CN106485560A - The method and apparatus that a kind of online affairs data processing model is issued - Google Patents
The method and apparatus that a kind of online affairs data processing model is issued Download PDFInfo
- Publication number
- CN106485560A CN106485560A CN201510530530.1A CN201510530530A CN106485560A CN 106485560 A CN106485560 A CN 106485560A CN 201510530530 A CN201510530530 A CN 201510530530A CN 106485560 A CN106485560 A CN 106485560A
- Authority
- CN
- China
- Prior art keywords
- online
- pending
- transaction information
- transaction
- process model
- 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
Landscapes
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
This application discloses a kind of method and apparatus of online affairs data processing model issue and a kind of online affairs data processing method and device.Relevant parameter according to pending Transaction Information online will pending Transaction Information cluster online, and the pending online Transaction Information after cluster is diverted to newly-built transaction data process model according to the algorithm setting and after existing transaction data process model processed, according to above two model, the Performance Evaluating Indexes of pending item data result online are calculated with the performance difference of newly-built transaction data process model and existing transaction data process model.
Description
Technical field
The application is related to the method and apparatus that a kind of online affairs data processing model is issued.The application further relates to
A kind of method and apparatus of online affairs data processing.
Background technology
The online online affairs data handling system of Internet, needs to ensure the safety of Transaction Information,
The transaction data process model of line transaction data process system is used for the safety that assessment judges online affairs data,
Its performance can increasingly be degenerated in passage in time, previously established for a period of time existing transaction data process
Model can not meet up-to-date true online affairs data cases, needs for this to develop newly-built transaction data process
Model is replacing existing model, and newly-built model is based on experimental data modeling.Generally, newly-built thing
The foundation of business data processing model is the related data according to the real online affairs data having occurred and that,
The behavior characteristicss in Transaction Information that needs are identified in newly-built transaction data process model use parameter or
The form of variable embodies, and that is, newly-built transaction data process model is set up based on off-line training mode.
And the modeling of newly-built transaction data process model is a time-consuming long project, general newly-built affairs
The foundation of data processing model uses, to formally reaching the standard grade, the time needing 1 month the soonest, because what model was set up
Cycle length, therefore newly-built transaction data process model based on true transaction data after Transaction Information
The behavior characteristicss of Transaction Information that identify of interior needs will not be included in newly-built transaction data process mould
In type, thus leading to the training result of model and the recognition result of true transaction data environment to have very big difference
Different.
Application content
The application provides a kind of method that online affairs data processing model is issued, can be using online true
The newly-built transaction data process model of Transaction Information environmental training, to solve existing transaction data process mould
The training result of type off-line training mode is had very with the recognition result of the most newly-built true online affairs data environment
The problem of big difference.Correspondingly, the application also provides a kind of online affairs data processing model real-time release
Device.In addition the application is also provided at a kind of method of online affairs data processing and a kind of online affairs data
The device of reason.
The method that a kind of online affairs data processing model that the application provides is issued, comprises the following steps:
Obtain the relevant parameter of pending Transaction Information online;
Relevant parameter according to pending Transaction Information online will pending Transaction Information classification online;
Pending online Transaction Information after cluster is diverted to newly-built transaction data process according to the algorithm setting
Model or existing transaction data process model;
According to above two model, the Performance Evaluating Indexes of pending item data result online are calculated new
Build transaction data process model and the performance difference of existing transaction data process model.
Alternatively, the relevant parameter of described pending online Transaction Information includes:Pending Transaction Information online
Sequence number, the pending affairs amount of money online, the place that pending transaction initiator logs in online, wait online to locate
The time of director's business generation or in real time pending transaction initiator account carry out the number of times of online affairs in one day.
Alternatively, the relevant parameter of described basis pending Transaction Information online will pending Transaction Information online
The method of cluster includes:
Relevant parameter according to pending Transaction Information online will pending Transaction Information gather uniformly at random online
Class.
Alternatively, uniformly at random online waited locate online by the relevant parameter of pending Transaction Information for described basis
The method of director's business data clusters includes:
The operation as divisor, the sequence number of pending Transaction Information online being remmed with the value setting, remainder
Pending Transaction Information is classified as same class to identical online.
Alternatively, described by cluster after pending online Transaction Information according to set algorithm be diverted to newly-built
The method of transaction data process model or existing transaction data process model includes:
Set the ratio of pending Transaction Information shunting online;
Ratio according to the pending online Transaction Information shunting setting is by different classes of pending online affairs
Data is diverted to newly-built transaction data process model respectively and existing transaction data process model is processed.
Alternatively, the relevant parameter of described basis pending Transaction Information online will pending Transaction Information online
The method of cluster includes:
According to the relevant parameter of pending Transaction Information online, the pending online thing of the condition setting will be met
Business data is classified as excessive risk class pending Transaction Information online, and the condition being unsatisfactory for setting is pending online
Transaction Information is classified as low-risk class pending Transaction Information online;
Correspondingly,
Described by cluster after pending online Transaction Information according to set algorithm be diverted to newly-built Transaction Information
Process the method that model and existing transaction data process model processed to include:
Pending Transaction Information simultaneously enters newly-built transaction data process model and existing thing online to make excessive risk class
Business data processing model is processed.
Alternatively, the condition of described setting includes:
The pending affairs amount of money exceeds the threshold value setting online, the place that pending transaction initiator logs in online
Belong to scope set in advance, the time that pending affairs occur online belongs to scope set in advance or online
Pending transaction initiator account carries out online affairs number of times in one day exceeds the threshold value setting.
Alternatively, described according to above two model, the performance of pending item data result online is commented
The step that valency index calculates the performance difference of newly-built transaction data process model and existing transaction data process model
Including:
Draw the precision ratio-recall curve of newly-built, existing transaction data process model respectively;
According to the precision ratio-recall curve of described newly-built, existing transaction data process model, compare and draw newly
Build transaction data process model and the performance difference of existing transaction data process model.
Alternatively, the described precision ratio-recall curve drawing newly-built, existing transaction data process model respectively
Method include:
The adjustment ratio that pending Transaction Information shunts online;
Ratio according to different pending online Transaction Information shuntings makes pending Transaction Information online respectively enter
Newly-built, existing transaction data process model;
Calculate difference respectively online at newly-built, existing Transaction Information in the case of pending Transaction Information shunt ratio
The precision ratio of reason model and recall ratio;
Set up a flat square using recall ratio and precision ratio as the coordinate axess of plane right-angle coordinate respectively to sit
Mark system, will be corresponding for newly-built transaction data process model under pending Transaction Information difference shunt ratio online
Recall ratio and precision ratio data are to as the different coordinate points in described plane right-angle coordinate;Treated online
Process the corresponding recall ratio of existing transaction data process model and precision ratio under Transaction Information difference shunt ratio
Data is to as the different coordinate points in described plane right-angle coordinate;
By transaction data process model newly-built in described plane right-angle coordinate and existing transaction data process model
Coordinate points be linked to be line respectively and obtain the precision ratio-recall curve of newly-built transaction data process model and existing
Precision ratio-the recall curve of transaction data process model.
Alternatively, described precision ratio is:Transaction data process model evaluation identifies correctly pending thing online
The ratio of the pending online Transaction Information total amount that business data bulk is assessed with this transaction data process model;
Described recall ratio is:Transaction data process model evaluation correctly pending Transaction Information quantity and this thing online
It is actually needed, in the transaction of business data processing model assessment, the pending online Transaction Information quantity that assessment identifies
Ratio.
Alternatively, described precision ratio-recall curve according to described newly-built, existing transaction data process model,
Comparison show that the method for the performance difference of newly-built transaction data process model and existing transaction data process model is:
In described plane right-angle coordinate, precision ratio according to newly-built, existing transaction data process model-look into
The distance of full rate curve distance point (1,1), draws to newly-built transaction data process model and existing thing
The performance difference of business data processing model.
A kind of online affairs data processing method that the application provides, comprises the steps:
Obtain the relevant parameter of pending Transaction Information online;
Relevant parameter according to pending Transaction Information online will pending Transaction Information cluster online;
Pending online Transaction Information after cluster is diverted to newly-built transaction data process according to the algorithm setting
Model and existing transaction data process model are processed.
Alternatively, the relevant parameter of described pending online Transaction Information includes:Pending Transaction Information online
Sequence number, the pending affairs amount of money online, the place that pending transaction initiator logs in online, wait online to locate
The time of director's business generation or in real time pending transaction initiator account carry out the number of times of online affairs in one day.
Alternatively, the relevant parameter of described basis pending Transaction Information online will pending Transaction Information online
The method of cluster includes:
Relevant parameter according to pending Transaction Information online will pending Transaction Information gather uniformly at random online
Class.
Alternatively, uniformly at random online waited locate online by the relevant parameter of pending Transaction Information for described basis
The method of director's business data clusters includes:
The operation as divisor, the sequence number of pending Transaction Information online being remmed with the value setting, remainder
Pending Transaction Information is classified as same class to identical online.
Alternatively, described by cluster after pending online Transaction Information according to set algorithm be diverted to newly-built
The method that transaction data process model and existing transaction data process model are processed includes:
Set the ratio of pending Transaction Information shunting online;
Ratio according to the pending online Transaction Information shunting setting is by different classes of pending online affairs
Data is diverted to newly-built transaction data process model respectively and existing transaction data process model is processed.
Alternatively, the relevant parameter of described basis pending Transaction Information online will pending Transaction Information online
The method of cluster includes:
According to the relevant parameter of pending Transaction Information online, the pending online thing of the condition setting will be met
Business data is classified as excessive risk class pending Transaction Information online, and the condition being unsatisfactory for setting is pending online
Transaction Information is classified as low-risk class pending Transaction Information online;
Correspondingly,
Described by cluster after pending online Transaction Information according to set algorithm be diverted to newly-built Transaction Information
Process the method that model and existing transaction data process model processed to include:
Pending Transaction Information simultaneously enters newly-built transaction data process model and existing thing online to make excessive risk class
Business data processing model is processed.
Alternatively, the condition of described setting includes:
The pending affairs amount of money exceeds the threshold value setting online, the place that pending transaction initiator logs in online
Belong to scope set in advance, the time that pending affairs occur online belongs to scope set in advance or online
Pending transaction initiator account carries out online affairs number of times in one day exceeds the threshold value setting.
The device that a kind of online affairs data processing model that the application provides is issued, including:
Acquiring unit, for obtaining the relevant parameter of pending Transaction Information online;
Taxon, will pending number of transactions online for the relevant parameter according to pending Transaction Information online
According to cluster;
Dividing cell, for being diverted to newly the pending online Transaction Information after cluster according to the algorithm setting
Build transaction data process model or existing transaction data process model;
Assessment unit, for the performance to pending item data result online according to above two model
Evaluation index calculates the performance difference of newly-built transaction data process model and existing transaction data process model.
Alternatively, described taxon, specifically for the relevant parameter according to pending Transaction Information online with
Machine equably will pending Transaction Information cluster online.
Alternatively, described taxon, specifically for the value that sets as divisor to pending number of transactions online
According to the operation that remmed of sequence number, pending Transaction Information is classified as same class to remainder identical online.
Alternatively, described dividing cell includes:
Shunt ratio sets subelement, for setting the ratio of pending Transaction Information shunting online;
Shunting performance element, for the ratio according to the pending online Transaction Information shunting setting by inhomogeneity
Other pending online Transaction Information is diverted at newly-built transaction data process model and existing Transaction Information respectively
Reason model is processed.Alternatively, described taxon, specifically for according to pending number of transactions online
According to relevant parameter, the pending online Transaction Information meeting the condition setting is classified as excessive risk class and treats online
Process Transaction Information, the pending online Transaction Information of the condition being unsatisfactory for setting is classified as low-risk class online
Pending Transaction Information;
Correspondingly,
Described dividing cell is specifically for pending Transaction Information simultaneously enters newly-built thing online to make excessive risk class
Business data processing model and existing transaction data process model are processed.
Preferably, described taxon, specifically for:The pending affairs amount of money online is exceeded the threshold setting
Value, the place that pending transaction initiator logs in online belongs to scope set in advance, pending affairs online
Occur time belong to scope set in advance or online pending transaction initiator account carry out in one day online
The pending online Transaction Information that the number of times of affairs exceeds the threshold value setting is classified as excessive risk class pending thing online
Business data.
Alternatively, described assessment unit includes:
Curve plotting subelement, for drawing the precision ratio of newly-built, existing transaction data process model-look into respectively
Full rate curve;
Difference assesses subelement, for the precision ratio according to described newly-built, existing transaction data process model-look into
Full rate curve, compares the poor performance drawing newly-built transaction data process model and existing transaction data process model
Different.
Alternatively, described curve plotting subelement includes:
Shunt ratio adjusts subelement, for adjusting the ratio of pending Transaction Information shunting online;
Shunting execution subelement, for making to treat online according to the ratio of different pending online Transaction Information shuntings
Process Transaction Information and respectively enter newly-built, existing transaction data process model;
Computation subunit, for calculate respectively difference newly-built in the case of pending Transaction Information shunt ratio online,
The precision ratio of existing transaction data process model and recall ratio;
Draw the first subelement, for respectively using recall ratio and precision ratio as the coordinate of plane right-angle coordinate
Axle sets up a plane right-angle coordinate, will newly-built affairs under pending Transaction Information difference shunt ratio online
The corresponding recall ratio of data processing model and precision ratio data to as in described plane right-angle coordinate not
Same coordinate points;To existing transaction data process model phase under the different shunt ratio of pending Transaction Information online
Corresponding recall ratio and precision ratio data are to as the different coordinate points in described plane right-angle coordinate;
Draw the second subelement, for by transaction data process model newly-built in described plane right-angle coordinate and
What the coordinate points of existing transaction data process model were linked to be that line obtains newly-built transaction data process model respectively looks into standard
Rate-recall curve and the precision ratio-recall curve of existing transaction data process model.
Alternatively, described difference assesses subelement, specifically in described plane right-angle coordinate, according to
The distance of the precision ratio of newly-built, existing transaction data process model-recall curve range points (1,1),
Draw the performance difference to newly-built transaction data process model and existing transaction data process model.
A kind of online affairs data processing equipment that the application provides, including:
Data capture unit, for obtaining the relevant parameter of pending Transaction Information online;
Data sorting unit, will pending thing online for the relevant parameter according to pending Transaction Information online
Business data clusters;
Data distribution unit, for shunting the pending online Transaction Information after cluster according to the algorithm setting
Processed to newly-built transaction data process model and existing transaction data process model.
Compared with prior art, the application has advantages below:Compare existing online affairs data processing model
Off-line training mode, the method that the online affairs data processing model that the application provides is issued can be using online
The true transaction data environment newly-built transaction data process model of training, solve at existing newly-built Transaction Information
The training result of reason model off-line training mode and the actual recognition result of up-to-date true online affairs data environment
There is the problem of very big difference.Reach make the training of newly-built transaction data process model more conform to up-to-date
The effect of line Transaction Information environment.
Brief description
Fig. 1 is that a kind of flow process of the embodiment of method of online affairs data processing model issue of the application is shown
It is intended to.
Fig. 2 is a kind of structural frames of the embodiment of device of online affairs data processing model issue of the application
Figure.
Fig. 3 is a kind of schematic flow sheet of the embodiment of online affairs data processing method of the application.
Fig. 4 is a kind of structured flowchart schematic diagram of the embodiment of online affairs data processing equipment of the application.
Specific embodiment
Elaborate a lot of details in order to fully understand the application in the following description.But the application
Can much to implement different from alternate manner described here, those skilled in the art can without prejudice to
Similar popularization is done, therefore the application is not embodied as being limited by following public in the case of the application intension.
The embodiment of the method that a kind of online affairs data processing model of the application is issued, as shown in figure 1,
Below taking online transaction affairs on Internet as a example, describe the online affairs data processing of the application in detail
The enforcement of the method for model real-time release.
Online trading system on Internet, in order to ensure the safety of transaction, needs the transaction carrying out is entered
Row risk assessment, the transaction data of transaction system processes the dependency number of the transaction risk that model is grasped according to itself
Risk according to assessment real-time deal.So that the assessment of transaction risk meets up-to-date real-time network environment,
Online trading system needs often to issue the existing number of deals that new transaction data processes model replacement performance degradation
According to process model.Send out in real time with described online affairs data processing model in the online trading system of the present embodiment
The corresponding method for online transaction data processing model real-time release of the method for cloth.Described affairs are in this reality
Apply to be specially in example and conclude the business, Transaction Information is transaction data, transaction data process model is processed for transaction data
Model, transaction initiator is the buyer of transaction.
Step S101:Obtain the relevant parameter of pending Transaction Information online;
In the present embodiment, obtain in online trading system the relevant parameter of pending transaction data continuing in real time
Continuous execution next step S102.
Step S102:Relevant parameter according to pending Transaction Information online will pending Transaction Information gather online
Class;
The method of pending Transaction Information cluster online is had multiple, provide two kinds in the present embodiment preferably
Relevant parameter according to transaction data is by the method for the cluster of pending transaction data online:
Method one:After online trading system receives transaction data, can uniformly at random transaction data be clustered.
This method can work out sequence number according to the order that real-time transaction data reaches system for transaction data, and numeric order is even
Continuous, plus one and be incremented by.According to the sequence number in the parameter of pending transaction data online by cluster of concluding the business, specifically may be used
With with the positive number numerical value that sets as divisor, calculating that the sequence number of transaction is remmed, for example with 10 be except
The operation that several sequence numbers to transaction are taken the remainder, obtained remainder can be 0,1,2,3,4,5,6,
7,8,9, the transaction of remainder identical is classified as same class.Continue executing with next step S103.Due to transaction sequence
Number be according to transaction data reach transaction system sequentially, can by transaction data classification according to sequence number
Ensure the randomness of pending classification of business transaction.
Method two:The incident time of the friendship in parameter according to online transaction data, the buyer of transaction logs in
Place, in the buyer one day of dealing money or transaction, the number of times of transaction is by pending classification of business transaction or root
According to above-mentioned several parameters condition combination in any by pending classification of business transaction.It is specifically as follows:To conclude the business
The transaction that the time occurring belongs to scope set in advance is classified as the transaction of excessive risk class, and other are classified as low-risk class
Conclude the business, or the transaction that the place of buyer's login of transaction is belonged to scope set in advance is classified as excessive risk class
Transaction, other are classified as the transaction of low-risk class;Or the amount of money of pending transaction is exceeded the friendship of the threshold value of setting
Easily it is classified as the transaction of excessive risk class, other are classified as the transaction of low-risk class;Buyer's account again or by pending transaction
The transaction that the number of times being traded in the time of family one exceeds the threshold value setting is classified as the transaction of excessive risk class, other
It is classified as the transaction of low-risk class.Continue executing with next step S103.Such pending transaction data sorting technique
The specific aim of transaction risk assessment can be improved.
Step S103:Pending online Transaction Information after cluster is diverted to newly-built thing according to the algorithm setting
Business data processing model and existing transaction data process model are processed.
Pending online transaction data after cluster is diverted to newly-built transaction data process according to the algorithm setting
The method that model and existing transaction data process model are processed can have multiple, in the present embodiment with previous
In step S102 by the method for pending transaction cluster accordingly, by the pending online number of deals after cluster
There are two kinds according to being diverted to newly-built trading processing model and existing transaction data and processing the method that model processed.
Method one:For the pending transaction data after the clustering using method in step S102 one, press
According to the ratio setting, pending transaction data is diverted to newly-built transaction data respectively and processes model and existing
Transaction data processes model.For example according to 2:8 ratio, you can sequence number will be concluded the business to 10 remainders to specify
The transaction that number operation remainder is 1 and remainder is 4 is diverted to newly-built transaction data process model and carries out risk assessment;
Transaction sequence number is taken the remainder the transaction that operation remainder is 2,3,5,6,7,8,9 and remainder is 0 to 10 divide
Flow to existing transaction data process model and carry out risk assessment.Continue executing with next step S104.So can
Enough ensure that newly-built transaction data processes the randomness of the pending transaction of model evaluation, can neatly specify again
Newly-built transaction data processes the quantity of the transaction of model evaluation identification, it is to avoid newly-built transaction data processes model
Excessive negative impact can be caused to online transaction system.
Method two:Pending transaction data for the clustering using method in step S102 two it is intended that
The pending transaction data of excessive risk class had both entered newly-built transaction data and has processed model and simultaneously enter existing transaction
Data processing model carries out risk assessment.Low-risk pending transaction then can divide according to setting ratio respectively
Flow to newly-built transaction data process model or existing transaction data processes model and carries out risk assessment.Continue executing with
Next step S104.The transaction data of specified excessive risk class passes through newly-built transaction data simultaneously and processes model with now
There is transaction data to process model and carry out risk assessment, it is possible to increase the reliability of online trading system, reduce not
The infringement to system and client for the good transaction.
Step S104:According to the performance evaluation to pending item data result online for the above two model
Index calculates the performance difference of newly-built transaction data process model and existing transaction data process model.
In the present embodiment, described transaction data process model processes model, described performance evaluation for transaction data
Characteristic evidences are precision ratio and recall ratio data.Wherein precision ratio processes model evaluation for transaction data and just identifies
True pending online transaction data quantity processes the pending online transaction of model assessment with this transaction data
The ratio of data total amount.When transaction data processes the risk of model evaluation transaction, some transaction itself are to need really
The transaction to be identified by model evaluation and model is processed by transaction data correctly assess and identify out,
These transaction are transaction data and process model evaluation identification correctly pending transaction online, and these are concluded the business
Quantity divided by this transaction data process model evaluation pending online transaction data total amount just can be derived that this
Transaction data processes the precision ratio data of model.
The recall ratio that transaction data processes model is correctly waited to locate for transaction data process model evaluation identification online
Reason transaction data quantity and this transaction data process be actually needed in the transaction of model evaluation that assessment identifies
The ratio of line pending transaction data quantity.When transaction data processes the risk of model evaluation transaction, not institute
The evaluated transaction identifying in need can be identified by model evaluation, and correct assessment is identified
The number of transaction that the quantity of the transaction coming identifies divided by this model needs assessment can obtain this transaction data
Process the recall ratio data of model.
The precision ratio of model is processed according to transaction data and recall ratio data can compare the poor performance of different models
Different, in the present embodiment, look into standard according to what newly-built transaction data processed that model and existing transaction data process model
Rate and recall ratio data, compare and show whether the performance of new established model is better than existing model.Specifically can adopt
Method as described below:
Pending transaction data is diverted to newly-built trading processing model and existing transaction data processes mould online for adjustment
The ratio of type, precision ratio as described above and recall ratio computational methods calculate different split ratios respectively
The precision ratio of existing model and looking under the precision ratio of new established model and recall ratio data and different shunt ratios under example
Full rate data.Using the precision ratio under same for newly-built trading processing model shunt ratio and recall ratio data as one
Individual data pair, using the precision ratio under same for existing trading processing model shunt ratio and recall ratio data as one
Individual data pair, obtains precision ratio-recall ratio data pair that newly-built transaction data under different shunt ratios processes model
Process the precision ratio-recall ratio data pair of model with transaction data existing under different shunt ratios.
Coordinate axess with precision ratio and recall ratio as plane right-angle coordinate respectively, by described newly-built transaction data
Precision ratio-recall ratio the data processing under the different shunt ratios of model is plotted to described seat to as coordinate points
In mark system and these points are connected into line and obtain precision ratio-recall curve that newly-built transaction data processes model;
Using the precision ratio-recall ratio data under the different shunt ratios of described existing transaction data process model to as seat
Punctuate is plotted in described coordinate system and these points is connected into line and obtains existing transaction data process model
Precision ratio-recall curve.
Model is processed according to newly-built transaction data and existing transaction data processes the precision ratio-recall curve of model
Position relationship, the performance difference judging model can be assessed.It is specially range coordinate point (1,1) near
Model performance more excellent, if newly-built transaction data process model precision ratio-recall curve with respect to
Precision ratio-recall curve range points (1,1) that existing transaction data processes model are farther, then show newly-built
The performance that transaction data processes model is inferior to existing transaction data process model, should not be with this newly-built number of deals
Substitute existing transaction data process model according to processing model.Now can be by the shunting of pending transaction data
Ratio is adjusted to all pending transaction data and all passes through existing transaction data process model, thus this is newly-built
Transaction data processes offline the process accordingly of model safety and does not affect the service of transaction system.If new
Establish diplomatic relations easy data processing model precision ratio-recall curve with respect to existing transaction data process model look into standard
Rate-recall curve range points (1,1) are closer to then showing that newly-built transaction data processes the performance of model and is better than
Existing transaction data processes model, can issue newly-built transaction data and process at the model existing transaction data of replacement
Reason model.At this moment the shunt ratio of pending transaction data can be adjusted to all that pending transaction data is all
Model is processed by newly-built transaction data, thus will existing transaction data to process model safety offline and do not affect
The service of transaction system.
According to above two model, the Performance Evaluating Indexes of pending item data result online are calculated new
Build transaction data process model and the performance difference of existing transaction data process model, thus being newly-built number of deals
There is provided foundation according to the replacement of data processing model existing transaction data processing system.And can not affect
In the case of the service of line transaction system, safe one transaction data of underground cable any of which processes model.Ensure
The reliable and stable operation of system.
In the above-described embodiment, there is provided a kind of a kind of online affairs data processing model of the application is issued
Method, corresponding, the application also provides the device that a kind of online affairs data processing model is issued.
Fig. 2 is a kind of schematic diagram of the embodiment of device of online affairs data processing model issue of the application.By
It is substantially similar to embodiment of the method in device embodiment, so describing fairly simple, referring to side in place of correlation
The part of method embodiment illustrates.Device embodiment described below is only schematically.
Described device includes acquiring unit U201, taxon U202, dividing cell U203 and assessment unit
U204.
Described acquiring unit U201, for obtaining the relevant parameter of pending Transaction Information online.
Described acquiring unit U202, for being waited online to locate according to the relevant parameter of pending Transaction Information online
Director's business data clusters.
Described acquiring unit U203, for by cluster after pending online Transaction Information according to set algorithm
It is diverted to newly-built transaction data process model and existing transaction data process model is processed.
Described acquiring unit U204, for processing knot according to above two model to pending item data online
The Performance Evaluating Indexes of fruit calculate the performance of newly-built transaction data process model and existing transaction data process model
Difference.
The application also provides a kind of online affairs data processing method, schematic flow sheet such as Fig. 3 of embodiment
Shown, due to the online affairs data processing method of the application and the online affairs data processing model of the application
Step S101 in the embodiment of the method for real-time release is similar to the method for step S103, and here is only briefly
Bright online affairs data processing method.Each step correlation specific description is please respectively with reference to one kind of the application
Step S101, step S102 and step in the embodiment of the method for online affairs data processing model real-time release
S103.
Step S301:Obtain the relevant parameter of pending Transaction Information online.A kind of online with reference to the application
Step S101 in the embodiment of the method for transaction data process model real-time release.
Step S302:Relevant parameter according to pending Transaction Information online will pending Transaction Information gather online
Class.With reference to step in a kind of embodiment of the method for online affairs data processing model real-time release of the application
S102.
Step S303:Pending online Transaction Information after cluster is diverted to newly-built thing according to the algorithm setting
Business data processing model and existing transaction data process model are processed.A kind of online thing with reference to the application
Step S103 in the embodiment of the method for business data processing model real-time release.
Accordingly, the structured flowchart schematic diagram of the embodiment of a kind of online affairs data processing equipment of the application
As shown in figure 4, including data capture unit U401, data sorting unit U402, data distribution unit U403.
Described data capture unit U401, for obtaining the relevant parameter of pending Transaction Information online.
Described data sorting unit U402, will be online for the relevant parameter according to pending Transaction Information online
Pending Transaction Information cluster.
Described data distribution unit U403, for by cluster after pending online Transaction Information according to set
Algorithm is diverted to newly-built transaction data process model and existing transaction data process model is processed.
Although the application is open as above with preferred embodiment, it is not for limiting the application, Ren Heben
Skilled person, without departing from spirit and scope, can make possible variation and modification,
The protection domain of therefore the application should be defined by the scope that the application claim is defined.
In a typical configuration, computing device includes one or more processors (CPU), input/output
Interface, network interface and internal memory.
Internal memory potentially includes the volatile memory in computer-readable medium, random access memory
(RAM) and/or the form such as Nonvolatile memory, such as read only memory (ROM) or flash memory (flash RAM).
Internal memory is the example of computer-readable medium.
1st, computer-readable medium include permanent and non-permanent, removable and non-removable media can be by
Any method or technique is realizing information Store.Information can be computer-readable instruction, data structure, journey
The module of sequence or other data.The example of the storage medium of computer includes, but are not limited to phase transition internal memory
(PRAM), static RAM (SRAM), dynamic random access memory (DRAM), its
The random access memory (RAM) of his type, read only memory (ROM), electrically erasable is read-only deposits
Reservoir (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc read only memory (CD-ROM),
Digital versatile disc (DVD) or other optical storage, magnetic cassette tape, tape magnetic rigid disk stores or other
Magnetic storage apparatus or any other non-transmission medium, can be used for storing the information that can be accessed by a computing device.
Define according to herein, computer-readable medium does not include non-temporary computer readable media (transitory
Media), as data signal and the carrier wave of modulation.
2 it will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer
Program product.Therefore, the application using complete hardware embodiment, complete software embodiment or can combine software
Form with the embodiment of hardware aspect.And, the application can adopt and wherein include meter one or more
Calculation machine usable program code computer-usable storage medium (including but not limited to disk memory, CD-ROM,
Optical memory etc.) the upper computer program implemented form.
Claims (28)
1. a kind of method that online affairs data processing model is issued is it is characterised in that comprise the following steps:
Obtain the relevant parameter of pending Transaction Information online;
According to the relevant parameter of pending Transaction Information online, will pending Transaction Information cluster online;
Pending online Transaction Information after cluster is diverted to newly-built transaction data process according to the algorithm setting
Model and existing transaction data process model are processed;
According to above two model, the Performance Evaluating Indexes of pending item data result online are calculated new
Build transaction data process model and the performance difference of existing transaction data process model.
2. online affairs data processing model according to claim 1 is issued method it is characterised in that
The relevant parameter of described pending online Transaction Information include set forth below at least one:Pending online
The sequence number of Transaction Information, the pending affairs amount of money online, the place that pending transaction initiator logs in online,
Pending affairs occur online time or online pending transaction initiator account carried out online affairs in one day
Number of times.
3. online affairs data processing model according to claim 1 is issued method it is characterised in that
The relevant parameter of described basis pending Transaction Information online is by the method bag of the cluster of pending Transaction Information online
Include:
Relevant parameter according to pending Transaction Information online will pending Transaction Information gather uniformly at random online
Class.
4. online affairs data processing model according to claim 3 is issued method it is characterised in that
The relevant parameter of described basis pending Transaction Information online will pending Transaction Information gather uniformly at random online
The method of class includes:
The operation as divisor, the sequence number of pending Transaction Information online being remmed with the value setting, remainder
Pending Transaction Information is classified as same class to identical online.
5. online affairs data processing model according to claim 1 is issued method it is characterised in that
Described by cluster after pending online Transaction Information according to set algorithm be diverted to newly-built transaction data process
The method that model and existing transaction data process model are processed includes:
Set the ratio of pending Transaction Information shunting online;
Ratio according to the pending online Transaction Information shunting setting is by different classes of pending online affairs
Data is diverted to newly-built transaction data process model respectively and existing transaction data process model is processed.
6. online affairs data processing model according to claim 1 is issued method it is characterised in that
The relevant parameter of described basis pending Transaction Information online is by the method bag of the cluster of pending Transaction Information online
Include:
According to the relevant parameter of pending Transaction Information online, the pending online thing of the condition setting will be met
Business data is classified as excessive risk class pending Transaction Information online, and the condition being unsatisfactory for setting is pending online
Transaction Information is classified as low-risk class pending Transaction Information online;
Correspondingly,
Described by cluster after pending online Transaction Information according to set algorithm be diverted to newly-built Transaction Information
Process the method that model and existing transaction data process model processed to include:
Pending Transaction Information simultaneously enters newly-built transaction data process model and existing thing online to make excessive risk class
Business data processing model is processed.
7. online affairs data processing model according to claim 6 is issued method it is characterised in that
The condition of described setting includes:
The pending affairs amount of money exceeds the threshold value setting online, the place that pending transaction initiator logs in online
Belong to scope set in advance, the time that pending affairs occur online belongs to scope set in advance or online
Pending transaction initiator account carries out online affairs number of times in one day exceeds the threshold value setting.
8. online affairs data processing model according to claim 1 is issued method it is characterised in that
Described new to the Performance Evaluating Indexes calculating of pending item data result online according to above two model
The step building the performance difference of transaction data process model and existing transaction data process model includes:
Draw the precision ratio-recall curve of newly-built, existing transaction data process model respectively;
According to the precision ratio-recall curve of described newly-built, existing transaction data process model, compare and draw newly
Build transaction data process model and the performance difference of existing transaction data process model.
9. online affairs data processing model according to claim 8 is issued method it is characterised in that
The method of the described precision ratio-recall curve drawing newly-built, existing transaction data process model respectively includes:
The adjustment ratio that pending Transaction Information shunts online;
Ratio according to different pending online Transaction Information shuntings makes pending Transaction Information online respectively enter
Newly-built, existing transaction data process model;
Calculate difference respectively online at newly-built, existing Transaction Information in the case of pending Transaction Information shunt ratio
The precision ratio of reason model and recall ratio;
Set up a flat square using recall ratio and precision ratio as the coordinate axess of plane right-angle coordinate respectively to sit
Mark system, will be corresponding for newly-built transaction data process model under pending Transaction Information difference shunt ratio online
Recall ratio and precision ratio data are to as the different coordinate points in described plane right-angle coordinate;Treated online
Process the corresponding recall ratio of existing transaction data process model and precision ratio under Transaction Information difference shunt ratio
Data is to as the different coordinate points in described plane right-angle coordinate;
By transaction data process model newly-built in described plane right-angle coordinate and existing transaction data process model
Coordinate points be linked to be line respectively and obtain the precision ratio-recall curve of newly-built transaction data process model and existing
Precision ratio-the recall curve of transaction data process model.
10. the method that online affairs data processing model according to claim 8 is issued, its feature exists
In described precision ratio is:Transaction data process model evaluation identifies correctly pending Transaction Information number online
The ratio of the pending online Transaction Information total amount that amount is assessed with this transaction data process model;Described look into complete
Rate is:Transaction data process model evaluation is correctly online at pending Transaction Information quantity and this Transaction Information
It is actually needed the ratio of the pending online Transaction Information quantity that assessment identifies in the transaction of reason model evaluation.
The method that 11. online affairs data processing models according to claim 9 are issued, its feature exists
In, described precision ratio-recall curve according to described newly-built, existing transaction data process model, compares
The method going out the performance difference of newly-built transaction data process model and existing transaction data process model is:
In described plane right-angle coordinate, precision ratio according to newly-built, existing transaction data process model-look into
The distance of full rate curve distance point (1,1), draws to newly-built transaction data process model and existing thing
The performance difference of business data processing model.
A kind of 12. online affairs data processing methods are it is characterised in that comprise the steps:
Obtain the relevant parameter of pending Transaction Information online;
Relevant parameter according to pending Transaction Information online will pending Transaction Information cluster online;
Pending online Transaction Information after cluster is diverted to newly-built transaction data process according to the algorithm setting
Model and existing transaction data process model are processed.
13. online affairs data processing methods according to claim 12 it is characterised in that described
The relevant parameter of the pending Transaction Information of line includes:The sequence number of pending Transaction Information, pending online online
The affairs amount of money, the place that pending transaction initiator logs in online, the time that pending affairs occur online or
Pending transaction initiator account carries out the number of times of online affairs in one day online.
14. online affairs data processing methods according to claim 12 are it is characterised in that described
The method of pending Transaction Information cluster online is included by the relevant parameter according to pending Transaction Information online:
Relevant parameter according to pending Transaction Information online will pending Transaction Information gather uniformly at random online
Class.
15. online affairs data processing methods according to claim 14 are it is characterised in that described
The method of pending Transaction Information cluster online is included by the relevant parameter according to pending Transaction Information online:
The operation as divisor, the sequence number of pending Transaction Information online being remmed with the value setting, remainder
Pending Transaction Information is classified as same class to identical online.
16. online affairs data processing methods according to claim 12 will be it is characterised in that described will
Pending online Transaction Information after cluster according to set algorithm be diverted to newly-built transaction data process model and
The method that existing transaction data process model is processed includes:
Set the ratio of pending Transaction Information shunting online;
Ratio according to the pending online Transaction Information shunting setting is by different classes of pending online affairs
Data is diverted to newly-built transaction data process model respectively and existing transaction data process model is processed.
17. online affairs data processing methods according to claim 12 are it is characterised in that described
The method of pending Transaction Information cluster online is included by the relevant parameter according to pending Transaction Information online:
According to the relevant parameter of pending Transaction Information online, the pending online thing of the condition setting will be met
Business data is classified as excessive risk class pending Transaction Information online, and the condition being unsatisfactory for setting is pending online
Transaction Information is classified as low-risk class pending Transaction Information online;
Correspondingly,
Described by cluster after pending online Transaction Information according to set algorithm be diverted to newly-built Transaction Information
Process the method that model and existing transaction data process model processed to include:
Pending Transaction Information simultaneously enters newly-built transaction data process model and existing thing online to make excessive risk class
Business data processing model is processed.
18. online affairs data processing methods according to claim 17 are it is characterised in that described set
Fixed condition includes:
The pending affairs amount of money exceeds the threshold value setting online, the place that pending transaction initiator logs in online
Belong to scope set in advance, the time that pending affairs occur online belongs to scope set in advance or online
Pending transaction initiator account carries out online affairs number of times in one day exceeds the threshold value setting.
The device that a kind of 19. online affairs data processing models are issued, including:
Acquiring unit, for obtaining the relevant parameter of pending Transaction Information online;
Taxon, will pending number of transactions online for the relevant parameter according to pending Transaction Information online
According to cluster;
Dividing cell, for being diverted to newly the pending online Transaction Information after cluster according to the algorithm setting
Build transaction data process model and existing transaction data process model is processed;
Assessment unit, for the performance to pending item data result online according to above two model
Evaluation index calculates the performance difference of newly-built transaction data process model and existing transaction data process model.
The device that 20. online affairs data processing models according to claim 19 are issued, its feature exists
In described taxon, specifically for according to the relevant parameter of pending Transaction Information online uniformly at random
To pending Transaction Information cluster online.
The device that 21. online affairs data processing models according to claim 20 are issued, its feature exists
In, described taxon, specifically for the sequence number to pending Transaction Information online with the value that sets as divisor
The operation being remmed, pending Transaction Information is classified as same class to remainder identical online.
The device that 22. online affairs data processing models according to claim 19 are issued, its feature exists
In described dividing cell includes:
Shunt ratio sets subelement, for setting the ratio of pending Transaction Information shunting online;
Shunting performance element, for the ratio according to the pending online Transaction Information shunting setting by inhomogeneity
Other pending online Transaction Information is diverted at newly-built transaction data process model and existing Transaction Information respectively
Reason model is processed.
The device that 23. online affairs data processing models according to claim 19 are issued, its feature exists
In,
Described taxon, specifically for according to the relevant parameter of pending Transaction Information online, meeting
The pending online Transaction Information of the condition setting is classified as excessive risk class pending Transaction Information online, will be discontented with
The pending online Transaction Information of the condition that foot sets is classified as low-risk class pending Transaction Information online;
Correspondingly,
Described dividing cell is specifically for pending Transaction Information simultaneously enters newly-built thing online to make excessive risk class
Business data processing model and existing transaction data process model are processed.
The device that 24. online affairs data processing models according to claim 23 are issued, its feature exists
In described taxon, specifically for:The pending affairs amount of money online is exceeded the threshold value setting, treats online
The place processing transaction initiator login belongs to scope set in advance, the time that pending affairs occur online
Belong to scope set in advance or online pending transaction initiator account carry out the number of times of online affairs in one day
Pending online Transaction Information beyond the threshold value setting is classified as excessive risk class pending Transaction Information online.
The device that 25. online affairs data processing models according to claim 19 are issued, its feature exists
In described assessment unit includes:
Curve plotting subelement, for drawing the precision ratio of newly-built, existing transaction data process model-look into respectively
Full rate curve;
Difference assesses subelement, for the precision ratio according to described newly-built, existing transaction data process model-look into
Full rate curve, compares the poor performance drawing newly-built transaction data process model and existing transaction data process model
Different.
The device that 26. online affairs data processing models according to claim 25 are issued, its feature exists
In described curve plotting subelement includes:
Shunt ratio adjusts subelement, for adjusting the ratio of pending Transaction Information shunting online;
Shunting execution subelement, for making to treat online according to the ratio of different pending online Transaction Information shuntings
Process Transaction Information and respectively enter newly-built, existing transaction data process model;
Computation subunit, for calculate respectively difference newly-built in the case of pending Transaction Information shunt ratio online,
The precision ratio of existing transaction data process model and recall ratio;
Draw the first subelement, for respectively using recall ratio and precision ratio as the coordinate of plane right-angle coordinate
Axle sets up a plane right-angle coordinate, will newly-built affairs under pending Transaction Information difference shunt ratio online
The corresponding recall ratio of data processing model and precision ratio data to as in described plane right-angle coordinate not
Same coordinate points;To existing transaction data process model phase under the different shunt ratio of pending Transaction Information online
Corresponding recall ratio and precision ratio data are to as the different coordinate points in described plane right-angle coordinate;
Draw the second subelement, for by transaction data process model newly-built in described plane right-angle coordinate and
What the coordinate points of existing transaction data process model were linked to be that line obtains newly-built transaction data process model respectively looks into standard
Rate-recall curve and the precision ratio-recall curve of existing transaction data process model.
The device that 27. online affairs data processing models according to claim 25 are issued, its feature exists
In described difference assesses subelement, specifically in described plane right-angle coordinate, according to newly-built, existing
There is the distance of the precision ratio-recall curve range points (1,1) of affairs data processing model, it is right to draw
Newly-built transaction data process model and the performance difference of existing transaction data process model.
A kind of 28. online affairs data processing equipments, including:
Data capture unit, for obtaining the relevant parameter of pending Transaction Information online;
Data sorting unit, will pending thing online for the relevant parameter according to pending Transaction Information online
Business data clusters;
Data distribution unit, for shunting the pending online Transaction Information after cluster according to the algorithm setting
Processed to newly-built transaction data process model and existing transaction data process model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510530530.1A CN106485560A (en) | 2015-08-26 | 2015-08-26 | The method and apparatus that a kind of online affairs data processing model is issued |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510530530.1A CN106485560A (en) | 2015-08-26 | 2015-08-26 | The method and apparatus that a kind of online affairs data processing model is issued |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106485560A true CN106485560A (en) | 2017-03-08 |
Family
ID=58233871
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510530530.1A Pending CN106485560A (en) | 2015-08-26 | 2015-08-26 | The method and apparatus that a kind of online affairs data processing model is issued |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106485560A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111126621A (en) * | 2019-12-17 | 2020-05-08 | 北京九章云极科技有限公司 | Online model training method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070112715A1 (en) * | 2005-11-07 | 2007-05-17 | Nec Laboratories America | System failure detection employing supervised and unsupervised monitoring |
CN101504745A (en) * | 2008-12-04 | 2009-08-12 | 阿里巴巴集团控股有限公司 | Risk regulation/model establishing and optimizing system and method based on network on-line service |
CN103973589A (en) * | 2013-09-12 | 2014-08-06 | 哈尔滨理工大学 | Network traffic classification method and device |
-
2015
- 2015-08-26 CN CN201510530530.1A patent/CN106485560A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070112715A1 (en) * | 2005-11-07 | 2007-05-17 | Nec Laboratories America | System failure detection employing supervised and unsupervised monitoring |
CN101504745A (en) * | 2008-12-04 | 2009-08-12 | 阿里巴巴集团控股有限公司 | Risk regulation/model establishing and optimizing system and method based on network on-line service |
CN103973589A (en) * | 2013-09-12 | 2014-08-06 | 哈尔滨理工大学 | Network traffic classification method and device |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111126621A (en) * | 2019-12-17 | 2020-05-08 | 北京九章云极科技有限公司 | Online model training method and device |
CN111126621B (en) * | 2019-12-17 | 2021-02-09 | 北京九章云极科技有限公司 | Online model training method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105512799B (en) | Power system transient stability evaluation method based on mass online historical data | |
CN110111113B (en) | Abnormal transaction node detection method and device | |
CN107622432A (en) | Trade company's evaluation method and system | |
CN103366123B (en) | Software hazard appraisal procedure based on defect analysis | |
CN105354210A (en) | Mobile game payment account behavior data processing method and apparatus | |
Li et al. | Research and application of random forest model in mining automobile insurance fraud | |
CN110874744B (en) | Data anomaly detection method and device | |
CN112860841A (en) | Text emotion analysis method, device and equipment and storage medium | |
CN103929330A (en) | Domain name service quality evaluation method and system | |
CN109299846A (en) | A kind of nuclear power plant equipment analysis method for reliability, system and terminal device | |
CN113177585B (en) | User classification method, device, electronic equipment and storage medium | |
Bayat et al. | Exchange rates and foreign exchange reserves in Turkey: nonlinear and frequency domain causality approach | |
CN115222303B (en) | Industry risk data analysis method and system based on big data and storage medium | |
CN104320271B (en) | A kind of network equipment safety evaluation method and device | |
CN105574666A (en) | Method and device for evaluating credit level of enterprise based on key data modeling | |
CN109583731A (en) | A kind of Risk Identification Method, device and equipment | |
CN108845177A (en) | A kind of lightning stroke monitoring for power distribution network and recognition methods, device and equipment | |
CN114139931A (en) | Enterprise data evaluation method and device, computer equipment and storage medium | |
CN107862205A (en) | One kind assesses accurate information security risk evaluation system | |
CN109977131A (en) | A kind of house type matching system | |
CN113487241A (en) | Method, device, equipment and storage medium for classifying enterprise environment-friendly credit grades | |
CN106485560A (en) | The method and apparatus that a kind of online affairs data processing model is issued | |
CN105224954A (en) | A kind of topic discover method removing the impact of little topic based on Single-pass | |
CN112579847A (en) | Method and device for processing production data, storage medium and electronic equipment | |
Wang et al. | Temperature forecast based on SVM optimized by PSO algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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: 20170308 |