, can also be into one
Step splits into multiple submodule.
Above-mentioned the application serial number is for illustration only, does not represent the superiority and inferiority of implement scene.Disclosed above is only the application
Several specific implementation scenes, still, the application is not limited to this, and the changes that any person skilled in the art can think of is all
The protection scope of the application should be fallen into.
Claims (10)
1. a kind of credit assessment method characterized by comprising
Obtain the credit data of user and the application business of user；
According to the credit data and anti-fraud rule, anti-fraud investigation is carried out to the user；
If the user is checked by the fraud, model is cheated according to the credit data and clique, to described
User carries out clique's fraud investigation；
If the user is checked by clique's fraud, according to the credit data and preceding credit scoring mould is borrowed
Type calculates credit scoring before the loan of the user；
If credit scoring is greater than or equal to the corresponding default threshold value that scores of making loans of the application business, root before the loan of the user
It makes loans according to the application business.
2. the method according to claim 1, wherein it is described made loans according to the application business after, institute
State method further include:
Within the survival phase of making loans, the behavioral data of the user is obtained；
According to the behavioral data, the risk class of the user is evaluated；
If the risk class of the user is greater than or equal to secondary marketing risk grade, according to the credit data, analysis with
The secondary sales service of user's adaptation.
3. according to the method described in claim 2, it is characterized in that, it is described made loans according to the application business after, institute
State method further include:
Within the refund self-healing phase, the debt data of the user are obtained；
According to the credit data, the behavioral data and self-healing Rating Model, the self-healing scoring of the user is calculated；
According to the self-healing of user scoring and the debt data, determine that described refund interior goes back the user from overdue
Money prompting mode；
Within the refund collection phase, according to the credit data, the behavioral data and collection Rating Model, the user is calculated
Collection scoring；
According to the collection of user scoring and the debt data, determines and the user is gone back in the refund collection phase
Money prompting mode.
4. specific to wrap according to the method described in claim 3, it is characterized in that, described make loans according to the application business
It includes:
It according to amount calculation formula of making loans, calculates and makes loans amount to the maximum of the user, the amount calculation formula of making loans is
Wherein, x indicates before the loan of the user that credit scoring credit scoring before the loan of sample of users is right in sorting from low to high
The quantile answered, y (x) indicate the corresponding amount of making loans of the user, AmeanIndicate amount of averagely making loans, [Amin,Amax] it is to put
Amount of money degree section,
According to interest rate calculation formula, the gathering interest rate of the user is calculated, the interest rate calculation formula is
Wherein, r indicates the corresponding gathering interest rate of the user, and p indicates the corresponding bad credit rate of credit scoring before the loan of the user,
ro=rmean(1-pmean)-pmean, rmeanIndicate interest rate of averagely collecting money, pmeanIndicate average bad credit rate；
According to the maximum of the user make loans amount, gathering interest rate and application business, make loans.
5. according to the method described in claim 4, it is characterized in that, establishing the loan respectively according to Rating Model method for building up
Preceding credit scoring model, the self-healing Rating Model and the collection Rating Model, the Rating Model method for building up include:
Obtain any model before the loan in credit scoring model, the self-healing Rating Model and the collection Rating Model
Corresponding training sample, wherein before any model is the loan when credit scoring model, the training sample includes sample
The credit data of this user, when any model is the self-healing Rating Model or the collection Rating Model, the instruction
Practice the behavioral data and debt data that sample includes the sample of users；
The multiple characteristic variables for including according to the training sample, carry out the derivative of characteristic variable, obtain new characteristic variable；
The new characteristic variable corresponding to the training sample carries out branch mailbox processing；
Each branch mailbox is calculated separately treated the corresponding separating capacity of characteristic variable, according to the separating capacity to the feature
Variable is screened；
According to the characteristic variable training Rating Model after screening.
6. according to the method described in claim 5, it is characterized in that, the credit data of the user includes at least the user's
Identity information, social information, geographical location information, consumption information, lend-borrow action information, it is described according to the credit data and
Model is cheated by clique, carries out clique's fraud investigation to the user, specifically includes:
According to the identity information of the user, social information, geographical location information, consumption information, lend-borrow action information, letter is obtained
With the credit data and credit attribute tags with the other users of the user-association in database；
According to the credit data of the user and the other users, the user and each described other users are calculated separately
The degree of association；
According to the degree of association and the credit attribute tags of the other users, clique's fraud of the user is carried out
Investigation.
7. according to the method described in claim 6, it is characterized in that, described according to the credit data and anti-fraud rule,
Anti- fraud investigation is carried out to the user, is specifically included:
Obtain the credit data of the sample of users, wherein the credit data includes a variety of；
According to default maximum case number, default minimum leaf segment points and default minimum leaf node accounting, to the sample of users
Information data carries out the processing of decision tree branch mailbox, obtains multiple branch mailbox；
It calculates each the negative of the branch mailbox and comments rate, filter out the negative branch mailbox for commenting rate to comment rate greater than preset negative；
According to the branch mailbox after screening, the corresponding anti-fraud rule of every kind of credit data is determined；
It calculates separately the negative of the corresponding anti-fraud rule of every kind of credit data and comments rate, and negative comment rate from big to small according to described
Sequence anti-fraud rule corresponding to credit data described in every kind be ranked up；
Successively the anti-fraud rule after sequence is added in anti-fraud rule set, and calculates and is newly put into the anti-fraud rule set
The anti-fraud rule and the anti-fraud rule set in the anti-fraud rule of others between related coefficients；
If the related coefficient is greater than preset correlation coefficient number, the anti-fraud being newly put into the anti-fraud rule set is advised
Then delete；
The anti-fraud rule for including in the anti-fraud rule set is intersected two-by-two, and the anti-fraud after intersection is advised
It is then put into the anti-fraud rule set；
It calculates separately the negative of each anti-fraud rule in the anti-fraud rule set and comments rate, and comment rate true according to described bear
The fixed final anti-fraud rule；
Anti- fraud investigation is carried out to the user according to the final anti-fraud rule.
8. a kind of credit appraisal device characterized by comprising
First obtains module, for obtaining the credit data of user and the application business of user；
Anti- fraud investigation module, for carrying out anti-fraud row to the user according to the credit data and anti-fraud rule
For investigation；
Clique fraud investigation module, if being checked for the user by the fraud, according to the credit data with
And model is cheated by clique, carries out clique's fraud investigation to the user；
Credit scoring computing module before borrowing, if being checked for the user by clique's fraud, according to the letter
With data and preceding credit scoring model is borrowed, calculates credit scoring before the loan of the user；
It makes loans module, if being greater than or equal to for credit scoring before the loan of the user, the application business is corresponding default to make loans
Score threshold value, then is made loans according to the application business.
9. a kind of storage medium, is stored thereon with computer program, which is characterized in that realization when described program is executed by processor
Credit assessment method described in any one of claims 1 to 7.
10. a kind of computer equipment, including storage medium, processor and storage can be run on a storage medium and on a processor
Computer program, which is characterized in that the processor is realized described in any one of claims 1 to 7 when executing described program
Credit assessment method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910156799.6A CN110009479B (en) | 2019-03-01 | 2019-03-01 | Credit evaluation method and device, storage medium and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910156799.6A CN110009479B (en) | 2019-03-01 | 2019-03-01 | Credit evaluation method and device, storage medium and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110009479A true CN110009479A (en) | 2019-07-12 |
| CN110009479B CN110009479B (en) | 2021-02-19 |
Family
ID=67166194
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910156799.6A Active CN110009479B (en) | 2019-03-01 | 2019-03-01 | Credit evaluation method and device, storage medium and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110009479B (en) |
Cited By (44)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110533190A (en) * | 2019-07-18 | 2019-12-03 | 武汉烽火众智数字技术有限责任公司 | A kind of data object analysis method and device based on machine learning |
| CN110728301A (en) * | 2019-09-09 | 2020-01-24 | 北京镭文科技有限公司 | Credit scoring method, device, terminal and storage medium for individual user |
| CN110727770A (en) * | 2019-08-30 | 2020-01-24 | 深圳壹账通智能科技有限公司 | Intelligent request screening method and device, computer equipment and storage medium |
| CN110727922A (en) * | 2019-10-11 | 2020-01-24 | 集奥聚合(北京)人工智能科技有限公司 | Anti-fraud decision model construction method based on multi-dimensional data flow |
| CN110766541A (en) * | 2019-09-25 | 2020-02-07 | 平安科技（深圳）有限公司 | Loan risk assessment method, loan risk assessment device, loan risk assessment equipment and computer-readable storage medium |
| CN110796539A (en) * | 2019-10-31 | 2020-02-14 | 中国工商银行股份有限公司 | Credit investigation evaluation method and device |
| CN110912874A (en) * | 2019-11-07 | 2020-03-24 | 苏宁云计算有限公司 | Method and system for effectively identifying machine access behaviors |
| 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 |
| CN110956303A (en) * | 2019-10-12 | 2020-04-03 | 未鲲(上海)科技服务有限公司 | Information prediction method, device, terminal and readable storage medium |
| CN111062444A (en) * | 2019-12-21 | 2020-04-24 | 湖南大学 | Credit risk prediction method, system, terminal and storage medium |
| CN111127026A (en) * | 2019-12-13 | 2020-05-08 | 深圳中兴飞贷金融科技有限公司 | Method, device, storage medium and electronic equipment for determining user fraud behavior |
| CN111144938A (en) * | 2019-12-20 | 2020-05-12 | 上海数策软件股份有限公司 | Sales lead rating method and system suitable for automobile industry |
| CN111178722A (en) * | 2019-12-20 | 2020-05-19 | 上海数策软件股份有限公司 | Machine learning systems, methods, and media suitable for sales lead rating and distribution |
| CN111353901A (en) * | 2020-02-24 | 2020-06-30 | 阳光保险集团股份有限公司 | Risk identification monitoring method and device and electronic equipment |
| CN111488338A (en) * | 2020-06-29 | 2020-08-04 | 上海冰鉴信息科技有限公司 | Model monitoring method and device applied to wind control decision flow |
| CN111582646A (en) * | 2020-04-09 | 2020-08-25 | 上海淇毓信息科技有限公司 | User policy risk early warning method and device and electronic equipment |
| CN111652711A (en) * | 2020-06-09 | 2020-09-11 | 深圳前海微众银行股份有限公司 | Pre-credit analysis method, device, equipment and storage medium based on collection data |
| CN111652712A (en) * | 2020-06-09 | 2020-09-11 | 深圳前海微众银行股份有限公司 | Pre-credit analysis method, device, equipment and storage medium based on geographic information |
| CN111709828A (en) * | 2020-06-12 | 2020-09-25 | 中国建设银行股份有限公司 | Resource processing method, device, equipment and system |
| CN111738819A (en) * | 2020-06-15 | 2020-10-02 | 中国建设银行股份有限公司 | Method, device and equipment for screening characterization data |
| CN111815439A (en) * | 2020-07-23 | 2020-10-23 | 睿智合创(北京)科技有限公司 | Credit scoring system based on cloud platform |
| CN111861703A (en) * | 2020-07-10 | 2020-10-30 | 深圳无域科技技术有限公司 | Data-driven wind control strategy rule generation method and system and risk control method and system |
| CN112396517A (en) * | 2020-12-04 | 2021-02-23 | 深圳前海微众银行股份有限公司 | ETC card management method, device and system and computer storage medium |
| CN112508677A (en) * | 2020-11-06 | 2021-03-16 | 无锡艺界科技有限公司 | Financial system based on big data wind accuse |
| CN112541706A (en) * | 2020-12-24 | 2021-03-23 | 四川享宇金信金融科技有限公司 | Big data wind control model construction method |
| CN112561685A (en) * | 2020-12-15 | 2021-03-26 | 建信金融科技有限责任公司 | Client classification method and device |
| CN112767136A (en) * | 2021-01-26 | 2021-05-07 | 天元大数据信用管理有限公司 | Credit anti-fraud identification method, credit anti-fraud identification device, credit anti-fraud identification equipment and credit anti-fraud identification medium based on big data |
| CN112819034A (en) * | 2021-01-12 | 2021-05-18 | 平安科技（深圳）有限公司 | Data binning threshold calculation method and device, computer equipment and storage medium |
| CN112862585A (en) * | 2020-12-02 | 2021-05-28 | 浙江惠瀜网络科技有限公司 | Personal loan type bad asset risk rating method based on LightGBM decision tree algorithm |
| CN113011971A (en) * | 2021-03-31 | 2021-06-22 | 深圳前海微众银行股份有限公司 | Risk measurement method, device and system and computer storage medium |
| CN113011752A (en) * | 2021-03-19 | 2021-06-22 | 天道金科股份有限公司 | Enterprise credit evaluation index system based on big data |
| CN113011751A (en) * | 2021-03-19 | 2021-06-22 | 天道金科股份有限公司 | Small and medium-sized micro enterprise credit evaluation method based on big data |
| CN113177585A (en) * | 2021-04-23 | 2021-07-27 | 上海晓途网络科技有限公司 | User classification method and device, electronic equipment and storage medium |
| CN113222632A (en) * | 2020-02-04 | 2021-08-06 | 北京京东振世信息技术有限公司 | Object mining method and device |
| CN113408923A (en) * | 2021-06-29 | 2021-09-17 | 中国平安人寿保险股份有限公司 | Premium collection method and device, computer equipment and storage medium |
| CN113590310A (en) * | 2021-07-02 | 2021-11-02 | 上海淇玥信息技术有限公司 | Resource allocation method and device based on rule touch rate scoring and electronic equipment |
| CN113610175A (en) * | 2021-08-16 | 2021-11-05 | 上海冰鉴信息科技有限公司 | Service strategy generation method and device and computer readable storage medium |
| CN113657481A (en) * | 2021-08-13 | 2021-11-16 | 上海晓途网络科技有限公司 | Model construction system and method |
| CN113902551A (en) * | 2021-10-20 | 2022-01-07 | 长安汽车金融有限公司 | Method and apparatus for vehicle loan application fraud identification |
| CN114155080A (en) * | 2021-09-29 | 2022-03-08 | 东方微银科技股份有限公司 | Fraud identification method, equipment and storage medium |
| CN114240215A (en) * | 2021-12-22 | 2022-03-25 | 中国建设银行股份有限公司 | User loss of contact grade acquisition method and device and storage medium |
| CN114926261A (en) * | 2022-04-26 | 2022-08-19 | 厦门大学 | Method and medium for predicting fraud probability of automobile financial user application |
| CN115034888A (en) * | 2022-06-16 | 2022-09-09 | 支付宝(杭州)信息技术有限公司 | Credit service providing method and device |
| CN115797044A (en) * | 2022-11-15 | 2023-03-14 | 东方微银科技股份有限公司 | Credit wind control early warning method and system based on cluster analysis |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9098699B1 (en) * | 2013-09-25 | 2015-08-04 | Emc Corporation | Smart television data sharing to provide security |
| CN106408413A (en) * | 2016-09-23 | 2017-02-15 | 快睿登信息科技（上海）有限公司 | Multi-cycle installment decision making method and system |
| CN107292424A (en) * | 2017-06-01 | 2017-10-24 | 四川新网银行股份有限公司 | A kind of anti-fraud and credit risk forecast method based on complicated social networks |
| CN108492173A (en) * | 2018-03-23 | 2018-09-04 | 上海氪信信息技术有限公司 | A kind of anti-Fraud Prediction method of credit card based on dual-mode network figure mining algorithm |
| CN108564286A (en) * | 2018-04-19 | 2018-09-21 | 天合泽泰（厦门）征信服务有限公司 | A kind of artificial intelligence finance air control credit assessment method and system based on big data reference |
| CN109035003A (en) * | 2018-07-04 | 2018-12-18 | 北京玖富普惠信息技术有限公司 | Anti- fraud model modelling approach and anti-fraud monitoring method based on machine learning |
-
2019
- 2019-03-01 CN CN201910156799.6A patent/CN110009479B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9098699B1 (en) * | 2013-09-25 | 2015-08-04 | Emc Corporation | Smart television data sharing to provide security |
| CN106408413A (en) * | 2016-09-23 | 2017-02-15 | 快睿登信息科技（上海）有限公司 | Multi-cycle installment decision making method and system |
| CN107292424A (en) * | 2017-06-01 | 2017-10-24 | 四川新网银行股份有限公司 | A kind of anti-fraud and credit risk forecast method based on complicated social networks |
| CN108492173A (en) * | 2018-03-23 | 2018-09-04 | 上海氪信信息技术有限公司 | A kind of anti-Fraud Prediction method of credit card based on dual-mode network figure mining algorithm |
| CN108564286A (en) * | 2018-04-19 | 2018-09-21 | 天合泽泰（厦门）征信服务有限公司 | A kind of artificial intelligence finance air control credit assessment method and system based on big data reference |
| CN109035003A (en) * | 2018-07-04 | 2018-12-18 | 北京玖富普惠信息技术有限公司 | Anti- fraud model modelling approach and anti-fraud monitoring method based on machine learning |
Cited By (58)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110533190B (en) * | 2019-07-18 | 2023-09-05 | 武汉烽火众智数字技术有限责任公司 | Data object analysis method and device based on machine learning |
| CN110533190A (en) * | 2019-07-18 | 2019-12-03 | 武汉烽火众智数字技术有限责任公司 | A kind of data object analysis method and device based on machine learning |
| CN110727770A (en) * | 2019-08-30 | 2020-01-24 | 深圳壹账通智能科技有限公司 | Intelligent request screening method and device, computer equipment and storage medium |
| CN110728301A (en) * | 2019-09-09 | 2020-01-24 | 北京镭文科技有限公司 | Credit scoring method, device, terminal and storage medium for individual user |
| CN110766541A (en) * | 2019-09-25 | 2020-02-07 | 平安科技（深圳）有限公司 | Loan risk assessment method, loan risk assessment device, loan risk assessment equipment and computer-readable storage medium |
| CN110727922A (en) * | 2019-10-11 | 2020-01-24 | 集奥聚合(北京)人工智能科技有限公司 | Anti-fraud decision model construction method based on multi-dimensional data flow |
| CN110727922B (en) * | 2019-10-11 | 2023-08-29 | 集奥聚合(北京)人工智能科技有限公司 | Anti-fraud decision model construction method based on multi-dimensional data flow |
| CN110956303A (en) * | 2019-10-12 | 2020-04-03 | 未鲲(上海)科技服务有限公司 | Information prediction method, device, terminal and readable storage medium |
| CN110796539A (en) * | 2019-10-31 | 2020-02-14 | 中国工商银行股份有限公司 | Credit investigation evaluation method and device |
| CN110912874A (en) * | 2019-11-07 | 2020-03-24 | 苏宁云计算有限公司 | Method and system for effectively identifying machine access behaviors |
| 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 |
| 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 |
| CN111127026A (en) * | 2019-12-13 | 2020-05-08 | 深圳中兴飞贷金融科技有限公司 | Method, device, storage medium and electronic equipment for determining user fraud behavior |
| CN111178722A (en) * | 2019-12-20 | 2020-05-19 | 上海数策软件股份有限公司 | Machine learning systems, methods, and media suitable for sales lead rating and distribution |
| CN111178722B (en) * | 2019-12-20 | 2023-05-02 | 上海数策软件股份有限公司 | Machine learning system, method and medium suitable for sales lead rating and distribution |
| CN111144938B (en) * | 2019-12-20 | 2023-06-20 | 上海数策软件股份有限公司 | Method and system for rating sales lead applicable to automobile industry |
| CN111144938A (en) * | 2019-12-20 | 2020-05-12 | 上海数策软件股份有限公司 | Sales lead rating method and system suitable for automobile industry |
| CN111062444B (en) * | 2019-12-21 | 2023-12-08 | 湖南大学 | Credit risk prediction method, credit risk prediction system, credit risk prediction terminal and storage medium |
| CN111062444A (en) * | 2019-12-21 | 2020-04-24 | 湖南大学 | Credit risk prediction method, system, terminal and storage medium |
| CN113222632A (en) * | 2020-02-04 | 2021-08-06 | 北京京东振世信息技术有限公司 | Object mining method and device |
| CN111353901A (en) * | 2020-02-24 | 2020-06-30 | 阳光保险集团股份有限公司 | Risk identification monitoring method and device and electronic equipment |
| CN111582646A (en) * | 2020-04-09 | 2020-08-25 | 上海淇毓信息科技有限公司 | User policy risk early warning method and device and electronic equipment |
| CN111652711A (en) * | 2020-06-09 | 2020-09-11 | 深圳前海微众银行股份有限公司 | Pre-credit analysis method, device, equipment and storage medium based on collection data |
| CN111652712B (en) * | 2020-06-09 | 2024-05-10 | 深圳前海微众银行股份有限公司 | Pre-loan analysis method, device, equipment and storage medium based on geographic information |
| CN111652712A (en) * | 2020-06-09 | 2020-09-11 | 深圳前海微众银行股份有限公司 | Pre-credit analysis method, device, equipment and storage medium based on geographic information |
| CN111709828A (en) * | 2020-06-12 | 2020-09-25 | 中国建设银行股份有限公司 | Resource processing method, device, equipment and system |
| CN111738819A (en) * | 2020-06-15 | 2020-10-02 | 中国建设银行股份有限公司 | Method, device and equipment for screening characterization data |
| CN111488338A (en) * | 2020-06-29 | 2020-08-04 | 上海冰鉴信息科技有限公司 | Model monitoring method and device applied to wind control decision flow |
| CN111861703B (en) * | 2020-07-10 | 2024-05-10 | 深圳无域科技技术有限公司 | Data-driven wind control strategy rule generation method and system and risk control method and system |
| CN111861703A (en) * | 2020-07-10 | 2020-10-30 | 深圳无域科技技术有限公司 | Data-driven wind control strategy rule generation method and system and risk control method and system |
| CN111815439A (en) * | 2020-07-23 | 2020-10-23 | 睿智合创(北京)科技有限公司 | Credit scoring system based on cloud platform |
| CN111815439B (en) * | 2020-07-23 | 2021-05-18 | 睿智合创(北京)科技有限公司 | Credit scoring system based on cloud platform |
| CN112508677A (en) * | 2020-11-06 | 2021-03-16 | 无锡艺界科技有限公司 | Financial system based on big data wind accuse |
| CN112862585A (en) * | 2020-12-02 | 2021-05-28 | 浙江惠瀜网络科技有限公司 | Personal loan type bad asset risk rating method based on LightGBM decision tree algorithm |
| CN112396517A (en) * | 2020-12-04 | 2021-02-23 | 深圳前海微众银行股份有限公司 | ETC card management method, device and system and computer storage medium |
| CN112561685A (en) * | 2020-12-15 | 2021-03-26 | 建信金融科技有限责任公司 | Client classification method and device |
| CN112561685B (en) * | 2020-12-15 | 2023-10-17 | 建信金融科技有限责任公司 | Customer classification method and device |
| CN112541706A (en) * | 2020-12-24 | 2021-03-23 | 四川享宇金信金融科技有限公司 | Big data wind control model construction method |
| CN112819034A (en) * | 2021-01-12 | 2021-05-18 | 平安科技（深圳）有限公司 | Data binning threshold calculation method and device, computer equipment and storage medium |
| CN112767136A (en) * | 2021-01-26 | 2021-05-07 | 天元大数据信用管理有限公司 | Credit anti-fraud identification method, credit anti-fraud identification device, credit anti-fraud identification equipment and credit anti-fraud identification medium based on big data |
| CN113011751A (en) * | 2021-03-19 | 2021-06-22 | 天道金科股份有限公司 | Small and medium-sized micro enterprise credit evaluation method based on big data |
| CN113011752A (en) * | 2021-03-19 | 2021-06-22 | 天道金科股份有限公司 | Enterprise credit evaluation index system based on big data |
| CN113011971B (en) * | 2021-03-31 | 2024-05-10 | 深圳前海微众银行股份有限公司 | Risk measure method, apparatus, system and computer storage medium |
| CN113011971A (en) * | 2021-03-31 | 2021-06-22 | 深圳前海微众银行股份有限公司 | Risk measurement method, device and system and computer storage medium |
| CN113177585B (en) * | 2021-04-23 | 2024-04-05 | 上海晓途网络科技有限公司 | User classification method, device, electronic equipment and storage medium |
| CN113177585A (en) * | 2021-04-23 | 2021-07-27 | 上海晓途网络科技有限公司 | User classification method and device, electronic equipment and storage medium |
| CN113408923A (en) * | 2021-06-29 | 2021-09-17 | 中国平安人寿保险股份有限公司 | Premium collection method and device, computer equipment and storage medium |
| CN113590310A (en) * | 2021-07-02 | 2021-11-02 | 上海淇玥信息技术有限公司 | Resource allocation method and device based on rule touch rate scoring and electronic equipment |
| CN113657481A (en) * | 2021-08-13 | 2021-11-16 | 上海晓途网络科技有限公司 | Model construction system and method |
| CN113610175A (en) * | 2021-08-16 | 2021-11-05 | 上海冰鉴信息科技有限公司 | Service strategy generation method and device and computer readable storage medium |
| CN114155080A (en) * | 2021-09-29 | 2022-03-08 | 东方微银科技股份有限公司 | Fraud identification method, equipment and storage medium |
| CN113902551A (en) * | 2021-10-20 | 2022-01-07 | 长安汽车金融有限公司 | Method and apparatus for vehicle loan application fraud identification |
| CN113902551B (en) * | 2021-10-20 | 2023-09-01 | 长安汽车金融有限公司 | Method and equipment for identifying fraud of vehicle credit application |
| CN114240215A (en) * | 2021-12-22 | 2022-03-25 | 中国建设银行股份有限公司 | User loss of contact grade acquisition method and device and storage medium |
| CN114926261A (en) * | 2022-04-26 | 2022-08-19 | 厦门大学 | Method and medium for predicting fraud probability of automobile financial user application |
| CN115034888A (en) * | 2022-06-16 | 2022-09-09 | 支付宝(杭州)信息技术有限公司 | Credit service providing method and device |
| CN115797044A (en) * | 2022-11-15 | 2023-03-14 | 东方微银科技股份有限公司 | Credit wind control early warning method and system based on cluster analysis |
| CN115797044B (en) * | 2022-11-15 | 2024-03-29 | 东方微银科技股份有限公司 | Credit wind control early warning method and system based on cluster analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110009479B (en) | 2021-02-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110009479A (en) | Credit assessment method and device, storage medium, computer equipment | |
| CN109035003A (en) | Anti- fraud model modelling approach and anti-fraud monitoring method based on machine learning | |
| Hamid et al. | Developing prediction model of loan risk in banks using data mining | |
| TWI248001B (en) | Methods and apparatus for automated underwriting of segmentable portfolio assets | |
| Eletter et al. | Neuro-based artificial intelligence model for loan decisions | |
| CN108564286A (en) | A kind of artificial intelligence finance air control credit assessment method and system based on big data reference | |
| CN109409677A (en) | Enterprise Credit Risk Evaluation method, apparatus, equipment and storage medium | |
| CN111080442A (en) | Credit scoring model construction method, device, equipment and storage medium | |
| JP2003535387A (en) | Rapid evaluation of asset portfolios such as financial products | |
| JP2003529138A (en) | Methods and systems for optimizing revenue and present value | |
| CN108711107A (en) | Intelligent financing services recommend method and its system | |
| JP2004500646A (en) | Competitive bidding revenue simulation method and system | |
| JP2004500641A (en) | Method and system for automatically estimating credit score evaluation value | |
| JP2004500642A (en) | Methods and systems for assessing cash flow recovery and risk | |
| JP2003526147A (en) | Cross-correlation tool to automatically calculate portfolio description statistics | |
| CN111985937A (en) | Method, system, storage medium and computer equipment for evaluating value information of transaction traders | |
| Tran et al. | Credit scoring model: A combination of genetic programming and deep learning | |
| CN108596758A (en) | A kind of credit rating method based on classification rule-based classification | |
| Eddy et al. | Credit scoring models: Techniques and issues | |
| TWM613536U (en) | Investment risk scoring system for fund commodities | |
| Cerny et al. | Data mining and neural networks from a commercial perspective | |
| CN111292188A (en) | Intelligent integrated return test system for pairing transaction | |
| Li et al. | Research on default risk of peer-to-peer online lending based on data mining algorithm | |
| CN109801162A (en) | A kind of credit rating method that social media data are merged with multi-standard cross-certification | |
| US20040225628A1 (en) | Data merging program, data merging method, and scoring system using data merging program |
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 | ||
| CB02 | Change of applicant information |
Address after: 100041 No.76, 5th floor, building 5, yard 30, Shixing street, Shijingshan District, Beijing Applicant after: Bairong Yunchuang Technology Co.,Ltd. Address before: 100041 No.76, 5th floor, building 5, yard 30, Shixing street, Shijingshan District, Beijing Applicant before: 100CREDIT FINANCE INFORMATION SERVICE Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |