CN106779272A - A kind of Risk Forecast Method and equipment - Google Patents
A kind of Risk Forecast Method and equipment Download PDFInfo
- Publication number
- CN106779272A CN106779272A CN201510825234.4A CN201510825234A CN106779272A CN 106779272 A CN106779272 A CN 106779272A CN 201510825234 A CN201510825234 A CN 201510825234A CN 106779272 A CN106779272 A CN 106779272A
- Authority
- CN
- China
- Prior art keywords
- resource
- training sample
- overdue
- release
- stock number
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
Abstract
This application discloses a kind of Risk Forecast Method and equipment, including：Obtain and pending user-related resource data；According to the described overdue time included in the resource data and the stock number of the overdue release resource, the characteristic value of the resource credit rating size for characterizing the pending user is calculated；According to the characteristic value, predict that the pending user uses the risk for distributing resource.By two angles of stock number from overdue time and overdue release resource, there is overdue degree using resource is distributed in comprehensive descision user, and then determine the resource credit rating of user, relatively accurate resource credit rating can so be obtained, and can according to obtain resource credit rating accurately predict user existing for risk, resource service platform is user resource allocation according to the resource credit rating being calculated simultaneously, the risk index of resource service platform can be effectively reduced, and then lifts the circulation efficiency of resource service platform resource.
Description
Technical field
The application is related to internet information processing technology field, more particularly to a kind of Risk Forecast Method and sets
It is standby.
Background technology
With developing rapidly for science and technology, various resource service platforms are occurred in that, for example：Resource
Shared platform, resource storage platform etc..These resource service platforms can be user point according to the demand of user
With resource so that user performs various business using the resource for obtaining, and is very easy to user's
Daily life.
In order to preferably provide the user resource service, resource service platform can be according to user in resource service
The user behavior data produced on platform determines the Resource Properties of the user, and true according to the Resource Properties of user
It is set to the quantity of user resource allocation, the quantity of resource characterizes user and can be obtained from resource service platform here
The resource being free to arrange by user number, it is generally the case that the Resource Properties of user are better, network clothes
Business business is that the resource quantity of user's distribution is more.
Resource service platform will produce a resource useful life when to user resource allocation, it means that
User can use the resource when the resource is got in resource useful life, and in resource validity period
Limit needs to discharge the resource when expiring.If user discharges the resource when resource useful life expires, then
Resource service platform will improve the Resource Properties of user, in order to corresponding subsequently when for the user resource allocation
Ground increases the resource quantity for distributing to the user；If user does not discharge the money when resource useful life expires
Source, illustrates there is the situation released after the sentence expires when in use for the resource user for distributing, then resource service is put down
Platform will reduce the Resource Properties of user when the situation generation for releasing resource after the sentence expires occurs in user, and be subsequently
The resource quantity for distributing to the user is correspondingly reduced during the user resource allocation.
Resource service platform distributes resource by user resource allocation quantity, it is necessary to use user for convenience
Risk be predicted, and according to the risk of user that prediction is obtained for user distributes corresponding resource.It is existing
The method of prediction consumer's risk is in technology：According to user discharge resource during produce it is overdue when
Between, determine the resource credit rating of user, and be that user distributes corresponding resource according to resource credit rating, here
Described resource credit rating refers to the probability that user can on time discharge distributed resource.
It has been investigated that, the resource credit rating of user determined by aforesaid way and the resource credit of actual user
Degree has differences, and the resource credit rating for determining in the manner described above is that user distributes corresponding resource, will be caused
The resource risk index of resource service platform is raised, and then influences the resource flow transfer efficient of resource service platform.
The content of the invention
In view of this, the embodiment of the present application provides a kind of Risk Forecast Method and equipment, existing for solving
The resource risk index of resource service platform is higher in technology and then influences the circulation of the resource of resource service platform
The problem of efficiency.
This application provides a kind of Risk Forecast Method, including：
Acquisition and pending user-related resource data, wherein, comprising for characterizing in the resource data
The pending user occur releasing after the sentence expires when using the resource for getting the resource the overdue time and
Occur releasing the resource after the sentence expires when using the resource for getting for characterizing the pending user
Overdue release resource stock number；
According to the described overdue time included in the resource data and the stock number of the overdue release resource,
It is calculated the characteristic value of the resource credit rating size for characterizing the pending user；
According to the characteristic value, predict that the pending user uses the risk for distributing resource.
This application provides a kind of risk profile equipment, including：
Acquiring unit, for obtain with pending user-related resource data, wherein, the resource data
In comprising occurring releasing the money after the sentence expires when using the resource for getting for characterizing the pending user
The overdue time in source and occur prolonging when using the resource for getting for characterizing the pending user
Phase discharges the stock number of the overdue release resource of the resource；
Computing unit, for according to the described overdue time included in the resource data and the overdue release
The stock number of resource, is calculated the feature of the resource credit rating size for characterizing the pending user
Value；
Predicting unit, resource is distributed for according to the characteristic value, predicting that the pending user uses
Risk.
The application has the beneficial effect that：
The embodiment of the present application acquisition and pending user-related resource data, comprising use in the resource data
Occur releasing the overdue of the resource after the sentence expires when using the resource for getting in the pending user is characterized
Time and occur releasing institute after the sentence expires when using the resource for getting for characterizing the pending user
State the stock number of the overdue release resource of resource；According to described overdue time for being included in the resource data and
The stock number of the overdue release resource, is calculated the resource credit rating for characterizing the pending user
The characteristic value of size；According to the characteristic value, predict that the pending user uses the risk for distributing resource.
By two angles of stock number from overdue time and overdue release resource, comprehensive descision user is using being distributed
The overdue degree of resource generation, and then determine the resource credit rating of user, can so obtain relatively accurate
Resource credit rating, and can according to obtain resource credit rating accurately predict user existing for risk, while money
Source service platform is user resource allocation according to the resource credit rating being calculated, and can effectively reduce resource clothes
The risk index of business platform, and then lift the circulation efficiency of resource service platform resource.
Brief description of the drawings
In order to illustrate more clearly of the technical scheme in the embodiment of the present application, institute in being described to embodiment below
The accompanying drawing for needing to use is briefly introduced, it should be apparent that, drawings in the following description are only the application's
Some embodiments, for one of ordinary skill in the art, are not paying the premise of creative labor
Under, other accompanying drawings can also be obtained according to these accompanying drawings.
A kind of Risk Forecast Method schematic flow sheet that Fig. 1 is provided for the embodiment of the present application；
A kind of risk profile device structure schematic diagram that Fig. 2 is provided for the embodiment of the present application.
Specific embodiment
A kind of Risk Forecast Method is provided in order to realize the purpose of the application, in the embodiment of the present application and is set
It is standby, obtain and pending user-related resource data, comprising for characterizing described treating in the resource data
There is the overdue time for releasing the resource after the sentence expires and for table when using the resource for getting in treatment user
The pending user is levied to occur releasing the overdue of the resource after the sentence expires when using the resource for getting
Discharge the stock number of resource；According to the described overdue time included in the resource data and the overdue release
The stock number of resource, is calculated the feature of the resource credit rating size for characterizing the pending user
Value；According to the characteristic value, predict that the pending user uses the risk for distributing resource.
By two angles of stock number from overdue time and overdue release resource, comprehensive descision user uses institute
The overdue degree of distribution resource generation, and then determine the resource credit rating of user, can so obtain essence relatively
True resource credit rating, and can according to obtain resource credit rating accurately predict user existing for risk, together
When resource service platform be user resource allocation according to the resource credit rating that is calculated, can effectively reduce money
The risk index of source service platform, and then lift the circulation efficiency of resource service platform resource.
It should be noted that the described overdue time in the embodiment of the present application refers to user being obtained in release
During the resource for arriving, relative to the time that the resource useful life that resource service platform is specified is exceeded；It is overdue to release
The stock number for putting resource refers to user when the resource useful life that resource service platform is specified is reached, for obtaining
The resource got, releases the stock number of resource after the sentence expires.
For example：Resource service platform is that user A distribution stock numbers are the resource of a, the resource validity period specified
It is limited to n days, when reaching the time limits of n days, the stock number of user A releases is b (b ＜ a), and in n
Release surplus yield (a-b) in m days after it, then user A can be determined for the money distributed
There is overdue phenomenon in source, wherein, the overdue time is (m-n), and the stock number of overdue release resource is (a-b).
If user discharges the resource for getting completely in specified resource useful life, illustrate that user makes
Do not occur delaying during with the resource for getting, the overdue time of user is 0, the stock number of overdue release resource
Also it is 0.
It should be noted that user once occurs overdue phenomenon when distributed resource is used, either
Overdue time or overdue release resource, illustrate that the resource credit of user is relatively poor, for the resource distributed
There is high risk；User does not occur overdue phenomenon when distributed resource is used, and illustrates the money of user
Source credit is relatively preferable, and the resource for distributing has relatively low risk.In other words, the resource letter of user
Expenditure is higher, and the risk produced by user is lower；Conversely, the resource credit rating of user is smaller, user is produced
Raw risk is higher.
There is overdue situation when using the resource for getting for user, can by user it is overdue when
Between to weigh user, using distributing, the risk of resource, i.e., overdue time are more long, the risk produced by user is got over
It is high；User can also be weighed by the stock number of the overdue release resource of user and uses the wind for distributing resource
Danger, i.e., the stock number of overdue release resource is more, and the risk produced by user is higher.
In the embodiment of the present application, two granularities pair of stock number from overdue time and overdue release resource are proposed
Risk produced by user is predicted, but is also not necessarily limited to the two granularities, it is also possible to enter from multiple granularities
Row analysis.
Each embodiment of the application is described in further detail with reference to Figure of description.Obviously,
Described embodiment is only some embodiments of the present application, rather than whole embodiments.Based on this Shen
Please in embodiment, the institute that those of ordinary skill in the art are obtained under the premise of creative work is not made
There are other embodiments, belong to the scope of the application protection.
A kind of Risk Forecast Method schematic flow sheet that Fig. 1 is provided for the embodiment of the present application, methods described is as follows
It is described.
Step 101：Obtain and pending user-related resource data.
Wherein, included in the resource data and using the resource for getting for characterizing the pending user
When occur the overdue time for releasing the resource after the sentence expires and for characterize the pending user using obtain
To the resource when there is the stock number of the overdue release resource for releasing the resource after the sentence expires.
In a step 101, pending user is obtained on resource service platform using distributing money in order to predict
The risk in source, server acquisition and pending user-related resource data, and according to the number of resources for getting
It is predicted that the risk of user.
Here resource service platform can be the platform for referring to provide the user resource service, money here
Source service can be the shared service of data resource, or the configuration service of channel resource, can also be
The transactional services of fund resources, here the form for resource described in the embodiment of the present application do not do specifically
Limit.
Because pending user is when using resource service platform, substantial amounts of resource data will be produced, for example：
Pending user discharges the stock number of resource in the resource useful life that resource service platform is specified, pending
User's release distributes time of resource etc., and these data are stored in the database of resource service platform,
That is can be from the data of resource service platform when needing to be predicted user with the presence or absence of risk
The resource data of pending user is got in storehouse.
When getting with pending user-related resource data, can also be taken according further to resource
Business platform is that the resource of stock number and resource service platform the synchronization generation of pending user resource allocation makes
With the time limit, determine whether the pending user overdue situation, example occurs when using the resource for getting
Such as：Whether there is the overdue time, and/or overdue release resource whether occur, it is assumed that overdue release resource occur
Situation, further determine it is overdue release resource stock number size.
Step 102：According to the described overdue time included in the resource data and the overdue release resource
Stock number, be calculated the characteristic value of the resource credit rating size for characterizing the pending user.
In a step 102, not only included with the pending user-related resource data due to getting
Overdue time and the stock number of overdue release resource, also comprising the user resources information of the pending user,
Here user resources information can refer to that pending user is currently owned by resource in resource service platform
Stock number, it is also possible to refer to stock number for the resource that pending user has used etc., these data equally quilt
Storage can synchronously be obtained in the database of resource service platform when the resource data of pending user is obtained
Get the user resources information of the pending user.
So, in a step 102, can also further exceed according to being included in the resource data
The user resources information of time phase, the stock number of the overdue release resource and pending user, calculates
To the characteristic value of the resource credit rating size for characterizing the pending user.
Specifically, the characteristic value of the pending user can be in the following manner calculated：
First, resource credit classification model is obtained based on training, according to being included in the resource data
The user resources information of overdue time, the stock number of the overdue release resource and the pending user, meter
Calculation obtains the Grad of the pending user；
Secondly, according to the Grad, it is determined that the resource credit rating size for characterizing the pending user
Characteristic value.
It should be noted that the resource credit classification model described in the embodiment of the present application can be base
Obtained in the training of LambdaMART sort algorithms.LambdaMART algorithms are based on GBDT
What (Gradient Boosting Decision Tree, i.e. Gradient Iteration decision tree) algorithm was realized, GBDT is calculated
Method is many superpositions of decision tree, and total algorithm is a kind of gradient descent algorithm, and each decision tree is one
Weak Classifier, the actual fitting of every decision tree be object function Grad, and LambdaMART is calculated
Method obtains the resource credit classification model by directly defining gradient to train, by the resource credit score
The result obtained after class model treatment can be represented by the Grad of the pending user.
It should be noted that LambdaMART algorithms can determine the Grad of user, and
Definition of the LambdaMART algorithms to the Grad of training sample i be：
λi=Σj:{i,j}∈Iλij-Σl:{l,i}∈Iλil；
Wherein, I represents training sample set, according to the instruction included in the pre-conditioned I to training sample set
Practice sample to be ranked up according to order from big to small.Assuming that the position of training sample j training sample i it
Afterwards, the position of training sample l is before training sample i, Σj:{i,j}∈IλijRepresent training sample i with training sample
The change sum of the Grad of training sample i, Σ caused by the location swap of this jl:{l,i}∈IλilRepresent training sample
The change sum of the Grad of training sample i caused by the location swap of this i and training sample l.
Wherein,ΔzijRepresent training sample i with training sample
Location swap between this j causes the granularity that the Grad of training sample i changes, Δ zilRepresent training sample
Location swap between this i and training sample l causes the granularity that the Grad of training sample i changes, si
Represent positions of the training sample i in training sample set I, sjRepresent training sample j in training sample set
Position in I, slRepresent positions of the training sample l in training sample set I.
Additionally, here cause the granularity that the Grad of training sample changes to be NDCG
(Normalized Discounted Cumulative Gain), or MAP (Mean Average
Precision), other specification is can also be, is not particularly limited here, the embodiment of the present application is mainly with NDCG
As a example by illustrate.
In the embodiment of the present application, training in advance obtains the resource credit classification model, is getting and institute
When stating pending user-related resource data, can be with using the resource credit classification model that obtains of training
It is calculated the Grad of the pending user.
Specifically describe how to train below and obtain resource credit classification model.
The first step：The resource data of N number of training sample is obtained, and determines the money of each training sample
The user resources of the overdue time, the stock number of overdue release resource and the training sample that are included in source data
Information.
Wherein, N is natural number.
In the embodiment of the present application, all users of resource service platform will can be used as training sample,
A portion user of resource service platform can also will be used as training sample, specific limit is not done here
It is fixed.
After the resource data for getting N number of training sample, the number of resources of each training sample data is determined
The user resources letter of the overdue time, the stock number of overdue release resource and the training sample that are included in
Breath, the resource credit classification is obtained in order to the resource data training according to each training sample for determining
Model.
Second step：According to the size of overdue time, N number of training sample is ranked up, and according to row
Sequence result, calculates the Grad of each training sample.
Specifically, it is determined that after the overdue time for getting N number of training sample, according to it is described overdue when
Between N number of training sample is ranked up, obtain ranking results.
For example：The N number of training sample for getting is respectively：A, B, C, D, E, F and G, it is assumed that A,
The overdue time of B, C, D, E, F and G is respectively：x1、x2、x3、x4、x5、x6、x7, wherein,
x1=x7.Assuming that the size order of the overdue time of this 7 training samples is：x6、x2、x5、x1、x7、
x3、x4, or x6、x2、x5、x7、x1、x3、x4, then the order according to the overdue time from big to small is right
The result that above-mentioned 7 training samples are ranked up is：F, B, E, A, G, C, D, or F, B, E,
G、A、C、D。
It should be noted that in the embodiment of the present application, order that can be according to the overdue time from big to small is right
The N number of training sample for getting is ranked up, it is also possible to according to order from small to large of overdue time to obtaining
To N number of training sample be ranked up, be not specifically limited.The embodiment of the present application with according to the overdue time from
Arrive greatly as a example by each training samples of N of the small order to getting is ranked up and illustrate..
After the ranking results for obtaining N number of training sample, can be existed according to each training sample
Position in ranking results, is calculated the Grad of each training sample.For example：Can utilize
LambdaMART algorithms are calculated the Grad of each training sample.
Still by taking above-mentioned 7 training samples A~G as an example, according to the sequence of F, B, E, A, G, C, D
As a result, being utilized respectively LambdaMART algorithms can obtain the gradient of F, B, E, A, G, C, D
Value is respectively：λ6, λ2, λ5, λ1, λ7, λ3, λ4。
3rd step：For training sample each described, respectively according to the give-and-take conditions of setting, adjust each
The Grad of the individual training sample.
Wherein, the give-and-take conditions of the setting are that the stock number based on overdue time and overdue release resource determines
's.
It should be noted that the Grad of each training sample obtained according to ranking results in second step
It can refer to the Initial Gradient value of each training sample.And in the embodiment of the present application, due to different training
Location swap will cause the Grad of training sample to convert between sample, then to each training sample
, it is necessary to determine the Initial Gradient value of each training sample to obtaining by location swap when being originally trained
The adjustment granularity being adjusted, and then obtain the Grad after the adjustment of each training sample.
Specifically, for training sample each described, respectively according to the give-and-take conditions of setting, by following
Mode adjusts the Grad of each training sample：
First, for one of them the first training sample, determine that the overdue time is not more than first instruction respectively
Practice at least one second training samples of sample and determine the overdue time more than first training sample
At least one the 3rd training samples.
By taking above-mentioned 7 training samples A~G as an example, for training sample A, training sample A's exceedes
Time phase is x1, wherein, no more than overdue time x1The overdue time be x7、x3、x4, corresponding training
Sample is respectively G, C, D, more than overdue time x1The overdue time be x6、x2、x5, corresponding training
Sample is respectively F, B, E, it may be determined that second training sample of training sample A is G, C, D, the
Three training samples are F, B, E.
Second training sample of training sample B can be determined for E, A, G, C, D using same method,
3rd training sample is F；Second training sample of training sample C be D, the 3rd training sample be F, B,
E、A、G；Training sample D do not exist the second training sample, the 3rd training sample be F, B, E, A,
G、C；Second training sample of training sample E is A, G, C, D, and the 3rd training sample is F, B；
Second training sample of training sample F is B, E, A, G, C, D, in the absence of the 3rd training sample；
Second training sample of training sample G is C, D, and the 3rd training sample is F, B, E, A.
Secondly, based on described at least one second training samples for determining, calculate respectively in the described first training
The ladder of first training sample that sample is triggered after being exchanged with the position of second training sample
First numerical value of angle value change.
Here illustrated by taking second training sample as an example.Because the second training sample belongs to the overdue time
The training sample of the overdue time of no more than the first training sample, in the first training sample and the second training sample
After this position exchanges, the first numerical value of the gradient value changes of first training sample for being triggered is big
Cause may be divided into following several situations：
The first situation：
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time more than setting the overdue time threshold value
When, will the parameter value corresponding with the overdue time that set and set weighted value product value as the first numerical value.
It should be noted that the threshold value of the overdue time of the setting refers to the server settings overdue time
In the maximum time limit, can be here not specifically limited by being set according to actual needs.
It is determined that the overdue time of first training sample it is overdue more than the second training sample data
During the time, the stock number and second training sample of the relatively more described overdue release resource of first training sample
The stock number of overdue release resource, it is determined that the stock number of the overdue release resource of first training sample is big
When the stock number of the overdue release resource of second training sample, the first training sample is determined whether
This overdue time whether more than setting the overdue time threshold value, and it is determined that first training sample
When the overdue time is more than the threshold value of the overdue time of setting, in the first training sample and second training sample
After position exchanges, the first numerical value of the gradient value changes of first training sample for being triggered can be
Will the parameter value corresponding with the overdue time that set and set weighted value product value as the first numerical value.
Wherein, the setting weighted value is greater than 1 natural number, is specifically set according to actual needs by server
It is fixed, it is not specifically limited, overdue time, corresponding parameter value can be according to actual LambdaMART
Algorithm determines, in the embodiment of the present application, it is assumed that parameter value corresponding with overdue time is t1, setting
Weighted value is ω1, then the first numerical value for obtaining is t1ω1。
By taking above-mentioned 7 training samples A~G as an example, for the first training sample A, the first training sample
Second training sample of this A is G, C, D.Wherein：The overdue time of A is x1, overdue release resource
Stock number be y1, the overdue time of G, C, D is respectively x7、x3、x4, the resource of overdue release resource
It is y to measure7、y3、y4。
Assuming that the overdue time x of A1Overdue time x more than C3, and y1More than y3, for the second training
For sample C, the overdue time of overdue time of A more than C is met, and the overdue release resource of A
The stock number of overdue release resource of the stock number more than C, determines whether whether the overdue time of A is more than
The threshold value of the overdue time of setting, it is assumed that the overdue time x of A1More than the threshold value of the overdue time of setting,
So after the sorting position of the sorting position of A and C is converted, the Grad of caused A occurs
The granularity of conversion is t1ω1。
Second case：
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time be not more than setting the overdue time threshold value
When, the parameter value corresponding with the overdue time that will be set is used as the first numerical value.
Specifically, it is determined that the exceeding more than second training sample overdue time of first training sample
During time phase, compare the resource of first training sample and the overdue release resource of second training sample
Amount, it is determined that the stock number of the overdue release resource of first training sample is more than second training sample
Overdue release resource stock number when, determine whether whether the overdue time of first training sample big
In setting the overdue time threshold value, and it is determined that the overdue time of first training sample be not more than setting
The overdue time threshold value when, after the first training sample is exchanged with the position of second training sample,
First numerical value of the gradient value changes of first training sample for being triggered can be will set with it is overdue
Time, corresponding parameter value was used as the first numerical value.
By taking above-mentioned 7 training samples A~G as an example, for the first training sample A, the first training sample
Second training sample of this A is G, C, D.
If overdue time of the overdue time of A more than C, it is assumed that y1More than y3, for the second training sample
For C, the overdue time of the overdue time more than C of A is met, the stock number of the overdue release resource of A is big
In the stock number of the overdue release resource of C, determine whether the overdue time of A whether more than exceeding for setting
The threshold value of time phase, it is assumed herein that the overdue time x of A1The threshold value of the overdue time for no more than setting,
So after the sorting position of the sorting position of A and C is converted, the Grad of caused A occurs
The granularity of conversion is t1。
The third mode：Exceeding for second training sample is equal to when the overdue time of first training sample
Time phase, and the stock number of the overdue release resource of first training sample is more than second training sample
Overdue release resource stock number when, the corresponding parameter value of stock number with overdue release resource that will be set
As the first numerical value.
Specifically, it is determined that the exceeding equal to second training sample overdue time of first training sample
During time phase, compare the resource of first training sample and the overdue release resource of second training sample
Amount, and it is determined that the stock number of the overdue release resource of first training sample is more than the described second training sample
During the stock number of this overdue release resource, occur in the position of the first training sample and second training sample
After exchange, the first numerical value of the gradient value changes of first training sample for being triggered can be to set
Parameter value corresponding with the stock number of overdue release resource is used as the first numerical value.
It should be noted that it is described it is overdue release resource the corresponding parameter value of stock number by server according to reality
LambdaMART algorithm in border determines, is not specifically limited, in the embodiment of the present application, it is assumed that with it is described
The corresponding parameter value of stock number of overdue release resource is p.
By taking above-mentioned 7 training samples A~G as an example, for the first training sample A, the first training sample
Second training sample of this A is G, C, D.
If the overdue time of A is equal to the overdue time of G, it is assumed that y1More than y4, using p as first
Numerical value.
It should be noted that for first training sample, being handed in the position with the second training sample
When changing, because the number of the second training sample is multiple, for the training sample of part second, in the first training
After sample is exchanged with the position of second training sample, the gradient of first training sample for being triggered
First numerical value of value changes belongs to the first situation；For the training sample of part second, in the first training sample
After being exchanged with the position of second training sample, the Grad of first training sample for being triggered becomes
The first numerical value changed belongs to second case；For the training sample of part second, the first training sample with should
After the position of the second training sample exchanges, the gradient value changes of first training sample for being triggered
First numerical value belongs to the third situation, that is to say, that sent out in the position of the first training sample and the second training sample
After raw exchange, multiple first numerical value will be obtained according to the number of the second training sample.
Again, based on described at least one the 3rd training samples for determining, calculate respectively in the 3rd training
The ladder of first training sample that sample is triggered after being exchanged with the position of first training sample
The second value of angle value change.
In the embodiment of the present application, when the second value is calculated, by the 3rd training sample
The overdue time of the stock number and first training sample of overdue time and expected release resource and expection are released
The stock number for putting resource is compared, and first training sample is to the described 3rd after being calculated exchange position
The Grad of training sample adjustment, and using result of calculation as second value.
It should be noted that calculating the method and the method phase for calculating first numerical value of the second value
Together, description is not repeated herein.
For the ease of the second value is made a distinction with first numerical value, the embodiment of the present application is being calculated
During the second value, it is assumed that parameter value corresponding with overdue time is t2, the weighted value for setting is ω2。
By taking above-mentioned 7 training samples A~G as an example, for training sample A, the of training sample A
Three training samples are F, B, E.Assuming that the overdue time of A is x1, it is overdue release resource stock number be
y1, the overdue time of F, B, E is respectively x6、x2、x5, the stock number of overdue release resource is y6、y2、
y5。
Assuming that x6More than the threshold value of the overdue time of setting, x2、x5、x1The overdue time for no more than setting
Threshold value, y1、y6、y2、y5Order from big to small is：y5＞ y6＞ y2＞ y1。
Because the overdue time of F, B, E is both greater than A, and the overdue time of only F is more than exceeding for setting
The threshold value of time phase, therefore, when the position of training sample A and training sample F changes, cause instruction
The first numerical value that the Grad of white silk sample F is adjusted is t2ω2, then cause the Grad of training sample A
The second value being adjusted is-t2ω2；When the position of training sample A and training sample B changes,
The first numerical value for causing the Grad of training sample B to be adjusted is t2, then cause the ladder of training sample A
The second value that angle value is adjusted is-t2；When training sample A changes with the position of training sample E
When, the first numerical value for causing the Grad of training sample E to be adjusted is t2, then cause training sample A
The second value that is adjusted of Grad be-t2。
Finally, using first numerical value and the second value, the gradient of first training sample is adjusted
Value.
After the first numerical value and the second value for obtaining each training sample, by first numerical value and
The second value is added in the Initial Gradient value of the training sample, after accumulated result is adjustment
The Grad of first training sample.
Still by taking above-mentioned 7 training samples A~G as an example, if A, B, C, D, E, F and G it is overdue when
Between be respectively：x1、x2、x3、x4、x5、x6、x7, the overdue release of A, B, C, D, E, F and G
The stock number of resource is respectively：y1、y2、y3、y4、y5、y6、y7, it is assumed that the size order of overdue time
For：x6、x2、x5、x1、x7、x3、x4, wherein, x1=x7, x6More than the threshold value of the overdue time of setting,
The size order of stock number of overdue release resource is：y5、y6、y4、y1、y3、y2、y7, according to above-mentioned
First numerical value and the computational methods of the second value recorded, can obtain, for F, B, E
For, the second value for adjusting A is respectively：-t2ω2、-t2、-t2, for G, C, D,
First numerical value of A is：P, t, t ω, then the size of the Grad of the training sample A after adjusting can be with
For：Initial Gradient value+p+t+t1ω1-t2ω2-t2-t2。
4th step：After the Grad to training sample each described is adjusted, after adjustment
The Grad of each training sample, classifies to N number of training sample, obtains resource credit
Disaggregated model, wherein, the money comprising overdue time and overdue release resource in the resource credit classification model
Source amount corresponding resource credit type and each corresponding Grad of resource credit type.
By the computational methods of the Grad of first training sample of above-mentioned record, institute can be calculated
State the corresponding N number of Grad of N number of sample data.
First, the N number of Grad that will be obtained is ranked up according to order from big to small, obtains ranking results,
According to ranking results, N number of Grad is divided into M different Grad interval, each ladder
Angle value is interval to include at least one gradient, and wherein M is natural number, and less than or equal to N, in the application reality
In applying example, M different Grad will obtaining is interval as M different resource credit type, by
At least one Grad is included in each resource credit type, therefore, each resource credit type correspondence
Different overdue time and the stock number of overdue release resource.
It is interval for each Grad after it is determined that the M Grad is interval, using preset algorithm
The interval corresponding characteristic value of described each Grad is obtained, and sets up the characteristic value with the Grad area
Between between corresponding relation.Here preset algorithm can be linear transformation, or other algorithms, no
It is specifically limited.
So, after the resource data for obtaining the pending user, based on the resource credit score that training is obtained
Class model, it is possible to obtain the Grad of the pending user, the size according to the Grad for obtaining determines
Grad where the Grad of the pending user is interval, interval corresponding with characteristic value according to Grad
Relation, can obtain the characteristic value of the pending user.
Step 103：According to the characteristic value, predict that the pending user uses the risk for distributing resource.
In step 103, according to the characteristic value of the described pending user for obtaining, can predict and obtain described
The risk of pending user.
It should be noted that the resource credit rating of the pending user is bigger, the wind of the pending user
Danger is lower, conversely, the resource credit rating of the pending user is smaller, the risk of the pending user is got over
It is high.In the embodiment of the present application, the characteristic value is used to characterize the resource credit rating of the pending user
Size, therefore, after the characteristic value that server determines the pending user, can be according to the characteristic value
The size of the resource credit rating of the described pending user for characterizing, prediction obtains the wind of the pending user
Danger.
Assuming that being that direct proportion is closed between the size of the resource credit rating of the characteristic value and the characteristic value
System, that is to say, that the characteristic value of the pending user is bigger, the resource credit rating of the pending user
Greatly, the risk of the pending user is smaller.
The scheme provided by the embodiment of the present application, is the optimization to prior art identifying user risk, from exceeding
Two granularities were set out and the risk of user is predicted time phase and overdue resource so that for it is different overdue when
Between and it is overdue release resource stock number user, can relatively accurately determine the resource credit rating of user.
If it should be noted that the overdue time of user more than the overdue time of setting threshold value and overdue release
The stock number of the resource put also than larger, increases the punishment to user by way of increasing user's Grad,
That is the Grad of user is bigger, the user's obtained in the resource credit classification model prediction obtained based on training
Characteristic value is smaller, and the risk of user is higher.
The scheme that the embodiment of the present application is provided, obtains and pending user-related resource data, the resource
Included in data and occur releasing institute after the sentence expires when using the resource for getting for characterizing the pending user
State the overdue time of resource and go out when using the resource for getting for characterizing the pending user
Now release the stock number of the overdue release resource of the resource after the sentence expires；According to the institute included in the resource data
The stock number of overdue time and the overdue release resource is stated, is calculated for characterizing the pending user
Resource credit rating size characteristic value；According to the characteristic value, predict that the pending user uses and divide
Risk with resource.By two angles of stock number from overdue time and overdue release resource, comprehensive descision
User uses and distributes resource generation overdue degree, and then determines the resource credit rating of user, so can
Obtain relatively accurate resource credit rating, and can be according to obtaining existing for resource credit rating accurately predicts user
Risk, while resource service platform is user resource allocation according to the resource credit rating that is calculated, can
The risk index of resource service platform is effectively reduced, and then lifts the circulation efficiency of resource service platform resource.
A kind of structural representation of risk identification equipment that Fig. 2 is provided for the embodiment of the present application.The risk is known
Other equipment includes：Acquiring unit 21, computing unit 22 and predicting unit 23, wherein：
Acquiring unit 21, for obtain with pending user-related resource data, wherein, the number of resources
In comprising for characterize the pending user occur releasing after the sentence expires when using the resource for getting it is described
The overdue time of resource and occur when using the resource for getting for characterizing the pending user
Release the stock number of the overdue release resource of the resource after the sentence expires；
Computing unit 22, for according to the described overdue time included in the resource data and described overdue releasing
The stock number of resource is put, the feature of the resource credit rating size for characterizing the pending user is calculated
Value；
Predicting unit 23, resource is distributed for according to the characteristic value, predicting that the pending user uses
Risk.
Alternatively, the user resources information of the pending user is also included in the resource data；
The computing unit 22, was additionally operable to according to described overdue time included in the resource data and described
The stock number of overdue release resource, is calculated the resource credit rating size for characterizing the pending user
Characteristic value, including：
Resource credit classification model is obtained based on training, according to included in the resource data it is described overdue when
Between, it is described it is overdue release resource stock number and the pending user user resources information, be calculated
The Grad of the pending user；
According to the Grad, it is determined that the feature of the resource credit rating size for characterizing the pending user
Value.
Specifically, the computing unit 22 is trained obtain resource credit classification model in the following manner, bag
Include：
The resource data of N number of training sample is obtained, and determines the resource data of each training sample
Comprising overdue time, the stock number of overdue release resource and the training sample user resources information, its
In, N is natural number；
According to the size of overdue time, N number of training sample is ranked up, and according to ranking results,
Calculate the Grad of each training sample；
For training sample each described, respectively according to the give-and-take conditions of setting, each described instruction is adjusted
Practice the Grad of sample, wherein, the give-and-take conditions of the setting are based on overdue time and overdue release resource
Stock number determine；
After the Grad to training sample each described is adjusted, based on each institute after adjustment
The Grad of training sample is stated, N number of training sample is classified, obtain resource credit classification model,
Wherein, corresponding to the stock number comprising overdue time and overdue release resource in the resource credit classification model
Resource credit type and each corresponding Grad of resource credit type.
Alternatively, the computing unit 22 adjusts each described training sample according to the give-and-take conditions of setting
This Grad, including：
For one of them the first training sample, determine that the overdue time is not more than first training sample respectively
At least one second training samples and determine overdue time more than at least the one of first training sample
Individual 3rd training sample；
Based on determine described at least one second training samples, calculate respectively first training sample with
The Grad of first training sample that the position of second training sample is triggered after exchanging becomes
The first numerical value changed；
Based on determine described at least one the 3rd training samples, calculate respectively the 3rd training sample with
The Grad of first training sample that the position of first training sample is triggered after exchanging becomes
The second value of change；
Using first numerical value and the second value, the Grad of first training sample is adjusted.
Alternatively, the computing unit 22 is calculated in first training sample and second training sample
Position exchange after the first numerical value of the gradient value changes of first training sample for being triggered, bag
Include：
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time more than setting the overdue time threshold value
When, will the parameter value corresponding with the overdue time that set and set weighted value product value as the first numerical value；
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time be not more than setting the overdue time threshold value
When, the parameter value corresponding with the overdue time that will be set is used as the first numerical value；
The overdue time of second training sample, and institute are equal to when the overdue time of first training sample
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
During the stock number of resource, the parameter value corresponding with the stock number of overdue release resource that will be set is counted as first
Value.
It should be noted that the risk profile equipment that the embodiment of the present application is provided can be by hardware mode reality
It is existing, it is also possible to be realized by software mode, do not limited here.
It will be understood by those skilled in the art that embodiments herein can be provided as method, device (equipment),
Or computer program product.Therefore, the application can using complete hardware embodiment, complete software embodiment,
Or the form of the embodiment in terms of combination software and hardware.And, the application can use at one or more it
In include computer-usable storage medium (the including but not limited to disk storage of computer usable program code
Device, CD-ROM, optical memory etc.) on implement computer program product form.
The application is with reference to the method according to the embodiment of the present application, device (equipment) and computer program product
Flow chart and/or block diagram describe.It should be understood that can by computer program instructions realize flow chart and/or
Flow in each flow and/or square frame and flow chart and/or block diagram and/or square frame in block diagram
With reference to.These computer program instructions to all-purpose computer, special-purpose computer, Embedded Processor can be provided
Or the processor of other programmable data processing devices is producing a machine so that by computer or other
The instruction of the computing device of programmable data processing device produce for realizing in one flow of flow chart or
The device of the function of being specified in one square frame of multiple flows and/or block diagram or multiple square frames.
These computer program instructions may be alternatively stored in can guide computer or the treatment of other programmable datas to set
In the standby computer-readable memory for working in a specific way so that storage is in the computer-readable memory
Instruction produce include the manufacture of command device, the command device realization in one flow of flow chart or multiple
The function of being specified in one square frame of flow and/or block diagram or multiple square frames.
These computer program instructions can be also loaded into computer or other programmable data processing devices, made
Obtain and series of operation steps is performed on computer or other programmable devices to produce computer implemented place
Reason, so as to the instruction performed on computer or other programmable devices is provided for realizing in flow chart one
The step of function of being specified in flow or multiple one square frame of flow and/or block diagram or multiple square frames.
Although having been described for the preferred embodiment of the application, those skilled in the art once know base
This creative concept, then can make other change and modification to these embodiments.So, appended right will
Ask and be intended to be construed to include preferred embodiment and fall into having altered and changing for the application scope.
Obviously, those skilled in the art can carry out various changes and modification without deviating from this Shen to the application
Scope please.So, if these modifications of the application and modification belong to the application claim and its be equal to
Within the scope of technology, then the application is also intended to comprising these changes and modification.
Claims (10)
1. a kind of Risk Forecast Method, it is characterised in that including：
Acquisition and pending user-related resource data, wherein, comprising for characterizing in the resource data
The pending user occur releasing after the sentence expires when using the resource for getting the resource the overdue time and
Occur releasing the resource after the sentence expires when using the resource for getting for characterizing the pending user
Overdue release resource stock number；
According to the described overdue time included in the resource data and the stock number of the overdue release resource,
It is calculated the characteristic value of the resource credit rating size for characterizing the pending user；
According to the characteristic value, predict that the pending user uses the risk for distributing resource.
2. Risk Forecast Method as claimed in claim 1, it is characterised in that in the resource data also
User resources information comprising the pending user；
According to the described overdue time included in the resource data and the stock number of the overdue release resource,
The characteristic value of the resource credit rating size for characterizing the pending user is calculated, including：
Resource credit classification model is obtained based on training, according to included in the resource data it is described overdue when
Between, it is described it is overdue release resource stock number and the pending user user resources information, be calculated
The Grad of the pending user；
According to the Grad, it is determined that the feature of the resource credit rating size for characterizing the pending user
Value.
3. Risk Forecast Method as claimed in claim 2, it is characterised in that train in the following manner
Resource credit classification model is obtained, including：
The resource data of N number of training sample is obtained, and determines the resource data of each training sample
Comprising overdue time, the stock number of overdue release resource and the training sample user resources information, its
In, N is natural number；
According to the size of overdue time, N number of training sample is ranked up, and according to ranking results,
Calculate the Grad of each training sample；
For training sample each described, respectively according to the give-and-take conditions of setting, each described instruction is adjusted
Practice the Grad of sample, wherein, the give-and-take conditions of the setting are based on overdue time and overdue release resource
Stock number determine；
After the Grad to training sample each described is adjusted, based on each institute after adjustment
The Grad of training sample is stated, N number of training sample is classified, obtain resource credit classification model,
Wherein, corresponding to the stock number comprising overdue time and overdue release resource in the resource credit classification model
Resource credit type and each corresponding Grad of resource credit type.
4. Risk Forecast Method as claimed in claim 3, it is characterised in that according to the exchange bar of setting
Part, adjusts the Grad of each training sample, including：
For one of them the first training sample, determine that the overdue time is not more than first training sample respectively
At least one second training samples and determine overdue time more than at least the one of first training sample
Individual 3rd training sample；
Based on determine described at least one second training samples, calculate respectively first training sample with
The Grad of first training sample that the position of second training sample is triggered after exchanging becomes
The first numerical value changed；
Based on determine described at least one the 3rd training samples, calculate respectively the 3rd training sample with
The Grad of first training sample that the position of first training sample is triggered after exchanging becomes
The second value of change；
Using first numerical value and the second value, the Grad of first training sample is adjusted.
5. Risk Forecast Method as claimed in claim 4, it is characterised in that calculate in the described first instruction
First training sample that white silk sample is triggered after being exchanged with the position of second training sample
First numerical value of gradient value changes, including：
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time more than setting the overdue time threshold value
When, will the parameter value corresponding with the overdue time that set and set weighted value product value as the first numerical value；
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time be not more than setting the overdue time threshold value
When, the parameter value corresponding with the overdue time that will be set is used as the first numerical value；
The overdue time of second training sample, and institute are equal to when the overdue time of first training sample
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
During the stock number of resource, the parameter value corresponding with the stock number of overdue release resource that will be set is counted as first
Value.
6. a kind of risk profile equipment, it is characterised in that including：
Acquiring unit, for obtain with pending user-related resource data, wherein, the resource data
In comprising occurring releasing the money after the sentence expires when using the resource for getting for characterizing the pending user
The overdue time in source and occur prolonging when using the resource for getting for characterizing the pending user
Phase discharges the stock number of the overdue release resource of the resource；
Computing unit, for according to the described overdue time included in the resource data and the overdue release
The stock number of resource, is calculated the feature of the resource credit rating size for characterizing the pending user
Value；
Predicting unit, resource is distributed for according to the characteristic value, predicting that the pending user uses
Risk.
7. risk profile equipment as claimed in claim 6, it is characterised in that in the resource data also
User resources information comprising the pending user；
The computing unit is according to the described overdue time included in the resource data and the overdue release
The stock number of resource, is calculated the feature of the resource credit rating size for characterizing the pending user
Value, including：
Resource credit classification model is obtained based on training, according to included in the resource data it is described overdue when
Between, it is described it is overdue release resource stock number and the pending user user resources information, be calculated
The Grad of the pending user；
According to the Grad, it is determined that the feature of the resource credit rating size for characterizing the pending user
Value.
8. risk profile equipment as claimed in claim 7, it is characterised in that the computing unit passes through
In the following manner training obtains resource credit classification model, including：
The resource data of N number of training sample is obtained, and determines the resource data of each training sample
Comprising overdue time, the stock number of overdue release resource and the training sample user resources information, its
In, N is natural number；
According to the size of overdue time, N number of training sample is ranked up, and according to ranking results,
Calculate the Grad of each training sample；
For training sample each described, respectively according to the give-and-take conditions of setting, each described instruction is adjusted
Practice the Grad of sample, wherein, the give-and-take conditions of the setting are based on overdue time and overdue release resource
Stock number determine；
After the Grad to training sample each described is adjusted, based on each institute after adjustment
The Grad of training sample is stated, N number of training sample is classified, obtain resource credit classification model,
Wherein, corresponding to the stock number comprising overdue time and overdue release resource in the resource credit classification model
Resource credit type and each corresponding Grad of resource credit type.
9. risk profile equipment as claimed in claim 8, it is characterised in that the computing unit according to
The give-and-take conditions of setting, adjust the Grad of each training sample, including：
For one of them the first training sample, determine that the overdue time is not more than first training sample respectively
At least one second training samples and determine overdue time more than at least the one of first training sample
Individual 3rd training sample；
Based on determine described at least one second training samples, calculate respectively first training sample with
The Grad of first training sample that the position of second training sample is triggered after exchanging becomes
The first numerical value changed；
Based on determine described at least one the 3rd training samples, calculate respectively the 3rd training sample with
The Grad of first training sample that the position of first training sample is triggered after exchanging becomes
The second value of change；
Using first numerical value and the second value, the Grad of first training sample is adjusted.
10. risk profile equipment as claimed in claim 9, it is characterised in that the computing unit is calculated
Described triggered after first training sample is exchanged with the position of second training sample
First numerical value of the gradient value changes of one training sample, including：
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time more than setting the overdue time threshold value
When, will the parameter value corresponding with the overdue time that set and set weighted value product value as the first numerical value；
When the overdue time of the overdue time more than second training sample of first training sample, and institute
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
The stock number of resource, and first training sample the overdue time be not more than setting the overdue time threshold value
When, the parameter value corresponding with the overdue time that will be set is used as the first numerical value；
The overdue time of second training sample, and institute are equal to when the overdue time of first training sample
State the overdue release of the stock number more than second training sample of the overdue release resource of the first training sample
During the stock number of resource, the parameter value corresponding with the stock number of overdue release resource that will be set is counted as first
Value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510825234.4A CN106779272A (en) | 2015-11-24 | 2015-11-24 | A kind of Risk Forecast Method and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510825234.4A CN106779272A (en) | 2015-11-24 | 2015-11-24 | A kind of Risk Forecast Method and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106779272A true CN106779272A (en) | 2017-05-31 |
Family
ID=58963722
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510825234.4A Pending CN106779272A (en) | 2015-11-24 | 2015-11-24 | A kind of Risk Forecast Method and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106779272A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108389125A (en) * | 2018-02-27 | 2018-08-10 | 挖财网络技术有限公司 | The overdue Risk Forecast Method and device of credit applications |
| CN108614737A (en) * | 2018-05-11 | 2018-10-02 | 阿里巴巴集团控股有限公司 | Recovery method as resource, device and server |
| TWI664535B (en) * | 2017-06-16 | 2019-07-01 | 香港商阿里巴巴集團服務有限公司 | Data type identification, model training, risk identification method, device and equipment |
| WO2019218748A1 (en) * | 2018-05-16 | 2019-11-21 | 阿里巴巴集团控股有限公司 | Insurance service risk prediction processing method, device and processing equipment |
| CN111428125A (en) * | 2019-01-10 | 2020-07-17 | 北京三快在线科技有限公司 | Sorting method and device, electronic equipment and readable storage medium |
| CN114926267A (en) * | 2022-05-28 | 2022-08-19 | 平安银行股份有限公司 | Risk data processing method, related equipment and medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8312460B1 (en) * | 2007-05-22 | 2012-11-13 | Hewlett-Packard Development Company, L.P. | Allocating computer resources to workloads using utilization based probability distributions |
| CN103093379A (en) * | 2012-05-09 | 2013-05-08 | 李葛亮 | Methods used for preventing and controlling credit risk |
| CN103123712A (en) * | 2011-11-17 | 2013-05-29 | 阿里巴巴集团控股有限公司 | Method and system for monitoring network behavior data |
| CN104168318A (en) * | 2014-08-18 | 2014-11-26 | 中国联合网络通信集团有限公司 | Resource service system and resource distribution method thereof |
| CN104270459A (en) * | 2014-10-20 | 2015-01-07 | 山东省计算中心（国家超级计算济南中心） | Method for enhancing fair cloud computing user resource quota allocation |
| CN104462611A (en) * | 2015-01-05 | 2015-03-25 | 五八同城信息技术有限公司 | Modeling method, ranking method, modeling device and ranking device for information ranking model |
-
2015
- 2015-11-24 CN CN201510825234.4A patent/CN106779272A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8312460B1 (en) * | 2007-05-22 | 2012-11-13 | Hewlett-Packard Development Company, L.P. | Allocating computer resources to workloads using utilization based probability distributions |
| CN103123712A (en) * | 2011-11-17 | 2013-05-29 | 阿里巴巴集团控股有限公司 | Method and system for monitoring network behavior data |
| CN103093379A (en) * | 2012-05-09 | 2013-05-08 | 李葛亮 | Methods used for preventing and controlling credit risk |
| CN104168318A (en) * | 2014-08-18 | 2014-11-26 | 中国联合网络通信集团有限公司 | Resource service system and resource distribution method thereof |
| CN104270459A (en) * | 2014-10-20 | 2015-01-07 | 山东省计算中心（国家超级计算济南中心） | Method for enhancing fair cloud computing user resource quota allocation |
| CN104462611A (en) * | 2015-01-05 | 2015-03-25 | 五八同城信息技术有限公司 | Modeling method, ranking method, modeling device and ranking device for information ranking model |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI664535B (en) * | 2017-06-16 | 2019-07-01 | 香港商阿里巴巴集團服務有限公司 | Data type identification, model training, risk identification method, device and equipment |
| CN108389125A (en) * | 2018-02-27 | 2018-08-10 | 挖财网络技术有限公司 | The overdue Risk Forecast Method and device of credit applications |
| CN108614737A (en) * | 2018-05-11 | 2018-10-02 | 阿里巴巴集团控股有限公司 | Recovery method as resource, device and server |
| WO2019218748A1 (en) * | 2018-05-16 | 2019-11-21 | 阿里巴巴集团控股有限公司 | Insurance service risk prediction processing method, device and processing equipment |
| CN111428125A (en) * | 2019-01-10 | 2020-07-17 | 北京三快在线科技有限公司 | Sorting method and device, electronic equipment and readable storage medium |
| CN111428125B (en) * | 2019-01-10 | 2023-05-30 | 北京三快在线科技有限公司 | Ordering method, ordering device, electronic equipment and readable storage medium |
| CN114926267A (en) * | 2022-05-28 | 2022-08-19 | 平安银行股份有限公司 | Risk data processing method, related equipment and medium |
| CN114926267B (en) * | 2022-05-28 | 2024-03-26 | 平安银行股份有限公司 | Risk data processing method, related equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106779272A (en) | A kind of Risk Forecast Method and equipment | |
| CN105869401B (en) | A kind of road network dynamic zoning method based on the different degree of crowding | |
| CN107705207A (en) | Method, apparatus, equipment and the computer-readable storage medium that customer value is assessed | |
| CN107682575A (en) | Business personnel's incoming call inlet wire distribution method, electronic installation, computer-readable recording medium | |
| CN105354595A (en) | Robust visual image classification method and system | |
| CN104427505B (en) | A kind of method and device of cell scenario division | |
| CN102722577B (en) | The determination method and device of index changeable weight | |
| CN106295351B (en) | A kind of Risk Identification Method and device | |
| CN107506350A (en) | A kind of method and apparatus of identification information | |
| CN110119948A (en) | Based on when variable weight dynamic combined power consumer credit assessment method and system | |
| CN107705011A (en) | Electricity market risk management system and method | |
| CN104581748A (en) | A method and device for identifying a scene in a wireless communication network | |
| CN107341512A (en) | A kind of method and device of transfer learning classification | |
| CN106454857A (en) | Evaluation method and device for network planning | |
| CN109784556A (en) | A kind of base station equipment installation project resource smoothing method based on previous data | |
| CN116109458A (en) | Reservoir flood discharge gate scheduling method, system, storage medium and computing equipment | |
| CN111506876A (en) | Data prediction analysis method, system, equipment and readable storage medium | |
| CN111061948A (en) | User label recommendation method and device, computer equipment and storage medium | |
| CN109685133A (en) | The data classification method of prediction model low cost, high discrimination based on building | |
| CN113869700A (en) | Performance index prediction method and device, electronic equipment and storage medium | |
| CN111932302B (en) | Method, device, equipment and system for determining number of service sites in area | |
| CN107886217A (en) | A kind of labor turnover Risk Forecast Method and device based on clustering algorithm | |
| Knežević et al. | A two-level approach for human resource planning towards organizational efficiency of a postal distribution system | |
| US11374869B2 (en) | Managing bandwidth based on user behavior | |
| CN111404974B (en) | Cloud computing efficiency evaluation method and device and evaluation equipment |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170531 |