CN108965951A - The playback method and device of advertisement - Google Patents
The playback method and device of advertisement Download PDFInfo
- Publication number
- CN108965951A CN108965951A CN201810762883.8A CN201810762883A CN108965951A CN 108965951 A CN108965951 A CN 108965951A CN 201810762883 A CN201810762883 A CN 201810762883A CN 108965951 A CN108965951 A CN 108965951A
- Authority
- CN
- China
- Prior art keywords
- preset
- model
- advertisement
- clicking rate
- data
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/266—Channel or content management, e.g. generation and management of keys and entitlement messages in a conditional access system, merging a VOD unicast channel into a multicast channel
- H04N21/2668—Creating a channel for a dedicated end-user group, e.g. insertion of targeted commercials based on end-user profiles
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44213—Monitoring of end-user related data
- H04N21/44222—Analytics of user selections, e.g. selection of programs or purchase activity
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4667—Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/81—Monomedia components thereof
- H04N21/812—Monomedia components thereof involving advertisement data
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Social Psychology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Business, Economics & Management (AREA)
- Marketing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The playback method and device of advertisement provided in an embodiment of the present invention play the FM model trained that the Current ad data of advertisement train by obtaining current time；By ad data to be selected concentrate each ad data to be selected be input in the FM model trained, the second clicking rate of each advertisement to be selected of output estimation；The highest advertisement to be selected of the second clicking rate is selected, is played out in the preset time after current time.The FM model that the present embodiment has been trained using Current ad data, real-time are higher；Go out the clicking rate of each advertisement to be selected using the FM model pre-estimating trained, accuracy is higher；Then the selection highest advertisement to be selected of clicking rate plays out.The more matching of the advertisement to be selected and the required product advertising of user that play.It is thus possible to improve playing the clicking rate of advertisement.
Description
Technical field
The present invention relates to field of computer technology, more particularly to the playback method and device of a kind of advertisement.
Background technique
The network platform passes through the clicking rate size of advertisement, plays to the playing duration of advertisement, the broadcasting time of advertisement, advertisement
Flow is reasonably distributed, and is formulated advertising expense charging standard, is then played advertisement.
In the prior art, the playing process of advertisement is as follows:
Use the history ad data in database as advertising copy, using advertising copy as comprising presetting parameters
FM (Factorization Machine, Factorization machine) algorithm input, ad click rate in history ad data is made
For training objective, trains and export the FM model for ad click rate.The each advertisement played will be needed as advertisement to be selected, so
Each ad data to be selected of acquisition is input to the FM model after optimization, the click of each advertisement to be selected of output estimation afterwards
Rate selects the highest advertisement to be selected of clicking rate as the advertisement played online.
When many times being had been subjected to due to the history ad data used in the prior art, but user is for the need of product
It asks and often changes.Therefore, in the prior art, the FM model trained by using history ad data is carried out in advance really
Property may be lower when survey, so that the clicking rate for the advertisement to be selected estimated out is often less accurate, in the advertisement played online
Product often may differ by with product needed for user very remote, user may not click the advertisement played online, to lead
Cause the clicking rate of the advertisement played online may be lower.
Summary of the invention
The playback method and device for being designed to provide a kind of advertisement of the embodiment of the present invention are played using current time
Advertisement characteristic data trains the FM model trained in real time, improves the FM model real-time trained, and raising is estimated to be selected
The clicking rate accuracy of advertisement, therefore the online clicking rate for playing advertisement can be improved.Specific technical solution is as follows:
In a first aspect, the embodiment of the invention provides a kind of playback methods of advertisement, comprising:
Obtain the Current ad data that current time plays advertisement；Current ad data include: current-user data, current
Advertisement characteristic data and the first clicking rate；First clicking rate is the clicking rate that user clicks that the advertisement that current time plays generates；
Using Current ad characteristic and current-user data as the input of preset FM model, the first clicking rate is made
For the training objective of preset FM model output, the preset FM model of training, using the preset FM model after the completion of training as
The FM model trained；
Each ad data to be selected that preset ad data to be selected is concentrated is input in the FM model trained, is exported
Second clicking rate of each advertisement to be selected estimated；Ad data to be selected includes: user data and advertisement characteristic data；
The highest advertisement to be selected of the second clicking rate is selected, is played out in the preset time after current time.
Optionally, using Current ad characteristic and current-user data as the input of preset FM model, by first
The training objective that clicking rate is exported as preset FM model, the preset FM model of training, by the preset FM after the completion of training
Model is as the FM model trained, comprising:
Using Current ad characteristic and current-user data as the input of preset FM model, by the first clicking rate with
Third clicking rate is made comparisons, and using gradient descent algorithm, determines that the first clicking rate and preset FM model export third clicking rate
Error amount whether be minimum；Third clicking rate are as follows: the input of preset FM model is Current ad characteristic and current use
User data, the clicking rate of preset FM model output；
If the error amount of the first clicking rate and third clicking rate is not minimum, the parameters of preset FM model are adjusted
Until the first clicking rate and third clicking rate error amount are minimum；
Using the preset FM model after adjustment parameters as the FM model trained.
Optionally, using Current ad characteristic and current-user data as the input of preset FM model, by first
The training objective that clicking rate is exported as preset FM model, the preset FM model of training, by the preset FM after the completion of training
Model is as the FM model trained, comprising:
Using Current ad characteristic and current-user data as the input of preset FM model, by the first clicking rate with
Third clicking rate is made comparisons, and the error amount for obtaining the first clicking rate and third clicking rate is calculated；
Using preset parameters in error amount and preset FM model as the defeated of preset Maximum-likelihood estimation function
Enter, the output of preset Maximum-likelihood estimation function is taken into logarithm, obtains the loss function of preset FM model output；
Using following regular leader's FTRL algorithm to calculate loss function, for the ladder of the parameters of preset FM model
Degree determines whether the loss function of preset FM model output is minimum；
If loss function value is not minimum, the parameters of preset FM model are adjusted until loss function value is minimum；
Using the preset FM model after adjustment parameters as the FM model trained.
Optionally, preset FM model is to train acquisition in advance as follows:
Using the history ad data of an advertisement as a training sample, training set is obtained；Training set includes: more
Training sample；Every training sample includes: that historical use data, history advertisement characteristic data and user's click advertisement generate
Historic click-through rate；
Using in training set, each training sample is as the input of initial FM model, by the history point in each training sample
Rate is hit as initial FM model training target；
The gradient descent direction for being intended the pseudo- gradient function of newton OWLQN algorithm setting using normal state finite memory, is calculated
The parameters of preset loss function value under the direction determine whether the loss function of preset FM model output is minimum；
If loss function value is not minimum, the parameters of preset FM model are adjusted until loss function value is minimum；
Using the initial FM model after adjustment parameters as preset FM model.
Optionally, following steps can be used, preset advertisement data set to be selected is obtained:
The advertisement characteristic data of one advertisement to be selected and a user data are combined, as a forecast sample；
Obtain preset advertisement data set to be selected；Preset advertisement data set to be selected includes: a plurality of forecast sample.
Optionally, the embodiment of the invention provides a kind of playback methods of advertisement further include:
Using Current ad data as an advertising copy, preset advertisement data set to be selected is added, update it is preset to
Select advertisement data set.
Second aspect, the embodiment of the invention provides a kind of playing devices of advertisement, comprising:
Module is obtained, the Current ad data of advertisement are played for obtaining current time；Current ad data include: current
User data, Current ad characteristic and the first clicking rate；First clicking rate is the advertisement that user clicks that current time plays
The clicking rate of generation；
Training module, for using Current ad characteristic and current-user data as the input of preset FM model,
The training objective that first clicking rate is exported as preset FM model, the preset FM model of training, will be pre- after the completion of training
If FM model as the FM model trained；
Module is estimated, each ad data to be selected for concentrating preset ad data to be selected, which is input to, has trained
In FM model, the second clicking rate of each advertisement to be selected of output estimation；Ad data to be selected includes: that user data and advertisement are special
Levy data；
Playing module, for selecting the highest advertisement to be selected of the second clicking rate, in the preset time after current time into
Row plays.
Optionally, training module includes:
First error unit, for using Current ad characteristic and current-user data as the defeated of preset FM model
Enter, the first clicking rate is made comparisons with third clicking rate, using gradient descent algorithm, determines the first clicking rate and preset FM mould
Whether the error amount of type output third clicking rate is minimum；Third clicking rate are as follows: the input of preset FM model is Current ad
Characteristic and current-user data, the clicking rate of preset FM model output；
The first adjustment unit, if the error amount for the first clicking rate and third clicking rate is not minimum, adjustment is default
FM model parameters until the first clicking rate and third clicking rate error amount are minimum；
First determination unit, for the preset FM model after parameters will to be adjusted as the FM model trained.
Optionally, training module includes:
Second error unit, for the first clicking rate to be made comparisons with third clicking rate, calculate obtain the first clicking rate with
The error amount of third clicking rate；
First-loss unit, for using preset parameters in error amount and preset FM model as preset very big
The output of preset Maximum-likelihood estimation function is taken logarithm by the input of possibility predication function, obtains preset FM model output
Loss function；
Computing unit, for calculating loss function using gradient descent method, for the parameters of preset FM model
Gradient determines whether the loss function of preset FM model output is minimum；
Second adjustment unit, if not being minimum for loss function value, the parameters for adjusting preset FM model are straight
It is minimum to loss function value；
Second determination unit, for the preset FM model after parameters will to be adjusted as the FM model trained.
Optionally, the playing device of a kind of advertisement provided in an embodiment of the present invention further include:
Sample module, for obtaining training set using the history ad data of an advertisement as a training sample；Instruction
Practicing set includes: a plurality of training sample；Every training sample includes: historical use data, history advertisement characteristic data and user
Click the historic click-through rate of advertisement generation；
Object module, for input of each training sample as initial FM model in set will to be trained, by each training
Historic click-through rate is as initial FM model training target in sample；
Module is lost, under the gradient of the pseudo- gradient function for intending the setting of newton OWLQN algorithm using normal state finite memory
Direction is dropped, the parameters of the preset loss function value under the direction are calculated, determines the loss letter of preset FM model output
Whether number is minimum；
Loss adjustment module, if not being minimum for loss function value, the parameters for adjusting preset FM model are straight
It is minimum to loss function value；
Model determining module, for the initial FM model after parameters will to be adjusted as preset FM model.
Optionally, the playing device of a kind of advertisement provided in an embodiment of the present invention further include:
Prediction module, for the advertisement characteristic data of an advertisement to be selected and a user data to be combined, as one
Forecast sample；
Module to be selected, for obtaining preset advertisement data set to be selected；Preset advertisement data set to be selected includes: a plurality of pre-
Test sample sheet.
Optionally, the playing device of a kind of advertisement provided in an embodiment of the present invention further include:
Update module, for preset advertisement data set to be selected to be added using Current ad data as an advertising copy,
Update preset advertisement data set to be selected.
At the another aspect that the present invention is implemented, a kind of computer readable storage medium is additionally provided, it is described computer-readable
Instruction is stored in storage medium, when run on a computer, so that computer executes any of the above-described advertisement
Playback method.
At the another aspect that the present invention is implemented, the embodiment of the invention also provides a kind of, and the computer program comprising instruction is produced
Product, when it is when counting operation, so that computer executes the playback method of any of the above-described advertisement.
The playback method and device of advertisement provided in an embodiment of the present invention play the current of advertisement by obtaining current time
Ad data；The Current ad data include: current-user data, Current ad characteristic and the first clicking rate；It is described
First clicking rate is the clicking rate that user clicks that the advertisement that the current time plays generates；By Current ad characteristic and work as
Input of the preceding user data as preset FM model, the training objective that the first clicking rate is exported as preset FM model,
The preset FM model of training, using the preset FM model after the completion of training as the FM model trained；It will be preset to be selected wide
The each ad data to be selected accused in data set is input in the FM model trained, and the of each advertisement to be selected of output estimation
Two clicking rates；Ad data to be selected includes: user data and advertisement characteristic data；Select the second clicking rate highest to be selected wide
It accuses, is played out in the preset time after current time.The present embodiment is obtained using currently playing ad data training
Trained FM model, real-time are higher；Go out the clicking rate of each advertisement to be selected using the FM model pre-estimating trained, accuracy compared with
It is high；Then the selection highest advertisement to be selected of clicking rate plays out.The advertisement to be selected played and the required product advertising of user
More match.It is thus possible to improve playing the clicking rate of advertisement.Certainly, it implements any of the products of the present invention or method must be different
It is fixed to need while reaching all the above advantage.
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 a kind of flow chart of the playback method of advertisement provided in an embodiment of the present invention；
Fig. 2 is the flow chart provided in an embodiment of the present invention for obtaining preset FM model；
Fig. 3 is the flow chart provided in an embodiment of the present invention for obtaining preset advertisement data set to be selected；
Fig. 4 is a kind of flow chart for obtaining the FM model trained provided in an embodiment of the present invention；
Fig. 5 is another flow chart for obtaining the FM model trained provided in an embodiment of the present invention；
Fig. 6 is a kind of structure chart of the playing device of advertisement provided in an embodiment of the present invention；
Fig. 7 is the structure chart of a kind of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention is described.
Firstly, to facilitate understanding of the present embodiment of the invention, first introducing the use in the embodiment of the present invention hereafter herein
Term " the first clicking rate ", " the second clicking rate ", " third clicking rate ", " historic click-through rate ".Deng.
First clicking rate is the clicking rate that user clicks that the advertisement that current time plays generates；Second clicking rate is to have trained
FM mode input be each ad data to be selected, the clicking rate for each advertisement to be selected that the FM model trained exports.Third
Clicking rate are as follows: third clicking rate are as follows: the input of preset FM model is Current ad characteristic and current-user data, is preset
FM model output clicking rate.Here relational terms such as first and second and the like are used merely to " first clicks
Rate " and " the second clicking rate " " third clicking rate " distinguish, and are not necessarily to or imply " the first clicking rate " and " second
There are any actual relationship or orders between clicking rate " " third clicking rate ".Specifically " the first clicking rate " and " second
Whether clicking rate " " third clicking rate " has sequence, can go to limit according to the actual situation.
Historic click-through rate is the clicking rate that user clicks that history advertisement generates.
The FM model that the embodiment of the present invention has been trained using the Current ad data that current time plays；It will be pre-
If ad data to be selected concentrate each ad data to be selected be input in the FM model trained, output estimation it is each to
Select the second clicking rate of advertisement；The highest advertisement to be selected of the second clicking rate is selected, is carried out in the preset time after current time
It plays.The FM model that the present embodiment has been trained using currently playing ad data, real-time are higher；Using having instructed
Experienced FM model pre-estimating goes out the clicking rate of each advertisement to be selected, and accuracy is higher；Then the highest advertisement to be selected of clicking rate is selected
It plays out.The more matching of the advertisement to be selected and the required product advertising of user that play.It is thus possible to improve playing advertisement
Clicking rate.
It continues with and a kind of playback method of advertisement provided in an embodiment of the present invention is briefly described.
A kind of playback method of advertisement provided in an embodiment of the present invention is applied to electronic equipment, and further electronic equipment can
Think mobile phone, computer, server, intelligent mobile terminal equipment, wearable intelligent mobile terminal equipment etc.；It can also be applied to
The advertising company of real time bid advertisement position profit.It is not limited here, any that electronic equipment of the invention may be implemented, it belongs to
Protection scope of the present invention.
As shown in Figure 1, the embodiment of the invention provides a kind of playback methods of advertisement, comprising:
S101 obtains the Current ad data that current time plays advertisement；Current ad data include: active user's number
According to, Current ad characteristic and the first clicking rate；First clicking rate is the advertisement production that user clicks that the current time plays
Raw clicking rate；
Wherein, current-user data includes: age, city, gender, hobby, and Current ad characteristic includes: current wide
The text of announcement, the format of Current ad, the picture of Current ad, the playing duration of Current ad, Current ad broadcasting time,
The broadcasting flow of Current ad.Current ad data further include: environmental data, environmental data include: client model, the time,
Geographical location, client screen size, the network IP of client and client and internetwork connection mode；The value of first clicking rate is
100%.
Wherein, current time is the period artificially set according to industry experience, and current time is less than or equal to current
The playing duration of advertisement, current time can be adaptively adjusted according to variety classes advertisement playing duration.
Such as: if currently playing advertisement is mobile phone advertisement, mobile phone advertisement playing duration was usually from 10 seconds；It can artificially set
Setting current time is 10 seconds or 5 seconds；If currently playing advertisement is shampoo advertisement, shampoo advertisement playing duration is usually from
8 seconds, it was 8 seconds or 7 seconds that current time, which artificially can be set,.
S102, using Current ad characteristic and current-user data as the input of preset FM model, by described
Training objective of one clicking rate as the preset FM model output, the training preset FM model, after the completion of training
Preset FM model as the FM model trained；
In order to realize the efficiency of FM model that raising has been trained, above-mentioned S102 can using it is following it is at least one may
Implementation, obtain preset FM model:
In one possible implementation, preset FM model can improve taking human as directly providing and obtain preset FM
The rate of model.
In alternatively possible implementation, preset FM model uses following steps, and training obtains in advance:
S201 obtains training set using the history ad data of an advertisement as a training sample；Training set packet
It includes: a plurality of training sample；Every training sample includes: that historical use data, history advertisement characteristic data and user click this extensively
Accuse the historic click-through rate generated；
Wherein, historical use data includes: age, city, gender, hobby, and history advertisement characteristic data includes: that history is wide
The text of announcement, the format of history advertisement, the picture of history advertisement, the playing duration of history advertisement, history advertisement broadcasting time,
The broadcasting flow of history advertisement.History ad data further include: environmental data, environmental data include: client model, the time,
Geographical location, client screen size, the network IP of client and client and internetwork connection mode；The value of historic click-through rate is
100%.
S202, will be in each training sample as the input of initial FM model using each training sample in training set
Historic click-through rate is as initial FM model training target；
Wherein, initial FM model is the mathematical formulae for artificially meeting ad click rate according to the selection of the data of advertising sector,
Parameters in the mathematical formulae can be artificial preset, and it is initially own to be also possible to the formula.
S203 intends the gradient descent direction of the pseudo- gradient function of newton OWLQN algorithm setting, meter using normal state finite memory
Whether most the parameters for calculating the preset loss function value under the direction determine the loss function of preset FM model output
It is small；
Using OWLQN algorithm set pseudo- gradient function, can each training sample in same quadrant, guarantee each training
The directional derivative that sample obtains is minimum, so that it is determined that the descent direction of gradient；Calculate the preset loss function value under the direction
Parameters, determine whether the loss function of preset FM model output minimum, to carry out to preset FM model multiple
Optimization determines the smallest efficiency of loss function to improve.
S204 adjusts the parameters of preset FM model until loss function value if loss function value is not minimum
It is minimum；
In a kind of possible embodiment, if loss function value is not minimum, adjusted to gradient descent direction preset
The parameters of FM model are until loss function value minimum, can save the time of adjustment parameters, quickly to lose
Function reaches minimum.
S205, using the initial FM model after adjustment parameters as preset FM model.
Compared to the mode of preset FM model is artificially directly provided, present embodiment passes through the history advertisement number of advertisement
According to, the initial FM model after parameters will be adjusted as preset FM model, using the parameters of preset FM model as
The start-up parameter for the FM model trained, make its need not move through cold start-up the time can immediately reach and preset FM mould
The same effect of type saves the time for obtaining training pattern, so that the accuracy rate of the training pattern trained is higher.
In a kind of possible embodiment, had partially, periodically using the preset FM model of FTRL algorithm training in order to prevent
The start-up parameter for the FM model trained is updated using the parameters of preset FM, improves the accuracy rate of training pattern.
Each ad data to be selected that preset ad data to be selected is concentrated is input to the FM model trained by S103
In, the second clicking rate of each advertisement to be selected of output estimation；Ad data to be selected includes: user data and characteristic of advertisement number
According to；
In order to realize that raising obtains the efficiency of the second clicking rate, above-mentioned S103 can be using following at least one possible reality
Existing mode, obtains preset advertisement data set to be selected:
In one possible implementation, preset advertisement data set to be selected can be taking human as by all history on website
Ad data is as preset advertisement data set to be selected, to improve the rate for obtaining preset advertisement data set to be selected.
In alternatively possible implementation, in conjunction with the embodiment of Fig. 1, as shown in figure 3, being obtained using following steps
Preset advertisement data set to be selected:
The advertisement characteristic data of one advertisement to be selected and a user data are combined by S301, as a forecast sample；
Wherein, the advertisement characteristic data and user data of advertisement to be selected can be provided by investment in advertising quotient or businessman.
Such as: advertisement characteristic data are as follows: shampoo advertisement, playing duration 10 seconds, loop play was primary in 3 seconds, advertising pictures
Size 10K；User data are as follows: female, 28, by the user data be arranged in after advertisement characteristic data or before be combined, as
One forecast sample, the forecast sample are as follows: shampoo advertisement, playing duration 10 seconds, loop play is primary within 3 seconds, and advertising pictures are big
Small 10K, female, 28.Or the forecast sample are as follows:, female, 28, shampoo advertisement, playing duration 10 seconds, loop play is primary within 3 seconds,
Advertising pictures size 10K.
S302 obtains preset advertisement data set to be selected；Preset advertisement data set to be selected includes: a plurality of forecast sample.
Compared to artificially using all history ad datas on website as the embodiment party of preset advertisement data set to be selected
Formula, present embodiment in view of taking in and user demand website because history ad data often with user demand not
Match, user's impression and website income can be improved in present embodiment.
S104 selects the highest advertisement to be selected of the second clicking rate, plays out in the preset time after current time.
The FM model that the embodiment of the present invention has been trained using currently playing ad data, real-time are higher；
Go out the clicking rate of each advertisement to be selected using the FM model pre-estimating trained, accuracy is higher；Then selection clicking rate is highest
Advertisement to be selected plays out.The more matching of the advertisement to be selected and the required product advertising of user that play.It is thus possible to improve broadcasting
Put the clicking rate of advertisement.
In order to realize the accuracy rate for improving the FM model trained, above-mentioned S102 can be using following at least one possible
Implementation, the FM model trained:
In a kind of possible embodiment, in conjunction with the embodiment of Fig. 1 and Fig. 2, as shown in figure 4, being obtained using following steps
Obtain the FM model trained:
S401, using Current ad characteristic and current-user data as the input of preset FM model, by first point
It hits rate to make comparisons with third clicking rate, using gradient descent algorithm, determines that the first clicking rate and preset FM model export third
Whether the error amount of clicking rate is minimum；Third clicking rate are as follows: the input of preset FM model be Current ad characteristic and
Current-user data, the clicking rate of preset FM model output；
In a kind of possible embodiment, Momentum (momentum) algorithm, Nesterov (newton momentum) can be used
Speed gradient algorithm, Adagrad (Ah Da Gray) algorithm, Adadelta (Ah 's Delta) algorithm, (root mean square is anti-by RMSprop
To propagation) algorithm, Adam (estimation of adaptability momentum) algorithm, determine that the first clicking rate and preset FM model output third are clicked
Whether the error amount of rate is minimum, the speed of raising gradient decline, so as to the FM model quickly trained.
S402 adjusts each of preset FM model if the error amount of the first clicking rate and third clicking rate is not minimum
Item parameter is until the first clicking rate and third clicking rate error amount are minimum；
Wherein, the first clicking rate and third clicking rate error amount can be minimum when 0, be also possible to the decimal no more than 1
When, reach minimum, herein with no restrictions, concrete condition is according to adjustable error of value of the first clicking rate and third clicking rate
Value reaches the measurement standard of minimum value.
S403, using the preset FM model after adjustment parameters as the FM model trained.
Present embodiment by adjusting preset FM model parameters so that the first clicking rate and third clicking rate
Error amount is minimum, using the preset FM model after adjustment parameters as the FM model trained, can be improved and has been instructed
Practice the efficiency of FM model.
In alternatively possible embodiment, in conjunction with the embodiment of Fig. 1 and Fig. 2, as shown in figure 5, using following steps,
Obtain the FM model trained:
S501, using Current ad characteristic and current-user data as the input of preset FM model, by first point
It hits rate to make comparisons with third clicking rate, calculates the error amount for obtaining the first clicking rate and third clicking rate；
S502, using preset parameters in error amount and preset FM model as preset Maximum-likelihood estimation function
Input, the output of preset Maximum-likelihood estimation function is taken into logarithm, obtains the loss function of preset FM model output；
In a kind of possible embodiment, using preset parameters in error amount and preset FM model as default
Maximum-likelihood estimation function input, the output of preset Maximum-likelihood estimation function is taken into logarithm, it is after logarithm, this is defeated
The loss function that preset Maximum-likelihood estimation function after taking logarithm out is exported as preset FM model is obtained in advance with improving
If the accuracy rate of FM model.
S503 is calculated using FTRL (Follow-the-regularized-Leader follows regular leader) algorithm and is damaged
Function is lost, for the gradient of the parameters of preset FM model, whether most to determine the loss function of preset FM model output
It is small；
In a kind of possible embodiment, using the formula of FTRL algorithm, loss function is iterated to calculate for preset
The parameters of FM model, it is cumulative no more than the gradient of difference threshold and, utilize the gradient it is cumulative and minimum when it is default
FM parameters, solve preset FM output third clicking rate and the first clicking rate error whether be it is minimum, improve
The accuracy for the FM model trained.
S504 adjusts the parameters of preset FM model until loss function value if loss function value is not minimum
It is minimum；
In a kind of possible embodiment, if loss function value is not minimum, adjusted to gradient descent direction preset
The parameters of FM model are until loss function value minimum, can save the time of adjustment parameters, quickly to lose
Function reaches minimum.
S505, using the preset FM model after adjustment parameters as the FM model trained.
Present embodiment adjusts the parameters of preset FM model using FTRL algorithm, solves the first clicking rate and third
The loss function of clicking rate is minimum, determines the FM model trained.There is FTRL algorithm model parameter rarefaction, data need to only change
Generation one time the advantages that, guarantees Current ad characteristic and active user in the preset FM model process of training using FTRL algorithm
The sparse features of data are improved and have been trained while the preset FM model of raising training has been trained the efficiency of FM model
The standard of FM model goes rate.
The clicking rate of advertisement in order to better improve, after S102, a kind of optional reality provided in an embodiment of the present invention
Existing mode, the method also includes:
Using Current ad data as an advertising copy, preset advertisement data set to be selected is added, update it is preset to
Select advertisement data set.
Present embodiment will be added current wide by the way that preset advertisement data set to be selected is added in Current ad characteristic
It is preset to be selected to reach update as preset advertisement data set to be selected for preset advertisement data set to be selected after accusing characteristic
The purpose of advertisement data set, to improve the accuracy rate of the second clicking rate.
It continues with and a kind of playing device of advertisement provided in an embodiment of the present invention is briefly described.
As shown in fig. 6, a kind of playing device of advertisement provided in an embodiment of the present invention, comprising:
Module 601 is obtained, the Current ad data of advertisement are played for obtaining current time；Current ad data include:
Current-user data, Current ad characteristic and the first clicking rate；First clicking rate is that user clicks what current time played
The clicking rate that advertisement generates；
Training module 602, for using Current ad characteristic and current-user data as the defeated of preset FM model
Enter, the training objective that the first clicking rate is exported as preset FM model, the preset FM model of training, after the completion of training
Preset FM model is as the FM model trained；
Module 603 is estimated, each ad data to be selected for concentrating preset ad data to be selected, which is input to, has instructed
In experienced FM model, the second clicking rate of each advertisement to be selected of output estimation；Ad data to be selected includes: user data and wide
Accuse characteristic；
Playing module 604, for the selection highest advertisement to be selected of the second clicking rate, in the preset time after current time
It plays out.
Optionally, training module includes:
First error unit, for using Current ad characteristic and current-user data as the defeated of preset FM model
Enter, the first clicking rate is made comparisons with third clicking rate, using gradient descent algorithm, determines the first clicking rate and preset FM mould
Whether the error amount of type output third clicking rate is minimum；Third clicking rate are as follows: the input of preset FM model is Current ad
Characteristic and current-user data, the clicking rate of preset FM model output；
The first adjustment unit, if the error amount for the first clicking rate and third clicking rate is not minimum, adjustment is default
FM model parameters until the first clicking rate and third clicking rate error amount are minimum；
First determination unit, for the preset FM model after parameters will to be adjusted as the FM model trained.
Optionally, training module includes:
Second error unit, for the first clicking rate to be made comparisons with third clicking rate, calculate obtain the first clicking rate with
The error amount of third clicking rate；
First-loss unit, for using preset parameters in error amount and preset FM model as preset very big
The output of preset Maximum-likelihood estimation function is taken logarithm by the input of possibility predication function, obtains preset FM model output
Loss function；
Computing unit, for calculating loss function using gradient descent method, for the parameters of preset FM model
Gradient determines whether the loss function of preset FM model output is minimum；
Second adjustment unit, if not being minimum for loss function value, the parameters for adjusting preset FM model are straight
It is minimum to loss function value；
Second determination unit, for the preset FM model after parameters will to be adjusted as the FM model trained.
Optionally, the playing device of a kind of advertisement provided in an embodiment of the present invention further include:
Sample module, for obtaining training set using the history ad data of an advertisement as a training sample；Instruction
Practicing set includes: a plurality of training sample；Every training sample includes: historical use data, history advertisement characteristic data and user
Click the historic click-through rate of advertisement generation；
Object module, for input of each training sample as initial FM model in set will to be trained, by each training
Historic click-through rate is as initial FM model training target in sample；
Module is lost, under the gradient of the pseudo- gradient function for intending the setting of newton OWLQN algorithm using normal state finite memory
Direction is dropped, the parameters of the preset loss function value under the direction are calculated, determines the loss letter of preset FM model output
Whether number is minimum；
Loss adjustment module, if not being minimum for loss function value, the parameters for adjusting preset FM model are straight
It is minimum to loss function value；
Model determining module, for the initial FM model after parameters will to be adjusted as preset FM model.
Optionally, the playing device of a kind of advertisement provided in an embodiment of the present invention further include:
Prediction module, for the advertisement characteristic data of an advertisement to be selected and a user data to be combined, as one
Forecast sample；
Module to be selected, for obtaining preset advertisement data set to be selected；Preset advertisement data set to be selected includes: a plurality of pre-
Test sample sheet.
Optionally, the playing device of a kind of advertisement provided in an embodiment of the present invention further include:
Update module, for preset advertisement data set to be selected to be added using Current ad data as an advertising copy,
Update preset advertisement data set to be selected.
The embodiment of the invention also provides a kind of electronic equipment, as shown in fig. 7, comprises processor 701, communication interface 702,
Memory 703 and communication bus 704, wherein processor 701, communication interface 702, memory 703 are complete by communication bus 704
At mutual communication,
Memory 703, for storing computer program；
Processor 701 when for executing the program stored on memory 703, realizes following steps:
Obtain the Current ad data that current time plays advertisement；Current ad data include: current-user data, current
Advertisement characteristic data and the first clicking rate；First clicking rate clicks the click that the advertisement that institute's current time plays generates for user
Rate；
Using Current ad characteristic and current-user data as the input of preset FM model, the first clicking rate is made
For the training objective of preset FM model output, the preset FM model of training, using the preset FM model after the completion of training as
The FM model trained；
Each ad data to be selected that preset ad data to be selected is concentrated is input in the FM model trained, is exported
Second clicking rate of each advertisement to be selected estimated；Ad data to be selected includes: user data and advertisement characteristic data；
The highest advertisement to be selected of the second clicking rate is selected, is played out in the preset time after current time.
The communication bus that above-mentioned electronic equipment is mentioned can be Peripheral Component Interconnect standard (Peripheral Component
Interconnect, abbreviation PCI) bus or expanding the industrial standard structure (Extended Industry Standard
Architecture, abbreviation EISA) bus etc..The communication bus can be divided into address bus, data/address bus, control bus etc..
Only to be indicated with a thick line in figure, it is not intended that an only bus or a type of bus convenient for indicating.
Communication interface is for the communication between above-mentioned electronic equipment and other equipment.
Memory may include random access memory (Random Access Memory, abbreviation RAM), also may include
Nonvolatile memory (non-volatile memory), for example, at least a magnetic disk storage.Optionally, memory may be used also
To be storage device that at least one is located remotely from aforementioned processor.
Above-mentioned processor can be general processor, including central processing unit (Central Processing Unit,
Abbreviation CPU), network processing unit (Network Processor, abbreviation NP) etc.；It can also be digital signal processor
(Digital Signal Processing, abbreviation DSP), specific integrated circuit (Application Specific
Integrated Circuit, abbreviation ASIC), field programmable gate array (Field-Programmable Gate Array,
Abbreviation FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware components.
In another embodiment provided by the invention, a kind of computer readable storage medium is additionally provided, which can
It reads to be stored with instruction in storage medium, when run on a computer, so that computer executes any institute in above-described embodiment
The playback method for the advertisement stated.
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 playback method of any advertisement 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, number from a web-site, computer, server or data center
User's line (DSL)) or wireless (such as infrared, wireless, microwave etc.) mode to another web-site, computer, server or
Data center is transmitted.The computer readable storage medium can be any usable medium that computer can access or
It is comprising data storage devices such as one or more usable mediums integrated server, data centers.The usable medium can be with
It is magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or semiconductor medium (such as solid state hard disk
Solid State Disk (SSD)) etc..
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
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 reality
For applying example, since it is substantially similar to the method embodiment, so being described relatively simple, related place is referring to embodiment of the method
Part explanation.
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 (13)
1. a kind of playback method of advertisement, which is characterized in that the described method includes:
Obtain the Current ad data that current time plays advertisement；The Current ad data include: current-user data, current
Advertisement characteristic data and the first clicking rate；First clicking rate is the advertisement generation that user clicks that the current time plays
Clicking rate；
Using the Current ad characteristic and the current-user data as the input of preset FM model, by described first
Training objective of the clicking rate as the preset FM model output, the training preset FM model, after the completion of training
Preset FM model is as the FM model trained；
Each ad data to be selected that preset ad data to be selected is concentrated is input in the FM model trained, output
Second clicking rate of each advertisement to be selected estimated；The ad data to be selected includes: user data and advertisement characteristic data；
The highest advertisement to be selected of the second clicking rate is selected, is played out in the preset time after current time.
2. the method according to claim 1, wherein described by the Current ad characteristic and described current
Input of the user data as the preset FM model, using first clicking rate as the preset FM model output
Training objective, the training preset FM model, using the preset FM model after the completion of training as the FM model trained, packet
It includes:
Using Current ad characteristic and current-user data as the input of preset FM model, by first clicking rate with
Third clicking rate is made comparisons, and using gradient descent algorithm, determines the first clicking rate and the preset FM model output thirdly
Whether the error amount for hitting rate is minimum；The third clicking rate are as follows: the input of preset FM model is the Current ad feature
Data and current-user data, the clicking rate of preset FM model output；
If the error amount of first clicking rate and the third clicking rate is not minimum, the items of preset FM model are adjusted
Parameter is until first clicking rate and third clicking rate error amount are minimum；
Using the preset FM model after adjustment parameters as the FM model trained.
3. according to the method described in claim 2, it is characterized in that, described by the Current ad characteristic and described current
Input of the user data as the preset FM model, using first clicking rate as the preset FM model output
Training objective, the training preset FM model, using the preset FM model after the completion of training as the FM model trained, packet
It includes:
Using Current ad characteristic and current-user data as the input of preset FM model, by first clicking rate with
The third clicking rate is made comparisons, and the error amount for obtaining first clicking rate and third clicking rate is calculated；
Using preset parameters in the error amount and the preset FM model as preset Maximum-likelihood estimation function
Input, the output of preset Maximum-likelihood estimation function is taken into logarithm, obtains the loss letter of the preset FM model output
Number；
Using following regular leader's FTRL algorithm to calculate loss function, for the ladder of the parameters of the preset FM model
Degree determines whether the loss function of the preset FM model output is minimum；
If the loss function value be not it is minimum, adjust the parameters of the preset FM model until loss function value most
It is small；
Using the preset FM model after adjustment parameters as the FM model trained.
4. the method according to claim 1, wherein the preset FM model is to instruct in advance as follows
Practice and obtain:
Using the history ad data of an advertisement as a training sample, training set is obtained；The training set includes: more
Training sample；Every training sample includes: that historical use data, history advertisement characteristic data and user click the advertisement production
Raw historic click-through rate；
Using in training set, each training sample is as the input of initial FM model, by the historic click-through rate in each training sample
As the initial FM model training target；
The gradient descent direction for being intended the pseudo- gradient function of newton OWLQN algorithm setting using normal state finite memory, is calculated in the party
The parameters of the downward preset loss function value determine whether the loss function of preset FM model output is minimum；
If loss function value is not minimum, the parameters of preset FM model are adjusted until loss function value is minimum；
Using the initial FM model after adjustment parameters as preset FM model.
5. the method according to claim 1, wherein obtaining preset ad data to be selected using following steps
Collection:
The advertisement characteristic data of one advertisement to be selected and a user data are combined, as a forecast sample；
Obtain preset advertisement data set to be selected；The preset advertisement data set to be selected includes: a plurality of forecast sample.
6. the method according to claim 1, wherein the method also includes:
Using Current ad data as an advertising copy, the preset advertisement data set to be selected is added, updates described default
Advertisement data set to be selected.
7. a kind of playing device of advertisement, which is characterized in that described device includes:
Module is obtained, the Current ad data of advertisement are played for obtaining current time；The Current ad data include: current
User data, Current ad characteristic and the first clicking rate；First clicking rate is that user's click current time is broadcast
The clicking rate that the advertisement put generates；
Training module, for using the Current ad characteristic and the current-user data as the defeated of preset FM model
Enter, using first clicking rate as the training objective of the preset FM model output, the training preset FM model will
Preset FM model after the completion of training is as the FM model trained；
Module is estimated, each ad data to be selected for concentrating preset ad data to be selected is input to described trained
In FM model, the second clicking rate of each advertisement to be selected of output estimation；The ad data to be selected includes: user data and wide
Accuse characteristic；
Playing module is broadcast in the preset time after current time for selecting the highest advertisement to be selected of the second clicking rate
It puts.
8. device according to claim 7, which is characterized in that the training module includes:
First error unit, for using Current ad characteristic and current-user data as the input of preset FM model,
First clicking rate is made comparisons with third clicking rate, using gradient descent algorithm, determines that the first clicking rate is preset with described
The error amount of FM model output third clicking rate whether be minimum；The third clicking rate are as follows: the input of preset FM model
For the Current ad characteristic and current-user data, the clicking rate of preset FM model output；
The first adjustment unit adjusts if the error amount for first clicking rate and the third clicking rate is not minimum
The parameters of preset FM model are until first clicking rate and third clicking rate error amount are minimum；
First determination unit, for the preset FM model after parameters will to be adjusted as the FM model trained.
9. device according to claim 8, which is characterized in that the training module includes:
Second error unit, for using Current ad characteristic and current-user data as the input of preset FM model,
First clicking rate is made comparisons with the third clicking rate, calculates the mistake for obtaining first clicking rate and third clicking rate
Difference；
First-loss unit, for using preset parameters in the error amount and the preset FM model as preset
The output of preset Maximum-likelihood estimation function is taken logarithm, obtains the preset FM by the input of Maximum-likelihood estimation function
The loss function of model output；
Computing unit, for calculating loss function using gradient descent method, for the parameters of the preset FM model
Gradient determines whether the loss function of the preset FM model output is minimum；
Second adjustment unit adjusts every ginseng of the preset FM model if not being minimum for the loss function value
Number is until loss function value is minimum；
Second determination unit, for the preset FM model after parameters will to be adjusted as the FM model trained.
10. device according to claim 7, which is characterized in that described device further include:
Sample module, for obtaining training set using the history ad data of an advertisement as a training sample；The instruction
Practicing set includes: a plurality of training sample；Every training sample includes: historical use data, history advertisement characteristic data and user
Click the historic click-through rate that the advertisement generates；
Object module, for input of each training sample as initial FM model in set will to be trained, by each training sample
Middle historic click-through rate is as the initial FM model training target；
Module is lost, the gradient decline side of the pseudo- gradient function for intending the setting of newton OWLQN algorithm using normal state finite memory
To the parameters of calculating preset loss function value under the direction determine the loss letter of preset FM model output
Whether number is minimum；
Loss adjustment module adjusts the parameters of preset FM model until damage if not being minimum for loss function value
It is minimum to lose functional value；
Model determining module, for the initial FM model after parameters will to be adjusted as preset FM model.
11. device according to claim 7, which is characterized in that described device further include:
Prediction module is predicted for being combined the advertisement characteristic data of an advertisement to be selected and a user data as one
Sample；
Module to be selected, for obtaining preset advertisement data set to be selected；The preset advertisement data set to be selected includes: a plurality of pre-
Test sample sheet.
12. device according to claim 7, which is characterized in that described device further include:
Update module, for the preset advertisement data set to be selected to be added using Current ad data as an advertising copy,
Update the preset advertisement data set to be selected.
13. 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 communication bus；
Memory, for storing computer program；
Processor when for executing the program stored on memory, realizes any method and step of claim 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810762883.8A CN108965951B (en) | 2018-07-12 | 2018-07-12 | Advertisement playing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810762883.8A CN108965951B (en) | 2018-07-12 | 2018-07-12 | Advertisement playing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108965951A true CN108965951A (en) | 2018-12-07 |
| CN108965951B CN108965951B (en) | 2020-10-02 |
Family
ID=64482821
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810762883.8A Active CN108965951B (en) | 2018-07-12 | 2018-07-12 | Advertisement playing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108965951B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825966A (en) * | 2019-10-31 | 2020-02-21 | 广州市百果园信息技术有限公司 | Information recommendation method and device, recommendation server and storage medium |
| CN111310988A (en) * | 2020-01-23 | 2020-06-19 | 湖南快乐阳光互动娱乐传媒有限公司 | Advertisement flow prediction method and device |
| CN112270571A (en) * | 2020-11-03 | 2021-01-26 | 中国科学院计算技术研究所 | Meta-model training method for cold-start advertisement click rate estimation model |
| CN114612167A (en) * | 2022-05-12 | 2022-06-10 | 杭州桃红网络有限公司 | Method for establishing automatic advertisement shutdown model and automatic advertisement shutdown model |
| CN117495459A (en) * | 2024-01-02 | 2024-02-02 | 蓝色火焰科技成都有限公司 | Man-machine interaction advertisement method, device, equipment and storage medium based on big data |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110231256A1 (en) * | 2009-07-25 | 2011-09-22 | Kindsight, Inc. | Automated building of a model for behavioral targeting |
| US20110264507A1 (en) * | 2010-04-27 | 2011-10-27 | Microsoft Corporation | Facilitating keyword extraction for advertisement selection |
| CN102346899A (en) * | 2011-10-08 | 2012-02-08 | 亿赞普(北京)科技有限公司 | Method and device for predicting advertisement click rate based on user behaviors |
| US20120136722A1 (en) * | 2010-11-30 | 2012-05-31 | Divy Kothiwal | Using Clicked Slate Driven Click-Through Rate Estimates in Sponsored Search |
| CN102663617A (en) * | 2012-03-20 | 2012-09-12 | 亿赞普(北京)科技有限公司 | Method and system for prediction of advertisement clicking rate |
| CN104536983A (en) * | 2014-12-08 | 2015-04-22 | 北京掌阔技术有限公司 | Method and device for predicting advertisement click rate |
| CN105046515A (en) * | 2015-06-26 | 2015-11-11 | 深圳市腾讯计算机系统有限公司 | Advertisement ordering method and device |
| CN105183772A (en) * | 2015-08-07 | 2015-12-23 | 百度在线网络技术（北京）有限公司 | Release information click rate estimation method and apparatus |
| CN105631711A (en) * | 2015-12-30 | 2016-06-01 | 合一网络技术(北京)有限公司 | Advertisement putting method and apparatus |
| CN106897892A (en) * | 2015-12-18 | 2017-06-27 | 北京奇虎科技有限公司 | Advertisement placement method and device |
-
2018
- 2018-07-12 CN CN201810762883.8A patent/CN108965951B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110231256A1 (en) * | 2009-07-25 | 2011-09-22 | Kindsight, Inc. | Automated building of a model for behavioral targeting |
| US20110264507A1 (en) * | 2010-04-27 | 2011-10-27 | Microsoft Corporation | Facilitating keyword extraction for advertisement selection |
| US20120136722A1 (en) * | 2010-11-30 | 2012-05-31 | Divy Kothiwal | Using Clicked Slate Driven Click-Through Rate Estimates in Sponsored Search |
| CN102346899A (en) * | 2011-10-08 | 2012-02-08 | 亿赞普(北京)科技有限公司 | Method and device for predicting advertisement click rate based on user behaviors |
| CN102663617A (en) * | 2012-03-20 | 2012-09-12 | 亿赞普(北京)科技有限公司 | Method and system for prediction of advertisement clicking rate |
| CN104536983A (en) * | 2014-12-08 | 2015-04-22 | 北京掌阔技术有限公司 | Method and device for predicting advertisement click rate |
| CN105046515A (en) * | 2015-06-26 | 2015-11-11 | 深圳市腾讯计算机系统有限公司 | Advertisement ordering method and device |
| CN105183772A (en) * | 2015-08-07 | 2015-12-23 | 百度在线网络技术（北京）有限公司 | Release information click rate estimation method and apparatus |
| CN106897892A (en) * | 2015-12-18 | 2017-06-27 | 北京奇虎科技有限公司 | Advertisement placement method and device |
| CN105631711A (en) * | 2015-12-30 | 2016-06-01 | 合一网络技术(北京)有限公司 | Advertisement putting method and apparatus |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825966A (en) * | 2019-10-31 | 2020-02-21 | 广州市百果园信息技术有限公司 | Information recommendation method and device, recommendation server and storage medium |
| WO2021082500A1 (en) * | 2019-10-31 | 2021-05-06 | 百果园技术(新加坡)有限公司 | Information recommendation method, apparatus, recommendation server and storage medium |
| CN111310988A (en) * | 2020-01-23 | 2020-06-19 | 湖南快乐阳光互动娱乐传媒有限公司 | Advertisement flow prediction method and device |
| CN111310988B (en) * | 2020-01-23 | 2023-04-07 | 湖南快乐阳光互动娱乐传媒有限公司 | Advertisement flow prediction method and device |
| CN112270571A (en) * | 2020-11-03 | 2021-01-26 | 中国科学院计算技术研究所 | Meta-model training method for cold-start advertisement click rate estimation model |
| CN112270571B (en) * | 2020-11-03 | 2023-06-27 | 中国科学院计算技术研究所 | Meta-model training method for cold-start advertisement click rate estimation model |
| CN114612167A (en) * | 2022-05-12 | 2022-06-10 | 杭州桃红网络有限公司 | Method for establishing automatic advertisement shutdown model and automatic advertisement shutdown model |
| CN114612167B (en) * | 2022-05-12 | 2022-08-19 | 杭州桃红网络有限公司 | Method for establishing automatic advertisement shutdown model and automatic advertisement shutdown model |
| CN117495459A (en) * | 2024-01-02 | 2024-02-02 | 蓝色火焰科技成都有限公司 | Man-machine interaction advertisement method, device, equipment and storage medium based on big data |
| CN117495459B (en) * | 2024-01-02 | 2024-03-19 | 蓝色火焰科技成都有限公司 | Man-machine interaction advertisement method, device, equipment and storage medium based on big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108965951B (en) | 2020-10-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108965951A (en) | The playback method and device of advertisement | |
| JP6707136B2 (en) | Distribution method, device and system for promotional information | |
| WO2015158149A1 (en) | Method and device for forecasting changes of feature information | |
| US10846613B2 (en) | System and method for measuring and predicting content dissemination in social networks | |
| CN110830812A (en) | Similar anchor classification model training method, anchor recommendation method and related device | |
| CN107341176B (en) | Sample weight setting method and device and electronic equipment | |
| TW201543238A (en) | Method and system for evaluating user satisfaction with respect to a user session | |
| CN104348892B (en) | A kind of information displaying method and device | |
| CN108460082A (en) | A kind of recommendation method and device, electronic equipment | |
| CN105590240A (en) | Discrete calculating method of brand advertisement effect optimization | |
| CN110362242A (en) | A kind of map methods of exhibiting and terminal device | |
| JP6946082B2 (en) | Advertisement distribution support device, advertisement distribution support method, and program | |
| US8234170B2 (en) | Online search advertising auction bid determination tools and techniques | |
| US20170330243A1 (en) | Bid recommendations for budget constrained campaigns | |
| CN109348260B (en) | Live broadcast room recommendation method, device, equipment and medium | |
| CN109690534A (en) | It is prefetched using dynamic user model to reduce the waiting time | |
| WO2015096742A1 (en) | Information processing method, device and system | |
| CN111369324B (en) | Target information determining method, device, equipment and readable storage medium | |
| CN113158550A (en) | Method and device for federated learning, electronic equipment and storage medium | |
| JP2011192224A (en) | Numerical calculation device, method, and program | |
| CN111353597B (en) | Target detection neural network training method and device | |
| CN113793181B (en) | Bid method, system and device for advertisement space | |
| JP6956564B2 (en) | Information processing equipment, information processing methods, and programs | |
| CN107248094B (en) | Electronic product activation amount prediction method and server cluster | |
| JP5805321B2 (en) | Determining the physical location of the client |
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 |