CN105809462A - Estimation method and device for estimating advertisement click rate - Google Patents
Estimation method and device for estimating advertisement click rate Download PDFInfo
- Publication number
- CN105809462A CN105809462A CN201410843845.7A CN201410843845A CN105809462A CN 105809462 A CN105809462 A CN 105809462A CN 201410843845 A CN201410843845 A CN 201410843845A CN 105809462 A CN105809462 A CN 105809462A
- Authority
- CN
- China
- Prior art keywords
- advertisement
- vector
- weight vector
- user
- characteristic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention is applicable to the information pushing field and provides an estimation method and a device for estimating the advertisement click rate. The method comprises the steps of acquiring the characteristics of a user and the advertisement characteristics of the to-be-pushed advertisement information, encoding the acquired characteristics to obtain a user characteristic vector and an advertisement characteristic vector, calculating a weight vector in the logistic regression algorithm based on the memory-limited BFGS algorithm through the manner of alternatively fixing parameters, and estimating the click rate of the advertisement information according to the weight vector, the user characteristic vector and the advertisement characteristic vector. Compared with the prior art, parameters are effectively constrained in the iterative way through alternatively fixing the parameters, so that the over-fitting phenomenon in the logistic regression algorithm can be avoided. Therefore, the estimation efficiency of the advertisement click rate is improved. Meanwhile, the estimation accuracy is improved.
Description
Technical field
The invention belongs to information pushing field, particularly relate to evaluation method and the device of a kind of ad click rate.
Background technology
Along with the development of television set intelligently, networking, television content views and admires experience to what user brought more horn of plenty, thus also makes intelligent television user also be continuously increased.Intelligent television, while playing TV programme to user, also can push and user-dependent advertising message, when the user that breath pushes to correspondence is produced in advertisement, then and can by obtaining the accuracy that the clicking rate of user evaluates the information of propelling movement.By improving degree of accuracy, thus improving the efficiency of transmitting advertisement information.
In order to improve the clicking rate of advertising message, current widely used logistic regression algorithm estimates the probability that user clicks.Wherein, described logistic regression algorithm is by estimating that user is classified by user characteristics weights.When advertised event is about to occur, logistic regression algorithm can calculate user characteristics and weights, and effectively estimation user clicks the probability of advertisement, it was predicted that the clicking rate of this advertisement, thus further guide design advertisement pushing strategy accurately.
But, owing to the parameter of logistic regression algorithm lacks constraint means, general regularization method, as lasso algorithm or ridge algorithm are bad to the effect of the constraint of the parameter of logistic regression, easily produce over-fitting problem, affect predictive efficiency and the accuracy of ad click rate.
Summary of the invention
It is an object of the invention to provide the evaluation method of a kind of ad click rate, to solve prior art due to the parameter shortage constraint means of logistic regression algorithm, general regularization method is bad to the effect of the constraint of the parameter of logistic regression, easily produce over-fitting problem, affect the predictive efficiency of ad click rate and the problem of accuracy.
The present invention is achieved in that the evaluation method of a kind of ad click rate, and described method includes:
Acquisition user characteristics, and treat the characteristic of advertisement of advertisement information, coding obtains user characteristics vector sum characteristic of advertisement vector；
By the BFGS algorithm based on limited memory, adopt the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter；
The clicking rate of described advertising message is estimated according to described weight vector, user characteristics vector and characteristic of advertisement vector.
Another object of the present invention is to provide the estimating device of a kind of ad click rate, described device includes:
Obtain coding unit, be used for obtaining user characteristics, and treat that the characteristic of advertisement of advertisement information, coding obtain user characteristics vector sum characteristic of advertisement vector；
Iterative computation unit, for by the BFGS algorithm based on limited memory, adopting the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter；
Clicking rate evaluation unit, for estimating the clicking rate of described advertising message according to described weight vector, user characteristics vector and characteristic of advertisement vector.
In the present invention, it is encoded by obtaining user characteristics and characteristic of advertisement, BFGS algorithm based on limited memory, adopt the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter, comparing with the method for existing calculating weight vector, parameter can effectively be retrained by the method being iterated by mutual preset parameter, it is to avoid occurs over-fitting problem in logistic regression algorithm, it is thus possible to improve the estimation efficiency of ad click rate, improve the precision of estimation.
Accompanying drawing explanation
Fig. 1 is the flowchart of the evaluation method of the ad click rate that the embodiment of the present invention provides；
Fig. 2 is the flowchart of the algorithm solving weight vector that the embodiment of the present invention provides；
Fig. 3 is the flowchart of the evaluation method of the another ad click rate that the embodiment of the present invention provides；
Fig. 4 is the flowchart of the evaluation method of the another ad click rate that the embodiment of the present invention provides；
Fig. 5 is the structural representation of the estimating device of the ad click rate that the embodiment of the present invention provides.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein is only in order to explain the present invention, is not intended to limit the present invention.
When the main purpose of the embodiment of the present invention is in that to solve to estimate ad click rate by Logic Regression Models in prior art, owing to the parameter of Logic Regression Models lacks effective constraint means, using general regularization method, when such as lasso algorithm or ridge algorithm are to the restriction on the parameters of logistic regression, the binding effect of parameter is bad, so that when carrying out ad click rate estimation, problem over-fitting easily occur, affects the estimation efficiency of ad click rate and the problem of the accuracy of estimation.
For solving this problem, the present invention proposes the evaluation method of a kind of ad click rate, and described method includes: acquisition user characteristics, and treats the characteristic of advertisement of advertisement information, and coding obtains user characteristics vector sum characteristic of advertisement vector；By the BFGS algorithm based on limited memory, adopt the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter；The clicking rate of described advertising message is estimated according to described weight vector, user characteristics vector and characteristic of advertisement vector.The present invention compares with the method for existing calculating weight vector, parameter can effectively be retrained by the method being iterated by mutual preset parameter, avoid logistic regression algorithm occurs over-fitting problem such that it is able to improve the estimation efficiency of ad click rate, improve the precision of estimation.Illustrate below in conjunction with accompanying drawing.
What Fig. 1 illustrated the evaluation method of the ad click rate that the embodiment of the present invention provides realizes flow process, and details are as follows:
In step S101, acquisition user characteristics, and treat the characteristic of advertisement of advertisement information, coding obtains user characteristics vector sum characteristic of advertisement vector.
Concrete, user characteristics described in the embodiment of the present invention, it is possible to include user property feature, user watches programs feature, user behavior feature.And the user characteristics described in the embodiment of the present invention and characteristic of advertisement, can be carried out sliding-model control, generate corresponding K dimensional feature vector according to the value number K of each feature in described user characteristics and described characteristic of advertisement.When one of them feature in such as user characteristics and characteristic of advertisement has 5 feature values, then need to generate the characteristic vector of 5 dimensions.Feature object concrete individually will be carried out respectively sliding-model control explanation below.
Wherein, described user property feature can include the information such as age of user, sex, education degree, income, certificate address information.Described user can be the mark that the interface accounts information of television terminal identifies as user.Described age of user carries out sliding-model control, it is possible to age of user is divided into 5 fields, respectively " child ", " teenager ", " youth ", " middle age ", " old age ", wherein only have signature identification position take 1, remaining flag all takes 0.If described user is middle age user, then namely the characteristic vector for age of user is represented by [0,0,1,0,0].
Equally, for education degree, can being divided into " primary school ", " middle school ", " university ", " master ", " doctor " five fields, wherein only have signature identification position to take 1, other flag all takes 0, if user is university education degree, then the characteristic vector of corresponding education degree is [0,0,1,0,0].
For the further feature equally possible coding carrying out characteristic vector according to above-mentioned sliding-model control method of vector.
Described user watches programs feature and includes the programm name of user's viewing, program category, program protagonist, PD program director, programme language etc..If program category can include parent-offspring's class program, history class program, war class program, science fiction class program etc..Described programme language can include such as national language, English, Japanese.
Described user behavior feature includes clicking advertisement and not clicking advertisement.
Described characteristic of advertisement includes advertisement classification, commodity price, the commodity place of production, commodity production enterprise sort etc..Described advertisement classification can include commodity series advertisements and service series advertisements, or can also adopt other more specifically advertisement classification dividing mode.
After system acquisition user characteristics and characteristic of advertisement, form primitive character set, after feature coding, form training data { xi, yi, i=1 ..., N is x whereini∈{0,1}d, vector x is user characteristics, advertisement characteristic data, and it is the characteristic dimension of vector x that element only takes 0 or 1, d.
yi∈-1, and 1}, represent that y value only takes-1 and 1.In the present invention, y=1 represents user and clicks advertisement, and y=-1 then represents that user does not click on.The target of ad click rate model is to find a weight vector w (w is a d dimensional vector) so that the user clicking advertisement and the user not clicking on advertisement distinguish as much as possible.When a new user watches certain program, system extracts user characteristics and characteristic of advertisement, is encoded according to above-mentioned encoding mechanism, forms characteristic vector x.
If the total Characteristic Number of user is F, the value of each feature is cfIndividual, then the feature space dimension c in Logic Regression Models definition can be calculated as:
In step s 102, by the BFGS algorithm based on limited memory, the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter is adopted.
Concrete, the BFGS algorithm of described limited memory, also it to be L-BFGS algorithm, be the more excellent algorithm of the one in Quasi-Newton algorithm, its title is to be obtained by the surname initial of its inventor Broyden, Fletcher, Goldfarb and Shanno.
Described by the BFGS algorithm based on limited memory, adopt the weight vector step in the mode iterative computation logistic regression algorithm of mutual preset parameter to include:
In step sl, according to logistic regression functionExpression show in solution formula corresponding for weight vector w, namely the object function of logistic regression algorithm model of the present invention is: First the value of parameter alpha to be optimized and λ is fixed, by the optimal value of the BFGS Algorithm for Solving weight vector w of limited memory.
Certainly, first because changing the value of described parameter alpha and λ, calculate the mode of the optimal value of weight vector, simply embodiment of the present invention one embodiment, persons skilled in the art are understandable that, can also first solidify weight vector, realize the purpose of the present invention by the optimal value of BFGS Algorithm for Solving Parameters variation amount α and the λ of limited memory is equally possible.Solidify weight vector in the ban, during by the optimal value of BFGS Algorithm for Solving Parameters variation amount α and the λ of limited memory, be adjusted correspondingly in subsequent steps.
In step s 2, the weight vector w optimal value solved is fixed, by the optimal value of the BFGS Algorithm for Solving parameter alpha of limited memory and λ, and the optimal value of described parameter alpha and λ is fixed the optimal value next time calculating w, so iterate, until solution formula convergence corresponding for weight vector w, wherein, y is user behavior, w is weight vector, α, λ are weight computing parameters of formula variable, and d is the dimension of weight vector.
In embodiments of the present invention, if using common L-BFGS algorithm, it can only calculate the optimal value of convex function, and the non-convex function for target of the present invention then can not calculate its optimal value.In order to solve this problem, the parameter that the present invention first standing part is to be optimized so that non-convex function becomes convex function, then run L-BFGS Algorithm for Solving optimal value.Then fixing this optimal value, then go to optimize another part parameter, iteration is until object function is restrained successively.
The algorithmic procedure solving weight vector describes as in figure 2 it is shown, include:
S201, input feature value data set { (xi, yi), i=1 ..., N}, wherein N is the number of characteristic vector.First preset parameter (α, λ), gradient g (w, the α of calculating target function, λ), approximate extra large gloomy matrix H (w, α, λ) is then calculated, make d (w, α, λ)=-1*H (w, α, λ) * g (w, α, λ, to function d (w, α, λ) carry out linear search, obtain best initial weights vector w1, and weight w 1 is updated.
S202, after obtaining described best initial weights vector, fixing described best initial weights vector w1, the gradient g (w of calculating target function, α, λ), and calculate approximate extra large gloomy matrix H (w, α, λ), d (w, α, λ)=-1*H (w is made, α, λ) * g (w, α, λ), function d (w, α, λ) is carried out linear search, obtains optimal value of the parameter (α, λ), and to weights (α, λ) update.
S203, it is judged that object function whether stable convergence, without convergence, then goes successively to step S1 and is iterated solving.After algorithmic stability is restrained, such as step S204, final model parameter (w, α, λ) can be obtained.Wherein w is the Model Weight value that we want.
In step S203, estimate the clicking rate of described advertising message according to described weight vector, user characteristics vector and characteristic of advertisement vector.
The flowchart of the evaluation method of the another ad click rate that Fig. 3 provides for the embodiment of the present invention, details are as follows:
Step S301, S303, S304 are identical to step S103 with the step S101 described in Fig. 1 respectively, are not repeated at this and repeat, are different in that with Fig. 1, before step S303, also include step S302.
In step s 302, by hash function by described user characteristics vector, and characteristic of advertisement vector carries out dimension-reduction treatment.
Owing to the coded method described in the embodiment of the present invention can cause the sharp increase of user characteristics dimension, cause dimension disaster problem, cause that efficiency of algorithm declines.In order to solve this problem, the present invention adopts the method for Hash dimensionality reduction, drops in a relatively low feature space by hash function by original user characteristics and characteristic of advertisement, preserves original characteristic information simultaneously as far as possible.
Wherein, for F user characteristics, it is necessary to use F hash function dimensionality reduction, the dimension after dimensionality reduction is d, if luv space dimension c is 200, d is 100, then and dimension 100 dimensions less of original feature space after dimensionality reduction.Need data volume to be processed it is thus possible to greatly reduce, improve the computational efficiency of logistic regression.
The evaluation method flowchart of the another ad click rate that Fig. 4 provides for the embodiment of the present invention, compared to Figure 1, Fig. 4 also includes step S404, the clicking rate of the described advertising message according to estimation, described advertising message is carried out successively sequence, preferentially pushes the advertising message to be pushed that the clicking rate of estimation is high.
Running parameter according to alternate curing object function of the present invention such that it is able to the more crucial value effectively calculating weight vector, by the weight vector value calculated, in conjunction with user characteristics vector and characteristic of advertisement vector, according to logistic regression functionThe ad click rate of estimation can be obtained.
After obtaining the clicking rate of advertisement of required propelling movement, according to user, advertising message can be ranked up, and according to described ranking results, select the transmitting advertisement information extremely described user that the clicking rate of estimation is the highest.It is thus possible to obtain higher clicking rate so that the advertising message of propelling movement is more accurate.
The structural representation of the estimating device of the ad click rate that Fig. 5 provides for the embodiment of the present invention, as it is shown in figure 5, the estimating device of described ad click rate includes:
Obtain coding unit 501, be used for obtaining user characteristics, and treat that the characteristic of advertisement of advertisement information, coding obtain user characteristics vector sum characteristic of advertisement vector；
Iterative computation unit 502, for by the BFGS algorithm based on limited memory, adopting the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter；
Clicking rate evaluation unit 503, for estimating the clicking rate of described advertising message according to described weight vector, user characteristics vector and characteristic of advertisement vector.
Preferably, described iterative computation unit includes:
First computation subunit, for according to logistic regression functionExpression show in solution formula corresponding for weight vector w: First the value of parameter alpha to be optimized and λ is fixed, by the optimal value of the BFGS Algorithm for Solving weight vector w of limited memory；
Second computation subunit, for the weight vector solved w optimal value being fixed, by the optimal value of the BFGS Algorithm for Solving parameter alpha of limited memory and λ, and the optimal value of described parameter alpha and λ is fixed the optimal value next time calculating w, so iterate, until solution formula convergence corresponding for weight vector w, wherein, y is user behavior, w is weight vector, α, λ are weight computing parameters of formula variable, and d is the dimension of weight vector.
Preferably, described device also includes:
Dimensionality reduction unit, for described user characteristics is vectorial by hash function, and characteristic of advertisement vector carries out dimension-reduction treatment.
Preferably, described user characteristics includes user property feature, user watches programs feature, user behavior feature.
Preferably, described user property feature includes age of user, sex, education degree, income, certificate address information, described user watches programs feature and includes the programm name of user's viewing, program category, program protagonist, PD program director, programme language, described user behavior feature includes clicking advertisement and not clicking advertisement, and described characteristic of advertisement includes advertisement classification, commodity price, the commodity place of production, commodity production enterprise sort.
Preferably, described device also includes:
Push unit, for the clicking rate of the described advertising message according to estimation, carries out successively sequence, preferentially pushes the advertising message to be pushed that the clicking rate of estimation is high described advertising message.
The estimating device of ad click rate described in the embodiment of the present invention is corresponding with the evaluation method of above-mentioned ad click rate, is not repeated at this and repeats.
In several embodiments provided by the present invention, it should be understood that disclosed apparatus and method, it is possible to realize by another way.Such as, device embodiment described above is merely schematic, such as, the division of described unit, being only a kind of logic function to divide, actual can have other dividing mode when realizing, for instance multiple unit or assembly can in conjunction with or be desirably integrated into another system, or some features can ignore, or do not perform.Another point, shown or discussed coupling each other or direct-coupling or communication connection can be through INDIRECT COUPLING or the communication connection of some interfaces, device or unit, it is possible to be electrical, machinery or other form.
The described unit illustrated as separating component can be or may not be physically separate, and the parts shown as unit can be or may not be physical location, namely may be located at a place, or can also be distributed on multiple NE.Some or all of unit therein can be selected according to the actual needs to realize the purpose of the present embodiment scheme.
It addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, it is also possible to be that unit is individually physically present, it is also possible to two or more unit are integrated in a unit.Above-mentioned integrated unit both can adopt the form of hardware to realize, it would however also be possible to employ the form of SFU software functional unit realizes.
If described integrated unit is using the form realization of SFU software functional unit and as independent production marketing or use, it is possible to be stored in a computer read/write memory medium.Based on such understanding, part or all or part of of this technical scheme that prior art is contributed by technical scheme substantially in other words can embody with the form of software product, this computer software product is stored in a storage medium, including some instructions with so that a computer equipment (can be personal computer, server, or the network equipment etc.) perform all or part of of method described in each embodiment of the present invention.And aforesaid storage medium includes: USB flash disk, portable hard drive, read only memory (ROM, Read-OnlyMemory), the various media that can store program code such as random access memory (RAM, RandomAccessMemory), magnetic disc or CD.
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all any amendment, equivalent replacement and improvement etc. made within the spirit and principles in the present invention, should be included within protection scope of the present invention.
Claims (10)
1. the evaluation method of an ad click rate, it is characterised in that described method includes:
Acquisition user characteristics, and treat the characteristic of advertisement of advertisement information, coding obtains user characteristics vector sum characteristic of advertisement vector；
By the BFGS algorithm based on limited memory, adopt the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter；
The clicking rate of described advertising message is estimated according to described weight vector, user characteristics vector and characteristic of advertisement vector.
2. method according to claim 1, it is characterised in that described by the BFGS algorithm based on limited memory, adopts the weight vector step in the mode iterative computation logistic regression algorithm of mutual preset parameter to include:
According to logistic regression functionExpression show in solution formula corresponding for weight vector w: First the value of parameter alpha to be optimized and λ is fixed, by the optimal value of the BFGS Algorithm for Solving weight vector w of limited memory；
The weight vector w optimal value solved is fixed, optimal value by the BFGS Algorithm for Solving parameter alpha of limited memory and λ, and the optimal value of described parameter alpha and λ is fixed the optimal value next time calculating w, so iterate, until solution formula convergence corresponding for weight vector w, wherein, y is user behavior, and w is weight vector, α, λ is weight computing parameters of formula variable, and d is the dimension of weight vector.
3. method according to claim 1, it is characterised in that described by the BFGS algorithm based on limited memory, before adopting the weight vector step in the mode iterative computation logistic regression algorithm of mutual preset parameter, described method also includes:
By hash function by described user characteristics vector, and characteristic of advertisement vector carries out dimension-reduction treatment.
4. method according to claim 1, it is characterised in that described coding obtains user characteristics vector sum characteristic of advertisement vector step and includes:
Corresponding K dimensional feature vector is generated according to the value number K of each feature in described user characteristics and described characteristic of advertisement.
5. method according to any one of claim 1-4, it is characterised in that described user characteristics includes user property feature, user watches programs feature, user behavior feature.
6. method according to claim 5, it is characterized in that, described user property feature includes age of user, sex, education degree, income, certificate address information, described user watches programs feature and includes the programm name of user's viewing, program category, program protagonist, PD program director, programme language, described user behavior feature includes clicking advertisement and not clicking advertisement, and described characteristic of advertisement includes advertisement classification, commodity price, the commodity place of production, commodity production enterprise sort.
7. method according to claim 1, it is characterised in that after the described clicking rate step estimating described advertising message according to described weight vector, user characteristics vector and characteristic of advertisement vector, described method also includes:
The clicking rate of the described advertising message according to estimation, carries out successively sequence to described advertising message, and what preferentially the clicking rate of propelling movement estimation was high treats advertisement information.
8. the estimating device of an ad click rate, it is characterised in that described device includes:
Obtain coding unit, be used for obtaining user characteristics, and treat that the characteristic of advertisement of advertisement information, coding obtain user characteristics vector sum characteristic of advertisement vector；
Iterative computation unit, for by the BFGS algorithm based on limited memory, adopting the weight vector in the mode iterative computation logistic regression algorithm of mutual preset parameter；
Clicking rate evaluation unit, for estimating the clicking rate of described advertising message according to described weight vector, user characteristics vector and characteristic of advertisement vector.
9. device according to claim 8, it is characterised in that described iterative computation unit includes:
First computation subunit, for according to logistic regression functionExpression show in solution formula corresponding for weight vector w: First the value of parameter alpha to be optimized and λ is fixed, by the optimal value of the BFGS Algorithm for Solving weight vector w of limited memory；
Second computation subunit, for the weight vector solved w optimal value being fixed, by the optimal value of the BFGS Algorithm for Solving parameter alpha of limited memory and λ, and the optimal value of described parameter alpha and λ is fixed the optimal value next time calculating w, so iterate, until solution formula convergence corresponding for weight vector w, wherein, y is user behavior, w is weight vector, α, λ are weight computing parameters of formula variable, and d is the dimension of weight vector.
10. device according to claim 8, it is characterised in that described device also includes:
Dimensionality reduction unit, for described user characteristics is vectorial by hash function, and characteristic of advertisement vector carries out dimension-reduction treatment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410843845.7A CN105809462A (en) | 2014-12-30 | 2014-12-30 | Estimation method and device for estimating advertisement click rate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410843845.7A CN105809462A (en) | 2014-12-30 | 2014-12-30 | Estimation method and device for estimating advertisement click rate |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105809462A true CN105809462A (en) | 2016-07-27 |
Family
ID=56420077
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410843845.7A Pending CN105809462A (en) | 2014-12-30 | 2014-12-30 | Estimation method and device for estimating advertisement click rate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105809462A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106997549A (en) * | 2017-02-14 | 2017-08-01 | 火烈鸟网络(广州)股份有限公司 | The method for pushing and system of a kind of advertising message |
| CN107613022A (en) * | 2017-10-20 | 2018-01-19 | 广州优视网络科技有限公司 | Content delivery method, device and computer equipment |
| CN108805611A (en) * | 2018-05-21 | 2018-11-13 | 北京小米移动软件有限公司 | Advertisement screening technique and device |
| CN109934629A (en) * | 2019-03-12 | 2019-06-25 | 重庆金窝窝网络科技有限公司 | A kind of information-pushing method and device |
| CN110321422A (en) * | 2018-03-28 | 2019-10-11 | 腾讯科技（深圳）有限公司 | Method, method for pushing, device and the equipment of on-line training model |
| CN110933499A (en) * | 2018-09-19 | 2020-03-27 | 飞狐信息技术（天津）有限公司 | Video click rate estimation method and device |
| CN111080377A (en) * | 2019-12-31 | 2020-04-28 | 苏宁云计算有限公司 | Method, system and device for generating business circle data |
| CN112819492A (en) * | 2019-11-15 | 2021-05-18 | 北京达佳互联信息技术有限公司 | Advertisement recommendation method and device and electronic equipment |
| CN113129046A (en) * | 2019-12-31 | 2021-07-16 | 上海哔哩哔哩科技有限公司 | Click rate prediction method and device and computer equipment |
| CN113139825A (en) * | 2020-01-20 | 2021-07-20 | 上海哔哩哔哩科技有限公司 | Method and device for determining distribution authority of advertisement space and computer equipment |
-
2014
- 2014-12-30 CN CN201410843845.7A patent/CN105809462A/en active Pending
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106997549A (en) * | 2017-02-14 | 2017-08-01 | 火烈鸟网络(广州)股份有限公司 | The method for pushing and system of a kind of advertising message |
| CN107613022A (en) * | 2017-10-20 | 2018-01-19 | 广州优视网络科技有限公司 | Content delivery method, device and computer equipment |
| CN107613022B (en) * | 2017-10-20 | 2020-10-16 | 阿里巴巴（中国）有限公司 | Content pushing method and device and computer equipment |
| CN110321422A (en) * | 2018-03-28 | 2019-10-11 | 腾讯科技（深圳）有限公司 | Method, method for pushing, device and the equipment of on-line training model |
| CN110321422B (en) * | 2018-03-28 | 2023-04-14 | 腾讯科技（深圳）有限公司 | Method for training model on line, pushing method, device and equipment |
| CN108805611A (en) * | 2018-05-21 | 2018-11-13 | 北京小米移动软件有限公司 | Advertisement screening technique and device |
| CN110933499A (en) * | 2018-09-19 | 2020-03-27 | 飞狐信息技术（天津）有限公司 | Video click rate estimation method and device |
| CN109934629A (en) * | 2019-03-12 | 2019-06-25 | 重庆金窝窝网络科技有限公司 | A kind of information-pushing method and device |
| CN112819492A (en) * | 2019-11-15 | 2021-05-18 | 北京达佳互联信息技术有限公司 | Advertisement recommendation method and device and electronic equipment |
| CN111080377A (en) * | 2019-12-31 | 2020-04-28 | 苏宁云计算有限公司 | Method, system and device for generating business circle data |
| CN113129046A (en) * | 2019-12-31 | 2021-07-16 | 上海哔哩哔哩科技有限公司 | Click rate prediction method and device and computer equipment |
| CN113139825A (en) * | 2020-01-20 | 2021-07-20 | 上海哔哩哔哩科技有限公司 | Method and device for determining distribution authority of advertisement space and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105809462A (en) | Estimation method and device for estimating advertisement click rate | |
| US10887640B2 (en) | Utilizing artificial intelligence to generate enhanced digital content and improve digital content campaign design | |
| Jain et al. | Extreme multi-label loss functions for recommendation, tagging, ranking & other missing label applications | |
| US11651286B2 (en) | Method and system for distributed machine learning | |
| CN111538912B (en) | Content recommendation method, device, equipment and readable storage medium | |
| US10650245B2 (en) | Generating digital video summaries utilizing aesthetics, relevancy, and generative neural networks | |
| US11868724B2 (en) | Generating author vectors | |
| CN105447730B (en) | Target user orientation method and device | |
| Lee et al. | Local collaborative ranking | |
| CN106776673B (en) | Multimedia document summarization | |
| US20210248449A1 (en) | Recommender system using bayesian graph convolution networks | |
| CA2941700A1 (en) | Clickstream purchase prediction using hidden markov models | |
| CN110796232A (en) | Attribute prediction model training method, attribute prediction method and electronic equipment | |
| CN111241394B (en) | Data processing method, data processing device, computer readable storage medium and electronic equipment | |
| US20220253722A1 (en) | Recommendation system with adaptive thresholds for neighborhood selection | |
| Sahu et al. | Movie popularity and target audience prediction using the content-based recommender system | |
| CN114202061A (en) | Article recommendation method, electronic device and medium based on generation of confrontation network model and deep reinforcement learning | |
| CN104915444A (en) | Information recommending method and device | |
| Sun et al. | Local residual similarity for image re-ranking | |
| CN115867919A (en) | Graph structure aware incremental learning for recommendation systems | |
| CN115344698A (en) | Label processing method, label processing device, computer equipment, storage medium and program product | |
| US11775813B2 (en) | Generating a recommended target audience based on determining a predicted attendance utilizing a machine learning approach | |
| Vančura et al. | Scalable linear shallow autoencoder for collaborative filtering | |
| Zhu et al. | Deep exploration for recommendation systems | |
| Luo et al. | HySAGE: A hybrid static and adaptive graph embedding network for context-drifting recommendations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160727 |