CN108965951B - Advertisement playing method and device - Google Patents
Advertisement playing method and device Download PDFInfo
- Publication number
- CN108965951B CN108965951B CN201810762883.8A CN201810762883A CN108965951B CN 108965951 B CN108965951 B CN 108965951B CN 201810762883 A CN201810762883 A CN 201810762883A CN 108965951 B CN108965951 B CN 108965951B
- Authority
- CN
- China
- Prior art keywords
- preset
- model
- advertisement
- click 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.)
- Active
Links
Images
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)
- Business, Economics & Management (AREA)
- Marketing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
According to the method and the device for playing the advertisement, which are provided by the embodiment of the invention, the trained FM model is trained by acquiring the current advertisement data of the advertisement played at the current time; inputting each advertisement data to be selected in the advertisement data set to be selected into a trained FM model, and outputting a second click rate of each estimated advertisement to be selected; and selecting the advertisement to be selected with the highest second click rate, and playing the advertisement within the preset time after the current time. The embodiment utilizes the current advertisement data to train to obtain a trained FM model, and the real-time performance is high; the click rate of each advertisement to be selected is estimated by using the trained FM model, so that the accuracy is high; and then selecting the advertisement to be selected with the highest click rate for playing. The played advertisements to be selected are matched with the advertisements of the products required by the user. Therefore, the click rate of playing the advertisement can be improved.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a method and a device for playing advertisements.
Background
The network platform reasonably distributes the playing time of the advertisement, the playing times of the advertisement and the playing flow of the advertisement according to the click rate of the advertisement, formulates an advertisement fee charging standard and then plays the advertisement.
In the prior art, the playing process of the advertisement is as follows:
historical advertisement data in the database are used as advertisement samples, the advertisement samples are used as input of an FM (factor decomposition) algorithm containing preset parameters, and the advertisement click rate in the historical advertisement data is used as a training target and is trained and output as an FM model of the advertisement click rate. And taking each advertisement to be played as an advertisement to be selected, inputting the obtained data of each advertisement to be selected into the optimized FM model, outputting the estimated click rate of each advertisement to be selected, and selecting the advertisement to be selected with the highest click rate as the advertisement to be played on line.
Since the historical advertising data employed in the prior art is often out of date, the user's demand for the product often changes. Therefore, in the prior art, the FM model trained by using the historical advertisement data may have a low prediction time, so that the estimated click rate of the advertisement to be selected is often inaccurate, the product in the online advertisement may be far from the product required by the user, and the user may not click the online advertisement, thereby causing the click rate of the online advertisement to be low.
Disclosure of Invention
The embodiment of the invention aims to provide an advertisement playing method and device, which train a trained FM model in real time by using advertisement characteristic data played at the current time, improve the real-time performance of the trained FM model, and improve the accuracy of estimating the click rate of an advertisement to be selected, so that the click rate of online advertisement playing can be improved. The specific technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides a method for playing an advertisement, including:
acquiring current advertisement data of an advertisement played at the current time; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated when the user clicks the advertisement played at the current time;
taking the current advertisement characteristic data and the current user data as the input of a preset FM model, taking the first click rate as a training target output by the preset FM model, training the preset FM model, and taking the preset FM model after training as a trained FM model;
inputting each piece of advertisement data to be selected in a preset advertisement data set to be selected into a trained FM model, and outputting a second estimated click rate of each advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data;
and selecting the advertisement to be selected with the highest second click rate, and playing the advertisement within the preset time after the current time.
Optionally, the current advertisement characteristic data and the current user data are used as input of a preset FM model, the first click rate is used as a training target of the preset FM model output, the preset FM model is trained, and the preset FM model after the training is completed is used as the trained FM model, including:
taking the current advertisement characteristic data and the current user data as the input of a preset FM model, comparing the first click rate with the third click rate, and determining whether the error value of the first click rate and the third click rate output by the preset FM model is the minimum or not by utilizing a gradient descent algorithm; the third click rate is: inputting current advertisement characteristic data and current user data into a preset FM model, and outputting a click rate by the preset FM model;
if the error value of the first click rate and the third click rate is not the minimum, adjusting all the parameters of the preset FM model until the error value of the first click rate and the third click rate is the minimum;
and taking the preset FM model after adjusting each parameter as the trained FM model.
Optionally, the current advertisement characteristic data and the current user data are used as input of a preset FM model, the first click rate is used as a training target of the preset FM model output, the preset FM model is trained, and the preset FM model after the training is completed is used as the trained FM model, including:
taking the current advertisement characteristic data and the current user data as the input of a preset FM model, comparing the first click rate with the third click rate, and calculating to obtain an error value of the first click rate and the third click rate;
taking the error value and various preset parameters in a preset FM model as the input of a preset maximum likelihood estimation function, and taking the logarithm of the output of the preset maximum likelihood estimation function to obtain a loss function of the preset FM model output;
calculating a loss function by utilizing a leader FTRL algorithm following the normality, and determining whether the loss function output by the preset FM model is minimum or not for the gradient of each parameter of the preset FM model;
if the loss function value is not the minimum, adjusting all parameters of a preset FM model until the loss function value is the minimum;
and taking the preset FM model after adjusting each parameter as the trained FM model.
Optionally, the preset FM model is obtained by pre-training through the following steps:
taking historical advertisement data of an advertisement as a training sample to obtain a training set; the training set includes: a plurality of training samples; each training sample comprises: historical user data, historical advertisement characteristic data and historical click rate generated by clicking advertisements by users;
taking each training sample in the training set as the input of an initial FM model, and taking the historical click rate in each training sample as the training target of the initial FM model;
calculating each parameter of a loss function value preset in the gradient descending direction of a pseudo-gradient function set by a normal finite memory quasi-Newton OWLQN algorithm, and determining whether the loss function output by a preset FM model is minimum or not;
if the loss function value is not the minimum, adjusting all parameters of a preset FM model until the loss function value is the minimum;
and taking the initial FM model after each parameter is adjusted as a preset FM model.
Optionally, the following steps may be adopted to obtain a preset advertisement data set to be selected:
combining advertisement characteristic data of a selected advertisement and user data to serve as a prediction sample;
acquiring a preset advertisement data set to be selected; the preset advertisement data set to be selected comprises: a plurality of prediction samples.
Optionally, an embodiment of the present invention provides a method for playing an advertisement, which further includes:
and taking the current advertisement data as an advertisement sample, adding a preset advertisement data set to be selected, and updating the preset advertisement data set to be selected.
In a second aspect, an embodiment of the present invention provides an advertisement playing apparatus, including:
the acquisition module is used for acquiring current advertisement data of the advertisement played at the current time; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated when the user clicks the advertisement played at the current time;
the training module is used for taking the current advertisement characteristic data and the current user data as the input of a preset FM model, taking the first click rate as a training target output by the preset FM model, training the preset FM model, and taking the preset FM model after training as the trained FM model;
the estimation module is used for inputting each piece of advertisement data to be selected in the preset advertisement data set to be selected into the trained FM model and outputting the estimated second click rate of each piece of advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data;
and the playing module is used for selecting the advertisement to be selected with the highest second click rate and playing the advertisement within a preset time after the current time.
Optionally, the training module includes:
the first error unit is used for taking the current advertisement characteristic data and the current user data as the input of a preset FM model, comparing the first click rate with the third click rate, and determining whether the error value of the first click rate and the third click rate output by the preset FM model is the minimum or not by utilizing a gradient descent algorithm; the third click rate is: inputting current advertisement characteristic data and current user data into a preset FM model, and outputting a click rate by the preset FM model;
the first adjusting unit is used for adjusting all the parameters of the preset FM model until the error value of the first click rate and the third click rate is the minimum if the error value of the first click rate and the third click rate is not the minimum;
and the first determining unit is used for taking the preset FM model after each parameter is adjusted as the trained FM model.
Optionally, the training module includes:
the second error unit is used for comparing the first click rate with the third click rate and calculating to obtain an error value of the first click rate and the third click rate;
the first loss unit is used for taking the error value and each preset parameter in the preset FM model as the input of a preset maximum likelihood estimation function, taking the logarithm of the output of the preset maximum likelihood estimation function and obtaining the loss function output by the preset FM model;
the calculation unit is used for calculating a loss function by using a gradient descent method, and determining whether the loss function output by the preset FM model is minimum or not for the gradient of each parameter of the preset FM model;
the second adjusting unit is used for adjusting each parameter of the preset FM model until the loss function value is minimum if the loss function value is not minimum;
and the second determining unit is used for taking the preset FM model after each parameter is adjusted as the trained FM model.
Optionally, the apparatus for playing an advertisement provided in the embodiment of the present invention further includes:
the sample module is used for taking historical advertisement data of an advertisement as a training sample to obtain a training set; the training set includes: a plurality of training samples; each training sample comprises: historical user data, historical advertisement characteristic data and historical click rate generated by clicking the advertisement by the user;
the target module is used for taking each training sample in the training set as the input of the initial FM model and taking the historical click rate in each training sample as the training target of the initial FM model;
the loss module is used for calculating parameters of a loss function value preset in the gradient descending direction by utilizing the gradient descending direction of a pseudo-gradient function set by a normal finite memory quasi-Newton OWLQN algorithm and determining whether the loss function output by the preset FM model is minimum or not;
the loss adjusting module is used for adjusting each parameter of the preset FM model until the loss function value is minimum if the loss function value is not minimum;
and the model determining module is used for taking the initial FM model after each parameter is adjusted as a preset FM model.
Optionally, the apparatus for playing an advertisement provided in the embodiment of the present invention further includes:
the prediction module is used for combining advertisement characteristic data of a selected advertisement and user data to serve as a prediction sample;
the system comprises a candidate module, a selection module and a selection module, wherein the candidate module is used for obtaining a preset candidate advertisement data set; the preset advertisement data set to be selected comprises: a plurality of prediction samples.
Optionally, the apparatus for playing an advertisement provided in the embodiment of the present invention further includes:
and the updating module is used for adding the preset advertisement data set to be selected into the current advertisement data serving as an advertisement sample, and updating the preset advertisement data set to be selected.
In yet another aspect of the present invention, there is also provided a computer-readable storage medium, having stored therein instructions, which when run on a computer, cause the computer to execute any one of the above-mentioned advertisement playing methods.
In another aspect of the present invention, there is also provided a computer program product including instructions, which when executed, cause a computer to execute any of the above advertisement playing methods.
According to the method and the device for playing the advertisement, the current advertisement data of the advertisement played at the current time is obtained; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated when the user clicks the advertisement played at the current time; taking the current advertisement characteristic data and the current user data as the input of a preset FM model, taking the first click rate as a training target output by the preset FM model, training the preset FM model, and taking the preset FM model after training as a trained FM model; inputting each piece of advertisement data to be selected in a preset advertisement data set to be selected into a trained FM model, and outputting a second estimated click rate of each advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data; and selecting the advertisement to be selected with the highest second click rate, and playing the advertisement within the preset time after the current time. In the embodiment, the trained FM model is obtained by training the currently played advertisement data, so that the real-time performance is higher; the click rate of each advertisement to be selected is estimated by using the trained FM model, so that the accuracy is high; and then selecting the advertisement to be selected with the highest click rate for playing. The played advertisements to be selected are matched with the advertisements of the products required by the user. Therefore, the click rate of playing the advertisement can be improved. Of course, it is not necessary for any product or method of practicing the invention to achieve all of the above-described advantages at the same time.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
Fig. 1 is a flowchart of a method for playing an advertisement according to an embodiment of the present invention;
FIG. 2 is a flowchart of obtaining a preset FM model according to an embodiment of the present invention;
fig. 3 is a flowchart of obtaining a preset advertisement data set to be selected according to an embodiment of the present invention;
FIG. 4 is a flowchart of obtaining a trained FM model according to an embodiment of the present invention;
FIG. 5 is a flowchart of another method for obtaining a trained FM model according to an embodiment of the present invention;
fig. 6 is a structural diagram of a device for playing an advertisement according to an embodiment of the present invention;
fig. 7 is a structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described below with reference to the drawings in the embodiments of the present invention.
First, for the convenience of understanding the embodiment of the present invention, the following terms "first click rate", "second click rate", "third click rate", "historical click rate" in the embodiment of the present invention will be described. And the like.
The first click rate is the click rate generated when the user clicks the advertisement played at the current time; and the second click rate is the click rate of each advertisement to be selected, which is input by the trained FM model and output by the trained FM model. The third click rate is: the third click rate is: the preset FM model inputs current advertisement characteristic data and current user data, and the click rate output by the preset FM model. Relational terms such as first and second, and the like, are used herein only to distinguish between "first click rate" and "second click rate", "third click rate", and do not necessarily require or imply any such actual relationship or order between "first click rate" and "second click rate", "third click rate". Whether the first click rate, the second click rate and the third click rate have an order or not can be defined according to actual conditions.
The historical click rate is the click rate generated when the user clicks the historical advertisement.
The embodiment of the invention uses the current advertising data played at the current time to train to obtain a trained FM model; inputting each piece of advertisement data to be selected in a preset advertisement data set to be selected into a trained FM model, and outputting a second estimated click rate of each advertisement to be selected; and selecting the advertisement to be selected with the highest second click rate, and playing the advertisement within the preset time after the current time. In the embodiment, the trained FM model is obtained by training the currently played advertisement data, so that the real-time performance is higher; the click rate of each advertisement to be selected is estimated by using the trained FM model, so that the accuracy is high; and then selecting the advertisement to be selected with the highest click rate for playing. The played advertisements to be selected are matched with the advertisements of the products required by the user. Therefore, the click rate of playing the advertisement can be improved.
The following provides a brief introduction to the advertisement playing method provided by the embodiment of the present invention.
The advertisement playing method provided by the embodiment of the invention is applied to electronic equipment, and further the electronic equipment can be a mobile phone, a computer, a server, intelligent mobile terminal equipment, wearable intelligent mobile terminal equipment and the like; but also to advertising companies bidding on ad spots for profit in real time. Without limitation, any electronic device that can implement the present invention is within the scope of the present invention.
As shown in fig. 1, an embodiment of the present invention provides a method for playing an advertisement, including:
s101, acquiring current advertisement data of an advertisement played at the current time; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated when the user clicks the advertisement played at the current time;
wherein the current user data comprises: age, city, gender, hobbies, current advertisement characteristic data includes: the method comprises the following steps of current advertisement characters, current advertisement formats, current advertisement pictures, current advertisement playing time, current advertisement playing times and current advertisement playing flow. The current advertisement data further includes: environmental data, the environmental data comprising: the client type, time, geographical position, client screen size, network IP of the client and a connection mode between the client and the network; the first click rate has a value of 100%.
The current time is a time period set manually according to industry experience, the current time is less than or equal to the playing time of the current advertisement, and the current time can be adaptively adjusted according to the playing time of different types of advertisements.
For example: if the current advertisement is the mobile phone advertisement, the playing time of the mobile phone advertisement is usually 10 seconds; the current time can be set to be 10 seconds or 5 seconds manually; if the current broadcast advertisement is a shampoo advertisement, the broadcast time of the shampoo advertisement is usually 8 seconds, and the current time can be set to be 8 seconds or 7 seconds by people.
S102, taking current advertisement characteristic data and current user data as input of a preset FM model, taking the first click rate as a training target output by the preset FM model, training the preset FM model, and taking the trained preset FM model as a trained FM model;
in order to improve the efficiency of obtaining the trained FM model, in step S102, the preset FM model may be obtained by using at least one of the following possible implementation manners:
in a possible implementation manner, the preset FM model may be directly provided by human, so as to increase the rate of obtaining the preset FM model.
In another possible implementation manner, the preset FM model is obtained by pre-training the following steps:
s201, taking historical advertisement data of an advertisement as a training sample to obtain a training set; the training set includes: a plurality of training samples; each training sample comprises: historical user data, historical advertisement characteristic data and historical click rate generated by clicking the advertisement by the user;
wherein the historical user data comprises: age, city, gender, hobbies, historical advertising characteristic data include: the method comprises the following steps of writing of the historical advertisement, format of the historical advertisement, picture of the historical advertisement, playing time of the historical advertisement, playing times of the historical advertisement and playing flow of the historical advertisement. The historical advertisement data further includes: environmental data, the environmental data comprising: the client type, time, geographical position, client screen size, network IP of the client and a connection mode between the client and the network; the value of the historical click rate is 100%.
S202, taking each training sample in the training set as input of an initial FM model, and taking a historical click rate in each training sample as a training target of the initial FM model;
the initial FM model is a mathematical formula which is artificially selected according with the advertisement click rate according to the data of the advertisement industry, and each parameter in the mathematical formula can be artificially preset or can be originally owned by the formula.
S203, calculating parameters of a loss function value preset in the gradient descending direction by utilizing the gradient descending direction of a pseudo-gradient function set by a normal finite memory quasi-Newton OWLQN algorithm, and determining whether the loss function output by a preset FM model is minimum or not;
by utilizing a pseudo gradient function set by an OWLQN algorithm, each training sample can be in the same quadrant, and the direction derivative obtained by each training sample is ensured to be minimum, so that the descending direction of the gradient is determined; and calculating each parameter of the loss function value preset in the direction, and determining whether the loss function output by the preset FM model is minimum or not, so that the preset FM model is optimized for multiple times to improve the efficiency of determining the minimum loss function.
S204, if the loss function value is not the minimum, adjusting each parameter of a preset FM model until the loss function value is the minimum;
in a possible implementation manner, if the loss function value is not the minimum, the parameters of the preset FM model are adjusted in the gradient descending direction until the loss function value is the minimum, so that the time for adjusting the parameters can be saved, and the loss function can be quickly minimized.
And S205, taking the initial FM model after each parameter is adjusted as a preset FM model.
Compared with a mode of directly providing the preset FM model, the method has the advantages that the initial FM model with various adjusted parameters is used as the preset FM model through historical advertisement data of advertisements, the various preset parameters of the FM model are used as the starting parameters for training the trained FM model, the effect same as that of the preset FM model can be achieved immediately without cold starting time, the time for obtaining the trained model is saved, and the accuracy of the trained model is higher.
In one possible implementation mode, in order to prevent the deviation of the preset FM model trained by using the FTRL algorithm, the starting parameters of the trained FM model are periodically updated by using various parameters of the preset FM, and the accuracy of the trained model is improved.
S103, inputting each piece of advertisement data to be selected in a preset advertisement data set to be selected into a trained FM model, and outputting an estimated second click rate of each advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data;
in order to improve the efficiency of obtaining the second click rate, in step S103, at least one possible implementation manner may be adopted to obtain a preset advertisement data set to be selected:
in a possible implementation manner, the preset candidate advertisement data set may artificially use all historical advertisement data on the website as the preset candidate advertisement data set, so as to increase a rate of obtaining the preset candidate advertisement data set.
In another possible implementation manner, in combination with the embodiment of fig. 1, as shown in fig. 3, the following steps are adopted to obtain a preset data set of advertisements to be selected:
s301, combining advertisement characteristic data of a selected advertisement and user data to serve as a prediction sample;
the advertisement characteristic data and the user data of the advertisement to be selected can be provided by an advertisement investor or a merchant.
For example: the advertisement characteristic data is as follows: shampoo advertisements are played for 10 seconds, the playing time is 3 seconds, the advertisement pictures are played once in a circulating mode, and the size of the advertisement pictures is 10K; the user data is: and the woman and the user 28 arrange the user data after or before the advertisement characteristic data to be combined to form a prediction sample, wherein the prediction sample is as follows: shampoo advertisement and playing time are 10 seconds, the shampoo advertisement and playing time is 3 seconds, and advertisement pictures are 10K, female and 28. Or the prediction sample is: the advertisement of women, 28 and shampoo is played for 10 seconds, the advertisement picture is played once in 3 seconds in a circulating mode, and the size of the advertisement picture is 10K.
S302, obtaining a preset advertisement data set to be selected; the preset advertisement data set to be selected comprises: a plurality of prediction samples.
Compared with an implementation mode of manually taking all historical advertisement data on a website as a preset advertisement data set to be selected, the implementation mode takes the website income and the user requirements into consideration, and the implementation mode can improve the user experience and the website income because the historical advertisement data are often not matched with the user requirements.
And S104, selecting the advertisement to be selected with the highest second click rate, and playing the advertisement within a preset time after the current time.
The embodiment of the invention utilizes the currently played advertisement data to train and obtain the trained FM model, and the real-time performance is higher; the click rate of each advertisement to be selected is estimated by using the trained FM model, so that the accuracy is high; and then selecting the advertisement to be selected with the highest click rate for playing. The played advertisements to be selected are matched with the advertisements of the products required by the user. Therefore, the click rate of playing the advertisement can be improved.
In order to improve the accuracy of the trained FM model, in step S102, the trained FM model may be obtained by training in at least one of the following possible implementation manners:
in one possible implementation, in combination with the embodiments of fig. 1 and fig. 2, as shown in fig. 4, the trained FM model is obtained by the following steps:
s401, taking current advertisement characteristic data and current user data as input of a preset FM model, comparing a first click rate with a third click rate, and determining whether an error value of the first click rate and the third click rate output by the preset FM model is minimum or not by using a gradient descent algorithm; the third click rate is: inputting current advertisement characteristic data and current user data into a preset FM model, and outputting a click rate by the preset FM model;
in one possible implementation, a Momentum (impulse) algorithm, a Nesterov (newton Momentum) acceleration gradient algorithm, an adagred (adagrel) algorithm, an adapelta (adadel ta) algorithm, an RMSprop (root mean square back propagation) algorithm, and an Adam (adaptive Momentum estimation) algorithm may be used to determine whether an error value between the first click rate and a third click rate output by the preset FM model is minimum, and increase the speed of gradient descent, so as to obtain the trained FM model quickly.
S402, if the error value of the first click rate and the third click rate is not the minimum, adjusting all parameters of a preset FM model until the error value of the first click rate and the third click rate is the minimum;
the error value of the first click rate and the third click rate may be the minimum when 0, or the minimum when a decimal number not greater than 1, and is not limited herein.
And S403, taking the preset FM model after each parameter is adjusted as the trained FM model.
In the embodiment, each parameter of the preset FM model is adjusted to minimize the error value of the first click rate and the third click rate, and the preset FM model after each parameter is adjusted is used as the trained FM model, so that the efficiency of obtaining the trained FM model can be improved.
In another possible implementation, in combination with the embodiments of fig. 1 and fig. 2, as shown in fig. 5, the trained FM model is obtained by the following steps:
s501, taking current advertisement characteristic data and current user data as input of a preset FM model, comparing a first click rate with a third click rate, and calculating to obtain an error value of the first click rate and the third click rate;
s502, taking the error value and each preset parameter in the preset FM model as the input of a preset maximum likelihood estimation function, and taking the logarithm of the output of the preset maximum likelihood estimation function to obtain a loss function of the preset FM model output;
in a possible implementation manner, the error value and each preset parameter in the preset FM model are used as inputs of a preset maximum likelihood estimation function, a logarithm is taken from an output of the preset maximum likelihood estimation function, and after the logarithm is taken, the output logarithmized preset maximum likelihood estimation function is used as a loss function of the preset FM model output, so as to improve the accuracy of obtaining the preset FM model.
S503, calculating a loss function by using an FTRL (Follow-the-normalized Leader) algorithm, and determining whether the loss function output by the preset FM model is minimum or not for the gradient of each parameter of the preset FM model;
in a possible implementation manner, the formula of the FTRL algorithm is utilized to iteratively calculate the sum of gradients of the loss function for each parameter of the preset FM model, the sum of gradients does not exceed the difference threshold, and whether the error between the third click rate and the first click rate of the preset FM output is the minimum or not is solved by utilizing each parameter of the preset FM when the sum of the gradients is the minimum, so that the accuracy of the trained FM model is improved.
S504, if the loss function value is not the minimum, adjusting each parameter of the preset FM model until the loss function value is the minimum;
in a possible implementation manner, if the loss function value is not the minimum, the parameters of the preset FM model are adjusted in the gradient descending direction until the loss function value is the minimum, so that the time for adjusting the parameters can be saved, and the loss function can be quickly minimized.
And S505, taking the preset FM model after each parameter is adjusted as the trained FM model.
In the embodiment, the FTRL algorithm is utilized to adjust various parameters of the preset FM model, the minimum loss function of the first click rate and the third click rate is solved, and the trained FM model is determined. The FTRL algorithm has the advantages that model parameters are sparse, data only need to be iterated once, and the like, the FTRL algorithm is used for guaranteeing sparse characteristics of current advertisement characteristic data and current user data in the process of training the preset FM model, the efficiency of training the preset FM model to obtain the trained FM model is improved, and the accurate elimination rate of the trained FM model is improved.
In order to better improve the click through rate of the advertisement, after S102, in an optional implementation manner provided by the embodiment of the present invention, the method further includes:
and taking the current advertisement data as an advertisement sample, adding a preset advertisement data set to be selected, and updating the preset advertisement data set to be selected.
In the embodiment, the current advertisement characteristic data is added into the preset advertisement data set to be selected, and the preset advertisement data set to be selected after the current advertisement characteristic data is added is used as the preset advertisement data set to be selected, so that the aim of updating the preset advertisement data set to be selected is fulfilled, and the accuracy of the second click rate is improved.
The following provides a brief description of an advertisement playing device according to an embodiment of the present invention.
As shown in fig. 6, an advertisement playing apparatus provided in an embodiment of the present invention includes:
an obtaining module 601, configured to obtain current advertisement data of an advertisement played at a current time; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated when the user clicks the advertisement played at the current time;
the training module 602 is configured to take the current advertisement feature data and the current user data as inputs of a preset FM model, take the first click rate as a training target of the preset FM model output, train the preset FM model, and take the preset FM model after training as a trained FM model;
the estimation module 603 is configured to input each piece of advertisement data to be selected in the preset advertisement data set to be selected into the trained FM model, and output an estimated second click rate of each piece of advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data;
the playing module 604 is configured to select the advertisement to be selected with the highest second click rate, and play the selected advertisement within a preset time after the current time.
Optionally, the training module includes:
the first error unit is used for taking the current advertisement characteristic data and the current user data as the input of a preset FM model, comparing the first click rate with the third click rate, and determining whether the error value of the first click rate and the third click rate output by the preset FM model is the minimum or not by utilizing a gradient descent algorithm; the third click rate is: inputting current advertisement characteristic data and current user data into a preset FM model, and outputting a click rate by the preset FM model;
the first adjusting unit is used for adjusting all the parameters of the preset FM model until the error value of the first click rate and the third click rate is the minimum if the error value of the first click rate and the third click rate is not the minimum;
and the first determining unit is used for taking the preset FM model after each parameter is adjusted as the trained FM model.
Optionally, the training module includes:
the second error unit is used for comparing the first click rate with the third click rate and calculating to obtain an error value of the first click rate and the third click rate;
the first loss unit is used for taking the error value and each preset parameter in the preset FM model as the input of a preset maximum likelihood estimation function, taking the logarithm of the output of the preset maximum likelihood estimation function and obtaining the loss function output by the preset FM model;
the calculation unit is used for calculating a loss function by using a gradient descent method, and determining whether the loss function output by the preset FM model is minimum or not for the gradient of each parameter of the preset FM model;
the second adjusting unit is used for adjusting each parameter of the preset FM model until the loss function value is minimum if the loss function value is not minimum;
and the second determining unit is used for taking the preset FM model after each parameter is adjusted as the trained FM model.
Optionally, the apparatus for playing an advertisement provided in the embodiment of the present invention further includes:
the sample module is used for taking historical advertisement data of an advertisement as a training sample to obtain a training set; the training set includes: a plurality of training samples; each training sample comprises: historical user data, historical advertisement characteristic data and historical click rate generated by clicking the advertisement by the user;
the target module is used for taking each training sample in the training set as the input of the initial FM model and taking the historical click rate in each training sample as the training target of the initial FM model;
the loss module is used for calculating parameters of a loss function value preset in the gradient descending direction by utilizing the gradient descending direction of a pseudo-gradient function set by a normal finite memory quasi-Newton OWLQN algorithm and determining whether the loss function output by the preset FM model is minimum or not;
the loss adjusting module is used for adjusting each parameter of the preset FM model until the loss function value is minimum if the loss function value is not minimum;
and the model determining module is used for taking the initial FM model after each parameter is adjusted as a preset FM model.
Optionally, the apparatus for playing an advertisement provided in the embodiment of the present invention further includes:
the prediction module is used for combining advertisement characteristic data of a selected advertisement and user data to serve as a prediction sample;
the system comprises a candidate module, a selection module and a selection module, wherein the candidate module is used for obtaining a preset candidate advertisement data set; the preset advertisement data set to be selected comprises: a plurality of prediction samples.
Optionally, the apparatus for playing an advertisement provided in the embodiment of the present invention further includes:
and the updating module is used for adding the preset advertisement data set to be selected into the current advertisement data serving as an advertisement sample, and updating the preset advertisement data set to be selected.
An embodiment of the present invention further provides an electronic device, as shown in fig. 7, including a processor 701, a communication interface 702, a memory 703 and a communication bus 704, where the processor 701, the communication interface 702, and the memory 703 complete mutual communication through the communication bus 704,
a memory 703 for storing a computer program;
the processor 701 is configured to implement the following steps when executing the program stored in the memory 703:
acquiring current advertisement data of an advertisement played at the current time; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated by the advertisement played at the current time when the user clicks;
taking the current advertisement characteristic data and the current user data as the input of a preset FM model, taking the first click rate as a training target output by the preset FM model, training the preset FM model, and taking the preset FM model after training as a trained FM model;
inputting each piece of advertisement data to be selected in a preset advertisement data set to be selected into a trained FM model, and outputting a second estimated click rate of each advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data;
and selecting the advertisement to be selected with the highest second click rate, and playing the advertisement within the preset time after the current time.
The communication bus mentioned in the electronic device may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface is used for communication between the electronic equipment and other equipment.
The Memory may include a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.
The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the device can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.
In another embodiment of the present invention, a computer-readable storage medium is further provided, in which instructions are stored, and when the instructions are executed on a computer, the computer is caused to execute the method for playing an advertisement according to any one of the above embodiments.
In another embodiment of the present invention, there is also provided a computer program product containing instructions, which when run on a computer, causes the computer to execute the method for playing an advertisement as described in any of the above embodiments.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, as for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.
Claims (13)
1. A method for playing an advertisement, the method comprising:
acquiring current advertisement data of an advertisement played at the current time; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated when the user clicks the advertisement played at the current time;
taking the current advertisement characteristic data and the current user data as input of a preset FM model, taking the first click rate as a training target output by the preset FM model, training the preset FM model, and taking the trained preset FM model as a trained FM model;
inputting each piece of advertisement data to be selected in a preset advertisement data set to be selected into the trained FM model, and outputting the estimated second click rate of each advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data;
and selecting the advertisement to be selected with the highest second click rate, and playing the advertisement within the preset time after the current time.
2. The method of claim 1, wherein the using the current advertisement feature data and the current user data as inputs of the preset FM model, using the first click rate as a training target of the preset FM model output, training the preset FM model, and using the trained preset FM model as a trained FM model comprises:
taking current advertisement characteristic data and current user data as input of a preset FM model, comparing the first click rate with a third click rate, and determining whether an error value of the first click rate and the third click rate output by the preset FM model is minimum or not by using a gradient descent algorithm; the third click rate is: inputting the current advertisement characteristic data and the current user data as well as the click rate output by the preset FM model;
if the error value of the first click rate and the third click rate is not the minimum, adjusting all parameters of a preset FM model until the error value of the first click rate and the third click rate is the minimum;
and taking the preset FM model after adjusting each parameter as the trained FM model.
3. The method of claim 1, wherein the using the current advertisement feature data and the current user data as inputs of the preset FM model, using the first click rate as a training target of the preset FM model output, training the preset FM model, and using the trained preset FM model as a trained FM model comprises:
taking the current advertisement characteristic data and the current user data as the input of a preset FM model, comparing the first click rate with a third click rate, and calculating to obtain an error value of the first click rate and the third click rate; the third click rate is: inputting the current advertisement characteristic data and the current user data as well as the click rate output by the preset FM model;
taking the error value and each preset parameter in the preset FM model as the input of a preset maximum likelihood estimation function, and taking the logarithm of the output of the preset maximum likelihood estimation function to obtain a loss function of the preset FM model output;
calculating a loss function by utilizing a leader FTRL algorithm following the normality, and determining whether the loss function output by the preset FM model is minimum or not for the gradient of each parameter of the preset FM model;
if the loss function value is not the minimum, adjusting all parameters of the preset FM model until the loss function value is the minimum;
and taking the preset FM model after adjusting each parameter as the trained FM model.
4. The method of claim 1, wherein the preset FM model is obtained by pre-training by:
taking historical advertisement data of an advertisement as a training sample to obtain a training set; the training set includes: a plurality of training samples; each training sample comprises: historical user data, historical advertisement characteristic data and historical click rate generated by clicking the advertisement by the user;
taking each training sample in a training set as the input of an initial FM model, and taking the historical click rate in each training sample as the training target of the initial FM model;
calculating each parameter of a loss function value preset in the gradient descending direction of a pseudo-gradient function set by a normal finite memory quasi-Newton OWLQN algorithm, and determining whether the loss function output by a preset FM model is minimum or not;
if the loss function value is not the minimum, adjusting all parameters of a preset FM model until the loss function value is the minimum;
and taking the initial FM model after each parameter is adjusted as a preset FM model.
5. The method of claim 1, wherein the preset ad dataset to be selected is obtained by:
combining advertisement characteristic data of a selected advertisement and user data to serve as a prediction sample;
acquiring a preset advertisement data set to be selected; the preset advertisement data set to be selected comprises: a plurality of prediction samples.
6. The method of claim 1, further comprising:
and taking the current advertisement data as an advertisement sample, adding the preset advertisement data set to be selected, and updating the preset advertisement data set to be selected.
7. An apparatus for playing an advertisement, the apparatus comprising:
the acquisition module is used for acquiring current advertisement data of the advertisement played at the current time; the current advertisement data includes: current user data, current advertisement characteristic data and a first click rate; the first click rate is the click rate generated when the user clicks the advertisement played at the current time;
the training module is used for taking the current advertisement characteristic data and the current user data as input of a preset FM model, taking the first click rate as a training target output by the preset FM model, training the preset FM model, and taking the preset FM model after training as a trained FM model;
the estimation module is used for inputting each piece of advertisement data to be selected in a preset advertisement data set to be selected into the trained FM model and outputting an estimated second click rate of each piece of advertisement to be selected; the advertisement data to be selected includes: user data and advertisement characteristic data;
and the playing module is used for selecting the advertisement to be selected with the highest second click rate and playing the advertisement within a preset time after the current time.
8. The apparatus of claim 7, wherein the training module comprises:
the first error unit is used for taking the current advertisement characteristic data and the current user data as the input of a preset FM model, comparing the first click rate with a third click rate, and determining whether the error value of the first click rate and the third click rate output by the preset FM model is the minimum or not by utilizing a gradient descent algorithm; the third click rate is: inputting the current advertisement characteristic data and the current user data as well as the click rate output by the preset FM model;
a first adjusting unit, configured to adjust each parameter of a preset FM model until an error value between the first click rate and the third click rate is minimum if the error value between the first click rate and the third click rate is not minimum;
and the first determining unit is used for taking the preset FM model after each parameter is adjusted as the trained FM model.
9. The apparatus of claim 7, wherein the training module comprises:
the second error unit is used for taking the current advertisement characteristic data and the current user data as the input of a preset FM model, comparing the first click rate with a third click rate, and calculating to obtain the error value of the first click rate and the third click rate; the third click rate is: inputting the current advertisement characteristic data and the current user data as well as the click rate output by the preset FM model;
the first loss unit is used for taking the error value and each preset parameter in the preset FM model as the input of a preset maximum likelihood estimation function, and taking the logarithm of the output of the preset maximum likelihood estimation function to obtain the loss function output by the preset FM model;
the calculation unit is used for calculating a loss function by utilizing an FTRL algorithm, and determining whether the loss function output by the preset FM model is minimum or not for the gradient of each parameter of the preset FM model;
a second adjusting unit, configured to adjust each parameter of the preset FM model until the loss function value is minimum if the loss function value is not minimum;
and the second determining unit is used for taking the preset FM model after each parameter is adjusted as the trained FM model.
10. The apparatus of claim 7, further comprising:
the sample module is used for taking historical advertisement data of an advertisement as a training sample to obtain a training set; the training set includes: a plurality of training samples; each training sample comprises: historical user data, historical advertisement characteristic data and historical click rate generated by clicking the advertisement by the user;
the target module is used for taking each training sample in the training set as the input of an initial FM model and taking the historical click rate in each training sample as the training target of the initial FM model;
the loss module is used for calculating parameters of a loss function value preset in the gradient descending direction by utilizing the gradient descending direction of a pseudo-gradient function set by a normal finite memory quasi-Newton OWLQN algorithm and determining whether the loss function output by the preset FM model is minimum or not;
the loss adjusting module is used for adjusting each parameter of the preset FM model until the loss function value is minimum if the loss function value is not minimum;
and the model determining module is used for taking the initial FM model after each parameter is adjusted as a preset FM model.
11. The apparatus of claim 7, further comprising:
the prediction module is used for combining advertisement characteristic data of a selected advertisement and user data to serve as a prediction sample;
the system comprises a candidate module, a selection module and a selection module, wherein the candidate module is used for obtaining a preset candidate advertisement data set; the preset advertisement data set to be selected comprises: a plurality of prediction samples.
12. The apparatus of claim 7, further comprising:
and the updating module is used for adding the current advertisement data serving as an advertisement sample into the preset advertisement data set to be selected and updating the preset advertisement data set to be selected.
13. An electronic device is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor and the communication interface are used for realizing mutual communication by the memory through the communication bus;
a memory for storing a computer program;
a processor for implementing the method steps of any of claims 1-6 when executing a program stored in the memory.
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 CN108965951A (en) | 2018-12-07 |
| CN108965951B true 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) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825966B (en) * | 2019-10-31 | 2022-03-04 | 广州市百果园信息技术有限公司 | Information recommendation method and device, recommendation server and storage medium |
| 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 |
| 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 |
| 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 |
| 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 |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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***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 |
Family Cites Families (3)
| 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 |
| US8364525B2 (en) * | 2010-11-30 | 2013-01-29 | Yahoo! Inc. | Using clicked slate driven click-through rate estimates in sponsored search |
-
2018
- 2018-07-12 CN CN201810762883.8A patent/CN108965951B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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***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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108965951A (en) | 2018-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108965951B (en) | Advertisement playing method and device | |
| US20200356572A1 (en) | Search ranking method and apparatus, electronic device and storage medium | |
| CN109359217B (en) | User interest degree calculation method, server and readable storage medium | |
| CN109685536B (en) | Method and apparatus for outputting information | |
| CN109640176B (en) | Method and apparatus for generating information | |
| CN111612377A (en) | Information pushing method and device, electronic equipment and computer readable medium | |
| CN110782286A (en) | Advertisement pushing method and device, server and computer readable storage medium | |
| CN109348260B (en) | Live broadcast room recommendation method, device, equipment and medium | |
| CN110121088B (en) | User attribute information determination method and device and electronic equipment | |
| CN111815373A (en) | Conversion cost optimization method and device, electronic equipment and storage medium | |
| CN112883275B (en) | Live broadcast room recommendation method, device, server and medium | |
| CN114119123A (en) | Information pushing method and device | |
| CN109618193B (en) | Method and apparatus for processing information | |
| CN110020129B (en) | Click rate correction method, prediction method, device, computing equipment and storage medium | |
| CN112287208A (en) | User portrait generation method and device, electronic equipment and storage medium | |
| CN114257521B (en) | Traffic prediction method, traffic prediction device, electronic equipment and storage medium | |
| US20160140454A1 (en) | User Interest Learning through Hierarchical Interest Graphs | |
| WO2018205642A1 (en) | Video revenue calculation modeling device and method, video recommendation device and method, server, and storage medium | |
| CN111080349B (en) | Method, device, server and medium for identifying multiple devices of same user | |
| CN115271866A (en) | Product recommendation method and device, electronic equipment and readable storage medium | |
| CN111666486B (en) | Method and device for acquiring heat value | |
| US11494710B1 (en) | Optimization constraint adaptation for long-term target achievement | |
| CN113780703A (en) | Index adjusting method and device | |
| CN113850416A (en) | Advertisement promotion cooperation object determining method and device | |
| CN113591999A (en) | End edge cloud federal learning model training system and method |
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 |