CN106296247A - The online sort method of network information resource and device - Google Patents
The online sort method of network information resource and device Download PDFInfo
- Publication number
- CN106296247A CN106296247A CN201510276519.7A CN201510276519A CN106296247A CN 106296247 A CN106296247 A CN 106296247A CN 201510276519 A CN201510276519 A CN 201510276519A CN 106296247 A CN106296247 A CN 106296247A
- Authority
- CN
- China
- Prior art keywords
- network information
- orientation type
- exposure
- crowd
- under
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of online sort method of network information resource and device.Said method comprising the steps of: obtain the network information estimates light exposure and actual exposure amount;Light exposure is estimated and predictor error ratio that actual exposure measures under orientation type according to described;The crowd's discreet after the crowd's discreet obtained in advance obtains the rectification under described orientation type is corrected according to the predictor error ratio under described orientation type;Exposure probability according to the network information under the crowd's discreet described orientation type of calculating after described rectification;Obtain expected revenus and expected revenus distribution probability density that the network information under described orientation type exposes;The expected revenus distribution probability density exposing probability and correspondence according to the network information under described orientation type obtains the offset parameter of the network information under described orientation type;According to described expected revenus and offset parameter, the network information is carried out at line ordering.Improve the accuracy of resource distribution.
Description
Technical field
The present invention relates to field of information processing, particularly relate to a kind of online sort method of network information resource and
Device.
Background technology
Along with the development of Internet technology, the Internet provides the various network information service.Network is believed
Breath service can include advertising service, Item Information displaying service etc..The network information service takes with Internet advertising
As a example by business, it provides multiple advertisement form, distinguishes according to charging way and can be divided into effect advertisement and displaying
Advertisement.The Internet display advertisement is to be sold by contract greatly, i.e. before advertising platform and advertiser
Reach an agreement, sell special time period, specific user, particular display amount.
Traditional advertising resource allocation algorithm be use HWM (High Water Mark) algorithm, its be divided into from
Line and online two stages, be that contract advertisement distributes service priority according to the difficult satisfaction degree of advertising resource,
Then being followed successively by advertisement according to service priority and calculate allotment ratio, HWM algorithm requires that exposure crowd estimates non-
Often accurately, and this precondition is difficult to meet, such as run into technical dates (such as weekend, festivals or holidays etc.),
The situations such as media side's flow drastic change, are difficult to accurately estimate in advance crowd's quantity on the same day, thus cause distributing not
Accurately.
Summary of the invention
Based on this, it is necessary to the problem being difficult to for traditional network information resource allocation algorithm accurately distribute,
A kind of online sort method of network information resource is provided, the accuracy of resource distribution can be improved.
Additionally, there is a need to provide a kind of online collator of network information resource, resource distribution can be improved
Accuracy.
A kind of online sort method of network information resource, comprises the following steps:
Obtain the network information estimates light exposure and actual exposure amount;
Light exposure is estimated and predictor error ratio that actual exposure measures under orientation type according to described;
Correct, according to the predictor error ratio under described orientation type, the crowd's discreet obtained in advance and obtain institute
State the crowd's discreet after the rectification under orientation type;
Exposure according to the network information under the crowd's discreet described orientation type of calculating after described rectification is general
Rate;
Obtain under described orientation type the network information exposure expected revenus and expected revenus distribution probability close
Degree;
Exposure probability and the expected revenus distribution probability of correspondence according to the network information under described orientation type are close
Degree obtains the offset parameter of the network information under described orientation type;
According to described expected revenus and offset parameter, the network information is carried out at line ordering.
A kind of online collator of network information resource, including:
Light exposure obtaining module, estimates light exposure and actual exposure amount for obtain the network information;
Predictor error ratio acquisition module, is used for described in basis estimating light exposure and actual exposure measures orientation
Predictor error ratio under type;
Rectification module, for correcting the crowd obtained in advance according to the predictor error ratio under described orientation type
Discreet obtains the crowd's discreet after the rectification under described orientation type;
Exposure probability evaluation entity, for calculating described orientation class according to the crowd's discreet after described rectification
The exposure probability of the network information under type;
Expected revenus acquisition module, for obtaining the expected revenus of the network information exposure under described orientation type
And expected revenus distribution probability density;
Offset parameter acquisition module, for according to the exposure probability of the network information under described orientation type and right
The expected revenus distribution probability density answered obtains the offset parameter of each bar network information under described orientation type;
Order module, for carrying out at line ordering the network information according to described expected revenus and offset parameter.
The online sort method of above-mentioned network information resource and device, estimate light exposure and reality by the network information
Border light exposure obtains predictor error ratio, corrects crowd's discreet according to predictor error ratio, further according to people
Group's discreet obtains the exposure probability of the network information, according to expected revenus distribution probability density, exposure probability
Obtain the offset parameter of the network information, according to expected revenus and offset parameter, the network information carried out at line ordering,
Because crowd's discreet is corrected, improve the accuracy of crowd's discreet, so calculated
The exposure probability of the network information, offset parameter are more accurate, further according to expected revenus and offset parameter to network
Information is ranked up, the exposure effect of maximization network information, improves the accuracy of resource distribution, every net
The exposure of network information all considers the clicking rate of estimating of the network information, and the network information all selects to click in potential user
The user that rate is the highest, improves the overall clicking rate of the network information.
Accompanying drawing explanation
Fig. 1 is the interior of the server of the online sort method of operational network information resources and device in an embodiment
Portion's structural representation;
Fig. 2 is the block schematic illustration of network information resource distribution;
Fig. 3 is the schematic flow sheet of the online sort method of network information resource in an embodiment;
Fig. 4 is the schematic diagram corrected in adjacent two moment;
Fig. 5 is the schematic diagram of network information allocation proportion;
Fig. 6 is ECPM distribution schematic diagram;
Fig. 7 is the structural representation of the online collator of network information resource in an embodiment.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with accompanying drawing and reality
Execute example, the present invention is further elaborated.Only should be appreciated that specific embodiment described herein
Only in order to explain the present invention, it is not intended to limit the present invention.
Fig. 1 is the interior of the server of the online sort method of operational network information resources and device in an embodiment
Portion's structural representation.As it is shown in figure 1, this server includes that the processor connected by system bus, storage are situated between
Matter, internal memory and network interface.Wherein, the storage medium storage of this server have operating system, data base and
The online collator of network information resource, in data base, storage has network information resource, this network information resource
Online collator is for realizing being applicable to the online sort method of a kind of network information resource of server.These clothes
The processor of business device is used for providing calculating and control ability, supports the operation of whole server.This server
The operation inside saving as the online collator of the network information resource in storage medium provides environment.This server
Network interface is connected communication with outside terminal by network for according to this, such as receives the displaying that terminal sends
Ask and return network information resource etc. to terminal.Server can be with independent server or multiple
The server cluster of server composition realizes.It will be understood by those skilled in the art that the structure shown in Fig. 1,
It is only the block diagram of the part-structure relevant to the application scheme, is not intended that and the application scheme is applied to
The restriction of server thereon, concrete server can include than shown in figure more or less of parts,
Or combine some parts, or there is different parts layouts.
Fig. 2 is the block schematic illustration of network information resource distribution.As in figure 2 it is shown, network information resource point
Join framework and include off-line and online two stages.Off-line phase include booking module (order module) and
Allocation module (distribution module), order module is directly thrown in end docking according to the network information, is used for doing
Stock estimates and estimates with expense, and ultimately forms contract order.Wherein, contract order includes that user to purchase
Buy the network information orientation type (such as 20 to 25 years old ages, male), buy exposure quantity (such as one
They 1,000,000 times), the network information resource such as position (browser end) bought throw in the necessary information needed.
Distribution module does assignment of traffic, off-line meter according to light exposure and the crowd's quantity estimated of the main purchase of the network information
Calculate allocative decision.The distribution engine of distribution module by contract order, the network information estimate light exposure and in real time
Light exposure is input to allocative decision generator, periodically calculates, the offset parameter of the output network information and statistics CPC
(Cost Per Click, single click on expense), as the allocative decision of the network information, is pushed to online net
The network information server that network information resources are play.The Serve (service module) of network information server loads
The offset parameter of the network information and statistics CPC, estimating clicking rate and adding up taking advantage of of CPC according to the network information
Amass and as the ranking factor of the network information, the network information is ranked up with offset parameter, thus according to sequence knot
Fruit shows the network information.
The network information is to sign contract under a kind of line, the network information main purchase special time period, specific crowd,
Particular exposure number of times, with every thousand times show charging, on network information platform line throw in network information form.
The network information can be advertising message, Item Information, news information, action message etc..Advertising message can be net
The advertising message etc. shown on network platform.Item Information can be the merchandise news etc. shown on shopping platform.Newly
News information can be the news information searched at search engine or the news shown on news information distribution platform
Information etc..Action message can be the action message etc. promoted in the network platform, such as, promotion message of soliciting articles, comment
Select promotion message, micro-film promotion message etc..
Offset parameter is used to adjust the exposure of the network information.
Fig. 3 is the schematic flow sheet of the online sort method of network information resource in an embodiment.In Fig. 2
The online sort method of network information resource runs under the framework in Fig. 2.As it is shown on figure 3, a kind of network letter
The breath online sort method of resource, comprises the following steps:
Step 302, obtain the network information estimates light exposure and actual exposure amount.
In one embodiment, the step estimating light exposure obtaining the network information includes: statistics orientation type
Under the scheduled time in history show number of times;According to this history show number of times estimation the network information estimate exposure
Light quantity.
Specifically, orientation type refers to the classification in the network information under orientation crowd set in advance, such as man 1
Year, 2 years old or male 1 to 6 year old etc..The scheduled time can set as required, such as 5 days, 7 days or one month etc..
History in the scheduled time can be shown the number of times light exposure of estimating as the network information, or by the scheduled time
In history show that number of times is weighted average estimating light exposure etc. as the network information.
Step 304, estimates light exposure according to this and predictor error ratio that actual exposure measures under orientation type
Example.
Specifically, can predictor error ratio of interval calculation to schedule, this predetermined time interval can
Be 5 minutes, 10 minutes etc..Every time according to the real-time light exposure of network information server and estimate light exposure with
And the orientation type of the network information corrects the deviation estimating light exposure under each orientation type.Concrete rectification is public
Formula such as formula (1):
In formula (1), ujRepresent network information j actual exposure amount, djRepresent that network information j is during this period of time
Estimate light exposure, σitRepresent the orientation type i exposure bias in t, i.e. predictor error under orientation type
Ratio.
Step 306, corrects the crowd's discreet obtained in advance according to the predictor error ratio under this orientation type
Obtain the crowd's discreet after the rectification under this orientation type.
In one embodiment, step 306 includes: by time attenuation function to the net under this orientation type
The predictor error ratio of network information is done and is expected, obtains the total predictor error ratio under this orientation type;According to this
Total predictor error ratio under orientation type carries out rectification to the crowd's discreet obtained in advance and obtains this orientation
Crowd's discreet after rectification under type.
Specifically, σi=Ef(t)[σit] (2)
F (t)=λ e-λt (3)
In formula (2) (3), f (t) is time attenuation function, represents by exponential herein, and λ value can root
Determining according to practical situation, the intensity of express time decay, the least expression of λ is the biggest to the dependence of historical data, this
In embodiment, λ can value be 0.5;T represents apart from the current time, and its span can be 0 to 48,
To be a unit half an hour.σiRepresenting the total predictor error ratio under orientation type i, it is by each time period
Interior predictor error ratio is done expectation with time attenuation function and is obtained.
Fig. 4 is the schematic diagram corrected in adjacent two moment.As shown in Figure 4, in the T-1 moment, by contract order
Obtaining allocative decision with estimating after light exposure is input to distribution engine, online network information server is according to distribution
After scheme is allocated, then obtain actual exposure amount, by the distribution engine in actual exposure amount input T moment,
The light exposure deviation in T moment is corrected, the like.
According to total predictor error ratio σ calculatediThe crowd's discreet obtained in advance is corrected,
Rectification formula such as formula (4):
In formula (4), SupplyiRepresent the crowd's discreet obtained in advance,Represent the people after correcting
Group's discreet.
Step 308, calculates the exposure of the network information under this orientation type according to the crowd's discreet after this rectification
Light probability.
In one embodiment, step 308 includes: obtain the orientation crowd's discreet meeting the network information;
The priority of the network information is determined according to satisfied orientation crowd's discreet;Depend on from high to low according to priority
The exposure probability of the network information under this orientation type of secondary calculating.
Such as, the number that first network information meets is 100W (ten thousand people)+80W, altogether 180W people;
The number that second network information meets is 80W+120W, altogether 200W people.Then first network information
Priority is higher than the priority of second network information.
After determining the priority of each bar network information, calculate each network letter the most successively according to priority
The allocation proportion of breath, i.e. exposes probability.It is 150/ (100+80) as calculated the allocation proportion of first network information
=0.833;When calculating the allocation proportion of second network information, first the crowd distributing to first network information
Quantity is removed, and remains 100* (1-0.833)=16.7W, second orientation people after first orientation crowd's distribution
Group's residue 80* (1-0.833)=13.36W, the allocation proportion of second network information is 0.167+
(80-13.36-0.167*120)/120=0.5553.
Fig. 5 is the schematic diagram of network information allocation proportion.As it is shown in figure 5, first network information estimates people
Group's quantity is 150W, and it is 80W that second network information estimates crowd's quantity, and first orientation crowd's quantity is
100W, second orientation crowd's quantity is 80W, and the 3rd orientation crowd's quantity is 120W, meets first
Crowd's quantity of the individual network information is first orientation crowd and second orientation crowd, altogether 180W, meets the
Crowd's quantity of two network informations is second orientation crowd and the 3rd orientation crowd, altogether 200W, first
The priority of the individual network information is higher than the priority of second network information, the distribution ratio of first network information
It is 0.833 that example is calculated, and the allocation proportion of second network information is 0.5553.
The specific algorithm of the exposure probability calculating the network information is as follows:
1) for all orientation type i, residue supply r is initializedi=si;
2) for network information j, according to prioritization, traversal is a) and b);
A) solve equation below (5) and obtain αj;
If cannot solve, then set αj=1.
B) for all i ∈ Γ (j), r is updatedi=ri-min{ri,siαj}。
Because of riRepresent remaining crowd can light exposure, be after a network information is assigned with light exposure every time, this
Remaining can light exposure can reduce, so iteration updates r every timei。
Enter in addition it is also possible to directly use the actual exposure amount of feedback on line that the network information is estimated light exposure
Row sum-equal matrix, when actual exposure amount is more than when estimating light exposure, reduces the exposure probability of map network resource, instead
Increase exposure probability.
Step 310, the expected revenus and the expected revenus that obtain the network information exposure under this orientation type are distributed general
Rate density.
In one embodiment, step 310 includes: obtains the statistics of the network information and single clicks on expense and each
The network information of secondary exposure estimate clicking rate;Calculate this and single click on expense and the product estimating clicking rate obtains
The expected revenus every time exposed to the network information;Add up according to the expected revenus that the network information exposes every time,
Obtain the expected revenus distribution probability density of the network information.
Specifically, when every network information just starts to reach the standard grade, the exposure probability of the network information is directly used to do
Select on line, be ranked up exposure according to the exposure probability of the network information.
When, after the exposure frequency and number of clicks of accumulative predetermined quantity, statistics CPC calculating the network information is (single
Secondary clicking cost), represent with stat_cpc, wherein:
Stat_cpc=cost/click (6)
In formula (6), cost represents the expense in this network information Preset Time, and click represents Preset Time
Touching quantity.The exposure frequency of predetermined quantity may be selected to be 10000 times, and number of clicks can be 100 times.
Calculate each network information exposure ECPM (Expect Cost Per Thousand Impressions,
1000 times the network information shows desired income), computing formula such as formula (7).
ECPM=stat_cpc*pCtr (7)
Wherein, pCtr represent when time exposure the network information estimate clicking rate.
The expected revenus that each bar network information exposes every time is added up, obtains the expectation of each bar network information
Benefit distribution probability density fcontract(Ecpm)。
Step 312, according to exposure probability and the expected revenus distribution of correspondence of the network information under this orientation type
Probability density obtains the offset parameter of the network information under this orientation type.
In one embodiment, step 312 includes: add up the network information of non-agreement under this orientation type
Expected revenus distribution probability density;Exposure probability according to the network information under this orientation type, non-agreement
The expected revenus distribution probability density of the network information and the expected revenus distribution probability of each bar network information close
The offset parameter of the network information that degree is calculated under this orientation type by cumulative probability function.
Specifically, the expected revenus distribution probability density of the non-agreement network information under statistics orientation type
fnon-contract(Ecpm), obtained by log statistic, such as ECPM at the exposure frequency of [0 yuan, 0.1 yuan],
ECPM is at the exposure frequency etc. of [0.1 yuan, 0.2 yuan].The network information of non-agreement refer to except contract it
The outer network information.
The offset parameter bias of the network information, computing formula such as formula (8) is asked for by cumulative probability function F.
F(Ecpmcontract+Ecpmnon-contract+ bias)=α (8)
In formula (8), α represents that the network information exposes probability.
In one embodiment, the expected revenus of the network information adding up the non-agreement under this orientation type is distributed
The step of probability density includes: add up the expectation of the network information of non-agreement under this orientation type from daily record
Benefit distribution probability density.
Step 314, desirably the network information is carried out at line ordering by income and offset parameter.
Using expected revenus and offset parameter as ranking factor, the network information is ranked up.
Fig. 6 is ECPM distribution schematic diagram.As shown in Figure 6, by the ECPM of bias Translate network information
Distribution makes the exposure probability in the competition naturally of the network information in the ranking be α, reaches the purpose of guarantor's amount.61
For overall ECPM distribution, 62 is the ECPM distribution of the network information, and 63 is the ECPM after network information adjustment
Distribution.
The online sort method of above-mentioned network information resource, estimates light exposure and actual exposure by the network information
Measure predictor error ratio, correct crowd's discreet according to predictor error ratio, estimate further according to crowd
Quantity obtains the exposure probability of the network information, obtains net according to expected revenus distribution probability density, exposure probability
The offset parameter of network information, carries out at line ordering to the network information, because of right according to expected revenus and offset parameter
Crowd's discreet is corrected, and improves the accuracy of crowd's discreet, and then calculated network
The exposure probability of information, offset parameter are more accurate, further according to expected revenus and offset parameter to the network information
Being ranked up, the exposure effect of maximization network information, the exposure of every network information all considers the network information
Estimate clicking rate, the network information all selects the user that in potential user, clicking rate is the highest, improve network letter
The overall clicking rate of breath.
Additionally, employing orientation type, it is achieved that the fine differentiation to user, distributes the network information from probability
Selection is improved to tournament selection, maximizes user and is worth, promotes overall result of broadcast.
In order to further illustrate the online sort method of network information resource, with network information resource in line ordering side
Method is applied to be described in detail as a example by advertising message sorts online.The online sort method of network information resource
Carry out advertising message including in the process of line ordering:
(1) obtain advertising message estimates light exposure and actual exposure amount.
(2) light exposure is estimated according to this and predictor error ratio that actual exposure measures under orientation type.
(3) by time attenuation function, the predictor error ratio of the advertising message under this orientation type is done the phase
Hope, obtain the total predictor error ratio under this orientation type.
(3) correct, according to the total predictor error ratio under this orientation type, the crowd's discreet obtained in advance to obtain
The crowd's discreet after rectification under this orientation type.
(4) exposure of advertising message under this orientation type is calculated according to the crowd's discreet after this rectification general
Rate.
Specifically, acquisition meets orientation crowd's discreet of advertising message;Pre-according to satisfied orientation crowd
Estimate quantity and determine the priority of advertising message;Calculate the most successively under this orientation type according to priority
The exposure probability of advertising message.
(5) obtain under this orientation type advertising message exposure expected revenus and expected revenus distribution probability close
Degree
Specifically, obtain the statistics of advertising message and single click on estimating of advertising message that expense and each time expose
Clicking rate;Calculate this and single click on expense and the product estimating clicking rate obtains the phase that advertising message exposes every time
Hope income;Adding up according to the expected revenus that advertising message exposes every time, the expectation obtaining advertising message is received
Benefit distribution probability density.
(6) the expected revenus distribution probability density of non-contract advertising message under this orientation type is added up.
Specifically, non-contract advertising message refers to the advertising message in addition to signing contract.
(7) according to the exposure probability of the advertising message under this orientation type, the expectation of the advertising message of non-agreement
The expected revenus distribution probability density of benefit distribution probability density and each bar advertising message passes through accumulated probability letter
The offset parameter of the advertising message that number is calculated under this orientation type.
(8) desirably advertising message is carried out at line ordering by income and offset parameter.
This sentence the online sort method of network information resource be applied to advertising message sequence be described, it is also
Can be applicable to Item Information, news information, action message etc. at line ordering, its processing procedure is identical,
This repeats no more.
Fig. 7 is the structural representation of the online collator of network information resource in an embodiment.Such as Fig. 7 institute
Showing, a kind of online collator of network information resource, including light exposure obtaining module 710, predictor error ratio
Acquisition module 720, rectification module 730, exposure probability evaluation entity 740, expected revenus acquisition module 750,
Offset parameter acquisition module 760 and order module 770.Wherein:
Light exposure obtaining module 710 estimates light exposure and actual exposure amount for obtain the network information.
In the present embodiment, light exposure obtaining module 710 is additionally operable to add up in the scheduled time under orientation type
History show number of times, and according to this history show number of times estimation the network information estimate light exposure.
Specifically, orientation type refers to the classification in the network information under orientation crowd set in advance, such as man 1
Year, 2 years old or male 1 to 6 year old etc..The scheduled time can set as required, such as 5 days, 7 days or one month etc..
History in the scheduled time can be shown the number of times light exposure of estimating as the network information, or by the scheduled time
In history show that number of times is weighted average estimating light exposure etc. as the network information.
Predictor error ratio acquisition module 720 is used for estimating light exposure according to this and actual exposure measures orientation
Predictor error ratio under type.
Specifically, can predictor error ratio of interval calculation to schedule, this predetermined time interval can
Be 5 minutes, 10 minutes etc..Every time according to the real-time light exposure of network information server and estimate light exposure with
And the orientation type of the network information corrects the deviation estimating light exposure under each orientation type.Concrete rectification is public
Formula such as formula (1):
In formula (1), ujRepresent network information j actual exposure amount, djRepresent that network information j is during this period of time
Estimate light exposure, σitRepresent the orientation type i exposure bias in t, i.e. predictor error under orientation type
Ratio.
Rectification module 730 for correcting the crowd obtained in advance according to the predictor error ratio under this orientation type
Discreet obtains the crowd's discreet after the rectification under this orientation type.
In the present embodiment, rectification module 730 is additionally operable to by each under this orientation type of time attenuation function
The predictor error ratio of the bar network information is done and is expected, obtains the total predictor error ratio under this orientation type;Root
According to the total predictor error ratio under this orientation type, the crowd's discreet obtained in advance is carried out rectification to be somebody's turn to do
Crowd's discreet after rectification under orientation type.
Specifically, σi=Ef(t)[σit] (2)
F (t)=λ e-λt (3)
In formula (2) (3), f (t) is time attenuation function, represents by exponential herein, and λ value can root
Determining according to practical situation, the intensity of express time decay, the least expression of λ is the biggest to the dependence of historical data, this
In embodiment, λ can value be 0.5;T represents apart from the current time, and its span can be 0 to 48,
To be a unit half an hour.σiRepresenting the total predictor error ratio under orientation type i, it is by each time period
Interior predictor error ratio is done expectation with time attenuation function and is obtained.
According to total predictor error ratio σ calculatediThe crowd's discreet obtained in advance is corrected,
Rectification formula such as formula (4):
In formula (4), SupplyiRepresent the crowd's discreet obtained in advance,Represent the people after correcting
Group's discreet.
Exposure probability evaluation entity 740 is for calculating this orientation type according to the crowd's discreet after this rectification
Under the exposure probability of the network information.
In the present embodiment, exposure probability evaluation entity 740 is additionally operable to obtain the orientation crowd meeting the network information
Discreet;The priority of the network information is determined according to satisfied orientation crowd's discreet;According to priority
Calculate the exposure probability of the network information under this orientation type the most successively.
Such as, the number that first network information meets is 100W (ten thousand people)+80W, altogether 180W people;
The number that second network information meets is 80W+120W, altogether 200W people.Then first network information
Priority is higher than the priority of second network information.
The detailed process of the exposure probability calculating the network information is as follows:
1) for all orientation type i, residue supply r is initializedi=si;
2) for network information j, according to prioritization, traversal is a) and b);
A) solve equation below (5) and obtain αj;
If cannot solve, then set αj=1.
B) for all i ∈ Γ (j), r is updatedi=ri-min{ri,siαj}。
Because of riRepresent remaining crowd can light exposure, be after a network information is assigned with light exposure every time, this
Remaining can light exposure can reduce, so iteration updates r every timei。
Expected revenus acquisition module 750 is for obtaining the expected revenus of the network information exposure under this orientation type
Distribution probability density.
In the present embodiment, it is desirable to income acquisition module 750 is additionally operable to obtain the statistics of the network information and single clicks on
Expense and each time exposure the network information estimate clicking rate;Calculate this single click on expense and estimate clicking rate
Product obtain the expected revenus that the network information exposes every time;The expected revenus every time exposed according to the network information
Add up, obtain the expected revenus distribution probability density of the network information.
Specifically, when every network information just starts to reach the standard grade, the exposure probability of the network information is directly used to do
Select on line, be ranked up exposure according to the exposure probability of the network information.
When, after the exposure frequency and number of clicks of accumulative predetermined quantity, statistics CPC calculating the network information is (single
Secondary clicking cost), represent with stat_cpc, wherein:
Stat_cpc=cost/click (6)
In formula (6), cost represents the expense in this network information Preset Time, and click represents Preset Time
Touching quantity.The exposure frequency of predetermined quantity may be selected to be 10000 times, and number of clicks can be 100 times.
Calculate each network information exposure ECPM (Expect Cost Per Thousand Impressions,
1000 times the network information shows desired income), computing formula such as formula (7).
ECPM=stat_cpc*pCtr (7)
Wherein, pCtr represent when time exposure the network information estimate clicking rate.
The expected revenus that each bar network information exposes every time is added up, obtains the expectation of each bar network information
Benefit distribution probability density fcontract(Ecpm)。
Offset parameter acquisition module 760 is for the exposure probability according to the network information under this orientation type and right
The expected revenus distribution probability density answered obtains the offset parameter of the network information under this orientation type.
In the present embodiment, the non-agreement that offset parameter acquisition module 760 is additionally operable to add up under this orientation type
The expected revenus distribution probability density of the network information;Exposure probability according to the network information under this orientation type,
The expected revenus distribution probability density of the network information of non-agreement and the expected revenus distribution probability of the network information
The offset parameter of the network information that density is calculated under this orientation type by cumulative probability function.
Specifically, the expected revenus distribution probability density of the network information of non-agreement under statistics orientation type
fnon-contract(Ecpm), obtained by log statistic, such as ECPM at the exposure frequency of [0 yuan, 0.1 yuan],
ECPM is at the exposure frequency etc. of [0.1 yuan, 0.2 yuan].
The offset parameter bias of the network information, computing formula such as formula (8) is asked for by cumulative probability function F.
F(Ecpmcontract+Ecpmnon-contract+ bias)=α (8)
In formula (8), α represents that the network information exposes probability.
Order module 770 is for carrying out at line ordering the network information according to this expected revenus and offset parameter.
Specifically, it would be desirable to the network information, as ranking factor, is ranked up by income and offset parameter.
The online collator of above-mentioned network information resource, estimates light exposure and actual exposure by the network information
Measure predictor error ratio, correct crowd's discreet according to predictor error ratio, estimate further according to crowd
Quantity obtains the exposure probability of the network information, obtains net according to expected revenus distribution probability density, exposure probability
The offset parameter of network information, carries out at line ordering to the network information, because of right according to expected revenus and offset parameter
Crowd's discreet is corrected, and improves the accuracy of crowd's discreet, and then calculated network
The exposure probability of information, offset parameter are more accurate, further according to expected revenus and offset parameter to the network information
Being ranked up, the exposure effect of maximization network information, the exposure of every network information all considers the network information
Estimate clicking rate, the network information all selects the user that in potential user, clicking rate is the highest, improve network letter
The overall clicking rate of breath.
One of ordinary skill in the art will appreciate that all or part of flow process realizing in above-described embodiment method,
Can be by computer program and complete to instruct relevant hardware, described program can be stored in one non-easily
In the property lost computer read/write memory medium, this program is upon execution, it may include such as the enforcement of above-mentioned each method
The flow process of example.Wherein, described storage medium can be magnetic disc, CD, read-only store-memory body (Read-Only
Memory, ROM) etc..
Embodiment described above only have expressed the several embodiments of the present invention, and it describes more concrete and detailed,
But therefore can not be interpreted as the restriction to the scope of the claims of the present invention.It should be pointed out that, for this area
Those of ordinary skill for, without departing from the inventive concept of the premise, it is also possible to make some deformation and
Improving, these broadly fall into protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be with appended
Claim is as the criterion.
Claims (16)
1. the online sort method of network information resource, comprises the following steps:
Obtain the network information estimates light exposure and actual exposure amount;
Light exposure is estimated and predictor error ratio that actual exposure measures under orientation type according to described;
Correct, according to the predictor error ratio under described orientation type, the crowd's discreet obtained in advance and obtain institute
State the crowd's discreet after the rectification under orientation type;
Exposure according to the network information under the crowd's discreet described orientation type of calculating after described rectification is general
Rate;
Obtain under described orientation type the network information exposure expected revenus and expected revenus distribution probability close
Degree;
Exposure probability and the expected revenus distribution probability of correspondence according to the network information under described orientation type are close
Degree obtains the offset parameter of the network information under described orientation type;
According to described expected revenus and offset parameter, the network information is carried out at line ordering.
Method the most according to claim 1, it is characterised in that described according under described orientation type
Predictor error ratio corrects the people after the crowd's discreet obtained in advance obtains the rectification under described orientation type
The step of group's discreet includes:
By time attenuation function the predictor error ratio of the network information under described orientation type done and expects,
Obtain the total predictor error ratio under described orientation type;
According to the total predictor error ratio under described orientation type, the crowd's discreet obtained in advance is rectified
Just obtaining the crowd's discreet after the rectification under described orientation type.
Method the most according to claim 1, it is characterised in that estimate according to the crowd after described rectification
The step of the exposure probability that quantity calculates the network information under described orientation type includes:
Obtain the orientation crowd's discreet meeting the network information;
The priority of the network information is determined according to satisfied orientation crowd's discreet;
The exposure probability of the network information under described orientation type is calculated the most successively according to priority.
Method the most according to claim 1, it is characterised in that under the described orientation type of described acquisition
The expected revenus of network information exposure and the step of expected revenus distribution probability density include:
Obtain that the statistics of the network information single clicks on the network information that expense and each time expose estimates clicking rate;
The product single clicing on expense described in calculating and estimate clicking rate obtains the expectation that the network information exposes every time
Income;
Adding up according to the expected revenus that the network information exposes every time, the expected revenus obtaining the network information is divided
Cloth probability density.
Method the most according to claim 4, it is characterised in that described according under described orientation type
The exposure probability of the network information and the expected revenus distribution probability density of correspondence obtain the net under described orientation type
The step of the offset parameter of network information includes:
Add up the expected revenus distribution probability density of the network information of non-agreement under described orientation type;
The expectation of the network information exposing probability, non-agreement according to the network information under described orientation type is received
The expected revenus distribution probability density of benefit distribution probability density and the network information is calculated by cumulative probability function
Obtain the offset parameter of the network information under described orientation type.
Method the most according to claim 5, it is characterised in that under the described orientation type of described statistics
The step of the expected revenus distribution probability density of the network information of non-agreement includes:
The expected revenus distribution probability of the network information adding up non-agreement under described orientation type from daily record is close
Degree.
Method the most according to claim 1, it is characterised in that the described acquisition network information estimate exposure
The step of light quantity includes:
The history in the scheduled time under statistics orientation type shows number of times;
According to described history show number of times estimation the network information estimate light exposure.
Method the most according to any one of claim 1 to 7, it is characterised in that the described network information
For advertising message or Item Information or news information.
9. the online collator of network information resource, it is characterised in that including:
Light exposure obtaining module, estimates light exposure and actual exposure amount for obtain the network information;
Predictor error ratio acquisition module, is used for described in basis estimating light exposure and actual exposure measures orientation
Predictor error ratio under type;
Rectification module, for correcting the crowd obtained in advance according to the predictor error ratio under described orientation type
Discreet obtains the crowd's discreet after the rectification under described orientation type;
Exposure probability evaluation entity, for calculating described orientation class according to the crowd's discreet after described rectification
The exposure probability of the network information under type;
Expected revenus acquisition module, for obtaining the expected revenus of the network information exposure under described orientation type
And expected revenus distribution probability density;
Offset parameter acquisition module, for according to the exposure probability of the network information under described orientation type and right
The expected revenus distribution probability density answered obtains the offset parameter of the network information under described orientation type;
Order module, for carrying out at line ordering the network information according to described expected revenus and offset parameter.
Device the most according to claim 9, it is characterised in that described rectification module is additionally operable to pass through
The predictor error ratio of the network information under described orientation type is done by time attenuation function to be expected, obtains described
Total predictor error ratio under orientation type, and according to the total predictor error ratio pair under described orientation type
The crowd that the crowd's discreet obtained in advance is carried out after correcting the rectification obtained under described orientation type estimates number
Amount.
11. devices according to claim 9, it is characterised in that described exposure probability evaluation entity is also
For obtaining the orientation crowd's discreet meeting the network information, true according to satisfied orientation crowd's discreet
Determine the priority of the network information, and calculate the net under described orientation type the most successively according to priority
The exposure probability of network information.
12. devices according to claim 9, it is characterised in that described expected revenus acquisition module is also
For obtain the statistics of the network information single click on expense and each time exposure the network information estimate clicking rate,
The product single clicing on expense described in calculating and estimate clicking rate obtains the expectation receipts that the network information exposes every time
Benefit, and the expected revenus every time exposed according to the network information adds up, the expectation obtaining the network information is received
Benefit distribution probability density.
13. devices according to claim 12, it is characterised in that described offset parameter acquisition module is also
For adding up the expected revenus distribution probability density of the network information of the non-agreement under described orientation type, and
The expected revenus of the network information exposing probability, non-agreement according to the network information under described orientation type is divided
The expected revenus distribution probability density of cloth probability density and the network information is calculated by cumulative probability function
The offset parameter of the network information under described orientation type.
14. devices according to claim 13, it is characterised in that described offset parameter acquisition module is also
Close for adding up the expected revenus distribution probability of the network information of the non-agreement under described orientation type from daily record
Degree.
15. devices according to claim 9, it is characterised in that described light exposure obtaining module is also used
The history in the scheduled time under statistics orientation type shows number of times, and shows number of times according to described history
That estimates the network information estimates light exposure.
16. according to the device according to any one of claim 9 to 15, it is characterised in that described network is believed
Breath is advertising message or Item Information or news information.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510276519.7A CN106296247B (en) | 2015-05-26 | 2015-05-26 | Network information resource online ordering method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510276519.7A CN106296247B (en) | 2015-05-26 | 2015-05-26 | Network information resource online ordering method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106296247A true CN106296247A (en) | 2017-01-04 |
CN106296247B CN106296247B (en) | 2020-12-22 |
Family
ID=57635331
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510276519.7A Active CN106296247B (en) | 2015-05-26 | 2015-05-26 | Network information resource online ordering method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106296247B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107105031A (en) * | 2017-04-20 | 2017-08-29 | 北京京东尚科信息技术有限公司 | Information-pushing method and device |
CN108965936A (en) * | 2017-05-17 | 2018-12-07 | 腾讯科技(深圳)有限公司 | A kind of control method for playing back of media information, server and computer storage medium |
CN108959324A (en) * | 2017-05-26 | 2018-12-07 | 腾讯科技(深圳)有限公司 | Predictor method, device and the storage medium of multimedia show resource quantity in stock |
CN109657132A (en) * | 2017-10-11 | 2019-04-19 | 腾讯科技(深圳)有限公司 | Recommendation information cost control method, device, computer equipment and storage medium |
CN110599250A (en) * | 2019-09-09 | 2019-12-20 | 腾讯科技(深圳)有限公司 | Resource playing control method and device and computer storage medium |
CN112150182A (en) * | 2019-06-28 | 2020-12-29 | 腾讯科技(深圳)有限公司 | Multimedia file pushing method and device, storage medium and electronic device |
WO2021063223A1 (en) * | 2019-09-30 | 2021-04-08 | 阿里巴巴集团控股有限公司 | Creative script object subscription and traffic allocation method and device, platform, and medium |
CN113516495A (en) * | 2020-09-30 | 2021-10-19 | 腾讯科技(深圳)有限公司 | Information pushing method and device, electronic equipment and computer readable medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102592235A (en) * | 2011-12-28 | 2012-07-18 | 北京品友互动信息技术有限公司 | Internet advertisement serving system |
US20140236710A1 (en) * | 2013-02-19 | 2014-08-21 | Congoo, Llc | On-line advertising valuation |
CN104182801A (en) * | 2013-05-22 | 2014-12-03 | 阿里巴巴集团控股有限公司 | Method and device for predicting website visits |
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 |
CN104424291A (en) * | 2013-09-02 | 2015-03-18 | 阿里巴巴集团控股有限公司 | Method and device for sorting search results |
CN104572734A (en) * | 2013-10-23 | 2015-04-29 | 腾讯科技(深圳)有限公司 | Question recommendation method, device and system |
-
2015
- 2015-05-26 CN CN201510276519.7A patent/CN106296247B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102592235A (en) * | 2011-12-28 | 2012-07-18 | 北京品友互动信息技术有限公司 | Internet advertisement serving system |
US20140236710A1 (en) * | 2013-02-19 | 2014-08-21 | Congoo, Llc | On-line advertising valuation |
CN104182801A (en) * | 2013-05-22 | 2014-12-03 | 阿里巴巴集团控股有限公司 | Method and device for predicting website visits |
CN104424291A (en) * | 2013-09-02 | 2015-03-18 | 阿里巴巴集团控股有限公司 | Method and device for sorting search results |
CN104572734A (en) * | 2013-10-23 | 2015-04-29 | 腾讯科技(深圳)有限公司 | Question recommendation method, device and system |
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 |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107105031A (en) * | 2017-04-20 | 2017-08-29 | 北京京东尚科信息技术有限公司 | Information-pushing method and device |
CN108965936B (en) * | 2017-05-17 | 2021-05-11 | 腾讯科技(深圳)有限公司 | Media information playing control method, server and computer storage medium |
CN108965936A (en) * | 2017-05-17 | 2018-12-07 | 腾讯科技(深圳)有限公司 | A kind of control method for playing back of media information, server and computer storage medium |
CN108959324A (en) * | 2017-05-26 | 2018-12-07 | 腾讯科技(深圳)有限公司 | Predictor method, device and the storage medium of multimedia show resource quantity in stock |
CN108959324B (en) * | 2017-05-26 | 2022-04-15 | 腾讯科技(深圳)有限公司 | Method and device for estimating multimedia display resource inventory and storage medium |
CN109657132A (en) * | 2017-10-11 | 2019-04-19 | 腾讯科技(深圳)有限公司 | Recommendation information cost control method, device, computer equipment and storage medium |
CN112150182A (en) * | 2019-06-28 | 2020-12-29 | 腾讯科技(深圳)有限公司 | Multimedia file pushing method and device, storage medium and electronic device |
CN112150182B (en) * | 2019-06-28 | 2023-08-29 | 腾讯科技(深圳)有限公司 | Multimedia file pushing method and device, storage medium and electronic device |
CN110599250A (en) * | 2019-09-09 | 2019-12-20 | 腾讯科技(深圳)有限公司 | Resource playing control method and device and computer storage medium |
CN110599250B (en) * | 2019-09-09 | 2023-12-19 | 腾讯科技(深圳)有限公司 | Resource playing control method and device and computer storage medium |
WO2021063223A1 (en) * | 2019-09-30 | 2021-04-08 | 阿里巴巴集团控股有限公司 | Creative script object subscription and traffic allocation method and device, platform, and medium |
CN113516495A (en) * | 2020-09-30 | 2021-10-19 | 腾讯科技(深圳)有限公司 | Information pushing method and device, electronic equipment and computer readable medium |
CN113516495B (en) * | 2020-09-30 | 2024-03-08 | 腾讯科技(深圳)有限公司 | Information pushing method, device, electronic equipment and computer readable medium |
Also Published As
Publication number | Publication date |
---|---|
CN106296247B (en) | 2020-12-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106296247A (en) | The online sort method of network information resource and device | |
Cai et al. | Real-time bidding by reinforcement learning in display advertising | |
US9183562B2 (en) | Method and system for determining touchpoint attribution | |
CN111651722B (en) | Risk assessment method and platform for advertisement putting effect of advertiser | |
US8566207B2 (en) | Systems and methods for determining bids for placing advertisements | |
US20170300939A1 (en) | Optimizing promotional offer mixes using predictive modeling | |
US20080082400A1 (en) | Advertisement Campaign Simulator | |
Kumar | Optimization issues in web and mobile advertising: past and future trends | |
US20170091810A1 (en) | Brand engagement touchpoint attribution using brand engagement event weighting | |
US20110191170A1 (en) | Similarity function in online advertising bid optimization | |
US20130268347A1 (en) | Yield management system and method for advertising inventory | |
JP2008529190A (en) | Advertisement management method, shadow campaign system, and advertisement management system | |
US20160210656A1 (en) | System for marketing touchpoint attribution bias correction | |
US20100106605A1 (en) | Inventory allocation with tradeoff between fairness and maximal value of remaining inventory | |
CN105205696A (en) | Internet advertisement real-time bidding multi-advertiser multi-factor flow distribution method | |
US20110191169A1 (en) | Kalman filter modeling in online advertising bid optimization | |
US20140019232A1 (en) | Impression-trend techniques for providing a display advertising supply forecast | |
US20170300832A1 (en) | Cross-channel predictive model | |
CN104035926A (en) | Internet information release method and system | |
US20220122118A1 (en) | Planning device and computer program | |
US11743679B2 (en) | Systems and methods for pacing information delivery to mobile devices | |
US20140257972A1 (en) | Method, computer readable medium and system for determining true scores for a plurality of touchpoint encounters | |
US20140257966A1 (en) | Method , computer readable medium and system for determining weights for attributes and attribute values for a plurality of touchpoint encounters | |
US20220027959A1 (en) | Real-Time Bidding | |
CN110717666A (en) | Market competitiveness evaluation method for enterprise products |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |