CN110060144A - Amount model training method, amount appraisal procedure, device, equipment and medium - Google Patents
Amount model training method, amount appraisal procedure, device, equipment and medium Download PDFInfo
- Publication number
- CN110060144A CN110060144A CN201910203514.XA CN201910203514A CN110060144A CN 110060144 A CN110060144 A CN 110060144A CN 201910203514 A CN201910203514 A CN 201910203514A CN 110060144 A CN110060144 A CN 110060144A
- Authority
- CN
- China
- Prior art keywords
- data
- user
- user data
- basic
- amount
- 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
- 238000012549 training Methods 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 69
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 27
- 238000012216 screening Methods 0.000 claims abstract description 26
- 238000012545 processing Methods 0.000 claims description 29
- 238000004590 computer program Methods 0.000 claims description 28
- 238000003066 decision tree Methods 0.000 claims description 25
- 238000003860 storage Methods 0.000 claims description 17
- 238000007781 pre-processing Methods 0.000 claims description 15
- 238000010606 normalization Methods 0.000 claims description 13
- 238000011156 evaluation Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 description 21
- 238000005457 optimization Methods 0.000 description 4
- 230000033228 biological regulation Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000013481 data capture Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000002939 conjugate gradient method Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The present invention discloses a kind of amount model training method, amount appraisal procedure, device, equipment and medium, amount model training method includes carrying out Screening Treatment to original user data, the basic user data for meeting training standard is obtained, basic user data includes the accrediting amount, essential information data, basic assets data and capital consumption data；Essential information data are pre-processed, corresponding division of life span's classification is obtained；Basic assets data are pre-processed, loan repayment capacity grade corresponding with basic user data is obtained；Capital consumption data are pre-processed, consuming capacity grade corresponding with basic user data is obtained；The accrediting amount of basic user data, division of life span's classification, loan repayment capacity grade and consuming capacity grade are labeled, training data is obtained；Training data is trained using GBDT algorithm, obtains pre- accrediting amount model, solves the problems, such as that the pre- accrediting amount obtained by pre- accrediting amount model is not accurate.
Description
Technical field
The present invention relates to intelligent decision field more particularly to a kind of amount model training method, amount appraisal procedure, dresses
It sets, equipment and medium.
Background technique
Currently, each big bank is evaluated according to personal information, it is simultaneously anti-to obtain the pre- accrediting amount (the i.e. maximum accrediting amount)
Feed user, and user can carry out business application according to the pre- accrediting amount and handle.Now industry is to the pre- credit volume of user's petty load
Degree, usually evaluates personal information by pre- accrediting amount model, obtains the pre- accrediting amount.Wherein, the pre- accrediting amount
Model is usually to be modeled according to user basic information and user's Assets, however the Assets of most of user are very
Hardly possible obtains, or the Assets got are more unilateral, cause user data integrity degree lower, so that passing through pre- credit volume
The pre- accrediting amount that degree model is got is not accurate.
Summary of the invention
The embodiment of the present invention provides a kind of amount model training method, amount appraisal procedure, device, equipment and medium, with
Solve the problems, such as that the pre- accrediting amount is not accurate.
A kind of amount model training method, comprising:
Original user data in database is obtained, Screening Treatment is carried out to the original user data, acquisition meets training
The basic user data of standard, the basic user data include the accrediting amount, essential information data, basic assets data and
Capital consumption data；
The essential information data are pre-processed, division of life span's class corresponding with the basic user data is obtained
Not；
The basic assets data are pre-processed, loan repayment capacity corresponding with the basic user data etc. is obtained
Grade；
The capital consumption data are pre-processed, consuming capacity corresponding with the basic user data etc. is obtained
Grade；
To the accrediting amount of the basic user data, division of life span's classification, loan repayment capacity grade and consuming capacity grade
It is labeled, obtains the training data formed based on each basic user data；
Each training data is trained using GBDT algorithm, obtains pre- accrediting amount model.
A kind of amount appraisal procedure, comprising:
It obtains the pre- accrediting amount and checks request, the pre- accrediting amount checks to include user attribute data and target in request
User data, the user attribute data include age of user, user location and user identifier；
According to pre-set pre- credit evaluation condition to the age of user, the user location and the user
Identified determine whether the user attribute data is pre-granted letter data；
If the user attribute data is pre-granted letter data, the target user data is input to the pre- accrediting amount
In model, the initial pre- accrediting amount corresponding with the target user data is obtained；
The table of comparisons is inquired according to the initial pre- accrediting amount, it is pre- to obtain target corresponding with the initially pre- accrediting amount
The accrediting amount.
A kind of amount model training apparatus, comprising:
Basic user data obtains module, for obtaining original user data in database, to the original user data
Screening Treatment is carried out, obtains the basic user data for meeting training standard, the basic user data includes the accrediting amount, base
This information data, basic assets data and capital consumption data；
Division of life span's classification obtains module, for pre-processing to the essential information data, obtains and described basic
The corresponding division of life span's classification of user data；
Loan repayment capacity grade obtains module, for pre-processing to the basic assets data, obtains and described basic
The corresponding loan repayment capacity grade of user data；
Consuming capacity grade obtains module, for pre-processing to the capital consumption data, obtains and described basic
The corresponding consuming capacity grade of user data；
Training data forms module, for the accrediting amount to the basic user data, division of life span's classification, refund energy
Power grade and consuming capacity grade are labeled, and obtain the training data formed based on each basic user data；
Pre- accrediting amount model obtains module, for being trained using GBDT algorithm to each training data, obtains
Take pre- accrediting amount model.
A kind of amount assessment rotary device, comprising:
Request module checks request for obtaining the pre- accrediting amount, and the pre- accrediting amount, which is checked in request, includes
User attribute data and target user data, the user attribute data include age of user, user location and user's mark
Know；
Pre-granted letter data determining module is used for according to pre-set pre- credit evaluation condition to the age of user, institute
It states user location and the user identifier identifies, determine whether the user attribute data is pre-granted letter data；
The initial pre- accrediting amount obtains module, if being pre-granted letter data for the user attribute data, by the target
User data input obtains initial pre- credit volume corresponding with the target user data into the pre- accrediting amount model
Degree；
The pre- accrediting amount of target obtains module, for inquiring the table of comparisons, acquisition and institute according to the initial pre- accrediting amount
State the corresponding pre- accrediting amount of target of the initial pre- accrediting amount.
A kind of computer equipment, including memory, processor and storage are in the memory and can be in the processing
The computer program run on device, the processor realize above-mentioned amount model training method when executing the computer program；
Alternatively, the processor realizes above-mentioned amount appraisal procedure when executing the computer program.
A kind of computer readable storage medium, the computer-readable recording medium storage have computer program, the meter
Calculation machine program realizes above-mentioned amount model training method when being executed by processor；Alternatively, the computer program is held by processor
Above-mentioned amount appraisal procedure is realized when row.
Above-mentioned amount model training method, device, equipment and medium, by carrying out Screening Treatment to original user data,
The basic user data for meeting training standard is obtained, to improve model training speed.Essential information data are pre-processed,
Division of life span's classification corresponding with basic user data, loan repayment capacity grade and consuming capacity grade are obtained, to elemental user number
According to the accrediting amount, division of life span's classification, loan repayment capacity grade and consuming capacity grade be labeled and formed training data,
So as to subsequent according to labeled data, pre- accrediting amount model parameter is adjusted.Each training data is instructed using GBDT algorithm
Practice, so that it is higher that the pre- accrediting amount model accuracy got after consumption data is added.
Above-mentioned amount appraisal procedure, device, equipment and medium, according to pre-set pre- credit evaluation condition to user year
Age, user location and user identifier identify determine whether user attribute data is pre-granted letter data, to realize just
It walks and determines whether the user is the user that can carry out credit.If user attribute data is pre-granted letter data, by target user data
It is input in pre- accrediting amount model, obtains the initial pre- accrediting amount corresponding with target user data, realize initial pre- credit
The determination of amount, the initial pre- accrediting amount exported by pre- accrediting amount model are more accurate.According to the initial pre- accrediting amount
The table of comparisons is inquired, the pre- accrediting amount of target corresponding with the initially pre- accrediting amount is obtained, so that feeding back to the target pre-granted of user
Letter amount is the defined accrediting amount, when so that subsequent user carrying out business application, the accrediting amount and the pre- accrediting amount phase of target
It is poor little.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by institute in the description to the embodiment of the present invention
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be attached according to these
Figure obtains other attached drawings.
Fig. 1 is that the application environment of amount model training method or amount appraisal procedure is illustrated in one embodiment of the invention
Figure；
Fig. 2 is the flow chart of amount model training method in one embodiment of the invention；
Fig. 3 is the flow chart of amount model training method in one embodiment of the invention；
Fig. 4 is the flow chart of amount model training method in one embodiment of the invention；
Fig. 5 is the flow chart of amount model training method in one embodiment of the invention；
Fig. 6 is the flow chart of amount model training method in one embodiment of the invention；
Fig. 7 is the flow chart of amount appraisal procedure in one embodiment of the invention；
Fig. 8 is the functional block diagram of amount model training apparatus in one embodiment of the invention；
Fig. 9 is the functional block diagram of amount assessment device in one embodiment of the invention；
Figure 10 is a schematic diagram of computer equipment in one embodiment of the invention.
Specific embodiment
Below by the attached drawing in knot and the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, every other reality obtained by those of ordinary skill in the art without making creative efforts
Example is applied, shall fall within the protection scope of the present invention.
Amount model training method provided in an embodiment of the present invention, can be applicable in the application environment such as Fig. 1, the amount mould
Type training method is applied in server-side, by obtaining the original user data of authorized user in database, passes through original use
User data determines division of life span's classification and loan repayment capacity grade and consuming capacity grade and is labeled and trains, to get
Pre- accrediting amount model, by pre- accrediting amount model can quick output effect good prediction result, to propose the pre- accrediting amount
The accuracy rate of model, and pre- accrediting amount model only needs once to construct, and Reusability improves pre- accrediting amount model
Recognition efficiency.Wherein, server-side can be with the server-side cluster of the either multiple server-side compositions of independent server-side come real
It is existing.
In one embodiment, it as shown in Fig. 2, providing a kind of amount model training method, applies in Fig. 1 in this way
It is illustrated, specifically comprises the following steps: for server-side
S10: obtaining original user data in database, carries out Screening Treatment to original user data, acquisition meets training
The basic user data of standard, basic user data include the accrediting amount, essential information data, basic assets data and basic
Consumption data.
Wherein, original user data refers to the corresponding user data of each user of credit of storage in the database.Base
This user data, which refers to, carries out the number of users after Screening Treatment to the original user data of the user of credit stored in database
According to.The accrediting amount refers to the accrediting amount corresponding to the user stored in database.Essential information data, which refer to, can determine that use
The corresponding data of all specific fields of family division of life span, for example, gender, age, marital status, education background, inhabitation shape
Condition, working condition, children's situation and spouse's information etc..Basic assets data refer to all of the loan repayment capacity that can determine that user
The corresponding data of specific fields, for example, deposit amount, caravan valuation, monthly deposit and declaration form etc..Capital consumption data refer to
Can determine that the corresponding data of all specific fields of the consuming capacity of user, for example, monthly average consumption data, the monthly average consumption frequency,
Wholesale consumption data and electronic product data.It should be noted that specific fields, which refer to, preconfigured can determine life rank
Section classification, the field of consuming capacity grade and loan repayment capacity grade, for example, telephone number cannot determine division of life span's classification, disappear
Take ability rating and loan repayment capacity grade, is not then specific fields.
Specifically, it is stored with the corresponding original user data of each user of credit in database, passes through preset rules pair
Original user data carries out Screening Treatment.For example, screening out some influence model training times does not influence model training precision
Data, if there are missing values and the missing values are influences model training and be accurately worth in original user data, to the missing
Value carries out interpolation processing.In the present embodiment, the basic user data that can be trained is obtained, includes in basic user data
The corresponding accrediting amount of credit user, essential information data, basic assets data and capital consumption data.By obtaining data
The corresponding original user data of the accrediting amount in library, and Screening Treatment is carried out to original user data, to get basic use
User data, to improve model training speed, by being trained to basic user data, to improve pre- accrediting amount model
Accuracy.
S20: pre-processing essential information data, obtains division of life span's classification corresponding with basic user data.
Pretreatment in the present embodiment refers to that specific fields are got from essential information data is corresponding to determine user
Division of life span's classification data.Division of life span's classification refers to the essential information data according to user to determine locating for user
Classification, division of life span's classification include stage of going to school, struggle stage, the stage of supporting the family and old stage etc..
Specifically, server-side obtains essential information data from basic user data, and carries out to essential information data pre-
Processing, to get the corresponding essential information data to determine division of life span in basic user data, passes through the essential information
Data are to determine the corresponding division of life span's classification of basic user data.It further, include essential information in basic user data
Three data, basic assets data and capital consumption data plates first obtain essential information data from basic user data,
Essential information data are pre-processed again, the corresponding data of specific fields is such as only obtained, filters out name and telephone number etc.
It is not the corresponding data of specific fields, then the corresponding data of specific fields is input in preparatory trained decision-tree model,
Categorised decision is carried out to data by decision-tree model, to obtain division of life span's classification corresponding with basic user data.Pass through
The corresponding division of life span's type of basic user data is determined, to determine the demand for loan of user, for example, the loan in struggle stage
Money demand will be to the demand for loan for being higher than the old stage.It is to be appreciated that if demand for loan is higher, corresponding pre- credit volume
It spends bigger.
S30: pre-processing basic assets data, obtains loan repayment capacity grade corresponding with basic user data.
Pretreatment in the present embodiment refers to that specific fields are got from basic assets data is corresponding to determine user
The data of loan repayment capacity, and be normalized, to determine its loan repayment capacity grade.Loan repayment capacity grade refer to according to
The basic assets data at family are to determine grade locating for user's loan repayment capacity.
Specifically, server-side obtains basic assets data from basic user data, and carries out to basic assets data pre-
Processing, to get the corresponding basic assets data to determine loan repayment capacity grade of basic user data, and provides this substantially
It produces data to be normalized, determines corresponding loan repayment capacity of user etc. by the basic assets data after normalized
Grade.It is to be appreciated that including essential information data, basic assets data and capital consumption data three in basic user data
Plate first obtains basic assets data from basic user data, then pre-processes to basic assets data, such as only obtains
The corresponding data of specific fields, then the corresponding basic assets data of specific fields are normalized, by normalized
Basic assets data afterwards are input in preparatory trained loan repayment capacity prediction model, pass through loan repayment capacity prediction model pair
Basic assets data are identified, loan repayment capacity grade corresponding with basic user data is exported.By determining elemental user number
According to corresponding loan repayment capacity grade, to determine the loan repayment capacity of user, it should be noted that the basic assets data of user
Corresponding assets value is bigger, then illustrates that loan repayment capacity is stronger；Conversely, the corresponding assets value of the basic assets data of user is smaller,
Then illustrate that loan repayment capacity is weaker.It is to be appreciated that the pre- accrediting amount is bigger if the corresponding loan repayment capacity stronger grade of user.
S40: pre-processing capital consumption data, obtains consuming capacity grade corresponding with basic user data.
Pretreatment in the present embodiment refer to from capital consumption plural number in get specific fields it is corresponding to determine use
The data of family consuming capacity, and be normalized.Consuming capacity grade refers to according to the capital consumption data of user with true
Determine grade locating for customer consumption ability.
Specifically, server-side obtains capital consumption data from basic user data, and carries out to capital consumption data pre-
Processing, to get the corresponding data capital consumption data to determine consuming capacity of basic user data, and to disappearing substantially
Expense data are normalized, and determine corresponding consuming capacity of user etc. by the capital consumption data after normalized
Grade.It is to be appreciated that including essential information data, basic assets data and capital consumption data three in basic user data
Plate first obtains capital consumption data from basic user data, then pre-processes to capital consumption data, such as only obtains
The corresponding data of specific fields, then the corresponding data of specific fields are normalized.It is stored in server-side and matches in advance
The corresponding target consumer value of storage capital consumption data and consuming capacity grade in table are assessed in the consumption assessment table set, consumption
Corresponding relationship.In the present embodiment, consumption is searched based on the corresponding target consumer value of capital consumption data after normalized
Table is assessed, consuming capacity grade is obtained.Consuming capacity grade may include four grades, be trend intelligent's label respectively, and wholesale disappears
Take label, electrical type label and economical intelligent's label.By determining the corresponding consuming capacity grade of basic user data, so as to
Determine the consuming capacity of user.It is to be appreciated that the pre- accrediting amount is bigger if the corresponding consuming capacity bigger grade of user.
S50: to the accrediting amount of basic user data, division of life span's classification, loan repayment capacity grade and consuming capacity grade
It is labeled, obtains the training data formed based on each basic user data.
Specifically, server-side obtains the corresponding accrediting amount of each basic user data, division of life span's classification, loan repayment capacity
Grade and consuming capacity grade, are labeled each basic user data, and each basic user data is formed after obtaining mark
Training data.By being labeled to each basic user data, when accrediting amount model pre- so as to subsequent training, according to
Labeled data to adjust corresponding weight so that by pre- accrediting amount model export predicted value it is identical as labeled data, mention
The accuracy of high pre- accrediting amount model.
S60: each training data is trained using GBDT algorithm, obtains pre- accrediting amount model.
Wherein, GBDT (Gradient Boost Decision Tree, gradient promote decision tree) is a kind of based on iteration
The decision Tree algorithms constructed, it can be referred to as again MART (Multiple Additive Regression Tree) or
GBRT (Gradient Boosting Regression Tree).EromeH-Friedman was promoted in proposition gradient in 1999
Decision tree can be used for classifying and returning, and GBDT algorithm can handle a plurality of types of data, including continuous type and discrete type, and right
The processing capacity of exceptional value is stronger.GBDT algorithm will generate more decision trees in practical problem, i.e. generation division of life span's class
Not, loan repayment capacity grade and the corresponding decision tree of consuming capacity grade, and the result of all decision trees is summarized to obtain
The final pre- accrediting amount.
Specifically, server-side obtains training data, comprising each basic user data and corresponding has awarded in training data
Believe amount, division of life span's classification corresponding with essential information data, loan repayment capacity corresponding with basic assets data etc.
Grade, consuming capacity grade corresponding with capital consumption data, are trained training data using GBDT algorithm, to obtain
To pre- accrediting amount model.Training data is trained using GBDT algorithm, is specifically included: (1) being used from training data
Initial weight trains an initial weak learner, compares true value (accrediting amount marked) and predicted value (by initial
The value of weak learner prediction), obtain learning error rate；According to the learning error rate of initial weak learner, training data is updated
Weight, so that the weight of the high training data of initial weak learner learning error rate is got higher, so that error rate is high initial weak
More paid attention in weak learner after learner；(2) second is trained based on the training data after adjustment weight
A weak learner；(3) by continuous iteration, until weak learner number reaches number T specified in advance；(4) finally weak by T
Learner is integrated by aggregation policy, obtains final strong learner, that is, gets pre- accrediting amount model.Wherein, it adopts
It is trained with GBDT algorithm, can effectively carry out feature selecting and processing abnormal point automatically, moreover it is possible to avoid to a certain extent
Model overfitting problem.Pre- accrediting amount model is constructed by CBDT algorithm, so that subsequent obtain according to pre- accrediting amount model
The prediction result (i.e. initial pre-granted amount) arrived is more acurrate.
In step S10-S60, by carrying out Screening Treatment to original user data, the basic use for meeting training standard is obtained
User data, to improve model training speed.Essential information data are pre-processed, are obtained corresponding with basic user data
Division of life span's classification, loan repayment capacity grade and consuming capacity grade, the accrediting amount, division of life span's class to basic user data
Not, loan repayment capacity grade and consuming capacity grade are labeled and are formed training data, so as to subsequent according to labeled data, adjust
Whole pre- accrediting amount model parameter.Each training data is trained using GBDT algorithm, so that being obtained after consumption data is added
The pre- accrediting amount model accuracy got is higher.It should be noted that obtaining division of life span's classification, loan repayment capacity grade and disappearing
Take ability rating part sequencing, can carry out simultaneously, to improve model training speed.
In one embodiment, as shown in figure 3, in step S10, that is, original user data in database is obtained, to original use
User data carries out Screening Treatment, obtains the basic user data for meeting training standard, specifically comprises the following steps:
S11: obtaining original user data in database, judges original user data with the presence or absence of missing values.
Wherein, missing values refer in original user data as lack information and caused by the value of some or certain fields be
It is incomplete.For example, the corresponding value of age field is empty or phone number field in the original user data of a certain user
Not exclusively, then age field and the corresponding value of phone number field are missing values to corresponding value.
Specifically, server-side judges each original user data got, determines each original user data
It is whether complete, that is, it whether there is missing values, wherein judging result can be divided into two kinds, be complete number one is original user data
According to that is, there is no missing values, and another kind is that original user data is not partial data, then original user data has missing
Value.
S12: if missing values are not present in original user data, using original user data as basic user data.
Specifically, server-side judges that missing values, i.e., each field in original user data is not present in original user data
Data be partial data, then using the original user data as basic user data.By the way that there will be no the original of missing values
User data is as basic user data, so that basic user data is complete data.
S13: if there are missing values for original user data, the corresponding field of missing values is obtained.
Wherein, field refers to the corresponding field of original user data intermediate value, for example, a certain field is the age, then value is
One specific value.
Specifically, server-side is judged to obtain each in original user data lack in original user data there are missing values
Mistake value, and corresponding field is obtained according to missing values.By determining the corresponding field of missing values, determine whether to need so as to subsequent
Interpolation processing is carried out to missing values.
S14: if field is specific fields, the processing of interpolation missing values is carried out to missing values, obtains basic user data.
Specifically, specific fields table is stored in database, wherein pre- accrediting amount model is stored in specific fields table
The field of the required data of training, the field of data needed for pre- accrediting amount model training is stored in as specific fields
In specific fields table.Specific fields table is searched by the corresponding field of missing values, determines whether the field is specific fields, i.e.,
Determine whether the corresponding missing values of the field are data needed for model training.If the field is specific fields, to the missing
Value, which carries out interpolation processing, specifically can be used mean value interpolation, similar mean value interpolation, Maximum-likelihood estimation and multiple interpolating method
The processing of interpolation missing values is carried out to missing values.In the present embodiment, carried out at interpolation missing values using similar mean value interpolation method
Reason.Wherein, similar mean value interpolation method is mainly the type for using hierarchical clustering model prediction to lack variable, then with the type
Mean value interpolation.For example, the corresponding field of a certain missing values is " age ", specific fields table is searched by " age ", if specific word
Illustrate that " age " may be used to determine division of life span in segment table, then " age " is specific fields, carries out interpolation missing to the missing values
Value processing, with completion missing values, so that basic user data is more complete.By refer to the corresponding missing values of specific fields into
Row interpolation processing only carries out interpolation processing to data needed for model training, to improve model training speed.
S15: if field is unspecified field, missing values are not handled.
Specifically, it if searching specific fields table according to the field, determines that the field is unspecified field, that is, determines the field
Corresponding missing values are not data needed for model training, then without handling missing values.For example, a certain missing values pair
Answering field is " name ", searches specific fields table by " name ", if illustrating in the specific table of the field, " name " is nonspecific word
Section, then without handling the missing values, to accelerate the speed screened to original user data.
In step S11-S15, original user data is judged with the presence or absence of missing values, to guarantee the complete of model training data
Property.If missing values are not present in original user data, which is complete data, can be instructed directly as model
Experienced basic user data.If there are missing values for original user data, the corresponding field of missing values is obtained；If field is spy
Determine field, then the processing of interpolation missing values is carried out to missing values, the data to guarantee model training improve mould for complete data
The accuracy of type training.If field is unspecified field, missing values are not handled, to improve model training speed.
In one embodiment, as shown in figure 4, in step S20, i.e., essential information data are pre-processed, acquisition and base
The corresponding division of life span's classification of this user data, specifically comprises the following steps:
S21: Screening Treatment is carried out to essential information data using screening rule relevant to division of life span's classification, obtains mesh
Mark information data.
Wherein, screening rule is preset rule, for extracting the mesh of specific field from essential information data
Mark information data.Object information data, which refers to, carries out needed for Screening Treatment model training obtained essential information data
Data.
Specifically, server-side obtains screening rule relevant to division of life span, by screening rule to each elemental user
Essential information data in data are screened, and meet the standards such as pre- accrediting amount model training content and format to get
Object information data.For example, it is capable of washing fall telephone number and name etc., to get the object information data of specific field.
S22: identifying object information data using decision-tree model corresponding with division of life span's classification, obtains base
The corresponding division of life span's classification of this user data.
Wherein, decision-tree model is that preparatory training instructs historical use data in database using decision Tree algorithms
Practice, with the model got.
Wherein, decision Tree algorithms specifically can be used ID3 algorithm and be trained to the historical use data in database,
In, ID3 (Iterative Dichotomiser 3, iteration decision tree) algorithm, is a kind of algorithm for constructing decision tree,
It carries out Attributions selection according to information gain.In every single-step iteration of algorithm, each of ergodic data library is not used
User data in characteristic dimension, calculate this feature dimension entropy or information gain, therefrom selection have minimum entropy or maximum
Root node of the characteristic dimension of information gain as decision tree, then with selected characteristic dimension by the characteristic value in characteristic dimension
It is divided into different attribute values, continues to carry out Recursion process to feature weight table by ID3 algorithm, not have before only considering every time
Selected attribute is completed until decision tree is established.Specifically, the historical use data in database is carried out using ID3 algorithm
Trained process, the i.e. growth course of decision tree, decision tree building are completed, to get determine corresponding with division of life span's classification
Plan tree-model.
Specifically, server-side obtains decision-tree model corresponding with division of life span's classification, by decision-tree model to target
Information data is identified, to get the corresponding division of life span's classification of object information data, that is, is got and elemental user number
According to corresponding division of life span's classification.By decision-tree model to determine the corresponding division of life span's classification of basic user data, so that
The division of life span's classification got is more accurate.
In step S21-S22, essential information data are screened using screening rule relevant to division of life span's classification
Processing, so that the object information data got is complete data.Using decision tree mould corresponding with division of life span's classification
Type identifies that quick obtaining is got to division of life span's classification corresponding with basic user data to object information data
Division of life span's classification is quickly and accurate.
In one embodiment, as shown in figure 5, in step S30, i.e., basic assets data are pre-processed, acquisition and base
The corresponding loan repayment capacity grade of this user data, specifically comprises the following steps:
S31: being normalized basic assets data, the desired asset data after obtaining normalized.
Specifically, server-side obtains the basic assets data in basic user data, and basic assets data include Deposit
Volume, caravan valuation, monthly deposit and declaration form etc., can determine the loan repayment capacity of the user according to basic assets data.By Deposit
Volume, caravan valuation, monthly deposit and the data such as declaration form are normalized, by normalized by basic assets data
It is mapped within the scope of 0 to 1, i.e., basic assets data is become into the decimal between (0,1), by the decimal conduct between (0,1)
Desired asset data.By the way that basic assets data are normalized, the data of different number grade are become into same number
Magnitude, so as to the determination of subsequent loan repayment capacity grade.
S32: carrying out prediction processing to desired asset data using preparatory trained loan repayment capacity prediction model, obtain with
The corresponding loan repayment capacity grade of basic user data.
Wherein, loan repayment capacity prediction model is pre- to first pass through a large amount of historical sample data and be trained obtained mould
Type.Specifically, the corresponding historical sample data of a large amount of different loan repayment capacity grades can be first obtained, and is each historical sample
Data carry out loan repayment capacity grade mark, are carried out curve fitting using multivariate logistic regression algorithm, pre- to get loan repayment capacity
Survey model.Wherein, multivariate logistic regression algorithm judges that a sample is to belong to for a possibility that estimating certain things in other words
In the probability of some grade.
Further, loan repayment capacity prediction model training step, specifically includes: (1) obtaining historical sample data, history
It include the corresponding deposit amount of each historical user, caravan valuation, monthly deposit and declaration form number in sample data.(2) by pre-
Function is surveyed to carry out curve fitting to the corresponding deposit amount of each historical user, caravan valuation, monthly deposit and declaration form number.Its
In, anticipation function are as follows:It enablesThenWhereinWherein, x indicates the history of input
Sample includes n dimensional feature, xiIndicate i-th sampling, θ indicates model parameter.(3) optimization is iterated using optimization formula,
To get loan repayment capacity prediction model.Wherein, optimization formula includes loss function, likelihood function, gradient decline iteration function
Deng.In iterative optimization procedure, first handled using loss function, which is
Make the numerical value of J (θ) minimum, be allowed to approach and 0, i.e. the accrediting amount more matches with predicted value, that is, use likelihood function into
Row maximal possibility estimation, likelihood function areLogarithm is taken to likelihood function,
It is iterated especially by gradient decline iterative formula, iterative formula isWherein α is learning rate,
The movement " stride " for being exactly θ is exactly gradient, the value of solve system of equation, the continuous iteration of θ, when model parameter convergence to a certain extent,
Stop iterative calculation, the θ obtained at this time is final model parameter, to get loan repayment capacity prediction model.It needs to illustrate
, it further includes the corresponding iterative formula such as conjugate gradient method and quasi-Newton method that gradient, which declines iterative formula,.In specific embodiment
In, the optimal model parameters of Logic Regression Models can be calculated by any of the above-described iterative algorithm, trained containing optimal models
The loan repayment capacity prediction model of parameter.
Specifically, the desired asset data that server-side will acquire are input to preparatory trained loan repayment capacity prediction model
In, prediction processing is carried out to desired asset data by loan repayment capacity prediction model, obtains go back corresponding with desired asset data
The probability of money ability rating corresponds to loan repayment capacity grade by the determine the probability desired asset data, that is, gets elemental user
The corresponding loan repayment capacity grade of data.
In step S31-S32, basic assets data are normalized, the desired asset after obtaining normalized
Data, by normalized, so that the basic assets data in different number grade become the desired asset in the same order of magnitude
Data, more quick and precisely so as to the loan repayment capacity grade that gets.Using preparatory trained loan repayment capacity prediction model pair
Desired asset data carry out prediction processing, can the loan repayment capacity grade that arrives of quick obtaining, interim, loan repayment capacity prediction model can
It recycles, improves the utilization rate of loan repayment capacity prediction model.
In one embodiment, capital consumption data include at least one consumer factors, wherein consumer factors refer to monthly disappear
Take the factors such as the amount of money, the consumption frequency, wholesale consumption data and daily purchase product.
As shown in fig. 6, being pre-processed to capital consumption data in step S40, obtain corresponding with basic user data
Consuming capacity grade, specifically comprise the following steps:
S41: being normalized at least one consumer factors, obtains corresponding normalization factor value.
Specifically, server-side obtains the capital consumption data in basic user data, includes at least in capital consumption data
One consumer factors, at least one consumer factors is normalized, and the consumption data of different number grade is become same
The consumption data of one order of magnitude, the normalization factor value corresponding at least one consumer factors after obtaining normalized, tool
Body is the value between 0 to 1.
S42: processing is weighted at least one normalization factor value, obtains corresponding target consumer value.
Specifically, be stored with the corresponding preset weights of each consumer factors in database, default weight be according to consumption because
The value that the importance of element determines.For example, weight corresponding to wholesale consumption data is larger, electrical type is produced in daily purchase product
Product default weight corresponding compared to daily necessity is larger.Obtain the corresponding preset weights of each consumer factors and consumer factors
Corresponding normalization factor value is weighted processing to each consumer factors by weighted formula, obtains target consumer value.Its
In, weighted formula isY is target consumer value, and n is the number of consumer factors, AiIndicate i-th of consumer factors pair
The normalization factor value answered, wiIndicate the corresponding default weight of i-th of consumer factors.
S43: based on target consumer value inquiry consumption assessment table, consuming capacity corresponding with basic user data etc. is obtained
Grade.
Specifically, the corresponding relationship of target consumer value Yu consuming capacity grade is stored in database, according to target consumer
Value searches the consumption in database and assesses table, to determine corresponding with target consumer value consuming capacity grade, that is, gets and base
The corresponding consuming capacity grade of this user data.
In step S41-S43, at least one consumer factors is normalized, obtains corresponding normalization factor
Value is convenient for subsequent determining consuming capacity grade.Processing is weighted at least one normalization factor value, obtains corresponding mesh
Consumption value is marked, and consumption assessment table is inquired according to target consumer value, obtains consuming capacity corresponding with basic user data etc.
Grade realizes the determination of consuming capacity grade.
In one embodiment, amount appraisal procedure provided in an embodiment of the present invention, can be applicable to the application environment such as Fig. 1
In, which applies in server-side, checks request, pre-granted by obtaining the pre- accrediting amount that user terminal is sent
It includes user attribute data and target user data that letter amount, which is checked in request, according to user attribute data and target user's number
According to the pre- accrediting amount of determination.User terminal can be, but not limited to various personal computers, laptop, smart phone, plate electricity
Brain and portable wearable device.Server-side can use the server-side collection of the either multiple server-side compositions of independent server-side
Group realizes.
In one embodiment, as shown in fig. 7, providing a kind of amount appraisal procedure, the service in Fig. 1 is applied in this way
It is illustrated, specifically comprises the following steps: for end
S101: it obtains the pre- accrediting amount and checks request, the pre- accrediting amount checks to include user attribute data and mesh in request
User data is marked, user attribute data includes age of user, user location and user identifier.
Wherein, user identifier refers to the corresponding mark of user, determines unique user, user identifier tool by user identifier
Body can be demonstrate,proved for user identity.
Specifically, server-side provides a data acquisition interface, the interface and user terminal network linking.User is based on user terminal
The pre- accrediting amount, which is sent, to server-side checks request, server-side obtains the pre- accrediting amount according to data acquisition interface and checks request,
The pre- accrediting amount checks that wherein user attribute data includes user comprising user attribute data and target user data in request
Age, user location and user identifier.Wherein, target user data include object information data, desired asset data and
Target consumer data.
Further, target consumer data are obtained specifically comprise the following steps: that the production of (1) pre- accrediting amount is seen in request and include
User account obtains historical consumption data corresponding with user account in the default time limit.Wherein, user account refers to and user
Corresponding account is identified, specifically can be the account of the third-party platform of user terminal authorization, use can be got by user account
The historical consumption data at family.Be provided with the default time limit in server-side in advance.Server-side obtains in the default time limit and user account pair
The historical consumption data answered, wherein the default time limit can be year consumption data and season consumption data etc..(2) number is consumed to history
Repeated data in carries out duplicate removal cleaning treatment, obtains effective transaction data.Wherein, it by duplicate removal cleaning treatment, avoids going through
History consumption data repeats, so that the effective transaction data got is more accurate, that is, the target consumer data got are more acurrate
(3) effective transaction data is analyzed, obtains the monthly average consumption amount of money, the consumption frequency, wholesale consumption data and daily purchase and produces
Product, as target consumer data.Target consumer data are got by user account, improve acquisition speed.
Further, server-side is preset with the filling region of user attribute data and target user data, and user can be based on
User attribute data and target user data carry out data filling to corresponding filling region, in the filling of target user data
It also may include that must fill out option and option is filled out in choosing in region.Specifically, user basic information data, basic assets number are predefined
According to the weight with filled in region corresponding in capital consumption data, using weight it is biggish fill in region as must fill out option, will
Weight is lesser to be used as choosing to fill out option.
S102: according to pre-set pre- credit evaluation condition to age of user, user location and user identifier into
Row identification, determines whether user attribute data is pre-granted letter data.
Wherein, pre- credit evaluation condition refers to preset assessment regulation, for example, the preset and age (over
18 one full year of life and with complete civil acts), region (whether be can credit area) and blacklist (each big bank is corresponding black
List and borrower or lineal relative are without bad credit record) corresponding assessment regulation.
Specifically, server-side is by pre-set pre- credit evaluation condition to the age of user got, where user
Area and user identifier judged, judges whether age of user over 18 years old had complete civil acts, where judging user
Area whether be can credit area, and search according to user identifier the blacklist of each big bank, judge that user identifier is corresponding
Whether user, borrower or lineal relative are without record of bad behavior, to determine whether user attribute data is pre-granted letter data.
S103: if user attribute data is pre-granted letter data, target user data being input in pre- accrediting amount model,
Obtain the initial pre- accrediting amount corresponding with target user data.
Wherein, the initial pre- accrediting amount refers to the awarding corresponding with target user data by the output of pre- accrediting amount model
Believe amount.
Specifically, if user attribute data meets pre- credit evaluation condition, user attribute data is pre-granted letter data, will
Target user data is input in pre- accrediting amount model, is analyzed by pre- accrediting amount model target user data
Processing obtains the initial pre- accrediting amount corresponding with target user data.It should be noted that the initial pre- accrediting amount is
One specific value.
Further, if user attribute data is not pre-granted letter data, feedback information is generated, and feed back to user terminal.Its
In, feedback information is the user that prompting user is not pre- credit.
S104: inquiring the table of comparisons according to the initial pre- accrediting amount, obtains target pre-granted corresponding with the initially pre- accrediting amount
Believe amount.
Wherein, the pre- accrediting amount of target refers to the accrediting amount determined according to the initial pre- accrediting amount, the pre- credit volume of target
Degree is prespecified integer value.
Specifically, the corresponding relationship of the initial pre- accrediting amount and the pre- accrediting amount of target is stored in the table of comparisons.By pre-
The initial pre- accrediting amount that accrediting amount model is got is a specific value, and server-side is searched according to the initial pre- accrediting amount
The table of comparisons obtains pre- as target corresponding with target user data with the initially corresponding pre- accrediting amount of target of the pre- accrediting amount
The accrediting amount.For example, being 2 Wan Yiqian by the initial pre- accrediting amount that pre- accrediting amount model is got, pass through initial pre-granted
Believe that amount searches the table of comparisons, the initial pre- accrediting amount corresponding target pre- credit volume of the regulation less than two Wan Wuqian in the table of comparisons
Degree is 20,000, then by this 20,000 as the pre- accrediting amount of target corresponding with target user data.
In step S101-S104, according to pre-set pre- credit evaluation condition to age of user, user location and
User identifier is identified determine whether user attribute data is pre-granted letter data, with realize primarily determine the user whether be
It can carry out the user of credit.If user attribute data is pre-granted letter data, target user data is input to pre- accrediting amount mould
In type, the initial pre- accrediting amount corresponding with target user data is obtained, the determination of the initial pre- accrediting amount is realized, by pre-
The initial pre- accrediting amount of accrediting amount model output is more accurate.Inquire the table of comparisons according to the initial pre- accrediting amount, obtain with
The initially corresponding pre- accrediting amount of target of the pre- accrediting amount, so that the pre- accrediting amount of the target for feeding back to user is defined awards
Believe amount, when so that subsequent user carrying out business application, the accrediting amount is not much different with the pre- accrediting amount of target.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process
Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit
It is fixed.
In one embodiment, a kind of amount model training apparatus is provided, the amount model training apparatus and above-described embodiment
Middle amount model training method corresponds.As shown in figure 8, the amount model training apparatus includes that basic user data obtains
Module 10, division of life span's classification obtain module 20, loan repayment capacity grade obtains module 30, consuming capacity grade obtains module 40,
Training data forms module 50 and pre- accrediting amount model obtains module 60.Detailed description are as follows for each functional module:
Basic user data obtain module 10, for obtaining original user data in database, to original user data into
Row Screening Treatment obtains the basic user data for meeting training standard, and basic user data includes the accrediting amount, essential information
Data, basic assets data and capital consumption data.
Division of life span's classification obtains module 20, for pre-processing to essential information data, obtains and elemental user number
According to corresponding division of life span's classification.
Loan repayment capacity grade obtains module 30, for pre-processing to basic assets data, obtains and elemental user number
According to corresponding loan repayment capacity grade.
Consuming capacity grade obtains module 40, for pre-processing to capital consumption data, obtains and elemental user number
According to corresponding consuming capacity grade.
Training data forms module 50, for the accrediting amount, the division of life span's classification, loan repayment capacity to basic user data
Grade and consuming capacity grade are labeled, and obtain the training data formed based on each basic user data.
Pre- accrediting amount model obtains module 60, for being trained using GBDT algorithm to each training data, obtains
Pre- accrediting amount model.
In one embodiment, it includes missing values judging unit, the first determination unit, that basic user data, which obtains module 10,
Two determination units, missing values processing unit and third determination unit.
Missing values judging unit judges that original user data whether there is for obtaining original user data in database
Missing values.
First determination unit, if missing values are not present for original user data, using original user data as basic
User data.
Second determination unit obtains the corresponding field of missing values if there are missing values for original user data.
Missing values processing unit carries out the processing of interpolation missing values to missing values, obtains if being specific fields for field
Basic user data.
Third determination unit is not handled missing values if being unspecified field for field.
In one embodiment, it includes object information data acquiring unit and division of life span that division of life span's classification, which obtains module 20,
Classification acquiring unit.
Object information data acquiring unit, for using screening rule relevant to division of life span's classification to essential information number
According to Screening Treatment is carried out, object information data is obtained.
Division of life span's classification acquiring unit, for being believed using decision-tree model corresponding with division of life span's classification target
Breath data are identified, the corresponding division of life span's classification of basic user data is obtained.
In one embodiment, it includes: desired asset data capture unit and refund energy that loan repayment capacity grade, which obtains module 30,
Power grade acquiring unit.
Desired asset data capture unit obtains normalized for basic assets data to be normalized
Desired asset data afterwards.
Loan repayment capacity grade acquiring unit, for using preparatory trained loan repayment capacity prediction model to desired asset number
According to prediction processing is carried out, loan repayment capacity grade corresponding with basic user data is obtained.
In one embodiment, capital consumption data include at least one consumer factors.
Consuming capacity grade obtain module 40 include: normalization factor value acquiring unit, target consumer value acquiring unit and
Consuming capacity grade acquiring unit.
Normalization factor value acquiring unit obtains corresponding at least one consumer factors to be normalized
Normalization factor value.
Target consumer value acquiring unit obtains corresponding for being weighted processing at least one normalization factor value
Target consumer value.
Consuming capacity grade acquiring unit is used for based on target consumer value inquiry consumption assessment table, acquisition and elemental user
The corresponding consuming capacity grade of data.
Specific about amount model training apparatus limits the limit that may refer to above for amount model training method
Fixed, details are not described herein.Modules in above-mentioned amount model training apparatus can fully or partially through software, hardware and
Its group and to realize.Above-mentioned each module can be embedded in the form of hardware or independently of in the processor in computer equipment, can also
Be stored in the memory in computer equipment in a software form, the above modules pair are executed in order to which processor calls
The operation answered.
In one embodiment, a kind of amount assessment device is provided, amount assessment device is commented with amount in above-described embodiment
Estimate method one-to-one correspondence.As shown in figure 9, amount assessment device includes that request module 101, pre-granted letter data determine mould
Block 102, the initial pre- accrediting amount obtain module 103 and the pre- accrediting amount of target obtains module 104.Each functional module is specifically
It is bright as follows:
Request module 101 checks request for obtaining the pre- accrediting amount, and the pre- accrediting amount checks to include using in request
Family attribute data and target user data, user attribute data include age of user, user location and user identifier.
Pre-granted letter data determining module 102 is used for according to pre-set pre- credit evaluation condition to age of user, user
Location and user identifier are identified determine whether user attribute data is pre-granted letter data.
The initial pre- accrediting amount obtains module 103, if being pre-granted letter data for user attribute data, by target user's number
According to being input in pre- accrediting amount model, the initial pre- accrediting amount corresponding with target user data is obtained.
The pre- accrediting amount of target obtains module 104, is used to inquire the table of comparisons according to the initial pre- accrediting amount, obtain and initial
The corresponding pre- accrediting amount of target of the pre- accrediting amount.
In one embodiment, a kind of computer equipment is provided, which can be server-side, internal junction
Composition can be as shown in Figure 10.The computer equipment includes processor, the memory, network interface connected by system bus
And database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The storage of the computer equipment
Device includes non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program
And database.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.
The database of the computer equipment is for storing original user data, specific fields table and consumption assessment table etc..The computer is set
Standby network interface is used to communicate with external terminal by network connection.To realize when the computer program is executed by processor
A kind of amount model training method；Alternatively, to realize a kind of amount appraisal procedure when the computer program is executed by processor.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory simultaneously
The computer program that can be run on a processor, processor realize amount model in above-described embodiment when executing computer program
The step of training method, for example, step S10 shown in Fig. 2 to step S60, alternatively, Fig. 3 is to step shown in fig. 6.Processing
Device realizes the function of each module in above-described embodiment in amount model training apparatus when executing computer program, for example, Fig. 8
Shown module 10 to module 60 function.To avoid repeating, details are not described herein again.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory simultaneously
The computer program that can be run on a processor, processor realize that amount is assessed in above-described embodiment when executing computer program
The step of method, for example, step S101 shown in Fig. 7 to step S104.Processor realizes above-mentioned reality when executing computer program
The function of applying each module in example in amount assessment device, for example, function of the module 101 shown in Fig. 9 to module 104.To avoid
It repeats, details are not described herein again.
In one embodiment, a kind of computer readable storage medium is provided, computer program, computer are stored thereon with
Amount model training method in above method embodiment is realized when program is executed by processor, for example, step S10 shown in Fig. 2
To step S60；Alternatively, Fig. 3 is to step shown in fig. 6.The computer program is realized in above-described embodiment when being executed by processor
The function of each module in amount model training apparatus, for example, function of the module 10 shown in Fig. 8 to module 60.To avoid repeating,
Details are not described herein again.
In one embodiment, a kind of computer readable storage medium is provided, computer program, computer are stored thereon with
Amount appraisal procedure in above method embodiment is realized when program is executed by processor, for example, step S101 shown in Fig. 7 is extremely
Step S104.Alternatively, the computer program realizes that amount assesses each module in device in above-described embodiment when being executed by processor
Function, for example, function of the module 101 shown in Fig. 9 to module 104.To avoid repeating, details are not described herein again.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, it is readable that computer program can be stored in a non-volatile computer
It takes in storage medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, this Shen
Please provided by any reference used in each embodiment to memory, storage, database or other media, can wrap
Include non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM in a variety of forms may be used
, such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM),
Enhanced SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (RambuS) are direct
RAM (RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
It is apparent to those skilled in the art that for convenience of description and succinctly, only with above-mentioned each function
Can unit, module division progress for example, in practical application, can according to need and by above-mentioned function distribution by difference
Functional unit, module complete, i.e., the internal structure of device is divided into different functional unit or module, with complete more than
The all or part of function of description.
The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although with reference to the foregoing embodiments
Invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each implementation
Technical solution documented by example is modified or equivalent replacement of some of the technical features；And these modification or
Replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all include
Within protection scope of the present invention.
Claims (10)
1. a kind of amount model training method characterized by comprising
Original user data in database is obtained, Screening Treatment is carried out to the original user data, acquisition meets training standard
Basic user data, the basic user data includes the accrediting amount, essential information data, basic assets data and disappears substantially
Take data；
The essential information data are pre-processed, division of life span's classification corresponding with the basic user data is obtained；
The basic assets data are pre-processed, loan repayment capacity grade corresponding with the basic user data is obtained；
The capital consumption data are pre-processed, consuming capacity grade corresponding with the basic user data is obtained；
The accrediting amount of the basic user data, division of life span's classification, loan repayment capacity grade and consuming capacity grade are carried out
Mark obtains the training data formed based on each basic user data；
Each training data is trained using GBDT algorithm, obtains pre- accrediting amount model.
2. amount model training method as described in claim 1, which is characterized in that original number of users in the acquisition database
According to original user data progress Screening Treatment, acquisition meets the basic user data of training standard, comprising:
Original user data in database is obtained, judges the original user data with the presence or absence of missing values；
If missing values are not present in the original user data, using the original user data as basic user data；
If there are missing values for the original user data, the corresponding field of the missing values is obtained；
If the field is specific fields, the processing of interpolation missing values is carried out to the missing values, obtains basic user data；
If the field is unspecified field, the missing values are not handled.
3. amount model training method as described in claim 1, which is characterized in that described to be carried out to the essential information data
Pretreatment obtains division of life span's classification corresponding with the basic user data, comprising:
Screening Treatment is carried out to the essential information data using screening rule relevant to division of life span's classification, obtains target letter
Cease data；
The object information data is identified using decision-tree model corresponding with division of life span's classification, obtains the base
The corresponding division of life span's classification of this user data.
4. amount model training method as described in claim 1, which is characterized in that described to be carried out to the basic assets data
Pretreatment obtains loan repayment capacity grade corresponding with the basic user data, comprising:
The basic assets data are normalized, the desired asset data after obtaining normalized；
Prediction processing is carried out to the desired asset data using preparatory trained loan repayment capacity prediction model, obtain with it is described
The corresponding loan repayment capacity grade of basic user data.
5. amount model training method as described in claim 1, which is characterized in that the capital consumption data include at least one
A consumer factors；
It is described that the capital consumption data are pre-processed, obtain consuming capacity corresponding with the basic user data etc.
Grade, comprising:
At least one described consumer factors is normalized, corresponding normalization factor value is obtained；
Processing is weighted at least one described normalization factor value, obtains corresponding target consumer value；
Consumption assessment table is inquired based on the target consumer value, obtains consuming capacity corresponding with the basic user data etc.
Grade.
6. a kind of amount appraisal procedure characterized by comprising
It obtains the pre- accrediting amount and checks request, the pre- accrediting amount checks to include user attribute data and target user in request
Data, the user attribute data include age of user, user location and user identifier；
According to pre-set pre- credit evaluation condition to the age of user, the user location and the user identifier
Identified determine whether the user attribute data is pre-granted letter data；
If the user attribute data is pre-granted letter data, the target user data is input to any one of claim 1 to 5
In the pre- accrediting amount model, the initial pre- accrediting amount corresponding with the target user data is obtained；
The table of comparisons is inquired according to the initial pre- accrediting amount, obtains the pre- credit of target corresponding with the initially pre- accrediting amount
Amount.
7. a kind of amount model training apparatus characterized by comprising
Basic user data obtains module, for obtaining original user data in database, carries out to the original user data
Screening Treatment obtains the basic user data for meeting training standard, and the basic user data includes the accrediting amount, essential information
Data, basic assets data and capital consumption data；
Division of life span's classification obtains module, for pre-processing to the essential information data, obtains and the elemental user
The corresponding division of life span's classification of data；
Loan repayment capacity grade obtains module, for pre-processing to the basic assets data, obtains and the elemental user
The corresponding loan repayment capacity grade of data；
Consuming capacity grade obtains module, for pre-processing to the capital consumption data, obtains and the elemental user
The corresponding consuming capacity grade of data；
Training data forms module, for the accrediting amount, division of life span's classification, the loan repayment capacity etc. to the basic user data
Grade and consuming capacity grade are labeled, and obtain the training data formed based on each basic user data；
Pre- accrediting amount model obtains module, for being trained using GBDT algorithm to each training data, obtains pre-
Accrediting amount model.
8. a kind of amount assesses device characterized by comprising
Request module checks request for obtaining the pre- accrediting amount, and the pre- accrediting amount checks to include user in request
Attribute data and target user data, the user attribute data include age of user, user location and user identifier；
Pre-granted letter data determining module is used for according to pre-set pre- credit evaluation condition to the age of user, the use
Family location and the user identifier are identified determine whether the user attribute data is pre-granted letter data；
The initial pre- accrediting amount obtains module, if being pre-granted letter data for the user attribute data, by the target user
Data are input in any one of claim 1 to 5 pre- accrediting amount model, are obtained corresponding with the target user data
The initial pre- accrediting amount；
The pre- accrediting amount acquisition module of target is used to inquire the table of comparisons according to the initial pre- accrediting amount, obtain and described first
The pre- accrediting amount of the corresponding target of the pre- accrediting amount that begins.
9. a kind of computer equipment, including memory, processor and storage are in the memory and can be in the processor
The computer program of upper operation, which is characterized in that the processor realized when executing the computer program as claim 1 to
The step of any one of 5 amount model training method；Alternatively, being realized when processor executes the computer program as right is wanted
The step of seeking the 6 amount appraisal procedure.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In realizing the amount model training method as described in any one of claim 1 to 5 when the computer program is executed by processor
Step；Alternatively, the computer program realizes the step of amount appraisal procedure as claimed in claim 6 when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910203514.XA CN110060144B (en) | 2019-03-18 | 2019-03-18 | Method for training credit model, method, device, equipment and medium for evaluating credit |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910203514.XA CN110060144B (en) | 2019-03-18 | 2019-03-18 | Method for training credit model, method, device, equipment and medium for evaluating credit |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110060144A true CN110060144A (en) | 2019-07-26 |
| CN110060144B CN110060144B (en) | 2024-01-30 |
Family
ID=67316279
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910203514.XA Active CN110060144B (en) | 2019-03-18 | 2019-03-18 | Method for training credit model, method, device, equipment and medium for evaluating credit |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110060144B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110659982A (en) * | 2019-09-27 | 2020-01-07 | 上海淇玥信息技术有限公司 | Method and device for large-volume business credit granting and electronic equipment |
| CN110689425A (en) * | 2019-09-30 | 2020-01-14 | 上海淇玥信息技术有限公司 | Method and device for pricing quota based on income and electronic equipment |
| CN110706096A (en) * | 2019-09-30 | 2020-01-17 | 上海淇玥信息技术有限公司 | Method and device for managing credit line based on salvage-back user and electronic equipment |
| CN110807527A (en) * | 2019-09-30 | 2020-02-18 | 北京淇瑀信息科技有限公司 | Line adjusting method and device based on guest group screening and electronic equipment |
| CN111275542A (en) * | 2020-01-17 | 2020-06-12 | 中国建设银行股份有限公司 | Loan request processing method, device and system |
| CN111353872A (en) * | 2019-12-20 | 2020-06-30 | 上海淇玥信息技术有限公司 | Credit granting processing method and device based on financial performance value and electronic equipment |
| CN111524001A (en) * | 2020-04-27 | 2020-08-11 | 中国银行股份有限公司 | Method and device for predicting user credit line and related equipment |
| CN111626883A (en) * | 2020-05-29 | 2020-09-04 | 上海商汤智能科技有限公司 | Authority verification method and device, electronic equipment and storage medium |
| CN112381314A (en) * | 2020-11-24 | 2021-02-19 | 京东数字科技控股股份有限公司 | Model training method, model training device, risk prediction method, risk prediction device, electronic equipment and storage medium |
| CN112561691A (en) * | 2020-12-24 | 2021-03-26 | 中国农业银行股份有限公司 | Customer credit prediction method, device, equipment and storage medium |
| CN112598500A (en) * | 2020-12-21 | 2021-04-02 | 中国建设银行股份有限公司 | Credit processing method and system for non-limit client |
| CN112950347A (en) * | 2021-02-01 | 2021-06-11 | 大箴(杭州)科技有限公司 | Resource data processing optimization method and device, storage medium and terminal |
| CN113516533A (en) * | 2021-06-24 | 2021-10-19 | 平安科技（深圳）有限公司 | Product recommendation method, device, equipment and medium based on improved BERT model |
| CN113781198A (en) * | 2020-06-09 | 2021-12-10 | 台北富邦商业银行股份有限公司 | Enterprise loan application evaluation system |
| CN112561691B (en) * | 2020-12-24 | 2024-07-05 | 中国农业银行股份有限公司 | Client trust prediction method, device, equipment and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160225073A1 (en) * | 2015-01-30 | 2016-08-04 | Wal-Mart Stores, Inc. | System, method, and non-transitory computer-readable storage media for predicting a customer's credit score |
| CN106373006A (en) * | 2016-09-07 | 2017-02-01 | 派生科技集团股份有限公司 | Method for evaluating user credit and repayment willingness through big-data modeling |
| CN107590737A (en) * | 2017-10-24 | 2018-01-16 | 厦门大学 | Personal credit scores and credit line measuring method |
| CN107633455A (en) * | 2017-09-04 | 2018-01-26 | 深圳市华傲数据技术有限公司 | Credit estimation method and device based on data model |
| CN107633030A (en) * | 2017-09-04 | 2018-01-26 | 深圳市华傲数据技术有限公司 | Credit estimation method and device based on data model |
| US20180232805A1 (en) * | 2016-06-12 | 2018-08-16 | Tencent Technology (Shenzhen) Company Limited | User credit rating method and apparatus, and storage medium |
| CN108460674A (en) * | 2018-02-01 | 2018-08-28 | 平安科技（深圳）有限公司 | Information processing method, device, computer equipment and storage medium |
| CN108596758A (en) * | 2018-05-03 | 2018-09-28 | 湖南大学 | A kind of credit rating method based on classification rule-based classification |
| CN108615191A (en) * | 2018-05-03 | 2018-10-02 | 湖南大学 | A kind of credit line intelligent evaluation method |
| CN108648074A (en) * | 2018-05-18 | 2018-10-12 | 深圳壹账通智能科技有限公司 | Loan valuation method, apparatus based on support vector machines and equipment |
| CN108898476A (en) * | 2018-06-14 | 2018-11-27 | 中国银行股份有限公司 | A kind of loan customer credit-graded approach and device |
-
2019
- 2019-03-18 CN CN201910203514.XA patent/CN110060144B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160225073A1 (en) * | 2015-01-30 | 2016-08-04 | Wal-Mart Stores, Inc. | System, method, and non-transitory computer-readable storage media for predicting a customer's credit score |
| US20180232805A1 (en) * | 2016-06-12 | 2018-08-16 | Tencent Technology (Shenzhen) Company Limited | User credit rating method and apparatus, and storage medium |
| CN106373006A (en) * | 2016-09-07 | 2017-02-01 | 派生科技集团股份有限公司 | Method for evaluating user credit and repayment willingness through big-data modeling |
| CN107633455A (en) * | 2017-09-04 | 2018-01-26 | 深圳市华傲数据技术有限公司 | Credit estimation method and device based on data model |
| CN107633030A (en) * | 2017-09-04 | 2018-01-26 | 深圳市华傲数据技术有限公司 | Credit estimation method and device based on data model |
| CN107590737A (en) * | 2017-10-24 | 2018-01-16 | 厦门大学 | Personal credit scores and credit line measuring method |
| CN108460674A (en) * | 2018-02-01 | 2018-08-28 | 平安科技（深圳）有限公司 | Information processing method, device, computer equipment and storage medium |
| CN108596758A (en) * | 2018-05-03 | 2018-09-28 | 湖南大学 | A kind of credit rating method based on classification rule-based classification |
| CN108615191A (en) * | 2018-05-03 | 2018-10-02 | 湖南大学 | A kind of credit line intelligent evaluation method |
| CN108648074A (en) * | 2018-05-18 | 2018-10-12 | 深圳壹账通智能科技有限公司 | Loan valuation method, apparatus based on support vector machines and equipment |
| CN108898476A (en) * | 2018-06-14 | 2018-11-27 | 中国银行股份有限公司 | A kind of loan customer credit-graded approach and device |
Non-Patent Citations (3)
| Title |
|---|
| WEI LI; JIBIAO LIAO: "An Empirical Study on Credit Scoring Model for Credit Card by Using Data Mining Technology", pages 1279 - 1282 * |
| 李刚;许传华;: "基于BP神经网络的个人信用评估体系研究", no. 01, pages 84 - 86 * |
| 王齐齐;: "个人信用卡发展现状及信用风险评估模型分析", no. 02, pages 28 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110659982A (en) * | 2019-09-27 | 2020-01-07 | 上海淇玥信息技术有限公司 | Method and device for large-volume business credit granting and electronic equipment |
| CN110689425A (en) * | 2019-09-30 | 2020-01-14 | 上海淇玥信息技术有限公司 | Method and device for pricing quota based on income and electronic equipment |
| CN110706096A (en) * | 2019-09-30 | 2020-01-17 | 上海淇玥信息技术有限公司 | Method and device for managing credit line based on salvage-back user and electronic equipment |
| CN110807527A (en) * | 2019-09-30 | 2020-02-18 | 北京淇瑀信息科技有限公司 | Line adjusting method and device based on guest group screening and electronic equipment |
| CN110807527B (en) * | 2019-09-30 | 2023-11-14 | 北京淇瑀信息科技有限公司 | Credit adjustment method and device based on guest group screening and electronic equipment |
| CN111353872A (en) * | 2019-12-20 | 2020-06-30 | 上海淇玥信息技术有限公司 | Credit granting processing method and device based on financial performance value and electronic equipment |
| CN111275542A (en) * | 2020-01-17 | 2020-06-12 | 中国建设银行股份有限公司 | Loan request processing method, device and system |
| CN111524001A (en) * | 2020-04-27 | 2020-08-11 | 中国银行股份有限公司 | Method and device for predicting user credit line and related equipment |
| CN111524001B (en) * | 2020-04-27 | 2024-02-02 | 中国银行股份有限公司 | User credit limit prediction method, device and related equipment |
| CN111626883A (en) * | 2020-05-29 | 2020-09-04 | 上海商汤智能科技有限公司 | Authority verification method and device, electronic equipment and storage medium |
| CN113781198A (en) * | 2020-06-09 | 2021-12-10 | 台北富邦商业银行股份有限公司 | Enterprise loan application evaluation system |
| CN112381314A (en) * | 2020-11-24 | 2021-02-19 | 京东数字科技控股股份有限公司 | Model training method, model training device, risk prediction method, risk prediction device, electronic equipment and storage medium |
| CN112598500A (en) * | 2020-12-21 | 2021-04-02 | 中国建设银行股份有限公司 | Credit processing method and system for non-limit client |
| CN112561691A (en) * | 2020-12-24 | 2021-03-26 | 中国农业银行股份有限公司 | Customer credit prediction method, device, equipment and storage medium |
| CN112561691B (en) * | 2020-12-24 | 2024-07-05 | 中国农业银行股份有限公司 | Client trust prediction method, device, equipment and storage medium |
| CN112950347A (en) * | 2021-02-01 | 2021-06-11 | 大箴(杭州)科技有限公司 | Resource data processing optimization method and device, storage medium and terminal |
| CN113516533A (en) * | 2021-06-24 | 2021-10-19 | 平安科技（深圳）有限公司 | Product recommendation method, device, equipment and medium based on improved BERT model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110060144B (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110060144A (en) | Amount model training method, amount appraisal procedure, device, equipment and medium | |
| CN107066616A (en) | Method, device and electronic equipment for account processing | |
| CN108648074A (en) | Loan valuation method, apparatus based on support vector machines and equipment | |
| CN110570111A (en) | Enterprise risk prediction method, model training method, device and equipment | |
| CN109858957A (en) | Finance product recommended method, device, computer equipment and storage medium | |
| CN108694673A (en) | A kind of processing method, device and the processing equipment of insurance business risk profile | |
| CN104679743A (en) | Method and device for determining preference model of user | |
| CN107230108A (en) | The processing method and processing device of business datum | |
| CN110020862A (en) | A kind of business risk appraisal procedure, device and computer readable storage medium | |
| CN109583966A (en) | A kind of high value customer recognition methods, system, equipment and storage medium | |
| CN116523262B (en) | Intelligent planning method, system and medium for production plan based on big data | |
| CN108182633A (en) | Loan data processing method, device, computer equipment and storage medium | |
| CN111489259A (en) | Stock market risk prediction intelligent implementation method based on deep learning | |
| CN116645132A (en) | Multi-factor variable-based time sequence prediction method and device, electronic equipment and medium | |
| Firdaus et al. | development of bank’s customer segmentation model based on rfm+ b approach | |
| CN112508734B (en) | Method and device for predicting power generation capacity of power enterprise based on convolutional neural network | |
| CN110310012A (en) | Data analysing method, device, equipment and computer readable storage medium | |
| CN107424026A (en) | Businessman's reputation evaluation method and device | |
| CN116523301A (en) | System for predicting risk rating based on big data of electronic commerce | |
| CN114926261A (en) | Method and medium for predicting fraud probability of automobile financial user application | |
| CN112884028A (en) | System resource adjusting method, device and equipment | |
| Mohammed et al. | Soft set decision/forecasting system based on hybrid parameter reduction algorithm | |
| CN110246026A (en) | A kind of output combination setting method, device and the terminal device of data transfer | |
| CN116954591B (en) | Generalized linear model training method, device, equipment and medium in banking field | |
| Brešić | Knowledge acquisition in databases |
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 |