Specific embodiment
In the following description, in order to make the reader understand this application better, many technical details are proposed.But this
The those of ordinary skill in field is appreciated that even if without these technical details and many variations based on the following respective embodiments
And modification, the application technical solution claimed also may be implemented.
Conceptual illustration:
It first enjoys and pays (Non-sufficient Fund, abbreviation NSF) afterwards: based on user credit grade, giving user and enjoy first
Service, the means of payment then withholdd again promote convenience and user experience.
Implementation to keep the purposes, technical schemes and advantages of the application clearer, below in conjunction with attached drawing to the application
Mode is described in further detail.
The first embodiment of the application is related to a kind of credit risk method of discrimination based on transaction.Fig. 1 is that this is based on handing over
The flow diagram of easy credit risk method of discrimination.
Specifically, as shown in Figure 1, should credit risk method of discrimination based on transaction the following steps are included:
In a step 101, obtain multiple users of credit payment platform at the appointed time the transaction history information in section and
Account liveness information.
In the present embodiment, designated time period can be set according to actual needs, for example, can be nearest one week,
Nearest one month, nearest half a year or nearest 1 year etc..
Preferably, the transaction history information includes: that historical trading number, the historical trading amount of money, history are first enjoyed and being paid into afterwards
Power and history are first enjoyed pays rate of violation afterwards.
The account liveness information include: Account Logon information, account aging, account balance, account expenditure information and
Account takes in information.
Then into step 102, the machine for having supervision is carried out according to the transaction history information and account liveness information
Learning model building obtains transaction credit risk model.
Preferably, in a step 102, it is carried out using the machine learning algorithm for having supervision of such as random forest, GBDT etc.
The modeling of transaction credit risk model.
By transaction history information to account, first enjoys and pay payment success rate, account aging, account Assets etc. afterwards and become
Amount information is excavated, and carries out transaction credit risk mould using the Supervised machine learning algorithm of such as random forest, GBDT etc.
Type modeling excavates transaction history information and (is referred to as: behavior liveness) information with account liveness and first enjoys to pay letter afterwards
Accurate wind is carried out to pay the credit situation under transaction scene after formerly enjoying to each user account with the corresponding relationship of situation
Danger judgement.
In step 101 and step 102, solution is how to extract Modelling feature to carry out the modeling of transaction credit risk model
The problem of.
For example, in step 101 and 102, using user's dimension as major key, according to it on credit payment platform
Specific behavior feature and transaction history data are as machine learning Modelling feature, such as extract user nearest one week or one nearest
Transaction count, transaction amount and specific channel such as credit card, flower expenditure behavioural habits in month, believe substantially in conjunction with individual subscriber
The static natures such as breath generate the wide table of behavior transaction based on user.Then there is intendant using such as random forest, GBDT etc.
Device learning algorithm learns the trading activity feature of user, after building transaction credit risk model prediction first the enjoying of user
Pay whether transaction can honour an agreement.
It should be noted that it will be apparent to those skilled in the art that the machine learning modeling for having supervision is in the prior art
Mature technology, be not unfolded yet further herein.
Then into step 103, target user first enjoyed pay afterwards transaction when, obtain the target user in the letter
With the transaction history information and account liveness information in the designated time period of payment platform, and input the transaction credit risk
Model differentiates the transaction credit risk of the target user.
It should be noted that in step 103, described first enjoy pays transaction afterwards and can be the target user in the credit
Transaction is paid in first enjoying for carrying out on payment platform afterwards.Transaction is paid in first enjoying for carrying out on the credit payment platform afterwards, is directly acquired
Transaction history information and account liveness information of the target user in the designated time period of the credit payment platform, so
The differentiation of transaction credit risk is carried out using the transaction credit risk model afterwards.
Alternatively, optionally, in step 103, described first enjoy pays transaction afterwards to can be the target user flat in third party
Transaction is paid in first enjoying for platform progress afterwards.In this case, the third-party platform obtains the target user in the credit branch
The transaction history information and account liveness information in the designated time period of platform are paid, the transaction credit risk mould is then inputted
Type differentiates the transaction credit risk of the target user.
That is, the transaction credit risk model based on the credit payment platform construction, is applicable not only to described
Transaction is paid in first enjoying for carrying out on credit payment platform afterwards, and first enjoying of being also applied for carrying out on the third-party platform is paid afterwards
Transaction, the transaction credit risk model can be suitable for multiple platforms, substantially increase the transaction credit risk model
Universality.
In particular, this point is particularly useful for the third-party platform for not having transaction credit risk discriminating power,
Third-party platform can call the target user to believe in the transaction history information and account liveness of the credit payment platform
Breath, then inputs the transaction credit risk model, differentiates to the transaction credit risk of the target user.
In the present embodiment, it is preferable that the third-party platform obtains institute according to the phone number of the target user
State transaction history information and account liveness information of the target user in the designated time period of the credit payment platform.
Specifically, the target user can be searched in the credit payment according to the phone number of the target user
With the account of phone number binding on platform, the target user then is transferred in the credit branch further according to the account
The transaction history information and account liveness information for paying platform carry out the differentiation of transaction credit risk.
Certainly, in the other embodiments of the application, can also according to other identity informations of the target user come
The target user is searched in the account information of the credit payment platform, as long as the identity information and the account have one
One corresponding relationship, is not limited in phone number.
The problem of what step 103 solved is how to judge credit risk using transaction credit risk model automatic identification, tool
It says to body, may include following sub-step in step 103:
Calculate the transaction credit point of the target user;
It determines whether that the target user provides according to the transaction credit point and first enjoys the service of paying afterwards.
It should be noted that calculate the transaction credit point of target user using transaction credit risk model, can there are two types of
Mode:
First method is target user first enjoyed pay transaction afterwards when, obtained the target user at that time when specified
Between transaction history information and account liveness information in section, and the transaction credit risk model is inputted, on the spot to the mesh
The transaction credit risk for marking user carries out Quantitative marking, the transaction credit point of the target user is obtained, then further according to the friendship
Easy credit score judges the transaction credit risk of target user.
Second method is after the building for completing transaction credit risk model, and it is daily that system generates self-timing task
Timing carries out Quantitative marking to the transaction credit risk of all users, obtains the transaction credit point of all users.Work as target user
When first enjoying at progress one and pay transaction afterwards, the transaction credit point of the daily newest scoring of the target user will be transferred, then
The transaction credit risk of the target user is judged further according to the transaction credit point.
It, can also be there are two types of mode when judging the transaction credit risk of target user according to transaction credit point:
First, the transaction credit of the target user point can be compared with preset threshold value: when being higher than threshold value, mesh
Mark user, which can enjoy, first enjoys the service of paying afterwards;When lower than threshold value, target user, which cannot enjoy, first enjoys the service of paying afterwards.
Second, the configuration of expertise rule and policy can also be carried out, the transaction is generated according to the transaction credit point
The corresponding air control strategy of credit score, the friendship then further according to the transaction credit point and corresponding air control strategy, to target user
Easy credit risk differentiated, decides whether can to allow the target user using first enjoying the service of paying afterwards.
The above-mentioned second way, by by transaction credit point and expertise carry out it is secondary merge, based on transaction credit wind
The transaction credit point that dangerous model exports carries out the configuration of expertise rule and policy, first enjoys rear Fu Ce for user's formulation personalization
Slightly, it so as to have merged expertise interpretation high, the characteristics of strong flexibility, is used convenient for business, while expert can also be passed through
The configuration of empirical rule is so that the solution can be applied to multiple industry scenes with universality.
Further, it is preferable to ground, in step 103, when the transaction credit risk to the target user differentiates,
Simultaneously according to the terminal device of the target user (for example, can be according to the mobile device identifier or cell-phone number of terminal device
Code etc. judges the whether legal necessary being of the terminal user), environment locating for the target user is (for example, according to the IP of user
Address, network environment etc. are judged), the account state of the target user (for example, the balance status of account, account whether
Freeze), the account value (for example, remaining sum of account and historical trading volume etc.) of the target user and trade order whether
Conflict (for example, can be worth according to the order history and History Order of user, judge the transaction hobby and trading capacity of user,
To judge this trade order with the presence or absence of conflict) carry out the differentiation of transaction credit risk.
Hereafter terminate this process.
It should be noted that the credit payment platform includes but is not limited to pay in each embodiment of the application
Treasured, correspondingly, the account include but is not limited to Alipay account.
Transaction credit risk model is constructed using advanced machine learning algorithm, it is automatically right using transaction credit risk model
The transaction credit risk of user differentiates, first enjoys rear Fu Hangwei so as to effectively intercept low credit crowd, promotion is first enjoyed
Pay afterwards transaction scene under bad debt risk recognition capability, have higher recognition accuracy and it is wider array of use coverage rate.
In order to more fully understand the technical solution of the application, it is illustrated below with reference to a preferred embodiment,
The details enumerated in the preferred embodiment is primarily to be easy to understand, not as the limitation to the application protection scope.
Below by the preferred embodiment, it is described in detail how to differentiate NSF risk, and letter using transaction credit risk model
Illustrate the building principle of the transaction credit risk model.Fig. 2 is the NSF flow diagram of the preferred embodiment.Fig. 3 is the NSF
The strategy system figure of solution.
1. the modeling of transaction credit risk model.
Transaction credit risk model will be based primarily upon transactions history of the user in system of account and construct.The change that model uses
Amount specifically includes that user transaction history information and account liveness.User transaction history: including in user's nearest a period of time
Transaction, expenditure, income information and relevant historical first enjoy and pay success rate and rate of violation information etc. afterwards.Account liveness: including with
The static informations such as all kinds of logins, aging, remaining sum of the family in system of account, reflection account use active degree.Become based on above
Information is measured, using there is the machine learning algorithm of supervision to be modeled, friendship can be embodied by excavating in transactions history and behavior liveness
The variable of easy risk credit, and realize that transaction credit is sentenced based on the Supervised machine learning algorithm of such as random forest, GBDT etc.
Not.
It first enjoys and is paid in scene afterwards using the transaction credit strategy system based on transaction credit risk model, base in NSF
Credit risk is portrayed in account historical transaction dimension, compared to the variable dimension that existing credit score in the market simplifies model,
To allow model to focus more on covering transaction credit risk scene, lift scheme applicability and ease of use, promoted first
The bad debt risk recognition capability under transaction scene is paid after enjoying.
2. utilizing transaction credit risk model, the transaction credit risk of automatic identification user.
As shown in Fig. 2, entire air control process is classified into four layers in scene service application flow, it is source data input first,
It is calculated subsequently into transaction credit risk model, the transaction credit point after completing model point output, then by strategic layer based on user
Configuration personalization is first enjoyed pays strategy afterwards.In actual scene, after user creates order first, system will call the mistake of the user immediately
Situation is enlivened toward transactions history and account, calculates the friendship of this of user order using transaction credit risk model based on information above
Easy credit risk, and different user group's formulation personalizations is divided into according to the transaction credit of user and first enjoys the service plan paid afterwards
Slightly, it ensures and minimizes bad debt risk while user experience.After completing the calculating of transaction credit point and passing through air control strategy, user
It can enter first to enjoy and pay afterwards, enjoy first service and then be automatically performed again by system and withholdd.
3. as shown in figure 3, there are mainly two types of modes for NSF credit risk solution in NSF solution system entirety:
(1) for the target user for using Alipay account, exempt from the close scene withheld for signing, with the Alipay of target user
Account ID first passes around the safe UCT risk engine of Alipay as main body, it is ensured that link of contracting it is secure and trusted.Later in mesh
Single link under user is marked, trade company can carry out risk consulting to Alipay, and NSF solution can be using transaction credit risk model as core
The heart, while from equipment, environment, user, account, the big dimension progress NSF credit risk strategy differentiation of conflict five, and return to risk etc.
Grade judges bad debt risk for trade company.(2) scene of Alipay account is not used, for third-party platform with the mobile phone of target user
Based on number, when target user merchant end initiation place an order request when, trade company according to phone number carry out risk consulting, NSF solution
Certainly scheme searches corresponding binding Alipay account or common trusted payment treasured account according to phone number, and according to its transaction
Credit Risk Model scoring carries out NSF risk differentiation to it, and returns to risk class and judge bad debt risk for trade company.
4. subsequent returned money traces system.For the NSF bad credit occurred, which is also provided that subsequent returned money chases after
Trace back ability, based on trading activity equally by target user in Alipay account, pass through the remaining sum to account change, fund
Transaction expenditure is monitored, and may be implemented when assets change, and reminds collection, or the side directly dynamically to withhold by short message
Formula helps the quick returned money of trade company.It realizes intelligent collection prompting and policy of dynamically withholing, provides NSF bad credit subsequent returned money retrospect energy
Power helps trade company utmostly to reduce loss on bad debt.
As can be seen that on the one hand the application constructs transaction credit risk model using advanced machine learning algorithm, utilize
Algorithm model identifies that the method will be significantly better than in the accuracy rate and coverage rate of identification to the credit quality of user automatically
Tradition relies on the regular identification method of expertise.On the other hand, transaction credit risk is exported in machine learning algorithm model
/ after, model result can be subjected to the secondary credit score progress expertise rule merged, exported based on model with expertise
Then tactful configuration the characteristics of strong flexibility, is used convenient for business so that it is high to have merged expertise interpretation, while
It can be by the configuration of expertise rule so that the solution can be applied to multiple industry scenes with universality.
Each method embodiment of the invention can be realized in a manner of software, hardware, firmware etc..Regardless of the present invention be with
Software, hardware or firmware mode realize that instruction code may be stored in any kind of computer-accessible memory
In (such as permanent perhaps revisable volatibility is perhaps non-volatile solid or non-solid, it is fixed or
The replaceable medium etc. of person).Equally, memory may, for example, be programmable logic array (Programmable Array
Logic, referred to as " PAL "), random access memory (Random Access Memory, referred to as " RAM "), it may be programmed read-only deposit
Reservoir (Programmable Read Only Memory, referred to as " PROM "), read-only memory (Read-Only Memory, letter
Claim " ROM "), electrically erasable programmable read-only memory (Electrically Erasable Programmable ROM, referred to as
" EEPROM "), disk, CD, digital versatile disc (Digital Versatile Disc, referred to as " DVD ") etc..
The second embodiment of the application is related to a kind of credit risk discriminating gear based on transaction.Fig. 4 is that this is based on handing over
The structural schematic diagram of easy credit risk discriminating gear.
Specifically, as shown in figure 4, being somebody's turn to do the credit risk discriminating gear based on transaction, comprising:
Obtain module, for obtain transaction history information at the appointed time in section of multiple users of credit payment platform and
Account liveness information.
Preferably,
The transaction history information includes: that historical trading number, the historical trading amount of money, history first enjoy and pays success rate afterwards and go through
History first enjoys and pays rate of violation afterwards.
The account liveness information include: Account Logon information, account aging, account balance, account expenditure information and
Account takes in information.
Modeling module, for carrying out the machine learning for having supervision according to the transaction history information and account liveness information
Modeling, obtains transaction credit risk model.
Discrimination module, for target user first enjoyed pay afterwards transaction when, obtain the target user in the credit
Transaction history information and account liveness information in the designated time period of payment platform, and input the transaction credit risk mould
Type differentiates the transaction credit risk of the target user.
In the present embodiment, it is described first enjoy pay afterwards transaction can be the target user on the credit payment platform
Transaction is paid in first enjoying for carrying out afterwards.Transaction is paid in first enjoying for carrying out on the credit payment platform afterwards, is directly acquired the target and is used
Transaction history information and account liveness information of the family in the designated time period of the credit payment platform, then using described
Transaction credit risk model carries out the differentiation of transaction credit risk.
Alternatively, optionally, it is described first enjoy paying transaction afterwards and can be the target user first enjoyed what third-party platform carried out
After pay transaction.In this case, the third-party platform obtains the target user in the specified of the credit payment platform
Transaction history information and account liveness information in period, then input the transaction credit risk model, to the mesh
The transaction credit risk of mark user differentiates.
That is, the transaction credit risk model based on the credit payment platform construction, is applicable not only to described
Transaction is paid in first enjoying for carrying out on credit payment platform afterwards, and first enjoying of being also applied for carrying out on the third-party platform is paid afterwards
Transaction, the transaction credit risk model can be suitable for multiple platforms, substantially increase the transaction credit risk model
Universality.
In particular, this point is particularly useful for the third-party platform for not having transaction credit risk discriminating power,
Third-party platform can call the target user to believe in the transaction history information and account liveness of the credit payment platform
Breath, then inputs the transaction credit risk model, differentiates to the transaction credit risk of the target user.
In the present embodiment, it is preferable that the third-party platform obtains institute according to the phone number of the target user
State transaction history information and account liveness information of the target user in the designated time period of the credit payment platform.
Specifically, the target user can be searched in the credit payment according to the phone number of the target user
With the account of phone number binding on platform, the target user then is transferred in the credit branch further according to the account
The transaction history information and account liveness information for paying platform carry out the differentiation of transaction credit risk.
Certainly, in the other embodiments of the application, can also according to other identity informations of the target user come
The target user is searched in the account information of the credit payment platform, as long as the identity information and the account have one
One corresponding relationship, is not limited in phone number.
Further, the discrimination module, comprising:
Computational submodule, for calculating the transaction credit point of the target user;
It determines submodule, pays clothes afterwards for determining whether that the target user provides first to enjoy according to the transaction credit point
Business.
The discrimination module is when the transaction credit risk to the target user differentiates, while according to described
Environment locating for the terminal device of target user, the target user, the account state of the target user, the target user
Account value and trade order whether conflict and carry out transaction credit risk differentiation.
It should be noted that the credit payment platform includes but is not limited to pay in each embodiment of the application
Treasured, correspondingly, the account include but is not limited to Alipay account.
Transaction credit risk model is constructed by advanced machine learning, using transaction credit risk model automatically to user
Transaction credit risk differentiated, first enjoy rear Fu Hangwei so as to effectively intercept low credit crowd, promotion is first enjoyed pays afterwards
Trade scene under bad debt risk recognition capability, have higher recognition accuracy and it is wider array of use coverage rate.
First embodiment is method implementation corresponding with present embodiment, and the technology in first embodiment is thin
Section can be applied to present embodiment, and the technical detail in present embodiment also can be applied to first embodiment.
It should be noted that it will be appreciated by those skilled in the art that the above-mentioned credit risk discriminating gear based on transaction
The correlation that the realization function of each module shown in embodiment can refer to the aforementioned credit risk method of discrimination based on transaction is retouched
It states and understands.The function of each module shown in the embodiment of the above-mentioned credit risk discriminating gear based on transaction can pass through fortune
Row is realized in the program (executable instruction) on processor, can also be realized by specific logic circuit.The application is implemented
The above-mentioned credit risk discriminating gear based on transaction is realized in the form of software function module and as independent product in if
When selling or using, it also can store in a computer readable storage medium.Based on this understanding, the application is implemented
Substantially the part that contributes to existing technology can be embodied in the form of software products the technical solution of example in other words,
The computer software product is stored in a storage medium, including some instructions are used so that computer equipment (can be with
Personal computer, server or network equipment etc.) execute each embodiment the method for the application all or part.
And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read Only Memory), magnetic disk or light
The various media that can store program code such as disk.In this way, the embodiment of the present application is not limited to any specific hardware and software
In conjunction with.In addition, in order to protrude innovative part of the invention, the above-mentioned each equipment embodiment of the present invention will not be with this hair of solution
The technical issues of bright proposed, the less close module of relationship introduced, this does not indicate above equipment embodiment and there is no it
Its module.
Correspondingly, the application embodiment also discloses a kind of credit risk discriminating device based on transaction, comprising:
Memory, for storing computer executable instructions;And
Processor, in each method embodiment for realizing the application when executing the computer executable instructions
Step.
Correspondingly, the application embodiment also discloses a kind of computer storage medium, wherein being stored with computer can hold
Row instruction, the computer executable instructions realize the step in each method embodiment of the application when being executed by processor.
It should be noted that relational terms such as first and second and the like are only in the application documents of this patent
For distinguishing one entity or operation from another entity or operation, without necessarily requiring or implying these entities
Or there are any actual relationship or orders between operation.Moreover, the terms "include", "comprise" or its any other
Variant is intended to non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only
It including those elements, but also including other elements that are not explicitly listed, or further include for this process, method, object
Product or the intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence " including one ", not
There is also other identical elements in the process, method, article or apparatus that includes the element for exclusion.The application of this patent
In file, if it is mentioned that certain behavior is executed according to certain element, then refers to the meaning for executing the behavior according at least to the element, wherein
Include two kinds of situations: executing the behavior according only to the element and the behavior is executed according to the element and other elements.Multiple,
Repeatedly, the expression such as a variety of include 2,2 times, 2 kinds and 2 or more, 2 times or more, two or more.
It is included in disclosure of this application with being considered as globality in all documents that the application refers to, so as to
It can be used as the foundation of modification if necessary.In addition, it should also be understood that, after having read the above disclosure of the application, this field
Technical staff can make various changes or modifications the application, and such equivalent forms equally fall within the application model claimed
It encloses.