CN109377280A - A kind of order ads mechanism generation method and generate system - Google Patents
A kind of order ads mechanism generation method and generate system Download PDFInfo
- Publication number
- CN109377280A CN109377280A CN201811260738.6A CN201811260738A CN109377280A CN 109377280 A CN109377280 A CN 109377280A CN 201811260738 A CN201811260738 A CN 201811260738A CN 109377280 A CN109377280 A CN 109377280A
- Authority
- CN
- China
- Prior art keywords
- service
- increment
- accounting
- business revenue
- sequence index
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- G06Q30/0244—Optimization
Abstract
The invention discloses a kind of order ads mechanism generation method and system is generated, is related to calculating order ads mechanism technical field, it is intended to solve the problems, such as that existing order ads mechanism cannot support that the quantity of advertisement show position is 0 scene.Its key points of the technical solution are that: a kind of characteristic parameter section step S1, is configured, service of a kind of characteristic parameter outside a kind of characteristic parameter section is filtered;Step S2, forecast sample is constructed, by categorised decision algorithm and forecast sample, obtains buying rate of the user to service and the conversion ratio to main demand;Step S3, the displaying number that the quantity for generating advertisement show position is 0 accounts for the list for not showing accounting of total displaying number, according to accounting is not shown, calculates corresponding service business revenue increment and generate the sequence index calculation formula that the main single amount increment/service business revenue increment of demand conversion reaches setting value.The scene that the quantity that the technical solution of the application has the effect of can adapt to advertisement show position is 0.
Description
Technical field
The present invention relates to order ads mechanism technical field is calculated, more particularly, to a kind of order ads mechanism generation method
And generate system.
Background technique
At present, it is more taken out by the ordering mechanism of order ads index of platform business revenue as Strategy mould
Block is individually studied, such as Taobao, from two advertiser's business revenue+α × platform business revenue, platform business revenue sequence index Design OCPC collators
System has reached the purpose for promoting advertiser's business revenue and platform business revenue to be ranked up to candidate service.
Based on the purpose of above-mentioned promotion advertiser business revenue and platform business revenue, some internet platforms and Internet enterprises are also pushed away
Go out advertisement on different lines and launches scheme, such as:
D1: Beijing Qihu Technology Co., Ltd. is application No. is the China of " 201510947002 " filed on December 16th, 2015
Patent, it discloses a kind of reminding window advertisement positions to launch control method and device, by multiple subwindows advertisement position and
Different types of child window advertisement position carries out dispensing waiting, so that the reminding window for popping up browser client is in multiple subwindows
Advertisement position carries out advertising display, it can supports multiple subwindows advertisement position to show, keeps the ad content showed richer, and can
To facilitate the ad data to different child window advertisement positions and different types of child window advertisement position to be adjusted, effectively improve
Displaying efficiency.
D2: Sina's network technology (China) Co., Ltd filed on December 29th, 2016 application No. is
The Chinese patent of " 201611242538 ", it discloses method and device is launched on a kind of displaying advertisement line, this method comprises: choosing
All advertisements for meeting current business demand are selected as the first candidate being shown on Current ad position to page browsing person
Set of advertisements;It obtains the first candidate locations and concentrates the clicking rate threshold value of each advertisement, and it is each wide to estimate the first candidate locations concentration
That accuses estimates clicking rate;Clicking rate and clicking rate threshold value are estimated according to what the first candidate locations concentrated each advertisement, and judgement is each
Whether advertisement selects current page viewer, using all advertisements for selecting current page viewer as the second candidate locations collection;
The dispensing anxiety coefficient of each advertisement is concentrated according to the second candidate locations and estimates clicking rate, chooses exhibition for current page viewer
Show advertisement;The displaying advertisement chosen is shown page browsing person, to improve the clicking rate of advertisement.
From above-mentioned document it is found that promoted currently based on advertising display efficiency and ad click rate is improved advertiser's business revenue with
Platform business revenue has all had more perfect technical solution, and still, the quantity of advertisement show position is a static state in the prior art
Value N(N >=1), to the quantity of advertisement show position can be 0 scene the considerations of, as the quantity of advertisement show position may be
One positive number N greater than 0, it is also possible to be 0, in the case where pairing unsalable goods up with goods that sell well service scenarios, the quantity that often will appear advertisement show position is 0
Situation, but the prior art is not supported, for this purpose, present applicant proposes a kind of new schemes.
Summary of the invention
The object of the present invention is to provide a kind of order ads mechanism generation method and system is generated, having can adapt to extensively
It accuses and shows that the quantity of position is the effect of 0 scene.
Foregoing invention purpose one of the invention has the technical scheme that
A kind of order ads mechanism generation method, comprising the following steps:
Step S1, configuration allows to show a kind of characteristic parameter section serviced, filters a kind of characteristic parameter in a category feature
Service outside parameter section, remaining service, which is put into, recalls pond;
Step S2, forecast sample is constructed according to historical data information, by categorised decision algorithm and the forecast sample, obtained
Buying rate of the user to service and the conversion ratio to main demand;
Step S3, the displaying number that the quantity of advertisement show position is 0 is generated according to historical data information and accounts for total displaying number
The list for not showing accounting does not show purchase of the user to service in accounting and step S2 according in the list for not showing accounting
Rate and the conversion ratio to main demand are bought, calculates the single amount increment of corresponding main demand conversion and service business revenue increment, and generate main need
The single amount increment/service business revenue increment of conversion is asked to reach the sequence index calculation formula of setting value.
By using above-mentioned technical proposal, by not showing that accounting can capture the dynamic of the quantity of advertisement show position,
Quantity suitable for advertisement show position can be 0 scene.On the other hand, use can directly be fed back by converting single amount by main demand
Family is experienced, and converting list amount increment/service business revenue increment class ROI index by main demand directly feedback user can experience and put down
Relationship between platform business revenue, it is more direct to the optimization of user experience.
The present invention is further arranged to: the step S1 includes following sub-step:
S11, configuration allow to show the category feature upper limit parameter serviced, filter a kind of characteristic parameter and be greater than or equal to a category feature
The service of upper limit parameter, remaining service, which is put into, recalls pond;
The number serviced in pond is recalled in S12, detection, if recalling the number serviced in pond is 0, one in the additional service filtered
Again investment recalls pond to the smallest service of category feature parameter.
By using above-mentioned technical proposal, data can effectively be screened, the quantity suitable for advertisement show position can
Think 0 scene.On the other hand, by the setting of a category feature upper limit parameter, user experience is also increased.
The present invention is further arranged to: the step S2 includes following sub-step:
S21, it obtains historical data and assembles them into characteristic data set, categorised decision algorithm is used on the characteristic data set
Practise out service buying rate prediction device and main demand conversion ratio prediction device;
The service in pond is recalled described in S22, poll, splicing the service with feature described in step S21 is forecast sample;
S23, using the forecast sample as entering ginseng, input service buying rate prediction device, obtain different user in different scenes and
To the buying rate of service when pairing unsalable goods up with goods that sell well different services;Using the forecast sample as ginseng is entered, main demand conversion ratio prediction device is inputted, is obtained
To different user different scenes with when pairing unsalable goods up with goods that sell well different services to the conversion ratio of main demand.
By using above-mentioned technical proposal, to clothes when conveniently estimating (different user pairs unsalable goods up with goods that sell well different services in different scenes)
The buying rate of business, to the conversion ratio of main demand, increase the reliability of data.
The present invention is further arranged to: the step S21 includes following sub-step:
The historical data that S211, acquisition are estimated, system assemble the use according to the user id in the historical data
The user tag of family id;
S212, according to the time class data assembling in the user tag at including practicing the characteristic data set of category feature;
S213, learnt to service buying rate prediction device and main demand conversion ratio out with categorised decision algorithm on the characteristic data set
Prediction device.
By using above-mentioned technical proposal, category feature is practiced by user id, so that the data precision is higher, more conducively
Optimize user experience, and then promotes advertiser's business revenue and platform business revenue.
The present invention is further arranged to: the step S3 includes following sub-step:
S31, historical data is obtained, generates the displaying number that the quantity of advertisement show position is 0 by mono- step-length of m% and accounts for total displaying
The list for not showing accounting of number is generated the list of the factor alpha in sequence index calculation formula by mono- step-length of n;
S32, according to buying rate of the user in step S23 to service and the conversion ratio to main demand, calculating does not show accounting and is
Main demand in the binary group of number α converts single amount increment and service business revenue increment, and acquisition does not show that accounting is not shown with this and accounts for
The single amount increment/service business revenue increment of winner's demand conversion is set to reach the mapping table of maximum α than under;
In S33, mapping table in step s 32, a α is randomly choosed, generates initial sequence index calculation formula;
S34, according to the initial sequence index calculation formula, calculate the sequence index value of each advertising display request on line simultaneously
Sequence index series is formed according to the sequence index value;
S35, the quantile of the sequence index series is obtained by mono- step-length of q, corresponding service is calculated according to the quantile
Business revenue increment, and export the mapping table of the service business revenue increment under quantile and the quantile;
S36, the service business revenue increment according to setting, obtain corresponding quantile in mapping table in step s 35, and by this point
Digit obtains corresponding α as corresponding do not show in the mapping table of accounting in step s 32;
S37, the α in the sequence index calculation formula is updated according to the α in step S36, according to updated sequence index meter
Formula is calculated, the mapping table in step S35 is updated.
By using above-mentioned technical proposal, can be adjusted to make main demand in linkage according to desired service business revenue increment
The ordering mechanism that the single amount increment/service business revenue increment class ROI index of conversion is optimal, strengthens to user experience and service
The control of relationship between business revenue, so that user experience is more preferably.
Foregoing invention purpose two of the invention has the technical scheme that
A kind of order ads mechanism generation system, comprising:
Limitation module is serviced, being used to configure allows to show that a kind of characteristic parameter in a kind of characteristic parameter section serviced, filtering exists
It service outside one kind characteristic parameter section and remaining service is put into recalls pond;
Module is estimated, is used to construct forecast sample according to historical data information, and pass through categorised decision algorithm and described pre-
Buying rate of this acquisition of test sample user to service and the conversion ratio to main demand;
Ordering mechanism generation module is used to generate the displaying number that the quantity of advertisement show position is 0 according to historical data information
Account for the list for not showing accounting of total displaying number;
Wherein, the ordering mechanism generation module is also used to not show accounting and user according to not showing in the list of accounting
Buying rate to service and the conversion ratio to main demand calculate the single amount increment of corresponding main demand conversion and service business revenue increment,
And generate the sequence index calculation formula that the main single amount increment/service business revenue increment of demand conversion reaches setting value.
By using above-mentioned technical proposal, ordering mechanism generation module can capture the dynamic of the quantity of advertisement show position
Property, the quantity suitable for advertisement show position can be 0 scene.On the other hand, converting single amount by main demand can be directly anti-
User experience is presented, converting single amount increment/service business revenue increment class ROI index by main demand being capable of directly feedback user experience
It is more direct to the optimization of user experience with the relationship between platform business revenue.
The present invention is further arranged to: the service limitation module includes:
Ceiling restriction unit, being used to configure allows to show that a kind of characteristic parameter of the category feature upper limit parameter serviced, filtering is big
In or equal to a category feature upper limit parameter service and by remaining service be put into described in recall pond;
Lower limit limiting unit, be used to detect it is described recall the number serviced in pond, and for being serviced in pond when described recall
When number is 0, the smallest service of a kind of characteristic parameter recalls pond described in investment again in the additional service filtered.
By using above-mentioned technical proposal, ceiling restriction unit and lower limit limiting unit are able to carry out effective sieve to data
Choosing, the quantity suitable for advertisement show position can be 0 scene, and due to the setting of a category feature upper limit parameter, filter not
The service for meeting condition, increases user experience.
The present invention is further arranged to: the module of estimating includes:
Prediction device training unit is used to obtain historical data and assembles them into characteristic data set, is also used in the spy
Learnt to service buying rate prediction device and main demand conversion ratio prediction device out with categorised decision algorithm on sign data set;
Sample architecture unit, the service for being used to recall described in poll in pond simultaneously splice the spy that the service is concentrated with characteristic
Sign is forecast sample;
Unit is estimated, is used to obtain different user using the forecast sample as ginseng input service buying rate prediction device is entered and exist
Different scenes with when pairing unsalable goods up with goods that sell well different services to the buying rate of service;It is also used to input main need using the forecast sample as ginseng is entered
Seek conversion ratio prediction device, obtain different user different scenes with when pairing unsalable goods up with goods that sell well different services to the conversion ratio of main demand.
By using above-mentioned technical proposal, to clothes when conveniently estimating (different user pairs unsalable goods up with goods that sell well different services in different scenes)
The buying rate of business, to the conversion ratio of main demand, facilitate and calculate the single amount increment of main demand conversion and service business revenue increment, increase number
According to reliability.
The present invention is further arranged to: the prediction device training unit is also used to according to the user id assembling in historical data
The user tag of the user id, the feature that the characteristic is concentrated includes according to the time class data assembling in user tag
At practice category feature.
By using above-mentioned technical proposal, sufficient crawl is carried out to the category feature of practicing of user id, so that data are quasi-
Exactness is higher, more conducively optimization user experience.
The present invention is further arranged to: the ordering mechanism generation module includes:
Accounting and factor alpha mapping table generation module are not shown, after being used to obtain historical data, generate advertisement by mono- step-length of m%
It shows that the displaying number that the quantity of position is 0 accounts for the list for not showing accounting of total displaying number, and generates row by mono- step-length of n
The list of factor alpha in sequence index calculation formula;It is also used to the conversion according to user to the buying rate of service and to main demand
Rate calculates the single amount increment of main demand conversion and the service business revenue increment not shown in the binary group of accounting and factor alpha, and obtains not
Show that accounting does not show the mapping for making the single amount increment/service business revenue increment of winner's demand conversion reach maximum α under accounting with this
Table one;
Accounting and service business revenue Rise Map table generation module are not shown, are used to randomly choose one in the mapping table one
α generates initial sequence index calculation formula, is asked according to each advertising display on the initial sequence index calculation formula calculating line
The sequence index value asked and sequence index series is formed according to the sequence index value;It is also used to by mono- step-length of q
It obtains the quantile of the sequence index series, calculate corresponding service business revenue increment and output quartile according to the quantile
Several mapping tables two with the service business revenue increment under the quantile;
Sort calculation formula update module, is used to be corresponded in the mapping table two according to the service business revenue increment of setting
Quantile, and do not show that accounting obtains corresponding α in mapping table one using the quantile as corresponding;It is also used to basis
α in the corresponding α more new sort index calculation formula, and mapping table is updated according to updated sequence index calculation formula
Two.
, can be according to desired service business revenue increment by using above-mentioned technical proposal, more new sort index meter in linkage
The α in formula is calculated, to make the main single amount increment/service business revenue of demand conversion according to the generation of updated sequence index calculation formula
The ordering mechanism that the class ROI index of increment is optimal strengthens the control to relationship between user experience and service business revenue, makes
Obtain user experience more preferably.
In conclusion advantageous effects of the invention are as follows:
1. can capture the dynamic of the quantity of advertisement show position by not showing the setting of accounting, adapt to advertisement show position
The scene that quantity is 0;
2. passing through the setting of conversion list amount increment/service business revenue increment class ROI index, directly feedback user it can experience and put down
Relationship between platform business revenue facilitates optimization user experience;
3. passing through the setting of two mapping tables, the control to relationship between user experience and service business revenue is strengthened, is conducive to improve
User experience and service business revenue.
Detailed description of the invention
Fig. 1 is the flow chart of the order ads mechanism generation method shown in the embodiment of the present invention one;
Fig. 2 is the flow chart of the step S1 shown in the embodiment of the present invention two;
Fig. 3 is shown in the embodiment of the present invention two for embodying the flow chart of service id filtering process;
Fig. 4 is the flow chart of the step S2 shown in the embodiment of the present invention two;
Fig. 5 is the flow chart of the step S21 shown in the embodiment of the present invention two;
Fig. 6 is the flow chart of the step S3 shown in the embodiment of the present invention two;
Fig. 7 is the structural schematic diagram of the order ads mechanism generation system shown in the embodiment of the present invention three.
In figure, 10, service limitation module;11, ceiling restriction unit;12, lower limit limiting unit;20, module is estimated;21,
Prediction device training unit;22, sample architecture unit;23, unit is estimated;30, ordering mechanism generation module;31, accounting is not shown
With factor alpha mapping table generation module;32, accounting and service business revenue Rise Map table generation module are not shown;33, sequence calculates public
Formula update module.
Specific embodiment
Below in conjunction with attached drawing, invention is further described in detail.
Embodiment one
Referring to Fig.1, it is a kind of order ads mechanism generation method disclosed by the invention comprising following steps:
Step S1, configuration allows to show a kind of characteristic parameter section serviced, filters a kind of characteristic parameter in a kind of characteristic parameter
Service outside section, remaining service, which is put into, recalls pond.Specifically, in the present embodiment, a kind of characteristic parameter section is service
Price range, the service in price range, which is placed into, recalls pond, for guaranteeing user experience.
Although in the application using price as a kind of characteristic parameter for the application scenarios of the application are introduced, this
Field technical staff is appreciated that the technical solution of the application applies also for a variety of different parameter scenes, as ratio is launched in advertisement
Example, advertisement pushing time etc., the application is not especially limited this.
Step S2, forecast sample is constructed according to historical data information, by categorised decision algorithm and forecast sample, obtained
Buying rate of the user to service and the conversion ratio to main demand.In the present embodiment, categorised decision algorithm is using XGBoost
Learner.
Step S3, the displaying number that the quantity of advertisement show position is 0 is generated according to historical data information and accounts for total displaying time
Several lists for not showing accounting, according to not showing in the list for not showing accounting, user is to service in accounting and step S2
Buying rate and conversion ratio to main demand, calculate the single amount increment of corresponding main demand conversion and service business revenue increment, and generate
The main single amount increment/service business revenue increment class ROI index of demand conversion reaches the sequence index calculation formula of setting value.
It should be noted that passing through main need by not showing that accounting can capture the dynamic of the quantity of advertisement show position
It asks the single amount increment of conversion that can directly feedback user experience, single amount increment/service business revenue increment class ROI is converted by main demand
Index directly can experience the relationship between platform business revenue by feedback user.Wherein, the single amount increment of main demand conversion=(show certain
Main demand conversion ratio-when service does not show main demand conversion ratio when task service) × 1;Service business revenue increment=(show certain
Service buying rate-when service does not show service buying rate when task service) × certain service price;Sort index calculation formula
For (1- α) × the single amount increment+α of the main demand that standardized conversion × the service business revenue increment standardized, α are one between 0
And the real number between 1.
The implementation principle of above-described embodiment are as follows:
By not showing that accounting captures the dynamic of the quantity of advertisement show position, single amount increment is converted by main demand and is directly fed back
User experience converts the direct feedback user experience of single amount increment/service business revenue increment class ROI index and platform by main demand
Relationship between business revenue.By the technical solution of the application, can be adjusted to make in linkage according to desired service business revenue increment
The ordering mechanism that class ROI index is optimal can not only adapt to the scene that the quantity of advertisement show position is 0, also strengthen pair
The control of relationship between user experience and service business revenue.
Embodiment two
Referring to Fig. 2, based on embodiment one, the difference between this embodiment and the first embodiment lies in, step S1 includes following sub-step
It is rapid:
S11, configuration allow to show the category feature upper limit parameter serviced, filter a kind of characteristic parameter and be greater than or equal to a category feature
The service of upper limit parameter, remaining service, which is put into, recalls pond.
The number serviced in pond is recalled in S12, detection, if recalling the number serviced in pond is 0, the additional service filtered
Again investment recalls pond for the middle the smallest service of one kind characteristic parameter.
Specifically, judge whether service price/main demand price is more than or equal to setting after input services id referring to Fig. 3
Upper limit parameter.If the service price of partial service/main demand price is more than or equal to the upper limit parameter of setting, the part is filtered
Id is serviced, and residue service id investment is recalled into pond.If service price/main demand price of service is respectively less than the upper limit ginseng being arranged
Number, then put into all service id and recall pond.Before the service id in pond is recalled in output, it can detect and recall service id in pond
Whether number is equal to 0.If being equal to 0, adds cheapest service id investment in all services and recall pond;If being not equal to 0, recall
Pond is not processed, and exports the service id recalled in pond.
Referring to Fig. 4, step S2 includes following sub-step:
S21, it obtains historical data and assembles them into characteristic data set, learnt on characteristic data set with XGBoost learner
Buying rate prediction device and main demand conversion ratio prediction device are serviced out.
S22, poll recall the service in pond, and splicing service is forecast sample with the feature in step S21.
S23, using forecast sample as entering ginseng, input service buying rate prediction device, obtain different user in different scenes and
To the buying rate of service when pairing unsalable goods up with goods that sell well different services;Using forecast sample as ginseng is entered, main demand conversion ratio prediction device is inputted, is obtained not
With user different scenes with when pairing unsalable goods up with goods that sell well different services to the conversion ratio of main demand.
Referring to Fig. 5, step S21 includes following sub-step:
The historical data that S211, acquisition are estimated, system assemble the use of user id according to the user id in historical data
Family label.Specifically, user tag includes near real-time label and history class label.
S212, according to the time class data assembling in user tag at including practicing the characteristic data set of category feature.Specifically
Ground, practice category feature include year, the moon, week, which day in day, one week, which day in one month, whether festivals or holidays, festivals or holidays which
It etc..
S213, learnt to service buying rate prediction device and the conversion of main demand out with XGBoost learner on characteristic data set
Rate prediction device.
Referring to Fig. 6, step S3 includes following sub-step:
S31, historical data is obtained, generates the displaying number that the quantity of advertisement show position is 0 by mono- step-length of m% and accounts for total displaying
The list for not showing accounting of number is generated the list of the factor alpha in sequence index calculation formula by mono- step-length of n.
S32, according to buying rate of the user in step S23 to service and the conversion ratio to main demand, calculating (does not show and accounts for
It is more single than the main demand conversion in α) binary group to measure increment and service business revenue increment, and export and (do not show accounting, do not show accounting
Down so that the maximum α of class ROI index) mapping table.
In S33, mapping table in step s 32, a α is randomly choosed, generates initial sequence index calculation formula.
The initial sequence index calculation formula of S34, basis calculates the sequence index value of each advertising display request on line,
And sequence index series is formed according to sequence index value.
S35, the quantile of sequence index series is obtained by mono- step-length of q, corresponding service business revenue is calculated according to quantile
Increment, and export (quantile services business revenue increment) mapping table.Specifically, divide when the sequence index value requested on line is less than
When digit numerical value, without the displaying of advertisement, under the operation, the quantile for the index series that sorts is not show the approximation of accounting
Estimation, i.e. (quantile services business revenue increment) mapping table are approximately (not showing accounting, service business revenue increment) mapping table.
S36, the service business revenue increment according to setting obtain corresponding quantile in mapping table in step s 35, and will
The quantile obtains corresponding α as corresponding do not show in the mapping table of accounting in step s 32.
S37, according to the α in the α in step S36 more new sort index calculation formula, according to updated sequence index meter
Formula is calculated, (not showing accounting, service business revenue increment) mapping table in step S35 is updated.
The implementation principle of above-described embodiment are as follows:
By not showing that accounting captures the dynamic of the quantity of advertisement show position, single amount increment is converted by main demand and is directly fed back
User experience converts the direct feedback user experience of single amount increment/service business revenue increment class ROI index and platform by main demand
Relationship between business revenue.By the technical solution of the application, it can update in real time and (not show accounting, service business revenue increment) and reflect
Firing table, and can be according to desired service business revenue increment, more new sort index calculation formula, generation make class ROI index in linkage
The ordering mechanism being optimal.The technical solution of the application can not only adapt to advertisement show position quantity be 0 scene, also plus
The strong control to relationship between user experience and service business revenue.
Embodiment three
Referring to Fig. 7, for a kind of order ads mechanism generation system disclosed by the invention comprising service limitation module 10 is estimated
Module 20 and ordering mechanism generation module 30.
Referring to Fig. 7, service limitation module 10, which is used to configure, allows to show that a kind of characteristic parameter section serviced, filtering are a kind of
It service of the characteristic parameter outside a kind of characteristic parameter section and remaining service is put into recalls pond.Specifically, in the present embodiment
In, a kind of characteristic parameter section is the price range of service, and the service in price range, which can be placed into, recalls pond, for guaranteeing
User experience.
Although in the application using price as a kind of characteristic parameter for the application scenarios of the application are introduced, this
Field technical staff is appreciated that the technical solution of the application applies also for a variety of different parameter scenes, as ratio is launched in advertisement
Example, advertisement pushing time etc., the application is not especially limited this.
Referring to Fig. 7, module 20 is estimated for constructing forecast sample according to historical data information, and passes through categorised decision algorithm
And forecast sample obtains buying rate of the user to service and the conversion ratio to main demand.In the present embodiment, categorised decision is calculated
Method is using XGBoost learner.
Referring to Fig. 7, the quantity that ordering mechanism generation module 30 is used to generate advertisement show position according to historical data information is 0
Displaying number account for the list for not showing accounting of total displaying number.Ordering mechanism generation module 30 is also used to basis and does not show
Accounting and user to the buying rate of service and to the conversion ratio of main demand, calculate corresponding main not showing in the list of accounting
The single amount increment of demand conversion and service business revenue increment, and generate the main single amount increment/service business revenue increment class ROI of demand conversion and refer to
Mark reaches the sequence index calculation formula of setting value.
Wherein, the single amount increment of main demand conversion=(when showing that the main demand conversion ratio-when certain service does not show task service
Main demand conversion ratio) × 1;Service business revenue increment=(show when the service buying rate-when certain service does not show task service
Service buying rate) × certain service price;Sequence index calculation formula is the single amount increment of main demand conversion of (1- α) × standardized
The service business revenue increment for+α × standardized, α is a real number between 0 and 1.
The implementation principle of above-described embodiment are as follows:
By not showing that accounting can capture the dynamic of the quantity of advertisement show position, converting single amount increment by main demand can be directly
Feedback user experience, converting single amount increment/service business revenue increment class ROI index by main demand can directly feedback user experience
With the relationship between platform business revenue.By the technical solution of the application, can be adjusted in linkage according to desired service business revenue increment
It is whole to adapt to the scene that the quantity of advertisement show position is 0 to the ordering mechanism for being optimal class ROI index, also plus
The strong control to relationship between user experience and service business revenue.
Example IV
Referring to Fig. 7, based on embodiment one, the difference between this embodiment and the first embodiment lies in, service limitation module 10 includes
Ceiling restriction unit 11 and lower limit limiting unit 12.Ceiling restriction unit 11, which is used to configure, to be allowed to show the category feature serviced
Limit parameter filters service of a kind of characteristic parameter more than or equal to a category feature upper limit parameter and is put into residue service and recalls
Pond.Lower limit limiting unit 12 is used for for detecting the number recalled and serviced in pond when recalling the number serviced in pond is 0,
Investment recalls pond again for a kind of the smallest service of characteristic parameter in the additional service filtered.
Specifically, ceiling restriction unit 11 is for judging whether service price/main demand price is more than or equal to the one of setting
Category feature upper limit parameter.If the service price of partial service/main demand price is more than or equal to the upper limit parameter of setting, filtering should
Partial service id, and residue service id investment is recalled into pond.If service price/main demand price of service is respectively less than setting
All service id are then put into and recall pond by upper limit parameter.Lower limit limiting unit 12 recalls id number of service in pond for detecting
Whether it is equal to 0, if being equal to 0, adds cheapest service id investment in all services and recall pond;If being not equal to 0, pond is recalled
It is not processed, and exports the service id recalled in pond.
Referring to Fig. 7, module 20 is estimated including prediction device training unit 21, sample architecture unit 22 and estimates unit 23.
Referring to Fig. 7, prediction device training unit 21 is also used for obtaining historical data and assembling them into characteristic data set
In being learnt to service buying rate prediction device and main demand conversion ratio prediction device out with XGBoost learner on characteristic data set.
Specifically, prediction device training unit 21 is also used to user's mark according to the user id assembling user id in historical data
Label.User tag includes near real-time label, history class label, and the feature that characteristic is concentrated includes practicing category feature, practices class
Feature include according to the time class data assembling in user tag at year, the moon, week, which day in day, one week, in one month which
It, whether the features such as festivals or holidays, which day of festivals or holidays.
Referring to Fig. 7, sample architecture unit 22 recalls the service in pond for poll and splices service and characteristic concentration
Feature be forecast sample.
Referring to Fig. 7, unit 23 is estimated for obtaining difference using forecast sample as ginseng input service buying rate prediction device is entered
User different scenes with when pairing unsalable goods up with goods that sell well different services to the buying rate of service.Unit 23 is estimated to be also used to using forecast sample as entering
Ginseng inputs main demand conversion ratio prediction device, obtains different user in different scenes and conversion when pairing unsalable goods up with goods that sell well different services to main demand
Rate.
Referring to Fig. 7, ordering mechanism generation module 30 includes not showing accounting and factor alpha mapping table generation module 31, not opening up
Show accounting and service business revenue Rise Map table generation module 32 and sequence calculation formula update module 33.
Referring to Fig. 7, after not showing that accounting and factor alpha mapping table generation module 31 are used to obtain historical data, by m% mono-
Step-length generates the displaying number that the quantity of advertisement show position is 0 and accounts for the list for not showing accounting of total displaying number, and presses n mono-
A step-length generates the list of the factor alpha in sequence index calculation formula.Accounting and factor alpha mapping table generation module 31 are not shown also
For, to the buying rate of service and to the conversion ratio of main demand, calculating the main need in (not showing accounting, α) binary group according to user
It asks and converts list amount increment and service business revenue increment, and acquisition does not show that accounting is not shown with this and makes the conversion of winner's demand single under accounting
Amount increment/service business revenue increment reaches the mapping table one of maximum α, i.e., (does not show accounting, do not show and make class ROI under accounting
The maximum α of index) mapping table.
Referring to Fig. 7, do not show accounting and service business revenue Rise Map table generation module 32 for (not showing accounting, no
Showing and make the maximum α of class ROI index under accounting) one α of random selection generates initial sequence index calculation formula in mapping table,
And according to the sequence index value of each advertising display request on initial sequence index calculation formula calculating line and according to sequence
Index value forms sequence index series.Do not show that accounting and service business revenue Rise Map table generation module 32 are also used to by q mono-
A step-length obtains the quantile of sequence index series, and calculates corresponding service business revenue increment and output quartile according to quantile
Several mapping tables two with the service business revenue increment under the quantile, i.e. (quantile services business revenue increment) mapping table.
Specifically, when the sequence index value requested on line is less than quantile numerical value, without the displaying of advertisement, the behaviour
Under work, the quantile for the index series that sorts is not show the approximate evaluation of accounting, i.e. (quantile services business revenue increment) mapping
Table is approximately (not showing accounting, service business revenue increment) mapping table.
Referring to Fig. 7, sequence calculation formula update module 33 is used for the service business revenue increment according to setting in (quantile, clothes
Business business revenue increment) corresponding quantile is obtained in mapping table, and do not show that accounting (is not being shown using the quantile as corresponding
Accounting does not show and makes the maximum α of class ROI index under accounting) corresponding α is obtained in mapping table.The calculation formula that sorts updates mould
Block 33 is also used to according to the α in corresponding α more new sort index calculation formula, and according to updated sequence index calculation formula
(accounting is not shown update, service business revenue increment) mapping table.
The implementation principle of above-described embodiment are as follows:
By not showing that accounting captures the dynamic of the quantity of advertisement show position, single amount increment is converted by main demand and is directly fed back
User experience converts the direct feedback user experience of single amount increment/service business revenue increment class ROI index and platform by main demand
Relationship between business revenue.By the technical solution of the application, it can update in real time and (not show accounting, service business revenue increment) and reflect
Firing table, and can be according to desired service business revenue increment, more new sort index calculation formula in linkage, so that generating makes class ROI
The ordering mechanism that index is optimal.The scene that the quantity of advertisement show position is 0 can not only be adapted to, is also strengthened to user's body
Test and service the control of relationship between business revenue.
The embodiment of present embodiment is presently preferred embodiments of the present invention, not limits protection of the invention according to this
Range, therefore: the equivalence changes that all structures under this invention, shape, principle are done, should all be covered by protection scope of the present invention it
It is interior.
Claims (10)
1. a kind of order ads mechanism generation method, which comprises the following steps:
Step S1, configuration allows to show a kind of characteristic parameter section serviced, filters a kind of characteristic parameter in a category feature
Service outside parameter section, remaining service, which is put into, recalls pond;
Step S2, forecast sample is constructed according to historical data information, by categorised decision algorithm and the forecast sample, obtained
Buying rate of the user to service and the conversion ratio to main demand;
Step S3, the displaying number that the quantity of advertisement show position is 0 is generated according to historical data information and accounts for total displaying number
The list for not showing accounting does not show purchase of the user to service in accounting and step S2 according in the list for not showing accounting
Rate and the conversion ratio to main demand are bought, calculates the single amount increment of corresponding main demand conversion and service business revenue increment, and generate main need
The single amount increment/service business revenue increment of conversion is asked to reach the sequence index calculation formula of setting value.
2. order ads mechanism generation method according to claim 1, which is characterized in that the step S1 includes following son
Step:
S11, configuration allow to show the category feature upper limit parameter serviced, filter a kind of characteristic parameter and be greater than or equal to a category feature
The service of upper limit parameter, remaining service, which is put into, recalls pond;
The number serviced in pond is recalled in S12, detection, if recalling the number serviced in pond is 0, one in the additional service filtered
Again investment recalls pond to the smallest service of category feature parameter.
3. order ads mechanism generation method according to claim 2, which is characterized in that the step S2 includes following son
Step:
S21, it obtains historical data and assembles them into characteristic data set, categorised decision algorithm is used on the characteristic data set
Practise out service buying rate prediction device and main demand conversion ratio prediction device;
The service in pond is recalled described in S22, poll, splicing the service with feature described in step S21 is forecast sample;
S23, using the forecast sample as entering ginseng, input service buying rate prediction device, obtain different user in different scenes and
To the buying rate of service when pairing unsalable goods up with goods that sell well different services;Using the forecast sample as ginseng is entered, main demand conversion ratio prediction device is inputted, is obtained
To different user different scenes with when pairing unsalable goods up with goods that sell well different services to the conversion ratio of main demand.
4. order ads mechanism generation method according to claim 3, which is characterized in that the step S21 includes following
Sub-step:
The historical data that S211, acquisition are estimated, system assemble the use according to the user id in the historical data
The user tag of family id;
S212, according to the time class data assembling in the user tag at including practicing the characteristic data set of category feature;
S213, learnt to service buying rate prediction device and main demand conversion ratio out with categorised decision algorithm on the characteristic data set
Prediction device.
5. order ads mechanism generation method according to claim 3 or 4, which is characterized in that the step S3 include with
Lower sub-step:
S31, historical data is obtained, generates the displaying number that the quantity of advertisement show position is 0 by mono- step-length of m% and accounts for total displaying
The list for not showing accounting of number is generated the list of the factor alpha in sequence index calculation formula by mono- step-length of n;
S32, according to buying rate of the user in step S23 to service and the conversion ratio to main demand, calculating does not show accounting and is
Main demand in the binary group of number α converts single amount increment and service business revenue increment, and acquisition does not show that accounting is not shown with this and accounts for
The single amount increment/service business revenue increment of winner's demand conversion is set to reach the mapping table of maximum α than under;
In S33, mapping table in step s 32, a α is randomly choosed, generates initial sequence index calculation formula;
S34, according to the initial sequence index calculation formula, calculate the sequence index value of each advertising display request on line simultaneously
Sequence index series is formed according to the sequence index value;
S35, the quantile of the sequence index series is obtained by mono- step-length of q, corresponding service is calculated according to the quantile
Business revenue increment, and export the mapping table of the service business revenue increment under quantile and the quantile;
S36, the service business revenue increment according to setting, obtain corresponding quantile in mapping table in step s 35, and by this point
Digit obtains corresponding α as corresponding do not show in the mapping table of accounting in step s 32;
S37, the α in the sequence index calculation formula is updated according to the α in step S36, according to updated sequence index meter
Formula is calculated, the mapping table in step S35 is updated.
6. a kind of order ads mechanism generates system characterized by comprising
Service limitation module (10), being used to configure allows to show a kind of characteristic parameter section serviced, filtering one category feature ginseng
It counts the service outside a kind of characteristic parameter section and is put into remaining service and recall pond;
Module (20) are estimated, are used to construct forecast sample according to historical data information, and pass through categorised decision algorithm and institute
It states forecast sample and obtains buying rate of the user to service and the conversion ratio to main demand;
Ordering mechanism generation module (30) is used to generate the displaying that the quantity of advertisement show position is 0 according to historical data information
Number accounts for the list for not showing accounting of total displaying number;
Wherein, the ordering mechanism generation module (30) be also used to according to do not show in the list of accounting do not show accounting and
Buying rate of the user to service and the conversion ratio to main demand, calculate the single amount increment of corresponding main demand conversion and service business revenue increases
Amount, and generate the sequence index calculation formula that the main single amount increment/service business revenue increment of demand conversion reaches setting value.
7. order ads mechanism according to claim 6 generates system, which is characterized in that the service limits module (10)
Include:
Ceiling restriction unit (11), being used to configure allows to show the category feature upper limit parameter serviced, filtering one category feature ginseng
Number be greater than or equal to a category feature upper limit parameter services and by remaining service be put into described in recall pond;
Lower limit limiting unit (12), be used to detect it is described recall the number serviced in pond, and for being taken when described recall in pond
When the number of business is 0, the smallest service of a kind of characteristic parameter recalls pond described in investment again in the additional service filtered.
8. order ads mechanism according to claim 7 generates system, which is characterized in that described to estimate module (20) packet
It includes:
Prediction device training unit (21) is used to obtain historical data and assembles them into characteristic data set, is also used in institute
It states and is learnt to service buying rate prediction device and main demand conversion ratio prediction device out with categorised decision algorithm on characteristic data set;
Sample architecture unit (22), the service for being used to recall described in poll in pond simultaneously splice the service and characteristic concentration
Feature be forecast sample;
Unit (23) are estimated, are used to obtain different use using the forecast sample as ginseng input service buying rate prediction device is entered
Family different scenes with when pairing unsalable goods up with goods that sell well different services to the buying rate of service;It is also used to input using the forecast sample as ginseng is entered
Main demand conversion ratio prediction device, obtain different user different scenes with when pairing unsalable goods up with goods that sell well different services to the conversion ratio of main demand.
9. order ads mechanism according to claim 8 generates system, which is characterized in that the prediction device training unit
(21) it is also used to assemble the user tag of the user id, the spy that the characteristic is concentrated according to the user id in historical data
Sign include according to the time class data assembling in user tag at practice category feature.
10. order ads mechanism according to claim 8 or claim 9 generates system, which is characterized in that the ordering mechanism generates
Module (30) includes:
Accounting and factor alpha mapping table generation module (31) are not shown, after being used to obtain historical data, are generated by mono- step-length of m%
The displaying number that the quantity of advertisement show position is 0 accounts for the list for not showing accounting of total displaying number, and raw by mono- step-length of n
At the list of the factor alpha in sequence index calculation formula;It is also used to according to user to the buying rate of service and to main demand
Conversion ratio calculates the single amount increment of main demand conversion and the service business revenue increment not shown in the binary group of accounting and factor alpha, and obtains
It obtains and does not show that accounting does not show that so that winner's demand conversion list is measured increment/service business revenue increment under accounting reaches maximum α's with this
Mapping table one;
Accounting and service business revenue Rise Map table generation module (32) are not shown, are used in the mapping table one randomly choose
One α generates initial sequence index calculation formula, calculates each advertisement exhibition on line according to the initial sequence index calculation formula
Show the sequence index value of request and sequence index series is formed according to the sequence index value;It is also used to by q mono-
Step-length obtains the quantile of the sequence index series, calculates corresponding service business revenue increment and output according to the quantile
The mapping table two of service business revenue increment under quantile and the quantile;
It sorts calculation formula update module (33), is used to be obtained in the mapping table two according to the service business revenue increment of setting
Corresponding quantile, and do not show that accounting obtains corresponding α in mapping table one using the quantile as corresponding;It is also used to
It is reflected according to the α in the corresponding α more new sort index calculation formula, and according to the update of updated sequence index calculation formula
Firing table two.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811260738.6A CN109377280B (en) | 2018-10-26 | 2018-10-26 | Advertisement sequencing mechanism generation method and generation system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811260738.6A CN109377280B (en) | 2018-10-26 | 2018-10-26 | Advertisement sequencing mechanism generation method and generation system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109377280A true CN109377280A (en) | 2019-02-22 |
CN109377280B CN109377280B (en) | 2020-07-17 |
Family
ID=65389955
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811260738.6A Active CN109377280B (en) | 2018-10-26 | 2018-10-26 | Advertisement sequencing mechanism generation method and generation system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109377280B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111324800A (en) * | 2020-02-12 | 2020-06-23 | 腾讯科技(深圳)有限公司 | Business item display method and device and computer readable storage medium |
CN112200610A (en) * | 2020-10-10 | 2021-01-08 | 苏州创旅天下信息技术有限公司 | Marketing information delivery method, system and storage medium |
CN112232857A (en) * | 2020-09-25 | 2021-01-15 | 上海淇毓信息科技有限公司 | Automatic advertisement sequencing method and device and electronic equipment |
CN113032445A (en) * | 2021-05-24 | 2021-06-25 | 武汉卓尔数字传媒科技有限公司 | Data conversion sorting method and device and electronic equipment |
CN113516519A (en) * | 2021-07-28 | 2021-10-19 | 北京字节跳动网络技术有限公司 | Model training method, advertisement putting method, device, equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103177371A (en) * | 2011-12-21 | 2013-06-26 | 阿里巴巴集团控股有限公司 | Method and device for exhibiting information |
US20160210671A1 (en) * | 2015-01-15 | 2016-07-21 | Appnexus, Inc. | Modifying bid price for online advertising auction based on user impression frequency |
CN106485529A (en) * | 2015-09-02 | 2017-03-08 | 北京国双科技有限公司 | The sort method of advertisement position and device |
US20180075475A1 (en) * | 2016-09-13 | 2018-03-15 | Adobe Systems Incorporated | Distributing online ads by targeting online ad requests |
CN108335137A (en) * | 2018-01-31 | 2018-07-27 | 北京三快在线科技有限公司 | Sort method and device, electronic equipment, computer-readable medium |
CN108510309A (en) * | 2018-02-27 | 2018-09-07 | 阿里巴巴集团控股有限公司 | The method and device that advertisement is recalled |
-
2018
- 2018-10-26 CN CN201811260738.6A patent/CN109377280B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103177371A (en) * | 2011-12-21 | 2013-06-26 | 阿里巴巴集团控股有限公司 | Method and device for exhibiting information |
US20160210671A1 (en) * | 2015-01-15 | 2016-07-21 | Appnexus, Inc. | Modifying bid price for online advertising auction based on user impression frequency |
CN106485529A (en) * | 2015-09-02 | 2017-03-08 | 北京国双科技有限公司 | The sort method of advertisement position and device |
US20180075475A1 (en) * | 2016-09-13 | 2018-03-15 | Adobe Systems Incorporated | Distributing online ads by targeting online ad requests |
CN108335137A (en) * | 2018-01-31 | 2018-07-27 | 北京三快在线科技有限公司 | Sort method and device, electronic equipment, computer-readable medium |
CN108510309A (en) * | 2018-02-27 | 2018-09-07 | 阿里巴巴集团控股有限公司 | The method and device that advertisement is recalled |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111324800A (en) * | 2020-02-12 | 2020-06-23 | 腾讯科技(深圳)有限公司 | Business item display method and device and computer readable storage medium |
CN112232857A (en) * | 2020-09-25 | 2021-01-15 | 上海淇毓信息科技有限公司 | Automatic advertisement sequencing method and device and electronic equipment |
CN112200610A (en) * | 2020-10-10 | 2021-01-08 | 苏州创旅天下信息技术有限公司 | Marketing information delivery method, system and storage medium |
CN113032445A (en) * | 2021-05-24 | 2021-06-25 | 武汉卓尔数字传媒科技有限公司 | Data conversion sorting method and device and electronic equipment |
CN113032445B (en) * | 2021-05-24 | 2021-08-17 | 武汉卓尔数字传媒科技有限公司 | Data conversion sorting method and device and electronic equipment |
CN113516519A (en) * | 2021-07-28 | 2021-10-19 | 北京字节跳动网络技术有限公司 | Model training method, advertisement putting method, device, equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN109377280B (en) | 2020-07-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109377280A (en) | A kind of order ads mechanism generation method and generate system | |
CN107977859A (en) | Advertisement placement method, device, computing device and storage medium | |
CN107220353B (en) | Automatic satisfaction evaluation method and system for intelligent customer service robot | |
CN106934498A (en) | The recommendation method and system of hotel's house type in OTA websites | |
CN106372959A (en) | Internet-based user access behavior digital marketing system and method | |
JP3788555B2 (en) | Load type analysis method and apparatus, consumption tendency diagnosis method, apparatus and system, and recording medium | |
CN106127504A (en) | The method for pushing of electronic article certificate, generation method, device, user terminal and server | |
CN109509039A (en) | Method for building up and system, the Method of Commodity Recommendation and system of price expectation model | |
CN109165763A (en) | A kind of potential complained appraisal procedure and device of 95598 customer service work order | |
CN101802856A (en) | Measuring a location based advertising campaign | |
CN107657476A (en) | Method and device is recommended in the evaluation method and device in shop, shop | |
CN106528147A (en) | Desktop icon display method, desktop icon display device and terminal | |
CN110363621A (en) | A kind of order information supplying system based on artificial intelligence technology | |
CN108429776A (en) | Method for pushing, device, client, interactive device and the system of network object | |
CN110689401A (en) | Service commodity recommendation method and device | |
JP2023508172A (en) | Resource allocation method, device, facility, storage medium and computer program | |
CN108053323A (en) | Method, apparatus, computer equipment and the storage medium of service plan generation | |
CN104238985A (en) | Method for evaluating LED (Light Emitting Diode) display screens | |
CN112884550A (en) | Commodity recommendation method and device based on customer purchasing ability | |
CN111460301B (en) | Object pushing method and device, electronic equipment and storage medium | |
CN108009842A (en) | The consumer price index based on online data determines system | |
CN107169844A (en) | A kind of Method of Commodity Recommendation and device | |
CN104484745A (en) | City supermarket information publishing system | |
CN115578134A (en) | Household appliance selling system based on big data | |
CN115345662A (en) | Live broadcast e-commerce data processing system based on block chain and big data |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |