CN105046532A - Bidding method and device - Google Patents
Bidding method and device Download PDFInfo
- Publication number
- CN105046532A CN105046532A CN201510484829.8A CN201510484829A CN105046532A CN 105046532 A CN105046532 A CN 105046532A CN 201510484829 A CN201510484829 A CN 201510484829A CN 105046532 A CN105046532 A CN 105046532A
- Authority
- CN
- China
- Prior art keywords
- model
- ctr
- estimation
- module
- bidding
- 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 aims to provide a bidding method and device. The method comprises the following steps that: a model establishing module establishes at least one model of a CTR estimation module, a CTR calibration module, a media quality estimation module, a CVR estimation module and a probability blacklist-whitelist model; a model fusing module fuses the established models; a traffic value estimation module acquires the history data and real-time data of advertisement putting relative to a given KPI (Key Performance Indicator) parameter, and estimates the traffic value of an advertisement to be put based on the fused models; and a biding module determines a proper biding strategy for bidding based on the estimated traffic value. Compared with the prior art, the bidding method and device have the advantages that corresponding parameter estimation models are fused, so that the phenomenon of over-fitting in the statistics of spare data can be avoided; the spare data is analyzed effectively; a proper biding strategy is determined specific to KPIs set by different advertisers; and real-time biding can be better performed.
Description
Technical field
The present invention relates to Internet technology, particularly relate to a kind of price competing method and device.
Background technology
Real time bid (RTB) is that pattern is thrown in a kind of emerging Internet advertising, the input mode of advertisement is made to change the fine granularity purchasing model by user into from the coarseness purchasing model of media, advertisement position, make advertiser in the suitable time, suitable advertising message can be passed to suitable user, thus promote advertisement putting efficiency.In RTB pattern, media, advertisement transaction platform (AdExchange), party in request's platform (DemandSidePlatform, DSP), advertiser are wherein important members.Wherein, the advertisement exposure of advertisement transaction platform aggregating media shows opportunity information, and described information is passed to many DSP, DSP then collects the advertisement putting demand of advertiser, advertisement exposure scene information, user profile and advertising message that collective advertising transaction platform provides, the value of advertisement exposure chance is each time estimated, selects suitable advertisement and price to bid, with unexpectedly chance for exposure.Task due to DSP platform continues to optimize the input effect of advertiser, namely corresponding K PI (KeyPerformanceIndicator) value set by different advertiser is continued to optimize, and for DSP platform, bidding algorithm is most important, and this algorithm needs integrated use crowd orientation analysis, clicking rate and conversion ratio estimation, cost budgeting etc. to continue to optimize described KPI (KeyPerformanceIndicator) value.Usually, under RTB pattern, the form of different KPI value set by advertiser is diversity, such as CTR (clicking rate), CPC (each pay-per-click), CPA (every cost of activities), ROI (rate of return on investment) and arrival rate, two jumping rates etc.For adapting to the KPI value set by different advertiser, selecting suitable advertisement and price to bid, needing the bidding algorithm of DSP to have sizable dirigibility.
And existing DSP bidding algorithm mainly use for reference search advertisements, alliance's advertisement clicking rate (CTR), click conversion ratio (CVR) carry out estimating computing and bid, the algorithm adopted is mainly logistic regression and mutation thereof, but these algorithms are not enough for the KPI adaptability of different advertiser, the KPI limited types that can optimize, usually can only be optimized CTR, CVR parameter, and, parameter for any one KPI is all calculated by Independent modeling, and the model independently set up can not merge usually.Particularly, when data being added up in existing bidding algorithm, easily over-fitting is occurred to sparse data, such as, impression and the few dimension CTR of clicks are carried out adding up there will be too high or too low value; And, existing bidding algorithm to unknown flow rate exploring ability difference (such as, may always can not by advertisement putting to website corresponding to this domain name for the domain name lacking data), easily make non-white and black simple judgement, be difficult to incorporate the experimental knowledge etc. of optimization personnel.
Summary of the invention
The object of this invention is to provide a kind of price competing method and device, by the multiple model of conbined usage, the advertiser meeting setting different K PI value carries out real time bid better.
According to an aspect of the present invention, provide a kind of price competing method, the method comprises:
Set up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model;
Merge the model set up;
For given KPI parameter, obtain historical data and the real time data of advertisement putting, based on merged model, ad traffic to be put is worth and estimates;
Based on the Flow Value that described estimation obtains, determine that suitable bidding strategy is bidded.
Wherein, the CTR estimation model set up is for estimating clicking rate based on Bayes statistical method.
Wherein, the overall corresponding relation of the CTR calibrating patterns set up for adopting Log linear regression algorithm to catch CTR value and the true CTR value obtained based on described CTR estimation model.Further, described CTR calibrating patterns also for: based on Log linear regression algorithm, use the method for local regression to catch local nonlinearity feature, to make Log linear regression can describe the corresponding relation predicted between CTR and actual CTR more accurately, and make described recurrence be monotone increasing.
Wherein, the media quality assessment models set up, for carrying out comprehensive grading based on several data to the quality of media.Further, the quality score value that obtains of described comprehensive grading is continuous.
Wherein, the probability black and white lists model set up, for setting a black and white regulation coefficient for each list, to distinguish the probability that described each list is blacklist or white list.
Wherein, in described bidding, at least bid based on following two factors: the difference of the difference of the actual CPC of different prediction CTR section, the CPM (advertisement bar often shows the expense of 1000 times) of different time sections, CPC, CPA and average competing yield.
Further, when described specific KPI is CPA, described price competing method specifically comprises:
At least set up CTR estimation model and CVR estimation model;
Chain rule based on probability merges the CTR model and CVR model set up.
According to another aspect of the present invention, additionally provide one and to bid device, this device comprises:
Model building module, for setting up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model, and sets up bid model;
Model Fusion module, for merging set up model;
Flow Value estimation module, for given KPI parameter, for obtaining historical data and the real time data of advertisement putting, based on merged model, being worth ad traffic to be put and estimating;
To bid module, based on the Flow Value that described estimation obtains, determine that suitable bidding strategy is bidded.
Wherein, the CTR estimation model set up is for estimating clicking rate based on Bayes statistical method.
Wherein, the overall corresponding relation of the CTR calibrating patterns set up for adopting Log linear regression algorithm to catch CTR value and the true CTR value obtained based on described CTR estimation model.Further, described CTR calibrating patterns also for: based on Log linear regression algorithm, use the method for local regression to catch local nonlinearity feature, and make described recurrence be monotone increasing.
Wherein, the media quality assessment models set up, for carrying out comprehensive grading based on several data to the quality of media.Further, the quality score value that obtains of described comprehensive grading is continuous.
Wherein, the probability black and white lists model set up, for setting a black and white regulation coefficient for each list, to distinguish the probability that described each list is blacklist or white list.
Wherein, module of bidding described in is at least bid based on following two factors: the difference of the actual CPC of different prediction CTR section, CPM, CPC, CPA of different time sections and the difference of average competing yield.
Further, when described specific KPI is CPA, bid in device described:
Described model building module at least sets up CTR estimation model and CVR estimation model;
Described Model Fusion module merges based on the chain rule of probability the CTR model and CVR model set up.
Compared with prior art, price competing method based on Bayes statistical method provided by the invention, by conbined usage and the parameterized model unlatching of multi-model, meet the diversified KPI demand of different advertiser, apply Bayes statistical method to solve traditional statistical method and be difficult to the over-fitting problem avoided simultaneously, and improve the non-white and black simple judgement present situation that prior art carries out result.Further, the present invention can also incorporate artificial experience and optimizes further this algorithm.
Accompanying drawing explanation
By reading the detailed description done non-limiting example done with reference to the following drawings, other features, objects and advantages of the present invention will become more obvious:
Fig. 1 illustrates and to bid device schematic diagram according to the one of one aspect of the invention;
Fig. 2 illustrates a kind of price competing method process flow diagram according to a further aspect of the present invention;
In accompanying drawing, same or analogous Reference numeral represents same or analogous parts.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described in further detail.
Before described, following explanation is carried out to the relational language that the application relates to:
CTR: clicking rate;
CVR: click conversion ratio;
KPI:KeyPerformanceIndicator KPI Key Performance Indicator;
CPC: each pay-per-click;
CPM: advertisement bar often shows the expense of 1000 times;
CPA: every cost of activities;
ROI: rate of return on investment;
Fig. 1 illustrates and to bid device schematic diagram according to the one of one aspect of the invention; Wherein, device of bidding described in comprises model building module 1, Model Fusion module 2, Flow Value estimation module 3, module 4 of bidding.Particularly, model building module 1 sets up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model; Model Fusion module 2 merges the model set up; Flow Value estimation module 3, for given KPI parameter, obtains historical data and the real time data of advertisement putting, based on merged model, is worth estimates ad traffic to be put; The Flow Value that module of bidding 4 obtains based on described estimation, determines that suitable bidding strategy is bidded.
At this, described in device of bidding include but not limited to that network equipment, user's set or network equipment and user's set are by the mutually integrated device formed of network.Wherein, described network equipment comprise a kind of can according in advance setting or the instruction stored, automatically carry out the electronic equipment of numerical evaluation and information processing, its hardware includes but not limited to microprocessor, special IC, programmable gate array, digital processing unit, embedded device etc.Described network equipment includes but not limited to the cloud that computing machine, network host, single network server, multiple webserver collection or multiple server are formed.Described user's set includes but not limited to that any one can to carry out the electronic product of man-machine interaction, such as computing machine, smart mobile phone, PAD etc. with user by keyboard, telepilot, touch pad or voice-operated device.Described network includes but not limited to internet, wide area network, Metropolitan Area Network (MAN), LAN (Local Area Network) etc.Those skilled in the art will be understood that other device of bidding is equally applicable to the present invention, within also should being included in scope, and are contained in this at this with way of reference.
Constant work between above-mentioned each module, at this, it will be understood by those skilled in the art that " continuing " refer to above-mentioned each device respectively in real time or according to setting or the mode of operation requirement that adjusts in real time, carry out work.
Model building module 1 sets up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model, and sets up bid model.
Wherein, the CTR estimation model set up is for the clicking rate based on Bayes statistical method estimated flow, and Bayes is carried out smoothly to the data of each dimensions such as domain name, advertisement position, operating system, browser, thus solve the problem of Deta sparseness and over-fitting.
Wherein, the CTR calibrating patterns set up is for calibrating the clicking rate value estimated by described CTR estimation model, particularly, Log linear regression algorithm is adopted to catch the overall corresponding relation of CTR value and the true CTR value obtained based on described CTR estimation model, and adopt true CTR value to realize the calibration to the clicking rate value estimated by described CTR estimation model as feedback, make the CTR value after calibrating and true CTR value closely, promote the accuracy rate that follow-up corresponding model carries out the estimation of Flow Value.Further, described CTR calibrating patterns also for: based on Log linear regression algorithm, use the method for local regression to catch local nonlinearity feature, and make described recurrence be monotone increasing.
Wherein, the media quality assessment models set up, for carrying out comprehensive grading based on several data to the quality of media.Particularly, quality for media carries out comprehensive assessment in conjunction with arrival rate, two jumping figures, advertising environments index etc., obtain an actual definition at arithmetic number collection and continuous print quality score value, the estimation carrying out Flow Value for follow-up corresponding model provides better data supporting.Wherein, this model does media quality estimation mainly for the optimization that the KPI parameter set by advertiser is arrival rate (CRR).
Wherein, the CVR estimation model set up, by combined with CT R estimation model, is estimated exposure conversion ratio.This model carries out exposure conversion ratio for click correlation (as CTR, CPC) and/or effect target with conversion correlation (as CPA, ROI etc.) mainly for the KPI parameter set by advertiser and estimates.
Wherein, the probability black and white lists model set up, for setting a black and white regulation coefficient for each list, to distinguish the probability that described each list is blacklist or white list, be integrated into a list by black and white two lists, only judge the probability of each concrete list as blacklist or white list according to black and white adjustment System tree.Wherein, so-called blacklist refers to the list comprised in the dimension (dimension such as such as domain name, advertisement position, browser) that relative efficacy is poor, so-called white list refer to comprise in the good dimension of relative efficacy list.For blacklist, described Competitive Bidding Model 4 is not bidded.And in fact, list due to a certain dimension is not non-white and black, but fall between, therefore, the present invention is by the black and white regulation coefficient of probability black and white lists model setting, make blacklist and white list no longer there is clear and definite boundary, the estimation carrying out Flow Value for follow-up corresponding model provides better data supporting.
Model Fusion module 2 merges each model set up.More specifically, when described model building module at least sets up CTR estimation model and CVR estimation model, described Model Fusion module 2 merges based on the chain rule of probability the CTR model and CVR model set up, become the model of complete set, for being that the Flow Value of CPA is estimated to KPI, instead of by based on each isolated model set up separately, the Flow Value that KPI is CPA is estimated.
Flow Value estimation module 3, for given KPI parameter, obtains historical data and the real time data of advertisement putting, based on merged model, is worth estimates ad traffic to be put.Particularly, described Flow Value estimation module 3 is according to various communication protocol, by Various types of data transmission interface, with storage or provide one or more database of the historical data of described advertisement putting and real time data or other third party devices to interact, to extract the historical data and real time data that described database stores, or directly obtain historical data and the real time data of all advertisement puttings that whole database comprises; Or the mode that described Flow Value estimation module 3 can also be crawled by network, after the whole network crawls, obtain historical data and the real time data of described advertisement putting, or by crawling the one or more database comprising described data of rear acquisition, and then therefrom obtain required historical data and real time data.Described Flow Value estimation module 3, by fully obtaining all relevant informations, adopts Bayesian formula to estimate Flow Value, its form of expression can be P (Y|X1, X2 ...).Wherein, described historical data and real time data include but not limited to the demand information that advertiser throws in advertisement, advertisement exposure scene information, audience information and other information relevant to advertisement of obtaining from advertisement transaction platform, and the form of described historical data and real time data includes but not limited to text, picture, audio frequency, video, other multimedia resources etc.
At this; those skilled in the art will be understood that the information that above-mentioned historical data and real time data comprise is only citing, and other relevant informations that are existing or that may occur from now on are as being applicable to the present invention; also within scope should being included in, and this is contained at this with way of reference.
Be still that CPA illustrates with KPI, after merging based on the chain method of probability the CTR model and CVR model set up by above-mentioned Model Fusion module 2, by formula CR=CTR*CVR estimated exposure conversion ratio, wherein, described CTR and CVR is exported by the CTR estimation model on basis and the CVR estimation model on basis respectively.
At this, those skilled in the art will be understood that the parameter of above-mentioned KPI is that CPA is only citing, can also comprise various parameter and the parameter combinations such as such as CPR, in this no limit.When described KPI parameter is CPR, CTR estimation model and CRR estimation model is set up by model building module 1, and merge described two models by described Model Fusion module 2, by Flow Value estimation module 3 based on merged model, Flow Value estimation is carried out to the parameter that described KPI is CPR.
The Flow Value that module of bidding 4 obtains based on described estimation, determines that suitable bidding strategy is bidded.Particularly, described in bid the value of the chance of advertisement exposure each time that module 4 estimates based on described Flow Value estimation module 3, select suitable advertisement and price to bid, with so competing that to expose display machine meeting at every turn.Wherein, the bid model set up, based on at least bidding based on following two factors: the difference of the actual CPC of different prediction CTR section, CPM, CPC, CPA of different time sections and the difference of average competing yield.
Those skilled in the art will be understood that, above-mentioned model building module, Model Fusion module, Flow Value estimation module, module of bidding are only example, in practice, they can be four independently modules, or any several module integration is in a module, also can all be integrated in a module.
Device of bidding provided by the invention is by setting up and the multiple parameterized model of conbined usage, meet the optimization demand of the different K PI set by diversified advertiser, apply simultaneously Bayes statistical method solve traditional statistical method be difficult to the Expired Drugs avoided and result is carried out simply, non-white and black judgement roughly, cause judged result inaccurate.Certainly, the present invention can also incorporate relevant artificial experience in parameterized model, optimizes further the KPI parameter of advertiser's setting, promotes the experience that advertiser throws in advertisement.
Fig. 2 illustrates a kind of price competing method process flow diagram according to a further aspect of the present invention.Particularly, in step s1, model building module sets up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model; In step s2, Model Fusion module merges the model set up; In step s3, Flow Value estimation module, for given KPI parameter, obtains historical data and the real time data of advertisement putting, based on merged model, is worth estimates ad traffic to be put; In step s4, the Flow Value that module of bidding obtains based on described estimation, determines that suitable bidding strategy is bidded.
Be constant work between above steps, at this, it will be understood by those skilled in the art that " continuing " refers to that above steps requires to carry out respectively in real time or according to the mode of operation of setting or real-time adjustment.
In step s1, model building module sets up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model, and sets up bid model.
Wherein, the CTR estimation model set up is for the clicking rate based on Bayes statistical method estimated flow, and Bayes is carried out smoothly to the data of each dimensions such as domain name, advertisement position, operating system, browser, thus solve the problem of Deta sparseness and over-fitting.
Wherein, the CTR calibrating patterns set up is for calibrating the clicking rate value estimated by described CTR estimation model, particularly, Log linear regression algorithm is adopted to catch the overall corresponding relation of CTR value and the true CTR value obtained based on described CTR estimation model, and adopt true CTR value to realize the calibration to the clicking rate value estimated by described CTR estimation model as feedback, make the CTR value after calibrating and true CTR value closely, promote the accuracy rate that follow-up corresponding model carries out the estimation of Flow Value.Further, described CTR calibrating patterns also for: based on Log linear regression algorithm, use the method for local regression to catch local nonlinearity feature, and make described recurrence be monotone increasing.
Wherein, the media quality assessment models set up, for carrying out comprehensive grading based on several data to the quality of media.Particularly, quality for media carries out comprehensive assessment in conjunction with arrival rate, two jumping figures, advertising environments index etc., obtain an actual definition at arithmetic number collection and continuous print quality score value, the estimation carrying out Flow Value for follow-up corresponding model provides better data supporting.Wherein, this model does media quality estimation mainly for the optimization that the KPI parameter set by advertiser is arrival rate (CRR).
Wherein, the probability black and white lists model set up, for setting a black and white regulation coefficient for each list, to distinguish the probability that described each list is blacklist or white list, be integrated into a list by black and white two lists, only judge the probability of each concrete list as blacklist or white list according to black and white adjustment System tree.Wherein, so-called blacklist refers to the list comprised in the dimension (dimension such as such as domain name, advertisement position, browser) that relative efficacy is poor, so-called white list refer to comprise in the good dimension of relative efficacy list.For blacklist, described Competitive Bidding Model 4 is not bidded.And in fact, list due to a certain dimension is not non-white and black, but fall between, therefore, the present invention is by the black and white regulation coefficient of probability black and white lists model setting, make blacklist and white list no longer there is clear and definite boundary, the estimation carrying out Flow Value for follow-up corresponding model provides better data supporting.
In step s2, described Model Fusion module merges each model set up.More specifically, when described model building module at least sets up CTR estimation model and CVR estimation model, described Model Fusion module merges based on the chain rule of probability the CTR model and CVR model set up, become the model of complete set, for being that the Flow Value of CPA is estimated to KPI, instead of by based on each isolated model set up separately, the Flow Value that KPI is CPA is estimated.
In step s3, Flow Value estimation module, for given KPI parameter, obtains historical data and the real time data of advertisement putting, based on merged model, is worth estimates ad traffic to be put.Particularly, described Flow Value estimation module is according to various communication protocol, by Various types of data transmission interface, with storage or provide one or more database of the historical data of described advertisement putting and real time data or other third party devices to interact, to extract the historical data and real time data that described database stores, or directly obtain historical data and the real time data of all advertisement puttings that whole database comprises; Or the mode that described Flow Value estimation module can also be crawled by network, after the whole network crawls, obtain historical data and the real time data of described advertisement putting, or by crawling the one or more database comprising described data of rear acquisition, and then therefrom obtain required historical data and real time data.Described Flow Value estimation module 3, by fully obtaining all relevant informations, adopts Bayesian formula to estimate Flow Value, its form of expression can be P (Y|X1, X2 ...).Wherein, described historical data and real time data include but not limited to the demand information that advertiser throws in advertisement, advertisement exposure scene information, audience information and other information relevant to advertisement of obtaining from advertisement transaction platform, and the form of described historical data and real time data includes but not limited to text, picture, audio frequency, video, other multimedia resources etc.
At this; those skilled in the art will be understood that the information that above-mentioned historical data and real time data comprise is only citing, and other relevant informations that are existing or that may occur from now on are as being applicable to the present invention; also within scope should being included in, and this is contained at this with way of reference.
In above-mentioned steps, be still that CPA illustrates with KPI, after merging based on the chain method of probability the CTR model and CVR model set up by above-mentioned Model Fusion module, by formula CR=CTR*CVR estimated exposure conversion ratio, wherein, described CTR and CVR is exported by the CTR estimation model on basis and the CVR estimation model on basis respectively.
At this, those skilled in the art will be understood that the parameter of above-mentioned KPI is that CPA is only citing, can also comprise various parameter and the parameter combinations such as such as CPR, in this no limit.When described KPI parameter is CPR, CTR estimation model and CRR estimation model is set up by above-mentioned model building module, and merge described two models by described Model Fusion module, by Flow Value estimation module based on merged model, Flow Value estimation is carried out to the parameter that described KPI is CPR.
In step s4, the Flow Value that module of bidding obtains based on described estimation, determines that suitable bidding strategy is bidded.Particularly, described in bid the value of the chance of advertisement exposure each time that module estimates based on described Flow Value estimation module, select suitable advertisement and price to bid, with so competing that to expose display machine meeting at every turn.Wherein, the bid model set up, based on at least bidding based on following two factors: the difference of the actual CPC of different prediction CTR section, CPM, CPC, CPA of different time sections and the difference of average competing yield.
To those skilled in the art, obviously the invention is not restricted to the details of above-mentioned one exemplary embodiment, and when not deviating from spirit of the present invention or essential characteristic, the present invention can be realized in other specific forms.Therefore, no matter from which point, all should embodiment be regarded as exemplary, and be nonrestrictive, scope of the present invention is limited by claims instead of above-mentioned explanation, and all changes be therefore intended in the implication of the equivalency by dropping on claim and scope are included in the present invention.Any Reference numeral in claim should be considered as the claim involved by limiting.
Claims (10)
1. a price competing method, the method comprises:
Set up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model;
Merge the model set up;
For given KPI parameter, obtain historical data and the real time data of advertisement putting, based on merged model, ad traffic to be put is worth and estimates;
Based on the Flow Value that described estimation obtains, determine that suitable bidding strategy is bidded.
2. price competing method according to claim 1, wherein,
The CTR estimation model set up is for estimating clicking rate based on Bayes statistical method; Or/and,
The overall corresponding relation of the CTR calibrating patterns set up for adopting Log linear regression algorithm to catch CTR value and the true CTR value obtained based on described CTR estimation model; Or/and,
The media quality assessment models set up, for carrying out comprehensive grading based on several data to the quality of media; Or/and,
The probability black and white lists model set up, for setting a black and white regulation coefficient for each list, to distinguish the probability that described each list is blacklist or white list; Or/and,
In described bidding, at least bid based on following two factors: CPM, CPC, CPA of the actual CPC of different prediction CTR section, different time sections and average competing yield.
3. price competing method according to claim 2, wherein, described CTR calibrating patterns also for: based on Log linear regression algorithm, use the method for local regression to catch local nonlinearity feature, and make described recurrence be monotone increasing.
4. price competing method according to claim 2, wherein, the media quality assessment models set up at least based on arrival rate, two jumping figures or/and the quality of advertising environments exponent pair media carries out comprehensive grading, obtain an actual definition at arithmetic number collection and continuous print quality score value.
5. the price competing method according to any one of claim 1-4, wherein,
When described specific KPI is CPA, described price competing method also comprises:
At least set up CTR estimation model and CVR estimation model;
Chain rule based on probability merges the CTR model and CVR model set up.
6. bid a device, this device comprises:
Model building module, for setting up at least one model in described CTR estimation model, CTR calibrating patterns, media quality assessment models, CVR estimation model and probability black and white lists model;
Model Fusion module, for merging set up model;
Flow Value estimation module, for given KPI parameter, for obtaining historical data and the real time data of advertisement putting, based on merged model, being worth ad traffic to be put and estimating;
To bid module, based on the Flow Value that described estimation obtains, determine that suitable bidding strategy is bidded.
7. device of bidding according to claim 6, wherein,
The CTR estimation model set up is for estimating clicking rate based on Bayes statistical method; Or/and,
The overall corresponding relation of the CTR calibrating patterns set up for adopting Log linear regression algorithm to catch CTR value and the true CTR value obtained based on described CTR estimation model; Or/and,
The media quality assessment models set up, for carrying out comprehensive grading based on several data to the quality of media; Or/and,
The probability black and white lists model set up, for setting a black and white regulation coefficient for each list, to distinguish the probability that described each list is blacklist or white list; Or/and,
The bid model set up, at least bidding based on following two factors: CPM, CPC, CPA of the actual CPC of different prediction CTR section, different time sections and average competing yield.
8. device of bidding according to claim 7, wherein, described CTR calibrating patterns also for: based on Log linear regression algorithm, use the method for local regression to catch local nonlinearity feature, and make described recurrence be monotone increasing.
9. device of bidding according to claim 7, wherein, the media quality assessment models set up at least based on arrival rate, two jumping figures or/and the quality of advertising environments exponent pair media carries out comprehensive grading, obtain an actual definition at arithmetic number collection and continuous print quality score value.
10. the device of bidding according to any one of claim 6-9, wherein, when described specific KPI is CPA, bid in device described:
Described model building module at least sets up CTR estimation model and CVR estimation model;
Described Model Fusion module merges based on the chain rule of probability the CTR model and CVR model set up.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510484829.8A CN105046532A (en) | 2015-08-07 | 2015-08-07 | Bidding method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510484829.8A CN105046532A (en) | 2015-08-07 | 2015-08-07 | Bidding method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105046532A true CN105046532A (en) | 2015-11-11 |
Family
ID=54453056
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510484829.8A Pending CN105046532A (en) | 2015-08-07 | 2015-08-07 | Bidding method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105046532A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104574160A (en) * | 2015-01-29 | 2015-04-29 | 深圳市英威诺科技有限公司 | Smooth advertisement traffic control method |
| CN105574741A (en) * | 2015-12-22 | 2016-05-11 | 北京奇虎科技有限公司 | Flow value assessment method, device, application method and system |
| CN105678317A (en) * | 2015-12-30 | 2016-06-15 | 腾讯科技（深圳）有限公司 | Information processing method and server |
| CN106779860A (en) * | 2016-12-27 | 2017-05-31 | 天津数集科技有限公司 | Advertising campaign simulation system |
| CN106844178A (en) * | 2017-01-22 | 2017-06-13 | 腾云天宇科技（北京）有限公司 | Prediction is presented method, computing device, server and the system of information transferring rate |
| CN106846046A (en) * | 2016-12-29 | 2017-06-13 | 腾讯科技（深圳）有限公司 | Heterogeneous flow is bidded equalization parameter method and system |
| CN107067274A (en) * | 2016-12-27 | 2017-08-18 | 北京掌阔移动传媒科技有限公司 | One DSP real time bid ad system based on blended learning model |
| CN107527128A (en) * | 2016-06-21 | 2017-12-29 | 阿里巴巴集团控股有限公司 | A kind of method and apparatus for determining resource parameters |
| CN108171545A (en) * | 2017-12-27 | 2018-06-15 | 银橙(上海)信息技术有限公司 | A kind of conversion ratio predictor method based on level of hierarchy data |
| CN108256881A (en) * | 2016-12-28 | 2018-07-06 | 北京奇虎科技有限公司 | A kind of traffic filtering method and apparatus |
| CN108681915A (en) * | 2018-04-18 | 2018-10-19 | 北京奇艺世纪科技有限公司 | A kind of clicking rate predictor method, device and electronic equipment |
| CN109426980A (en) * | 2017-08-31 | 2019-03-05 | 腾讯科技（深圳）有限公司 | Determine method, apparatus, server and the storage medium of advertisement bidding |
| CN110889725A (en) * | 2019-11-26 | 2020-03-17 | 深圳市随手金服信息科技有限公司 | Online advertisement CTR estimation method, device, equipment and storage medium |
| CN111192094A (en) * | 2020-01-06 | 2020-05-22 | 武汉稀云科技有限公司 | Intelligent advertising method and device |
| CN111199454A (en) * | 2019-12-16 | 2020-05-26 | 北京淇瑀信息科技有限公司 | Real-time user transformation evaluation method and device and electronic equipment |
| CN111383047A (en) * | 2020-01-06 | 2020-07-07 | 武汉稀云科技有限公司 | Intelligent advertising method and device with real-time bidding function |
| CN112396475A (en) * | 2020-12-23 | 2021-02-23 | 上海苍苔信息技术有限公司 | CPM system and method for controlling CPA value |
| CN112446717A (en) * | 2019-08-27 | 2021-03-05 | 百度在线网络技术（北京）有限公司 | Advertisement putting method and device |
| CN113222656A (en) * | 2021-04-29 | 2021-08-06 | 西安点告网络科技有限公司 | Programmed advertisement putting method, system, device, equipment and storage medium |
| CN113807891A (en) * | 2021-09-16 | 2021-12-17 | 北京沃东天骏信息技术有限公司 | Advertisement putting processing method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206758A (en) * | 2006-12-24 | 2008-06-25 | 洪富和 | Method and system for searching bid ranking |
| KR101324106B1 (en) * | 2012-07-26 | 2013-10-31 | 김수현 | Advertise exposure ranking management apparatus and method using client position |
| CN104809144A (en) * | 2014-01-28 | 2015-07-29 | 祁勇 | Personalized advertisement serving method based on auxiliary keywords |
-
2015
- 2015-08-07 CN CN201510484829.8A patent/CN105046532A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206758A (en) * | 2006-12-24 | 2008-06-25 | 洪富和 | Method and system for searching bid ranking |
| KR101324106B1 (en) * | 2012-07-26 | 2013-10-31 | 김수현 | Advertise exposure ranking management apparatus and method using client position |
| CN104809144A (en) * | 2014-01-28 | 2015-07-29 | 祁勇 | Personalized advertisement serving method based on auxiliary keywords |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104574160A (en) * | 2015-01-29 | 2015-04-29 | 深圳市英威诺科技有限公司 | Smooth advertisement traffic control method |
| CN105574741A (en) * | 2015-12-22 | 2016-05-11 | 北京奇虎科技有限公司 | Flow value assessment method, device, application method and system |
| CN105678317A (en) * | 2015-12-30 | 2016-06-15 | 腾讯科技（深圳）有限公司 | Information processing method and server |
| CN107527128B (en) * | 2016-06-21 | 2021-03-23 | 阿里巴巴集团控股有限公司 | Resource parameter determination method and equipment for advertisement platform |
| CN107527128A (en) * | 2016-06-21 | 2017-12-29 | 阿里巴巴集团控股有限公司 | A kind of method and apparatus for determining resource parameters |
| CN107067274A (en) * | 2016-12-27 | 2017-08-18 | 北京掌阔移动传媒科技有限公司 | One DSP real time bid ad system based on blended learning model |
| CN106779860A (en) * | 2016-12-27 | 2017-05-31 | 天津数集科技有限公司 | Advertising campaign simulation system |
| CN108256881A (en) * | 2016-12-28 | 2018-07-06 | 北京奇虎科技有限公司 | A kind of traffic filtering method and apparatus |
| CN106846046A (en) * | 2016-12-29 | 2017-06-13 | 腾讯科技（深圳）有限公司 | Heterogeneous flow is bidded equalization parameter method and system |
| CN106844178A (en) * | 2017-01-22 | 2017-06-13 | 腾云天宇科技（北京）有限公司 | Prediction is presented method, computing device, server and the system of information transferring rate |
| CN106844178B (en) * | 2017-01-22 | 2019-11-15 | 腾云天宇科技（北京）有限公司 | Prediction is presented the method for information transferring rate, calculates equipment, server and system |
| CN109426980B (en) * | 2017-08-31 | 2022-11-11 | 腾讯科技（深圳）有限公司 | Method, device, server and storage medium for determining advertisement bidding |
| CN109426980A (en) * | 2017-08-31 | 2019-03-05 | 腾讯科技（深圳）有限公司 | Determine method, apparatus, server and the storage medium of advertisement bidding |
| CN108171545A (en) * | 2017-12-27 | 2018-06-15 | 银橙(上海)信息技术有限公司 | A kind of conversion ratio predictor method based on level of hierarchy data |
| CN108681915A (en) * | 2018-04-18 | 2018-10-19 | 北京奇艺世纪科技有限公司 | A kind of clicking rate predictor method, device and electronic equipment |
| CN112446717A (en) * | 2019-08-27 | 2021-03-05 | 百度在线网络技术（北京）有限公司 | Advertisement putting method and device |
| CN112446717B (en) * | 2019-08-27 | 2023-08-11 | 百度在线网络技术（北京）有限公司 | Advertisement putting method and device |
| CN110889725B (en) * | 2019-11-26 | 2023-08-22 | 深圳市财象云科技有限公司 | Online advertisement CTR estimation method, device, equipment and storage medium |
| CN110889725A (en) * | 2019-11-26 | 2020-03-17 | 深圳市随手金服信息科技有限公司 | Online advertisement CTR estimation method, device, equipment and storage medium |
| CN111199454A (en) * | 2019-12-16 | 2020-05-26 | 北京淇瑀信息科技有限公司 | Real-time user transformation evaluation method and device and electronic equipment |
| CN111199454B (en) * | 2019-12-16 | 2024-04-26 | 北京淇瑀信息科技有限公司 | Real-time user conversion evaluation method and device and electronic equipment |
| CN111192094A (en) * | 2020-01-06 | 2020-05-22 | 武汉稀云科技有限公司 | Intelligent advertising method and device |
| CN111192094B (en) * | 2020-01-06 | 2023-05-05 | 武汉稀云科技有限公司 | Intelligent advertising method and device |
| CN111383047B (en) * | 2020-01-06 | 2023-06-13 | 武汉稀云科技有限公司 | Intelligent advertisement method and device with real-time bidding |
| CN111383047A (en) * | 2020-01-06 | 2020-07-07 | 武汉稀云科技有限公司 | Intelligent advertising method and device with real-time bidding function |
| CN112396475A (en) * | 2020-12-23 | 2021-02-23 | 上海苍苔信息技术有限公司 | CPM system and method for controlling CPA value |
| CN113222656A (en) * | 2021-04-29 | 2021-08-06 | 西安点告网络科技有限公司 | Programmed advertisement putting method, system, device, equipment and storage medium |
| CN113222656B (en) * | 2021-04-29 | 2023-11-24 | 西安点告网络科技有限公司 | Programmed advertisement putting method, system, device, equipment and storage medium |
| CN113807891A (en) * | 2021-09-16 | 2021-12-17 | 北京沃东天骏信息技术有限公司 | Advertisement putting processing method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105046532A (en) | Bidding method and device | |
| TWI476716B (en) | Integrated and comprehensive advertising campaign management and optimization | |
| US10037543B2 (en) | Estimating conversion rate in display advertising from past performance data | |
| Wang et al. | Understanding customer needs through quantitative analysis of Kano's model | |
| Yang et al. | Analyzing the relationship between organic and sponsored search advertising: Positive, negative, or zero interdependence? | |
| US7827128B1 (en) | System identification, estimation, and prediction of advertising-related data | |
| US20160210658A1 (en) | Determining touchpoint attributions in a segmented media campaign | |
| US20170337578A1 (en) | Dynamic media buy optimization using attribution-informed media buy execution feeds | |
| US20120046996A1 (en) | Unified data management platform | |
| US20160132935A1 (en) | Systems, methods, and apparatus for flexible extension of an audience segment | |
| US20110040613A1 (en) | Learning system for advertising bidding and valuation of third party data | |
| KR20210006505A (en) | Methods and apparatus to estimate demographics of users employing social media | |
| US20160210656A1 (en) | System for marketing touchpoint attribution bias correction | |
| CN104252679A (en) | Construction method and system of brand advertisement evaluation system | |
| CN111768244A (en) | Advertisement delivery recommendation method and device | |
| CN110109901B (en) | Method and device for screening target object | |
| TW201142733A (en) | Placeholder bids in online advertising | |
| KR20080086454A (en) | Data independent relevance evaluation utilizing cognitive concept relationship | |
| US20160098524A1 (en) | Medical inventory tracking | |
| Perumal et al. | Social media in food and beverages industry: case of Klang Valley, Malaysia | |
| CN110473043B (en) | Article recommendation method and device based on user behaviors | |
| US20180096379A1 (en) | Methods and systems for estimating Click-Through-Rate for a SERP layout | |
| CN107067282B (en) | Consumer product rebate sale marketing management system and use method thereof | |
| US20160292196A1 (en) | Methods and Systems for Collaborated Change Point Detection in Time Series | |
| CN110766456A (en) | Commodity recommendation method and device |
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 |
Application publication date: 20151111 |
|
| RJ01 | Rejection of invention patent application after publication |