CN109635952A - A kind of prediction technique of regularization, device, electronic equipment and medium - Google Patents
A kind of prediction technique of regularization, device, electronic equipment and medium Download PDFInfo
- Publication number
- CN109635952A CN109635952A CN201811285269.3A CN201811285269A CN109635952A CN 109635952 A CN109635952 A CN 109635952A CN 201811285269 A CN201811285269 A CN 201811285269A CN 109635952 A CN109635952 A CN 109635952A
- Authority
- CN
- China
- Prior art keywords
- regularization
- frequency
- feature
- occurrence
- training
- 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
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses prediction technique, device, electronic equipment and the medium of a kind of regularization, and prediction technique includes: the frequency of occurrence for obtaining each feature in characteristic set；It is calculated using regularization, training sample is generated according to the characteristic set；Wherein, the frequency of occurrence inverse correlation of the penalty factor in the regularization calculating and each feature；Training sample input machine learning model is subjected to model training；It is predicted by the machine learning model that training is completed.When method and apparatus provided by the present application are to solve to be predicted using machine learning model in the prior art, the technical problem larger to loss of significance existing for traditional regularization calculation method realizes the technical effect for improving precision of prediction.
Description
Technical field
The present invention relates to field of computer technology more particularly to a kind of prediction technique of regularization, device, electronic equipment and
Medium.
Background technique
With the development of science and technology various Predicting Techniques emerge one after another, wherein using machine learning model carry out prediction by
In its precision of prediction height, and can be in the case where sample size continues to increase, it is perfect to carry out lasting training, has obtained largely answering
With.
In order to improve the training accuracy of machine learning model, often go to avoid the mistake in model prediction using regularization
Fitting.However, the feature distribution of sample is uneven, and long-tail is obvious by taking the region clicking rate prediction that advertisement is launched as an example, use
Existing regularization calculation method, can large effect precision of prediction.
As it can be seen that traditional regularization calculation method, which exists, damages precision when currently employed machine learning model is predicted
Lose larger, the lower technical problem of precision of prediction.
Summary of the invention
In view of the above problems, it proposes on the present invention overcomes the above problem or at least be partially solved in order to provide one kind
State prediction technique, device, electronic equipment and the medium of the regularization of problem.
In a first aspect, providing a kind of prediction technique of regularization, comprising:
Obtain the frequency of occurrence of each feature in characteristic set；
It is calculated using regularization, training sample is generated according to the characteristic set；Wherein, punishing in the regularization calculating
The frequency of occurrence inverse correlation of penalty factor and each feature；
Training sample input machine learning model is subjected to model training；
It is predicted by the machine learning model that training is completed.
Optionally, described to be calculated using regularization, training sample is generated according to the characteristic set, comprising: use canonical
The calculation formula ∑ λ of the factoriWiWi, training sample is generated according to the characteristic set, wherein WiFor the characteristic variable, λiFor
The penalty factor, the number that subscript i is characterized.
Optionally, when the characteristic set is the set of different geographical clicking rate feature after advertisement is launched, the feature
Variable is the clicking rate of different geographical, and the frequency of occurrence is characterized the frequency of occurrence of each region in set.
Optionally, described to be calculated using regularization, training sample is generated according to the characteristic set, comprising: use logic
The calculation formula of regression modelWith regular factor ∑ λiwiwi, training sample is generated according to the characteristic set, wherein
XiIt is characterized value.
Optionally, when the characteristic set is the set of different geographical clicking rate feature after advertisement is launched, the feature
Value is the code value of region, and the characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized various regions in set
The frequency of occurrence in domain.
Optionally, the machine learning model completed by training is predicted, comprising: receives target area
Code value exports the corresponding clicking rate in the target area as input parameter.
Optionally, the calculation formula of the penalty factor are as follows: λi=λ/NiOr λi=λ/Ni 2, wherein λiFor the punishment
The factor, λ are conventional penalty factor, NiThe number being characterized for the frequency of occurrence, subscript i.
Optionally, the machine learning model is that logic-based regression model or depth model generate.
Second aspect provides a kind of prediction meanss of regularization, comprising:
Module is obtained, for obtaining the frequency of occurrence of each feature in characteristic set；
Generation module generates training sample according to the characteristic set for calculating using regularization；Wherein, it is described just
Then change the frequency of occurrence inverse correlation of the penalty factor and each feature in calculating；
Training module, for training sample input machine learning model to be carried out model training；
Prediction module, the machine learning model for being completed by training are predicted.
Optionally, the generation module is also used to: using the calculation formula ∑ λ of regular factoriWiWi, according to the feature
Set generates training sample, wherein WiFor the characteristic variable, λiThe number being characterized for the penalty factor, subscript i.
Optionally, when the characteristic set is the set of different geographical clicking rate feature after advertisement is launched, the feature
Variable is the clicking rate of different geographical, and the frequency of occurrence is characterized the frequency of occurrence of each region in set.
Optionally, the generation module is also used to: using the calculation formula of Logic Regression ModelsWith canonical because
Sub- ∑ λiWiWi, training sample is generated according to the characteristic set, wherein XiIt is characterized value.
Optionally, when the characteristic set is the set of different geographical clicking rate feature after advertisement is launched, the feature
Value is the code value of region, and the characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized various regions in set
The frequency of occurrence in domain.
Optionally, prediction module is also used to: receiving the code value of target area as input parameter, with exporting the target
The corresponding clicking rate in domain.
Optionally, the calculation formula of the penalty factor are as follows: λi=λ/NiOr λi=λ/Ni 2, wherein λiFor the punishment
The factor, λ are conventional penalty factor, NiThe number being characterized for the frequency of occurrence, subscript i.
Optionally, the machine learning model is that logic-based regression model or depth model generate.
The third aspect, provides a kind of electronic equipment, including memory, processor and storage on a memory and can handled
The computer program run on device, the processor realize first aspect any method when executing described program.
Fourth aspect provides a kind of computer readable storage medium, is stored thereon with computer program, and the program is processed
First aspect any method is realized when device executes.
The technical solution provided in the embodiment of the present application, has at least the following technical effects or advantages:
Prediction technique, device, electronic equipment and the medium of regularization provided by the embodiments of the present application, using regularization meter
Each feature in the penalty factor and characteristic set calculated, characteristic set is generated training sample, and be arranged in regularization calculating
Frequency of occurrence inverse correlation, avoid in traditional regularization using identical penalty factor, do not met with actual prediction scene,
The low technical problem of caused precision.And will be calculated using new regularization the training sample generated input machine learning model into
Row model training is predicted by the machine learning model that training is completed, to improve prediction accuracy.
The above description is only an overview of the technical scheme of the present invention, in order to better understand the technical means of the present invention,
And it can be implemented in accordance with the contents of the specification, and in order to allow above and other objects of the present invention, feature and advantage can
It is clearer and more comprehensible, the followings are specific embodiments of the present invention.
Detailed description of the invention
By reading the following detailed description of the preferred embodiment, various other advantages and benefits are common for this field
Technical staff will become clear.The drawings are only for the purpose of illustrating a preferred embodiment, and is not considered as to the present invention
Limitation.And throughout the drawings, the same reference numbers will be used to refer to the same parts.In the accompanying drawings:
Fig. 1 is the flow chart of the prediction technique of regularization in the embodiment of the present invention；
Fig. 2 is the structural schematic diagram of the prediction meanss of regularization in the embodiment of the present invention；
Fig. 3 is the structural schematic diagram of electronic equipment in the embodiment of the present invention；
Fig. 4 is the structural schematic diagram of storage medium in the embodiment of the present invention.
Specific embodiment
Technical solution in the embodiment of the present application, general thought are as follows:
The frequency of occurrence of each feature in characteristic set is first obtained, then is calculated using regularization, characteristic set is generated
Training sample, and be arranged regularization calculate in penalty factor and characteristic set in each feature frequency of occurrence inverse correlation, with
Meet the setting of penalty factor more with actual prediction scene.And it is the training sample for calculating generation using new regularization is defeated
Enter machine learning model and carry out model training, to improve the prediction accuracy of model.
Exemplary embodiments of the present disclosure are described in more detail below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
It is limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
It is fully disclosed to those skilled in the art.
Embodiment one
A kind of prediction technique of regularization is present embodiments provided, referring to FIG. 1, Fig. 1 is canonical in the embodiment of the present application
The flow chart of the prediction technique of change, comprising:
Step S101 obtains the frequency of occurrence of each feature in characteristic set；
Step S102, is calculated using regularization, generates training sample according to the characteristic set；Wherein, the regularization
The frequency of occurrence inverse correlation of penalty factor and each feature in calculating；
Training sample input machine learning model is carried out model training by step S103；
Step S104 is predicted by the machine learning model that training is completed.
It should be noted that the prediction technique of regularization provided in this embodiment can be applied to ad click rate prediction,
Duration prediction, harassing and wrecking information prediction interception etc. prediction field are watched in advertisement.It is unevenly distributed each that it is suitable for sample characteristics
Kind prediction scene, improves precision of prediction by optimization penalty factor and the frequency of occurrence inverse correlation of feature.Its equipment applied
It can be user terminal or server-side, herein also with no restriction.
In the following, the specific implementation step of method provided by the embodiments of the present application is discussed in detail in conjunction with Fig. 1:
Step S101 obtains the frequency of occurrence of each feature in characteristic set.
Specifically, different prediction scenes has different characteristic sets.
For example, need to predict that advertisement is launched in the clicking rate of different geographical, after this feature collection is combined into advertisement dispensing
The set of different geographical clicking rate feature, each characteristic element in this feature set may include: the code value for launching region
With the clicking rate of different geographical.
For another example needing to predict that the playing duration in distinct device (smart phone, plate or desktop computer) is launched in advertisement
When, this feature collection is combined into the set of distinct device playing duration feature after advertisement is launched, each characteristic element in this feature set
Element may include: the code value of dispensing device and the playing duration of distinct device.
The frequency of occurrence of the present embodiment can refer to the frequency of occurrence of same characteristic value in characteristic set.For example, with this
After characteristic set is launched for advertisement for the set of different geographical clicking rate feature, including sample: (Beijing number of clicks 20 times,
Chengdu number of clicks 10 times, Beijing number of clicks 32 times, Shanghai number of clicks 27 times, Beijing number of clicks 25 times, Chengdu is clicked
Number 7 times), wherein " Beijing " occurs 3 times, then corresponding frequency of occurrence is 3, and " Chengdu " occurs 2 times, corresponding appearance
The frequency is 2.
Step S102, is calculated using regularization, generates training sample according to the characteristic set；Wherein, the regularization
The frequency of occurrence inverse correlation of penalty factor and each feature in calculating.
In the embodiment of the present application, the calculation formula of the regular factor in regularization calculating is ∑ λiWiWi, wherein WiFor
The characteristic variable, λiThe number being characterized for the penalty factor, subscript i.
For example, when the characteristic set is the set of different geographical clicking rate feature after advertisement is launched, feature becomes
Measure WiFor the clicking rate of different geographical.For example, it is assumed that characteristic set includes sample: (Beijing number of clicks 20 times, Chengdu is clicked secondary
Number 10 times, Beijing number of clicks 32 times, Shanghai number of clicks 27 times, Beijing number of clicks 25 times, Chengdu number of clicks 7 times), W1
It is 20, W2It is 10, W3It is 32, W4It is 27, W5It is 25, W6It is 7.
Penalty factor λiWith the frequency of occurrence inverse correlation of each feature, refer to that frequency of occurrence is bigger, corresponding penalty factor
It is smaller.Specifically λ can be arranged according to usage scenarioi=λ/NiOr λi=λ/Ni 2, this is not restricted.Wherein, λ is conventional punishes
Penalty factor, NiThe number being characterized for the frequency of occurrence, subscript i.
For example, it is assumed that λi=λ/Ni, characteristic set includes sample: (Beijing number of clicks 20 times, Chengdu number of clicks
10 times, Beijing number of clicks 32 times, Shanghai number of clicks 27 times, Beijing number of clicks 25 times, Chengdu number of clicks 7 times), north
The frequency of occurrence in capital " is 3, and the frequency of occurrence in " Chengdu " is 2, and the frequency of occurrence in " Shanghai " is 1.λ1=λ/3, λ2=λ/2, λ3=
λ/3, λ4=λ/1, λ5=λ/3, λ6=λ/2.
Lesser penalty factor is given by the above-mentioned feature high to frequency of occurrence, is given to the lower feature of frequency of occurrence
Biggish penalty factor carrys out meeting market's demand scene unevenly distributed, improves precision of prediction.
In the specific implementation process, the corresponding regularization calculation method of different machine learning models is not identical, but all relates to
And regular factor ∑ λ provided in this embodimentiWiWi, wherein the machine learning model can be raw with logic-based regression model
At, can also based on depth model generate, be also based on other models generation, this is not restricted.
Assuming that use the regularization calculation formula of Logic Regression Models forWith regular factor ∑ λiWiWi, whereinThe formula of characterization prediction clicking rate, ∑ λiWiWiThe regularization factors improved for the present embodiment, wherein XiIt is characterized
Value.
For example, when the characteristic set is the set of different geographical clicking rate feature after advertisement is launched, characteristic value
XiFor the code value of region, characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized each region in set
Frequency of occurrence.For example, it is assumed that the code value in " Beijing " is 10, the code value in " Chengdu " is 20, and the code value in " Shanghai " is 30, right
Characteristic set: (Beijing number of clicks 20 times, Chengdu number of clicks 10 times, Beijing number of clicks 32 times, Shanghai number of clicks 27
It is secondary, Beijing number of clicks 25 times, Chengdu number of clicks 7 times) i.e. X1It is 10, X2It is 20, X3It is 10, X4It is 30, X5It is 10, X6For
20。λiAnd WiValue as previously mentioned, being not repeated herein.
Specific use has new penalty factor λiRegularization calculate the method that generates training sample, with it is existing just
Then metaplasia is identical at the method for training sample, does not do tired state herein.
Training sample input machine learning model is carried out model training by step S103 and step S104；Pass through instruction
Practice the machine learning model completed to be predicted.
Specifically, the method for training pattern and the method for use model are similar with existing machine learning model, herein
Do not do tired state.
For example, for needing to predict that advertisement is launched in the clicking rate of different geographical, in the machine completed by training
When device learning model is predicted, using the code value of target area as input parameter, made with the corresponding clicking rate in target area
For output.For needing to predict that advertisement is launched in the playing duration of distinct device, in the engineering completed by training
Model is practised when being predicted, using the code value of target device as inputting parameter, using the corresponding playing duration of target device as
Output.
Based on the same inventive concept, the embodiment of the invention also provides the corresponding device of method in embodiment one, see implementation
Example two.
Embodiment two
As shown in Fig. 2, providing a kind of prediction meanss of regularization, comprising:
Module 201 is obtained, for obtaining the frequency of occurrence of each feature in characteristic set；
Generation module 202 generates training sample according to the characteristic set for calculating using regularization；Wherein, described
The frequency of occurrence inverse correlation of penalty factor and each feature in regularization calculating；
Training module 203, for training sample input machine learning model to be carried out model training；
Prediction module 204, the machine learning model for being completed by training are predicted.
In the embodiment of the present application, the generation module 202 is also used to:
Using the calculation formula ∑ λ of regular factoriWiWi, training sample is generated according to the characteristic set, wherein WiFor
The characteristic variable, λiThe number being characterized for the penalty factor, subscript i.
In the embodiment of the present application, when the set that the characteristic set is different geographical clicking rate feature after advertisement is launched
When, the characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized the frequency of occurrence of each region in set.
In the embodiment of the present application, the generation module 202 is also used to:
Using the calculation formula of Logic Regression ModelsWith regular factor ∑ λiWiWi, according to the characteristic set
Generate training sample, wherein XiIt is characterized value.
In the embodiment of the present application, when the set that the characteristic set is different geographical clicking rate feature after advertisement is launched
When, the characteristic value is the code value of region, and the characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized
The frequency of occurrence of each region in set.
In the embodiment of the present application, prediction module 204 is also used to: the code value for receiving target area is used as input parameter,
Export the corresponding clicking rate in the target area.
In the embodiment of the present application, the calculation formula of the penalty factor are as follows:
λi=λ/NiOr λi=λ/Ni 2, wherein λiFor the penalty factor, λ is conventional penalty factor, NiFor it is described go out
The existing frequency, the number that subscript i is characterized.
In the embodiment of the present application, the machine learning model is that logic-based regression model or depth model generate.
By the device that the embodiment of the present invention two is introduced, filled used by the method to implement the embodiment of the present invention one
It sets, so based on the method that the embodiment of the present invention one is introduced, the affiliated personnel in this field can understand the specific structure of the device
And deformation, so details are not described herein.Device used by the method for all embodiment of the present invention one belongs to the present invention and is intended to
The range of protection.
Based on the same inventive concept, the embodiment of the invention also provides the corresponding equipment of method in embodiment one, see implementation
Example three.
Embodiment three
As shown in figure 3, the present embodiment provides a kind of electronic equipment, including memory 310, processor 320 and it is stored in
On reservoir 310 and the computer program 311 that can run on the processor 320, the processor 320 execute the computer program
It is performed the steps of when 311
Obtain the frequency of occurrence of each feature in characteristic set；
It is calculated using regularization, training sample is generated according to the characteristic set；Wherein, punishing in the regularization calculating
The frequency of occurrence inverse correlation of penalty factor and each feature；
Training sample input machine learning model is subjected to model training；
It is predicted by the machine learning model that training is completed.
In the embodiment of the present application, the application reality may be implemented when the processor 320 executes the computer program 311
Apply any embodiment in example one.
By the equipment that the embodiment of the present invention three is introduced, set used by the method to implement the embodiment of the present invention one
It is standby, so based on the method that the embodiment of the present invention one is introduced, the affiliated personnel in this field can understand the specific structure of the equipment
And deformation, so details are not described herein.Equipment used by the method for all embodiment of the present invention one belongs to the present invention and is intended to
The range of protection.
Based on the same inventive concept, the embodiment of the invention also provides the corresponding storage medium of method in embodiment one, see
Example IV.
Example IV
The present embodiment provides a kind of computer readable storage mediums 400, as shown in figure 4, being stored thereon with computer program
411, which is characterized in that the computer program 411 performs the steps of when being executed by processor
Obtain the frequency of occurrence of each feature in characteristic set；
It is calculated using regularization, training sample is generated according to the characteristic set；Wherein, punishing in the regularization calculating
The frequency of occurrence inverse correlation of penalty factor and each feature；
Training sample input machine learning model is subjected to model training；
It is predicted by the machine learning model that training is completed.
In the specific implementation process, when which is executed by processor, the embodiment of the present application one may be implemented
Middle any embodiment.
The technical solution provided in the embodiment of the present application, has at least the following technical effects or advantages:
Prediction technique, device, electronic equipment and the medium of regularization provided by the embodiments of the present application, using regularization meter
Each feature in the penalty factor and characteristic set calculated, characteristic set is generated training sample, and be arranged in regularization calculating
Frequency of occurrence inverse correlation, avoid in traditional regularization using identical penalty factor, do not met with actual prediction scene,
The low technical problem of caused precision.And will be calculated using new regularization the training sample generated input machine learning model into
Row model training is predicted by the machine learning model that training is completed, to improve prediction accuracy.
Algorithm and display are not inherently related to any particular computer, virtual system, or other device provided herein.
Various general-purpose systems can also be used together with teachings based herein.As described above, it constructs required by this kind of system
Structure be obvious.In addition, the present invention is also not directed to any particular programming language.It should be understood that can use various
Programming language realizes summary of the invention described herein, and the description done above to language-specific is to disclose this hair
Bright preferred forms.
In the instructions provided here, numerous specific details are set forth.It is to be appreciated, however, that implementation of the invention
Example can be practiced without these specific details.In some instances, well known method, structure is not been shown in detail
And technology, so as not to obscure the understanding of this specification.
Similarly, it should be understood that in order to simplify the disclosure and help to understand one or more of the various inventive aspects,
Above in the description of exemplary embodiment of the present invention, each feature of the invention is grouped together into single implementation sometimes
In example, figure or descriptions thereof.However, the disclosed method should not be interpreted as reflecting the following intention: i.e. required to protect
Shield the present invention claims features more more than feature expressly recited in each claim.More precisely, as following
Claims reflect as, inventive aspect is all features less than single embodiment disclosed above.Therefore,
Thus the claims for following specific embodiment are expressly incorporated in the specific embodiment, wherein each claim itself
All as a separate embodiment of the present invention.
Those skilled in the art will understand that can be carried out adaptively to the module in the equipment in embodiment
Change and they are arranged in one or more devices different from this embodiment.It can be the module or list in embodiment
Member or component are combined into a module or unit or component, and furthermore they can be divided into multiple submodule or subelement or
Sub-component.Other than such feature and/or at least some of process or unit exclude each other, it can use any
Combination is to all features disclosed in this specification (including adjoint claim, abstract and attached drawing) and so disclosed
All process or units of what method or apparatus are combined.Unless expressly stated otherwise, this specification is (including adjoint power
Benefit require, abstract and attached drawing) disclosed in each feature can carry out generation with an alternative feature that provides the same, equivalent, or similar purpose
It replaces.
In addition, it will be appreciated by those of skill in the art that although some embodiments in this include institute in other embodiments
Including certain features rather than other feature, but the combination of the feature of different embodiment means in the scope of the present invention
Within and form different embodiments.For example, in the following claims, embodiment claimed it is any it
One can in any combination mode come using.
Various component embodiments of the invention can be implemented in hardware, or to run on one or more processors
Software module realize, or be implemented in a combination thereof.It will be understood by those of skill in the art that can be used in practice
Microprocessor or digital signal processor (DSP) realize gateway according to an embodiment of the present invention, proxy server, in system
Some or all components some or all functions.The present invention is also implemented as executing side as described herein
Some or all device or device programs (for example, computer program and computer program product) of method.It is such
It realizes that program of the invention can store on a computer-readable medium, or can have the shape of one or more signal
Formula.Such signal can be downloaded from an internet website to obtain, and perhaps be provided on the carrier signal or with any other shape
Formula provides.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and ability
Field technique personnel can be designed alternative embodiment without departing from the scope of the appended claims.In the claims,
Any reference symbol between parentheses should not be configured to limitations on claims.Word "comprising" does not exclude the presence of not
Component or step listed in the claims.Word "a" or "an" before component does not exclude the presence of multiple such
Component.The present invention can be by means of including the hardware of several different components and being come by means of properly programmed computer real
It is existing.In the unit claims listing several devices, several in these devices can be through the same hardware branch
To embody.The use of word first, second, and third does not indicate any sequence.These words can be explained and be run after fame
Claim.
The invention discloses A1, a kind of prediction technique of regularization, comprising:
Obtain the frequency of occurrence of each feature in characteristic set；
It is calculated using regularization, training sample is generated according to the characteristic set；Wherein, punishing in the regularization calculating
The frequency of occurrence inverse correlation of penalty factor and each feature；
Training sample input machine learning model is subjected to model training；
It is predicted by the machine learning model that training is completed.
A2, method according to a1, which is characterized in that it is described to be calculated using regularization, according to the feature set symphysis
At training sample, comprising:
Using the calculation formula ∑ λ of regular factoriWiWi, training sample is generated according to the characteristic set, wherein WiFor
The characteristic variable, λiThe number being characterized for the penalty factor, subscript i.
A3, the method according to A2, which is characterized in that when the characteristic set is that different geographical is clicked after advertisement is launched
When the set of rate feature, the characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized each region in set
Frequency of occurrence.
A4, the method according to A2, which is characterized in that it is described to be calculated using regularization, according to the feature set symphysis
At training sample, comprising:
Using the calculation formula of Logic Regression ModelsWith regular factor ∑ λiWiWi, according to the characteristic set
Generate training sample, wherein XiIt is characterized value.
A5, method according to a4, which is characterized in that when the characteristic set is that different geographical is clicked after advertisement is launched
When the set of rate feature, the characteristic value be region code value, the characteristic variable be different geographical clicking rate, it is described go out
The existing frequency is characterized the frequency of occurrence of each region in set.
A6, method according to a5, which is characterized in that the machine learning model completed by training carries out
Prediction, comprising: receive the code value of target area as input parameter, export the corresponding clicking rate in the target area.
A7, method according to a1, which is characterized in that the calculation formula of the penalty factor are as follows:
λi=λ/NiOr λi=λ/Ni 2, wherein λ i is the penalty factor, and λ is conventional penalty factor, NiFor it is described go out
The existing frequency, the number that subscript i is characterized.
A8, method according to a1, which is characterized in that the machine learning model is logic-based regression model or depth
Model is spent to generate.
B9, a kind of prediction meanss of regularization, comprising:
Module is obtained, for obtaining the frequency of occurrence of each feature in characteristic set；
Generation module generates training sample according to the characteristic set for calculating using regularization；Wherein, it is described just
Then change the frequency of occurrence inverse correlation of the penalty factor and each feature in calculating；
Training module, for training sample input machine learning model to be carried out model training；
Prediction module, the machine learning model for being completed by training are predicted.
B10, the device according to B9, which is characterized in that the generation module is also used to:
Using the calculation formula ∑ λ of regular factoriWiWi, training sample is generated according to the characteristic set, wherein Wi is
The characteristic variable, λ i are the penalty factor, the number that subscript i is characterized.
B11, device according to b10, which is characterized in that when the characteristic set is different geographical point after advertisement is launched
When hitting the set of rate feature, the characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized various regions in set
The frequency of occurrence in domain.
B12, device according to b10, which is characterized in that the generation module is also used to:
Using the calculation formula of Logic Regression ModelsWith regular factor ∑ λiWiWi, according to the characteristic set
Generate training sample, wherein XiIt is characterized value.
B13, device according to b12, which is characterized in that when the characteristic set is different geographical point after advertisement is launched
When hitting the set of rate feature, the characteristic value is the code value of region, and the characteristic variable is the clicking rate of different geographical, described
Frequency of occurrence is characterized the frequency of occurrence of each region in set.
B14, device according to b13, which is characterized in that prediction module is also used to: the code value of target area is received
As input parameter, the corresponding clicking rate in the target area is exported.
B15, the device according to B9, which is characterized in that the calculation formula of the penalty factor are as follows:
λi=λ/Ni or λi=λ/Ni 2, wherein λiFor the penalty factor, λ is conventional penalty factor, NiFor it is described go out
The existing frequency, the number that subscript i is characterized.
B16, the device according to B9, which is characterized in that the machine learning model be logic-based regression model or
Depth model generates.
C17, a kind of electronic equipment, including memory, processor and storage can be run on a memory and on a processor
Computer program, which is characterized in that the processor realizes A1-A8 any method when executing described program.
D18, a kind of computer readable storage medium, are stored thereon with computer program, which is characterized in that the program is located
It manages and realizes A1-A8 any method when device executes.
Claims (10)
1. a kind of prediction technique of regularization characterized by comprising
Obtain the frequency of occurrence of each feature in characteristic set；
It is calculated using regularization, training sample is generated according to the characteristic set；Wherein, the regularization calculate in punishment because
The sub frequency of occurrence inverse correlation with each feature；
Training sample input machine learning model is subjected to model training；
It is predicted by the machine learning model that training is completed.
2. the method as described in claim 1, which is characterized in that it is described to be calculated using regularization, according to the feature set symphysis
At training sample, comprising:
Using the calculation formula ∑ λ of regular factoriWiWi, training sample is generated according to the characteristic set, wherein WiFor the spy
Levy variable, λiThe number being characterized for the penalty factor, subscript i.
3. method according to claim 2, which is characterized in that when the characteristic set is that different geographical is clicked after advertisement is launched
When the set of rate feature, the characteristic variable is the clicking rate of different geographical, and the frequency of occurrence is characterized each region in set
Frequency of occurrence.
4. method according to claim 2, which is characterized in that it is described to be calculated using regularization, according to the feature set symphysis
At training sample, comprising:
Using the calculation formula of Logic Regression ModelsWith regular factor ∑ λiwiwi, generated and instructed according to the characteristic set
Practice sample, wherein XiIt is characterized value.
5. method as claimed in claim 4, which is characterized in that when the characteristic set is that different geographical is clicked after advertisement is launched
When the set of rate feature, the characteristic value be region code value, the characteristic variable be different geographical clicking rate, it is described go out
The existing frequency is characterized the frequency of occurrence of each region in set.
6. method as claimed in claim 5, which is characterized in that the machine learning model completed by training carries out
Prediction, comprising: receive the code value of target area as input parameter, export the corresponding clicking rate in the target area.
7. the method as described in claim 1, which is characterized in that the calculation formula of the penalty factor are as follows:
λi=λ/NiOr λi=λ/Ni 2, wherein λ i is the penalty factor, and λ is conventional penalty factor, NiFor the appearance frequency
Number secondary, that subscript i is characterized.
8. a kind of prediction meanss of regularization characterized by comprising
Module is obtained, for obtaining the frequency of occurrence of each feature in characteristic set；
Generation module generates training sample according to the characteristic set for calculating using regularization；Wherein, the regularization
The frequency of occurrence inverse correlation of penalty factor and each feature in calculating；
Training module, for training sample input machine learning model to be carried out model training；
Prediction module, the machine learning model for being completed by training are predicted.
9. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that the processor realizes method as claimed in claim 1 to 7 when executing described program.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Method as claimed in claim 1 to 7 is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811285269.3A CN109635952A (en) | 2018-10-31 | 2018-10-31 | A kind of prediction technique of regularization, device, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811285269.3A CN109635952A (en) | 2018-10-31 | 2018-10-31 | A kind of prediction technique of regularization, device, electronic equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109635952A true CN109635952A (en) | 2019-04-16 |
Family
ID=66067008
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811285269.3A Pending CN109635952A (en) | 2018-10-31 | 2018-10-31 | A kind of prediction technique of regularization, device, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109635952A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111464690A (en) * | 2020-02-27 | 2020-07-28 | 华为技术有限公司 | Application preloading method and electronic equipment |
-
2018
- 2018-10-31 CN CN201811285269.3A patent/CN109635952A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111464690A (en) * | 2020-02-27 | 2020-07-28 | 华为技术有限公司 | Application preloading method and electronic equipment |
| CN111464690B (en) * | 2020-02-27 | 2021-08-31 | 华为技术有限公司 | Application preloading method, electronic equipment, chip system and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109583594B (en) | Deep learning training method, device, equipment and readable storage medium | |
| CN109949219B (en) | Reconstruction method, device and equipment of super-resolution image | |
| CN108109011A (en) | A kind of anti-cheat method of advertisement and computing device | |
| CN110830234B (en) | User traffic distribution method and device | |
| US20090312992A1 (en) | Computer-Implemented Systems And Methods For Executing Stochastic Discrete Event Simulations For Design Of Experiments | |
| CN104025048B (en) | Method and system for estimating installation time of software product in operating environment | |
| CN110114784A (en) | Recurrent neural network | |
| CN109155005A (en) | Learnt using the enhancing of spurious count | |
| CN105117953A (en) | User sign-in processing method and device | |
| CN108876893A (en) | Method, apparatus, system and the computer storage medium of three-dimensional facial reconstruction | |
| CN109672909A (en) | Data processing method, device, electronic equipment and readable storage medium storing program for executing | |
| CN108052670A (en) | A kind of recommendation method and device of camera special effect | |
| CN109218769A (en) | A kind of recommended method and relevant device of direct broadcasting room | |
| CN109345317A (en) | The determination method and device of user's taste preference information | |
| CN109660612A (en) | A kind of request processing method and server | |
| US10943691B2 (en) | Cost of healthcare analytics platform | |
| CN106981099B (en) | Method and apparatus for manipulating three-dimensional animated characters | |
| CN109992500A (en) | Interface test method, device and the computer readable storage medium of application | |
| CN109635952A (en) | A kind of prediction technique of regularization, device, electronic equipment and medium | |
| CN108958855A (en) | The method, apparatus and system of page jump | |
| CN106557416B (en) | The implementation method and device of software cloud test | |
| Stefansson et al. | Orders of magnitude speed increase in partial least squares feature selection with new simple indexing technique for very tall data sets | |
| CN108062709A (en) | Personal behavior model training method and device based on semi-supervised learning | |
| CN113626688B (en) | Intelligent medical data acquisition method and system based on software definition | |
| CN108198128A (en) | A kind of method and device of alpha channel boundary corrosions |
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 |