CN110196797A - Automatic optimization method and system suitable for credit scoring card system - Google Patents
Automatic optimization method and system suitable for credit scoring card system Download PDFInfo
- Publication number
- CN110196797A CN110196797A CN201910491304.5A CN201910491304A CN110196797A CN 110196797 A CN110196797 A CN 110196797A CN 201910491304 A CN201910491304 A CN 201910491304A CN 110196797 A CN110196797 A CN 110196797A
- Authority
- CN
- China
- Prior art keywords
- scoring
- card system
- variable
- value
- accuracy rate
- 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
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3447—Performance evaluation by modeling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Technology Law (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Hardware Design (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention discloses a kind of automatic optimization methods and system suitable for credit scoring card system, periodically credit scoring card system is assessed in terms of accuracy rate, types of variables two, different optimization methods is executed according to assessment result, optimal variable segmentation is carried out by way of computer iterations, so that variable separating capacity is most strong, adaptability is most strong, while the accuracy of the credit card points-scoring system ensured is best.The present invention can Automatic Optimal credit scoring card system, deduction and exemption artificial screening variable, manual analysis variable, merge variable, statistics score value and storage workload, reject the interference of artificial cognition, it is more accurate just；Credit card points-scoring system after optimization can discover extraneous variation according to sensitivity, and dynamic updates, and the newest demand to adapt to user is reduced as far as operand, improves optimal speed under the premise of meeting credit scoring actual demand.
Description
Technical field
The present invention relates to credit scoring technology fields, in particular to a kind of suitable for the automatic excellent of credit scoring card system
Change method and system.
Background technique
In the examination & approval link of credit financing, need to assess the credit standing of client, credit scoring card technique is mesh
The technological means of preceding mainstream.As the financial institution supervised by the Banking Supervision Commission, used credit scoring model will have interpretable
Property.
Application No. is 201810566473.6 patent of invention " based on maximize IV data grouping method, apparatus, storage
Deposit medium and equipment " in, disclose it is a kind of based on the data grouping method for maximizing IV, by being carried out pair to sample according to variable
This grouping, calculating are grouped corresponding IV value every time, and the corresponding packet mode of maximum IV value is selected to model for credit scoring card,
Improve the prediction accuracy of credit scoring card mold type.But it is grouped just for own variable, then creates credit scoring
Card mold type, packet mode is also relatively fixed, do not account for because variable dynamic change caused by group mode update,
Even the optimization of credit scoring card mold type updates.
As the fast development of information age and the extensive use of big data technology relate in terms of credit financing service
And types of variables, each variable becomes the interval division mode of the weighing factor of credit scoring, variable being constantly occurring
Change.In practical applications, with the variation of external environment, for example, there is the new variable being affected to credit scoring, or
The original scoring parameter of person gradually give up do not have to perhaps change or it is unchanged even if scoring parameter, the segmentation that uses now
Mode is not suitable for current environment etc., is likely to result in practical scoring accuracy rate and declines or on a declining curve.
A kind of automatic optimization method is needed at present at present, can be continued to optimize credit scoring card system, be adapted it to the epoch
Growth requirement.
Summary of the invention
It is an object of that present invention to provide a kind of automatic optimization methods and system suitable for credit scoring card system, periodically from standard
Two true rate, types of variables aspects assess credit scoring card system, and different optimization methods is executed according to assessment result,
Optimal variable segmentation is carried out by way of computer iterations, so that variable separating capacity is most strong, adaptability is most strong, while really
The accuracy for protecting obtained credit card points-scoring system is best.The present invention can Automatic Optimal credit scoring card system, deduction and exemption are artificial
Selection variables, manual analysis variable, the workload for merging variable, statistics score value and storage, reject the interference of artificial cognition, more
It is precisely just；Credit card points-scoring system after optimization can discover extraneous variation according to sensitivity, and dynamic updates, to adapt to user's
Newest demand is reduced as far as operand, improves optimal speed under the premise of meeting credit scoring actual demand.
To reach above-mentioned purpose, in conjunction with Fig. 1, the present invention proposes a kind of Automatic Optimal side suitable for credit scoring card system
Method, the automatic optimization method include:
S1: assessing credit scoring card system according to the setting period, and evaluation item includes practical scoring accuracy rate, becomes
Type is measured, if assessment is qualified, S10 is entered step, otherwise, enters step S2.
S2: unqualified reason is assessed in judgement, unqualified reason include any one scoring parameter be rejected, variable data library
It is middle there is new variable, practical scoring accuracy rate is unsatisfactory for sets requirement.
If unqualified reason is that any one scoring parameter is rejected, S9 is entered step, otherwise, enters step S3.
S3: creation and/or updating variable data library, be stored in the variable data library several numeric type variables, with
And the corresponding sample information of each numeric type variable.
S4: arbitrarily selecting one of numeric type variable from the variable data library, and the numeric type variable is corresponding
After numerical value is according to setting ordering rule sequence, numerical value is segmented by N number of Minimum Area section according to preset chopping rule, statistics is every
The corresponding sample information of a area segments.
S5: calculating the WOE value of the numeric type variable for each area segments, calculates the number under the segmented model
The IV value of value type variable.
S6: according to preset merging rule to merge the adjacent Minimum Area section in part, the numeric type variable is iterated to calculate
IV value to judge optimal segmentation mode, wherein the IV value of the numeric type variable is maximum under optimal segmentation mode, if most optimal sorting
Stage mode is divided into P optimal region section.
S7: repeating step S3- step S5, until the optimal of all numeric type variables in the variable data library is calculated
Segmented model and corresponding IV value filter out M numeric type variable as new scoring parameter according to the value of IV value.
S8: optimizing credit scoring card system using new scoring parameter and corresponding optimal segmentation mode, calculates
Each optimal region section corresponding score value of each new scoring parameter under optimal segmentation mode is obtained, S10 is entered step.
S9: the scoring parameter being rejected described in rejecting, using new scoring parameter and corresponding optimal segmentation mode to letter
It is optimized with scorecard system, credit scoring card system is assessed, assessment is qualified, enters step S10, otherwise, returns
Step S3.
S10: terminate this Optimizing Flow.
Based on the aforementioned automatic optimization method suitable for credit scoring card system, the present invention further mentions a kind of suitable for credit scoring
The Automatic Optimal system of card system, the Automatic Optimal system include evaluation subsystem, variable data library, the first update module,
Assess sample database, optimum management subsystem.
The variable data library is to store several numeric type variables and the corresponding sample letter of each numeric type variable
Breath.
The assessment sample database assesses sample information used by assessment credit scoring card system to store.
First update module, in real time/periodically update data in variable data library.
The evaluation subsystem includes the second update module, detection module, accuracy rate evaluation module, the first setting module.
First setting module is to be arranged setting difference threshold and setting accuracy rate threshold value.
Second update module is to the data in real time/periodic evaluation sample database.
The detection module in real time/be periodically detected in variable data library new variable whether occur.
The accuracy rate evaluation module is to real-time/periodically invoked credit scoring card system in assessment sample database
Sample carry out credit scoring, statistics obtains the corresponding practical scoring accuracy rate of current credit scoring card system, and
By the standard of assessment corresponding to the practical scoring accuracy rate being calculated and current credit scoring card system optimization version
True rate and setting accuracy rate threshold value are made comparisons.
Wherein, if any one in the following conditions is set up: 1) actually detected accuracy rate is less than assessment accuracy rate, and the two
Between difference be greater than setting difference threshold, 2) actually detected accuracy rate is less than setting accuracy rate threshold value, determine that practical scoring is quasi-
True rate is unsatisfactory for sets requirement.
The accuracy rate evaluation module is set up in response to any one in the following conditions: 1) practical scoring accuracy rate is unsatisfactory for
Sets requirement, 2) occur new variable in variable data library, 3) any one scoring parameter is rejected, and sends optimization signal extremely
Optimum management subsystem includes assessing unqualified reason in the optimization signal.
The optimum management subsystem receives optimization signal, optimizes to credit scoring card system.
Automatic optimization method proposed by the present invention suitable for credit scoring card system includes following components.
One, credit scoring card system is assessed
Periodically credit scoring card system is assessed, evaluation item includes practical scoring accuracy rate, types of variables.
Purpose is that timely assessment has the unreasonable place of credit scoring card system, is executed according to assessment result different excellent
Change method is reduced as far as operand, improves optimal speed under the premise of meeting credit scoring actual demand.
From the foregoing it will be appreciated that unqualified reason includes that any one scoring parameter is rejected, occurs in variable data library newly
Variable, practical scoring accuracy rate be unsatisfactory for sets requirement etc..
If is there is new variable in unqualified reason in variable data library, practical scoring accuracy rate is unsatisfactory for setting and wants
It asks, calls variable data library, screening scoring parameter and corresponding optimal segmentation mode, carry out credit scoring card system again
Optimization.
If unqualified reason is that any one scoring parameter is rejected, the scoring parameter being first rejected described in rejecting,
Credit scoring card system is optimized using new scoring parameter and corresponding optimal segmentation mode, to credit scoring card system
It is assessed, assessment is qualified, terminates Optimizing Flow, and assessment is unqualified, illustrates that there is also other unqualified reasons, recalls variable
Database, screening scoring parameter and corresponding optimal segmentation mode, optimize credit scoring card system again.Pass through the party
Method can reduce operand, save the optimization time.
Two, screening scoring parameter and corresponding optimal segmentation mode
Firstly, one of numeric type variable is arbitrarily selected in the variable data library, the numeric type variable is corresponding
After numerical value is according to setting ordering rule sequence, numerical value is segmented by N number of Minimum Area section according to preset chopping rule, statistics is every
The corresponding sample information of a area segments.The WOE value that the numeric type variable is calculated for each area segments, calculates the segmentation
The IV value of the numeric type variable under mode.
Secondly, iterating to calculate the numeric type to merge the adjacent Minimum Area section in part according to preset merging rule and becoming
The IV value of amount is to judge optimal segmentation mode, wherein the IV value of the numeric type variable is maximum under optimal segmentation mode, if optimal
Segmented model is divided into P optimal region section.
Finally, abovementioned steps are repeated, until the most optimal sorting of all numeric type variables in the variable data library is calculated
Stage mode and corresponding IV value filter out M numeric type variable as new scoring parameter according to the value of IV value.
By division Minimum Area section, then adjacent Minimum Area section is successively merged, iterates to calculate the mode of IV value, it can be true
Protecting each screening process includes the segmented model as much as possible with continuity, regularity, on the one hand, convenient for using computer
System automatic screening, on the other hand, the maximum IV value and optimal segmentation mode made is relatively reasonable.
Three, using new scoring parameter and corresponding optimal segmentation mode, credit scoring card system is optimized
Using machine learning algorithm, in conjunction with new scoring parameter and corresponding optimal segmentation mode, to credit scoring card system
System optimizes, and the credit scoring card system after optimization is updated based on newest variable data library, meets the newest demand of user.
The above technical solution of the present invention, compared with existing, significant beneficial effect is:
(1) Automatic Optimal credit scoring card system, deduction and exemption artificial screening variable, merge variable, statistics at manual analysis variable
The workload of score value and storage rejects the interference of artificial cognition, more accurate just.
(2) optimal variable segmentation is carried out by way of computer iterations, so that variable separating capacity is most strong, adaptability
It is most strong, while the accuracy of the credit card points-scoring system ensured is best.
(3) the different unqualified reasons of assessment are combined, is optimized on the basis of original credit card points-scoring system, reduces operation
Amount.
(4) credit card points-scoring system assessed in terms of accuracy rate, types of variables two, optimized, after making optimization
Credit card points-scoring system can discover extraneous variation according to sensitivity, and dynamic updates, to adapt to the newest demand of user.
It should be appreciated that as long as aforementioned concepts and all combinations additionally conceived described in greater detail below are at this
It can be viewed as a part of the subject matter of the disclosure in the case that the design of sample is not conflicting.In addition, required guarantor
All combinations of the theme of shield are considered as a part of the subject matter of the disclosure.
Can be more fully appreciated from the following description in conjunction with attached drawing present invention teach that the foregoing and other aspects, reality
Apply example and feature.The features and/or benefits of other additional aspects such as illustrative embodiments of the invention will be below
Description in it is obvious, or learnt in practice by the specific embodiment instructed according to the present invention.
Detailed description of the invention
Attached drawing is not intended to drawn to scale.In the accompanying drawings, identical or nearly identical group each of is shown in each figure
It can be indicated by the same numeral at part.For clarity, in each figure, not each component part is labeled.
Now, example will be passed through and the embodiments of various aspects of the invention is described in reference to the drawings, in which:
Fig. 1 is the flow chart of method of the invention.
Specific embodiment
In order to better understand the technical content of the present invention, special to lift specific embodiment and institute's accompanying drawings is cooperated to be described as follows.
In conjunction with Fig. 1, the present invention refers to a kind of automatic optimization method suitable for credit scoring card system, the Automatic Optimal side
Method includes:
S1: assessing credit scoring card system according to the setting period, and evaluation item includes practical scoring accuracy rate, becomes
Type is measured, if assessment is qualified, S10 is entered step, otherwise, enters step S2.
S2: unqualified reason is assessed in judgement, unqualified reason include any one scoring parameter be rejected, variable data library
It is middle there is new variable, practical scoring accuracy rate is unsatisfactory for sets requirement.
If unqualified reason is that any one scoring parameter is rejected, S9 is entered step, otherwise, enters step S3.
S3: creation and/or updating variable data library, be stored in the variable data library several numeric type variables, with
And the corresponding sample information of each numeric type variable.
S4: arbitrarily selecting one of numeric type variable from the variable data library, and the numeric type variable is corresponding
After numerical value is according to setting ordering rule sequence, numerical value is segmented by N number of Minimum Area section according to preset chopping rule, statistics is every
The corresponding sample information of a area segments.
S5: calculating the WOE value of the numeric type variable for each area segments, calculates the number under the segmented model
The IV value of value type variable.
S6: according to preset merging rule to merge the adjacent Minimum Area section in part, the numeric type variable is iterated to calculate
IV value to judge optimal segmentation mode, wherein the IV value of the numeric type variable is maximum under optimal segmentation mode, if most optimal sorting
Stage mode is divided into P optimal region section.
S7: repeating step S3- step S5, until the optimal of all numeric type variables in the variable data library is calculated
Segmented model and corresponding IV value filter out M numeric type variable as new scoring parameter according to the value of IV value.
S8: optimizing credit scoring card system using new scoring parameter and corresponding optimal segmentation mode, calculates
Each optimal region section corresponding score value of each new scoring parameter under optimal segmentation mode is obtained, S10 is entered step.
S9: the scoring parameter being rejected described in rejecting, using new scoring parameter and corresponding optimal segmentation mode to letter
It is optimized with scorecard system, credit scoring card system is assessed, assessment is qualified, enters step S10, otherwise, returns
Step S3.
S10: terminate this Optimizing Flow.
Scheme proposed by the invention is described in detail from the following aspects below.
Automatic optimization method proposed by the present invention suitable for credit scoring card system includes following components.
One, credit scoring card system is assessed
In step S2, the practical scoring accuracy rate is unsatisfactory for sets requirement and refers to:
S21: the practical scoring accuracy rate of credit scoring card system is calculated according to the setting period.
S22: it will be commented corresponding to the practical scoring accuracy rate being calculated and current credit scoring card system optimization version
Estimate accuracy rate and setting accuracy rate threshold value is made comparisons.
Wherein, if any one in the following conditions is set up: 1) actually detected accuracy rate is less than assessment accuracy rate, and the two
Between difference be greater than setting difference threshold, 2) actually detected accuracy rate is less than setting accuracy rate threshold value, determine that practical scoring is quasi-
True rate is unsatisfactory for sets requirement.
Preferably, the method also includes:
Credit scoring card system after optimization is assessed, the assessment accuracy rate corresponding to it is obtained.
For example, assessing after certain suboptimization the credit scoring card system after optimization, the assessment obtained corresponding to it is quasi-
True rate is 97%, sets accuracy rate threshold value as 95%.
Example one
If obtained actually detected accuracy rate is 94% in optimizing certain periodic evaluation after a period of time,
Due to its be less than setting accuracy rate threshold value, determine that the credit scoring card system is no longer satisfied actual demand, need at once into
Row optimization.
Example two
If in optimizing certain periodic evaluation after a period of time, obtained actually detected accuracy rate is
95.5%, that is, be larger than setting accuracy rate threshold value, but the assessment accuracy rate difference of itself and just optimization when finishing is larger, present compared with
For apparent downward trend, this downward trend is likely due to the variation of scoring parameter or unreasonable causes.It can be pre-
It measures, in a short period of time, actually detected accuracy rate would be possible to drop under setting accuracy rate threshold value, so that credit is commented
Card system is divided not to be able to satisfy actual demand, at this point it is possible to which the optimization of giving a forecast property updates.
Preferably, in step S21, the actually detected accuracy rate packet that credit scoring card system is calculated according to the setting period
Include following steps:
S211: creation assessment sample database in real time or periodically updates assessment sample database.
S212: it calls credit scoring card system to carry out credit to the sample in assessment sample database according to the setting period and comments
Point, statistics obtains the corresponding practical scoring accuracy rate of current credit scoring card system.
Two, screening scoring parameter and corresponding optimal segmentation mode
Firstly, one of numeric type variable is arbitrarily selected in the variable data library, the numeric type variable is corresponding
After numerical value is according to setting ordering rule sequence, numerical value is segmented by N number of Minimum Area section according to preset chopping rule, statistics is every
The corresponding sample information of a area segments.The WOE value that the numeric type variable is calculated for each area segments, calculates the segmentation
The IV value of the numeric type variable under mode.
Secondly, iterating to calculate the numeric type to merge the adjacent Minimum Area section in part according to preset merging rule and becoming
The IV value of amount is to judge optimal segmentation mode, wherein the IV value of the numeric type variable is maximum under optimal segmentation mode, if optimal
Segmented model is divided into P optimal region section.
Finally, abovementioned steps are repeated, until the most optimal sorting of all numeric type variables in the variable data library is calculated
Stage mode and corresponding IV value filter out M numeric type variable as new scoring parameter according to the value of IV value.
The WOE value that the numeric type variable is calculated for each area segments calculates the numerical value under the segmented model
The process of the IV value of type variable includes:
S31: the corresponding WOE value of each area segments is calculated according to following formula:
Wherein, woeiIt is WOE value, py corresponding to ith zone sectioniIt is good sample number corresponding to ith zone section
Amount, pniIt is bad sample size corresponding to ith zone section, ∑ pyjIt is that numeric type variable described in variable data library is corresponding
Good total sample number, ∑ pnjIt is the corresponding bad total sample number of numeric type variable, i=1,2 ..., p described in variable data library.
S32: the IV value of the numeric type variable under current fragment mode is calculated according to following formula:
Below as an example with one of numeric type variable, aforementioned screening process is described in detail.
Step 1: a numeric type variable is divided into N sections.
Wherein, x1Indicate the minimum value of the numerical variable, xNIndicate the maximum value of the numerical variable, remaining yjPoint is minute
Duan Dian, j=1,2 ..., N-1.
Calculate corresponding WOE value and IV value under the segmented model.
Step 2: merging automatically to adjacent piecewise interval, corresponding WOE value is iterated to calculate.
This transformation can carry out adjacent interval merging to the section after segmentation, judge optimal waypoint by iterative calculation,
So that final IV value is maximum, i.e., the variable is maximum by the resolution capability for screening fine or not client after such segmentation.Pass through the step
Make (N-1) a waypoint reconsolidate into (p-1) a waypoint: [x suddenly1,x2),[x2,x3)…[xp,xN)。
Corresponding IV value is calculated according to the WOE value of calculating, formula is as follows:
Step 3: constantly adjusting to variable waypoint, optimal IV value is iterated to calculate, so that the variable has fine or not client
There is optimal separating capacity.
In step S7, the value according to IV value filters out M numeric type variable and refers to as scoring parameter,
IV value is sorted according to descending sequence, the maximum preceding M numeric type variable of value is selected to join as scoring
Amount, or IV value is greater than all numeric type variables of default IV value threshold value as scoring parameter.Preferably, the default IV value
Threshold value is 0.1.
The IV value of one variable is bigger, and the information value contained is bigger, and the ability as scoring, prediction is bigger.Choosing
IV value is selected to be greater than the numeric type variable of default IV value threshold value or the maximum M numeric type variable of IV value is selected to join as scoring
Amount, it can be ensured that the reasonability for the parameter selection that scores.
In such credit scoring card system, it is segmented by screen into mould to variable automatically, and to the difference of variable
Combination is iterated, and selects optimal segmented mode, guarantees that variable has best separating capacity to target variable.
Table 1 is two kinds of segmented model examples of one of variable.
Table 1
From table 1, by taking 6 months credit card amounts as an example, automatic segmentation obtains the segmented mode of A, in segmented model A
Area segments be Minimum Area section, the WOE and IV of each area segments corresponding to segmented model A is calculated.
Again by the merging of adjacent fields, segmented model B is obtained, calculates WOE value again, ceaselessly iteration changes adjacent words
Section recalculates WOE value.Judge whether segmented model B is better than segmented model A by calculating whole IV value.
The present invention obtains maximum IV value by obtaining optimal segmented mode to the merging different field.It is maximum
Segmented model corresponding to IV value is exactly the optimal segmentation mode of the field.
All models in variable data library are entered after moding amount carries out above-mentioned processing, IV value size is ranked up,
The IV value (default IV value be greater than 0.1) strong to separating capacity enters mould and participates in scoring calculating, every section after segmentation is given certain
Score value accumulates these score values and can get the client and finally scores.This process is counted when realizing using automation SQL statement
According to the writing in library.This modeling procedure, which only needs to be ready to data, to be automatically completed, and the artificial link for participating in adjustment is saved,
It is more accurate just, reject the interference of artificial cognition.
Three, using new scoring parameter and corresponding optimal segmentation mode, credit scoring card system is optimized
Using machine learning algorithm, in conjunction with new scoring parameter and corresponding optimal segmentation mode, to credit scoring card system
System optimizes, and the credit scoring card system after optimization is updated based on newest variable data library, meets the newest demand of user.
In step S8, it is described credit scoring card system is optimized using new scoring parameter the following steps are included:
S81: using the new corresponding sample information of scoring parameter as modeling sample, use machine learning algorithm with excellent
Change credit scoring card system.
S82: the credit scoring card system after export optimization obtains each new scoring parameter under optimal segmentation mode
The corresponding score value of each optimal region section.
Based on the aforementioned automatic optimization method suitable for credit scoring card system, the present invention further mentions a kind of suitable for credit scoring
The Automatic Optimal system of card system, the Automatic Optimal system include evaluation subsystem, variable data library, the first update module,
Assess sample database, optimum management subsystem.
The variable data library is to store several numeric type variables and the corresponding sample letter of each numeric type variable
Breath.
The assessment sample database assesses sample information used by assessment credit scoring card system to store.
First update module, in real time/periodically update data in variable data library.
The evaluation subsystem includes the second update module, detection module, accuracy rate evaluation module, the first setting module.
First setting module is to be arranged setting difference threshold and setting accuracy rate threshold value.
Second update module is to the data in real time/periodic evaluation sample database.
The detection module in real time/be periodically detected in variable data library new variable whether occur.
The accuracy rate evaluation module is to real-time/periodically invoked credit scoring card system in assessment sample database
Sample carry out credit scoring, statistics obtains the corresponding practical scoring accuracy rate of current credit scoring card system, and
By the standard of assessment corresponding to the practical scoring accuracy rate being calculated and current credit scoring card system optimization version
True rate and setting accuracy rate threshold value are made comparisons.
Wherein, if any one in the following conditions is set up: 1) actually detected accuracy rate is less than assessment accuracy rate, and the two
Between difference be greater than setting difference threshold, 2) actually detected accuracy rate is less than setting accuracy rate threshold value, determine that practical scoring is quasi-
True rate is unsatisfactory for sets requirement.
The accuracy rate evaluation module is set up in response to any one in the following conditions: 1) practical scoring accuracy rate is unsatisfactory for
Sets requirement, 2) occur new variable in variable data library, 3) any one scoring parameter is rejected, and sends optimization signal extremely
Optimum management subsystem includes assessing unqualified reason in the optimization signal.
The optimum management subsystem receives optimization signal, optimizes to credit scoring card system.
Further, the optimum management subsystem includes the second setting module, scoring parameter acquisition module, optimization mould
Block, evaluation module, export module.
Second setting module is to set the ordering rule of each variable, chopping rule, merging in variable data library
Rule.
The scoring parameter obtains ordering rule, chopping rule, merging rule of the module to use each variable, will be every
A variable partitions calculate the IV value under every kind of segmented model at a variety of segmented models, to judge that optimal segmentation mode and its institute are right
The maximum IV value answered filters out M scoring parameter according to the maximum IV value of each variable.
The optimization module is to use machine learning algorithm to comment to optimize credit according to the M scoring parameter filtered out
Divide card system.
For the evaluation module to assess the credit scoring card system after optimization, the assessment obtained corresponding to it is quasi-
True rate.
The export module is to export the credit scoring card system after optimizing and each new scoring parameter optimal
The corresponding score value of each optimal region section under segmented model.
Various aspects with reference to the accompanying drawings to describe the present invention in the disclosure, shown in the drawings of the embodiment of many explanations.
Embodiment of the disclosure need not be defined on including all aspects of the invention.It should be appreciated that a variety of designs and reality presented hereinbefore
Those of apply example, and describe in more detail below design and embodiment can in many ways in any one come it is real
It applies, this is because conception and embodiment disclosed in this invention are not limited to any embodiment.In addition, disclosed by the invention one
A little aspects can be used alone, or otherwise any appropriately combined use with disclosed by the invention.
Although the present invention has been disclosed as a preferred embodiment, however, it is not to limit the invention.Skill belonging to the present invention
Has usually intellectual in art field, without departing from the spirit and scope of the present invention, when can be used for a variety of modifications and variations.Cause
This, the scope of protection of the present invention is defined by those of the claims.
Claims (10)
1. a kind of automatic optimization method suitable for credit scoring card system, which is characterized in that the automatic optimization method includes:
S1: assessing credit scoring card system according to the setting period, and evaluation item includes practical scoring accuracy rate, variable class
Type enters step S10, otherwise, enters step S2 if assessment is qualified；
S2: unqualified reason is assessed in judgement, and unqualified reason includes that any one scoring parameter is rejected, goes out in variable data library
New variable is showed, practical scoring accuracy rate is unsatisfactory for sets requirement；
If unqualified reason is that any one scoring parameter is rejected, S9 is entered step, otherwise, enters step S3；
S3: creation and/or updating variable data library, several numeric type variables and every are stored in the variable data library
The corresponding sample information of a numeric type variable；
S4: arbitrarily selecting one of numeric type variable from the variable data library, by the corresponding numerical value of numeric type variable
After setting ordering rule sequence, numerical value is segmented by N number of Minimum Area section according to preset chopping rule, counts each area
The corresponding sample information of domain section；
S5: calculating the WOE value of the numeric type variable for each area segments, calculates the numeric type under the segmented model
The IV value of variable；
S6: according to preset merging rule to merge the adjacent Minimum Area section in part, the IV of the numeric type variable is iterated to calculate
Value is to judge optimal segmentation mode, wherein the IV value of the numeric type variable is maximum under optimal segmentation mode, if optimal segmentation mould
Formula is divided into P optimal region section；
S7: repeating step S3- step S5, until the optimal segmentation of all numeric type variables in the variable data library is calculated
Mode and corresponding IV value filter out M numeric type variable as new scoring parameter according to the value of IV value；
S8: credit scoring card system is optimized using new scoring parameter and corresponding optimal segmentation mode, is calculated
Each optimal region section corresponding score value of each new scoring parameter under optimal segmentation mode, enters step S10；
S9: the scoring parameter being rejected described in rejecting comments credit using new scoring parameter and corresponding optimal segmentation mode
Divide card system to optimize, credit scoring card system is assessed, assessment is qualified, enters step S10, otherwise, return step
S3；
S10: terminate this Optimizing Flow.
2. the automatic optimization method according to claim 1 suitable for credit scoring card system, which is characterized in that step S2
In, the practical scoring accuracy rate is unsatisfactory for sets requirement and refers to:
S21: the practical scoring accuracy rate of credit scoring card system is calculated according to the setting period；
S22: by the standard of assessment corresponding to the practical scoring accuracy rate being calculated and current credit scoring card system optimization version
True rate and setting accuracy rate threshold value are made comparisons；
Wherein, if any one in the following conditions is set up: 1) actually detected accuracy rate is less than assessment accuracy rate, and between the two
Difference be greater than setting difference threshold, 2) actually detected accuracy rate is less than setting accuracy rate threshold value, determine practical scoring accuracy rate
It is unsatisfactory for sets requirement.
3. the automatic optimization method according to claim 2 suitable for credit scoring card system, which is characterized in that step S21
In, it is described according to setting the period calculate credit scoring card system actually detected accuracy rate the following steps are included:
Creation assessment sample database, in real time or periodically updates assessment sample database；
It calls credit scoring card system to carry out credit scoring to the sample in assessment sample database according to the setting period, counts
To the corresponding practical scoring accuracy rate of current credit scoring card system.
4. the automatic optimization method according to claim 1 or 2 suitable for credit scoring card system, which is characterized in that institute
State method further include:
Credit scoring card system after optimization is assessed, the assessment accuracy rate corresponding to it is obtained.
5. being suitable for the automatic optimization method of credit scoring card system, feature described in -3 any one according to claim 1
It is, in step S7, the value according to IV value filters out M numeric type variable and refers to as scoring parameter,
IV value is sorted according to descending sequence, selects the maximum preceding M numeric type variable of value as scoring parameter.
6. being suitable for the automatic optimization method of credit scoring card system, feature described in -3 any one according to claim 1
It is, in step S7, the value according to IV value filters out M numeric type variable and refers to as scoring parameter,
IV value is greater than all numeric type variables of default IV value threshold value as scoring parameter.
7. the automatic optimization method according to claim 6 suitable for credit scoring card system, which is characterized in that described default
IV value threshold value is 0.1.
8. being suitable for the automatic optimization method of credit scoring card system, feature described in -3 any one according to claim 1
Be, in step S8, it is described credit scoring card system is optimized using new scoring parameter the following steps are included:
Using the new corresponding sample information of scoring parameter as modeling sample, machine learning algorithm is used to comment to optimize credit
Divide card system；
Credit scoring card system after export optimization, it is optimal each of under optimal segmentation mode to obtain each new scoring parameter
The corresponding score value of area segments.
9. a kind of Automatic Optimal system suitable for credit scoring card system, which is characterized in that the Automatic Optimal system includes commenting
Estimate subsystem, variable data library, the first update module, assessment sample database, optimum management subsystem；
The variable data library is to store several numeric type variables and the corresponding sample information of each numeric type variable；
The assessment sample database assesses sample information used by assessment credit scoring card system to store；
First update module, in real time/periodically update data in variable data library；
The evaluation subsystem includes the second update module, detection module, accuracy rate evaluation module, the first setting module；
First setting module is to be arranged setting difference threshold and setting accuracy rate threshold value；
Second update module is to the data in real time/periodic evaluation sample database；
The detection module in real time/be periodically detected in variable data library new variable whether occur；
The accuracy rate evaluation module is to real-time/periodically invoked credit scoring card system to the sample in assessment sample database
This progress credit scoring, statistics obtain the corresponding practical scoring accuracy rate of current credit scoring card system, and
By assessment accuracy rate corresponding to the practical scoring accuracy rate being calculated and current credit scoring card system optimization version,
And setting accuracy rate threshold value is made comparisons；
Wherein, if any one in the following conditions is set up: 1) actually detected accuracy rate is less than assessment accuracy rate, and between the two
Difference be greater than setting difference threshold, 2) actually detected accuracy rate is less than setting accuracy rate threshold value, determine practical scoring accuracy rate
It is unsatisfactory for sets requirement；
The accuracy rate evaluation module is set up in response to any one in the following conditions: 1) practical scoring accuracy rate is unsatisfactory for setting
It is required that 2) occur new variable in variable data library, 3) any one scoring parameter is rejected, optimization signal is sent to optimizing
Management subsystem includes assessing unqualified reason in the optimization signal；
The optimum management subsystem receives optimization signal, optimizes to credit scoring card system.
10. the Automatic Optimal system according to claim 9 suitable for credit scoring card system, which is characterized in that described excellent
Changing management subsystem includes the second setting module, scoring parameter acquisition module, optimization module, evaluation module, export module；
Second setting module is to set the ordering rule of each variable in variable data library, chopping rule, merge rule；
The scoring parameter obtains module to regular using the ordering rule of each variable, chopping rule, merging, by each change
Amount is divided into a variety of segmented models, calculates the IV value under every kind of segmented model, with judge optimal segmentation mode and its corresponding to
Maximum IV value filters out M scoring parameter according to the maximum IV value of each variable；
The optimization module is to use machine learning algorithm to optimize credit scoring card according to the M scoring parameter filtered out
System；
For the evaluation module to assess the credit scoring card system after optimization, the assessment obtained corresponding to it is accurate
Rate；
The export module is to export the credit scoring card system after optimizing and each new scoring parameter in optimal segmentation
The corresponding score value of each optimal region section under mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910491304.5A CN110196797B (en) | 2019-06-06 | 2019-06-06 | Automatic optimization method and system suitable for credit scoring card system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910491304.5A CN110196797B (en) | 2019-06-06 | 2019-06-06 | Automatic optimization method and system suitable for credit scoring card system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110196797A true CN110196797A (en) | 2019-09-03 |
| CN110196797B CN110196797B (en) | 2022-08-02 |
Family
ID=67754030
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910491304.5A Active CN110196797B (en) | 2019-06-06 | 2019-06-06 | Automatic optimization method and system suitable for credit scoring card system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110196797B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111861734A (en) * | 2020-07-31 | 2020-10-30 | 重庆富民银行股份有限公司 | Test evaluation system and method for three-party data source |
| 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 |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109087196A (en) * | 2018-08-20 | 2018-12-25 | 北京玖富普惠信息技术有限公司 | Credit-graded approach, system, computer equipment and readable medium |
| CN109325639A (en) * | 2018-12-06 | 2019-02-12 | 南京安讯科技有限责任公司 | A kind of credit scoring card automation branch mailbox method for credit forecast assessment |
| CN109325792A (en) * | 2017-07-31 | 2019-02-12 | 北京嘀嘀无限科技发展有限公司 | The branch mailbox method and box separation device of credit evaluation variable, equipment and storage medium |
-
2019
- 2019-06-06 CN CN201910491304.5A patent/CN110196797B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325792A (en) * | 2017-07-31 | 2019-02-12 | 北京嘀嘀无限科技发展有限公司 | The branch mailbox method and box separation device of credit evaluation variable, equipment and storage medium |
| CN109087196A (en) * | 2018-08-20 | 2018-12-25 | 北京玖富普惠信息技术有限公司 | Credit-graded approach, system, computer equipment and readable medium |
| CN109325639A (en) * | 2018-12-06 | 2019-02-12 | 南京安讯科技有限责任公司 | A kind of credit scoring card automation branch mailbox method for credit forecast assessment |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111861734A (en) * | 2020-07-31 | 2020-10-30 | 重庆富民银行股份有限公司 | Test evaluation system and method for three-party data source |
| CN111861734B (en) * | 2020-07-31 | 2024-05-03 | 重庆富民银行股份有限公司 | Test evaluation system and method for three-party data source |
| 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 |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110196797B (en) | 2022-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110543616B (en) | SMT solder paste printing volume prediction method based on industrial big data | |
| CN108846526A (en) | A kind of CO2 emissions prediction technique | |
| CN110909963A (en) | Credit scoring card model training method and taxpayer abnormal risk assessment method | |
| CN107633030A (en) | Credit estimation method and device based on data model | |
| CN111639882B (en) | Deep learning-based electricity risk judging method | |
| CN114048436A (en) | Construction method and construction device for forecasting enterprise financial data model | |
| CN111126865B (en) | Technology maturity judging method and system based on technology big data | |
| CN110135587A (en) | It is hesitated based on section and obscures the Multiple Attribute Group Decision of more granularity decision rough sets | |
| CN107633455A (en) | Credit estimation method and device based on data model | |
| CN111709826A (en) | Target information determination method and device | |
| CN112884590A (en) | Power grid enterprise financing decision method based on machine learning algorithm | |
| CN107766500A (en) | The auditing method of fixed assets card | |
| CN110196797A (en) | Automatic optimization method and system suitable for credit scoring card system | |
| CN110634060A (en) | User credit risk assessment method, system, device and storage medium | |
| CN116468536A (en) | Automatic risk control rule generation method | |
| CN108304975A (en) | A kind of data prediction system and method | |
| CN112037006A (en) | Credit risk identification method and device for small and micro enterprises | |
| CN103942604A (en) | Prediction method and system based on forest discrimination model | |
| CN113656707A (en) | Financing product recommendation method, system, storage medium and equipment | |
| CN109359850A (en) | A kind of method and device generating risk assessment scale | |
| CN108920428B (en) | Fuzzy distance discrimination method based on joint fuzzy expansion principle | |
| CN108197740A (en) | Business failure Forecasting Methodology, electronic equipment and computer storage media | |
| CN112906765A (en) | RBF neural network-based customer money laundering risk grading method and system | |
| CN114663102A (en) | Method, equipment and storage medium for predicting debt subject default based on semi-supervised model | |
| CN106301880A (en) | One determines that cyberrelationship degree of stability, Internet service recommend method and apparatus |
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 | ||
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: No.88, Huaihai Road, Qinhuai District, Nanjing City, Jiangsu Province, 210000 Patentee after: Nanyin Faba Consumer Finance Co.,Ltd. Address before: No.88, Huaihai Road, Qinhuai District, Nanjing City, Jiangsu Province, 210000 Patentee before: SUNING CONSUMER FINANCE Co.,Ltd. |