CN106997472A - User characteristics sorting technique, user credit appraisal procedure and the device of user credit model - Google Patents
User characteristics sorting technique, user credit appraisal procedure and the device of user credit model Download PDFInfo
- Publication number
- CN106997472A CN106997472A CN201610049452.8A CN201610049452A CN106997472A CN 106997472 A CN106997472 A CN 106997472A CN 201610049452 A CN201610049452 A CN 201610049452A CN 106997472 A CN106997472 A CN 106997472A
- Authority
- CN
- China
- Prior art keywords
- user
- numerical value
- value
- user characteristics
- linear
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/245—Classification techniques relating to the decision surface
- G06F18/2453—Classification techniques relating to the decision surface non-linear, e.g. polynomial classifier
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Nonlinear Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
This application discloses the user characteristics sorting technique of user credit model, user credit appraisal procedure and device.The user credit appraisal procedure includes：Receive the user characteristic data of each user characteristics；The user characteristic data of each unspecified persons feature in the unspecified persons characteristic set is subjected to Nonlinear Processing, the first numerical value is determined according to the result of the Nonlinear Processing；The user characteristic data of each specified user characteristics in the specified user characteristics set is subjected to Linear Mapping, second value is determined according to the result of the Linear Mapping；By first numerical value and the second value, the output valve of the user credit model is determined；User credit is assessed according to the output valve.So as to solve in the prior art, whole characteristics is handled in the same way, the problem of causing to assess obtained user credit and larger actual deviation according to the result.
Description
Technical field
The application is related to technical field of data processing, more particularly to user credit model user characteristics classification side
Method, user credit appraisal procedure and device.
Background technology
With flourishing for internet, machine learning (Machine Learning, ML) technology is used as one kind
New technology, is increasingly valued by people.Behavior of the machine learning techniques commonly used to prediction user,
It can be realized in actual applications by various user models.For example, can by user's Credit Model,
Pair user characteristic data related to individual subscriber credit is handled, according to the prediction of result user's of processing
The behaviors such as personal consumption credit.
Prior art using user credit model when being handled characteristic, typically by the way that user is believed
Summation is weighted with each characteristic of model with the corresponding parameter of the model, so as to obtain data processing
Result, and the personal credit of user is estimated according to the result.However, problem of the prior art is
It is this to be handled whole characteristics in the same way, it will usually to cause according to the result
Assess obtained user credit and actual deviation is larger.
The content of the invention
The embodiment of the present application provides user characteristics sorting technique, the user credit appraisal procedure of user credit model
And device, for solve the user credit assessed in the prior art generally with actual deviation it is larger the problem of.
The embodiment of the present application provides a kind of user characteristics sorting technique, and this method includes：
Extract each user characteristics in user credit model；
Each user characteristics in each user characteristics is worked as active user's feature with described respectively
Preceding user characteristics is independent variable, and the output valve using the user credit model is determined described current as dependent variable
User characteristics and the mapping relations of the user credit model output valve；
Whether be specify mapping relations, if so, then by active user's feature point if judging the mapping relations
Class is to specify user characteristics；If it is not, being then unspecified persons feature by active user's tagsort.
It is preferred that, the specified mapping relations specifically include linear mapping relation；Then, reflected described in the judgement
Whether be specify mapping relations, if so, being then to specify user special by active user's tagsort if penetrating relation
Levy；If it is not, being then that unspecified persons feature is specifically included by active user's tagsort：
Whether be linear mapping relation, if so, then by active user's feature point if judging the mapping relations
Class is linear user feature；If it is not, being then non-linear customer feature by active user's tagsort.
The embodiment of the present application also provides a kind of user credit appraisal procedure of user credit model, by the user
Each user characteristics in Credit Model is divided into specified user characteristics set according to a kind of user characteristics sorting technique
With unspecified persons characteristic set, a unspecified persons are comprised at least in the unspecified persons characteristic set
Feature, the user credit appraisal procedure of the user credit model includes：
Receive the user characteristic data of each user characteristics；
The user characteristic data of each unspecified persons feature in the unspecified persons characteristic set is carried out non-
Linear process, the first numerical value is determined according to the result of the Nonlinear Processing, and first numerical value refers to first
Determine in span；
The user characteristic data of each specified user characteristics in the specified user characteristics set is linearly reflected
Penetrate, second value is determined according to the result of the Linear Mapping, the second value specifies value model second
In enclosing；
By first numerical value and the second value, the output valve of the user credit model is determined；
User credit is assessed according to the output valve.
It is preferred that, the user characteristics by each unspecified persons feature in the unspecified persons characteristic set
Data carry out Nonlinear Processing, determine that the first numerical value is specifically included according to the result of the Nonlinear Processing：
The user characteristic data of each unspecified persons feature in the unspecified persons characteristic set is entered into line
Property mapping, determine the first subnumber value；
The first subnumber value is subjected to Nonlinear Mapping, first is determined according to the result of the Nonlinear Mapping
Numerical value.
It is preferred that, the user credit model includes Linear processing module and Nonlinear processing module, the line
Property processing module be used for the user characteristic data of each specified user characteristics in the specified user characteristics set
Linear Mapping is carried out, the Nonlinear processing module is used for each non-finger in the unspecified persons characteristic set
The user characteristic data for determining user characteristics carries out Nonlinear Processing.
It is preferred that, the Nonlinear Mapping specifically includes excitation function conversion；Then,
It is described that the first subnumber value is subjected to Nonlinear Mapping, determined according to the result of the Nonlinear Mapping
First numerical value is specifically included：The first subnumber value is subjected to excitation function conversion, according to the excitation function
The result of conversion determines the first numerical value, and the excitation function is the nonlinear function of specified type.
It is preferred that, it is described by first numerical value and the second value, determine the user credit model
Output valve specifically include：
By the way that first numerical value and the second value are carried out into linear transformation, according to the linear transformation
As a result the output valve of the user credit model is determined.
It is preferred that, the specified user characteristics set specifically includes linear user characteristic set；It is described non-designated
User characteristics set specifically includes non-linear customer characteristic set；The specified user characteristics is specifically included linearly
User characteristics；The unspecified persons feature specifically includes non-linear customer feature.
The embodiment of the present application also provides a kind of user characteristics sorter, and the device includes：
Extraction unit, determining unit and taxon, wherein：
Extraction unit, for extracting each user characteristics in user credit model；
Determining unit, for regarding each user characteristics in each user characteristics as active user respectively
Feature, independent variable is characterized as with the active user, the output valve using the user credit model as dependent variable,
Determine the mapping relations of active user's feature and the user credit model output valve；
Taxon, for judging whether the mapping relations are to specify mapping relations, if so, then will be described
Active user's tagsort is to specify user characteristics；If it is not, then active user's tagsort is referred to be non-
Determine user characteristics.
It is preferred that, the taxon specifically includes the first taxon, for judging that the mapping relations are
No is linear mapping relation, if so, being then linear user feature by active user's tagsort；If it is not,
It is then non-linear customer feature by active user's tagsort.
The embodiment of the present application also provides a kind of user credit apparatus for evaluating of user credit model, by the user
Each user characteristics in Credit Model is divided into specified user characteristics set according to a kind of user characteristics sorter
With unspecified persons characteristic set, a unspecified persons are comprised at least in the unspecified persons characteristic set
Feature, the user credit apparatus for evaluating of the user credit model includes：
Receiving unit, the first numerical value determining unit, second value determining unit, output valve determining unit and comment
Estimate unit, wherein：
Receiving unit, the user characteristic data for receiving each user characteristics；
First numerical value determining unit, for by each unspecified persons feature in the unspecified persons characteristic set
User characteristic data carry out Nonlinear Processing, the first numerical value is determined according to the result of the Nonlinear Processing,
First numerical value is specified in span first；
Second value determining unit, for by the use of each specified user characteristics in the specified user characteristics set
Family characteristic carries out Linear Mapping, and second value, described second are determined according to the result of the Linear Mapping
Numerical value is specified in span second；
Output valve determining unit, for by first numerical value and the second value, determining the user
The output valve of Credit Model；
Assessment unit, for assessing user credit according to the output valve.
It is preferred that, the first numerical value determining unit determines the first subelement and the first numerical value including the first numerical value
The second subelement is determined, wherein：
First numerical value determines the first subelement, for by the unspecified persons characteristic set it is each it is non-refer to
The user characteristic data for determining user characteristics carries out Linear Mapping, determines the first subnumber value；
First numerical value determines the second subelement, for the first subnumber value to be carried out into Nonlinear Mapping,
First numerical value is determined according to the result of the Nonlinear Mapping.
It is preferred that, first numerical value determines that the second subelement determines the second submodule including the first numerical value, uses
In the first subnumber value is carried out into excitation function conversion, the result converted according to the excitation function determines the
One numerical value, the excitation function is the nonlinear function of specified type.
It is preferred that, the output valve determining unit includes output valve determination subelement, for by by described the
One numerical value and the second value carry out linear transformation, and the user is determined according to the result of the linear transformation
The output valve of Credit Model.
At least one above-mentioned technical scheme that the embodiment of the present application is used can reach following beneficial effect：
, will be each non-in unspecified persons characteristic set after the user characteristics in user credit model is classified
The user characteristic data of user characteristics is specified to carry out Nonlinear Processing, and by the specified user characteristics set
The user characteristic data of each specified user characteristics carries out Linear Mapping so that according to Nonlinear Processing in the model
Result and Linear Mapping result determine output valve, by output valve assess user credit when can obtain
To more preferable result.So as to solve in the prior art, whole characteristics is passed through into identical side
Formula is handled, the problem of causing to assess obtained user credit and larger actual deviation according to the result.
Brief description of the drawings
Accompanying drawing described herein is used for providing further understanding of the present application, constitutes the part of the application,
The schematic description and description of the application is used to explain the application, does not constitute the improper limit to the application
It is fixed.In the accompanying drawings：
Fig. 1 is a kind of tool of the user characteristics sorting technique for user credit model that the embodiment of the present application 1 is provided
Body implementation process schematic diagram；
Fig. 2 is a kind of tool of the user credit appraisal procedure for user credit model that the embodiment of the present application 2 is provided
Body implementation process schematic diagram；
Fig. 3 is to be connected inside a kind of user credit model for deep neural network that the embodiment of the present application 3 is provided
Schematic diagram；
Fig. 4 is a kind of user credit appraisal procedure for user credit model that the embodiment of the present application 3 is provided one
Plant the implementation process schematic diagram under concrete application scene；
Fig. 5 is a kind of tool of the user characteristics sorter for user credit model that the embodiment of the present application 4 is provided
Body structural representation；
Fig. 6 is a kind of tool of the user credit apparatus for evaluating for user credit model that the embodiment of the present application 5 is provided
Body structural representation.
Embodiment
In the embodiment of the present application, user credit model can include input layer, hidden layer and output layer.Its
In, input layer is used for the data for receiving user characteristics, and hidden layer is used to handle received data,
Output layer exports the result of the user credit model.In addition, any one use in the embodiment of the present application
Family Credit Model, the user credit model is used for the credit for assessing user.
For user credit model, in the prior art simply by each user characteristic data and user credit mould
The corresponding parameter (namely weight coefficient) of type is weighted summation, and user credit model is obtained so as to calculate
Output valve.This data by user characteristics carry out the mode that linear combination obtains output valve, due to linear group
The effect of contraction that expression formula itself is closed to data is inadequate, such as when running into the data of user characteristics of morbid state,
The data of the ill user characteristics can further be amplified, cause to be generally difficult to draw preferable result,
Namely predict that user credit and actual deviation are larger according to result.
For example, borrow or lend money the user credit model of credit for above-mentioned prediction user, the user credit model its
In the input value of a user characteristics length of service be 90, but the length of service is 90 simultaneously under normal circumstances
It is unreasonable, therefore the data are ill data.In this case, by the user characteristic data and weight system
Number is multiplied, by can further amplify the ill data after summation, so as to cause by the defeated of the model
When going out value assessment user credit, differ larger with actual.
In order to solve this problem, the present invention is handled each user characteristic data of user credit model,
By introducing nonlinear data processing method during processing so that final data processed result is more
Plus the personal credit of user can be reflected.
It is specifically real below in conjunction with the application to make the purpose, technical scheme and advantage of the application clearer
Apply example and technical scheme is clearly and completely described corresponding accompanying drawing.Obviously, it is described
Embodiment is only some embodiments of the present application, rather than whole embodiments.Based on the implementation in the application
Example, the every other implementation that those of ordinary skill in the art are obtained under the premise of creative work is not made
Example, belongs to the scope of the application protection.
Below in conjunction with accompanying drawing, the technical scheme that each embodiment of the application is provided is described in detail.
Embodiment 1
Embodiment 1 provides a kind of user characteristics sorting technique of user credit model, for by user credit
Each user characteristics in model is classified, so as to be laid the foundation for accurate evaluation user credit.This method
Idiographic flow schematic diagram is as shown in figure 1, comprise the steps：
Step S11：Extract each user characteristics in user credit model.
Herein, user credit model refer to the model be used for assess user personal credit, it is necessary to explanation
It is that selected user characteristics can also be different in actual applications for different user credit models.
In addition, user characteristics is the abstract result of each user property related to user credit, these users category
Property can include age, sex, income etc..It is generally same in practical application, when setting up user credit model,
First user characteristics is handled by Feature Engineering, including user characteristics is deleted, merged.Here
Described user characteristics can be the user characteristics not handled by Feature Engineering or pass through feature
User characteristics after project treatment.For example, user credit model is by the age of user, sex, income, letter
With history, this four user characteristicses are set up, then step S11 can extract this four user characteristicses；Can also
To extract identity speciality and credit history the two user characteristicses, wherein identity speciality be by the age, sex,
User characteristics after three user characteristicses merging of income.
The step is used to extract each user characteristics related to the user credit model set up.
Step S12：Respectively using each user characteristics in each user characteristics as active user's feature,
Independent variable is characterized as with the active user, the output valve using the user credit model as dependent variable, it is determined that
The mapping relations of active user's feature and the user credit model output valve.
During the step is each user characteristics for being extracted step S11, each user characteristics respectively as
Active user's feature, using the current user characteristics as independent variable, using the output valve of the user credit model as because
Variable, determines the mapping relations between the current user characteristics and the user credit model output valve.
In actual applications, generally by changing the value of the current user characteristics, record and to analyze this corresponding
User credit model output valve, then passes through the value and corresponding user credit model of the current user characteristics
Mapping relations between them are studied by output valve.Generally can be by the technique study of curve matching
Mapping relations between them, that is, appropriate curve type is selected to be fitted the value of user characteristics and its right
The user credit model output valve answered.For example, one group of data (xi, yi) (i=1,2 ... m), wherein xi
Represent i-th of value of active user's feature, yiRepresent xiThe output valve of corresponding user credit model, passes through
The appropriate curve type of selection (linear, conic section, index etc.) is to this group of data (xi, yi) (i=1,
2 ... m) are fitted, so that it is determined that the mapping relations between them.
Step S13：Whether be specify mapping relations, if so, then performing step S14 if judging the mapping relations；
If it is not, then performing step S15.
Step S14：It is to specify class user characteristics by active user's tagsort；
Step S15：It is non-designated class user characteristics by active user's tagsort.
Step S13 is into step S15, when it is determined that active user's feature is exported with corresponding user credit model
After mapping relations between value, by analyzing whether the mapping relations are to specify mapping relations, to the current use
Family feature is classified.
In actual applications, this can be selected to specify mapping relations according to actual needs.If for example, needed
Study in user credit model, into the user characteristics of Quadratic Map relation between output valve, then can select
This specifies mapping relations to be Quadratic Map relation.By analyzing the current user characteristics and corresponding user credit
Whether it is Quadratic Map relation between model output valve, the current user characteristics is classified.Herein,
It is secondary, such as y that Quadratic Map relation, which refers to highest number of times in mapping function,i=axi 2+ b, wherein, a, b are
Preset constant and a are not 0, xiFor i-th of value of active user's feature, yiRepresent xiCorresponding user's letter
With the output valve of model.
Particularly, this specifies mapping relations to be usually linear mapping relation in actual applications, that is to say, that sentence
Whether the mapping relations of breaking are linear mapping relation, if so, being then line by active user's tagsort
Property class user characteristics；If it is not, being non-linear class user characteristics by active user's tagsort.Herein,
Linear mapping relation refers to highest number of times in mapping function, such as yi=pxi+ q, wherein, p, q are
Preset constant and q are not 0, xiFor i-th of value of active user's feature, yiRepresent xiCorresponding user's letter
With the output valve of model.
This method provided using embodiment 1, by analyzing each user characteristics in user credit model,
Mapping relations between the user credit model output valve, and judge whether the mapping relations are specified respectively
Mapping relations, so that each user characteristics be classified.So it is to pass through user's Credit Model accurate evaluation
User credit provides the foundation.
It should be noted that the executive agent that embodiment 1 provides each step of method may each be same and set
It is standby, or, each step of this method can also be used as executive agent by distinct device.Such as, step 11
Executive agent with step 12 can be equipment 1；Again such as, the executive agent of step 11 can be equipment
1, the executive agent of step 12 sum can be equipment 2；Etc..
Embodiment 2
Embodiment 2 provides the user in a kind of user credit appraisal procedure of user credit model, this method
Each user characteristics in Credit Model is to specify user characteristics set and non-according to the classification of embodiment 1
Specify and a unspecified persons feature is comprised at least in user characteristics set, the unspecified persons characteristic set.
The idiographic flow schematic diagram of this method is as shown in Fig. 2 comprise the steps：
Step S21：Receive the user characteristic data of each user characteristics.
Herein, each user characteristics refers to each user characteristics corresponding with user credit model.For example, with
Family Credit Model is built by three user characteristicses, respectively age of user, marital status and wage income, then
Each user characteristics is specially age of user, marital status and wage income these three user characteristicses.
In actual applications, the user characteristic data received, can be receive extracted from database it is each
The data of the data of individual user characteristics or each user characteristics obtained immediately by network, here
Restriction is not made to the source of user characteristic data, while also not done to the mode for receiving user characteristic data
Go out to limit.
Step S22：By the user characteristics number of each unspecified persons feature in the unspecified persons characteristic set
According to Nonlinear Processing is carried out, the first numerical value is determined according to the result of the Nonlinear Processing.
Herein, first numerical value is specified in span first, wherein first specifies span can
Determined with according to actual needs.For example, generally for causing the output result of user credit model to fall 0
To in the range of 100, the first numerical value that can at this time limit the model falls in the interval of (0,100),
That is first span is specified to be interval (0,100)；Under other application scenarios, generally use
In the model for assessing user credit, it is thus necessary to determine that be user's promise breaking (or not breaking a contract) probability, it is therefore logical
Often need the output valve of user credit model to fall in the range of 0 to 1, can also at this time limit the first finger
It is interval (0,1) to determine span；Certainly, the first numerical value can also be using 5 points of systems according to actual needs
Give a mark, either ten point system or other modes are given a mark, be at this time required to limit the first numerical value
It is scheduled in corresponding particular range, this corresponding particular range is exactly the first specified span.
A unspecified persons feature is comprised at least in unspecified persons characteristic set, by unspecified persons feature
The user characteristic data of the unspecified persons feature of each in set carries out Nonlinear Processing, according to the result of processing
Determine the first numerical value.In actual applications, have much to the mode that user characteristic data carries out Nonlinear Processing,
For example it is squared and either summation square or even multiply.
It should be further stated that, in actual applications, the user credit model set up according to actual needs
Multilayer hidden layer can generally be included, (non-thread can be included to carry out repeatedly processing to user characteristic data
Property processing and Linear Mapping etc.), so as to improve the accuracy of the user credit model.In the user credit mould
In type, the input of i-th layer of hidden layer is the output of the i-th -1 layer hidden layer, and i-th layer of hidden layer be output as the
The input of i+1 layers of hidden layer, i is the positive integer more than or equal to 2, and the input of first layer hidden layer in addition is
The output of mode input layer.Therefore, when user credit model has multilayer hidden layer, step S22 is by institute
The user characteristic data for stating each unspecified persons feature in unspecified persons characteristic set carries out Nonlinear Processing,
Determine that the first numerical value can be optimized for according to the result of the Nonlinear Processing：By the unspecified persons feature
The user characteristic data of each unspecified persons feature carries out Nonlinear Processing at least one times in set, according to each institute
The result for stating Nonlinear Processing determines the first numerical value.
In addition, in a kind of characteristic set by unspecified persons each unspecified persons feature user characteristic data
The preferred scheme for carrying out Nonlinear Processing is step S221.
Step S221：By the user characteristics number of each unspecified persons feature in the unspecified persons characteristic set
According to Linear Mapping is carried out, the first subnumber value is determined；The first subnumber value is subjected to Nonlinear Mapping, according to
The result of the Nonlinear Mapping determines the first numerical value.
In step S221, first by the user of each unspecified persons feature in unspecified persons characteristic set
Characteristic carries out Linear Mapping, and the result of Linear Mapping then is carried out into Nonlinear Mapping, so that it is determined that the
One numerical value, this method causes the number of times of Nonlinear Mapping to reduce, and computational methods are easier.Actually should
In, Linear Mapping (linear map, LP) is typically referred to from a vector space V to another vector
The mapping of space W, Linear Mapping generally includes add operation and quantity multiplying.Herein, linearly
Mapping can include summation operation and/or sum operation with coefficient etc., that is to say, that by unspecified persons characteristic set
In each unspecified persons characteristic carry out summation operation and/or sum operation with coefficient, the power in weighted sum
Again be typically user credit model in model parameter.
Herein, Nonlinear Mapping refers to other mapping modes outside Linear Mapping, can generally include secondary
Functional transformation, exponential function conversion, hyperbolic tangent function conversion, sigmoid functional transformations and/or Softplus
Functional transformation etc..Particularly in user credit model, the mode of Nonlinear Mapping is generally used for by swashing
Function living enters line translation, that is to say, that the first subnumber value is carried out into Nonlinear Mapping, according to the non-thread
Property mapping result determine that the first numerical value is specifically included：The first subnumber value is subjected to excitation function conversion,
The result converted according to the excitation function determines the first numerical value, and the excitation function is the non-thread of specified type
Property function.In actual applications, the activation primitive generally used includes：Hyperbolic tangent function, sigmoid
Function and/or Softplus functions etc..For example, in actual applications when first specifies in span in (0,1)
Interval when, generally can by activation primitive selection be sigmoid functions, that is to say, that can by first son
Numerical value carries out sigmoid functional transformations, and the first numerical value is determined according to the result of the sigmoid functional transformations.
By the way that the first subnumber value is carried out into sigmoid functional transformations, because the codomain of sigmoid functions is (0,1),
Therefore the output valve after the Nonlinear Mapping can be limited in the range of (0,1).
Generally can also be when user characteristics to be classified, the specified mapping relations of selection are linear mapping relation,
Then corresponding specified user characteristics collection is combined into linear user characteristic set, and unspecified persons characteristic set is non-thread
Property user characteristics set, it is linear user feature to specify user characteristics, and unspecified persons are characterized as non-linear use
Family feature.So, in actual process, processing procedure can be caused more to facilitate.
Step S23：The user characteristic data of each specified user characteristics in the specified user characteristics set is entered
Row Linear Mapping, second value is determined according to the result of the Linear Mapping.
Herein, the second value is specified in span second, wherein second specifies span
Determine according to actual needs.Other second specifies span with first span can also be specified identical,
With first span can also be specified different.In actual applications, in order that obtaining the defeated of user's Credit Model
Go out value more rationally simultaneously for simplified model, usual second specifies span and first to specify span
It is identical.
Explanation is needed also exist for, in actual applications, because user credit model can be implicit including multilayer
Layer, repeatedly to be handled user characteristic data, so as to improve the accuracy of the user credit model.
Wherein, the input and output of each hidden layer are similar with step S22, just no longer illustrate here.When user's letter
When having multilayer hidden layer with model, step S23 is by each specified user characteristics in the specified user characteristics set
User characteristic data carry out Linear Mapping, determine that second value can be with excellent according to the result of the Linear Mapping
Turn to：The user characteristic data of each specified user characteristics in the specified user characteristics set is carried out at least one
Sublinear function, second value is determined according to the result of each Linear Mapping.
In addition, the Linear Mapping in the step is identical with the Linear Mapping being previously mentioned in step S22, it can wrap
Include summation operation and/or sum operation with coefficient etc..By the use of each specified user characteristics in specified user characteristics set
Family characteristic carries out Linear Mapping, and second value is determined according to the result of the Linear Mapping.
Step S24：By first numerical value and the second value, the user credit model is determined
Output valve.
By the first numerical value and second value determined by above-mentioned steps S22 and step S23, the user is determined
The output valve of Credit Model.In actual applications, can be according to the side that the first numerical value is multiplied with second value
Formula determines that the output valve or other modes of the model determine output valve.It is a kind of preferably,
By the way that first numerical value and the second value are carried out into linear transformation, according to the result of the linear transformation
Determine the output valve of the family Credit Model.Described linear transformation herein can also include summation operation and
/ or sum operation with coefficient etc., that is to say, that the first numerical value and second value are subjected to summation operation and/or weighting
Summation operation, so that it is determined that the output valve of family Credit Model.
Step S25：User credit is assessed according to the output valve.
The user credit model exports the defeated of the model by handling user characteristic data from output layer
Go out value, user credit can be assessed by the output valve.Generally, in order that obtaining the output of user's Credit Model
Value more can easily explain user credit, and the output valve and user credit that can cause user credit model are in
Monotonic relationshi (monotonic increase or monotone decreasing), the output valve of such as model is bigger, represents user credit
It is higher, or the output valve of the model is bigger, represents that user credit is lower.
This method provided using embodiment 2, will be non-after the user characteristics classification in user credit model
The user characteristic data of each unspecified persons feature in user characteristics set is specified to carry out Nonlinear Processing, and will
The user characteristic data of each specified user characteristics carries out Linear Mapping in the specified user characteristics set so that
The output valve determined in the model according to the result of the result of Nonlinear Processing and Linear Mapping, is assessing user
More preferable result can be obtained during credit.So as to solve in the prior art, by whole characteristics
Handled in the same way, cause to assess obtained user credit according to the result with reality partially
The problem of difference is larger.
Embodiment 3
Embodiment 1 and embodiment 2 to be classified by user's Credit Model to user characteristics, then to
Family characteristic is handled, so as to assess user credit according to the output valve after processing.In order to make it easy to understand,
The embodiment of the present application is additionally provided under a kind of practical application scene, and the user credit model is embodied
User credit appraisal procedure, first can also explain to the user credit model here.
The user credit model includes Linear processing module and Nonlinear processing module, the linear process mould
Block is used to carry out the user characteristic data of each specified user characteristics in the specified user characteristics set linearly
Mapping, the Nonlinear processing module is used for each unspecified persons in the unspecified persons characteristic set are special
The user characteristic data levied carries out Nonlinear Processing.
In actual applications, in order to consider the accuracy of user credit model, the inside of user credit model is also
The connected mode of deep neural network can be used, as shown in figure 3, so whole user credit model can be with
Regard a big deep neural network as.The deep neural network functionally can be by inputting from each several part
Layer, hidden layer, output layer, continue to be subdivided into input layer, Internet, interpretation layer, output layer.
Wherein, output layer：Output valve for exporting the user credit model.In actual applications, according to
The output valve assesses user credit, for example, can judge whether user has overdue 30 days according to the output valve.
Interpretation layer：Interpretation layer can include how each point of interpretation layer, and each point interpretation layer is solved from different grain size
Release the output valve of output layer.For example, a kind of user credit solution to model in practical application releases layer including two
Divide interpretation layer, first point of interpretation layer is close to the output layer of the model, and this point of interpretation layer can be from big granularity
(for example：Identity speciality, contractual capacity, credit history, relationship among persons, Behavior preference etc.) to output layer
Output valve explain, second point of interpretation layer can be from small grain size (for example：Back ground Information, education
Degree, occupational information, marriage information etc.) output valve of output layer is explained, wherein, identity speciality
Including Back ground Information, education degree, occupational information, marriage information etc..
Internet：Using locally-attached network structure, between Linear processing module and Nonlinear processing module
It is disconnected from each other, the interference of other uncorrelated user characteristicses on the one hand so can be reduced, on the other hand can also
Greatly reduce the parameter of the model, reduce the complexity of the model, may also speed up the training speed of the model.
In actual applications, Linear processing module and Nonlinear processing module can also continue to be subdivided into multiple submodules
Block, can also increase Dropout layers, in the training process of user credit model in each submodule
Random allows a part of neuron not work, so that the interdependency between reducing user characteristics so that learn
The user credit model practised has more preferable performance on test set.
Input layer：User characteristic data from input layer can be input to user credit model, in practical application
In user characteristic data can also be handled by Feature Engineering, then from input layer be input to user letter
Use model.
In the training process of user credit model, it can be trained using stochastic gradient descent method；Due to
Generally be present huge disequilibrium in training set sample, can be replicated the quantity of bad sample so that good
The quantity of sample and bad sample is basically identical；Furthermore it is also possible to add height behind the input layer of the model
This noise floor, to be filtered to ill data, so as to reduce user credit model over-fitting, further
Increase the accuracy of the model.
It should be further stated that, user credit model usually requires to comment from multiple dimensions in practical application
Estimate the personal credit of user, and each dimension needs multiple user characteristicses and is described.It is general to count,
In the design process of user credit model, each dimension can be designed as a standalone module.Therefore,
The user credit model includes M standalone module, at least one independence in each standalone module
Module can be entered to the user characteristic data of the unspecified persons feature in the unspecified persons characteristic set
Row Nonlinear Processing, the M is the positive integer more than or equal to 2.
For example, for a kind of user credit model, the user credit model can be from identity speciality, energy of honouring an agreement
The dimensions such as power, credit history, relationship among persons and Behavior preference assess the personal credit of user.Wherein, also
Marital status, occupational information, age, length of service etc. can be chosen from identity speciality this dimension multiple
User characteristics.When setting up user credit model, the model can include a standalone module, the independent mould
Block assesses the personal credit of user from identity speciality this dimension, and the user characteristics for describing the standalone module can be with
Including marital status, occupational information, age, length of service, if wherein the length of service is unspecified persons
Feature, then the standalone module is to unspecified persons feature progress Nonlinear Processing.
In practical application, user credit is estimated by above-mentioned user credit model, as shown in figure 4,
The method of the assessment user credit specifically includes following steps：
Step S31：Each user characteristics in user credit model is categorized as linear user feature and non-linear
User characteristics, linear user is characterized as that the input value and the output valve of the user credit model of the user characteristics are in
Linear relationship, non-linear customer is characterized as the input value of the user characteristics and the output valve of the user credit model
In non-linear relation.
Step S32：Receive the user characteristic data of each user characteristics.
Step S33：The user characteristic data of each non-linear customer feature is weighted summation and obtains the first son
Numerical value, weight is the parameter preset of the user credit model.
Step S34：First subnumber value progress sigmoid functional transformations are obtained into the first numerical value, the first numerical value
In the interval of (0,1).
Step S35：The user characteristic data of each linear user feature is weighted summation and obtains second value,
Second value is in the interval of (0,1), and weight is the parameter preset of the user credit model.
Step S36：First numerical value and second value are weighted summation and obtain the defeated of the user credit model
Go out value, weight is the parameter preset of the user credit model.
Step S37：User credit is assessed according to the output valve of the user credit model.
In this embodiment, by user characteristics according to the mapping relations with output valve be divided into linear user feature and
Non-linear customer feature.The characteristic of non-linear customer feature therein is weighted summation, and will be added
The result of power summation carries out sigmoid functional transformations and obtains the first numerical value；By the characteristic of linear user feature
Second value is obtained according to summation is directly weighted；The first numerical value and second value are finally weighted summation
The output valve of the model is obtained, and user credit is assessed according to output valve.This method for assessing user credit,
Overcome in the prior art, assessed simply by the characteristic of user characteristics is directly weighted into summation
During user credit, the problem of assessment result and actual deviation are larger.
Embodiment 4
Based on inventive concept same as Example 1, embodiments herein 4 provides a kind of user credit
The user characteristics sorter of model.As shown in figure 5, the device 40 includes：
Extraction unit 401, determining unit 402 and taxon 403, wherein：
Extraction unit 401, for extracting each user characteristics in user credit model；
Determining unit 402, for respectively using each user characteristics in each user characteristics as current
User characteristics, independent variable is characterized as with the active user, using the output valve of the user credit model as because
Variable, determines the mapping relations of active user's feature and the user credit model output valve；
Taxon 403, for judging whether the mapping relations are to specify mapping relations, if so, then will
Active user's tagsort is to specify user characteristics；If it is not, being by active user's tagsort then
Unspecified persons feature.
Device provided using embodiment 4, by determining unit to each user in user credit model
Feature is analyzed, and determines the mapping relations between each user characteristics user credit model output valve, then lead to
Whether be specify mapping relations, so that each user characteristics be divided if crossing the classification unit judges mapping relations
Class.So it is to be provided the foundation by user's Credit Model accurate evaluation user credit.
In actual applications, in order to simplify between problem under many application scenarios, this specifies mapping relations to lead to
Often it is linear mapping relation, the taxon specifically includes the first taxon, for judging the mapping
Whether relation is linear mapping relation, if so, being then linear user feature by active user's tagsort；
If it is not, being then non-linear customer feature by active user's tagsort.
Embodiment 5
Based on inventive concept same as Example 2, embodiment 5 provides a kind of use of user credit model
Family credit evaluation device, each user characteristics in the user credit model is divided into according to the device of embodiment 4
Specify and at least wrapped in user characteristics set and unspecified persons characteristic set, the unspecified persons characteristic set
Containing a unspecified persons feature.As shown in fig. 6, the device 50 includes：
Receiving unit 501, the first numerical value determining unit 502, second value determining unit 503, output valve are true
Order member 504 and assessment unit 505, wherein：
Receiving unit 501, the user characteristic data for receiving each user characteristics；
First numerical value determining unit 502, for by each unspecified persons in the unspecified persons characteristic set
The user characteristic data of feature carries out Nonlinear Processing, and the first number is determined according to the result of the Nonlinear Processing
Value, first numerical value is specified in span first；
Second value determining unit 503, for by each specified user characteristics in the specified user characteristics set
User characteristic data carry out Linear Mapping, second value is determined according to the result of the Linear Mapping, it is described
Second value is specified in span second；
Output valve determining unit 504, for by first numerical value and the second value, it is determined that described
The output valve of user credit model；
Assessment unit 505, for assessing user credit according to the output valve.
The device provided using embodiment 5, after the user characteristics classification in user credit model, is passed through
First numerical value determining unit is by the user characteristics number of each unspecified persons feature in unspecified persons characteristic set
According to progress Nonlinear Processing so that the output valve of the model can obtain more preferable when assessing user credit
Result.Predict that the behavior of user is inclined generally with reality according to result in the prior art so as to solve
The problem of difference is larger.
The first numerical value determining unit 502 determines the first subelement 5021 and the first number including the first numerical value
Value determines the second subelement 5022, wherein：
First numerical value determines the first subelement 5021, for will in the unspecified persons characteristic set it is each
The user characteristic data of unspecified persons feature carries out Linear Mapping, determines the first subnumber value；
First numerical value determines the second subelement 5022, for the first subnumber value to be carried out into non-linear reflect
Penetrate, the first numerical value is determined according to the result of the Nonlinear Mapping.
Determine that user characteristic data is carried out Linear Mapping by the first subelement 5021 by the first numerical value, then
Determine that the result of Linear Mapping is carried out Nonlinear Mapping by the second subelement by the first numerical value, due to non-linear
The number of times of mapping is reduced so that the computational methods are easier.
First numerical value determines that the second subelement 5021 determines the second submodule 50211 including the first numerical value,
For the first subnumber value to be carried out into excitation function conversion, the result converted according to the excitation function is determined
First numerical value, the excitation function is the nonlinear function of specified type.In actual applications, generally use
Activation primitive include：Hyperbolic tangent function, sigmoid functions and/or Softplus functions etc..For example,
, generally can be by activation primitive in actual applications when first specifies in span in the interval of (0,1)
Select as sigmoid functions, that is to say, that the first subnumber value can be subjected to sigmoid functional transformations, root
The first numerical value is determined according to the result of the sigmoid functional transformations.By the way that the first subnumber value is carried out into sigmoid
Functional transformation, because the codomain of sigmoid functions is (0,1), therefore can be by after the Nonlinear Mapping
Output valve is limited in the range of (0,1).
The output valve determining unit 504 includes output valve determination subelement 5041, for by by described the
One numerical value and the second value carry out linear transformation, and the user is determined according to the result of the linear transformation
The output valve of Credit Model.By the way that the first numerical value and second value are carried out into linear transformation, determine that the user believes
With the output valve of model, can cause the output valve of the user credit model has monotonicity, explanatory enhancing.
It should be understood by those skilled in the art that, embodiments herein can be provided as method, system or meter
Calculation machine program product.Therefore, the application can be using complete hardware embodiment, complete software embodiment or knot
The form of embodiment in terms of conjunction software and hardware.Wherein wrapped one or more moreover, the application can be used
Containing computer usable program code computer-usable storage medium (include but is not limited to magnetic disk storage,
CD-ROM, optical memory etc.) on the form of computer program product implemented.
The application is produced with reference to according to the method, equipment (system) and computer program of the embodiment of the present application
The flow chart and/or block diagram of product is described.It should be understood that can by computer program instructions implementation process figure and
/ or each flow and/or square frame in block diagram and the flow in flow chart and/or block diagram and/
Or the combination of square frame.These computer program instructions can be provided to all-purpose computer, special-purpose computer, insertion
Formula processor or the processor of other programmable data processing devices are to produce a machine so that pass through and calculate
The instruction of the computing device of machine or other programmable data processing devices is produced for realizing in flow chart one
The device for the function of being specified in individual flow or multiple flows and/or one square frame of block diagram or multiple square frames.
These computer program instructions, which may be alternatively stored in, can guide computer or the processing of other programmable datas to set
In the standby computer-readable memory worked in a specific way so that be stored in the computer-readable memory
Instruction produce include the manufacture of command device, the command device realization in one flow or multiple of flow chart
The function of being specified in one square frame of flow and/or block diagram or multiple square frames.
These computer program instructions can be also loaded into computer or other programmable data processing devices, made
Obtain and perform series of operation steps on computer or other programmable devices to produce computer implemented place
Reason, so that the instruction performed on computer or other programmable devices is provided for realizing in flow chart one
The step of function of being specified in flow or multiple flows and/or one square frame of block diagram or multiple square frames.
In a typical configuration, computing device includes one or more processors (CPU), input/defeated
Outgoing interface, network interface and internal memory.
Internal memory potentially includes the volatile memory in computer-readable medium, random access memory
And/or the form, such as read-only storage (ROM) or flash memory (flash RAM) such as Nonvolatile memory (RAM).
Internal memory is the example of computer-readable medium.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by appointing
What method or technique realizes that information is stored.Information can be computer-readable instruction, data structure, program
Module or other data.The example of the storage medium of computer includes, but are not limited to phase transition internal memory
(PRAM), static RAM (SRAM), dynamic random access memory (DRAM), its
Random access memory (RAM), read-only storage (ROM), the electrically erasable of his type are read-only
Memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc read-only storage
(CD-ROM), digital versatile disc (DVD) or other optical storages, magnetic cassette tape, tape magnetic
Disk storage or other magnetic storage apparatus or any other non-transmission medium, can be calculated available for storage
The information that equipment is accessed.Defined according to herein, computer-readable medium does not include temporary computer-readable matchmaker
The data-signal and carrier wave of body (transitory media), such as modulation.
It should also be noted that, term " comprising ", "comprising" or its any other variant be intended to it is non-
It is exclusive to include, so that process, method, commodity or equipment including a series of key elements are not only wrapped
Include those key elements, but also other key elements including being not expressly set out, or also include for this process,
Method, commodity or the intrinsic key element of equipment.In the absence of more restrictions, by sentence " including
One ... " limit key element, it is not excluded that in the process including key element, method, commodity or equipment
Also there is other identical element.
It will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer journey
Sequence product.Therefore, the application can using complete hardware embodiment, complete software embodiment or combine software and
The form of the embodiment of hardware aspect.Moreover, the application can be used wherein includes calculating one or more
Machine usable program code computer-usable storage medium (include but is not limited to magnetic disk storage, CD-ROM,
Optical memory etc.) on the form of computer program product implemented.
Embodiments herein is these are only, the application is not limited to.For people in the art
For member, the application can have various modifications and variations.It is all to be made within spirit herein and principle
Any modification, equivalent substitution and improvements etc., should be included within the scope of claims hereof.
Claims (14)
1. a kind of user characteristics sorting technique of user credit model, it is characterised in that including：
Extract each user characteristics in user credit model；
Each user characteristics in each user characteristics is worked as active user's feature with described respectively
Preceding user characteristics is independent variable, and the output valve using the user credit model is determined described current as dependent variable
User characteristics and the mapping relations of the user credit model output valve；
Whether be specify mapping relations, if so, then by active user's feature point if judging the mapping relations
Class is to specify user characteristics；If it is not, being then unspecified persons feature by active user's tagsort.
2. method as claimed in claim 1, it is characterised in that the specified mapping relations specifically include line
Property mapping relations；Then, it is described to judge whether the mapping relations are to specify mapping relations, if so, then by institute
It is to specify user characteristics to state active user's tagsort；If it is not, being then non-by active user's tagsort
Specified user characteristics is specifically included：
Whether be linear mapping relation, if so, then by active user's feature point if judging the mapping relations
Class is linear user feature；If it is not, being then non-linear customer feature by active user's tagsort.
3. a kind of user credit appraisal procedure of user credit model, it is characterised in that believe the user
Specified user spy is divided into according to any one methods described of claim 1 and 2 with each user characteristics in model
Collection is closed and unspecified persons characteristic set, and a non-finger is comprised at least in the unspecified persons characteristic set
Determine user characteristics, including：
Receive the user characteristic data of each user characteristics；
The user characteristic data of each unspecified persons feature in the unspecified persons characteristic set is carried out non-
Linear process, the first numerical value is determined according to the result of the Nonlinear Processing, and first numerical value refers to first
Determine in span；
The user characteristic data of each specified user characteristics in the specified user characteristics set is linearly reflected
Penetrate, second value is determined according to the result of the Linear Mapping, the second value specifies value model second
In enclosing；
By first numerical value and the second value, the output valve of the user credit model is determined；
User credit is assessed according to the output valve.
4. method as claimed in claim 3, it is characterised in that described by the unspecified persons feature set
The user characteristic data of each unspecified persons feature carries out Nonlinear Processing in conjunction, according to the Nonlinear Processing
Result determine that the first numerical value is specifically included：
The user characteristic data of each unspecified persons feature in the unspecified persons characteristic set is entered into line
Property mapping, determine the first subnumber value；
The first subnumber value is subjected to Nonlinear Mapping, first is determined according to the result of the Nonlinear Mapping
Numerical value.
5. such as any one methods described of claim 3 and 4, it is characterised in that the user credit mould
Type includes Linear processing module and Nonlinear processing module, and the Linear processing module is used for the specified use
The user characteristic data of each specified user characteristics carries out Linear Mapping, the Nonlinear Processing in the characteristic set of family
Module is used to enter the user characteristic data of each unspecified persons feature in the unspecified persons characteristic set
Row Nonlinear Processing.
6. method as claimed in claim 4, it is characterised in that the Nonlinear Mapping specifically includes excitation
Functional transformation；Then,
It is described that the first subnumber value is subjected to Nonlinear Mapping, determined according to the result of the Nonlinear Mapping
First numerical value is specifically included：The first subnumber value is subjected to excitation function conversion, according to the excitation function
The result of conversion determines the first numerical value, and the excitation function is the nonlinear function of specified type.
7. method as claimed in claim 3, it is characterised in that described by first numerical value and described
Second value, determines that the output valve of the user credit model is specifically included：
By the way that first numerical value and the second value are carried out into linear transformation, according to the linear transformation
As a result the output valve of the user credit model is determined.
8. method as claimed in claim 3, it is characterised in that the specified user characteristics set is specifically wrapped
Include linear user characteristic set；The unspecified persons characteristic set specifically includes non-linear customer characteristic set；
The specified user characteristics specifically includes linear user feature；The unspecified persons feature specifically includes non-thread
Property user characteristics.
9. a kind of user characteristics sorter of user credit model, it is characterised in that including：
Extraction unit, determining unit and taxon, wherein：
Extraction unit, for extracting each user characteristics in user credit model；
Determining unit, for regarding each user characteristics in each user characteristics as active user respectively
Feature, independent variable is characterized as with the active user, the output valve using the user credit model as dependent variable,
Determine the mapping relations of active user's feature and the user credit model output valve；
Taxon, for judging whether the mapping relations are to specify mapping relations, if so, then will be described
Active user's tagsort is to specify user characteristics；If it is not, then active user's tagsort is referred to be non-
Determine user characteristics.
10. device as claimed in claim 9, it is characterised in that the taxon specifically includes first point
Class unit, for judging whether the mapping relations are linear mapping relation, if so, then by the current use
Family tagsort is linear user feature；If it is not, being then non-linear customer by active user's tagsort
Feature.
11. a kind of user credit apparatus for evaluating of user credit model, it is characterised in that believe the user
It is divided into specified user characteristics set and non-designated with each user characteristics device according to claim 7 in model
A unspecified persons feature, bag are comprised at least in user characteristics set, the unspecified persons characteristic set
Include：
Receiving unit, the first numerical value determining unit, second value determining unit, output valve determining unit and comment
Estimate unit, wherein：
Receiving unit, the user characteristic data for receiving each user characteristics；
First numerical value determining unit, for by each unspecified persons feature in the unspecified persons characteristic set
User characteristic data carry out Nonlinear Processing, the first numerical value is determined according to the result of the Nonlinear Processing,
First numerical value is specified in span first；
Second value determining unit, for by the use of each specified user characteristics in the specified user characteristics set
Family characteristic carries out Linear Mapping, and second value, described second are determined according to the result of the Linear Mapping
Numerical value is specified in span second；
Output valve determining unit, for by first numerical value and the second value, determining the user
The output valve of Credit Model；
Assessment unit, for assessing user credit according to the output valve.
12. device as claimed in claim 11, it is characterised in that the first numerical value determining unit includes
First numerical value determines that the first subelement and the first numerical value determine the second subelement, wherein：
First numerical value determines the first subelement, for by the unspecified persons characteristic set it is each it is non-refer to
The user characteristic data for determining user characteristics carries out Linear Mapping, determines the first subnumber value；
First numerical value determines the second subelement, for the first subnumber value to be carried out into Nonlinear Mapping,
First numerical value is determined according to the result of the Nonlinear Mapping.
13. device as claimed in claim 12, it is characterised in that first numerical value determines that the second son is single
Member determines the second submodule including the first numerical value, for the first subnumber value to be carried out into excitation function conversion,
The result converted according to the excitation function determines the first numerical value, and the excitation function is the non-thread of specified type
Property function.
14. device as claimed in claim 11, it is characterised in that the output valve determining unit includes defeated
Go out to be worth determination subelement, for by the way that first numerical value and the second value are carried out into linear transformation, root
The output valve of the user credit model is determined according to the result of the linear transformation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610049452.8A CN106997472B (en) | 2016-01-25 | 2016-01-25 | User feature classification method of user credit model, user credit evaluation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610049452.8A CN106997472B (en) | 2016-01-25 | 2016-01-25 | User feature classification method of user credit model, user credit evaluation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106997472A true CN106997472A (en) | 2017-08-01 |
| CN106997472B CN106997472B (en) | 2020-06-02 |
Family
ID=59428610
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610049452.8A Active CN106997472B (en) | 2016-01-25 | 2016-01-25 | User feature classification method of user credit model, user credit evaluation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106997472B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108256995A (en) * | 2018-01-09 | 2018-07-06 | 秒钱（北京）信息服务有限公司 | A kind of refund wish Forecasting Methodology and system based on contiguous function |
| CN112507204A (en) * | 2019-09-16 | 2021-03-16 | 北京智联云海科技有限公司 | Method for automatically constructing user portrait by utilizing data analysis |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7881994B1 (en) * | 2003-09-11 | 2011-02-01 | Fannie Mae | Method and system for assessing loan credit risk and performance |
| CN102436622A (en) * | 2011-12-28 | 2012-05-02 | 浙江汇信科技有限公司 | Method for evaluating network market operator credit status |
| CN102663643A (en) * | 2012-03-16 | 2012-09-12 | 暨南大学 | Method of farmers' multi-layer one-way network piecewise linear credit rating |
-
2016
- 2016-01-25 CN CN201610049452.8A patent/CN106997472B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7881994B1 (en) * | 2003-09-11 | 2011-02-01 | Fannie Mae | Method and system for assessing loan credit risk and performance |
| CN102436622A (en) * | 2011-12-28 | 2012-05-02 | 浙江汇信科技有限公司 | Method for evaluating network market operator credit status |
| CN102663643A (en) * | 2012-03-16 | 2012-09-12 | 暨南大学 | Method of farmers' multi-layer one-way network piecewise linear credit rating |
Non-Patent Citations (1)
| Title |
|---|
| 熊志斌: "基于非线性主成分分析的信用评估模型研究", 《数量经济技术经济研究》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108256995A (en) * | 2018-01-09 | 2018-07-06 | 秒钱（北京）信息服务有限公司 | A kind of refund wish Forecasting Methodology and system based on contiguous function |
| CN112507204A (en) * | 2019-09-16 | 2021-03-16 | 北京智联云海科技有限公司 | Method for automatically constructing user portrait by utilizing data analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106997472B (en) | 2020-06-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163261A (en) | Unbalanced data disaggregated model training method, device, equipment and storage medium | |
| CN107203774A (en) | The method and device that the belonging kinds of data are predicted | |
| CA2485263A1 (en) | Electronic data processing system and method of using an electronic data processing system for automatically determining a risk indicator value | |
| CN108573358A (en) | A kind of overdue prediction model generation method and terminal device | |
| CN109615280A (en) | Employee's data processing method, device, computer equipment and storage medium | |
| CN110288350A (en) | User's Value Prediction Methods, device, equipment and storage medium | |
| CN110930038A (en) | Loan demand identification method, loan demand identification device, loan demand identification terminal and loan demand identification storage medium | |
| CN107545038A (en) | A kind of file classification method and equipment | |
| CN111199469A (en) | User payment model generation method and device and electronic equipment | |
| CN110222733A (en) | The high-precision multistage neural-network classification method of one kind and system | |
| CN111210332A (en) | Method and device for generating post-loan management strategy and electronic equipment | |
| CN106997472A (en) | User characteristics sorting technique, user credit appraisal procedure and the device of user credit model | |
| CN112508684B (en) | Collecting-accelerating risk rating method and system based on joint convolutional neural network | |
| CN111160929B (en) | Method and device for determining client type | |
| CN116611911A (en) | Credit risk prediction method and device based on support vector machine | |
| Kadam et al. | Loan approval prediction system using logistic regression and CIBIL score | |
| CN106874286A (en) | A kind of method and device for screening user characteristics | |
| CN115358330A (en) | Client user loss prediction method, device, equipment and storage medium | |
| CN115049484A (en) | Risk account prediction method and device and electronic equipment | |
| CN114170000A (en) | Credit card user risk category identification method, device, computer equipment and medium | |
| CN113177733A (en) | Medium and small micro-enterprise data modeling method and system based on convolutional neural network | |
| CN112150276A (en) | Training method, using method, device and equipment of machine learning model | |
| CN114693325A (en) | User public praise intelligent guarantee method and device based on neural network | |
| CN111581068A (en) | Terminal workload calculation method and device, storage medium, terminal and cloud service system | |
| CN111784503B (en) | Operation rendering method, system and storage medium of communication credit investigation data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |