CN107886361A - A kind of method and server for assessing ad conversion rates prediction model - Google Patents
A kind of method and server for assessing ad conversion rates prediction model Download PDFInfo
- Publication number
- CN107886361A CN107886361A CN201711120819.1A CN201711120819A CN107886361A CN 107886361 A CN107886361 A CN 107886361A CN 201711120819 A CN201711120819 A CN 201711120819A CN 107886361 A CN107886361 A CN 107886361A
- Authority
- CN
- China
- Prior art keywords
- prediction model
- conversion
- advertisement
- conversion rates
- edition
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention discloses a kind of method, server and computer-readable recording medium for assessing ad conversion rates prediction model, wherein method includes:Ad conversion rates prediction model call request is received, random call new edition ad conversion rates prediction model or old edition ad conversion rates prediction model calculate the conversion ratio of estimating of candidate locations, record model corresponding to candidate locations and call mark;The conversion ratio of estimating of candidate locations is sent to advertising platform;Model calls mark according to corresponding to having exposed advertisement, calculates the second actual advertisement conversion ratio corresponding to the first actual advertisement conversion ratio corresponding to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model;According to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and default assessment strategy, good and bad assessment is carried out to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model.The embodiment of the present invention can improve the good and bad accuracy assessed of ad conversion rates prediction model.
Description
Technical field
The present invention relates to communication technical field, more particularly to a kind of method for assessing ad conversion rates prediction model, service
Device and computer-readable recording medium.
Background technology
Ad conversion rates prediction model is used for advertising platform before advertising display is carried out, to the conversion ratio of advertisement to be presented
(such as clicking rate or download rate) is estimated.When the version of ad conversion rates prediction model has renewal, it usually needs first right
New edition ad conversion rates prediction model carries out test assessment, to determine if that being really better than old edition ad conversion rates estimates mould
Type, then determine whether new edition ad conversion rates prediction model of reaching the standard grade further according to assessment result.
Existing ad conversion rates prediction model appraisal procedure is typically the reality according to new edition ad conversion rates prediction model
When data and the historical data of old edition ad conversion rates prediction model new and old edition ad conversion rates prediction model commented
Estimate, i.e., the ad conversion rates data estimated out during use before according to old edition ad conversion rates prediction model with it is new
Version ad conversion rates prediction model is assessed in the ad conversion rates data that test phase is estimated out in real time, that is to say, that existing
Some ad conversion rates prediction model appraisal procedures when the ad conversion rates prediction model to new and old edition is assessed institute according to
According to be different time dimension data, can so cause the actual good and bad difference of assessment result and new and old edition, assess
As a result it is inaccurate.
The content of the invention
The embodiment of the present invention, which provides, a kind of to be assessed the method for ad conversion rates prediction model, server and computer-readable deposits
Storage media, it is possible to increase the good and bad accuracy assessed of ad conversion rates prediction model.
In a first aspect, the embodiments of the invention provide a kind of method for assessing ad conversion rates prediction model, this method bag
Include:
Ad conversion rates prediction model call request is received, random call new edition ad conversion rates prediction model or old edition are wide
That accuses conversion ratio prediction model calculating candidate locations estimates conversion ratio, records model corresponding to the candidate locations and calls mark;
The conversion ratio of estimating of the candidate locations is sent to advertising platform, makes the advertising platform wide according to the candidate
Accuse estimate conversion ratio and default advertisement exposure strategy is exposed to the candidate locations;
Model calls mark according to corresponding to having exposed advertisement, calculates corresponding to the new edition ad conversion rates prediction model
Second actual advertisement conversion ratio corresponding to first actual advertisement conversion ratio and the old edition ad conversion rates prediction model;
According to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and default assessment strategy,
Good and bad assessment is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model.
Second aspect, the embodiments of the invention provide a kind of server, the server includes being used to perform above-mentioned first party
The unit of the method in face.
The third aspect, the embodiments of the invention provide another server, including processor, input equipment, output equipment
And memory, the processor, input equipment, output equipment and memory are connected with each other, wherein, the memory is used to store
Support server to perform the computer program of the above method, the computer program includes programmed instruction, the processor by with
Put for calling described program to instruct, the method for performing above-mentioned first aspect.
Fourth aspect, the embodiments of the invention provide a kind of computer-readable recording medium, the computer-readable storage
Media storage has computer program, and the computer program includes programmed instruction, and described program instructs when being executed by a processor
Make the method for the above-mentioned first aspect of the computing device.
The embodiment of the present invention is excellent to new edition ad conversion rates prediction model and the progress of old edition ad conversion rates prediction model
During bad assessment, due to after ad conversion rates prediction model call request is received, being random call new edition model or old edition mould
Type is estimated to the conversion ratio of candidate locations, so that new edition model and old edition model are carried out in advance to ad conversion rates
The ad data of institute's foundation belongs to same time dimension when estimating, for example, the ad data belonged in assessment cycle, then advertisement
Platform is estimated after conversion ratio is exposed to advertisement in the advertisement gone out according to new edition model or old edition model pre-estimating, according to having exposed
The model of advertisement, which calls, identifies the new edition model being calculated and each self-corresponding actual advertisement conversion ratio of old edition model also necessarily
It is the data for belonging to same time dimension, according to the data in same time dimension to new and old edition ad conversion rates prediction model
Assessed, it is possible to increase the good and bad accuracy assessed of ad conversion rates prediction model.
Brief description of the drawings
Technical scheme in order to illustrate the embodiments of the present invention more clearly, it is required in being described below to embodiment to use
Accompanying drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the present invention, general for this area
For logical technical staff, on the premise of not paying creative work, other accompanying drawings can also be obtained according to these accompanying drawings.
Fig. 1 is a kind of schematic flow diagram of method for assessing ad conversion rates prediction model provided in an embodiment of the present invention;
Fig. 2 is a kind of exemplary flow of the method for assessment ad conversion rates prediction model that another embodiment of the present invention provides
Figure;
Fig. 3 is a kind of schematic block diagram of server provided in an embodiment of the present invention;
Fig. 4 is a kind of schematic block diagram for server that another embodiment of the present invention provides.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is part of the embodiment of the present invention, rather than whole embodiments.Based on this hair
Embodiment in bright, the every other implementation that those of ordinary skill in the art are obtained under the premise of creative work is not made
Example, belongs to the scope of protection of the invention.
It should be appreciated that ought be in this specification and in the appended claims in use, term " comprising " and "comprising" instruction
Described feature, entirety, step, operation, the presence of element and/or component, but it is not precluded from one or more of the other feature, whole
Body, step, operation, element, component and/or its presence or addition for gathering.
It is also understood that the term used in this description of the invention is merely for the sake of the mesh for describing specific embodiment
And be not intended to limit the present invention.As used in description of the invention and appended claims, unless on
Other situations are hereafter clearly indicated, otherwise " one " of singulative, "one" and "the" are intended to include plural form.
It will be further appreciated that the term "and/or" used in description of the invention and appended claims is
Refer to any combinations of one or more of the associated item listed and be possible to combine, and including these combinations.
As used in this specification and in the appended claims, term " if " can be according to context quilt
Be construed to " when ... " or " once " or " in response to determining " or " in response to detecting ".Similarly, phrase " if it is determined that " or
" if detecting [described condition or event] " can be interpreted to mean according to context " once it is determined that " or " in response to true
It is fixed " or " once detecting [described condition or event] " or " in response to detecting [described condition or event] ".
In the specific implementation, the server described in the embodiment of the present invention is including but not limited to such as with touch sensitive surface
The mobile phone, laptop computer or tablet PC of (for example, touch-screen display and/or touch pad) etc it is other just
Portable device.It is to be further understood that in certain embodiments, the equipment is not portable communication device, but with tactile
Touch the desktop computer of sensing surface (for example, touch-screen display and/or touch pad).
In discussion below, the server including display and touch sensitive surface is described.It is, however, to be understood that
, server can include such as physical keyboard, mouse and/or control-rod one or more of the other physical user interface set
It is standby.
Server supports various application programs, such as one or more of following:Drawing application program, demonstration application journey
Sequence, word-processing application, website create application program, disk imprinting application program, spreadsheet applications, game application
Program, telephony application, videoconference application, email application, instant messaging applications, exercise
Support application program, photo management application program, digital camera application program, digital camera application program, web-browsing application
Program, digital music player application and/or video frequency player application program.
The various application programs that can be performed on the server can use at least one public affairs of such as touch sensitive surface
Physical user-interface device altogether.It can adjust and/or change among applications and/or in corresponding application programs and touch sensitivity
The corresponding information shown in the one or more functions and server on surface.So, server public physical structure (for example,
Touch sensitive surface) the various application programs with user interface directly perceived and transparent for a user can be supported.
Referring to Fig. 1, Fig. 1 is a kind of signal of method for assessing ad conversion rates prediction model provided in an embodiment of the present invention
Flow chart.The executive agent that the method for ad conversion rates prediction model is assessed in the present embodiment is server.As shown in Figure 1 comments
The method for estimating ad conversion rates prediction model may comprise steps of:
S11:Receive ad conversion rates prediction model call request, random call new edition ad conversion rates prediction model or old
Version ad conversion rates prediction model calculating candidate locations estimate conversion ratio, record model corresponding to the candidate locations and call mark
Know.
When advertiser have launch advertisement demand when, can to want advertisement side's platform (Demand-Side Platform,
DSP) advertisement is launched in request.Advertisement DSP is a kind of online advertisement platform, and it serves advertiser, is used to help advertiser mutual
Advertisement putting is carried out on networking or mobile Internet.Advertisement DSP platform can be supported by click charging (Cost Per
Click, CPC) type advertisement putting, can also support by the charging of download rate (Cost Per Download, CPD) type
Advertisement putting, or, the advertisement putting of other charge mode can also be supported, is not limited herein.
In order to obtain higher ad revenue, advertisement DSP is generally needed before being launched to candidate (to be put) advertisement
The conversion ratio of candidate locations is estimated, then self-corresponding estimate conversion ratio and advertiser further according to candidate locations are each
The parameters such as the bid to candidate locations determine the exposure information of each candidate locations to integrate, and last advertisement DSP can be according to each
The exposure information of candidate locations is exposed to candidate locations.Wherein, exposure information includes but is not limited to exposure order, exposure time
Number or exposure position etc..
In embodiments of the present invention, candidate locations can be one, or at least two, not be limited herein.Wait
CPC types or CPD types etc. can be included but is not limited to by selecting the type of advertisement.For advertisement for CPC types, the point of advertisement
The rate of hitting is the conversion ratio of advertisement;For advertisement for CPD types, the clicking rate of advertisement and the product for estimating download rate are wide
The conversion ratio of announcement.Wherein, the clicking rate of advertisement accounts for the ratio of the light exposure of advertisement for the click volume of advertisement;The download rate of advertisement is
By click on advertisement into corresponding product download interface after, the download for downloading product corresponding to the advertisement accounts for ad click amount
Ratio.
When advertisement DSP needs to estimate the conversion ratio of candidate locations, ad conversion rates can be sent to server
Prediction model call request, server, can be with when receiving the ad conversion rates prediction model call request of advertisement DSP transmissions
Ad conversion rates prediction model is called to estimate the conversion ratio of candidate locations.
In embodiments of the present invention, before new edition ad conversion rates prediction model is formally reached the standard grade, in order to determine that new edition is wide
Conversion ratio prediction model is accused whether really better than old edition ad conversion rates prediction model, it is necessary to estimate mould to new edition ad conversion rates
Type and old edition ad conversion rates prediction model carry out test experiments, with according to test experiments to new edition ad conversion rates prediction model
Assessed with the quality of old edition ad conversion rates prediction model, determine whether new edition advertising conversion of reaching the standard grade further according to assessment result
Rate prediction model.
Specifically, when testing new edition ad conversion rates prediction model and old edition ad conversion rates prediction model,
Within test period (i.e. assessment cycle), if the ad conversion rates prediction model calling that server receives advertisement DSP transmissions please
Ask, then it is pre- according to default random call strategy random call new edition ad conversion rates prediction model or old edition ad conversion rates
That estimates model calculating candidate locations estimates conversion ratio, and records model corresponding to candidate locations and call mark.Implement in the present invention
In example, for every candidate locations, advertisement DSP platform can send an ad conversion rates prediction model calling to server please
Ask, each ad conversion rates prediction model call request sent for advertisement DSP, server all can be according to default random
One of in regulative strategy random call new edition ad conversion rates prediction model and old edition ad conversion rates prediction model, meter
The conversion ratio of estimating of the targeted candidate locations of this ad conversion rates prediction model call request is calculated, meanwhile, server is also remembered
Record model corresponding to every candidate locations and call mark.
Wherein, default randomized policy need to ensure to estimate new edition ad conversion rates prediction model or old edition ad conversion rates
The calling of model is uniform on the whole, i.e., need to ensure to the call number of new edition ad conversion rates prediction model and to old edition
The call number of ad conversion rates prediction model is roughly equal on the whole.Default random call strategy can be according to reality
Border demand is set, and is not limited herein.
The ad conversion rates called during conversion ratio of estimating that model calls mark to be used to represent to calculate candidate locations are estimated
The version of model, that is, the conversion ratio of estimating for being used to represent candidate locations is by calling new edition ad conversion rates prediction model to calculate
It is obtaining or being calculated by calling old edition ad conversion rates prediction model.
In embodiment of the present invention, ad conversion rates prediction model can specifically estimate clicking rate and the download of advertisement
Rate.If server detects the advertisement that the targeted candidate locations of ad conversion rates prediction model call request are CPC types,
Random call new edition ad conversion rates prediction model or old edition ad conversion rates prediction model calculate pre- corresponding to the candidate locations
Estimate clicking rate, and by the candidate locations estimate that clicking rate is identified as the candidate locations estimate conversion ratio;If server detects
The candidate locations targeted to ad conversion rates prediction model call request are the advertisement of CPD types, then random call new edition is wide
Accuse conversion ratio prediction model or old edition ad conversion rates prediction model calculates and clicking rate is estimated corresponding to the candidate locations and is estimated
Download rate, and the clicking rate of estimating of the candidate locations is identified as that the candidate locations are corresponding to be estimated with estimating the product of download rate
Conversion ratio.
In embodiments of the present invention, ad conversion rates predictive algorithm and old edition corresponding to new edition ad conversion rates prediction model
Ad conversion rates predictive algorithm corresponding to ad conversion rates prediction model is different, and advertisement corresponding to ad conversion rates prediction model turns
Rate predictive algorithm can be set according to the actual requirements, not repeated herein.
S12:The conversion ratio of estimating of the candidate locations is sent to advertising platform, makes the advertising platform according to the time
Select advertisement estimate conversion ratio and default advertisement exposure strategy is exposed to the candidate locations.
Server be calculated every candidate locations estimate conversion ratio after, by every candidate locations to estimate conversion ratio equal
Send to advertising platform.Wherein, advertising platform can be advertisement DSP, or advertisement bidding engine etc., with specific reference to reality
Demand determines, is not limited herein.
Advertising platform receive every candidate locations estimate conversion ratio after, estimate conversion ratio according to every candidate locations
And default advertisement exposure strategy is exposed to candidate locations.Wherein, default advertisement exposure strategy can be according to reality
Demand is set, and is not limited herein.For example, default advertisement exposure strategy can be, turned according to every estimating for candidate locations
The parameters such as the bid of rate and advertiser to every candidate locations are integrated ordered to candidate locations progress, true according to ranking results
The exposure information of fixed every candidate locations, and candidate locations are exposed according to the exposure information of every candidate locations.Generally,
Estimate that conversion ratio is higher, advertisement bid is higher corresponding to candidate locations, then sort more forward.The sequence of advertisement is more forward, then its
The priority of corresponding exposure order is higher, and exposure frequency is more, and exposure position is more excellent.
S13:Model calls mark according to corresponding to having exposed advertisement, calculates the new edition ad conversion rates prediction model pair
Second actual advertisement conversion ratio corresponding to the first actual advertisement conversion ratio and the old edition ad conversion rates prediction model answered.
After advertising platform is exposed to candidate locations, server can call mark by model according to corresponding to having exposed advertisement
Know, count the first accumulative advertisement exposure amount corresponding to new edition ad conversion rates prediction model and the first accumulative advertising conversion amount, with
And the second accumulative advertisement exposure amount corresponding to statistics old edition ad conversion rates prediction model and the second accumulative advertising conversion amount.Service
Device is according to corresponding to the first accumulative advertisement exposure amount and the first accumulative advertising conversion amount calculate new edition ad conversion rates prediction model
First actual advertisement conversion ratio, and calculate old edition advertisement according to the second accumulative advertisement exposure amount and the second accumulative advertising conversion amount and turn
Second actual advertisement conversion ratio corresponding to rate prediction model.It should be noted that for advertisement for CPC types, advertisement
Inversion quantity refer to the click volume of advertisement;For advertisement for CPD types, the inversion quantity of advertisement refers to the download of advertisement
Amount.
Specifically, server calculates the ratio of the first accumulative advertising conversion amount and the first accumulative advertisement exposure amount, that is, obtain
First actual advertisement conversion ratio;Server calculates the ratio of the second accumulative advertising conversion amount and the second accumulative advertisement exposure amount, i.e.,
Obtain the second actual advertisement conversion ratio.
In embodiments of the present invention, server can be wide from test experiments start time to current time new edition with real-time statistics
The first accumulative advertisement exposure amount corresponding to conversion ratio prediction model and the first accumulative advertising conversion amount are accused, and is gone out according to real-time statistics
The first accumulative advertisement exposure amount and the first accumulative advertising conversion amount, calculate corresponding to new edition ad conversion rates prediction model in real time
First actual advertisement conversion ratio;And real-time statistics are estimated from test experiments start time to current time old edition ad conversion rates
Second accumulative advertisement exposure amount corresponding to model and the second accumulative advertising conversion amount, and second gone out according to real-time statistics is accumulative wide
Light exposure and the second accumulative advertising conversion amount are accused, calculates real-time second advertising conversion corresponding to old edition ad conversion rates prediction model
Rate.
Server can also count new edition ad conversion rates in preset period of time and estimate mould according to default time dimension information
3rd accumulative advertisement exposure amount corresponding to type and the 3rd accumulative advertising conversion amount, and according to the 3rd accumulative advertisement exposure amount and the 3rd
Accumulative advertising conversion amount, calculate offline first actual advertisement conversion corresponding to new edition ad conversion rates prediction model in preset period of time
Rate;And the 4th accumulative advertisement exposure amount and the 4th adds up corresponding to old edition ad conversion rates prediction model in statistics preset period of time
Advertising conversion amount, and according to the 4th accumulative advertisement exposure amount and the 4th accumulative advertising conversion amount, it is wide to calculate old edition in preset period of time
Accuse offline second actual advertisement conversion ratio corresponding to conversion ratio prediction model.
Wherein, default time dimension information, which is used for expression, will be divided at least one preset period of time test period.It is default
The duration of period can be set according to the actual requirements, be not limited herein.For example, the duration of preset period of time can be 1 hour,
It can be 1 day etc..If preset period of time when a length of 1 day, default time dimension information will divide test period for expression
For 7 days.Server is counted in units of day and is calculated offline first reality corresponding to new edition ad conversion rates prediction model every day
Offline second actual advertisement conversion ratio corresponding to border ad conversion rates and old edition ad conversion rates prediction model.
S14:According to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and default assessment plan
Slightly, good and bad assessment is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model.
It is wide that server calculates the first actual advertisement conversion ratio and old edition corresponding to new edition ad conversion rates prediction model
, can be according to the first actual advertisement conversion ratio, second in fact after accusing the second actual advertisement conversion ratio corresponding to conversion ratio prediction model
Border ad conversion rates and default assessment strategy, mould is estimated to new edition ad conversion rates prediction model and old edition ad conversion rates
Type carries out good and bad assessment.
Wherein, default assessment strategy can be configured according to the actual requirements, be not limited herein.It is for example, default
Assessment strategy can be, respectively according to real time data (including real-time first actual advertisement conversion ratio and real-time second counted
Actual advertisement conversion ratio) and off-line data (including offline first actual advertisement conversion ratio and offline second actual advertisement conversion ratio)
Good and bad assessment is carried out to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model, obtains the first assessment result
With the second assessment result, according to the first assessment result and the second assessment result come comprehensive assessment new edition ad conversion rates prediction model
With the quality of the old edition ad conversion rates prediction model.Used data were measured within experimental period during due to assessing
, therefore, it is generally the case that the first assessment result and the second assessment result are identicals.For example, if the first assessment result is new
Version ad conversion rates prediction model is better than old edition ad conversion rates prediction model, then the second assessment result is generally also new edition advertisement
Conversion ratio prediction model is better than old edition ad conversion rates prediction model.
Specifically, for the real time data counted, server can be according to real-time first actual advertisement conversion ratio and reality
When the second actual advertisement conversion ratio, determine new edition ad conversion rates prediction model relative to old edition ad conversion rates prediction model
Enhancing rate.Wherein, enhancing rate is used to represent that the ad conversion rates of new edition ad conversion rates prediction model to turn relative to old edition advertisement
The ratio that the ad conversion rates of rate prediction model are lifted.
Server can also count real-time participation number corresponding to new edition ad conversion rates prediction model and in real time conversion people
Number, and number and conversion number, and to count in real time are participated in real time corresponding to statistics old edition ad conversion rates prediction model
New edition ad conversion rates prediction model corresponding to participation number and real-time conversion number, old edition ad conversion rates estimate mould in real time
Number is participated in corresponding to type in real time and conversion number is sample data in real time, it is wide to new edition ad conversion rates prediction model and old edition
Accuse conversion ratio prediction model and carry out the first hypothesis testing.Wherein, the first hypothesis testing is used to examine new edition ad conversion rates to estimate
Difference between model and each self-corresponding overall ad conversion rates of old edition ad conversion rates prediction model.First hypothesis testing can
To set according to the actual requirements, it is not limited herein.For example, the first hypothesis testing can be T inspections or Z test etc..
Server can determine the first assessment result according to the enhancing rate calculated, can also be according to the first hypothesis testing
Assay determines the first assessment result, or can also be integrated according to the result and conversion ratio of the first hypothesis testing and determine first
Assessment result, it is not limited herein.
For the off-line data counted, server can be with offline first actual advertisement conversion ratio and offline second reality
Ad conversion rates are sample data, and second is carried out to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model
Hypothesis testing.Wherein, the second hypothesis testing is used to examine new edition ad conversion rates prediction model and old edition ad conversion rates to estimate
Difference between overall ad conversion rates corresponding to model.Second hypothesis testing can be set according to the actual requirements, not done herein
Limitation.For example, the second hypothesis testing can be rank tests.Rank tests can be Wilcoxen (wilcoxon) rank tests,
It can be other kinds of rank tests, not be limited herein.
Such scheme, server is to new edition ad conversion rates prediction model and the progress of old edition ad conversion rates prediction model
When quality is assessed, due to being random call new edition model or old edition after ad conversion rates prediction model call request is received
Model is estimated to the conversion ratio of candidate locations, so that new edition model and old edition model are carried out to ad conversion rates
The ad data of institute's foundation belongs to same time dimension when estimating, for example, the ad data belonged in assessment cycle, then wide
Accuse platform to estimate after conversion ratio is exposed advertisement in the advertisement gone out according to new edition model or old edition model pre-estimating, according to having exposed
The new edition model and each self-corresponding actual advertisement conversion ratio of old edition model that the model calling mark of light advertisement is calculated also must
It is so the data for belonging to same time dimension, mould is estimated to new and old edition ad conversion rates according to the data in same time dimension
Type is assessed, it is possible to increase the good and bad accuracy assessed of ad conversion rates prediction model.
Referring to Fig. 2, Fig. 2 is a kind of method for assessment ad conversion rates prediction model that another embodiment of the present invention provides
Schematic flow diagram.The executive agent that the method for ad conversion rates prediction model is assessed in the present embodiment is server.As shown in Figure 2
The method of assessment ad conversion rates prediction model may comprise steps of:
S21:Receive ad conversion rates prediction model call request, random call new edition ad conversion rates prediction model or old
Version ad conversion rates prediction model calculating candidate locations estimate conversion ratio, record model corresponding to the candidate locations and call mark
Know.
S21 in the present embodiment is identical with the S11 in a upper embodiment, referring specifically to the S11's in a upper embodiment
Associated description, do not repeat herein.
In embodiments of the present invention, new edition ad conversion rates prediction model can include new edition ad click rate prediction model
With new edition advertisement download rate prediction model;Old edition ad conversion rates prediction model can include old edition ad click rate prediction model
With old edition advertisement download rate prediction model.Ad click rate prediction model is used for the clicking rate for estimating advertisement, and advertisement download rate is pre-
Estimate the download rate that model is used to estimate advertisement.Ad click rate predictive algorithm corresponding to new edition ad click rate prediction model with it is old
Ad click rate predictive algorithm corresponding to version ad click rate prediction model is different, corresponding to new edition advertisement download rate prediction model
Advertisement download rate predictive algorithm advertisement download rate corresponding with old edition advertisement download rate prediction model estimates difference in advance.Ad click
Rate predictive algorithm and advertisement download rate predictive algorithm can be set according to the actual requirements, be not limited herein.
In the present embodiment, server can the ad click rate prediction model to new and old edition and new and old edition respectively
Advertisement download rate prediction model carries out good and bad assessment.For example, it can will be divided into for the first default test period and second test period
Default test period, and be arranged in the first default test period and the ad click rate prediction model of new and old edition is tested
Assess, the advertisement download rate prediction model to new and old edition within the second default test period carries out test assessment.Or can be with
It is arranged on the advertisement download rate prediction model in the first default test period to new and old edition and carries out test assessment, it is default second
Test assessment is carried out to the ad click rate prediction model of new and old edition in test period, is not limited herein.Exemplary, if
Test period is 14 days, then the first default test period can be first 7 days, and the second default test period can be latter 7 days.
In embodiments of the present invention, when the ad click rate prediction model to new and old edition is tested, need to ensure wide
The version for accusing download rate prediction model is consistent.When the advertisement download rate prediction model to new and old edition is tested, need to ensure
The version of ad click rate prediction model is consistent.
Exemplary, the ad click rate prediction model of new and old edition is carried out if being arranged in the first default test period
Test is assessed, and the advertisement download rate prediction model to new and old edition within the second default test period carries out test assessment, then S21
S211~S213 can specifically be included.
S211:Ad conversion rates prediction model call request is received, if being currently in the first default test period, is examined
Survey the type of candidate locations corresponding to the ad conversion rates prediction model call request.
In the present embodiment, if server receives the ad conversion rates prediction model call request of advertisement DSP transmissions,
Current temporal information is obtained, if current time was in the first default test period, server detection ad conversion rates are pre-
Estimate the type of candidate locations corresponding to model call request.
S212:If the type of the candidate locations is the first preset kind, new edition ad click rate described in random call
What prediction model or the old edition ad click rate prediction model calculated the candidate locations estimates clicking rate, and the candidate is wide
Clicking rate is identified as the candidate locations estimates conversion ratio for estimating of accusing.
In the present embodiment, the first preset kind is CPC types, and the second preset kind is CPD types.
If server detects that the type of candidate locations corresponding to ad conversion rates prediction model call request is pre- for first
If type, then according to default random call strategy random call new edition ad click rate prediction model or old edition ad click rate
Prediction model calculates the clicking rate of estimating of the candidate locations, and the clicking rate of estimating of the candidate locations is identified as into the candidate locations
Estimate conversion ratio.Meanwhile server also records model corresponding to the candidate locations and calls mark.
For example, it is assumed that new edition ad click rate prediction model is identified as a, the mark of old edition ad click rate prediction model
For b.If server calling new edition ad click rate prediction model the first candidate locations of calculating estimate clicking rate, the first candidate
Model corresponding to advertisement calls and is identified as a, if server calls old edition ad click rate prediction model to calculate the second candidate locations
Estimate clicking rate, then corresponding to the second candidate locations model call be identified as b.
S213:If the type of the candidate locations is the second preset kind, new edition ad click rate described in random call
Prediction model or the old edition ad click rate prediction model calculate the clicking rate of estimating of the candidate locations, call first to preset
What advertisement download rate prediction model calculated the candidate locations estimates download rate, by the candidate locations estimate clicking rate with it is pre-
What the product for estimating download rate was identified as the candidate locations estimates conversion ratio.
If server detects that the type of candidate locations corresponding to ad conversion rates prediction model call request is pre- for second
If type, then according to default random call strategy random call new edition ad click rate prediction model or old edition ad click rate
Prediction model calculates the clicking rate of estimating of the candidate locations, and calls the first default advertisement download rate prediction model to calculate the candidate
Download rate is estimated in advertisement, and the product estimated clicking rate and estimate download rate of the candidate locations is identified as into the candidate locations
Estimate conversion ratio.Meanwhile server also records model corresponding to the candidate locations and calls mark.
Wherein, the first default advertisement download rate prediction model can be old edition advertisement download rate prediction model.Or first
Default advertisement download rate prediction model can be new edition advertisement download rate prediction model, be determined with specific reference to actual demand, herein
It is not limited.
Further, S21 can also specifically include S214~S216.
S214:Ad conversion rates prediction model call request is received, if being currently in the second default test period, is examined
Survey the type of candidate locations corresponding to the ad conversion rates prediction model call request.
In the present embodiment, server is receiving the ad conversion rates prediction model call request of advertisement DSP transmissions, and
When detecting that current time was in the second default test period, waited corresponding to detection ad conversion rates prediction model call request
Select the type of advertisement.
S215:If the type of the candidate locations is the first preset kind, the first default ad click rate is called to estimate
The model calculating candidate locations estimate clicking rate, and the clicking rate of estimating of the candidate locations is identified as into the candidate locations
Estimate conversion ratio.
If server detects that the type of candidate locations is the first preset kind, call the first default ad click rate pre-
That estimates model calculating candidate locations estimates clicking rate, and the clicking rate of estimating of the candidate locations is identified as into the pre- of the candidate locations
Estimate conversion ratio.
Wherein, the first default ad click rate prediction model can be according to the new and old edition in the first default test period
The good and bad assessment result that ad click rate estimates pattern determines.If the assessment result in the first default test period is new edition advertisement
Clicking rate prediction model is better than old edition ad click rate prediction model, then the first default ad click rate prediction model can be new
Version ad conversion rates prediction model;If the assessment result in the first default test period is excellent for old edition ad click rate prediction model
In new edition ad click rate prediction model, then the first default ad click rate prediction model can be that old edition ad conversion rates are estimated
Model.
S216:If the type of the candidate locations is the second preset kind, the described first default ad click rate is called
The prediction model calculating candidate locations estimate clicking rate, new edition advertisement download rate prediction model or described described in random call
The old edition advertisement download rate prediction model calculating candidate locations estimate download rate, and the candidate locations are estimated into clicking rate
What the product with estimating download rate was identified as the candidate locations estimates conversion ratio.
If server detects that the type of candidate locations corresponding to ad conversion rates prediction model call request is pre- for second
If type, then the first default ad click rate prediction model is called to calculate the clicking rate of estimating of the candidate locations, and according to default
Random call strategy to call new edition advertisement download rate prediction model or old edition advertisement download rate prediction model to calculate the candidate wide
That accuses estimates download rate, and the product estimated clicking rate and estimate download rate of the candidate locations is identified as into the candidate locations
Estimate conversion ratio.Meanwhile server also records model corresponding to the candidate locations and calls mark.
For example, it is assumed that new edition advertisement download rate prediction model is identified as c, the mark of old edition advertisement download rate prediction model
For d.If server calling new edition advertisement download rate prediction model the 3rd candidate locations of calculating estimate download rate, the 3rd candidate
Model corresponding to advertisement calls and is identified as c, if server calls old edition advertisement download rate prediction model to calculate the 4th candidate locations
Estimate download rate, then corresponding to the 4th candidate locations model call be identified as d.
It should be noted that in the present embodiment, S211~S213 and S214~S216 is step arranged side by side.If server is held
Go S211~S213, then no longer perform S214~S216;If server performs S214~S216, no longer execution S211~
S213。
S22:The conversion ratio of estimating of the candidate locations is sent to advertising platform, makes the advertising platform according to the time
Select advertisement estimate conversion ratio and default advertisement exposure strategy is exposed to the candidate locations.
S22 in the present embodiment is identical with the S12 in a upper embodiment, referring specifically to the S12's in a upper embodiment
Associated description, do not repeat herein.
S23:Model calls mark according to corresponding to having exposed advertisement, calculates the new edition ad conversion rates prediction model pair
Second actual advertisement conversion ratio corresponding to the first actual advertisement conversion ratio and the old edition ad conversion rates prediction model answered.
After advertising platform is exposed to candidate locations, server can call mark by model according to corresponding to having exposed advertisement
Know, count the first accumulative advertisement exposure amount corresponding to new edition ad conversion rates prediction model and the first accumulative advertising conversion amount, with
And the second accumulative advertisement exposure amount corresponding to statistics old edition ad conversion rates prediction model and the second accumulative advertising conversion amount.Service
Device is according to corresponding to the first accumulative advertisement exposure amount and the first accumulative advertising conversion amount calculate new edition ad conversion rates prediction model
First actual advertisement conversion ratio, and calculate old edition advertisement according to the second accumulative advertisement exposure amount and the second accumulative advertising conversion amount and turn
Second actual advertisement conversion ratio corresponding to rate prediction model.
It should be noted that for advertisement for CPC types, the inversion quantity of advertisement refers to the click volume of advertisement;It is right
For the advertisement of CPD types, the inversion quantity of advertisement refers to the download of advertisement.
Specifically, server calculates the ratio of the first accumulative advertising conversion amount and the first accumulative advertisement exposure amount, that is, obtain
First actual advertisement conversion ratio;Server calculates the ratio of the second accumulative advertising conversion amount and the second accumulative advertisement exposure amount, i.e.,
Obtain the second actual advertisement conversion ratio.
In the present embodiment, the first actual advertisement conversion ratio can include real-time first actual advertisement conversion ratio and offline the
One actual advertisement conversion ratio.Real-time first actual advertisement conversion ratio be the accumulative advertisement exposure amount of first gone out according to real-time statistics and
What the first accumulative advertising conversion amount was calculated;Offline first actual advertisement conversion ratio is according in the preset period of time counted
What the first accumulative advertisement exposure amount and the first accumulative advertising conversion amount were calculated.Second ad conversion rates can also include real-time
Second actual advertisement conversion ratio and offline second actual advertisement conversion ratio.Real-time second actual advertisement conversion ratio is according to system in real time
What the second accumulative advertisement exposure amount and the second accumulative advertising conversion amount counted out were calculated;Offline second actual advertisement conversion ratio
It is to be calculated according to the second accumulative advertisement exposure amount in the preset period of time counted and the second accumulative advertising conversion amount.
Further, S23 can specifically include S231 and S232.
S231:Model calls mark according to corresponding to having exposed advertisement, and new edition ad conversion rates are estimated described in real-time statistics
First accumulative advertisement exposure amount corresponding to model and the first accumulative advertising conversion amount, according to the described first accumulative advertisement exposure amount and
Described first accumulative advertising conversion amount, calculates real-time first actual advertisement corresponding to the new edition ad conversion rates prediction model and turns
Rate.
S232:Model calls mark according to corresponding to having exposed advertisement, and old edition ad conversion rates are estimated described in real-time statistics
Second accumulative advertisement exposure amount corresponding to model and the second accumulative advertising conversion amount, according to the described second accumulative advertisement exposure amount and
Described second accumulative advertising conversion amount, calculates real-time second actual advertisement corresponding to the old edition ad conversion rates prediction model and turns
Rate.
In the present embodiment, server can call mark by model according to corresponding to having exposed advertisement, and real-time statistics are from survey
First accumulative advertisement exposure amount and first corresponding to examination experiment start time to current time new edition ad conversion rates prediction model
Accumulative advertising conversion amount, and the first accumulative advertisement exposure amount and the first accumulative advertising conversion amount gone out according to real-time statistics, are calculated
Real-time first actual advertisement conversion ratio corresponding to new edition ad conversion rates prediction model, and real-time statistics are since test experiments
Second accumulative advertisement exposure amount corresponding to moment to current time old edition ad conversion rates prediction model and the second accumulative advertisement turn
Change amount, and the second accumulative advertisement exposure amount and the second accumulative advertising conversion amount gone out according to real-time statistics, calculate old edition advertisement and turn
Real-time second ad conversion rates corresponding to rate prediction model.
Further, S23 can also specifically include S233 and S234.
S233:Model calls mark according to corresponding to having exposed advertisement, counts the new edition advertising conversion in preset period of time
3rd accumulative advertisement exposure amount corresponding to rate prediction model and the 3rd accumulative advertising conversion amount, expose according to the described 3rd accumulative advertisement
Light quantity and the 3rd accumulative advertising conversion amount, calculate in preset period of time corresponding to the new edition ad conversion rates prediction model from
Line the first actual advertisement conversion ratio.
S234:Model calls mark according to corresponding to having exposed advertisement, counts the old edition advertising conversion in preset period of time
4th accumulative advertisement exposure amount corresponding to rate prediction model and the 4th accumulative advertising conversion amount, expose according to the described 4th accumulative advertisement
Light quantity and the 4th accumulative advertising conversion amount, calculate in preset period of time corresponding to the old edition ad conversion rates prediction model from
Line the second actual advertisement conversion ratio.
In the present embodiment, server can call mark and default time by model according to corresponding to having exposed advertisement
Dimensional information, count in preset period of time the 3rd accumulative advertisement exposure amount corresponding to new edition ad conversion rates prediction model and the 3rd tired
Advertising conversion amount is counted, and according to the 3rd accumulative advertisement exposure amount and the 3rd accumulative advertising conversion amount, calculates new edition in preset period of time
Offline first actual advertisement conversion ratio corresponding to ad conversion rates prediction model;And old edition advertising conversion in statistics preset period of time
4th accumulative advertisement exposure amount corresponding to rate prediction model and the 4th accumulative advertising conversion amount, and according to the 4th accumulative advertisement exposure
Amount and the 4th accumulative advertising conversion amount, calculate offline second reality corresponding to old edition ad conversion rates prediction model in preset period of time
Ad conversion rates.
Wherein, default time dimension information, which is used for expression, will be divided at least one preset period of time test period.It is default
The duration of period can be set according to the actual requirements, be not limited herein.For example, the duration of preset period of time can be 1 hour,
It can be 1 day etc..If preset period of time when a length of 1 day, default time dimension information will divide test period for expression
For 7 days.Server is counted in units of day and is calculated offline first reality corresponding to new edition ad conversion rates prediction model every day
Offline second actual advertisement conversion ratio corresponding to border ad conversion rates and old edition ad conversion rates prediction model.
It should be noted that in the present embodiment, S231, S232, S233 and S234 sequentially, can be held simultaneously in no particular order
OK.
S24:According to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and default assessment plan
Slightly, good and bad assessment is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model.
S24 in the present embodiment is identical with the S14 in a upper embodiment, referring specifically to the S14's in a upper embodiment
Associated description, do not repeat herein.
Further, after S232, S24 may comprise steps of:
According to the real-time first actual advertisement conversion ratio, the real-time second actual advertisement conversion ratio and default carry
Rate calculative strategy is risen, calculates the new edition ad conversion rates prediction model relative to the old edition ad conversion rates prediction model
Enhancing rate;
Real-time participation number and real-time conversion people according to corresponding to the new edition ad conversion rates prediction model counted
Number is participated in real time corresponding to several, described old edition ad conversion rates prediction model and converts number in real time, default Z score calculates
Z score under the conditions of policy calculation null hypothesis;
The new edition ad conversion rates prediction model and the old edition advertisement are turned according to the enhancing rate and the Z score
Rate prediction model carries out good and bad assessment.
In the present embodiment, server can turn according to real-time first actual advertisement conversion ratio, real-time second actual advertisement
Rate and default enhancing rate calculative strategy, it is pre- relative to old edition ad conversion rates to calculate new edition ad conversion rates prediction model
Estimate the enhancing rate of model.
Default enhancing rate calculative strategy can be A=(a1-a2)/a2.Wherein, A is that new edition ad conversion rates estimate mould
For type relative to the enhancing rate of old edition ad conversion rates prediction model, a1 is real-time first actual advertisement conversion ratio, and a2 is real-time the
Two actual advertisement conversion ratios.In the present embodiment, if the value of enhancing rate is on the occasion of illustrating new edition ad conversion rates prediction model
Better than old edition ad conversion rates prediction model, if the value of enhancing rate is negative value, illustrate that old edition ad conversion rates prediction model is excellent
In new edition ad conversion rates prediction model.And according to the absolute value of enhancing rate a model can be determined better than another model
Degree.
Server can also count real-time participation number corresponding to new edition ad conversion rates prediction model and in real time conversion people
Number, and number and conversion number, and to count in real time are participated in real time corresponding to statistics old edition ad conversion rates prediction model
New edition ad conversion rates prediction model corresponding to participation number and real-time conversion number, old edition ad conversion rates estimate mould in real time
Number is participated in corresponding to type in real time and conversion number is sample data in real time, it is wide to new edition ad conversion rates prediction model and old edition
Accuse conversion ratio prediction model and carry out Z test, to calculate the Z score under the conditions of null hypothesis.Z score is used to identify new edition advertising conversion
Difference between rate prediction model and old edition ad conversion rates prediction model.
Wherein, null hypothesis condition is specially:Assuming that new edition ad conversion rates prediction model is estimated with old edition ad conversion rates
There is no significant difference between model.
Specifically, server can be according to formulaCalculate null hypothesis
Under the conditions of Z score.Wherein, p1=x1/n1, p2=x2/n2, q1=1-p1, q2=1-p2, x1 are that new edition ad conversion rates are pre-
Estimate real-time participation number corresponding to model, n1 is that conversion number, x2 are old in real time corresponding to new edition ad conversion rates prediction model
Number is participated in corresponding to version ad conversion rates prediction model in real time, n2 is to turn in real time corresponding to old edition ad conversion rates prediction model
Change number.In the present embodiment, if the value of Z score be on the occasion of, illustrate new edition ad conversion rates prediction model better than old edition it is wide
Conversion ratio prediction model is accused, if the value of Z score is negative value, illustrates that old edition ad conversion rates prediction model turns better than new edition advertisement
Rate prediction model.
In the present embodiment, server can be estimated according to enhancing rate and Z score is calculated to new edition ad conversion rates
Model and the old edition ad conversion rates prediction model carry out good and bad assessment.Specifically, if the value of enhancing rate and the value of Z score are equal
For on the occasion of then illustrating that new edition ad conversion rates prediction model is better than old edition ad conversion rates prediction model;If the value and Z of enhancing rate
The value of fraction is negative value, then illustrates that old edition ad conversion rates prediction model is better than new edition ad conversion rates prediction model.
In the present embodiment, server can be according to the real time data counted within the first default test period, to new
The ad click rate prediction model of legacy version carries out good and bad assessment.Specifically, server can be preset in test period first,
Mark, starting of the real-time statistics from the first default test period are called according to model corresponding to the advertisement of the CPC types exposed
Moment started untill current time, and first of the advertisement of CPC types corresponding to new edition ad click rate prediction model is accumulative wide
Light exposure and the first accumulative ad click amount are accused, and the first accumulative ad click amount and first are added up to the ratio of advertisement exposure amount
It is identified as the real-time first actual advertisement conversion ratio of new edition ad click rate prediction model.Server can be in the first default test
In cycle, mark is called according to model corresponding to the advertisement of the CPC types exposed, real-time statistics preset test period from first
Initial time start untill current time, the second of the advertisement of CPC types corresponding to old edition ad click rate prediction model
Accumulative advertisement exposure amount and the second accumulative ad click amount, and the second accumulative ad click amount is added up into advertisement exposure amount with second
Ratio be identified as the real-time second actual advertisement conversion ratio of old edition ad click rate prediction model.
Server can be according to the real time data counted within the second default test period, under the advertisement of new and old edition
Load rate prediction model carries out good and bad assessment.Specifically, server can be within the second default test period, according to what is exposed
Model corresponding to the advertisement of CPD types calls mark, and real-time statistics are since the initial time of the second default test period to working as
Untill the preceding moment, the first accumulative advertisement exposure amount of the advertisement of CPD types corresponding to new edition advertisement download rate prediction model and the
One accumulative advertisement download, and it is wide that with the first ratio for adding up advertisement exposure amount the first accumulative advertisement download is identified as into new edition
Accuse the real-time first actual advertisement conversion ratio of download rate prediction model.Server can be preset in test period second, according to
Model corresponding to the advertisement of the CPD types exposed calls mark, initial time of the real-time statistics from the second default test period
Start untill current time, the second accumulative advertisement of the advertisement of CPD types corresponding to old edition advertisement download rate prediction model exposes
Light quantity and the second accumulative advertisement download, and the ratio of the second accumulative advertisement download and the second accumulative advertisement exposure amount is identified
For the real-time second actual advertisement conversion ratio of old edition advertisement download rate prediction model.
It should be noted that after server calculates the first actual advertisement conversion ratio and the second actual advertisement conversion ratio, root
Good and bad assessment is carried out to new edition ad click rate prediction model and old edition light ad click rate prediction model according to above-mentioned appraisal procedure,
Here is omitted.
Further, after S234, S24 may comprise steps of:
According to offline first actual advertisement conversion ratio, the offline second actual advertisement conversion ratio and the default order
Maintain strategy, quality is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model and commented
Estimate.
In this embodiment, an offline first actual advertisement conversion can be calculated in server in each preset period of time
Rate and an offline second actual advertisement conversion ratio, i.e. server can be calculated in each preset period of time one group it is offline
First actual advertisement conversion ratio and offline second actual advertisement conversion ratio.
Server can be turned with the multigroup offline first actual advertisement conversion ratio being calculated and offline second actual advertisement
Rate is sample data, and wilcoxon orders are carried out to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model
Examine, and according to wilcoxon rank tests result to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model
Carry out good and bad assessment.
Specifically, the step of wilcoxon rank tests, can be as follows:
(1) null hypothesis is established:Do not have between new edition ad conversion rates prediction model and old edition ad conversion rates prediction model
Significant difference;Establish alternative hypothesis:New edition ad conversion rates prediction model is better than old edition ad conversion rates prediction model;Set
Significance.
Wherein, significance is used to identify the probability that null hypothesis condition may make mistakes.Significance can basis
Actual demand is set, and is not limited herein.For example, significance can be 0.05.
(2) difference of the offline first actual advertisement conversion ratio and offline second actual advertisement conversion ratio in every group is sought, is obtained
To multiple differences, be ranked up according to the absolute value order from small to large of difference, and volume order carried out according to sequence, rank it
The sign of preceding holding original error value is constant.Wherein, the sequence number of sequence is rank.
(3) compile order and run into and then cast out the difference when difference is zero and do not compile order, the rank of the difference equal to absolute value is made even
Average, and before rank keep original error value sign.
(4) all positive rank sum W+ and all negative rank sum W- are calculated.
If positive rank sum W+ and negative rank sum W- difference is less than predetermined threshold value, receives null hypothesis establishment, that is, recognize
There is no significant difference between new edition ad conversion rates prediction model and old edition ad conversion rates prediction model;If positive rank it
With one of W+ and negative rank sum W- very little, then refuse null hypothesis, that is, think new edition ad conversion rates prediction model with
Old edition ad conversion rates prediction model has significant difference.Wherein, predetermined threshold value can be set according to the actual requirements, not done herein
Limitation.
(5) statistic W=min (W+, W-) is taken, is obtained according to the distribution table of statistic W and wilcoxon rank tests
P values under null hypothesis.Assessment result is determined according to the size of P values.
Wherein, if P values are less than or equal to default significance (such as 0.05), refuse null hypothesis, receive alternative
It is assumed that think that new edition ad conversion rates prediction model is better than old edition ad conversion rates prediction model.If P values are more than default aobvious
Work property is horizontal, then refuses null hypothesis without adequate cause.
In the present embodiment, server can the ad click rate prediction model to new and old edition and new and old edition respectively
Download rate prediction model carries out wilcoxon rank tests, and pre- to the ad click rate of new and old edition respectively according to rank tests result
The download rate prediction model for estimating model and new and old edition carries out good and bad assessment.
It should be noted that in the present embodiment, server can according to Z score, enhancing rate and rank tests result come
The quality of comprehensive assessment new edition ad conversion rates prediction model and old edition ad conversion rates prediction model.
Such scheme, server is to new edition ad conversion rates prediction model and the progress of old edition ad conversion rates prediction model
When quality is assessed, due to being random call new edition model or old edition after ad conversion rates prediction model call request is received
Model is estimated to the conversion ratio of candidate locations, so that new edition model and old edition model are carried out to ad conversion rates
The ad data of institute's foundation belongs to same time dimension when estimating, for example, the ad data belonged in assessment cycle, then wide
Accuse platform to estimate after conversion ratio is exposed advertisement in the advertisement gone out according to new edition model or old edition model pre-estimating, according to having exposed
The new edition model and each self-corresponding actual advertisement conversion ratio of old edition model that the model calling mark of light advertisement is calculated also must
It is so the data for belonging to same time dimension, mould is estimated to new and old edition ad conversion rates according to the data in same time dimension
Type is assessed, it is possible to increase the good and bad accuracy assessed of ad conversion rates prediction model.
Server can be according to the real time data counted, from many aspects to new edition ad conversion rates prediction model and old
The carry out quality assessment of version ad conversion rates prediction model, and the real time data counted and the offline number counted can be combined
According to integrate the carry out quality assessment to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model, entering one
Step improves the good and bad accuracy assessed of ad conversion rates prediction model.
The embodiment of the present invention also provides a kind of server, and the server includes being used to perform the assessment described in foregoing any one
The unit of the method for ad conversion rates prediction model.Specifically, referring to Fig. 3, Fig. 3 is a kind of service provided in an embodiment of the present invention
The schematic block diagram of device.Server 300 can be the Mobile Serveies such as smart mobile phone, tablet personal computer.The server of the present embodiment
300 include can be with model call unit 301, transmitting element 302, computing unit 303 and assessment unit 304.
Model call unit 301 is used to receive ad conversion rates prediction model call request, and the advertisement of random call new edition turns
What rate prediction model or old edition ad conversion rates prediction model calculated candidate locations estimates conversion ratio, records the candidate locations
Corresponding model calls mark.
Transmitting element 302 is used to send the conversion ratio of estimating of the candidate locations to advertising platform, puts down the advertisement
Platform according to the candidate locations estimate conversion ratio and default advertisement exposure strategy is exposed to the candidate locations.
Computing unit 303 is used to, according to model calling mark corresponding to advertisement has been exposed, calculate the new edition advertising conversion
Second is actual corresponding to first actual advertisement conversion ratio corresponding to rate prediction model and the old edition ad conversion rates prediction model
Ad conversion rates.
Assessment unit 304 be used for according to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and
Default assessment strategy, the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model are carried out excellent
Bad assessment.
Optionally, model call unit 301 can include the first detection unit 3011, the first model call unit 3012 and
Second model call unit 3013.
First detection unit 3011 is used to receive ad conversion rates prediction model call request, if it is default to be currently at first
In test period, then the type of candidate locations corresponding to the ad conversion rates prediction model call request is detected.
If the type that the first model call unit 3012 is used for the candidate locations is the first preset kind, random call
The new edition ad click rate prediction model or the old edition ad click rate prediction model calculate estimating for the candidate locations
Clicking rate, by the candidate locations estimate that clicking rate is identified as the candidate locations estimate conversion ratio.
If the type that the second model call unit 3013 is used for the candidate locations is the second preset kind, random call
The new edition ad click rate prediction model or the old edition ad click rate prediction model calculate estimating for the candidate locations
Clicking rate, that calls the first default advertisement download rate prediction model calculating candidate locations estimates download rate, by the candidate
What the product estimated clicking rate and estimate download rate of advertisement was identified as the candidate locations estimates conversion ratio.
Further, the first detection unit 3011 is additionally operable to receive ad conversion rates prediction model call request, if currently
Within the second default test period, then the class of candidate locations corresponding to the ad conversion rates prediction model call request is detected
Type.
If the type that the first model call unit 3012 is additionally operable to the candidate locations is the first preset kind, is called
The one default ad click rate prediction model calculating candidate locations estimate clicking rate, and the candidate locations are estimated into click
What rate was identified as the candidate locations estimates conversion ratio.
If the type that the second model call unit 3013 is additionally operable to the candidate locations is the second preset kind, is called
The one default ad click rate prediction model calculating candidate locations estimate clicking rate, and new edition advertisement described in random call is downloaded
What rate prediction model or the old edition advertisement download rate prediction model calculated the candidate locations estimates download rate, by the candidate
What the product estimated clicking rate and estimate download rate of advertisement was identified as the candidate locations estimates conversion ratio.
Further, computing unit 303 includes the first computing unit 3031 and the second computing unit 3032.
First computing unit 3031 is used for basis and has exposed model calling mark, new edition described in real-time statistics corresponding to advertisement
First accumulative advertisement exposure amount corresponding to ad conversion rates prediction model and the first accumulative advertising conversion amount, it is tired according to described first
Advertisement exposure amount and the first accumulative advertising conversion amount are counted, is calculated corresponding to the new edition ad conversion rates prediction model in real time
First actual advertisement conversion ratio.
Second computing unit 3032 is used for basis and has exposed model calling mark, old edition described in real-time statistics corresponding to advertisement
Second accumulative advertisement exposure amount corresponding to ad conversion rates prediction model and the second accumulative advertising conversion amount, it is tired according to described second
Advertisement exposure amount and the second accumulative advertising conversion amount are counted, is calculated corresponding to the old edition ad conversion rates prediction model in real time
Second actual advertisement conversion ratio.
Further, assessment unit 304 includes enhancing rate computing unit, Z score computing unit and the first assessment unit.
Enhancing rate computing unit is used for according to the real-time first actual advertisement conversion ratio, real-time second actual advertisement
Conversion ratio and default enhancing rate calculative strategy, it is wide relative to the old edition to calculate the new edition ad conversion rates prediction model
Accuse the enhancing rate of conversion ratio prediction model.
Z score computing unit is used to be participated in real time according to corresponding to the new edition ad conversion rates prediction model counted
Real-time participation number corresponding to number and real-time conversion number, the old edition ad conversion rates prediction model and people is converted in real time
Several, default Z score calculative strategy calculates the Z score under the conditions of null hypothesis.
First assessment unit is used for according to the enhancing rate and the Z score to the new edition ad conversion rates prediction model
Good and bad assessment is carried out with the old edition ad conversion rates prediction model.
Further, computing unit 303 also includes the 3rd statistic unit 3033 and the 4th computing unit 3034.
3rd statistic unit 3033 is used to, according to model calling mark corresponding to advertisement has been exposed, count institute in preset period of time
The 3rd accumulative advertisement exposure amount corresponding to new edition ad conversion rates prediction model and the 3rd accumulative advertising conversion amount are stated, according to described
3rd accumulative advertisement exposure amount and the 3rd accumulative advertising conversion amount, it is pre- to calculate the new edition ad conversion rates in preset period of time
Estimate offline first actual advertisement conversion ratio corresponding to model.
4th computing unit 3034 is used to, according to model calling mark corresponding to advertisement has been exposed, count institute in preset period of time
The 4th accumulative advertisement exposure amount corresponding to old edition ad conversion rates prediction model and the 4th accumulative advertising conversion amount are stated, according to described
4th accumulative advertisement exposure amount and the 4th accumulative advertising conversion amount, it is pre- to calculate the old edition ad conversion rates in preset period of time
Estimate offline second actual advertisement conversion ratio corresponding to model.
Further, assessment unit 304 is additionally operable to according to the offline first actual advertisement conversion ratio, described offline second
Actual advertisement conversion ratio and default rank tests strategy, to the new edition ad conversion rates prediction model and the old edition advertisement
Conversion ratio prediction model carries out good and bad assessment.
Such scheme, server is to new edition ad conversion rates prediction model and the progress of old edition ad conversion rates prediction model
When quality is assessed, due to being random call new edition model or old edition after ad conversion rates prediction model call request is received
Model is estimated to the conversion ratio of candidate locations, so that new edition model and old edition model are carried out to ad conversion rates
The ad data of institute's foundation belongs to same time dimension when estimating, for example, the ad data belonged in assessment cycle, then wide
Accuse platform to estimate after conversion ratio is exposed advertisement in the advertisement gone out according to new edition model or old edition model pre-estimating, according to having exposed
The new edition model and each self-corresponding actual advertisement conversion ratio of old edition model that the model calling mark of light advertisement is calculated also must
It is so the data for belonging to same time dimension, mould is estimated to new and old edition ad conversion rates according to the data in same time dimension
Type is assessed, it is possible to increase the good and bad accuracy assessed of ad conversion rates prediction model.
Server can be according to the real time data counted, from many aspects to new edition ad conversion rates prediction model and old
The carry out quality assessment of version ad conversion rates prediction model, and the real time data counted and the offline number counted can be combined
According to integrate the carry out quality assessment to new edition ad conversion rates prediction model and old edition ad conversion rates prediction model, entering one
Step improves the good and bad accuracy assessed of ad conversion rates prediction model.
Referring to Fig. 4, Fig. 4 is a kind of schematic block diagram for server that yet another embodiment of the invention provides.Sheet as shown in Figure 4
Server 400 in embodiment can include:One or more processors 401, one or more input equipments 402, one or
Multiple then output equipments 403 and one or more memories 404.Above-mentioned processor 401, then input equipment 402, output equipment
403 and memory 404 mutual communication is completed by communication bus 405.Memory 404 is used to store computer program, institute
Stating computer program includes programmed instruction.Processor 401 is used for the programmed instruction for performing the storage of memory 404.Wherein, processor
401 are arranged to call described program instruction to perform following operate:
Ad conversion rates prediction model call request is received, random call new edition ad conversion rates prediction model or old edition are wide
That accuses conversion ratio prediction model calculating candidate locations estimates conversion ratio, records model corresponding to the candidate locations and calls mark;
The conversion ratio of estimating of the candidate locations is sent to advertising platform, makes the advertising platform wide according to the candidate
Accuse estimate conversion ratio and default advertisement exposure strategy is exposed to the candidate locations;
Model calls mark according to corresponding to having exposed advertisement, calculates corresponding to the new edition ad conversion rates prediction model
Second actual advertisement conversion ratio corresponding to first actual advertisement conversion ratio and the old edition ad conversion rates prediction model;
According to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and default assessment strategy,
Good and bad assessment is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model.
Further, the new edition ad conversion rates prediction model includes new edition ad click rate prediction model and new edition is wide
Accuse download rate prediction model;The old edition ad conversion rates prediction model includes old edition ad click rate prediction model and old edition is wide
Accuse download rate prediction model;Processor 401 is specifically arranged to call described program instruction to perform following operate:
Ad conversion rates prediction model call request is received, if being currently in the first default test period, detects institute
State the type of candidate locations corresponding to ad conversion rates prediction model call request;
If the type of the candidate locations is the first preset kind, new edition ad click rate estimates mould described in random call
What type or the old edition ad click rate prediction model calculated the candidate locations estimates clicking rate, by the pre- of the candidate locations
That estimates that clicking rate is identified as the candidate locations estimates conversion ratio;
If the type of the candidate locations is the second preset kind, new edition ad click rate estimates mould described in random call
Type or the old edition ad click rate prediction model calculate the clicking rate of estimating of the candidate locations, call under the first default advertisement
What load rate prediction model calculated the candidate locations estimates download rate, estimating the candidate locations clicking rate and estimate download
What the product of rate was identified as the candidate locations estimates conversion ratio.
Further, processor 401 is specific is arranged to call described program instruction to perform following operate:
Ad conversion rates prediction model call request is received, if being currently in the second default test period, detects institute
State the type of candidate locations corresponding to ad conversion rates prediction model call request;
If the type of the candidate locations is the first preset kind, the first default ad click rate prediction model meter is called
That calculates the candidate locations estimates clicking rate, and the clicking rate of estimating of the candidate locations is identified as into estimating for the candidate locations
Conversion ratio;
If the type of the candidate locations is the second preset kind, the first default ad click rate prediction model meter is called
The clicking rate of estimating of the candidate locations is calculated, under new edition advertisement download rate prediction model described in random call or the old edition advertisement
What load rate prediction model calculated the candidate locations estimates download rate, estimating the candidate locations clicking rate and estimate download
What the product of rate was identified as the candidate locations estimates conversion ratio.
Further, processor 401 is specific is arranged to call described program instruction to perform following operate:
Model calls mark, new edition ad conversion rates prediction model pair described in real-time statistics according to corresponding to having exposed advertisement
The the first accumulative advertisement exposure amount answered and the first accumulative advertising conversion amount, according to the described first accumulative advertisement exposure amount and described the
One accumulative advertising conversion amount, calculates real-time first actual advertisement conversion ratio corresponding to the new edition ad conversion rates prediction model;
Model calls mark, old edition ad conversion rates prediction model pair described in real-time statistics according to corresponding to having exposed advertisement
The the second accumulative advertisement exposure amount answered and the second accumulative advertising conversion amount, according to the described second accumulative advertisement exposure amount and described the
Two accumulative advertising conversion amounts, calculate real-time second actual advertisement conversion ratio corresponding to the old edition ad conversion rates prediction model.
Further, processor 401 is specific is arranged to call described program instruction to perform following operate:
According to the real-time first actual advertisement conversion ratio, the real-time second actual advertisement conversion ratio and default carry
Rate calculative strategy is risen, calculates the new edition ad conversion rates prediction model relative to the old edition ad conversion rates prediction model
Enhancing rate;
Real-time participation number and real-time conversion people according to corresponding to the new edition ad conversion rates prediction model counted
Number is participated in real time corresponding to several, described old edition ad conversion rates prediction model and converts number in real time, default Z score calculates
Z score under the conditions of policy calculation null hypothesis;
The new edition ad conversion rates prediction model and the old edition advertisement are turned according to the enhancing rate and the Z score
Rate prediction model carries out good and bad assessment.
Further, processor 401 is specific is arranged to call described program instruction to perform following operate:
Model calls mark according to corresponding to having exposed advertisement, counts the new edition ad conversion rates in preset period of time and estimates
3rd accumulative advertisement exposure amount corresponding to model and the 3rd accumulative advertising conversion amount, according to the described 3rd accumulative advertisement exposure amount and
Described 3rd accumulative advertising conversion amount, is calculated in preset period of time offline first corresponding to the new edition ad conversion rates prediction model
Actual advertisement conversion ratio;
Model calls mark according to corresponding to having exposed advertisement, counts the old edition ad conversion rates in preset period of time and estimates
4th accumulative advertisement exposure amount corresponding to model and the 4th accumulative advertising conversion amount, according to the described 4th accumulative advertisement exposure amount and
Described 4th accumulative advertising conversion amount, is calculated in preset period of time offline second corresponding to the old edition ad conversion rates prediction model
Actual advertisement conversion ratio.
Further, processor 401 is specific is arranged to call described program instruction to perform following operate:
According to offline first actual advertisement conversion ratio, the offline second actual advertisement conversion ratio and the default order
Maintain strategy, quality is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model and commented
Estimate.
It should be appreciated that in embodiments of the present invention, alleged processor 401 can be CPU (Central
Processing Unit, CPU), the processor can also be other general processors, digital signal processor (Digital
Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit,
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other FPGAs
Device, discrete gate or transistor logic, discrete hardware components etc..General processor can be microprocessor or this at
It can also be any conventional processor etc. to manage device.
Input equipment 402 can include Trackpad, fingerprint adopt sensor (finger print information that is used to gathering user and fingerprint
Directional information), microphone etc., output equipment 403 can include display (LCD etc.), loudspeaker etc..
The memory 404 can include read-only storage and random access memory, and to processor 401 provide instruction and
Data.The a part of of memory 404 can also include nonvolatile RAM.For example, memory 404 can also be deposited
Store up the information of device type.
In the specific implementation, processor 401, input equipment 402, the output equipment 403 described in the embodiment of the present invention can
In the first embodiment and second embodiment that perform the method for assessment ad conversion rates prediction model provided in an embodiment of the present invention
Described implementation, the implementation of the server described by the embodiment of the present invention is also can perform, will not be repeated here.
A kind of computer-readable recording medium, the computer-readable storage medium are provided in another embodiment of the invention
Matter is stored with computer program, and the computer program includes programmed instruction, and described program instruction is realized when being executed by processor:
Ad conversion rates prediction model call request is received, random call new edition ad conversion rates prediction model or old edition are wide
That accuses conversion ratio prediction model calculating candidate locations estimates conversion ratio, records model corresponding to the candidate locations and calls mark;
The conversion ratio of estimating of the candidate locations is sent to advertising platform, makes the advertising platform wide according to the candidate
Accuse estimate conversion ratio and default advertisement exposure strategy is exposed to the candidate locations;
Model calls mark according to corresponding to having exposed advertisement, calculates corresponding to the new edition ad conversion rates prediction model
Second actual advertisement conversion ratio corresponding to first actual advertisement conversion ratio and the old edition ad conversion rates prediction model;
According to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and default assessment strategy,
Good and bad assessment is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model.
Further, the new edition ad conversion rates prediction model includes new edition ad click rate prediction model and new edition is wide
Accuse download rate prediction model;The old edition ad conversion rates prediction model includes old edition ad click rate prediction model and old edition is wide
Accuse download rate prediction model;The computer program implements when being executed by processor:
Ad conversion rates prediction model call request is received, if being currently in the first default test period, detects institute
State the type of candidate locations corresponding to ad conversion rates prediction model call request;
If the type of the candidate locations is the first preset kind, new edition ad click rate estimates mould described in random call
What type or the old edition ad click rate prediction model calculated the candidate locations estimates clicking rate, by the pre- of the candidate locations
That estimates that clicking rate is identified as the candidate locations estimates conversion ratio;
If the type of the candidate locations is the second preset kind, new edition ad click rate estimates mould described in random call
Type or the old edition ad click rate prediction model calculate the clicking rate of estimating of the candidate locations, call under the first default advertisement
What load rate prediction model calculated the candidate locations estimates download rate, estimating the candidate locations clicking rate and estimate download
What the product of rate was identified as the candidate locations estimates conversion ratio.
Further, implemented when the computer program is executed by processor:
Ad conversion rates prediction model call request is received, if being currently in the second default test period, detects institute
State the type of candidate locations corresponding to ad conversion rates prediction model call request;
If the type of the candidate locations is the first preset kind, the first default ad click rate prediction model meter is called
That calculates the candidate locations estimates clicking rate, and the clicking rate of estimating of the candidate locations is identified as into estimating for the candidate locations
Conversion ratio;
If the type of the candidate locations is the second preset kind, the first default ad click rate prediction model meter is called
The clicking rate of estimating of the candidate locations is calculated, under new edition advertisement download rate prediction model described in random call or the old edition advertisement
What load rate prediction model calculated the candidate locations estimates download rate, estimating the candidate locations clicking rate and estimate download
What the product of rate was identified as the candidate locations estimates conversion ratio.
Further, implemented when the computer program is executed by processor:
Model calls mark, new edition ad conversion rates prediction model pair described in real-time statistics according to corresponding to having exposed advertisement
The the first accumulative advertisement exposure amount answered and the first accumulative advertising conversion amount, according to the described first accumulative advertisement exposure amount and described the
One accumulative advertising conversion amount, calculates real-time first actual advertisement conversion ratio corresponding to the new edition ad conversion rates prediction model;
Model calls mark, old edition ad conversion rates prediction model pair described in real-time statistics according to corresponding to having exposed advertisement
The the second accumulative advertisement exposure amount answered and the second accumulative advertising conversion amount, according to the described second accumulative advertisement exposure amount and described the
Two accumulative advertising conversion amounts, calculate real-time second actual advertisement conversion ratio corresponding to the old edition ad conversion rates prediction model.
Further, implemented when the computer program is executed by processor:
According to the real-time first actual advertisement conversion ratio, the real-time second actual advertisement conversion ratio and default carry
Rate calculative strategy is risen, calculates the new edition ad conversion rates prediction model relative to the old edition ad conversion rates prediction model
Enhancing rate;
Real-time participation number and real-time conversion people according to corresponding to the new edition ad conversion rates prediction model counted
Number is participated in real time corresponding to several, described old edition ad conversion rates prediction model and converts number in real time, default Z score calculates
Z score under the conditions of policy calculation null hypothesis;
The new edition ad conversion rates prediction model and the old edition advertisement are turned according to the enhancing rate and the Z score
Rate prediction model carries out good and bad assessment.
Further, implemented when the computer program is executed by processor:
Model calls mark according to corresponding to having exposed advertisement, counts the new edition ad conversion rates in preset period of time and estimates
3rd accumulative advertisement exposure amount corresponding to model and the 3rd accumulative advertising conversion amount, according to the described 3rd accumulative advertisement exposure amount and
Described 3rd accumulative advertising conversion amount, is calculated in preset period of time offline first corresponding to the new edition ad conversion rates prediction model
Actual advertisement conversion ratio;
Model calls mark according to corresponding to having exposed advertisement, counts the old edition ad conversion rates in preset period of time and estimates
4th accumulative advertisement exposure amount corresponding to model and the 4th accumulative advertising conversion amount, according to the described 4th accumulative advertisement exposure amount and
Described 4th accumulative advertising conversion amount, is calculated in preset period of time offline second corresponding to the old edition ad conversion rates prediction model
Actual advertisement conversion ratio.
Further, implemented when the computer program is executed by processor:
According to offline first actual advertisement conversion ratio, the offline second actual advertisement conversion ratio and the default order
Maintain strategy, quality is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model and commented
Estimate.
The computer-readable recording medium can be the internal storage unit of the server described in foregoing any embodiment,
Such as the hard disk or internal memory of server.The computer-readable recording medium can also be that the external storage of the server is set
Plug-in type hard disk that is standby, such as being equipped with the server, intelligent memory card (Smart Media Card, SMC), secure digital
(Secure Digital, SD) blocks, flash card (Flash Card) etc..Further, the computer-readable recording medium is also
The internal storage unit of the server can both be included or including External memory equipment.The computer-readable recording medium is used
In other programs and data needed for the storage computer program and the server.The computer-readable recording medium is also
It can be used for temporarily storing the data that has exported or will export.
Those of ordinary skill in the art are it is to be appreciated that the list of each example described with reference to the embodiments described herein
Member and algorithm steps, it can be realized with electronic hardware, computer software or the combination of the two, in order to clearly demonstrate hardware
With the interchangeability of software, the composition and step of each example are generally described according to function in the above description.This
A little functions are performed with hardware or software mode actually, application-specific and design constraint depending on technical scheme.Specially
Industry technical staff can realize described function using distinct methods to each specific application, but this realization is not
It is considered as beyond the scope of this invention.
It is apparent to those skilled in the art that for convenience of description and succinctly, the clothes of foregoing description
The specific work process of business device and unit, may be referred to the corresponding process in preceding method embodiment, will not be repeated here.
In the several embodiments provided are originally provided, it should be understood that disclosed server and method, can pass through
Other modes are realized.For example, device embodiment described above is only schematical, for example, the division of the unit,
Only a kind of division of logic function, can there is an other dividing mode when actually realizing, such as multiple units or component can be with
With reference to or be desirably integrated into another system, or some features can be ignored, or not perform.It is in addition, shown or discussed
Mutual coupling or direct-coupling or communication connection can be the INDIRECT COUPLINGs or logical by some interfaces, device or unit
Letter connection or electricity, the connection of mechanical or other forms.
The unit illustrated as separating component can be or may not be physically separate, show as unit
The part shown can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple
On NE.Some or all of unit therein can be selected to realize scheme of the embodiment of the present invention according to the actual needs
Purpose.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, can also
It is that unit is individually physically present or two or more units are integrated in a unit.It is above-mentioned integrated
Unit can both be realized in the form of hardware, can also be realized in the form of SFU software functional unit.
If the integrated unit is realized in the form of SFU software functional unit and is used as independent production marketing or use
When, it can be stored in a computer read/write memory medium.Based on such understanding, technical scheme is substantially
The part to be contributed in other words to prior art, or all or part of the technical scheme can be in the form of software product
Embody, the computer software product is stored in a storage medium, including some instructions are causing a computer
Equipment (can be personal computer, server, or network equipment etc.) performs the complete of each embodiment methods described of the present invention
Portion or part steps.And foregoing storage medium includes:USB flash disk, mobile hard disk, read-only storage (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can store journey
The medium of sequence code.
The foregoing is only a specific embodiment of the invention, but protection scope of the present invention is not limited thereto, any
Those familiar with the art the invention discloses technical scope in, various equivalent modifications can be readily occurred in or replaced
Change, these modifications or substitutions should be all included within the scope of the present invention.Therefore, protection scope of the present invention should be with right
It is required that protection domain be defined.
Claims (10)
- A kind of 1. method for assessing ad conversion rates prediction model, it is characterised in that including:Ad conversion rates prediction model call request is received, random call new edition ad conversion rates prediction model or old edition advertisement turn Rate prediction model calculating candidate locations estimate conversion ratio, record model corresponding to the candidate locations and call mark;The conversion ratio of estimating of the candidate locations is sent to advertising platform, makes the advertising platform according to the candidate locations Estimate conversion ratio and default advertisement exposure strategy is exposed to the candidate locations;Model calls mark according to corresponding to having exposed advertisement, calculates first corresponding to the new edition ad conversion rates prediction model Second actual advertisement conversion ratio corresponding to actual advertisement conversion ratio and the old edition ad conversion rates prediction model;According to the first actual advertisement conversion ratio, the second actual advertisement conversion ratio and default assessment strategy, to institute State new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model carries out good and bad assessment.
- 2. according to the method for claim 1, it is characterised in that it is wide that the new edition ad conversion rates prediction model includes new edition Accuse clicking rate prediction model and new edition advertisement download rate prediction model;It is wide that the old edition ad conversion rates prediction model includes old edition Accuse clicking rate prediction model and old edition advertisement download rate prediction model;The reception ad conversion rates prediction model call request, random call new edition ad conversion rates prediction model or old edition are wide That accuses conversion ratio prediction model calculating candidate locations estimates conversion ratio, records model corresponding to the candidate locations and calls mark, Including:Ad conversion rates prediction model call request is received, if being currently in the first default test period, is detected described wide Accuse the type of candidate locations corresponding to conversion ratio prediction model call request;If the type of the candidate locations is the first preset kind, new edition ad click rate prediction model described in random call or The old edition ad click rate prediction model calculating candidate locations estimate clicking rate, by estimating a little for the candidate locations What the rate of hitting was identified as the candidate locations estimates conversion ratio;If the type of the candidate locations is the second preset kind, new edition ad click rate prediction model described in random call or The old edition ad click rate prediction model calculates the clicking rate of estimating of the candidate locations, calls the first default advertisement download rate What prediction model calculated the candidate locations estimates download rate, estimating the candidate locations clicking rate and estimate download rate What product was identified as the candidate locations estimates conversion ratio.
- 3. according to the method for claim 2, it is characterised in that the reception ad conversion rates prediction model call request, What random call new edition ad conversion rates prediction model or old edition ad conversion rates prediction model calculated candidate locations estimates conversion Rate, record model corresponding to the candidate locations and call mark, in addition to:Ad conversion rates prediction model call request is received, if being currently in the second default test period, is detected described wide Accuse the type of candidate locations corresponding to conversion ratio prediction model call request;If the type of the candidate locations is the first preset kind, the first default ad click rate prediction model is called to calculate institute That states candidate locations estimates clicking rate, by the candidate locations estimate that clicking rate is identified as the candidate locations estimate conversion Rate;If the type of the candidate locations is the second preset kind, the first default ad click rate prediction model is called to calculate institute The clicking rate of estimating of candidate locations is stated, new edition advertisement download rate prediction model described in random call or the old edition advertisement download rate What prediction model calculated the candidate locations estimates download rate, estimating the candidate locations clicking rate and estimate download rate What product was identified as the candidate locations estimates conversion ratio.
- 4. according to the method for claim 1, it is characterised in that the basis has exposed model corresponding to advertisement and called mark Know, calculate the first actual advertisement conversion ratio corresponding to the new edition ad conversion rates prediction model and the old edition ad conversion rates Second actual advertisement conversion ratio corresponding to prediction model, including:The model according to corresponding to having exposed advertisement, which calls, to be identified, corresponding to new edition ad conversion rates prediction model described in real-time statistics First accumulative advertisement exposure amount and the first accumulative advertising conversion amount, tire out according to the described first accumulative advertisement exposure amount and described first Advertising conversion amount is counted, calculates real-time first actual advertisement conversion ratio corresponding to the new edition ad conversion rates prediction model;The model according to corresponding to having exposed advertisement, which calls, to be identified, corresponding to old edition ad conversion rates prediction model described in real-time statistics Second accumulative advertisement exposure amount and the second accumulative advertising conversion amount, tire out according to the described second accumulative advertisement exposure amount and described second Advertising conversion amount is counted, calculates real-time second actual advertisement conversion ratio corresponding to the old edition ad conversion rates prediction model.
- 5. according to the method for claim 4, it is characterised in that it is described according to the first actual advertisement conversion ratio, it is described Second actual advertisement conversion ratio and default assessment strategy, it is wide to the new edition ad conversion rates prediction model and the old edition Accuse conversion ratio prediction model and carry out good and bad assessment, including:According to real-time first actual advertisement conversion ratio, the real-time second actual advertisement conversion ratio and the default enhancing rate Calculative strategy, calculate lifting of the new edition ad conversion rates prediction model relative to the old edition ad conversion rates prediction model Rate;Real-time participation number and real-time conversion number, institute according to corresponding to the new edition ad conversion rates prediction model counted State real-time participation number corresponding to old edition ad conversion rates prediction model and real-time conversion number, default Z score calculative strategy Calculate the Z score under the conditions of null hypothesis;According to the enhancing rate and the Z score to the new edition ad conversion rates prediction model and the old edition ad conversion rates Prediction model carries out good and bad assessment.
- 6. according to the method for claim 1, it is characterised in that the basis has exposed model corresponding to advertisement and called mark Know, calculate the first actual advertisement conversion ratio corresponding to the new edition ad conversion rates prediction model and the old edition ad conversion rates Second actual advertisement conversion ratio corresponding to prediction model, including:Model calls mark according to corresponding to having exposed advertisement, counts the new edition ad conversion rates prediction model in preset period of time Corresponding 3rd accumulative advertisement exposure amount and the 3rd accumulative advertising conversion amount, according to the described 3rd accumulative advertisement exposure amount and described 3rd accumulative advertising conversion amount, calculate offline first reality corresponding to the new edition ad conversion rates prediction model in preset period of time Ad conversion rates;Model calls mark according to corresponding to having exposed advertisement, counts the old edition ad conversion rates prediction model in preset period of time Corresponding 4th accumulative advertisement exposure amount and the 4th accumulative advertising conversion amount, according to the described 4th accumulative advertisement exposure amount and described 4th accumulative advertising conversion amount, calculate offline second reality corresponding to the old edition ad conversion rates prediction model in preset period of time Ad conversion rates.
- 7. according to the method for claim 6, it is characterised in that it is described according to the first actual advertisement conversion ratio, it is described Second actual advertisement conversion ratio and default assessment strategy, it is wide to the new edition ad conversion rates prediction model and the old edition Accuse conversion ratio prediction model and carry out good and bad assessment, including:According to offline first actual advertisement conversion ratio, the offline second actual advertisement conversion ratio and the default rank tests Strategy, good and bad assessment is carried out to the new edition ad conversion rates prediction model and the old edition ad conversion rates prediction model.
- 8. a kind of server, it is characterised in that including for performing the method as described in claim 1-7 any claims Unit.
- A kind of 9. server, it is characterised in that including processor, input equipment, output equipment and memory, the processor, Input equipment, output equipment and memory are connected with each other, wherein, the memory is used to store computer program, the calculating Machine program includes programmed instruction, and the processor is arranged to call described program instruction, performed as claim 1-7 is any Method described in.
- 10. a kind of computer-readable recording medium, it is characterised in that the computer-readable recording medium storage has computer journey Sequence, the computer program include programmed instruction, and described program instruction makes the computing device such as when being executed by a processor Method described in claim any one of 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711120819.1A CN107886361A (en) | 2017-11-14 | 2017-11-14 | A kind of method and server for assessing ad conversion rates prediction model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711120819.1A CN107886361A (en) | 2017-11-14 | 2017-11-14 | A kind of method and server for assessing ad conversion rates prediction model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107886361A true CN107886361A (en) | 2018-04-06 |
Family
ID=61777064
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711120819.1A Pending CN107886361A (en) | 2017-11-14 | 2017-11-14 | A kind of method and server for assessing ad conversion rates prediction model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107886361A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108921624A (en) * | 2018-07-27 | 2018-11-30 | 百度在线网络技术(北京)有限公司 | advertisement fusion method, device, storage medium and terminal device |
CN108985823A (en) * | 2018-06-27 | 2018-12-11 | 腾讯科技(深圳)有限公司 | A kind of information distribution method, device, server and storage medium |
CN109087136A (en) * | 2018-07-25 | 2018-12-25 | 上海驰游信息技术有限公司 | A kind of method of adjustment and device of advertising resource value |
CN110930193A (en) * | 2019-11-22 | 2020-03-27 | 浙江大搜车软件技术有限公司 | Advertisement conversion rate evaluation method and device, computer equipment and storage medium |
CN111131356A (en) * | 2018-10-31 | 2020-05-08 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
CN112232853A (en) * | 2020-09-22 | 2021-01-15 | 北京明略昭辉科技有限公司 | Conversion rate calculation method and device, storage medium and electronic equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102385729A (en) * | 2011-10-25 | 2012-03-21 | 北京亿赞普网络技术有限公司 | Method and device for evaluating advertisement serving policy |
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 |
CN107301247A (en) * | 2017-07-14 | 2017-10-27 | 广州优视网络科技有限公司 | Set up the method and device, terminal, storage medium of clicking rate prediction model |
-
2017
- 2017-11-14 CN CN201711120819.1A patent/CN107886361A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102385729A (en) * | 2011-10-25 | 2012-03-21 | 北京亿赞普网络技术有限公司 | Method and device for evaluating advertisement serving policy |
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 |
CN107301247A (en) * | 2017-07-14 | 2017-10-27 | 广州优视网络科技有限公司 | Set up the method and device, terminal, storage medium of clicking rate prediction model |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108985823A (en) * | 2018-06-27 | 2018-12-11 | 腾讯科技(深圳)有限公司 | A kind of information distribution method, device, server and storage medium |
CN108985823B (en) * | 2018-06-27 | 2021-06-15 | 腾讯科技(深圳)有限公司 | Information delivery method, device, server and storage medium |
CN109087136A (en) * | 2018-07-25 | 2018-12-25 | 上海驰游信息技术有限公司 | A kind of method of adjustment and device of advertising resource value |
CN109087136B (en) * | 2018-07-25 | 2022-06-14 | 上海驰游信息技术有限公司 | Method and device for adjusting advertisement resource value |
CN108921624A (en) * | 2018-07-27 | 2018-11-30 | 百度在线网络技术(北京)有限公司 | advertisement fusion method, device, storage medium and terminal device |
CN111131356A (en) * | 2018-10-31 | 2020-05-08 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
CN110930193A (en) * | 2019-11-22 | 2020-03-27 | 浙江大搜车软件技术有限公司 | Advertisement conversion rate evaluation method and device, computer equipment and storage medium |
CN110930193B (en) * | 2019-11-22 | 2023-08-29 | 浙江大搜车软件技术有限公司 | Advertisement conversion rate evaluation method, advertisement conversion rate evaluation device, computer equipment and storage medium |
CN112232853A (en) * | 2020-09-22 | 2021-01-15 | 北京明略昭辉科技有限公司 | Conversion rate calculation method and device, storage medium and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107886361A (en) | A kind of method and server for assessing ad conversion rates prediction model | |
US8095612B2 (en) | Ranking messages in an electronic messaging environment | |
CN108470253B (en) | User identification method and device and storage equipment | |
CN107357718A (en) | One kind structure method of testing and device | |
US20210090017A1 (en) | Feedback-based management of delivery orders | |
CN109474483A (en) | A kind of detection method, detection device and the terminal device of unit exception situation | |
CN108230010A (en) | A kind of method and server for estimating ad conversion rates | |
CN107508860A (en) | One kind service current-limiting method, server and terminal | |
CN107507036A (en) | The method and terminal of a kind of data prediction | |
CN108038112A (en) | Document handling method, mobile terminal and computer-readable recording medium | |
CN109034867B (en) | Click traffic detection method and device and storage medium | |
CN109600344A (en) | Identify the method, apparatus and electronic equipment of risk group | |
CN110750530B (en) | Service system and data checking method thereof | |
CN103164505A (en) | System and method for conducting ranking on internet applications based on actual use frequency | |
CN109274639A (en) | The recognition methods of open platform abnormal data access and device | |
WO2023000491A1 (en) | Application recommendation method, apparatus and device, and computer-readable storage medium | |
CN107506494A (en) | Document handling method, mobile terminal and computer-readable recording medium | |
CN108694588B (en) | Virtual resource request method and terminal equipment | |
CN107679222A (en) | Image processing method, mobile terminal and computer-readable recording medium | |
CN105989103A (en) | Method for recommending application program and terminal | |
CN107330058A (en) | Application searches method, mobile terminal and computer-readable recording medium | |
CN105938606B (en) | Apparatus and method for providing account book service | |
CN107332988A (en) | Information processing method, mobile terminal and computer-readable recording medium | |
WO2023066258A1 (en) | Data processing method and apparatus for private data, computer device and medium | |
CN115496609A (en) | Data processing method, device and equipment based on financial product and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180406 |
|
WD01 | Invention patent application deemed withdrawn after publication |