CN102663617A - Method and system for prediction of advertisement clicking rate - Google Patents
Method and system for prediction of advertisement clicking rate Download PDFInfo
- Publication number
- CN102663617A CN102663617A CN201210074541XA CN201210074541A CN102663617A CN 102663617 A CN102663617 A CN 102663617A CN 201210074541X A CN201210074541X A CN 201210074541XA CN 201210074541 A CN201210074541 A CN 201210074541A CN 102663617 A CN102663617 A CN 102663617A
- Authority
- CN
- China
- Prior art keywords
- data
- sample
- user
- advertisement
- click
- 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
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method and a system for prediction of advertisement clicking rate to solve a problem of accuracy of the clicking rate prediction affected by seriously imbalanced sample data in the original sample set. The method comprises: extracting sample data to construct an original sample set, wherein the sample data comprises clicked data and unclicked data of users; constructing a training sample set by carrying out sampling on the original sample set; constructing a prediction model by using the sample data in the training sample set as model parameters; and predicting the user clicking rate of each kind of advertisement by utilizing the prediction model to predict the testing sample set. The method and the system in the invention can eliminate the problem of serious imbalance of the proportion between the clicked data and the unclicked data in the original sample set and construct a relatively balanced training sample set. The method and the system improve recognition rate of the prediction model on the clicked rate and improve the accuracy of clicking rate prediction.
Description
Technical field
The application relates to network technology, particularly relates to a kind of clicking rate Forecasting Methodology and system of advertisement.
Background technology
The rise of internet makes people can when browsing the identical page, see different advertisement, can realize that the personalization of advertisement is showed.Through clicking rate is tested, can understand the interested advertisement of different user, thereby show corresponding advertisement more accurately to each user, to improve the clicking rate of advertisement, improve the visit capacity of the advertisement delivery effect and the page.
When being tested, clicking rate need carry out analysis modeling to the input effect of history.At first to extract sample data and make up original sample set, clicking rate is predicted, so comprise user's click data and click data not in the sample data.In this process, the unbalanced problem of sample data is a big problem of restriction modeling effect.Data shows that Internet advertising is at present thrown on average has only the user about 0.3% to click, i.e. the every displaying of advertisement 1000 times has only the click about 3 times.
Thereby original sample set is fashionable as training sample set, can cause training sample set mid point hit data and not between the click data ratio seriously unbalance.The sample characteristics of click data has not occupied the overwhelming majority in this serious unbalanced sample data; In the prior art; Directly said original sample set is gathered the structure forecast model as training sample; Cause predicting the outcome of forecast model to be partial to not click data, it is inaccurate to predict the outcome.
Sample data skewness weighing apparatus generally can cause certain type of sample data amount very rare, and is very deficient like click data in application.And in real data was excavated, the existence of noise data was inevitable, and to a certain extent prediction or disaggregated model was exerted an influence.In this unbalanced problem, because click data itself is very rare, the statistics that is difficult to provide enough is distinguished it with noise data mutually, anti-noise ability relatively a little less than.Thereby a spot of noise sample will influence training pattern foundation and predict the outcome.
Usually can contain noise in the original sample set; Because click data and not between the click data ratio seriously unbalance; For example click data and not the ratio of click data be 3: 997; Wherein contain 1 noise data, then noise data is just bigger to the influence of click data, and smaller to the influence of click data not.
Thereby when carrying out the training of forecast model based on this kind sample data; Click data does not exert an influence bigger than normal in model; And then cause forecast model to be partial to not click data more; Predicting the outcome of forecast model according to sample data obtains is partial to not click data, and click data only accounts for the fraction space, when based on said forecast model test sample book being tested; The forecast model of deviation can produce predicting the outcome of deviation, thereby the feasible situation of being partial to not click that predicts the outcome has influenced the clicking rate prediction accuracy.
Summary of the invention
The application provides a kind of clicking rate Forecasting Methodology and system of advertisement, to solve the serious unbalanced problem that can influence the clicking rate prediction accuracy of sample data in the original sample set.
In order to address the above problem, the application discloses a kind of clicking rate Forecasting Methodology of advertisement, comprising:
Extract sample data and make up the original sample set, wherein said sample data comprises user's click data and click data not;
Through said original sample set is sampled, make up the training sample set;
With the sample data in the said training sample set is that model parameter makes up forecast model;
Utilize said forecast model that test sample book is gathered and predict, dope the clicking rate of user to every kind of advertisement.
Preferably, said through structure training sample set that said original sample set is sampled, comprising:
Sample with the sampling of presetting comparison original training sample, and structure gathers than corresponding training sample with said sampling, wherein, said sampling of presetting is than being the click data that draws through statistics and the ratio of click data not.
Preferably,, make up the training sample set, comprising through said original sample set is sampled:
In the sampling, all click datas are all added in the said training sample set.
Preferably, said extraction sample data makes up the original sample set, comprising:
Make up the original sample set from throwing in the data of extracting data in the period as sample data;
And extracting each sample data corresponding sample characteristic in the original sample set, said sample characteristics is used to describe sample data;
Wherein, as click data, the conduct of user's not number of clicks is click data not with the number of clicks of throwing in user in the data.
Preferably, test sample book set is: to the advertisement putting page, extract click the advertisement putting page the user as the samples tested data after, the test sample book set of structure.
Preferably, described method also comprises:
User in the said test sample book set shows the advertisement that clicking rate is the highest to said user in the page.
Accordingly, disclosed herein as well is a kind of clicking rate prognoses system of advertisement, comprising:
Make up the original sample collection modules, be used to extract sample data and make up the original sample set, wherein said sample data comprises user's click data and click data not;
Make up the training sample collection modules, be used for making up the training sample set through said original sample set is sampled;
Make up the forecast model module, the sample data that is used for said training sample set is that model parameter makes up forecast model;
The clicking rate prediction module is used to utilize said forecast model that test sample book is gathered and predicts, dopes the clicking rate of user to every kind of advertisement.
Preferably; Said structure training sample collection modules be used for sampling with the sampling comparison original training sample that presets, and structure is gathered than corresponding training sample with said sampling; Wherein, said sampling of presetting is than being the click data that draws through statistics and the ratio of click data not.
Preferably, said structure original sample collection modules is used for making up the original sample set from throwing in the data of extracting data in the period as sample data; And extracting each sample data corresponding sample characteristic, said sample characteristics is used to describe sample data; Wherein, as click data, the conduct of user's not number of clicks is click data not with the number of clicks of throwing in user in the data.
Preferably, described system also comprises:
Make up the test sample book collection modules, be used for, extract the user who clicks the advertisement putting page and gather as samples tested data construct test sample book to the advertisement putting page.
Display module is used for the user to said test sample book set, in the page, shows the advertisement that clicking rate is the highest to said user.
Compared with prior art, the application comprises following advantage:
The application extracts sample data and makes up the original sample set, and wherein said sample data comprises user's click data and click data not, makes up the training sample set through said original sample set is sampled then.The application does not directly gather the original sample set as training sample; Make up the training sample set but original sample set is optimized; So just can eliminate in the original sample set click data and the serious unbalance problem of ratio between the click data not; Make up the training sample set of relative equilibrium, and the influence that this moment, noise data hit data to said training sample set mid point is less than the influence of noise data to click data in the original sample set.Be that model parameter makes up forecast model with the sample data in the said training sample set again; The sample data of ratio relative equilibrium make forecast model to the discrimination of click data than higher; Utilizing said forecast model that test sample book is gathered predicts; Predictive user is to the clicking rate of every kind of advertisement, and this moment, forecast model can seriously be partial to not click data, had improved the clicking rate prediction accuracy.
Secondly; The described sampling of the application is than the pairing click data and the ratio of click data not; Be to draw according to the statistics of repeatedly testing; Have statistical accuracy and objectivity, the data during the training sample that therefore makes up than sampling to sampling is gathered also have accuracy and objectivity, have further improved the clicking rate prediction accuracy.
Once more, the application can compare said original sample set with the sampling of presetting and sample, and in sampling, all click datas is all added in the said training sample set; Guaranteeing to click under the constant situation of data bulk; Reduce the data in the training sample set, made that the data of execution model training structure forecast model are fewer, reduced the burden of system; Accelerate processing speed of data, improved the efficient of model training.
Once more, the application than being the foundation of quantity of sampling quantity, therefore when making up the test sample book set, can be chosen at the sample that meets said set characteristic most with the sampling of presetting, and improves the accuracy and the specific aim of sample, has further improved the accuracy of test.
Description of drawings
Fig. 1 is the clicking rate Forecasting Methodology process flow diagram of the said a kind of advertisement of the application embodiment;
Fig. 2 is the clicking rate Forecasting Methodology process flow diagram of the said a kind of advertisement of the application's preferred embodiment;
Fig. 3 is based on feature space decision surface synoptic diagram in the clicking rate Forecasting Methodology of the said a kind of advertisement of the application's preferred embodiment;
Fig. 4 is the clicking rate Forecasting Methodology synoptic diagram of the said a kind of advertisement of the application's preferred embodiment;
Fig. 5 is the clicking rate prognoses system structural drawing of the said a kind of advertisement of the application embodiment.
Embodiment
For above-mentioned purpose, the feature and advantage that make the application can be more obviously understandable, the application is done further detailed explanation below in conjunction with accompanying drawing and embodiment.
In the prior art, in the original sample set click data and not between the click data ratio seriously unbalance, cause predicting the outcome of forecast model to be partial to not click data, thereby influenced the performance and the accuracy of clicking rate prediction.
The application is optimized original sample set and makes up the training sample set; Sample data in the training sample set of employing ratio relative equilibrium makes up forecast model; Utilizing said forecast model that test sample book is gathered predicts; Can predictive user to the clicking rate of every kind of advertisement, this moment, forecast model can be partial to not click data, had improved the clicking rate prediction accuracy.
With reference to Fig. 1, it has provided the clicking rate Forecasting Methodology process flow diagram of the said a kind of advertisement of the application embodiment.
At first to extract sample data and make up original sample set, clicking rate is predicted, so comprise user's click data and click data not in the sample data.
Wherein, there are click data and the serious unbalance situation of ratio between the click data not in the set of said original sample.
Above-mentioned argumentation can be known, if directly said original sample set is gathered the structure forecast model as training sample, can cause predicting the outcome of forecast model to be partial to click data.
In order to address this problem, be not directly said original sample set to be gathered as training sample among the application, but set is sampled to original sample, use the sample data after sampling to make up the training sample set.Therefore the sample data in the said training sample set also comprises click data and click data not.
Sample data during the training sample that makes up after the sampling is gathered distributes relatively more balanced, and then as if the noisy data, then noise data is to click data and the not also relatively equilibrium of influence of click data.Click data itself just can provide enough statisticss that it is distinguished with noise data mutually, and anti-noise ability is higher relatively.Even thereby contain a spot of noise sample, can not influence training pattern yet and set up and predict the outcome, the forecast model texture ratio that therefore makes up is more accurate.
After training sample set structure is accomplished, can be model parameter, make up forecast model with the sample data in the said training sample set.Wherein forecast model has multiplely, can select according to real needs, and the application does not do qualification to this.
When set is predicted to test sample book, can dope the clicking rate of user, for example to every kind of advertisement; Clicking rate to the ecommerce series advertisements is 50%; Clicking rate to the network game series advertisements is 20%, is 15% to the clicking rate of recommendation and introduction of Web site series advertisements, and other is 15%.
In sum, the application extracts sample data and makes up the original sample set, and wherein said sample data comprises user's click data and click data not, makes up the training sample set through said original sample set is sampled then.The application does not directly gather the original sample set as training sample; Make up the training sample set but original sample set is optimized; So just can eliminate in the original sample set click data and the serious unbalance problem of ratio between the click data not; Make up the training sample set of relative equilibrium, and the influence that this moment, noise data hit data to said training sample set mid point is less than the influence of noise data to click data in the original sample set.Be that model parameter makes up forecast model with the sample data in the said training sample set again; The sample data of ratio relative equilibrium make forecast model to the discrimination of click data than higher; Utilizing said forecast model that test sample book is gathered predicts; Predictive user is to the clicking rate of every kind of advertisement, and this moment, forecast model can seriously be partial to not click data, had improved the clicking rate prediction accuracy.
With reference to Fig. 2, it has provided the clicking rate Forecasting Methodology process flow diagram of the said a kind of advertisement of the application's preferred embodiment.
For example, when testing, there are a data of throwing in to the clicking rate of advertisement, comprising the impressions of certain advertisement, in these impressions, the number of clicks of respective user and number of clicks not.
Therefore can from throw in data, obtain a certain section data in the time; Statistics is to certain advertisement; User's number of clicks is as click data, and user's not number of clicks is as click data not, with said click data and not click data make up sample set as sample data.
For example, if the clicking rate of prediction advertisement then comprises whether clicking the advertisement of being thrown in the sample data.If the clicking rate of entertainment news in the prediction webpage, then comprise in the sample data to the number of clicks of said entertainment news and number of clicks not.
Add up and throw in a certain period whether the user has clicked the advertisement of being thrown in the data, can count has not click data of how many click datas and what in the sample data.As, in nearest 3 months, to certain advertisement putting 1000 times, possibly to obtain click data be 2 through statistics, click data is not 998.
Certainly also comprise other data in the sample data, the application does not do qualification to this.For example, predict to the clicking rate of advertisement:
At first from the advertisement putting daily record, extract the sample data of the advertisement that needs input; A sample data can comprise the ID ID of this time input or user's IP (Internet Protocol; The agreement that interconnects between the network) url (the Uniform Resource Locator of the release time of address, advertisement, institute's advertisement delivery; URL), reach the user and whether clicked the advertisement of being thrown in.
Make up a sample set, except extracting sample data, also will extract the corresponding sample characteristic, wherein sample characteristics is used to describe sample data, and each dimensional feature is all quantized by a characteristic of correspondence value, can distinguish different samples through eigenwert.
Wherein, A lot of to the method that sample characteristics quantizes through eigenwert, for example, the frequency that adopts sample characteristics to occur is measured; And for example the access characteristic of certain website can use visit capacity in the set time divided by registered user's sum etc., and the application does not do qualification to this.
From the network message daily record, extract and throw in sample data corresponding sample characteristic, wherein sample characteristics comprises with the next item down characteristic at least:
The correlated characteristic of user's correlated characteristic, input url correlated characteristic and advertisement.
Wherein, user's correlated characteristic can comprise: the query word information that historical website of being visited of user or webpage, user use, the advertising message that the user was clicked in the past etc.
Throwing in the url correlated characteristic comprises: content characteristic, anchor text feature, the pairing query word characteristic of url and hyperlink characteristic etc.
The correlated characteristic of advertisement can comprise: advertisement dimension, the landing page of advertisement (Landing Page is called as sometimes and primarily catches user page) correlated characteristic, the speech of bidding, descriptor, advertising sector etc.
For example, in the prediction of the clicking rate of advertisement, sample characteristics is the ecommerce advertisement, can also comprise corresponding link information and bid speech etc.And for example, the clicking rate of entertainment news prediction in the webpage, then sample characteristics can be characteristic speech and content characteristic etc.
The application can be provided with a sampling in advance than P, can sample according to the sampling ratio that presets when original training sample is sampled.Wherein said sampling of presetting makes the two relative original sample set that distributes balanced more than being the click data that draws through statistics and the ratio of click data not.
Definite method of said sampling ratio can comprise:
1, i sampling ratio is set, to each sampling ratio, set is sampled and is made up the training sample set to original sample respectively, makes up forecast model again.The corresponding a prediction index of each forecast model, said prediction index is about the accuracy rate of advertisement putting and the index of recall rate.
Wherein, first sampling is A than corresponding prediction index
1..., i sampling is A than corresponding prediction index
i
2, directly use the original sample set to make up forecast model, the prediction index that this forecast model is corresponding is B.
Sample than corresponding i prediction index A to i
i, with each prediction index A
iCompare with prediction index B respectively, repeatedly repeat said process, the result after the statistical.
Wherein, if prediction index A
iMore than or equal to prediction index B, then be suitable sampling ratio, can be used for original sample set is sampled.Otherwise, being inappropriate sampling ratio, can not be used for original sample set is sampled.
To click data in the original sample set and the serious unbalance problem of ratio between the click data not; The said sampling of choosing of the application can be 1: 2～1: 10 than the span of P, be applicable to click data in the original sample set and not between the click data ratio greater than 1: 10 situation.
After sampling than P according to the sampling of presetting, the click data of sampling out and not click data promptly can be used as sample data and make up corresponding training sample set.
In sampling, can all click datas in the original sample set all be added in the said training sample set, can keep click data to greatest extent this moment in the training sample set.If click data is n in the original sample set, then to hit data be n to training sample set mid point, and click data is not n*P.
For example; Sample data is 1000 in the original sample set, and wherein click data is 2: 998 with the ratio of click data not, and can be according to the sampling of presetting than P=1 when then sampling: 10 sample; All click datas in the raw sample data are added in the training sample set that makes up; Then sample data is 22 in the training sample set, and wherein click data is 2, and click data is not 20.
Kept whole click datas this moment, and improved the very unbalanced problem of sample data distribution in the original sample set.The sampling of wherein presetting draws than through statistics, has objectivity and accuracy.
Sample data with said training sample set is that model parameter makes up forecast model; Select corresponding forecast model according to demand in for example making up; Like BT (behavioral targeting; The user behavior orientation) model, CM (contextual match, content match) model or search trigger model.
For example; Said forecast model is the Bayesian model based on probability estimate; If the sample data with in the original sample set is a model parameter; Click data more after a little while, the accuracy rate of the characteristic that may produce ad click being carried out probability estimate will reduce, and causes the discrimination that possibly produce click data is descended.
And be model parameter with the sample data in the set of said training sample; Click data is with the click data distribution is relatively more not balanced; Can improve the accuracy rate of the characteristic that may produce ad click being carried out probability estimate, and promote producing the discrimination of click data.
With reference to Fig. 3, provided in the said a kind of clicking rate Forecasting Methodology of the application's preferred embodiment based on feature space decision surface synoptic diagram.
And for example, said forecast model is based on the feature space decision surface, and this forecast model purpose is to seek the optimizing decision face that makes structure risk minimum.When training sample is unbalanced; Selected support vector distributes and also can weigh by skewness; In the minimum process of computation structure risk; Model can neglect the influence of click data to structure risk, thereby has enlarged the decision boundary of non-click data, and the actual decision surface and the optimizing decision face that cause model to obtain produce deviation.
Fig. 3 is 2 o'clock for characteristic dimension, click data and non-click data exemplary plot, and wherein circle is represented not click data basis, and box indicating click data, dotted line are the optimizing decision face, and solid line is represented actual decision surface.
Wherein figure (a) is for when sample distribution is unbalanced; Because the existence of noise data; Near click data and click data is not overlapped the optimizing decision face, because click data is not more preponderated on sample size, the actual decision surface (shown in the solid line) that causes model to obtain is partial to not click data.
Figure (b) is after sampling is sampled than the ratio that is 1: 2 for process, the sample data distribution instance of the training sample set of structure.Through sampling, effectively suppressed the influence that click data produces decision surface, the actual decision surface that obtains more approaches the optimizing decision face.
When carrying out the clicking rate prediction, forecast model need be tested the test sample book set, and therefore the clicking rate that just can obtain predicting needs to make up the test sample book set.
Advertisement once throw in corresponding the concrete advertisement putting page, for example, in webpage A, thrown in advertisement, then webpage A is the advertisement putting page.The user has clicked the page of certain website; Can generate and send page request and give the server of said website; If this page is the advertisement putting page, then the server of this website also can send ad-request and give Advertisement Server, and Advertisement Server can be predicted the advertisement that the user possibly click; Therefore gather as samples tested data construct test sample book extracting the user who clicks the input page, also can extract sample data corresponding sample characteristic simultaneously.
Wherein, the sample data in the set of said test sample book can with the sample data basically identical of training sample set, comprise ID ID or user's IP address, the url of the release time of advertisement, institute's advertisement delivery etc.
For example; Said forecast model is the linear decision surface based on feature space; Then can distinguish the weight that each characteristic can produce click data, therefore when using forecast model that test sample book is predicted, just can amass through feature weight on calculating each dimensional feature value of this sample and this dimension is; And to the long-pending summation on all characteristic dimension; Thereby dope comprise some sample characteristics the sample data respective user for possibly click the user of certain advertisement, and dope the user, and the sample data respective user that does not comprise these sample characteristics is not for clicking the user of this advertisement to the clicking rate of said advertisement.
For example, to the network game series advertisements, the user who then once clicked the network game website in the sample characteristics be for possibly click the user of network game series advertisements, and the user who did not click the network game website in the sample characteristics is not for clicking the user of network game series advertisements.
Advertisement that the user that can dope through above-mentioned method possibly click and the clicking rate that is directed against this advertisement; And the advertisement that the user possibly click has multiple; For example, be 50% to the clicking rate of ecommerce series advertisements, be 20% to the clicking rate of network game series advertisements; Clicking rate to the recommendation and introduction of Web site series advertisements is 15%, and other is 15%.
Step 27, the user in the said test sample book set shows the advertisement that clicking rate is the highest to said user in the page.
The above-mentioned clicking rate that dopes the user to every kind of advertisement; Can the clicking rate of all advertisements be sorted; Choose the highest advertisement of clicking rate, when the user opens a page, can show the highest advertisement of said clicking rate; The most possible advertisement of clicking of i.e. this visit behavior of this user, and then can improve the clicking rate of corresponding advertisement.Even what this moment, different user was opened is the same page, the advertisement of demonstration also possibly be different.
For example; The explicit user 1 that predicts the outcome possibly clicked the ecommerce series advertisements, and user 2 possibly click the network game advertisement, therefore when user 1 clicks the homepage of same website with user 2; What user 1 saw is the ecommerce series advertisements, and what user 2 saw is the network game series advertisements.
And for example, the above-mentioned user of doping is 50% to the clicking rate of ecommerce series advertisements, is 20% to the clicking rate of network game series advertisements, is 15% to the clicking rate of recommendation and introduction of Web site series advertisements, and other is 15%.After then the clicking rate of all advertisements being sorted, this user is the highest to the clicking rate of ecommerce series advertisements, therefore can be to user's showing electronic business series advertisements.
To, the clicking rate prediction of advertisement, the main policies of carrying out the accurate advertisement input at present comprises search trigger (sponsored search), content match, user behavior directed (behavioral targeting, BT) several kinds of modes.
Wherein the advertisement of search trigger is to carry out the advertisement retrieval according to the keyword that the user submits to search engine, because keyword has directly reflected the interest that the user is current, so can push the advertisement relevant with the current search content to the user.
Content match then is that the content of the webpage of user's browsing is carried out modeling analysis, shows the advertisement close with web page contents to the user.
The user behavior orientation can be at the historical behavior record according to the user; Search history, web page browsing historical record, advertisement displaying and click record etc. like the user; Modeling and prediction are carried out in interest and behavior to the user, choose the advertisement that meets this user interest and show.
Can select corresponding strategy to set up forecast model according to demand, carry out the clicking rate prediction.
With reference to Fig. 4, provided the clicking rate Forecasting Methodology synoptic diagram of the said a kind of advertisement of the application's preferred embodiment.
The sample data of the click data of advertisement distributes seriously unbalance in the ad click rate prediction.In real work, in order to guarantee the group sample data (click data) of q.s, this unbalance one side causes the training sample data sharply to expand, and increases to calculate reality and storage resources.On the other hand, sample data is unbalance may have a negative impact to the training performance of model.
Based on this observation, this patent provides a kind of sampling policy of owing, under the unbalance situation of this sample data; Reduce multiclass sample data quantity on the one hand; Thereby reduce training sample set scale, save needed storage space and computational resource, training effectiveness is provided; On the other hand, because sample data distribution relative equilibrium can effectively be avoided sample set and the negative effect that itself model performance is produced, improve prediction effect.
Discuss the clicking rate Forecasting Methodology of advertisement below.
To the prediction of the clicking rate of advertisement, at first can from the network message daily record, extract sample characteristics, and from the advertisement putting daily record, extract the sample of advertisement putting, the sample of said advertisement putting comprises click data and click data not.Make up the original sample set according to said sample and sample characteristics then, click data is with the click data distribution is very not unbalanced in the common said original sample set.Therefore said original sample set is sampled, make up the training sample set, said training sample set mid point hits the relatively equilibrium that distributes of data and click data.According to being combined into model parameter, make up forecast model then, use forecast model that test sample book is gathered and carry out the clicking rate test with said training sample set.
In sum; The described sampling of the application is than the pairing click data and the ratio of click data not; Be to draw according to the statistics of repeatedly testing; Have statistical accuracy and objectivity, the data during the training sample that therefore makes up than sampling to sampling is gathered also have accuracy and objectivity, have further improved the clicking rate prediction accuracy.
Secondly because click data and not between the click data ratio seriously unbalance, the data in the training sample set are many accordingly usually, so the data of execution model training are many, system burden is bigger, processing speed is slower, influences the efficient of model training.The application can compare said original sample set with the sampling of presetting and sample; In sampling, all click datas are all added in the said training sample set, guaranteeing to have reduced the data in the training sample set under the constant situation of click data bulk; Make that the data of execution model training structure forecast model are fewer; Reduce the burden of system, accelerated processing speed of data, improved the efficient of model training.
Once more, the application than being the foundation of quantity of sampling quantity, therefore when making up the test sample book set, can be chosen at the sample that meets said set characteristic most with the sampling of presetting, and improves the accuracy and the specific aim of sample, has further improved the accuracy of test.
With reference to Fig. 5, provided the said a kind of clicking rate prognoses system structural drawing of the application embodiment.
Accordingly, the application also provides a kind of prognoses system of clicking rate, and each module in this system all can be by computer realization.Described system comprises: makes up original sample collection modules 11, makes up training sample collection modules 12, makes up forecast model module 13 and clicking rate prediction module 15, wherein:
Make up original sample collection modules 11, be used to extract sample data and make up the original sample set, wherein said sample data comprises user's click data and click data not;
Make up training sample collection modules 12, be used for making up the training sample set through said original sample set is sampled;
Make up forecast model module 13, the sample data that is used for said training sample set is that model parameter makes up forecast model;
Clicking rate prediction module 15 is used to utilize said forecast model that test sample book is gathered and predicts, dopes the clicking rate of user to every kind of advertisement.
Preferably, said structure original sample collection modules 11 is used for making up the original sample set from throwing in the data of extracting data in the period as sample data; And extracting each sample data corresponding sample characteristic, said sample characteristics is used to describe sample data.
Wherein, as click data, the conduct of user's not number of clicks is click data not with the number of clicks of throwing in user in the data.
Preferably, said structure training sample collection modules 12 be used for sampling with the sampling comparison original training sample that presets, and structure is gathered than corresponding training sample with said sampling.
Wherein, said sampling of presetting is than being the click data that draws through statistics and the ratio of click data not.In the sampling, all click datas are all added in the said training sample set.
Preferably, described system also comprises:
Make up test sample book collection modules 14, be used for, extract the user who clicks the advertisement putting page and gather as samples tested data construct test sample book to the advertisement putting page.
Display module 16 is used for the user to said test sample book set, in the page, shows the advertisement that clicking rate is the highest to said user.
For system embodiment, because it is similar basically with method embodiment, so description is fairly simple, relevant part gets final product referring to the part explanation of method embodiment.
Each embodiment in this instructions all adopts the mode of going forward one by one to describe, and what each embodiment stressed all is and the difference of other embodiment that identical similar part is mutually referring to getting final product between each embodiment.
Although described the application's preferred embodiment, in a single day those skilled in the art get the basic inventive concept could of cicada, then can make other change and modification to these embodiment.So accompanying claims is intended to be interpreted as all changes and the modification that comprises preferred embodiment and fall into the application's scope.
Those skilled in the art should understand that the application's embodiment can be provided as method, system or computer program.Therefore, the application can adopt the form of the embodiment of complete hardware embodiment, complete software implementation example or combination software and hardware aspect.And the application can be employed in the form that one or more computer-usable storage medium (including but not limited to magnetic disk memory, CD-ROM, optical memory etc.) that wherein include computer usable program code go up the computer program of implementing.
The application is that reference is described according to the process flow diagram and/or the block scheme of method, equipment (system) and the computer program of the application embodiment.Should understand can be by the flow process in each flow process in computer program instructions realization flow figure and/or the block scheme and/or square frame and process flow diagram and/or the block scheme and/or the combination of square frame.Can provide these computer program instructions to the processor of multi-purpose computer, special purpose computer, Embedded Processor or other programmable data processing device to produce a machine, make the instruction of carrying out through the processor of computing machine or other programmable data processing device produce to be used for the device of the function that is implemented in flow process of process flow diagram or a plurality of flow process and/or square frame of block scheme or a plurality of square frame appointments.
These computer program instructions also can be stored in ability vectoring computer or the computer-readable memory of other programmable data processing device with ad hoc fashion work; Make the instruction that is stored in this computer-readable memory produce the manufacture that comprises command device, this command device is implemented in the function of appointment in flow process of process flow diagram or a plurality of flow process and/or square frame of block scheme or a plurality of square frame.
These computer program instructions also can be loaded on computing machine or other programmable data processing device; Make on computing machine or other programmable devices and to carry out the sequence of operations step producing computer implemented processing, thereby the instruction of on computing machine or other programmable devices, carrying out is provided for being implemented in the step of the function of appointment in flow process of process flow diagram or a plurality of flow process and/or square frame of block scheme or a plurality of square frame.
The application can describe in the general context of the computer executable instructions of being carried out by computing machine, for example program module.Usually, program module comprises the routine carrying out particular task or realize particular abstract, program, object, assembly, data structure or the like.Also can in DCE, put into practice the application, in these DCEs, by through communication network connected teleprocessing equipment execute the task.In DCE, program module can be arranged in this locality and the remote computer storage medium that comprises memory device.
At last; Also need to prove; In this article; Relational terms such as first and second grades only is used for an entity or operation are made a distinction with another entity or operation, and not necessarily requires or hint relation or the order that has any this reality between these entities or the operation.And; Term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability; Thereby make and comprise that process, method, commodity or the equipment of a series of key elements not only comprise those key elements; But also comprise other key elements of clearly not listing, or also be included as this process, method, commodity or equipment intrinsic key element.Under the situation that do not having much more more restrictions, the key element that limits by statement " comprising ... ", and be not precluded within process, method, commodity or the equipment that comprises said key element and also have other identical element.
More than to the clicking rate Forecasting Methodology and the system of a kind of advertisement that the application provided; Carried out detailed introduction; Used concrete example among this paper the application's principle and embodiment are set forth, the explanation of above embodiment just is used to help to understand the application's method and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to the application's thought, the part that on embodiment and range of application, all can change, in sum, this description should not be construed as the restriction to the application.
Claims (11)
1. the clicking rate Forecasting Methodology of an advertisement is characterized in that, comprising:
Extract sample data and make up the original sample set, wherein said sample data comprises user's click data and click data not;
Through said original sample set is sampled, make up the training sample set;
With the sample data in the said training sample set is that model parameter makes up forecast model;
Utilize said forecast model that test sample book is gathered and predict, dope the clicking rate of user to every kind of advertisement.
2. method according to claim 1 is characterized in that, and is said through structure training sample set that said original sample set is sampled, and comprising:
Sample with the sampling of presetting comparison original training sample, and structure gathers than corresponding training sample with said sampling, wherein, said sampling of presetting is than being the click data that draws through statistics and the ratio of click data not.
3. method according to claim 1 is characterized in that, through said original sample set is sampled, makes up the training sample set, comprising:
In the sampling, all click datas are all added in the said training sample set.
4. method according to claim 1 is characterized in that, said extraction sample data makes up the original sample set, comprising:
Make up the original sample set from throwing in the data of extracting data in the period as sample data;
And extracting each sample data corresponding sample characteristic in the original sample set, said sample characteristics is used to describe sample data;
Wherein, as click data, the conduct of user's not number of clicks is click data not with the number of clicks of throwing in user in the data.
5. method according to claim 1 is characterized in that, the test sample book set is:
To the advertisement putting page, the user who extract to click the advertisement putting page as the samples tested data after, the test sample book set of structure.
6. method according to claim 5 is characterized in that, also comprises:
User in the said test sample book set shows the advertisement that clicking rate is the highest to said user in the page.
7. the clicking rate prognoses system of an advertisement is characterized in that, comprising:
Make up the original sample collection modules, be used to extract sample data and make up the original sample set, wherein said sample data comprises user's click data and click data not;
Make up the training sample collection modules, be used for making up the training sample set through said original sample set is sampled;
Make up the forecast model module, the sample data that is used for said training sample set is that model parameter makes up forecast model;
The clicking rate prediction module is used to utilize said forecast model that test sample book is gathered and predicts, dopes the clicking rate of user to every kind of advertisement.
8. system according to claim 7 is characterized in that, comprising:
Said structure training sample collection modules; Be used for sampling with the sampling comparison original training sample that presets; And structure gathers than corresponding training sample with said sampling, and wherein, said sampling of presetting is than being the click data that draws through statistics and the ratio of click data not.
9. system according to claim 7 is characterized in that, comprising:
Said structure original sample collection modules is used for making up the original sample set from throwing in the data of extracting data in the period as sample data; And extracting each sample data corresponding sample characteristic, said sample characteristics is used to describe sample data; Wherein, as click data, the conduct of user's not number of clicks is click data not with the number of clicks of throwing in user in the data.
10. according to the said system of claim 9, it is characterized in that, also comprise:
Make up the test sample book collection modules, be used for, extract the user who clicks the advertisement putting page and gather as samples tested data construct test sample book to the advertisement putting page.
11. according to the said system of claim 10, it is characterized in that, also comprise:
Display module is used for the user to said test sample book set, in the page, shows the advertisement that clicking rate is the highest to said user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210074541XA CN102663617A (en) | 2012-03-20 | 2012-03-20 | Method and system for prediction of advertisement clicking rate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210074541XA CN102663617A (en) | 2012-03-20 | 2012-03-20 | Method and system for prediction of advertisement clicking rate |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102663617A true CN102663617A (en) | 2012-09-12 |
Family
ID=46773097
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210074541XA Pending CN102663617A (en) | 2012-03-20 | 2012-03-20 | Method and system for prediction of advertisement clicking rate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102663617A (en) |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102880688A (en) * | 2012-09-14 | 2013-01-16 | 北京百度网讯科技有限公司 | Method, device and equipment for evaluating webpage |
| CN103235893A (en) * | 2013-05-06 | 2013-08-07 | 重庆大学 | User-commodity click rate self-adaptive prediction device and method |
| CN103246985A (en) * | 2013-04-26 | 2013-08-14 | 北京亿赞普网络技术有限公司 | Advertisement click rate predicting method and device |
| CN103310003A (en) * | 2013-06-28 | 2013-09-18 | 华东师范大学 | Method and system for predicting click rate of new advertisement based on click log |
| CN103345512A (en) * | 2013-07-06 | 2013-10-09 | 北京品友互动信息技术有限公司 | Online advertising click-through rate forecasting method and device based on user attribute |
| CN103746898A (en) * | 2013-12-25 | 2014-04-23 | 新浪网技术(中国)有限公司 | Sampling analysis based e-mail sending method and system |
| CN103853711A (en) * | 2012-11-28 | 2014-06-11 | 中国移动通信集团广西有限公司 | Text information processing method and device |
| CN103914475A (en) * | 2013-01-05 | 2014-07-09 | 腾讯科技（北京）有限公司 | Method, system and device for predicting video views |
| CN104090919A (en) * | 2014-06-16 | 2014-10-08 | 华为技术有限公司 | Advertisement recommending method and advertisement recommending server |
| CN104268644A (en) * | 2014-09-23 | 2015-01-07 | 新浪网技术(中国)有限公司 | Method and device for predicting click frequency of advertisement at advertising position |
| 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 |
| CN104778608A (en) * | 2015-04-13 | 2015-07-15 | 合一信息技术（北京）有限公司 | N+ advertisement putting and optimizing method |
| CN104933075A (en) * | 2014-03-20 | 2015-09-23 | 百度在线网络技术（北京）有限公司 | User attribute predicting platform and method |
| CN104951965A (en) * | 2015-06-26 | 2015-09-30 | 深圳市腾讯计算机系统有限公司 | Advertisement delivery method and device |
| CN105095625A (en) * | 2014-05-14 | 2015-11-25 | 阿里巴巴集团控股有限公司 | Click Through Ratio (CTR) prediction model establishing method and device, information providing method and information providing system |
| CN105654200A (en) * | 2015-12-30 | 2016-06-08 | 上海珍岛信息技术有限公司 | Deep learning-based advertisement click-through rate prediction method and device |
| CN105915438A (en) * | 2016-04-15 | 2016-08-31 | 北京奇虎科技有限公司 | Message pushing method, apparatus, and system |
| CN106227743A (en) * | 2016-07-12 | 2016-12-14 | 精硕世纪科技（北京）有限公司 | Advertisement target group touches and reaches ratio estimation method and device |
| CN106886915A (en) * | 2017-01-17 | 2017-06-23 | 华南理工大学 | A kind of ad click predictor method based on time decay sampling |
| WO2017107571A1 (en) * | 2015-12-24 | 2017-06-29 | 北京大学 | Method and system for determining quality of application on basis of user behaviors of application management |
| CN107124320A (en) * | 2017-06-30 | 2017-09-01 | 北京金山安全软件有限公司 | Traffic data monitoring method and device and server |
| CN107346464A (en) * | 2016-05-06 | 2017-11-14 | 腾讯科技（深圳）有限公司 | Operational indicator Forecasting Methodology and device |
| CN107369052A (en) * | 2017-08-29 | 2017-11-21 | 北京小度信息科技有限公司 | User's registration behavior prediction method, apparatus and electronic equipment |
| CN107463580A (en) * | 2016-06-06 | 2017-12-12 | 腾讯科技（深圳）有限公司 | Train clicking rate prediction model method and apparatus, clicking rate predictor method and device |
| CN107527243A (en) * | 2017-08-29 | 2017-12-29 | 有米科技股份有限公司 | A kind of method that prediction APP rushes spending on ads needed for list |
| CN108921604A (en) * | 2018-06-22 | 2018-11-30 | 华南理工大学 | A kind of ad click rate prediction technique integrated based on Cost-Sensitive Classifiers |
| CN108965951A (en) * | 2018-07-12 | 2018-12-07 | 北京奇艺世纪科技有限公司 | The playback method and device of advertisement |
| CN109408724A (en) * | 2018-11-06 | 2019-03-01 | 北京达佳互联信息技术有限公司 | Multimedia resource estimates the determination method, apparatus and server of clicking rate |
| CN110188942A (en) * | 2019-05-27 | 2019-08-30 | 北京金山安全软件有限公司 | Click through rate prediction method, device and equipment |
| CN112907284A (en) * | 2021-02-09 | 2021-06-04 | 上海明略人工智能(集团)有限公司 | Method and device for constructing target feature combination |
| 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 |
| CN113706228A (en) * | 2021-11-01 | 2021-11-26 | 腾讯科技（深圳）有限公司 | Multimedia information playing control method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101203875A (en) * | 2005-03-30 | 2008-06-18 | 谷歌公司 | Adjusting an advertising cost, such as a per-ad impression cost, using a likelihood that the ad will be sensed or perceived by users |
| CN101385018A (en) * | 2005-12-30 | 2009-03-11 | 谷歌公司 | Using estimated ad qualities for ad filtering, ranking and promotion |
| CN101390118A (en) * | 2005-12-30 | 2009-03-18 | 谷歌公司 | Predicting ad quality |
| CN102110265A (en) * | 2009-12-23 | 2011-06-29 | 深圳市腾讯计算机系统有限公司 | Network advertisement effect estimating method and network advertisement effect estimating system |
| 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 |
-
2012
- 2012-03-20 CN CN201210074541XA patent/CN102663617A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101203875A (en) * | 2005-03-30 | 2008-06-18 | 谷歌公司 | Adjusting an advertising cost, such as a per-ad impression cost, using a likelihood that the ad will be sensed or perceived by users |
| CN101385018A (en) * | 2005-12-30 | 2009-03-11 | 谷歌公司 | Using estimated ad qualities for ad filtering, ranking and promotion |
| CN101390118A (en) * | 2005-12-30 | 2009-03-18 | 谷歌公司 | Predicting ad quality |
| CN102110265A (en) * | 2009-12-23 | 2011-06-29 | 深圳市腾讯计算机系统有限公司 | Network advertisement effect estimating method and network advertisement effect estimating system |
| 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 |
Cited By (47)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102880688B (en) * | 2012-09-14 | 2016-07-27 | 北京百度网讯科技有限公司 | A kind of method for webpage is estimated, device and equipment |
| CN102880688A (en) * | 2012-09-14 | 2013-01-16 | 北京百度网讯科技有限公司 | Method, device and equipment for evaluating webpage |
| CN103853711A (en) * | 2012-11-28 | 2014-06-11 | 中国移动通信集团广西有限公司 | Text information processing method and device |
| CN103853711B (en) * | 2012-11-28 | 2017-02-08 | 中国移动通信集团广西有限公司 | Text information processing method and device |
| CN103914475A (en) * | 2013-01-05 | 2014-07-09 | 腾讯科技（北京）有限公司 | Method, system and device for predicting video views |
| CN103914475B (en) * | 2013-01-05 | 2018-05-04 | 腾讯科技（北京）有限公司 | A kind of Forecasting Methodology, system and the device of video playing amount |
| CN103246985B (en) * | 2013-04-26 | 2016-12-28 | 北京亿赞普网络技术有限公司 | A kind of ad click rate Forecasting Methodology and device |
| CN103246985A (en) * | 2013-04-26 | 2013-08-14 | 北京亿赞普网络技术有限公司 | Advertisement click rate predicting method and device |
| CN103235893B (en) * | 2013-05-06 | 2016-03-23 | 重庆大学 | A kind of user-commodity clicking rate adaptive prediction device and Forecasting Methodology |
| CN103235893A (en) * | 2013-05-06 | 2013-08-07 | 重庆大学 | User-commodity click rate self-adaptive prediction device and method |
| CN103310003A (en) * | 2013-06-28 | 2013-09-18 | 华东师范大学 | Method and system for predicting click rate of new advertisement based on click log |
| CN103345512A (en) * | 2013-07-06 | 2013-10-09 | 北京品友互动信息技术有限公司 | Online advertising click-through rate forecasting method and device based on user attribute |
| CN103746898A (en) * | 2013-12-25 | 2014-04-23 | 新浪网技术(中国)有限公司 | Sampling analysis based e-mail sending method and system |
| CN104933075A (en) * | 2014-03-20 | 2015-09-23 | 百度在线网络技术（北京）有限公司 | User attribute predicting platform and method |
| CN105095625A (en) * | 2014-05-14 | 2015-11-25 | 阿里巴巴集团控股有限公司 | Click Through Ratio (CTR) prediction model establishing method and device, information providing method and information providing system |
| WO2015192667A1 (en) * | 2014-06-16 | 2015-12-23 | 华为技术有限公司 | Advertisement recommending method and advertisement recommending server |
| CN104090919B (en) * | 2014-06-16 | 2017-04-19 | 华为技术有限公司 | Advertisement recommending method and advertisement recommending server |
| CN104090919A (en) * | 2014-06-16 | 2014-10-08 | 华为技术有限公司 | Advertisement recommending method and advertisement recommending server |
| CN104268644A (en) * | 2014-09-23 | 2015-01-07 | 新浪网技术(中国)有限公司 | Method and device for predicting click frequency of advertisement at advertising position |
| 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 |
| CN104778608A (en) * | 2015-04-13 | 2015-07-15 | 合一信息技术（北京）有限公司 | N+ advertisement putting and optimizing method |
| CN104951965A (en) * | 2015-06-26 | 2015-09-30 | 深圳市腾讯计算机系统有限公司 | Advertisement delivery method and device |
| WO2017107571A1 (en) * | 2015-12-24 | 2017-06-29 | 北京大学 | Method and system for determining quality of application on basis of user behaviors of application management |
| CN105654200A (en) * | 2015-12-30 | 2016-06-08 | 上海珍岛信息技术有限公司 | Deep learning-based advertisement click-through rate prediction method and device |
| CN105915438B (en) * | 2016-04-15 | 2019-02-19 | 北京奇虎科技有限公司 | Information push method, apparatus and system |
| CN105915438A (en) * | 2016-04-15 | 2016-08-31 | 北京奇虎科技有限公司 | Message pushing method, apparatus, and system |
| CN107346464B (en) * | 2016-05-06 | 2021-04-16 | 腾讯科技（深圳）有限公司 | Service index prediction method and device |
| CN107346464A (en) * | 2016-05-06 | 2017-11-14 | 腾讯科技（深圳）有限公司 | Operational indicator Forecasting Methodology and device |
| CN107463580B (en) * | 2016-06-06 | 2020-07-24 | 腾讯科技（深圳）有限公司 | Click rate estimation model training method and device and click rate estimation method and device |
| CN107463580A (en) * | 2016-06-06 | 2017-12-12 | 腾讯科技（深圳）有限公司 | Train clicking rate prediction model method and apparatus, clicking rate predictor method and device |
| CN106227743A (en) * | 2016-07-12 | 2016-12-14 | 精硕世纪科技（北京）有限公司 | Advertisement target group touches and reaches ratio estimation method and device |
| CN106227743B (en) * | 2016-07-12 | 2019-09-24 | 精硕科技（北京）股份有限公司 | Advertisement target group touching reaches ratio estimation method and device |
| CN106886915B (en) * | 2017-01-17 | 2020-07-28 | 华南理工大学 | Advertisement click estimation method based on time attenuation sampling |
| CN106886915A (en) * | 2017-01-17 | 2017-06-23 | 华南理工大学 | A kind of ad click predictor method based on time decay sampling |
| CN107124320A (en) * | 2017-06-30 | 2017-09-01 | 北京金山安全软件有限公司 | Traffic data monitoring method and device and server |
| CN107527243A (en) * | 2017-08-29 | 2017-12-29 | 有米科技股份有限公司 | A kind of method that prediction APP rushes spending on ads needed for list |
| CN107369052A (en) * | 2017-08-29 | 2017-11-21 | 北京小度信息科技有限公司 | User's registration behavior prediction method, apparatus and electronic equipment |
| CN108921604B (en) * | 2018-06-22 | 2022-03-29 | 华南理工大学 | Advertisement click rate prediction method based on cost-sensitive classifier integration |
| CN108921604A (en) * | 2018-06-22 | 2018-11-30 | 华南理工大学 | A kind of ad click rate prediction technique integrated based on Cost-Sensitive Classifiers |
| CN108965951A (en) * | 2018-07-12 | 2018-12-07 | 北京奇艺世纪科技有限公司 | The playback method and device of advertisement |
| CN108965951B (en) * | 2018-07-12 | 2020-10-02 | 北京奇艺世纪科技有限公司 | Advertisement playing method and device |
| CN109408724A (en) * | 2018-11-06 | 2019-03-01 | 北京达佳互联信息技术有限公司 | Multimedia resource estimates the determination method, apparatus and server of clicking rate |
| CN110188942A (en) * | 2019-05-27 | 2019-08-30 | 北京金山安全软件有限公司 | Click through rate prediction method, device and equipment |
| 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 |
| CN112907284A (en) * | 2021-02-09 | 2021-06-04 | 上海明略人工智能(集团)有限公司 | Method and device for constructing target feature combination |
| CN113706228A (en) * | 2021-11-01 | 2021-11-26 | 腾讯科技（深圳）有限公司 | Multimedia information playing control method and device, electronic equipment and storage medium |
| CN113706228B (en) * | 2021-11-01 | 2022-02-08 | 腾讯科技（深圳）有限公司 | Multimedia information playing control method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102663617A (en) | Method and system for prediction of advertisement clicking rate | |
| US11100178B2 (en) | Method and device for pushing information | |
| CN102609474B (en) | A kind of visit information supplying method and system | |
| CN103886090B (en) | Content recommendation method and device based on user preferences | |
| AU2006332534B2 (en) | Predicting ad quality | |
| CN102760124B (en) | Pushing method and system for recommended data | |
| CN103914468B (en) | A kind of method and apparatus of impression information search | |
| CN103221951B (en) | Predictive query suggestion caching | |
| CN103797474B (en) | The method, apparatus and system of the data related to conversion pathway are provided | |
| US8615514B1 (en) | Evaluating website properties by partitioning user feedback | |
| US8856146B2 (en) | Device for determining internet activity | |
| US20110119267A1 (en) | Method and system for processing web activity data | |
| WO2015175384A1 (en) | Query categorizer | |
| CN102262661B (en) | Web page access forecasting method based on k-order hybrid Markov model | |
| US8886650B2 (en) | Algorithmically choosing when to use branded content versus aggregated content | |
| Zhong et al. | Time-aware service recommendation for mashup creation in an evolving service ecosystem | |
| CN106251174A (en) | Information recommendation method and device | |
| CN102932206B (en) | The method and system of monitoring website access information | |
| WO2019161027A1 (en) | System and method for bot detection | |
| CN104216881A (en) | Method and device for recommending individual labels | |
| CN104462293A (en) | Search processing method and method and device for generating search result ranking model | |
| EP1969486A2 (en) | Using estimated ad qualities for ad filtering, ranking and promotion | |
| US20110054999A1 (en) | System and method for predicting user navigation within sponsored search advertisements | |
| CN103150663A (en) | Method and device for placing network placement data | |
| CN104899229A (en) | Swarm intelligence based behavior clustering system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20120912 |