CN103942603A - Advertisement click rate prediction method and device - Google Patents
Advertisement click rate prediction method and device Download PDFInfo
- Publication number
- CN103942603A CN103942603A CN201310016785.7A CN201310016785A CN103942603A CN 103942603 A CN103942603 A CN 103942603A CN 201310016785 A CN201310016785 A CN 201310016785A CN 103942603 A CN103942603 A CN 103942603A
- Authority
- CN
- China
- Prior art keywords
- click rate
- fiducial interval
- value
- advertisement
- degree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The invention provides an advertisement click rate prediction method and device. The method comprises the steps that the number of times of presentation and the number of times of being clicked of an advertisement are counted; according to the number of times of presentation and the number of times of being clicked, the confidence interval of the advertisement click rate is calculated; and according to normal distribution, a value is sampled from the confidence interval and is used as a prediction value of the advertisement click rate. According to the invention, an adverse effect caused by the prediction deviation of the click rate can be solved.
Description
Technical field
The present invention relates to order ads technical field, relate in particular to a kind of predictor method and device of ad click rate.
Background technology
At present, effect ad system in the industry, in the time carrying out order ads, is all to have used the method for estimating based on ad click rate.Concrete grammar is, first estimate the clicking rate of candidate's advertisement by the method for statistics or machine learning (as logistic regression), forecast click-through-rate (pCTR, predicted click-through rate), then calculate the quality degree (Quality) of advertisement based on pCTR, finally advertisement is arranged and represented according to bid (Bid) * Quality backward, the order ads higher and that quality degree is higher of bidding is more forward.In the time of sequence, follow general secondary price auction (GSP, Generalized Second Price Auction) mechanism, this mechanism can maximize the income of search engine, reaches GSP equilibrium.In above-mentioned computation process, it is one of most crucial part that ad click rate is estimated.
The method of estimating ad click rate is mainly point estimation method (point estimation), and weighed value is estimated again.The method of point estimation has a lot, has the method based on statistics, also have the method based on machine learning, but its general character is to estimate all to export a fixing probable value at every turn.Illustrate, if an advertisement A represented in history 10 times, clicked mistake 1 time, the method based on traditional so, estimating its ad click rate is 1/10=10%.
Representing all sparse with click data in the situation that, adopt which kind of method to estimate ad click rate and all can have larger deviation; And the existing method of estimating ad click rate estimates and all export a fixing probable value at every turn, therefore cannot solve this deviation adverse effect.
Summary of the invention
The invention provides a kind of predictor method and device of ad click rate, estimate thereby solve clicking rate the adverse effect that deviation is brought.
Technical scheme of the present invention is achieved in that
A predictor method for ad click rate, comprising:
Statistics advertisement represent number of times and clicked number of times; According to the described fiducial interval that represents number of times and clicked number of times and calculate described ad click rate;
According to normal distribution from the discreet value of a value as ad click rate of sampling in described fiducial interval.
In said method, after the discreet value of the value of sampling in fiducial interval according to normal distribution as ad click rate, further comprise:
Adopt the discreet value of described ad click rate to calculate the quality degree of advertisement, determine the collating sequence of described advertisement according to described quality degree.
The concrete mode of calculating ad quality degree can be: by the discreet value of described ad click rate, correlativity and the weighted sum of page object quality degree, obtain ad quality degree.
Above-mentioned fiducial interval is the fiducial interval of 95% degree of confidence;
The mode of calculating the fiducial interval of ad click rate 95% degree of confidence is the calculating of employing following formula:
Wherein, n is for representing number of times;
for clicked number of times is divided by representing number of times;
α=0.05;
z
1-α/2=1.96。
In said method, according to normal distribution from the value of sampling in fiducial interval as the mode of the discreet value of ad click rate can be:
According to average be
standard deviation is
normal distribution from the discreet value of a value as ad click rate of sampling in fiducial interval.
An estimating device for ad click rate, comprising:
Statistical module, represents number of times and clicked number of times for what add up advertisement;
Computing module, calculates the fiducial interval of described ad click rate for representing number of times and clicked number of times described in basis;
Estimate module, for according to normal distribution from the discreet value of a value as ad click rate of sampling in described fiducial interval.
Said apparatus also comprises order module, for adopting the discreet value of described ad click rate to calculate the quality degree of advertisement, determines the collating sequence of described advertisement according to described quality degree.
In said apparatus, order module, by the discreet value of described ad click rate, correlativity and the weighted sum of page object quality degree, obtains ad quality degree.
Above-mentioned fiducial interval is the fiducial interval of 95% degree of confidence;
Described computing module adopts following formula to calculate the fiducial interval of ad click rate 95% degree of confidence:
Wherein, n is for representing number of times;
for clicked number of times is divided by representing number of times;
α=0.05;
z
1-α/2=1.96。
In said apparatus, estimate module and can be according to average
standard deviation is
normal distribution from the discreet value of a value as ad click rate of sampling in fiducial interval.
Visible, the present invention proposes predictor method and the device of ad click rate, can in the time estimating, produce numerical value in fiducial interval and according to the discreet value of normal distribution variation at every turn, estimates thereby solve the adverse effect that deviation is brought.
Brief description of the drawings
Fig. 1 is the ad click rate predictor method process flow diagram that the present invention proposes;
Fig. 2 is the ad click rate estimating device structural representation that the present invention proposes.
Embodiment
The present invention proposes a kind of predictor method of ad click rate, if Fig. 1 is the method process flow diagram, comprising:
Step 101: statistics advertisement represent number of times and clicked number of times;
Step 102: according to the described fiducial interval that represents number of times and clicked number of times and calculate described ad click rate;
Step 103: according to normal distribution from the discreet value of a value as ad click rate of sampling in fiducial interval.
In said method, after step 103, may further include:
Adopt the discreet value of ad click rate to calculate the quality degree of advertisement, determine the collating sequence of described advertisement according to quality degree.
Particularly, the mode of calculating ad quality degree can be: by the weighted sum of the discreet value of described ad click rate, correlativity and page object (Landing Page) quality degree, obtain ad quality degree.
In said method, fiducial interval can be the fiducial interval of 95% degree of confidence;
The mode of calculating the fiducial interval of ad click rate 95% degree of confidence is to adopt following formula to calculate:
Wherein, n is for representing number of times;
for clicked number of times is divided by representing number of times;
α=1-95%=0.05;
z
1-α/2=1.96。
In said method, can be specially as the mode of the discreet value of ad click rate from the value of sampling in fiducial interval according to normal distribution: according to average be
standard deviation is
normal distribution from the discreet value of a value as ad click rate of sampling in fiducial interval.
Below lifting specific embodiment introduces in detail.
Embodiment mono-:
In the present embodiment, advertisement A represented 100 times in history, and clicked mistake 10 times is calculated 95% fiducial interval [min_pCTR, max_pCTR] of this advertisement pCTR, adopted following formula to calculate:
Wherein,
for clicked number of times is divided by representing number of times,
N is for representing number of times, n=100;
α=1-95%=0.05;
z
1-α/2=1.96。
According to above-mentioned formula, calculate min_pCTR=0.1-0.0588=0.0412, max_pCTR=0.1+0.0588=0.1588,95% fiducial interval of this advertisement pCTR is [0.0412,0.1588].
While estimating, according to average be at every turn
standard deviation is
normal distribution in the fiducial interval [0.0412,0.1588] value of sampling as the discreet value of this advertisement pCTR.The method of sampling of normal distribution has a lot, and in addition the inverse function of the normal state cumulative distribution function that a most basic method is use standard also has additive method, as Box-Muller conversion, Ziggurat algorithm etc.
Afterwards, calculate the quality degree of this advertisement according to the discreet value of this advertisement pCTR, specifically can adopt discreet value, correlativity and page object (Landing Page) the quality degree three's of this advertisement pCTR weighted sum.Determine the collating sequence of described advertisement according to the quality degree calculating.
Due to result the on-fixed of each sampling, but be distributed in fiducial interval [0.0412,0.1588], therefore can solve and estimate the adverse effect that deviation is brought.
In the time of the representing number of times or clicked number of times and change of this advertisement, recalculate 95% fiducial interval [min_pCTR, max_pCTR] of this advertisement pCTR, can be found out by computing formula, historical data is more abundant, and 95% fiducial interval of pCTR is less; Historical data is more sparse, and 95% fiducial interval of pCTR is larger, thereby has solved the large problem of point estimation error in Sparse situation.
The present invention also proposes a kind of estimating device of ad click rate, if Fig. 2 is this apparatus structure schematic diagram, comprising:
Statistical module 201, represents number of times and clicked number of times for what add up advertisement;
Computing module 202, calculates the fiducial interval of described ad click rate for representing number of times and clicked number of times described in basis;
Estimate module 203, for according to normal distribution from the discreet value of a value as ad click rate of sampling in described fiducial interval.
Said apparatus can also comprise order module 204, for adopting the discreet value of described ad click rate to calculate the quality degree of advertisement, determines the collating sequence of described advertisement according to described quality degree.
Order module 204, by the discreet value of described ad click rate, correlativity and the weighted sum of page object quality degree, obtains ad quality degree.
In said apparatus, fiducial interval can be the fiducial interval of 95% degree of confidence;
Computing module 202 can adopt following formula to calculate the fiducial interval of ad click rate 95% degree of confidence:
Wherein, n is for representing number of times;
for clicked number of times is divided by representing number of times;
α=0.05;
z
1-α/2=1.96。
In said apparatus, estimate module 203 and can be according to average
standard deviation is
normal distribution from the discreet value of a value as ad click rate of sampling in fiducial interval.
As fully visible, predictor method and the device of the ad click rate that the present invention proposes, first determine the fiducial interval of ad click rate, the discreet value as ad click rate according to a value in normal distribution sampling fiducial interval in the time at every turn estimating afterwards, thereby make the value of at every turn estimating meet normal distribution, instead of a value of fixing output, solve and estimated deviation adverse effect, and more represented chance can to new advertisement.
The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any amendment of making, be equal to replacement, improvement etc., within all should being included in the scope of protection of the invention.
Claims (10)
1. a predictor method for ad click rate, is characterized in that, described method comprises:
Statistics advertisement represent number of times and clicked number of times; According to the described fiducial interval that represents number of times and clicked number of times and calculate described ad click rate;
According to normal distribution from the discreet value of a value as ad click rate of sampling in described fiducial interval.
2. method according to claim 1, is characterized in that, after the described discreet value of a value as ad click rate of sampling in fiducial interval according to normal distribution, further comprises:
Adopt the discreet value of described ad click rate to calculate the quality degree of advertisement, determine the collating sequence of described advertisement according to described quality degree.
3. method according to claim 2, is characterized in that, the mode of described calculating ad quality degree is:
By the discreet value of described ad click rate, correlativity and the weighted sum of page object quality degree, obtain ad quality degree.
4. according to the method described in claim 1,2 or 3, it is characterized in that, described fiducial interval is the fiducial interval of 95% degree of confidence;
The mode of calculating the fiducial interval of ad click rate 95% degree of confidence is the calculating of employing following formula:
Wherein, n is for representing number of times;
for clicked number of times is divided by representing number of times;
α=0.05;
z
1-α/2=1.96。
5. method according to claim 4, is characterized in that, describedly according to normal distribution from the value of sampling in fiducial interval as the mode of the discreet value of ad click rate is:
According to average be
standard deviation is
normal distribution from the discreet value of a value as ad click rate of sampling in fiducial interval.
6. an estimating device for ad click rate, is characterized in that, described device comprises:
Statistical module, represents number of times and clicked number of times for what add up advertisement;
Computing module, calculates the fiducial interval of described ad click rate for representing number of times and clicked number of times described in basis;
Estimate module, for according to normal distribution from the discreet value of a value as ad click rate of sampling in described fiducial interval.
7. device according to claim 6, is characterized in that, described device also comprises order module, for adopting the discreet value of described ad click rate to calculate the quality degree of advertisement, determines the collating sequence of described advertisement according to described quality degree.
8. device according to claim 7, is characterized in that, described order module, by the discreet value of described ad click rate, correlativity and the weighted sum of page object quality degree, obtains ad quality degree.
9. according to the device described in claim 6,7 or 8, it is characterized in that, described fiducial interval is the fiducial interval of 95% degree of confidence;
Described computing module adopts following formula to calculate the fiducial interval of ad click rate 95% degree of confidence:
Wherein, n is for representing number of times;
for clicked number of times is divided by representing number of times;
α=0.05;
z
1-α/2=1.96。
10. device according to claim 9, is characterized in that, described in estimate module and according to average be
standard deviation is
normal distribution from the discreet value of a value as ad click rate of sampling in fiducial interval.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310016785.7A CN103942603B (en) | 2013-01-17 | 2013-01-17 | Method and device for estimating advertisement click rate |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310016785.7A CN103942603B (en) | 2013-01-17 | 2013-01-17 | Method and device for estimating advertisement click rate |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103942603A true CN103942603A (en) | 2014-07-23 |
CN103942603B CN103942603B (en) | 2020-05-22 |
Family
ID=51190265
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310016785.7A Active CN103942603B (en) | 2013-01-17 | 2013-01-17 | Method and device for estimating advertisement click rate |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103942603B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104463498A (en) * | 2014-12-24 | 2015-03-25 | 小米科技有限责任公司 | Method and device for counting operational indicators |
CN105260379A (en) * | 2015-09-08 | 2016-01-20 | 百度在线网络技术(北京)有限公司 | Information push method and device |
CN105760400A (en) * | 2014-12-19 | 2016-07-13 | 阿里巴巴集团控股有限公司 | Method and device for ranking push messages based on search behavior |
CN108171570A (en) * | 2017-12-15 | 2018-06-15 | 北京小度信息科技有限公司 | A kind of data screening method, apparatus and terminal |
CN108710635A (en) * | 2018-04-08 | 2018-10-26 | 达而观信息科技(上海)有限公司 | A kind of content recommendation method and device |
CN109214847A (en) * | 2017-07-05 | 2019-01-15 | 高文中 | Page clicking rate data processing method, apparatus and system |
CN109274987A (en) * | 2018-08-30 | 2019-01-25 | 武汉斗鱼网络科技有限公司 | A kind of video collection sort method, server and readable storage medium storing program for executing |
CN109978606A (en) * | 2019-03-04 | 2019-07-05 | 北京达佳互联信息技术有限公司 | Processing method, device and the computer readable storage medium of ad click rate data |
CN111917809A (en) * | 2019-05-09 | 2020-11-10 | 腾讯科技(深圳)有限公司 | Multimedia data pushing method and device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1860496A (en) * | 2003-07-22 | 2006-11-08 | Google公司 | Content-targeted advertising using collected user behavior data |
US20070094072A1 (en) * | 2005-10-26 | 2007-04-26 | Etica Entertainment, Inc., Dba Position Research | Apparatus and method of identifying critical factors in a pay-for-performance advertising network |
US7406434B1 (en) * | 2000-12-15 | 2008-07-29 | Carl Meyer | System and method for improving the performance of electronic media advertising campaigns through multi-attribute analysis and optimization |
CN101385018A (en) * | 2005-12-30 | 2009-03-11 | 谷歌公司 | Using estimated ad qualities for ad filtering, ranking and promotion |
-
2013
- 2013-01-17 CN CN201310016785.7A patent/CN103942603B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7406434B1 (en) * | 2000-12-15 | 2008-07-29 | Carl Meyer | System and method for improving the performance of electronic media advertising campaigns through multi-attribute analysis and optimization |
CN1860496A (en) * | 2003-07-22 | 2006-11-08 | Google公司 | Content-targeted advertising using collected user behavior data |
US20070094072A1 (en) * | 2005-10-26 | 2007-04-26 | Etica Entertainment, Inc., Dba Position Research | Apparatus and method of identifying critical factors in a pay-for-performance advertising network |
CN101385018A (en) * | 2005-12-30 | 2009-03-11 | 谷歌公司 | Using estimated ad qualities for ad filtering, ranking and promotion |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105760400A (en) * | 2014-12-19 | 2016-07-13 | 阿里巴巴集团控股有限公司 | Method and device for ranking push messages based on search behavior |
CN105760400B (en) * | 2014-12-19 | 2019-06-21 | 阿里巴巴集团控股有限公司 | A kind of PUSH message sort method and device based on search behavior |
CN104463498B (en) * | 2014-12-24 | 2017-11-28 | 小米科技有限责任公司 | The method and device of statistical service index |
CN104463498A (en) * | 2014-12-24 | 2015-03-25 | 小米科技有限责任公司 | Method and device for counting operational indicators |
CN105260379B (en) * | 2015-09-08 | 2019-02-01 | 北京小熊博望科技有限公司 | A kind of method and device of information push |
CN105260379A (en) * | 2015-09-08 | 2016-01-20 | 百度在线网络技术(北京)有限公司 | Information push method and device |
CN109214847A (en) * | 2017-07-05 | 2019-01-15 | 高文中 | Page clicking rate data processing method, apparatus and system |
CN108171570A (en) * | 2017-12-15 | 2018-06-15 | 北京小度信息科技有限公司 | A kind of data screening method, apparatus and terminal |
CN108710635A (en) * | 2018-04-08 | 2018-10-26 | 达而观信息科技(上海)有限公司 | A kind of content recommendation method and device |
CN109274987A (en) * | 2018-08-30 | 2019-01-25 | 武汉斗鱼网络科技有限公司 | A kind of video collection sort method, server and readable storage medium storing program for executing |
CN109978606A (en) * | 2019-03-04 | 2019-07-05 | 北京达佳互联信息技术有限公司 | Processing method, device and the computer readable storage medium of ad click rate data |
CN109978606B (en) * | 2019-03-04 | 2022-09-30 | 北京达佳互联信息技术有限公司 | Method and device for processing advertisement click rate data and computer readable storage medium |
CN111917809A (en) * | 2019-05-09 | 2020-11-10 | 腾讯科技(深圳)有限公司 | Multimedia data pushing method and device |
CN111917809B (en) * | 2019-05-09 | 2021-11-26 | 腾讯科技(深圳)有限公司 | Multimedia data pushing method and device |
Also Published As
Publication number | Publication date |
---|---|
CN103942603B (en) | 2020-05-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103942603A (en) | Advertisement click rate prediction method and device | |
CN109697522B (en) | Data prediction method and device | |
TWI521453B (en) | An estimation method and a device for estimating an estimated value of a keyword | |
CN104268644A (en) | Method and device for predicting click frequency of advertisement at advertising position | |
CN111651722B (en) | Risk assessment method and platform for advertisement putting effect of advertiser | |
US20120253927A1 (en) | Machine learning approach for determining quality scores | |
CN105825297A (en) | Markov-model-based position prediction method | |
WO2011035156A2 (en) | Advertisee-history-based bid generation system and method for multi-channel advertising | |
US20120253945A1 (en) | Bid traffic estimation | |
CN103207948A (en) | Wind farm anemograph wind speed missing data interpolation method based on wind speed correlation | |
CN109145446B (en) | Structural damage identification method based on modal strain energy and convolutional neural network | |
WO2019196427A1 (en) | Method and device for optimizing backup on the basis of supporting fault event constrained unit commitment | |
CN103761266A (en) | Click rate predicting method and system based on multistage logistic regression | |
CN108133390A (en) | For predicting the method and apparatus of user behavior and computing device | |
US20160055494A1 (en) | Booking based demand forecast | |
JP2006092058A (en) | Flow rate estimation device | |
CN111986490A (en) | Road condition prediction method and device, electronic equipment and storage medium | |
CN116307215A (en) | Load prediction method, device, equipment and storage medium of power system | |
CN111353620A (en) | Method, device and equipment for constructing network point component prediction model and storage medium | |
CN103617146A (en) | Machine learning method and device based on hardware resource consumption | |
US20140379460A1 (en) | Real-time updates to digital marketing forecast models | |
CN110135114B (en) | Method and device for determining minimum navigation water level of river design | |
CN105654326A (en) | Information processing system and information processing method | |
JP6135454B2 (en) | Estimation program, estimation apparatus, and estimation method | |
CN104731788A (en) | Processing method and equipment for promote information |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |