CN110223108A - Click through rate prediction method, device and equipment - Google Patents
Click through rate prediction method, device and equipment Download PDFInfo
- Publication number
- CN110223108A CN110223108A CN201910444789.2A CN201910444789A CN110223108A CN 110223108 A CN110223108 A CN 110223108A CN 201910444789 A CN201910444789 A CN 201910444789A CN 110223108 A CN110223108 A CN 110223108A
- Authority
- CN
- China
- Prior art keywords
- characteristic
- prediction model
- prediction
- click
- rate
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The method, the device and the equipment for predicting the click through rate provided by the embodiment of the invention are used for acquiring at least one characteristic data of the network advertisement to be delivered; aiming at each acquired feature data, searching a preset prediction model corresponding to the feature data from the corresponding relation between the preset feature data and the preset prediction model; the preset prediction model is obtained by utilizing sample characteristic data and a prediction result label of the sample characteristic data to train in advance, and the sample characteristic data is the same as the type of characteristic data corresponding to the preset prediction model; and inputting the characteristic data into a corresponding preset prediction model aiming at each acquired characteristic data to obtain the predicted click through rate corresponding to the characteristic data. The method and the device can determine the characteristic data of the to-be-delivered network advertisement which influences the acquired predicted click through rate, so that the characteristics reflected by the characteristic data can be adjusted in a targeted manner, and convenience and efficiency of characteristic adjustment are improved.
Description
Technical field
The present invention relates to web advertisement technical field, more particularly to a kind of prediction technique of click-through-rate, device and
Equipment.
Background technique
In web advertisement dispensing, CTR (Click-Through-Rate, click-through-rate), i.e., user is to the web advertisement
Actual click number and showing advertisement amount ratio, be measure advertisement delivery effect an important indicator.In order to rationally into
The dispensing of the row web advertisement can carry out click-through-rate prediction to the web advertisement to be put.It is clicked through obtaining prediction
After rate, according to the click-through-rate of prediction, release time, placement position and the user information of the web advertisement to be put are adjusted
Etc. feature, feature by adjusting after improves the click-through-rate of the web advertisement, to optimize the dispensing effect of the web advertisement.
Illustratively, in the feature of the web advertisement to be put, user information is 20 years old to 30 years old women, is led at this point, prediction is clicked
Cross that rate is in contrast lower, after user information to be adjusted to 30 years old to 40 years old women, prediction click-through-rate gets a promotion.
In the related technology, advance with sample characteristics data and sample characteristics data prediction result label training obtain it is pre-
If click-through-rate prediction model.Wherein, the feature of sample characteristics data are reflected feature and the web advertisement to be predicted
Identical, therefore, after the characteristic of the web advertisement is inputted preset click-through-rate prediction model, the available model is defeated
Prediction click-through-rate out.
But the above method is the feature of the web advertisement to be put to the basis for forecasting of click-through-rate, and network is wide
Announcement have the characteristic that it is diversified, after obtained prediction click-through-rate is the entire effect for combining various features data
Obtained prediction result.According to the prediction click-through-rate obtained after the entire effect for combining various features data, Wu Fabiao
Bright is any or which kind feature causes click-through-rate in contrast lower.Therefore, the web advertisement to be put is adjusted
Feature when, need operation maintenance personnel tentatively to adjust the feature of the web advertisement to be put one by one, lead to network to be put
The Character adjustment of advertisement is relatively complicated, and efficiency is lower.
Summary of the invention
The prediction technique for being designed to provide a kind of click-through-rate, device and the equipment of the embodiment of the present invention, to realize
Show in the web advertisement to be put, to the feature that has an impact of prediction click-through-rate, improve Character adjustment convenience and
The effect of efficiency.Specific technical solution is as follows:
In a first aspect, the embodiment of the invention provides a kind of prediction techniques of click-through-rate, this method comprises:
Obtain at least one characteristic of the web advertisement to be put;The type of characteristic is according to characteristic institute
The type that the feature of the web advertisement of reflection divides;
For every kind of acquired characteristic, from the corresponding relationship of preset characteristic and preset prediction model
In, search the corresponding preset prediction model of this kind of characteristic;The preset prediction model is to advance with sample characteristics
The model that the prediction result label training of data and sample characteristics data obtains, and sample characteristics data and the preset prediction mould
The type of the corresponding characteristic of type is identical;
For every kind of acquired characteristic, this kind of characteristic is inputted into corresponding preset prediction model, is obtained
The corresponding prediction click-through-rate of this kind of characteristic.
Optionally, at least one characteristic for obtaining the web advertisement to be put, comprising:
Obtain the type of the characteristic of the web advertisement to be put and multiple spies of the web advertisement to be put
Levy data;
According to the type of acquired characteristic, the building rule of every kind of characteristic is determined;
Respectively according to the building rule of every kind of characteristic, using the multiple characteristic, this kind of feature is constructed
Data;
At least one feature by constructed at least one obtained characteristic, as the web advertisement to be put
Data.
Optionally, described respectively according to the building rule of every kind of characteristic, using the multiple characteristic,
After constructing this kind of characteristic, the method also includes:
At least one characteristic that building is obtained respectively inputs preset hashing algorithm, obtains every kind of characteristic point
Not corresponding hashed value;
Using obtained hashed value as at least one characteristic of the web advertisement to be put.
Optionally, each preset prediction model respectively corresponds a thread;The thread is corresponding for loading the thread
Preset prediction model;
It is described to be directed to every kind of acquired characteristic, this kind of characteristic is inputted into corresponding preset prediction model,
Obtain the corresponding prediction click-through-rate of this kind of characteristic, comprising:
For every kind of acquired characteristic, detect whether the corresponding thread of this kind of characteristic has terminated to run;
If it has ended, it is corresponding preset pre- to load this kind of characteristic using the corresponding thread of this kind of characteristic
Model is surveyed, and this kind of characteristic is inputted into corresponding preset prediction model, obtains the corresponding future position of this kind of characteristic
Percent of pass is hit, otherwise, executes whether the corresponding thread of this kind of characteristic of the detection has terminated to run.
Optionally, described to be directed to every kind of acquired characteristic, this kind of characteristic input is corresponding preset pre-
Model is surveyed, the corresponding prediction click-through-rate of this kind of characteristic is obtained, comprising:
For every kind of acquired characteristic, the second mark of the corresponding preset prediction model of this kind of characteristic is obtained
Know;
The corresponding second identifier of this kind of characteristic is inputted into preset prediction model interface function, obtains this kind of characteristic
According to corresponding prediction click-through-rate.
Second aspect, the embodiment of the invention provides a kind of prediction meanss of click-through-rate, which includes:
Characteristic obtains module, for obtaining at least one characteristic of the web advertisement to be put;Characteristic
The type that is divided for the feature of the web advertisement that is reflected according to characteristic of type;
Prediction model obtains module, for being directed to every kind of acquired characteristic, from preset characteristic and presets
Prediction model corresponding relationship in, search the corresponding preset prediction model of this kind of characteristic;The preset prediction mould
Type is the model for advancing with the prediction result label training of sample characteristics data and sample characteristics data and obtaining, and sample characteristics
The type of data characteristic corresponding with the preset prediction model is identical;
Prediction module, it is for being directed to every kind of acquired characteristic, this kind of characteristic input is corresponding preset
Prediction model obtains the corresponding prediction click-through-rate of this kind of characteristic.
Optionally, the characteristic obtains module, is specifically used for:
Obtain the type of the characteristic of the web advertisement to be put and multiple spies of the web advertisement to be put
Levy data;
According to the type of acquired characteristic, the building rule of every kind of characteristic is determined;
Respectively according to the building rule of every kind of characteristic, using the multiple characteristic, this kind of feature is constructed
Data;
At least one feature by constructed at least one obtained characteristic, as the web advertisement to be put
Data.
Optionally, the characteristic obtains module, is also used to advising according to the building of every kind of characteristic respectively
Then, using the multiple characteristic, after constructing this kind of characteristic, at least one obtained characteristic will be constructed respectively
Preset hashing algorithm is inputted, the corresponding hashed value of every kind of characteristic is obtained;
Using obtained hashed value as at least one characteristic of the web advertisement to be put.
Optionally, each preset prediction model respectively corresponds a thread;The thread is corresponding for loading the thread
Preset prediction model;
The prediction module, including judging submodule and load submodule;
The judging submodule detects the corresponding line of this kind of characteristic for being directed to every kind of acquired characteristic
Whether journey, which has terminated, runs;If it has ended, the triggering load submodule operation;
The load submodule, for utilizing the corresponding line of this kind of characteristic under the triggering of the judging submodule
Journey loads the corresponding preset prediction model of this kind of characteristic, and this kind of characteristic is inputted corresponding preset prediction
Model obtains the corresponding prediction click-through-rate of this kind of characteristic, otherwise, it is corresponding to execute this kind of characteristic of the detection
Whether thread, which has terminated, runs.
Optionally, the prediction module, is specifically used for:
For every kind of acquired characteristic, the second mark of the corresponding preset prediction model of this kind of characteristic is obtained
Know;
The corresponding second identifier of this kind of characteristic is inputted into preset prediction model interface function, obtains this kind of characteristic
According to corresponding prediction click-through-rate.
The third aspect, the embodiment of the invention provides a kind of electronic equipment, which includes:
Processor, communication interface, memory and communication bus, wherein processor, communication interface, memory pass through bus
Complete mutual communication;Memory, for storing computer program;Processor, for executing the journey stored on memory
Sequence, the step of realizing the prediction technique for the click-through-rate that above-mentioned first aspect provides.
Fourth aspect is stored in the storage medium the embodiment of the invention provides a kind of computer readable storage medium
Computer program, the computer program realize the prediction side for the click-through-rate that above-mentioned first aspect provides when being executed by processor
The step of method.
In scheme provided in an embodiment of the present invention, when carrying out click-through-rate prediction, preset prediction model is preparatory
The model obtained using the prediction result label training of sample characteristics data and sample characteristics data, and sample characteristics data with and
The type of sample characteristics data characteristic corresponding with the preset prediction model is identical.Therefore, for acquired every kind
It is corresponding can to obtain this kind of characteristic after this kind of characteristic is inputted corresponding preset prediction model for characteristic
Predict click-through-rate.Due to being directed to different types of characteristic of the web advertisement, it is corresponding pre- to obtain this kind of characteristic
Click-through-rate is surveyed, when adjusting the feature of the web advertisement to be put according to prediction click-through-rate, is capable of determining that institute
The characteristic for the web advertisement to be put that the prediction click-through-rate of acquisition has an impact, therefore, operation maintenance personnel can be with needle
The feature that characteristic is reflected is adjusted to property.The web advertisement to be put is tentatively adjusted one by one with operation maintenance personnel is needed
Feature compare, save on it is no influence prediction click-through-rate feature tentative adjustment the step of, can be improved wait throw
The Character adjustment convenience and efficiency for the web advertisement put.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described.
Fig. 1 is the flow diagram of the prediction technique for the click-through-rate that one embodiment of the invention provides;
Fig. 2 be another embodiment of the present invention provides click-through-rate prediction technique flow diagram;
Fig. 3 is the structural schematic diagram of the prediction meanss for the click-through-rate that one embodiment of the invention provides;
Fig. 4 be another embodiment of the present invention provides click-through-rate prediction meanss structural schematic diagram;
Fig. 5 is the structural schematic diagram for the electronic equipment that one embodiment of the invention provides.
Specific embodiment
In order to make those skilled in the art more fully understand the technical solution in the present invention, implement below in conjunction with the present invention
Attached drawing in example, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment
Only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, the common skill in this field
Art personnel every other embodiment obtained without creative efforts belongs to the model that the present invention protects
It encloses.
The prediction technique of the click-through-rate of one embodiment of the invention is introduced first below.
The prediction technique of click-through-rate provided in an embodiment of the present invention can be applied to the electricity for being able to carry out data processing
Sub- equipment, which may include desktop computer, portable computer, internet television, intelligent mobile terminal, wearable
Intelligent terminal and server etc., are not limited thereto, any electronic equipment that the embodiment of the present invention may be implemented, and belong to this
The protection scope of inventive embodiments.
Further, since the dispensing of the web advertisement specifically can be in all kinds of Internet-related applications, therefore, having
In body application, the executing subject of the prediction technique of click-through-rate provided in an embodiment of the present invention can be the clothes of various applications
Business device.Illustratively, it can be the server of live streaming application, social application or Video Applications etc. application.
As shown in Figure 1, the process of the prediction technique of the click-through-rate of one embodiment of the invention, this method may include:
S101 obtains at least one characteristic of the web advertisement to be put;The type of characteristic is according to feature
The type that the feature for the web advertisement that data are reflected divides.
It, can in order to guarantee the subsequent prediction click-through-rate corresponding to certain characteristic obtained in step s 103
Show the feature having an impact to prediction click-through-rate, characteristic can be obtained according to the type of characteristic, also, need
The type of characteristic is divided according to the feature for the web advertisement that characteristic is reflected.Since the feature of the web advertisement divides
It can be a variety of, therefore, the type of the characteristic of the web advertisement can be a variety of.
Illustratively, the type of characteristic can be the source division by the feature of the web advertisement.For example, by source
User information, such as user's gender, user geographical location and age of user spy are divided into for the characteristic of user information
The type for levying data is user information;The characteristic that source is advertising information is divided into advertising information, such as the web advertisement
The type of the release time of content-form, the putting mode of the web advertisement and the web advertisement etc. characteristic is advertisement letter
Breath.
Alternatively, illustrative, the type of characteristic can be the building regular partition by the feature of the web advertisement.Example
Such as, when building rule is " each characteristic is as a kind of characteristic ", each characteristic can be divided into one kind
Characteristic, for example, age of user can be the first characteristic, user's gender can be second of characteristic;Work as structure
It, can be with when building rule as " after at least two specific characteristic data of splicing, obtained characteristic is as a kind of characteristic "
Each obtained characteristic of splicing is divided into a kind of characteristic, for example, what splicing age of user and user's gender obtained
Data can be the first characteristic, it is wide to splice the content-form of the web advertisement, the putting mode of the web advertisement and network
The data that the release time of announcement obtains can be second of characteristic.
The division mode of the type of any web advertisement is used equally for the present invention, the present embodiment to this with no restriction.Having
In body application, the form of characteristic can be a variety of.Illustratively, characteristic can be the feature sheet of the web advertisement
Body, for example, " image " and " women " etc..Alternatively, illustrative, characteristic can be the feature of the feature of the web advertisement
Value, for example, the characteristic value etc. of the characteristic value of feature " image " and feature " women ".
Corresponding to the different type division modes of features described above data, at least one feature of the web advertisement to be put
The acquisition modes of data can be a variety of.Illustratively, when the source of the feature by the web advertisement divides the kind of characteristic
When class, at least one characteristic of the web advertisement to be put can be directly acquired.For example, from the web advertisement to be put
At least one of advertising information and user information are extracted in dropping apparatus information.Alternatively, it is illustrative, when by the web advertisement
When the type of the building regular partition characteristic of feature, at least one characteristic of the available web advertisement to be put
And data class, and then determine that corresponding building rule utilizes at least one characteristic according to building rule according to data class
Different types of characteristic is obtained according to building.In order to facilitate understanding and rational deployment, subsequent in the form of alternative embodiment pair
It is specifically described according to the case where type for constructing regular partition characteristic.
The acquisition modes of at least one characteristic of any web advertisement to be put are used equally for the present invention, this implementation
Example to this with no restriction.
S102, it is corresponding with preset prediction model from preset characteristic for every kind of acquired characteristic
In relationship, the corresponding preset prediction model of this kind of characteristic is searched.Wherein, preset prediction model is to advance with sample
The model that the prediction result label training of characteristic and sample characteristics data obtains, and sample characteristics data are preset pre- with this
The type for surveying the corresponding characteristic of model is identical.
Wherein, the corresponding relationship of preset characteristic and preset prediction model can be a variety of.Illustratively, in advance
If characteristic can be with the corresponding relationship of preset prediction model and indicate the corresponding preset prediction model of characteristic
Pointer.Alternatively, illustrative, the corresponding relationship of preset characteristic and preset prediction model can be characteristic with
The mapping table of preset prediction model, be stored in the relation table characteristic first identifier and preset prediction model
Second identifier.
It, can in order to guarantee the subsequent prediction click-through-rate corresponding to certain characteristic obtained in step s 103
Show to the feature that has an impact of prediction click-through-rate, the sample characteristics data for obtaining preset prediction model need with
The type of the characteristic of the web advertisement to be put is identical.Illustratively, it is user that acquired characteristic, which includes type,
The first characteristic and type of information are second of characteristic of advertising information, then for training to the first characteristic
Type according to the sample characteristics data for the preset prediction model predicted is user information, for training to second of feature
The type of the sample characteristics data for the preset prediction model that data are predicted is advertising information.
Also, due to the type for trained sample characteristics data characteristic corresponding with the preset prediction model
Identical, therefore, the type of the obtained preset prediction model of training and characteristic corresponds.Illustratively, preset LR
Model M 1 is used to calculate the characteristic of reflection advertising information, and preset LR model M 2 is used for reflection user information
Characteristic is calculated.Alternatively, illustrative, preset LR model M 3 to by age of user and user's gender for constructing
To the first characteristic calculated, preset LR model M 4 is used for the content-form of the web advertisement, the web advertisement
Second of characteristic that the release time of putting mode and the web advertisement constructs is calculated.In addition, specifically answering
In, preset prediction model can be a variety of.Illustratively, preset prediction model can be preset LR model,
It can be preset neural network model.Any type phase that can be advanced with the characteristic of the web advertisement to be put
The prediction model that the prediction result label training of same sample characteristics data and sample characteristics data obtains, is used equally for this
Invention, the present embodiment to this with no restriction.
This kind of characteristic is inputted corresponding preset prediction model for every kind of acquired characteristic by S103,
Obtain the corresponding prediction click-through-rate of this kind of characteristic.
Illustratively, acquired characteristic includes characteristic " user's gender and age of user ", this feature data
Type be " type T1 ", characteristic " placement position of the web advertisement and release time ", the types of this feature data is " kind
Class T2 ".The first characteristic is inputted into preset prediction model M1, it is right to obtain characteristic " user's gender and age of user "
The prediction click-through-rate P1 answered;Second of characteristic is inputted into preset prediction model M2, obtaining characteristic, " network is wide
The placement position of announcement and release time " corresponding prediction click-through-rate P2.Wherein, preset prediction model is obtained for training
The type of the sample characteristics data of M1 is " type T1 ";The sample characteristics data of preset prediction model M2 are obtained for training
Type is " type T2 ".
In a particular application, for every kind of acquired characteristic, this kind of characteristic input is corresponding preset
The concrete mode of prediction model can be a variety of.Illustratively, it can be serial input, specifically: from acquired a variety of
It in characteristic, chooses one and inputs the corresponding preset prediction model of this kind of characteristic, it is corresponding to obtain this kind of characteristic
Prediction click-through-rate;After obtaining the corresponding prediction click-through-rate of characteristic of last selection, never predicted
In the various features data of click-through-rate, chooses next characteristic and obtain click-through-rate, circuit sequentially, until obtaining
The corresponding prediction click-through-rate of every kind of acquired characteristic.Alternatively, it is illustrative, it can be parallel input, this
When, each preset prediction model respectively corresponds a thread, and thread is used to load the corresponding preset prediction model of the thread,
Using multiple threads, the input of different types of characteristic can be carried out simultaneously, and the prediction of every kind of characteristic clicks through
The acquisition of rate is mutually indepedent, after obtaining a kind of corresponding prediction click-through-rate of characteristic without waiting, then obtains another kind
The corresponding prediction click-through-rate of characteristic can effectively improve the acquisition efficiency of prediction click-through-rate.
In addition, characteristic type be it is a variety of, when having obtained multiple prediction click-through-rates, for convenience maintenance people
Member determines the corresponding characteristic of obtained each prediction click-through-rate, the prediction click-through-rate addition that can be
It can show that the first identifier of corresponding characteristic, illustratively, the mark of addition can be the type of characteristic, can be with
It is the first identifier of every kind of characteristic, can also be third mark corresponding with every kind of characteristic.Alternatively, can will obtain
Prediction click-through-rate store to storage location corresponding with characteristic.Illustratively, every kind can be previously stored with
In the storage table of characteristic, prediction click-through-rate is subjected to corresponding storage with characteristic;The prediction that can also will be obtained
Click-through-rate is stored to designated storage location corresponding with characteristic, for example, type is the characteristic pair of " type T1 "
The prediction click-through-rate P1 answered can store in designated storage location S1, and type is that the characteristic of " type T2 " is corresponding pre-
Surveying click-through-rate P2 can store in designated storage location S2.
In scheme provided in an embodiment of the present invention, when carrying out click-through-rate prediction, preset prediction model is preparatory
The model obtained using the prediction result label training of sample characteristics data and sample characteristics data, and sample characteristics data with to
The type of the characteristic of the web advertisement of dispensing is identical, and the type of characteristic is wide for the network reflected according to characteristic
The type that the feature of announcement divides.Therefore, for every kind of acquired characteristic, this kind of characteristic input is corresponding default
Prediction model after, the corresponding prediction click-through-rate of this kind of characteristic can be obtained.Due to for the web advertisement
Different types of characteristic obtains the corresponding prediction click-through-rate of this kind of characteristic, according to prediction click-through-rate
When adjusting the feature of the web advertisement to be put, can determine acquired prediction click-through-rate is had an impact it is to be put
The web advertisement characteristic, therefore, operation maintenance personnel can pointedly adjust the feature that characteristic is reflected.With needs
The feature that operation maintenance personnel tentatively adjusts the web advertisement to be put one by one is compared, and is saved and is clicked through on no influence prediction
The Character adjustment convenience and efficiency of the web advertisement to be put can be improved in the step of tentative adjustment of the feature of rate.
Optionally, at least one characteristic of the above-mentioned acquisition web advertisement to be put, can specifically include following step
It is rapid:
Obtain the type of the characteristic of the web advertisement to be put and multiple characteristics of the web advertisement to be put
According to;
According to the type of acquired characteristic, the building rule of every kind of characteristic is determined;
Respectively according to the building rule of every kind of characteristic, using multiple characteristics, this kind of characteristic is constructed;
At least one characteristic by constructed at least one obtained characteristic, as the web advertisement to be put
According to.
In a particular application, allow the feature of the web advertisement to be put adjusted may not under different application scenarios
Together, for example, only this is a kind of in full clothes broadcast for the dispensing form of the web advertisement in certain game application, then click-through-rate is being carried out
When prediction, without being predicted for this feature of dispensing form.Therefore, the characteristic of the available web advertisement to be put
According to type, with obtain it is corresponding with current application scenarios at least one characteristic, avoid the feature for obtaining fixed type
When data, the redundancy for the characteristic that current application scenarios will not adjust is obtained, the wasting of resources is avoided.Wherein, to be put
The web advertisement characteristic type acquisition modes can be it is a variety of.Illustratively, it is defeated to can receive maintenance personnel
The type of the characteristic of the web advertisement to be put entered.Alternatively, it is illustrative, the web advertisement pair to be put can be read
The configuration file answered extracts the type of the characteristic recorded in the configuration file.It is any that can to obtain network to be put wide
The mode of the type of the characteristic of announcement, is used equally for the present invention, the present embodiment to this with no restriction.
In addition, the acquisition modes of multiple characteristics of the web advertisement to be put can be it is a variety of.Illustratively, when
When web advertisement to be put itself carries characteristic, at least one directly can be extracted from the web advertisement to be put
Characteristic.Alternatively, it is illustrative, when characteristic carries the dropping apparatus information or attribute in the web advertisement to be put
When in information, at least one characteristic can be extracted from the dropping apparatus information or attribute information of the web advertisement to be put
According to.For example, extracting in release time, placement position and the dispensing form of the web advertisement to be put etc. characteristic extremely
It is one few.
The building rule of different types of characteristic can be a variety of.Illustratively, building rule may include by
The characteristic found is spliced into a kind of feature as a kind of characteristic, by least two characteristics found
At least two specified characteristics using specific characteristic data as a kind of characteristic, and are spliced into a kind of spy by data
Levy data etc..Wherein, splicing form may include that at least two characteristics are spliced according to preset sequence, Huo Zhezhi
Few two characteristics carry out splicing in any order etc..Building rule it is specific be arranged can according to application scenarios into
Row, the building rule of any different types of characteristic are used equally for the present invention, the present embodiment to this with no restriction.
In addition, determining the building rule of every kind of characteristic according to the type of acquired characteristic, specifically can wrap
It includes: for the type of acquired each characteristic, from the type of preset characteristic and the corresponding relationship of building rule
In, search the corresponding building rule of type of this feature data.Illustratively, the type of preset characteristic and building rule
Corresponding relationship, comprising: " type T1 " corresponding building rule " splicing user's gender and age of user ", " type T2 " corresponding building
Rule " placement position of the splicing web advertisement and release time ", " type T3 " corresponding building rule is " by the content of the web advertisement
Form is as a kind of characteristic ".The type of acquired characteristic is " type T1 " and " type T2 ", then from preset spy
It levies in the type of data and the corresponding relationship of building rule, the building rule for finding the characteristic of " type T1 " is " splicing
User's gender and age of user ", the building rule of the characteristic of " type T2 " are " placement position and throwing of the splicing web advertisement
Put the time ".
It, can be respectively according to the building rule of every kind of characteristic, benefit after the building rule for determining every kind of characteristic
With multiple characteristics, this kind of characteristic is constructed, thus by constructed at least one obtained characteristic, as to be put
The web advertisement at least one characteristic.Illustratively, the building according to the characteristic of " type T1 " is regular, utilizes spy
Levy data " user's gender female " and " age of user 26 ", building " type T1 " characteristic " female, 26 ";According to type T2 "
The building rule of characteristic, using characteristic " advertisement position that the placement position of the web advertisement is TV play A " and " when dispensing
Between TV play A start play at the time of ", construct " type T2 " characteristic " advertisement position of TV play A, TV play A start to broadcast
At the time of putting ".
For above-mentioned alternative embodiment, and the building without characteristic, but multiple characteristics are fully entered
Preset prediction model is compared, the above-mentioned building to different types of characteristic, will screen the function of characteristic from default
Prediction model in remove, realize to characteristic obtain and click-through-rate prediction between content decouple.Realize content solution
After coupling, the function that the function of preset prediction model is obtained with characteristic is independent from each other, only need to be to one of function
When can be carried out extension, maintenance personnel can be extended directly against the corresponding content of the function, the corresponding content of another function
It is sightless to maintenance personnel.Compared with when not decoupling, when maintenance personnel carries out Function Extension, without being looked for from a large amount of contents
The corresponding content of the function to be extended out, can be improved easy to maintain and efficiency.
Optionally, using multiple characteristics, this kind is constructed respectively according to the building rule of every kind of characteristic above-mentioned
After characteristic, the prediction technique of click-through-rate provided in an embodiment of the present invention can also include:
At least one characteristic that building is obtained respectively inputs preset hashing algorithm, obtains every kind of characteristic point
Not corresponding hashed value;
Using obtained hashed value as at least one characteristic of the web advertisement to be put.
It in a particular application, is that multiple characteristics splice to obtain due to constructing at least one obtained characteristic,
Therefore, compared with not splicing characteristic, in contrast more memory space may be occupied.It is obtained to reduce building
Characteristic can use preset hashing algorithm, such as hash algorithm to the occupancy of memory space, and building is obtained at least
The form that one characteristic is converted to Key-Value (key-value) is stored, and Value is characterized data the application, and then will
At least one characteristic of the obtained hashed value as the web advertisement to be put.The subsequent preset prediction model of input
Characteristic can be the Key to this kind of characteristic of input.Illustratively, by the characteristic of " type T1 ", " female, 26 " is defeated
Enter preset hashing algorithm, obtains the Key " 1 " of this feature data;By the characteristic of type T2 " " advertisement position of TV play A,
At the time of TV play A starts to play ", obtain the Key " 2 " of this feature data.
As shown in Fig. 2, the process of the prediction technique of the click-through-rate of another embodiment of the present invention, this method can wrap
It includes:
S201 obtains at least one characteristic of the web advertisement to be put.
S201 is identical step with the S101 of Fig. 1 embodiment of the present invention, and details are not described herein, and it is real to be detailed in Fig. 1 of the present invention
Apply the description of example.
S202, it is corresponding with preset prediction model from preset characteristic for every kind of acquired characteristic
In relationship, the corresponding preset prediction model of this kind of characteristic is searched.Each preset prediction model respectively corresponds a line
Journey, thread is for loading the corresponding preset prediction model of the thread.
S202 is similar step to the S102 of Fig. 1 embodiment of the present invention, and difference is each preset prediction in S202
Model respectively corresponds a thread, and thread is for loading the corresponding preset prediction model of the thread.Herein for same section
It repeats no more, is detailed in the description of Fig. 1 embodiment of the present invention.
In order to improve various features data are carried out with forecasting efficiency when click-through-rate prediction, it can be to used more
A preset prediction model carries out loaded in parallel, for this purpose, a thread, line can be respectively set for each preset prediction model
Journey is for loading the corresponding preset prediction model of the thread.Also, when in order to guarantee multiple thread concurrent activities, per thread
The normal operation of the task of execution needs to execute step before characteristic is inputted corresponding preset prediction model
S203.
S203 detects whether the corresponding thread of this kind of characteristic has terminated to transport for every kind of acquired characteristic
Row;If it has ended, otherwise executing step S204 returns to step S203.
Wherein, the detection mode that whether the corresponding thread of every kind of characteristic has terminated operation can be a variety of.
Illustratively, this kind of characteristic can be judged after the preset duration for finding the corresponding thread of every kind of characteristic
Whether corresponding thread returns to first state word, if returned, determines that testing result is that the thread has terminated to run;Wherein first
Status word is to show that thread terminates the status word of operation.Alternatively, the corresponding thread of every kind of characteristic can found
Afterwards, directly judge whether the corresponding thread of this kind of characteristic returns to the second status word, if returned, determine that testing result is to be somebody's turn to do
Operation is not finished in thread;Wherein the second status word is to show that thread is the status word of active state.Whether any pair of thread has been tied
The detection mode of Shu Yunhang is used equally for the present invention, the present embodiment to this with no restriction.
When being directed to every kind of acquired characteristic, detecting the corresponding thread of this kind of characteristic has terminated to run, and shows
The task that the thread is not carried out, obtaining prediction click-through-rate using the thread not will cause thread activity exception, or number
It is abnormal according to obtaining, therefore, step S204 can be executed.When being directed to every kind of acquired characteristic, this kind of characteristic is detected
Operation is not finished in corresponding thread, and showing the thread, there are also executing for tasks, obtains prediction click-through-rate meeting using the thread
It causes thread activity abnormal or data acquisition exception, therefore, S203 can be returned to step, to wait the thread to run
Terminate, also, carries out the acquisition of prediction click-through-rate using the thread in time in end of run.
S204 loads the corresponding preset prediction model of this kind of characteristic using the corresponding thread of this kind of characteristic,
And this kind of characteristic is inputted into corresponding preset prediction model, it obtains the corresponding prediction of this kind of characteristic and clicks through
Rate.
S204 and the S103 of Fig. 1 embodiment of the present invention are similar step, and difference is in S204 for acquired every
Kind characteristic loads the corresponding preset prediction model of this kind of characteristic, energy using the corresponding thread of this kind of characteristic
It is enough to guarantee that the task of the thread corresponds to characteristic for per thread, caused by can be avoided data input exception
Thread executes confusion.For same section, details are not described herein, is detailed in the description of Fig. 1 embodiment of the present invention.
In above-mentioned Fig. 2 embodiment, using multiple threads, the input of different types of characteristic can be carried out simultaneously,
The acquisition of the prediction click-through-rate of every kind of characteristic is mutually indepedent, compared with the mode serially predicted, obtains without waiting
After a kind of corresponding prediction click-through-rate of characteristic, then the corresponding prediction click-through-rate of another characteristic is obtained,
It can effectively improve the acquisition efficiency of prediction click-through-rate.For example, it in serial prediction mode, completes a certain to be put
The web advertisement click-through-rate prediction, probably need 14105813 microseconds;And in the mode of multiple thread parallels prediction
In, the prediction of the click-through-rate of a certain web advertisement to be put is completed, 8988260 microseconds are probably needed.
Optionally, above-mentioned to be directed to every kind of acquired characteristic, this kind of characteristic input is corresponding preset pre-
Model is surveyed, the corresponding prediction click-through-rate of this kind of characteristic is obtained, can specifically include following steps:
For every kind of acquired characteristic, the second mark of the corresponding preset prediction model of this kind of characteristic is obtained
Know;
The corresponding second identifier of this kind of characteristic is inputted into preset prediction model interface function, obtains this kind of characteristic
According to corresponding prediction click-through-rate.
In a particular application, preset prediction model can be packaged, characteristic is obtained and is clicked to realize
Content decoupling between percent of pass prediction.To be called by prediction model interface function to preset prediction model.It is real
After existing content decoupling, the function that the function of preset prediction model is obtained with characteristic is independent from each other, only need to be to it
In function when being extended, maintenance personnel can be extended directly against the corresponding content of the function, another function pair
The content answered is sightless to maintenance personnel.Compared with when not decoupling, when maintenance personnel carries out Function Extension, it is not necessarily to from a large amount of
The corresponding content of the function to be extended is found out in content, and easy to maintain and efficiency can be improved.
Wherein, it is called to realize, the corresponding relationship of preset characteristic and preset prediction model can be utilization
The mapping table that the first identifier of characteristic and the second identifier of preset prediction model are established.Correspondingly, for being obtained
The every kind of characteristic taken obtains the second identifier of the corresponding preset prediction model of this kind of characteristic, specifically can be from
In the corresponding relationship of preset characteristic and preset prediction model, the first identifier of this kind of acquired characteristic is searched
Corresponding second identifier.By the second identifier input prediction model interface function of the preset prediction model found
The corresponding preset prediction model of the second identifier is called, so that the preset prediction model called calculates this kind of characteristic
Corresponding prediction click-through-rate.
Corresponding to above method embodiment, one embodiment of the invention additionally provides the prediction meanss of click-through-rate.
As shown in figure 3, the structure of the prediction meanss of the click-through-rate of one embodiment of the invention, the apparatus may include:
Characteristic obtains module 301, for obtaining at least one characteristic of the web advertisement to be put;Characteristic
According to the type that is divided for the feature of the web advertisement that is reflected according to characteristic of type;
Prediction model obtains module 302, for being directed to every kind of acquired characteristic, from preset characteristic and in advance
If prediction model corresponding relationship in, search the corresponding preset prediction model of this kind of characteristic;The preset prediction
Model is the model for advancing with the prediction result label training of sample characteristics data and sample characteristics data and obtaining, and sample is special
The type for levying data characteristic corresponding with the preset prediction model is identical;
Prediction module 303, it is for being directed to every kind of acquired characteristic, this kind of characteristic input is corresponding default
Prediction model, obtain the corresponding prediction click-through-rate of this kind of characteristic.
In scheme provided in an embodiment of the present invention, when carrying out click-through-rate prediction, preset prediction model is preparatory
The model obtained using the prediction result label training of sample characteristics data and sample characteristics data, and sample characteristics data with to
The type of the characteristic of the web advertisement of dispensing is identical, and the type of characteristic is wide for the network reflected according to characteristic
The type that the feature of announcement divides.Therefore, for every kind of acquired characteristic, this kind of characteristic input is corresponding default
Prediction model after, the corresponding prediction click-through-rate of this kind of characteristic can be obtained.Due to for the web advertisement
Different types of characteristic obtains the corresponding prediction click-through-rate of this kind of characteristic, according to prediction click-through-rate
When adjusting the feature of the web advertisement to be put, can determine acquired prediction click-through-rate is had an impact it is to be put
The web advertisement characteristic, therefore, operation maintenance personnel can pointedly adjust the feature that characteristic is reflected.With needs
The feature that operation maintenance personnel tentatively adjusts the web advertisement to be put one by one is compared, and is saved and is clicked through on no influence prediction
The Character adjustment convenience and efficiency of the web advertisement to be put can be improved in the step of tentative adjustment of the feature of rate.
Optionally, the characteristic obtains module 401, is specifically used for:
Obtain the type of the characteristic of the web advertisement to be put and multiple spies of the web advertisement to be put
Levy data;
According to the type of acquired characteristic, the building rule of every kind of characteristic is determined;
Respectively according to the building rule of every kind of characteristic, using the multiple characteristic, this kind of feature is constructed
Data;
At least one feature by constructed at least one obtained characteristic, as the web advertisement to be put
Data.
Optionally, the characteristic obtains module 301, is also used to respectively according to the building of every kind of characteristic
Rule after constructing this kind of characteristic, will construct at least one obtained characteristic using the multiple characteristic respectively
According to preset hashing algorithm is inputted, the corresponding hashed value of every kind of characteristic is obtained;
Using obtained hashed value as at least one characteristic of the web advertisement to be put.
Optionally, the prediction module 403, is specifically used for:
For every kind of acquired characteristic, the second mark of the corresponding preset prediction model of this kind of characteristic is obtained
Know;
The corresponding second identifier of this kind of characteristic is inputted into preset prediction model interface function, obtains this kind of characteristic
According to corresponding prediction click-through-rate.
As shown in figure 4, the structure of the prediction meanss of the click-through-rate of another embodiment of the present invention, which be can wrap
It includes:
Characteristic obtains module 401, for obtaining at least one characteristic of the web advertisement to be put;Characteristic
According to the type that is divided for the feature of the web advertisement that is reflected according to characteristic of type;
Prediction model obtains module 402, for being directed to every kind of acquired characteristic, from preset characteristic and in advance
If prediction model corresponding relationship in, search the corresponding preset prediction model of this kind of characteristic;The preset prediction
Model is the model for advancing with the prediction result label training of sample characteristics data and sample characteristics data and obtaining, and sample is special
The type for levying data characteristic corresponding with the preset prediction model is identical;Each preset prediction model respectively corresponds one
A thread;The thread is for loading the corresponding preset prediction model of the thread;
Prediction module 403, it is for being directed to every kind of acquired characteristic, this kind of characteristic input is corresponding default
Prediction model, obtain the corresponding prediction click-through-rate of this kind of characteristic.
The prediction module 403, including judging submodule 4031 and load submodule 4032;
It is corresponding to detect this kind of characteristic for being directed to every kind of acquired characteristic for the judging submodule 4031
Thread whether terminated to run;If it has ended, the triggering load submodule 4032 is run;
The load submodule 4032, for utilizing this kind of characteristic under the triggering of the judging submodule 4031
Corresponding thread loads the corresponding preset prediction model of this kind of characteristic, and this kind of characteristic input is corresponding pre-
If prediction model, obtain the corresponding prediction click-through-rate of this kind of characteristic, otherwise, execute this kind of characteristic of the detection
Whether terminate to run according to corresponding thread.
Corresponding to above-described embodiment, the embodiment of the invention also provides a kind of electronic equipment, as shown in figure 5, the equipment can
To include:
Processor 501, communication interface 502, memory 503 and communication bus 504, wherein processor 501, communication interface
502, memory logical 503 crosses communication bus 504 and completes mutual communication;
Memory 503, for storing computer program;
Processor 501 when for executing the computer program stored on above-mentioned memory 503, realizes above-described embodiment
In any click-through-rate prediction technique the step of.
In scheme provided in an embodiment of the present invention, when carrying out click-through-rate prediction, preset prediction model is preparatory
The model obtained using the prediction result label training of sample characteristics data and sample characteristics data, and sample characteristics data with to
The type of the characteristic of the web advertisement of dispensing is identical, and the type of characteristic is wide for the network reflected according to characteristic
The type that the feature of announcement divides.Therefore, for every kind of acquired characteristic, this kind of characteristic input is corresponding default
Prediction model after, the corresponding prediction click-through-rate of this kind of characteristic can be obtained.Due to for the web advertisement
Different types of characteristic obtains the corresponding prediction click-through-rate of this kind of characteristic, according to prediction click-through-rate
When adjusting the feature of the web advertisement to be put, can determine acquired prediction click-through-rate is had an impact it is to be put
The web advertisement characteristic, therefore, operation maintenance personnel can pointedly adjust the feature that characteristic is reflected.With needs
The feature that operation maintenance personnel tentatively adjusts the web advertisement to be put one by one is compared, and is saved and is clicked through on no influence prediction
The Character adjustment convenience and efficiency of the web advertisement to be put can be improved in the step of tentative adjustment of the feature of rate.
Above-mentioned memory may include RAM (Random Access Memory, random access memory), also may include
NVM (Non-Volatile Memory, nonvolatile memory), for example, at least a magnetic disk storage.Optionally, memory
It can also be that at least one is located away from the storage device of above-mentioned processor.
Above-mentioned processor can be general processor, including CPU (Central Processing Unit, central processing
Device), NP (Network Processor, network processing unit) etc.;Can also be DSP (Digital Signal Processor,
Digital signal processor), ASIC (Application Specific Integrated Circuit, specific integrated circuit),
FPGA (Field-Programmable Gate Array, field programmable gate array) or other programmable logic device are divided
Vertical door or transistor logic, discrete hardware components.
The computer readable storage medium that one embodiment of the invention provides, is contained in electronic equipment, this is computer-readable to deposit
It is stored with computer program in storage media, when which is executed by processor, realizes any click in above-described embodiment
The step of prediction technique of percent of pass.
In scheme provided in an embodiment of the present invention, when carrying out click-through-rate prediction, preset prediction model is preparatory
The model obtained using the prediction result label training of sample characteristics data and sample characteristics data, and sample characteristics data with to
The type of the characteristic of the web advertisement of dispensing is identical, and the type of characteristic is wide for the network reflected according to characteristic
The type that the feature of announcement divides.Therefore, for every kind of acquired characteristic, this kind of characteristic input is corresponding default
Prediction model after, the corresponding prediction click-through-rate of this kind of characteristic can be obtained.Due to for the web advertisement
Different types of characteristic obtains the corresponding prediction click-through-rate of this kind of characteristic, according to prediction click-through-rate
When adjusting the feature of the web advertisement to be put, can determine acquired prediction click-through-rate is had an impact it is to be put
The web advertisement characteristic, therefore, operation maintenance personnel can pointedly adjust the feature that characteristic is reflected.With needs
The feature that operation maintenance personnel tentatively adjusts the web advertisement to be put one by one is compared, and is saved and is clicked through on no influence prediction
The Character adjustment convenience and efficiency of the web advertisement to be put can be improved in the step of tentative adjustment of the feature of rate.
In another embodiment provided by the invention, a kind of computer program product comprising instruction is additionally provided, when it
When running on computers, so that computer executes the prediction technique of any click-through-rate in above-described embodiment.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or any combination thereof real
It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.The computer program
Product includes one or more computer instructions.When loading on computers and executing the computer program instructions, all or
It partly generates according to process or function described in the embodiment of the present invention.The computer can be general purpose computer, dedicated meter
Calculation machine, computer network or other programmable devices.The computer instruction can store in computer readable storage medium
In, or from a computer readable storage medium to the transmission of another computer readable storage medium, for example, the computer
Instruction can pass through wired (such as coaxial cable, optical fiber, DSL from a web-site, computer, server or data center
(Digital Subscriber Line, digital operation maintenance personnel line) or wireless (such as: infrared ray, radio, microwave etc.) mode
It is transmitted to another web-site, computer, server or data center.The computer readable storage medium can be
Any usable medium that computer can access either includes the integrated server of one or more usable mediums, data center
Equal data storage devices.The usable medium can be magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (such as:
DVD (Digital Versatile Disc, digital versatile disc)) or semiconductor medium (such as: SSD (Solid State
Disk, solid state hard disk)) etc..
Herein, relational terms such as first and second and the like be used merely to by an entity or operation with it is another
One entity or operation distinguish, and without necessarily requiring or implying between these entities or operation, there are any this reality
Relationship or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to the packet of nonexcludability
Contain, so that the process, method, article or equipment for including a series of elements not only includes those elements, but also including
Other elements that are not explicitly listed, or further include for elements inherent to such a process, method, article, or device.
In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including the element
Process, method, article or equipment in there is also other identical elements.
Each embodiment in this specification is all made of relevant mode and describes, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for device and
For electronic equipment embodiment, since it is substantially similar to the method embodiment, so be described relatively simple, related place referring to
The part of embodiment of the method illustrates.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all
Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention
It is interior.
Claims (10)
1. a kind of prediction technique of click-through-rate, which is characterized in that the described method includes:
Obtain at least one characteristic of the web advertisement to be put;The type of characteristic is is reflected according to characteristic
The web advertisement feature divide type;
It is looked into for every kind of acquired characteristic from the corresponding relationship of preset characteristic and preset prediction model
Look for the corresponding preset prediction model of this kind of characteristic;The preset prediction model be advance with sample characteristics data and
The model that the prediction result label training of sample characteristics data obtains, and sample characteristics data are corresponding with the preset prediction model
Characteristic type it is identical;
For every kind of acquired characteristic, this kind of characteristic is inputted into corresponding preset prediction model, obtains this kind
The corresponding prediction click-through-rate of characteristic.
2. the method according to claim 1, wherein at least one for obtaining the web advertisement to be put is special
Levy data, comprising:
Obtain the type of the characteristic of the web advertisement to be put and multiple characteristics of the web advertisement to be put
According to;
According to the type of acquired characteristic, the building rule of every kind of characteristic is determined;
Respectively according to the building rule of every kind of characteristic, using the multiple characteristic, this kind of characteristic is constructed;
At least one characteristic by constructed at least one obtained characteristic, as the web advertisement to be put
According to.
3. the method according to claim 1, wherein each preset prediction model respectively corresponds a thread;
The thread is for loading the corresponding preset prediction model of the thread;
It is described to be directed to every kind of acquired characteristic, this kind of characteristic is inputted into corresponding preset prediction model, is obtained
The corresponding prediction click-through-rate of this kind of characteristic, comprising:
For every kind of acquired characteristic, detect whether the corresponding thread of this kind of characteristic has terminated to run;
If it has ended, loading the corresponding preset prediction mould of this kind of characteristic using the corresponding thread of this kind of characteristic
Type, and this kind of characteristic is inputted into corresponding preset prediction model, it is logical to obtain the corresponding prediction click of this kind of characteristic
Rate is crossed, otherwise, executes whether the corresponding thread of this kind of characteristic of the detection has terminated to run.
4. the method according to claim 1, wherein described be directed to every kind of acquired characteristic, by this kind
Characteristic inputs corresponding preset prediction model, obtains the corresponding prediction click-through-rate of this kind of characteristic, comprising:
For every kind of acquired characteristic, the second identifier of the corresponding preset prediction model of this kind of characteristic is obtained;
The corresponding second identifier of this kind of characteristic is inputted into preset prediction model interface function, obtains this kind of characteristic pair
The prediction click-through-rate answered.
5. a kind of prediction meanss of click-through-rate, which is characterized in that described device includes:
Characteristic obtains module, for obtaining at least one characteristic of the web advertisement to be put;The kind of characteristic
The type that class divides for the feature of the web advertisement reflected according to characteristic;
Prediction model obtain module, for be directed to every kind of acquired characteristic, from preset characteristic with it is preset pre-
It surveys in the corresponding relationship of model, searches the corresponding preset prediction model of this kind of characteristic;The preset prediction model is
Advance with the model that the prediction result label training of sample characteristics data and sample characteristics data obtains, and sample characteristics data
The type of characteristic corresponding with the preset prediction model is identical;
This kind of characteristic is inputted corresponding preset prediction for being directed to every kind of acquired characteristic by prediction module
Model obtains the corresponding prediction click-through-rate of this kind of characteristic.
6. device according to claim 5, which is characterized in that the characteristic obtains module, is specifically used for:
Obtain the type of the characteristic of the web advertisement to be put and multiple characteristics of the web advertisement to be put
According to;
According to the type of acquired characteristic, the building rule of every kind of characteristic is determined;
Respectively according to the building rule of every kind of characteristic, using the multiple characteristic, this kind of characteristic is constructed;
At least one characteristic by constructed at least one obtained characteristic, as the web advertisement to be put
According to.
7. device according to claim 5, which is characterized in that each preset prediction model respectively corresponds a thread;
The thread is for loading the corresponding preset prediction model of the thread;
The prediction module, including judging submodule and load submodule;
The judging submodule, for being directed to every kind of acquired characteristic, detecting the corresponding thread of this kind of characteristic is
It is no to have terminated to run;If it has ended, the triggering load submodule operation;
The load submodule, for using the corresponding thread of this kind of characteristic, adding under the triggering of the judging submodule
The corresponding preset prediction model of this kind of characteristic is carried, and this kind of characteristic is inputted into corresponding preset prediction model,
The corresponding prediction click-through-rate of this kind of characteristic is obtained, otherwise, executes the corresponding thread of this kind of characteristic of the detection
Whether terminate to run.
8. device according to claim 5, which is characterized in that the prediction module is specifically used for:
For every kind of acquired characteristic, the second identifier of the corresponding preset prediction model of this kind of characteristic is obtained;
The corresponding second identifier of this kind of characteristic is inputted into preset prediction model interface function, obtains this kind of characteristic pair
The prediction click-through-rate answered.
9. a kind of electronic equipment, which is characterized in that including processor, communication interface, memory and communication bus, wherein processing
Device, communication interface, memory complete mutual communication by bus;Memory, for storing computer program;Processor,
For executing the program stored on memory, the method and step as described in claim 1-4 is any is realized.
10. a kind of computer readable storage medium, which is characterized in that computer program is stored in the storage medium, it is described
The method and step as described in claim 1-4 is any is realized when computer program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910444789.2A CN110223108B (en) | 2019-05-27 | 2019-05-27 | Click through rate prediction method, device and equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910444789.2A CN110223108B (en) | 2019-05-27 | 2019-05-27 | Click through rate prediction method, device and equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110223108A true CN110223108A (en) | 2019-09-10 |
CN110223108B CN110223108B (en) | 2022-05-24 |
Family
ID=67818015
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910444789.2A Active CN110223108B (en) | 2019-05-27 | 2019-05-27 | Click through rate prediction method, device and equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110223108B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112115365A (en) * | 2020-09-25 | 2020-12-22 | 贝壳技术有限公司 | Model collaborative optimization method, device, medium and electronic equipment |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150178783A1 (en) * | 2012-05-14 | 2015-06-25 | Iqzone, Inc. | Systems and methods for providing timely advertising to portable devices |
CN105227975A (en) * | 2015-09-29 | 2016-01-06 | 北京奇艺世纪科技有限公司 | Advertisement placement method and device |
CN105630996A (en) * | 2015-12-25 | 2016-06-01 | 腾讯科技(深圳)有限公司 | Information processing method and server |
CN106803190A (en) * | 2017-01-03 | 2017-06-06 | 北京掌阔移动传媒科技有限公司 | A kind of ad personalization supplying system and method |
WO2017219548A1 (en) * | 2016-06-20 | 2017-12-28 | 乐视控股(北京)有限公司 | Method and device for predicting user attributes |
CN107657486A (en) * | 2017-10-19 | 2018-02-02 | 厦门美柚信息科技有限公司 | A kind of advertisement placement method and device |
CN107742221A (en) * | 2016-08-23 | 2018-02-27 | 腾讯科技(深圳)有限公司 | A kind of processing method of promotion message, device and system |
CN108629630A (en) * | 2018-05-08 | 2018-10-09 | 广州太平洋电脑信息咨询有限公司 | A kind of feature based intersects the advertisement recommendation method of joint deep neural network |
CN109034880A (en) * | 2018-07-09 | 2018-12-18 | 麒麟合盛网络技术股份有限公司 | revenue prediction method and device |
CN109740068A (en) * | 2019-01-29 | 2019-05-10 | 腾讯科技(北京)有限公司 | Media data recommended method, device and storage medium |
CN110188942A (en) * | 2019-05-27 | 2019-08-30 | 北京金山安全软件有限公司 | Click through rate prediction method, device and equipment |
-
2019
- 2019-05-27 CN CN201910444789.2A patent/CN110223108B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150178783A1 (en) * | 2012-05-14 | 2015-06-25 | Iqzone, Inc. | Systems and methods for providing timely advertising to portable devices |
CN105227975A (en) * | 2015-09-29 | 2016-01-06 | 北京奇艺世纪科技有限公司 | Advertisement placement method and device |
CN105630996A (en) * | 2015-12-25 | 2016-06-01 | 腾讯科技(深圳)有限公司 | Information processing method and server |
WO2017219548A1 (en) * | 2016-06-20 | 2017-12-28 | 乐视控股(北京)有限公司 | Method and device for predicting user attributes |
CN107742221A (en) * | 2016-08-23 | 2018-02-27 | 腾讯科技(深圳)有限公司 | A kind of processing method of promotion message, device and system |
CN106803190A (en) * | 2017-01-03 | 2017-06-06 | 北京掌阔移动传媒科技有限公司 | A kind of ad personalization supplying system and method |
CN107657486A (en) * | 2017-10-19 | 2018-02-02 | 厦门美柚信息科技有限公司 | A kind of advertisement placement method and device |
CN108629630A (en) * | 2018-05-08 | 2018-10-09 | 广州太平洋电脑信息咨询有限公司 | A kind of feature based intersects the advertisement recommendation method of joint deep neural network |
CN109034880A (en) * | 2018-07-09 | 2018-12-18 | 麒麟合盛网络技术股份有限公司 | revenue prediction method and device |
CN109740068A (en) * | 2019-01-29 | 2019-05-10 | 腾讯科技(北京)有限公司 | Media data recommended method, device and storage medium |
CN110188942A (en) * | 2019-05-27 | 2019-08-30 | 北京金山安全软件有限公司 | Click through rate prediction method, device and equipment |
Non-Patent Citations (2)
Title |
---|
赵晗等: "基于XGBoost的搜索结果智能排序系统", 《软件导刊》 * |
顾智宇等: "一种基于因子图的搜索广告转化预测模型", 《中文信息学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112115365A (en) * | 2020-09-25 | 2020-12-22 | 贝壳技术有限公司 | Model collaborative optimization method, device, medium and electronic equipment |
CN112115365B (en) * | 2020-09-25 | 2021-09-14 | 贝壳找房(北京)科技有限公司 | Model collaborative optimization method, device, medium and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN110223108B (en) | 2022-05-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10769669B2 (en) | Determining an invitational content item type based on predicted user attention | |
CN103577658B (en) | The appraisal procedure and assessment system of intelligent terminal hardware | |
US8041652B2 (en) | Measuring web site satisfaction of information needs using page traffic profile | |
US20140279787A1 (en) | Systems And Methods for an Adaptive Application Recommender | |
US20090287989A1 (en) | Presenting tag clouds based on user information | |
CN109992498B (en) | Test case generation method and system and computer system | |
CN106557424A (en) | Internal storage testing method, measured terminal, test client and system | |
CN107872534B (en) | Information pushing method and device, server and readable storage medium | |
US20150269615A1 (en) | Predicting Visibility of Content Items | |
CN105389325A (en) | Content search method and electronic device implementing same | |
US20120096030A1 (en) | Method and apparatus for providing search results by using previous query | |
CN106294719B (en) | Information display method and device | |
CN110704677B (en) | Program recommendation method and device, readable storage medium and terminal equipment | |
CN110262918A (en) | Process collapses analysis method and device, distributed apparatus and storage medium | |
CN107783886A (en) | A kind of method and terminal for obtaining operation frame per second | |
US20160036878A1 (en) | Stream processing with context data affinity | |
CN108154024A (en) | A kind of data retrieval method, device and electronic equipment | |
CN110223108A (en) | Click through rate prediction method, device and equipment | |
CN109885729A (en) | A kind of method, apparatus and system showing data | |
CN109688217A (en) | A kind of information push method, device and electronic equipment | |
CN106233284A (en) | The method and system stablizing identifier is generated for the node potentially including main contents in information resources | |
CN107257358A (en) | A kind of information-pushing method, terminal and server | |
CN110189171A (en) | Feature data generation method, device and equipment | |
CN110245072A (en) | A kind of test method and device | |
US11256859B2 (en) | Extending a classification database by user interactions |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |