CN109299161A - A kind of data selecting method and device - Google Patents
A kind of data selecting method and device Download PDFInfo
- Publication number
- CN109299161A CN109299161A CN201811286327.4A CN201811286327A CN109299161A CN 109299161 A CN109299161 A CN 109299161A CN 201811286327 A CN201811286327 A CN 201811286327A CN 109299161 A CN109299161 A CN 109299161A
- Authority
- CN
- China
- Prior art keywords
- data
- sample
- data side
- training
- evaluation index
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
Abstract
This specification embodiment provides a kind of data selecting method and device, wherein method may include: to enter moding amount and label, training machine learning model according in training sample；Training sample further includes not entering moding amount；The moding amount input machine learning model that enters in test sample is obtained into predicted value；Test sample further includes label；According to the label and predicted value of test sample, the corresponding residual error of test sample is obtained；Residual error is respectively sent at least two second data sides, so that each second data side is respectively using the second data regression regression criterion possessed, and obtains returning evaluation index；The recurrence evaluation index that at least two second data sides return respectively is received, by comparing the second data side of recurrence evaluation index selected section of at least two second data sides.
Description
Technical field
This disclosure relates to big data technical field, in particular to a kind of data selecting method and device.
Background technique
With the rapid development of Internet technology, entire society is forcibly pushed into " big data " epoch.No matter whether people
It is ready, our personal data are just inadvertently passively collected and used by enterprise, individual.The networking of personal data and
Transparence has become irresistible main trend.At the same time, user data is also dangerous " box of Pandora ", data one
Denier leakage, the privacy of user will be invaded.In recent years, a lot of privacy of user leakage events are had occurred that, the individual's of citizen
Private data guard encounters stern challenge.Big data bring globality is changed, so that individual consumer is difficult confrontation individual
The risk that privacy is exposed comprehensively.In face of the privacy leakage event to take place frequently, Privacy Protection needs to obtain effective solution.
In practical business, we are likely encountered such scene: needing to come by the variable data of third party's channel
The effect of existing model is promoted, only when these data, which model us, understands helpful, just buys corresponding third number formulary
According to.It would therefore be desirable to judge its validity in advance in the case where not obtaining third party's data, and in this process not
The private data of our user can be revealed.
Summary of the invention
In view of this, this specification one or more embodiment provides a kind of data selecting method and device, by more
Internal data privacy is protected in a external data while selected section data.
Specifically, this specification one or more embodiment is achieved by the following technical solution:
In a first aspect, providing a kind of data selecting method, the method is applied to by at least two second of offer data
The second data side of selected section in data side；The method is executed by the first data side, the first data side possess first
Data include: the training set and test set of machine learning model；The training set includes multiple training samples, the test set packet
Include multiple test samples；The described method includes:
Enter moding amount and label, the training machine learning model according in training sample；The training sample also wraps
Include have neither part nor lot in machine learning model training do not enter moding amount；
It moding amount will be entered described in the test sample inputs the machine learning model to obtain predicted value；The test
Sample further includes label, the expection predicted value for entering moding amount input machine learning model of the tag representation test sample；
According to the label of the test sample and the predicted value, the corresponding residual error of the test sample is obtained；
The residual error is respectively sent at least two second data side, so that each second data side makes respectively
It is fitted the residual error with the second data regression possessed, and obtains returning evaluation index；
The recurrence evaluation index that at least two second data side returns respectively is received, by comparing described at least two
The second data side of recurrence evaluation index selected section of a second data side.
Second aspect, provides a kind of verification method of data validity, and the method is executed by the second data side, comprising:
The residual error of the first data side transmission is received, the residual error is that the first data root according in test sample enters moding amount
The label of predicted value and test sample that input machine learning model obtains obtains；The data packet that the first data side possesses
Include: training set and test set, the training set include multiple training samples, and the test set includes multiple test samples；It is described
Machine learning model be according in training sample enter moding amount and label training obtains；It further include not entering in the training sample
Moding amount；
The sample identification of the first data side transmission is received, and sample matches are carried out according to the sample identification and are obtained for joining
With the second data of regression fit；
It is fitted the residual error based on second data regression, obtains returning evaluation index；
The recurrence evaluation index is returned into the first data side so that the first data side by comparing it is to be selected extremely
The second data side of recurrence evaluation index selected section of few two the second data sides.
The third aspect, provides a kind of verifying device of data validity, and described device is applied to by offer data at least
The second data side of selected section in two the second data sides；Described device is applied to the first data side, and the first data side is gathered around
The first data having include: the training set and test set of machine learning model；The training set includes multiple training samples, described
Test set includes multiple test samples；Described device includes:
Model training module, for entering moding amount and label, the training machine learning according in the training sample
Model；The training sample further include have neither part nor lot in machine learning model training do not enter moding amount；
Model prediction module inputs the machine learning model for will enter moding amount described in the test sample and obtains
To predicted value；The test sample further includes label, and the tag representation test sample enters moding amount input machine learning mould
The expection predicted value of type；
Residual computations module, for according to the test sample label and the predicted value, it is corresponding to obtain test sample
Residual error；
Data transmission blocks, for the residual error to be respectively sent at least two second data side, so that respectively
A second data side is fitted the residual error using the second data regression possessed respectively, and obtains returning evaluation index；
Verification processing module, the recurrence evaluation index returned respectively for receiving at least two second data side, with
By comparing the second data side of recurrence evaluation index selected section of at least two second data side.
Fourth aspect, provides a kind of verifying device of data validity, and described device is applied to the second data side, the device
Include:
Residual error receiving module, for receiving the residual error of the first data side transmission, the residual error is the first data root according to survey
The predicted value that moding amount input machine learning model obtains of entering in sample sheet and label obtain；What the first data side possessed
Data include: training set and test set, and the training set includes multiple training samples, and the test set includes multiple test specimens
This；The machine learning model be according in training sample enter moding amount and label training obtains；In the training sample also
Including not entering moding amount；
Data match module, the sample identification sent for receiving the first data side, and according to the sample identification
It carries out sample matches and obtains the second data for participating in regression fit；
Processing module is returned, for being fitted the residual error based on second data regression, obtains returning evaluation index；
Index feedback module returns to the first data side for that will return evaluation index, so that the first data side passes through
Compare the second data side of recurrence evaluation index selected section of at least two second data sides to be selected.
5th aspect, provides a kind of verifying equipment of data validity, the equipment includes memory, processor and storage
On a memory and the computer program that can run on a processor, the processor realize following step when executing described program
It is rapid:
Enter moding amount and label, training machine learning model according in training sample；The training sample further includes not
Participate in machine learning model training does not enter moding amount；
It moding amount will be entered described in the test sample inputs the machine learning model to obtain predicted value；The test
Sample further includes label, the expection predicted value for entering moding amount input machine learning model of the tag representation test sample；
According to the label of the test sample and the predicted value, the corresponding residual error of the test sample is obtained；
The residual error is respectively sent at least two second data side, so that each second data side makes respectively
It is fitted the residual error with the second data regression possessed, and obtains returning evaluation index；
The recurrence evaluation index that at least two second data side returns respectively is received, by comparing described at least two
The second data side of recurrence evaluation index selected section of a second data side.
6th aspect, provides a kind of verifying equipment of data validity, the equipment includes memory, processor and storage
On a memory and the computer program that can run on a processor, the processor realize following step when executing described program
It is rapid:
The residual error of the first data side transmission is received, the residual error is that the first data root according in test sample enters moding amount
The label of predicted value and test sample that input machine learning model obtains obtains；The data packet that the first data side possesses
Include: training set and test set, the training set include multiple training samples, and the test set includes multiple test samples；It is described
Machine learning model be according in training sample enter moding amount and label training obtains；It further include not entering in the training sample
Moding amount；
The sample identification that the first data side is sent is received, and sample matches are carried out according to the sample identification and are used
In the second data for participating in regression fit；
It is fitted the residual error based on second data regression, obtains returning evaluation index；
The recurrence evaluation index is returned into the first data side so that the first data side by comparing it is to be selected extremely
The second data side of recurrence evaluation index selected section of few two the second data sides.
The data selecting method and device of this specification one or more embodiment pass through two data side's interactive modelings
Residual sum returns evaluation index, and the private data of non-user, therefore appointing for user can not be revealed in both sides' interactive process
What private data.Also, it can also be according to the recurrence evaluation index that multiple data sides return by selected section in multiple data sides
Data protect internal data privacy while realizing the selected section data in by multiple external datas.
Detailed description of the invention
In order to illustrate more clearly of this specification one or more embodiment or technical solution in the prior art, below will
A brief introduction will be made to the drawings that need to be used in the embodiment or the description of the prior art, it should be apparent that, it is described below
Attached drawing is only some embodiments recorded in this specification one or more embodiment, and those of ordinary skill in the art are come
It says, without any creative labor, is also possible to obtain other drawings based on these drawings.
Fig. 1 is the data set schematic diagram that this specification one or more embodiment provides；
Fig. 2 is the data selecting method that this specification one or more embodiment provides；
Fig. 3 is a kind of data selection means that this specification one or more embodiment provides；
Fig. 4 is another data selection means that this specification one or more embodiment provides.
Specific embodiment
In order to make those skilled in the art more fully understand the technical solution in this specification one or more embodiment,
Below in conjunction with the attached drawing in this specification one or more embodiment, to the technology in this specification one or more embodiment
Scheme is clearly and completely described, it is clear that described embodiment is only a part of the embodiment, rather than whole realities
Apply example.Based on this specification one or more embodiment, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present application.
In practical business, be likely encountered such scene: data side A possesses own data, it is desirable to if evaluation and test
By the data of data side B, the modelling effect of itself can be promoted.For example, it is assumed that data side A utilizes owned number
It according to a machine learning model M is had trained, still, is found in model measurement, the prediction effect of the model is not satisfactory, and pre-
Phase predicted value has a certain distance.If participating in the training and optimization of model M using the data of data side B, mould can be made
The effect of type M is promoted, then can choose the data of purchase data side B to assist modeling.
In above-mentioned scene, it is involved in a problem i.e.: how to determine whether data side B is effective, if data side B
Data it is helpful to the modeling of model M, facilitate the effect of lift scheme M, then confirm that the data of data side B are effective.
And the data validity of which kind of mode verify data side B is used, it will be at least one embodiment of this specification content to be described,
Also, in the verification method of data validity, will realize: data side A does not obtain the data of data side B, and data side A is not let out
That reveals itself possesses data.
As follows by taking data side A and data side B as an example, the verification method of data validity is described, and this method will verify number
It is whether effective according to the data of square B.
For example, data side A can be known as to the first data side, data side B is known as the second data side.
Firstly, shown in Figure 1, the data that the first data side possesses are properly termed as the first data.In first data
It may include: the training set and test set of machine learning model.
Wherein, training set is used for the training of machine learning model, for example, the training sample D in the training setA(XA, YA)
In, XAIt is variable, YAIt is label.The label YAIndicate above-mentioned variable XABy the expection predicted value of the machine learning model,
It is equivalent to a kind of model for having supervision.
Test set is used for the prediction of machine learning model, for example, the test sample D in test setB(XB, YB) equally include
Variable and label.
For example, the variable of above-mentioned training sample and test sample, it can include " entering moding amount " and " do not enter moding
Amount ".Wherein, the training for entering moding amount and taking part in model in training sample, and the moding amount that enters in test sample takes part in mould
Type prediction, and do not enter training and prediction that moding amount is not engaged in model.
Be exemplified below: for judging that some user is high-quality user or user inferior, which can use multiple changes
Amount indicates, for example, age, address, length of service, annual income etc..Assuming that a user can be indicated with 8 variables, U f1,
F2, f3, f4 ... .f8 } it be a user U include this eight variables of f1 to f8.It, can be first using wherein in training pattern
Five variable f1 to f5, and f6 to f8 is temporarily first not involved in the training of model.
So, in training sample DA(XA, YA) in, it may include multiple user's samples, for example, user U1, user U2, use
Family U3 etc..Each user's sample is DA(XA, YA), including variable and label, variable X thereinAIt may include above-mentioned use
Five variable f1 to f5 at family, the variable in each user's sample are this five variables, and variate-value can be different；And it is described
Label YACan be the user is high-quality user or user inferior, for example, high-quality user is indicated with 11,00 table of user inferior
Show.
The test sample D of prediction for machine learning modelB(XB, YB) it equally include variable and label, carrying out model
When prediction, DBThe variable used includes five variable f1 to f5 of user, and f6 to f8 has neither part nor lot in prediction, and label is that the user is excellent
Matter user or user inferior.Test set is the moding amount that enters of test sample to be inputted trained model, and sentence in prediction
Whether the output result of disconnected model is consistent with label.
By 1 example training sample of table, test sample and therein it can enter moding amount and do not enter moding amount as follows.
As shown in table 1, these samples of U1, U2 and U3 will participate in the training of model, be properly termed as training set.But participating in model instruction
When practicing, f1 only therein to f5 variable is participated in, and is properly termed as into moding amount, and f6 is temporarily not engaged in model to f8 variable and instructs
Practice, does not enter moding amount referred to as.YAIt is label.For another example, these samples of the U7 in test set and U8 are the predictions for model, will
The moding amount that enters in these test samples inputs trained model, and obtains the output result of model.Likewise, U7 and U8 exist
When input model, and only f1 is participated in f5 variable, and f6 to f8 variable has neither part nor lot in.If the following table 1 is only example, in actual implementation
It is not limited thereto, the variable for including in each sample can change.
1 first data D of tableA(XA, YA)
In the above example, when testing using the test sample in table 1 model, the effect of model is found less
Ideal, at this time, it is assumed that have at least two second data sides such as data side B that can provide data, this multiple data side B can possess
Different variables, or possess the different variate-values of identical variable.Can by selected section data side B in multiple data side B Lai
Assist Optimized model.For example, can be by one optimal data side B of selection in three data side B, or also can choose two
Or multiple data side B, decision can be considered according to actual business requirement.And it is more excellent which how to be assessed in these three data sides B,
The data selecting method of at least one embodiment of this specification then can be used.
Fig. 2 describes the data selecting method of at least one embodiment of this specification offer, and this method may include as follows
It handles, do not limit each step in specific implementation executes sequence:
In step 200, according to training sample, training machine learning model.
This step, which can be used in training sample, enters moding amount and label training pattern.For example, can be in table 1
The data training pattern of U1, U2 and U3, U1, U2 and U3 therein are user's samples, and each user's sample may include eight changes
Amount, and in training, five variables of f1 therein to f5 can be used.
In step 202, the moding amount input machine learning model that enters in test sample is obtained into predicted value.
For example, the test sample U7 and U8 in table 1 are not engaged in the training of model, but it can be used for the test of model.
It in the model that training is completed in input step 200, can be obtained using five variables of f1 to f5 in test sample as inputting
Model exports result, that is, predicted value.The moding amount that enters of tag representation test sample in the test sample inputs machine learning mould
The expection predicted value of type.
In step 204, according to the label in predicted value and test sample, the corresponding residual error of the test sample is obtained.
For example, the corresponding label of U7 and U8 is the Y in table 1A7And YA8, and residual error can be the difference between predicted value and label, the residual error
It can be used to indicate that the difference between the reality output result of model and desired output result, so as to for measuring model
Prediction effect.
In step 206, the residual error is sent respectively at least two second data sides to be selected by data side A.This
The corresponding residual error of the test sample of data side A can be sent to data side B by step, also that training sample and test sample is corresponding
Sample identification be sent to data side B.For example, the sample identification may include the User ID of U1 to U3.
For example, User ID can be encrypted by Encryption Algorithm such as MD5, to avoid user information leakage.What is transmitted is residual
Difference is that the gap between original tag is measured, and can also play the purpose of protection privacy of user.
Wherein, it should be noted that in following step, in step 206 to step 214, with data side A to two data
It is described for square B, can there is greater number of data side B in actual implementation.In the signal of Fig. 2, to two data side B
Sample matches and the regression fit processing that transmission sample identification and residual error and the two data sides B are respectively carried out, have used phase
Same step serial number, for example, being all step 206, however, it will be understood that this is that two data side B are respectively executed
Operation.
In a step 208, data side B carries out sample matches according to the sample identification, obtains the second data.
For example, data side B can carry out sample matches according to the User ID of U1 and U3, obtains and intend for participating in subsequent return
The second data closed.For example, may refer to above-mentioned table 2, the data of U1 and U3 that data side B possesses are obtained, and obtain variable
F9 to f11.
May include in second data corresponding data side A training sample and test sample sample ID data.It can
User's sample of the sample identification of the training sample of corresponding data side A is also referred to as the training sample in data side B, will correspond to
User's sample of the sample identification of the test sample of data side A is known as the test sample in data side B.
In step 210, data side B is fitted the residual error based on the variable regression in second data, is returned
Evaluation index.
For example, multiple user's samples in test sample, each sample can correspond to a residual error, and multiple samples can
To obtain multiple residual errors.Each variable regression in the training sample of data side B can be used and be fitted above-mentioned multiple residual errors.It is quasi-
The purpose of conjunction is to fit a polynomial function according to training sample, this function can be good at being fitted above-mentioned
Multiple residual errors.
For example, it is assumed that above-mentioned multiple residual errors may include y1、y2……yn.Wherein, n is natural number.
Variable in each training sample may include: x1、x2……xi.Wherein, i is natural number.
y1=a1*x11+a2*x12+…….ai*x1i；……(1)
y2=a1*x21+a2*x22+…….ai*x2i；……(2)
……………
yn=a1*xn1+a2*xn2+…….ai*xni；……(n)
Wherein, each residual error y1To ynBe it is known, the value of the variable in each training sample is also known, for example,
{ x in above-mentioned formula (1)11、x12……x1nBe each variable in a training sample value, { the x in formula (2)21、
x22……x2nBe each variable in another training sample value.It, can be with by above-mentioned formula (1) to formula (n)
Obtain coefficient a1、a2……aiValue, finally obtain regression equation y=a1*x1+a2*x2+…….ai*xi。
The corresponding variable importance weight of the available each variable of the regression equation acquired, above-mentioned a1、a2……
aiValue be the corresponding variable importance weight of each variable.
It should be noted that above-mentioned citing is by taking linear regression as an example, however, it is not limited to this.Other can also be used
Recurrence mode, e.g., polynomial regression.
Also, the recurrence evaluation index of this recurrence can also be calculated.Return evaluation index can there are many, for example, can
To be mean square error, root-mean-square error (Root Mean Squard Error, RMSE), mean absolute error etc..Return evaluation
Index can be used for measuring the effect of regression fit.
For example, returning evaluation index by taking mean square error as an example:
In formula (5), m indicates the quantity of test sample, yiIndicate true value, ynIndicate predicted value, true value and prediction
Value makes the difference, then square after sum-average arithmetic.For example, for each test sample, the corresponding residual error of each test sample, with
For one of test sample, the corresponding residual error of the test sample is exactly true value, and uses the variable in the test sample
Value substitute into regression equation obtained above, obtained residual values are exactly predicted value.According to above-mentioned formula (5), to each survey
The true value and predicted value of sample sheet make the difference, and square after sum-average arithmetic, it can obtain return evaluation index mean square error.
In the step 212, the second data side returns to the first data side for evaluation index is returned.It is described in this step
The recurrence evaluation index that oneself is calculated can be returned to data side A respectively by least two data side B.
In addition, data side B can also obtain at least one of following parameter: sample matches rate and the variable missing of the second data
Rate.Wherein, the sample matches rate, which can be understood as data side B, can find the data that the data side A of much ratios is required,
For example, the sample identification that data side A is transmitted to data side B there are eight, that is, data side B is required to provide user's sample of eight users.
And data side B only has 6, then sample matching rate can be 6/8*100%=75%.The variable miss rate is understood that
Are as follows: data side B can find some variable of data side A requirement, only some missings of variate-value.For example, the data side side B has 10
The data of a user's sample, all there are also variable f10 for this 10 user's samples, but wherein there are two variable of the user at f10
Value is sky, that is, variable missing occurs, variable miss rate can be 20%.
Data side B can will return evaluation index and return to data side A, can also lack the sample matches rate and variable
At least one of mistake rate returns to data side A, so that the first data side is in conjunction with recurrence evaluation index, the sample matches rate
The selection of data side is carried out with variable miss rate.
In step 214, the first data side is selected by comparing the recurrence evaluation index of multiple second data sides to determine
The second data side of part.
In this step, data side A can be individually according to the comparison for returning evaluation index, for example, can be by two data side B
The recurrence evaluation index of return compares, which index is more excellent just to select by which data side B.It can certainly select to return and evaluate
The preferably multiple data side B of index.
Alternatively, sample matches rate, variable miss rate can also be comprehensively considered and return evaluation index, for example, can first select
The second data side that sample matches rate is higher than preset threshold is selected out, matching rate is lower to be given up.It is high by sample matches rate again
The selective goal preferably data side B in the second data of preset threshold is ranked up for example, evaluation index can will be returned,
Data side B of the selected and sorted at former.Certainly, in other examples, can also comprehensively consider again variable miss rate etc. its
His index.For example, can be sample matches rate and sample miss rate given threshold, no matter the second data recurrence lower than threshold value is commented
How is valence index, not reselection.
A variety of regression algorithms can be used in above-mentioned recurrence, but are the need to ensure that each data side B uses unification
Regression algorithm and unified recurrence evaluation index, avoid due to each data side B select Different Effects subsequent contrast it is just.
In addition, the judgement of the data validity of this step, can be computer automatic execution, it is also possible to manually perform,
For example, data side B by sample matches rate, sample miss rate and is being returned after evaluation index returns to data side A, by data side A
Administrative staff judged according to these indexs returned, to carry out the selection of data side B.
The residual error of modeling is only sent to by the data selecting method of this specification one or more embodiment, data side A
Multiple data side B, multiple data side B also will only return evaluation index and return to data side A, and the interaction of data side is that modeling is residual
Difference and recurrence evaluation index, and the private data of non-user, therefore any of user can not be revealed in both sides' interactive process
Private data.Also, it can also be according to the recurrence evaluation index that multiple data side B are returned by selected section in multiple data side B
Data protect internal data privacy while realizing the selected section data in by multiple external datas.
Fig. 3 is the data selection means that at least one embodiment of this specification provides, and described device is applied to by offer number
According at least two second data sides in the second data side of selected section；Described device be applied to the first data side, described first
The first data that data side possesses include: the training set and test set of machine learning model；The training set includes multiple training
Sample, the test set include multiple test samples.As shown in figure 3, the apparatus may include: model training module 31, model
Prediction module 32, residual computations module 33, data transmission blocks 34 and verification processing module 35.
Model training module 31, for entering moding amount and label, the training engineering according in the training sample
Practise model；The training sample further include have neither part nor lot in machine learning model training do not enter moding amount.
Model prediction module 32 inputs the machine learning model for will enter moding amount described in the test sample
Obtain predicted value；The test sample further includes label, and the tag representation test sample enters the input machine learning of moding amount
The expection predicted value of model.
Residual computations module 33, for according to the test sample label and the predicted value, obtain test sample pair
The residual error answered.
Data transmission blocks 34, for the residual error to be respectively sent at least two second data side, so that
Each second data side is fitted the residual error using the second data regression possessed respectively, and obtains returning evaluation index；
Verification processing module 35, the recurrence evaluation index returned respectively for receiving at least two second data side,
By comparing the second data side of recurrence evaluation index selected section of at least two second data sides.
In one example, verification processing module 35 is also used to receive the sample matches rate of the second data side return；By sample
This matching rate is higher than in the second data of preset threshold, according to recurrence evaluation index by selector at least two second data sides
The second data for dividing the second data side to possess.
Fig. 4 is another data selection means that provide of at least one embodiment of this specification, and described device is applied to the
Two data sides, as shown in figure 4, the apparatus may include: residual error receiving module 41, returns processing module at data match module 42
43 and index feedback module 44.
Residual error receiving module 41, for receiving the residual error of the first data side transmission, the residual error is the first data root evidence
The predicted value that moding amount input machine learning model obtains of entering in test sample and label obtain；The first data side possesses
Data include: training set and test set, the training set includes multiple training samples, and the test set includes multiple test specimens
This；The machine learning model be according in training sample enter moding amount and label training obtains；In the training sample also
Including not entering moding amount.
Data match module 42, for receiving the corresponding sample identification of the training sample, and according to the sample identification
It carries out sample matches and obtains the second data for participating in regression fit.
Processing module 43 is returned, for being fitted the residual error based on second data regression, obtains returning evaluation index；
Index feedback module 44, for the recurrence evaluation index to be returned to the first data side, so that the first data
Square the second data side of recurrence evaluation index selected section by comparing at least two second data sides to be selected.
This specification embodiment additionally provides a kind of verifying equipment of data validity, and the equipment application is in the first data
Side, the equipment include memory, processor and storage on a memory and the computer program that can run on a processor, institute
It states when processor executes described program and performs the steps of
Enter moding amount and label, training machine learning model according in training sample；The training sample further includes not
Participate in machine learning model training does not enter moding amount；
It moding amount will be entered described in the test sample inputs the machine learning model to obtain predicted value；The test
Sample further includes label, the expection predicted value for entering moding amount input machine learning model of the tag representation test sample；
According to the label of the test sample and the predicted value, the corresponding residual error of the test sample is obtained；
The residual error is respectively sent at least two second data side, so that each second data side makes respectively
It is fitted the residual error with the second data regression possessed, and obtains returning evaluation index；
The recurrence evaluation index that at least two second data side returns respectively is received, by comparing described at least two
The second data side of recurrence evaluation index selected section of a second data side.
This specification embodiment additionally provides a kind of verifying equipment of data validity, and the equipment application is in the second data
Side, the equipment include memory, processor and storage on a memory and the computer program that can run on a processor, institute
It states when processor executes described program and performs the steps of
The residual error of the first data side transmission is received, the residual error is that the first data root according in test sample enters moding amount
The label of predicted value and test sample that input machine learning model obtains obtains；The data packet that the first data side possesses
Include: training set and test set, the training set include multiple training samples, and the test set includes multiple test samples；It is described
Machine learning model be according in training sample enter moding amount and label training obtains；It further include not entering in the training sample
Moding amount；
The corresponding sample identification of the training sample is received, and sample matches are carried out according to the sample identification and are used for
Participate in the second data of regression fit；
It is fitted the residual error based on second data regression, obtains returning evaluation index；
The recurrence evaluation index is returned into the first data side so that the first data side by comparing it is to be selected extremely
The second data side of recurrence evaluation index selected section of few two the second data sides.
Each step in process shown in above method embodiment, execution sequence are not limited to suitable in flow chart
Sequence.In addition, the description of each step, can be implemented as software, hardware or its form combined, for example, those skilled in the art
Member can implement these as the form of software code, can be can be realized the computer of the corresponding logic function of the step can
It executes instruction.When it is realized in the form of software, the executable instruction be can store in memory, and by equipment
Processor execute.
The device or module that above-described embodiment illustrates can specifically realize by computer chip or entity, or by having
The product of certain function is realized.A kind of typically to realize that equipment is computer, the concrete form of computer can be personal meter
Calculation machine, laptop computer, cellular phone, camera phone, smart phone, personal digital assistant, media player, navigation are set
It is any several in standby, E-mail receiver/send equipment, game console, tablet computer, wearable device or these equipment
The combination of equipment.
For convenience of description, it is divided into various modules when description apparatus above with function to describe respectively.Certainly, implementing this
The function of each module can be realized in the same or multiple software and or hardware when specification one or more embodiment.
It should be understood by those skilled in the art that, this specification one or more embodiment can provide for method, system or
Computer program product.Therefore, complete hardware embodiment can be used in this specification one or more embodiment, complete software is implemented
The form of example or embodiment combining software and hardware aspects.Moreover, this specification one or more embodiment can be used one
It is a or it is multiple wherein include computer usable program code computer-usable storage medium (including but not limited to disk storage
Device, CD-ROM, optical memory etc.) on the form of computer program product implemented.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap
Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described want
There is also other identical elements in the process, method of element, commodity or equipment.
This specification one or more embodiment can computer executable instructions it is general on
It hereinafter describes, such as program module.Generally, program module includes executing particular task or realization particular abstract data type
Routine, programs, objects, component, data structure etc..Can also practice in a distributed computing environment this specification one or
Multiple embodiments, in these distributed computing environments, by being executed by the connected remote processing devices of communication network
Task.In a distributed computing environment, the local and remote computer that program module can be located at including storage equipment is deposited
In storage media.
All the embodiments in this specification are described in a progressive manner, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.It is adopted especially for data
For collecting equipment or data processing equipment embodiment, since it is substantially similar to the method embodiment, so the comparison of description is simple
Single, the relevent part can refer to the partial explaination of embodiments of method.
It is above-mentioned that this specification specific embodiment is described.Other embodiments are in the scope of the appended claims
It is interior.In some cases, the movement recorded in detail in the claims or step can be come according to the sequence being different from embodiment
It executes and desired result still may be implemented.In addition, process depicted in the drawing not necessarily require show it is specific suitable
Sequence or consecutive order are just able to achieve desired result.In some embodiments, multitasking and parallel processing be also can
With or may be advantageous.
The foregoing is merely the preferred embodiments of this specification one or more embodiment, not to limit this public affairs
It opens, all within the spirit and principle of the disclosure, any modification, equivalent substitution, improvement and etc. done should be included in the disclosure
Within the scope of protection.
Claims (10)
1. a kind of data selecting method, the method is applied to by selected section at least two second data sides of offer data
Second data side；The method is executed by the first data side, and the first data that the first data side possesses include: machine learning
The training set and test set of model；The training set includes multiple training samples, and the test set includes multiple test samples；
The described method includes:
Enter moding amount and label, the training machine learning model according in training sample；The training sample further includes not
Participate in machine learning model training does not enter moding amount；
It moding amount will be entered described in the test sample inputs the machine learning model to obtain predicted value；The test sample
It further include label, the expection predicted value for entering moding amount input machine learning model of the tag representation test sample；
According to the label of the test sample and the predicted value, the corresponding residual error of the test sample is obtained；
The residual error is respectively sent at least two second data side, so that each second data side is respectively using gathering around
The second data regression having is fitted the residual error, and obtains returning evaluation index；
The recurrence evaluation index that at least two second data side returns respectively is received, by comparing described at least two the
The second data side of recurrence evaluation index selected section of two data sides.
2. according to the method described in claim 1, the method also includes:
The sample identification of the training sample and test sample is sent to the second data side, so that the second data root is according to institute
It states sample identification progress sample matches and obtains second data.
3. according to the method described in claim 1, the method also includes:
Receive the sample matches rate of the second data side return；
It is higher than in the second data of preset threshold by sample matches rate, according to recurrence evaluation index by least two second data sides
The second data that middle the second data side of selected section possesses.
4. a kind of data selecting method, the method is executed by the second data side, comprising:
The residual error of the first data side transmission is received, the residual error is that the first data root is inputted according to the moding amount that enters in test sample
The label of predicted value and test sample that machine learning model obtains obtains；The data that the first data side possesses include:
Training set and test set, the training set include multiple training samples, and the test set includes multiple test samples；The machine
Learning model be according in training sample enter moding amount and label training obtains；It further include not entering moding in the training sample
Amount；
The sample identification of the first data side transmission is received, and sample matches are carried out according to the sample identification and are obtained for participating in back
Return the second data of fitting；
It is fitted the residual error based on second data regression, obtains returning evaluation index；
The recurrence evaluation index is returned into the first data side, so that the first data side is by comparing to be selected at least two
The second data side of recurrence evaluation index selected section of a second data side.
5. according to the method described in claim 4, the method also includes:
Obtain at least one following parameter of second data: sample matches rate and variable miss rate；
The parameter is returned into the first data side, so that the first data side is in conjunction with the parameter and returns evaluation index selection
The second data side of part.
6. a kind of data selection means, described device is applied to by selected section at least two second data sides of offer data
Second data side；Described device is applied to the first data side, and the first data that the first data side possesses include: machine learning
The training set and test set of model；The training set includes multiple training samples, and the test set includes multiple test samples；Institute
Stating device includes:
Model training module, for entering moding amount and label, the training machine learning model according in the training sample；
The training sample further include have neither part nor lot in machine learning model training do not enter moding amount；
Model prediction module, for will enter described in the test sample moding amount input the machine learning model obtain it is pre-
Measured value；The test sample further includes label, and the moding amount that enters of the tag representation test sample inputs machine learning model
It is expected that predicted value；
Residual computations module, for according to the test sample label and the predicted value, it is corresponding residual to obtain test sample
Difference；
Data transmission blocks, for the residual error to be respectively sent at least two second data side, so that each
Two data sides are fitted the residual error using the second data regression possessed respectively, and obtain returning evaluation index；
Verification processing module, the recurrence evaluation index returned respectively for receiving at least two second data side, to pass through
Compare the second data side of recurrence evaluation index selected section of at least two second data side.
7. device according to claim 6,
The verification processing module is also used to receive the sample matches rate of the second data side return；It is higher than by sample matches rate pre-
If in the second data of threshold value, being gathered around according to evaluation index is returned by the second data side of selected section at least two second data sides
The second data having.
8. a kind of data selection means, described device is applied to the second data side, which includes:
Residual error receiving module, for receiving the residual error of the first data side transmission, the residual error is the first data root according to test specimens
The predicted value that moding amount input machine learning model obtains of entering in this and label obtain；The data that the first data side possesses
It include: training set and test set, the training set includes multiple training samples, and the test set includes multiple test samples；Institute
State machine learning model be according in training sample enter moding amount and label training obtains；It further include not in the training sample
Enter moding amount；
Data match module, the sample identification sent for receiving the first data side, and carried out according to the sample identification
Sample matches obtain the second data for participating in regression fit；
Processing module is returned, for being fitted the residual error based on second data regression, obtains returning evaluation index；
Index feedback module, for the recurrence evaluation index to be returned to the first data side, so that the first data side passes through
Compare the second data side of recurrence evaluation index selected section of at least two second data sides to be selected.
9. a kind of data selection equipment, the equipment include memory, processor and storage on a memory and can be in processor
The computer program of upper operation, the processor perform the steps of when executing described program
Enter moding amount and label, training machine learning model according in training sample；The training sample further includes having neither part nor lot in
Machine learning model training does not enter moding amount；
It moding amount will be entered described in the test sample inputs the machine learning model to obtain predicted value；The test sample
It further include label, the expection predicted value for entering moding amount input machine learning model of the tag representation test sample；
According to the label of the test sample and the predicted value, the corresponding residual error of the test sample is obtained；
The residual error is respectively sent at least two second data side, so that each second data side is respectively using gathering around
The second data regression having is fitted the residual error, and obtains returning evaluation index；
The recurrence evaluation index that at least two second data side returns respectively is received, by comparing described at least two the
The second data side of recurrence evaluation index selected section of two data sides.
10. a kind of data selection equipment, the equipment include memory, processor and storage on a memory and can be in processor
The computer program of upper operation, the processor perform the steps of when executing described program
The residual error of the first data side transmission is received, the residual error is that the first data root is inputted according to the moding amount that enters in test sample
The label of predicted value and test sample that machine learning model obtains obtains；The data that the first data side possesses include:
Training set and test set, the training set include multiple training samples, and the test set includes multiple test samples；The machine
Learning model be according in training sample enter moding amount and label training obtains；It further include not entering moding in the training sample
Amount；
The sample identification that the first data side is sent is received, and sample matches are carried out according to the sample identification and are obtained for joining
With the second data of regression fit；
It is fitted the residual error based on second data regression, obtains returning evaluation index；
The recurrence evaluation index is returned into the first data side, so that the first data side is by comparing to be selected at least two
The second data side of recurrence evaluation index selected section of a second data side.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811286327.4A CN109299161B (en) | 2018-10-31 | 2018-10-31 | Data selection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811286327.4A CN109299161B (en) | 2018-10-31 | 2018-10-31 | Data selection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109299161A true CN109299161A (en) | 2019-02-01 |
| CN109299161B CN109299161B (en) | 2022-01-28 |
Family
ID=65145327
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811286327.4A Active CN109299161B (en) | 2018-10-31 | 2018-10-31 | Data selection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109299161B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109961098A (en) * | 2019-03-22 | 2019-07-02 | 中国科学技术大学 | A kind of training data selection method of machine learning |
| CN110968886A (en) * | 2019-12-20 | 2020-04-07 | 支付宝(杭州)信息技术有限公司 | Method and system for screening training samples of machine learning model |
| CN111401483A (en) * | 2020-05-15 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | Sample data processing method and device and multi-party model training system |
| CN111612167A (en) * | 2019-02-26 | 2020-09-01 | 京东数字科技控股有限公司 | Joint training method, device, equipment and storage medium of machine learning model |
| CN112149834A (en) * | 2019-06-28 | 2020-12-29 | 北京百度网讯科技有限公司 | Model training method, device, equipment and medium |
| CN112183757A (en) * | 2019-07-04 | 2021-01-05 | 创新先进技术有限公司 | Model training method, device and system |
| US11295242B2 (en) | 2019-11-13 | 2022-04-05 | International Business Machines Corporation | Automated data and label creation for supervised machine learning regression testing |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101719147A (en) * | 2009-11-23 | 2010-06-02 | 合肥兆尹信息科技有限责任公司 | Rochester model-naive Bayesian model-based data classification system |
| US9501749B1 (en) * | 2012-03-14 | 2016-11-22 | The Mathworks, Inc. | Classification and non-parametric regression framework with reduction of trained models |
| CN108280462A (en) * | 2017-12-11 | 2018-07-13 | 北京三快在线科技有限公司 | A kind of model training method and device, electronic equipment |
| CN108375808A (en) * | 2018-03-12 | 2018-08-07 | 南京恩瑞特实业有限公司 | Dense fog forecasting procedures of the NRIET based on machine learning |
| CN108596757A (en) * | 2018-04-23 | 2018-09-28 | 大连火眼征信管理有限公司 | A kind of personal credit file method and system of intelligences combination |
-
2018
- 2018-10-31 CN CN201811286327.4A patent/CN109299161B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101719147A (en) * | 2009-11-23 | 2010-06-02 | 合肥兆尹信息科技有限责任公司 | Rochester model-naive Bayesian model-based data classification system |
| US9501749B1 (en) * | 2012-03-14 | 2016-11-22 | The Mathworks, Inc. | Classification and non-parametric regression framework with reduction of trained models |
| CN108280462A (en) * | 2017-12-11 | 2018-07-13 | 北京三快在线科技有限公司 | A kind of model training method and device, electronic equipment |
| CN108375808A (en) * | 2018-03-12 | 2018-08-07 | 南京恩瑞特实业有限公司 | Dense fog forecasting procedures of the NRIET based on machine learning |
| CN108596757A (en) * | 2018-04-23 | 2018-09-28 | 大连火眼征信管理有限公司 | A kind of personal credit file method and system of intelligences combination |
Non-Patent Citations (1)
| Title |
|---|
| 程求江 等: ""基于用户行为模型的客流量分析与预测"", 《计算机系统应用》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612167A (en) * | 2019-02-26 | 2020-09-01 | 京东数字科技控股有限公司 | Joint training method, device, equipment and storage medium of machine learning model |
| CN111612167B (en) * | 2019-02-26 | 2024-04-16 | 京东科技控股股份有限公司 | Combined training method, device, equipment and storage medium of machine learning model |
| CN109961098A (en) * | 2019-03-22 | 2019-07-02 | 中国科学技术大学 | A kind of training data selection method of machine learning |
| CN109961098B (en) * | 2019-03-22 | 2022-03-01 | 中国科学技术大学 | Training data selection method for machine learning |
| CN112149834A (en) * | 2019-06-28 | 2020-12-29 | 北京百度网讯科技有限公司 | Model training method, device, equipment and medium |
| CN112149834B (en) * | 2019-06-28 | 2023-11-07 | 北京百度网讯科技有限公司 | Model training method, device, equipment and medium |
| CN112183757A (en) * | 2019-07-04 | 2021-01-05 | 创新先进技术有限公司 | Model training method, device and system |
| CN112183757B (en) * | 2019-07-04 | 2023-10-27 | 创新先进技术有限公司 | Model training method, device and system |
| US11295242B2 (en) | 2019-11-13 | 2022-04-05 | International Business Machines Corporation | Automated data and label creation for supervised machine learning regression testing |
| CN110968886A (en) * | 2019-12-20 | 2020-04-07 | 支付宝(杭州)信息技术有限公司 | Method and system for screening training samples of machine learning model |
| CN111401483A (en) * | 2020-05-15 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | Sample data processing method and device and multi-party model training system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109299161B (en) | 2022-01-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109299161A (en) | A kind of data selecting method and device | |
| Jabbari et al. | What is DevOps? A systematic mapping study on definitions and practices | |
| Luna et al. | Quantitative reasoning about cloud security using service level agreements | |
| Babaioff et al. | Combinatorial agency | |
| Ling et al. | Selection of model in developing information security criteria on smart grid security system | |
| EP2201491A1 (en) | Apparatus for reconfiguration of a technical system based on security analysis and a corresponding technical decision support system and computer program product | |
| CN109508558A (en) | A kind of verification method and device of data validity | |
| US20220058266A1 (en) | Methods and systems of a cybersecurity scoring model | |
| Wei et al. | Two plane camera calibration: a unified model | |
| Maghrabi et al. | Improved software vulnerability patching techniques using CVSS and game theory | |
| Molka et al. | Conformance checking for BPMN-based process models | |
| Lin et al. | Project reliability interval for a stochastic project network subject to time and budget constraints | |
| Zalazar et al. | Analyzing requirements engineering for cloud computing | |
| Alturkistani et al. | A review of security risk assessment methods in cloud computing | |
| Jansen | Research directions in security metrics | |
| Ramireddy et al. | Privacy and Security Practices in the Arena of Cloud Computing-A Research in Progress | |
| Lauster et al. | Literature review linking blockchain and business process management | |
| Khrisna | Risk management framework with COBIT 5 and risk management framework for cloud computing integration | |
| Younis et al. | Towards the Impact of Security Vunnerabilities in Software Design: A Complex Network-Based Approach | |
| Baars et al. | Analysing the Security Risks of Cloud Adoption Using the SeCA Model: A Case Study. | |
| Kang et al. | Process Mining-based Understanding and Analysis of Volvo IT's Incident and Problem Management Processes. | |
| Hidayat et al. | Process model extension using heuristics miner:(Case study: Incident management of Volvo IT Belgium) | |
| Eftekhar et al. | Towards the development of a widely accepted cloud trust model | |
| Chorppath et al. | Risk management for it security: When theory meets practice | |
| CN109657482B (en) | Data validity verification method, device and 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 | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40003742 Country of ref document: HK |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200923 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Innovative advanced technology Co.,Ltd. Address before: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant before: Advanced innovation technology Co.,Ltd. Effective date of registration: 20200923 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Advanced innovation technology Co.,Ltd. Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Applicant before: Alibaba Group Holding Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |