CN110363590A - A kind of advertisement recommended method, device, terminal and storage medium - Google Patents
A kind of advertisement recommended method, device, terminal and storage medium Download PDFInfo
- Publication number
- CN110363590A CN110363590A CN201910640133.8A CN201910640133A CN110363590A CN 110363590 A CN110363590 A CN 110363590A CN 201910640133 A CN201910640133 A CN 201910640133A CN 110363590 A CN110363590 A CN 110363590A
- Authority
- CN
- China
- Prior art keywords
- advertisement
- measured
- target user
- probability
- user
- 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
- G06Q30/0244—Optimization
-
- 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
- G06Q30/0251—Targeted advertisements
- G06Q30/0255—Targeted advertisements based on user history
-
- 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
- G06Q30/0277—Online advertisement
Landscapes
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (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)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a kind of advertisement recommended method, device, terminal and storage mediums, this method comprises: obtaining target user and the first object input feature vector of advertisement to be measured and the second target input feature vector of target user and the advertisement to be measured now of emphasis class；First object input feature vector is inputted to the click prediction model constructed in advance, obtains target user to the click probability of each advertisement to be measured；According to target user to the click probability of each advertisement to be measured, the initial sequence of advertisement to be measured is determined；Lower single prediction model that the input of second target input feature vector is constructed in advance determines target user to lower single probability of advertisement to be measured each under emphasis classification；According to target user to lower single probability of advertisement to be measured each under emphasis classification, initial sequence is modified, determines the recommendation sequence of advertisement to be measured.The embodiment of the present invention can reduce the probability exposed in vain on the basis of improving clicking rate, and then promote the overall exposing income of ad system.
Description
Technical field
The present embodiments relate to technical field of data processing more particularly to a kind of advertisement recommended method, device, terminal and
Storage medium.
Background technique
Advertisement intelligent dispensing is a big application field of personalized recommendation system.During runing advertisement dispensing, advertisement
Activity bring clicking rate (Click Through Rate, CTR) and rate of return on investment (Return On Investment,
ROI) it is important judgment criteria.How the recommender system of advertisement intelligent release platform is improved, thus to greatest extent by light exposure
Ground validation seems most important.
The recommendation process of conventional ads intelligence release platform often faces the problem of two aspects: personalization is launched by numerous
The limitation of the factor such as factors such as user preference, advertising campaign and user's amount；The commodity classification of Advertisement association is more, and different classifications
Commodity generate income it is unbalanced.In view of the above-mentioned problems, at present advertisement recommendation be usually according to user's history consumption or
The record such as click is summarized user preference and is launched to match advertisement classification；The exposure higher commercial paper of income is counted based on historical data
Mesh gives the more chances for exposure of respective class purpose advertisement, even packet station.But the mode that above-mentioned advertisement is recommended is lacked there are following
It falls into: considering feature in terms of user's folk prescription region feature or handbill, only realize the personalization for user or advertisement, accurately
Property is relatively low；It concentrates on and launches the high advertisement classification of income, operation way is too extreme and relies on human configuration, causes no demand
User's corresponding exposure waste.
Summary of the invention
The embodiment of the present invention provides a kind of advertisement recommended method, device, terminal and storage medium, to optimize advertisement recommendation side
Case improves the accuracy of recommendation, reduces exposure waste.
In a first aspect, the embodiment of the invention provides a kind of advertisement recommended methods, comprising:
The first object input feature vector and the target user and emphasis class of acquisition target user and advertisement to be measured are now
Second target input feature vector of the advertisement to be measured；
The first object input feature vector is inputted to the click prediction model constructed in advance, obtains the target user to every
The click probability of a advertisement to be measured；
According to the target user to the click probability of each advertisement to be measured, the initial row of the advertisement to be measured is determined
Sequence；
Lower single prediction model that second target input feature vector input is constructed in advance, determines target user's counterweight
Lower single probability of the point class advertisement to be measured each now；
According to the target user to lower single probability of the advertisement to be measured each under emphasis classification, to the initial sequence
It is modified, determines the recommendation sequence of the advertisement to be measured.
Second aspect, the embodiment of the invention also provides a kind of advertisement recommendation apparatus, which includes:
Feature obtains module, for obtaining the first object input feature vector and the mesh of target user Yu advertisement to be measured
Mark the second target input feature vector of user and the advertisement to be measured described now of emphasis class；
Prediction module is clicked, for the first object input feature vector to be inputted the click prediction model constructed in advance, is obtained
To the target user to the click probability of each advertisement to be measured；
Initial sorting module determines institute for the click probability according to the target user to each advertisement to be measured
State the initial sequence of advertisement to be measured；
Lower list prediction module, lower single prediction model for constructing the second target input feature vector input in advance, really
Lower single probability of the fixed target user to the advertisement to be measured each under emphasis classification；
Sort correction module, for general to placing an order for the advertisement to be measured each under emphasis classification according to the target user
Rate is modified the initial sequence, determines the recommendation sequence of the advertisement to be measured.
Further, the initial sorting module is specifically used for:
According to the target user to the click probability of each advertisement to be measured, according to sequence from big to small to described to be measured
Advertisement is ranked up, and obtains the initial sequence.
Further, the sequence correction module is specifically used for:
For belonging to the emphasis classification, and lower single probability of the target user is less than the described to be measured of probability threshold value
Its position in the initial sequence is suppressed according to setting rule, and corrects the initial sequence by advertisement.
Further, the device further include:
Model construction module, for constructing before obtaining the first object input feature vector of target user and advertisement to be measured
The click prediction model and lower single prediction model.
Further, the model construction module includes clicking prediction model unit, the click prediction model unit tool
Body is used for:
Obtain the first advertisement log data in the first default historical time；
First sample data are generated according to first advertisement log data, the first sample data include first sample
Input feature vector and corresponding click label, the first sample input feature vector include user characteristics, characteristic of advertisement and user
With the cross feature of advertisement；
Higher-order logic regression model is trained according to the first sample data, obtains clicking prediction model.
Further, the model construction module includes lower single prediction model unit, lower single prediction model unit tool
Body is used for:
It obtains in the second default historical time, the second advertisement log data of the emphasis class now；
The second sample data is generated according to second advertisement log data, second sample data includes the second sample
Input feature vector and corresponding lower single label, the second sample input feature vector include user characteristics, user's air control feature and
The cross feature of user and the emphasis class advertisement now；
Xgboost model is trained according to second sample data, obtains lower single prediction model.
Further, the first object input feature vector includes target user's feature, characteristic of advertisement to be measured and the mesh
The cross feature of user and the advertisement to be measured are marked, the second target input feature vector includes target user's feature, target user
The cross feature of air control feature and the target user and the emphasis class advertisement to be measured described now, the target user are
Based on the user that hot spot rule determines, the advertisement to be measured is the effective advertisement set in the object time.
The third aspect, the embodiment of the invention also provides a kind of terminal, the terminal includes:
One or more processors；
Storage device, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes advertisement recommended method as described above.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program, the program realize advertisement recommended method as described above when being executed by processor.
The first object input feature vector and target user that the embodiment of the present invention passes through acquisition target user and advertisement to be measured
With the second target input feature vector of emphasis class advertisement to be measured now, the click that the input of first object input feature vector is constructed in advance is pre-
Model is surveyed, obtains target user to the click probability of each advertisement to be measured, and determine the initial sequence of advertisement to be measured；By the second mesh
Lower single prediction model that mark input feature vector input constructs in advance, determines target user under advertisement to be measured each under emphasis classification
Single probability, and lower single probability of advertisement to be measured each under emphasis classification is modified initial sequence, really according to target user
The recommendation sequence of fixed advertisement to be measured.Technical solution provided in an embodiment of the present invention, it is general by clicking the click that prediction model determines
Rate determines the initial sequence of advertisement to be measured, and is modified by lower single probability that lower single prediction model determines to initial sequence,
The recommendation sequence for finally obtaining advertisement to be measured, is considered by the multi-angle of forward and reverse, recommends most probable to click simultaneously to user
And the advertisement that most probable places an order, the probability exposed in vain can be reduced on the basis of improving clicking rate, and then promote advertisement system
The overall exposing income of system.
Detailed description of the invention
Fig. 1 is the flow chart of the advertisement recommended method in the embodiment of the present invention one；
Fig. 2 is the schematic diagram of the advertisement recommended method in the embodiment of the present invention one；
Fig. 3 is the schematic diagram of the cross feature in the embodiment of the present invention one；
Fig. 4 is the flow chart of the advertisement recommended method in the embodiment of the present invention two；
Fig. 5 is the flow chart of the advertisement recommended method in the embodiment of the present invention three；
Fig. 6 is the structural schematic diagram of the advertisement recommendation apparatus in the embodiment of the present invention four；
Fig. 7 is the structural schematic diagram of the terminal in the embodiment of the present invention five.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is the flow chart of the advertisement recommended method in the embodiment of the present invention one, and the present embodiment is applicable to realize advertisement
The case where recommendation, this method can be executed by advertisement recommendation apparatus, which can be real by the way of software and/or hardware
Existing, which is configured in terminal, such as the terminal can be smart phone, computer and tablet computer etc..
Fig. 2 is the schematic diagram of the advertisement recommended method in the embodiment of the present invention one, and referring to fig. 2, ad system may include
Advertisement backstage 11 and advertisement front end 12, wherein advertisement front end 12 can be the terminal of user, and dress is recommended in the advertisement in the present embodiment
Setting can be only fitted in advertisement front end 12, and advertisement backstage 11 can be the backstage of progress advertisement dispensing, include in advertisement backstage 11
Advertising resource pond can store whole advertising resources and advertisement phase that advertisement front end 12 needs to show user in advertising resource pond
Close information, such as show the time etc., advertising resource pond can be updated according to the actual situation, such as timing updates.
Specifically, advertisement front end 12 can carry out online log splicing and user data is reported to server；Work as advertisement
When advertisement recommendation apparatus in front end 12 receives advertisement recommendation request, user data in available server will pass through
It is trained in the model that user data input after Feature Engineering processing constructs in advance, obtains trained model, wherein
The model can be set according to the actual situation, in figure by taking LR model (Logic Regression Models) as an example；Pass through trained mould
The sequence of the available advertisement to be presented of type exposes the sequencing table of clicking rate, advertisement can be generated according to the sequence and recommend column
Table；The advertisement recommendation list can be sent to advertisement front end 12 and carry out online advertisement displaying by advertisement recommendation apparatus.Wherein, feature
May include in engineering user basic information, amount and refund information, user's commodity interactive information, advertisement essential information, air control,
The features such as purchase intention and ad content feature.It is understood that after generating advertisement recommendation list, it can also be according to correlation
Rule is then forwarded to advertisement front end 12 after being filtered.Further, the advertisement recommendation apparatus in figure in advertisement front end 12 is also
A control group can be set, by cold start-up so that the advertisement recommended is random, and then to above-mentioned advertisement recommendation list
Accuracy is verified.
As shown in Figure 1, this method can specifically include:
S110, the first object input feature vector for obtaining target user and advertisement to be measured and target user and emphasis classification
Under advertisement to be measured the second target input feature vector.
Wherein, target user can be the user determining based on hot spot rule, which can be according to the actual situation
It is set, such as target user can be the user that clicking rate is greater than clicking rate threshold value in history one month.Advertisement to be measured can
Think effective advertisement in the setting object time, wherein the setting object time can be set according to the actual situation, such as sets
Setting the goal the time can be for one day, one week or one month etc., and effective advertisement refers to that the setting object time interior interacts with user
Advertisement, such as the advertisement etc. that the advertisement clicked of user or user shared.The volume cost of target user and advertisement to be measured is real
It applies in example and is not construed as limiting, can be set according to the actual situation.
First object input feature vector and the second target input feature vector are the related data and advertisement to be measured according to target user
Related data, the feature handled by Feature Engineering.Emphasis classification can pass through history for Current ad release platform
Data analysis, having of obtaining concentrate the trend launched or have the advertisement classification of high coverage rate, such as some platform,
Emphasis classification can be 3C classification, including computer (Computer), communication (Communication) and consumer electronics product
(Consumer Electronics) etc..Specifically, first object input feature vector may include target user's feature, advertisement to be measured
The cross feature of feature and target user and advertisement to be measured, the second target input feature vector may include target user's feature, mesh
Mark user's air control feature and the cross feature of target user and the advertisement to be measured now of emphasis class.
Target user's feature may include essential information, air control and the amount information of user to be measured, history preference (such as classification
With brand etc.) information etc., characteristic of advertisement to be measured may include the corresponding classification of Advertisement association commodity to be measured and brand etc., and target is used
The cross feature of family and advertisement to be measured may include that the matched classification of advertisement to be measured and brand are clicked or bought to user's history to be measured
Deng.Specific feature can be set according to the actual situation.
Illustratively, Fig. 3 is the schematic diagram of the cross feature in the embodiment of the present invention one, and the cross feature in figure is user
With the cross feature of the advertisement in 3C classification, a nearest month 3C exposure frequency in figure is specifically included, a nearest month 3C is clicked
Number, nearest month 3C exposure clicking rate, nearest three months 3C buy number, a nearest month 3C extra bus number and nearest three
A month 3C purchase second trial number of pass times etc..
S120, first object input feature vector is inputted to the click prediction model constructed in advance, obtains target user to each
The click probability of advertisement to be measured.
Wherein, clicking probability is the probability that user clicks advertisement, can be indicated by percentage, such as user clicks extensively
The probability of announcement can be 50%.After getting first object input feature vector, which can be inputted pre-
The click prediction model first constructed, click probability of the output target user to each advertisement to be measured.Wherein, the click prediction model
It can be to be trained to obtain based on deep learning model, be not construed as limiting in specific deep learning model the present embodiment, such as
Clicking prediction model can be obtained based on higher-order logic recurrence (Logistic Regression, LR) model training.
S130, according to target user to the click probability of each advertisement to be measured, determine the initial sequence of advertisement to be measured.
Specifically, after getting target user to the click probability of each advertisement to be measured, can according to the click probability,
Advertisement to be measured is ranked up according to sequence from big to small, is initially sorted.In the initial sequence, maximum probability is clicked
Order ads to be measured click the smallest order ads to be measured of probability near rear near preceding.
S140, the lower single prediction model for constructing the input of the second target input feature vector in advance, determine target user to emphasis
Lower single probability of class advertisement to be measured each now.
After getting the second target input feature vector in S110, the second target input feature vector input can be placed an order pre-
It surveys in model, lower single probability of the output target user to advertisement to be measured each under emphasis classification.Wherein, which can
Think and be trained to obtain based on deep learning model, is not construed as limiting in specific deep learning model the present embodiment, such as under
Single prediction model can be obtained based on Xgboost model training.
S150, according to target user to lower single probability of advertisement to be measured each under emphasis classification, initial sequence is repaired
Just, the recommendation sequence of advertisement to be measured is determined.
Specifically, on the basis of determining the initial sequence of advertisement to be measured in S130, it can be according to target user to emphasis
Lower single probability of class advertisement to be measured each now, is adjusted the initial sequence, and the initial sequence after adjustment is determined as
Final recommendation sequence.
Optionally, lower single probability of advertisement to be measured each under emphasis classification carries out initial sequence according to target user
Amendment may include: for belonging to emphasis classification, and lower single probability of target user is less than the advertisement to be measured of probability threshold value,
Its position in initial sequence is suppressed according to setting rule, and corrects initial sequence.Wherein probability threshold value and setting
Rule can be set according to the actual situation, such as probability threshold value can be 0.5, and setting rule can be current location * 10,
It is considered that lower single probability is less than position of the advertisement to be measured of probability threshold value without recommendation meaning, in initial sequence in the present embodiment
It can be to suppress to the end.Illustratively, if the position of advertisement m to be measured is at the 2nd in initial sequence, and the advertisement m to be measured
Belong to emphasis classification and target user is less than probability threshold value to its lower single probability, then passes through the position of the advertisement m to be measured
The mode of current location * 10 is suppressed, obtain advertisement m to be measured position adjusted after initial sequence.
Optionally, the purchase probability of the advertisement to be predicted according to target user for each emphasis class now, to initial row
Sequence is modified, and can also include: the currently available volume according to target user for each advertisement to be measured for belonging to emphasis classification
Degree, the comparison result with history concluded price threshold value are modified initial sequence.Specifically, based on flowing water number single under history
According to, under available target user in single emphasis classification Advertisement association commodity to be measured knock-down price sequence, history can be struck a bargain
Price thresholds are set as the 90% of concluded price, compare target for the currently available of each advertisement to be measured for belonging to emphasis classification
Amount and history concluded price threshold value are less than currently available amount the advertisement to be measured of history concluded price threshold value, can incite somebody to action
It is suppressed according to setting rule its position in initial sequence.Wherein setting rule can be set according to the actual situation
It is fixed.
By the amendment to initially sorting, for target user, launched by the recommendation sequence of advertisement to be measured wide
Announcement is precisely to be recommended based on user individual, and be more likely to generate the advertisement of order income.
The present embodiment by obtain target user and advertisement to be measured first object input feature vector and target user with again
The click that the input of first object input feature vector constructs in advance is predicted mould by the second target input feature vector of point class advertisement to be measured now
Type obtains target user to the click probability of each advertisement to be measured, and determines the initial sequence of advertisement to be measured；Second target is defeated
Enter lower single prediction model that feature input constructs in advance, determines that target user is general to placing an order for advertisement to be measured each under emphasis classification
Rate, and lower single probability of advertisement to be measured each under emphasis classification being modified initial sequence according to target user, determine to
Survey the recommendation sequence of advertisement.Technical solution provided in this embodiment is waited for by clicking the click determine the probability that prediction model determines
The initial sequence of advertisement is surveyed, and initial sequence is modified by lower single probability that lower single prediction model determines, is finally obtained
The recommendation of advertisement to be measured is sorted, and is considered by the multi-angle of forward and reverse, recommends most probable click and most probable to user
The advertisement to place an order can reduce the probability exposed in vain on the basis of improving click-through-rate, and then promote ad system
Overall exposing income.
Embodiment two
Fig. 4 is the flow chart of the advertisement recommended method in the embodiment of the present invention two.Base of the present embodiment in above-described embodiment
On plinth, above-mentioned advertisement recommended method has been advanced optimized.Correspondingly, as shown in figure 4, the method for the present embodiment specifically includes:
S210, building click prediction model and lower single prediction model.
Specifically, prediction model is clicked in building, it may include: the first ad log obtained in the first default historical time
Data；Generate first sample data according to the first advertisement log data, first sample data include first sample input feature vector with
And corresponding click label, first sample input feature vector includes user characteristics, characteristic of advertisement and user and advertisement intersects spy
Sign；Higher-order logic regression model is trained according to first sample data, obtains clicking prediction model.
Wherein, the first default historical time can be set according to the actual situation, preset and gone through with first in the present embodiment
The history time is to be illustrated for one week.First advertisement log data be data relevant to advertisement, may include user data,
The data such as the interaction data of ad data and user and advertisement.First sample data are to click the sample data of prediction model, the
It include the training sample data for training pattern and the test sample data for test model in one sample data.First sample
Include multiple samples in notebook data, is not construed as limiting in quantity the present embodiment of specific sample.Each sample may include the first sample
This input feature vector and the corresponding click label of the first sample input feature vector, the click label is for marking user default first
Corresponding advertisement whether was clicked in historical time, if clicking, is clicked label labeled as 1, if not clicking on, is clicked label
Labeled as -1.For example, if user A is 10 to number of clicks of the advertisement A in one week, first in the corresponding sample of user A
Sample input feature vector includes the cross feature of the feature of user A, the feature of advertisement A and user A and advertisement A, and clicking label is 1.
The lower single prediction model of building, may include: to obtain in the second default historical time, the second advertisement of emphasis class now
Daily record data；The second sample data is generated according to the second advertisement log data, the second sample data includes that the input of the second sample is special
Sign and corresponding lower single label, the second sample input feature vector includes user characteristics, user's air control feature and user and emphasis
The cross feature of class advertisement now；Xgboost model is trained according to the second sample data, obtains lower single prediction model.
Wherein, user's air control feature can may include to indicate the feature to user's finance level risk control, user's air control feature
This month capital to be gone back, user's capital to be gone back, gone back expense, this month should go back capital, total credit line, can with spending limit, used
Spending limit, remaining interest-free amount and credit number of days etc..
Wherein, the second default historical time can be set according to the actual situation, preset and gone through with second in the present embodiment
The history time is to be illustrated for one month.Second advertisement log data is emphasis class data relevant to advertisement now, can be with
The data such as the interaction data including user data, ad data and user and advertisement.Second sample data is to click prediction model
Sample data, include for the training sample data of training pattern and for the test specimens of test model in the second sample data
Notebook data.Include multiple samples in second sample data, is not construed as limiting in quantity the present embodiment of specific sample.Each sample can
To include that the second sample input feature vector and the corresponding lower single label of the first sample input feature vector, lower single label are used for marking
Whether family placed an order in the second default historical time corresponds to the commodity of advertisement, if placing an order, lower list label is labeled as 1, if
It does not click on, then lower single label is labeled as -1.For example, if user B is to the corresponding commodity of advertisement B under emphasis classification in one month
Purchase number be 5, the second sample input feature vector in the corresponding sample of user B includes the wind of the feature of user B, user B
The cross feature of feature and user B and advertisement B is controlled, lower list label is 1.
Further, when being trained according to first sample data to higher-order logic regression model and according to the second sample
When data are trained Xgboost model, it can pass through AUC (Area Under the Curve) and KS
The indexs such as (Kolmogorov Smirnov) evaluation index measure the availability of model, if available, obtain clicking prediction model
With lower single prediction model.Wherein, AUC refers under ROC curve (Receiver Operating Characteristic curve)
Area, be one of deep learning model evaluation index.
S220, the first object input feature vector for obtaining target user and advertisement to be measured and target user and emphasis classification
Under advertisement to be measured the second target input feature vector.
Wherein, first object input feature vector include target user's feature, characteristic of advertisement to be measured and target user with it is to be measured
The cross feature of advertisement, the second target input feature vector include target user's feature, target user's air control feature, characteristic of advertisement to be measured
And the cross feature of target user and the advertisement to be measured now of emphasis class, target user are the user determined based on hot spot rule,
Advertisement to be measured is the effective advertisement set in the object time.
S230, first object input feature vector is inputted to the click prediction model constructed in advance, obtains target user to each
The click probability of advertisement to be measured.
S240, according to target user to the click probability of each advertisement to be measured, determine the initial sequence of advertisement to be measured.
Specifically, after getting target user to the click probability of each advertisement to be measured, can according to the click probability,
Advertisement to be measured is ranked up according to sequence from big to small, is initially sorted.
S250, the lower single prediction model for constructing the input of the second target input feature vector in advance, determine target user to emphasis
Lower single probability of class advertisement to be measured each now.
S260, according to target user to lower single probability of advertisement to be measured each under emphasis classification, initial sequence is repaired
Just, the recommendation sequence of advertisement to be measured is determined.
Specifically, on the basis of determining the initial sequence of advertisement to be measured in S240, it can be according to target user to emphasis
Lower single probability of class advertisement to be measured each now, is adjusted the initial sequence, and the initial sequence after adjustment is determined as
Final recommendation sequence.
Optionally, lower single probability of advertisement to be measured each under emphasis classification carries out initial sequence according to target user
Amendment may include: for belonging to emphasis classification, and lower single probability of target user is less than the advertisement to be measured of probability threshold value,
Its position in initial sequence is suppressed according to setting rule, and corrects initial sequence.
S270, being sorted according to the recommendation of advertisement to be measured carries out advertisement recommendation.
Specifically, after determining the final recommendation sequence of advertisement to be measured, can be arranged according to the recommendation for target user
Sequence recommends the advertisement to be measured of setting quantity to target user.Wherein, setting quantity can be set according to the actual situation, example
Such as setting quantity can be 5 or 10.
Prediction model and lower single prediction model are clicked in the present embodiment building, by obtain target user and advertisement to be measured the
One target input feature vector and the second target input feature vector of target user and the advertisement to be measured now of emphasis class, by first object
Input feature vector inputs the click prediction model constructed in advance, obtains target user to the click probability of each advertisement to be measured, and really
The initial sequence of fixed advertisement to be measured；Lower single prediction model that the input of second target input feature vector is constructed in advance, determines that target is used
Family is to lower single probability of advertisement to be measured each under emphasis classification, and according to target user to advertisement to be measured each under emphasis classification
Lower list probability, is modified initial sequence, determines the recommendation sequence of advertisement to be measured.Technical solution provided in this embodiment is led to
The initial sequence for clicking the click determine the probability advertisement to be measured of prediction model determination is crossed, and by under lower single prediction model determination
Single probability is modified initial sequence, finally obtains the recommendation sequence of advertisement to be measured, is examined by the multi-angle of forward and reverse
Consider, recommends the advertisement that most probable is clicked and most probable places an order to user, can be reduced on the basis of improving click-through-rate
The probability exposed in vain, and then promote the overall exposing income of ad system；Also, it is instructed by the cross feature of user and advertisement
Practice and click prediction model and lower single prediction model, further improves the accuracy of model.
Embodiment three
Fig. 5 is the flow chart of the advertisement recommended method in the embodiment of the present invention three.The present embodiment can be with above-described embodiment
Basis is illustrated advertisement recommended method by a specific example.The first default historical time is in the present embodiment with one
It is illustrated for week, the second default historical time is illustrated for one month.Referring to Fig. 5, this method specifically be can wrap
It includes:
S301, beginning.
S302, exposure and click data in nearly one week, are handled by Feature Engineering.
Specifically, the exposure and click data, i.e. the first advertisement log data extracted in history nearest one week pass through feature
Project treatment obtains first sample data, the training set and test set first sample data being divided into Fig. 5.
S303, the matching same day effective advertisement.
Effective advertisement that the same day is determined according to the real time data on the same day, that is, determine advertisement to be measured.
S304, initialization high-order LR model.
S305, training high-order LR model.
Training set input higher-order logic in first sample data is returned into (Logistic Regression, LR) model
In, it is trained.
S306, test high-order LR model.
Trained high-order LR model is tested by the test set in first sample data.
Whether S307, AUC are subjected to.
Judge whether the AUC (Area Under the Curve) of high-order LR model meets the requirements, if meeting the requirements,
To prediction model is clicked, S308 is executed, if being unsatisfactory for requiring, returns and executes S302.
S308, predict target user to the click probability of the same day effective advertisement.
By the related data of target user and the same day effective advertisement, by Feature Engineering, treated that first object input is special
Sign, the high-order LR model that input test is completed, click probability of the available target user to the same day effective advertisement.
S309, initial sequence.
According to target user to the click probability of the same day effective advertisement, according to sequence from big to small to the same day effective advertisement
It is ranked up, is initially sorted.
S310, advertising platform preference classification is determined.
It determines the Matrix classification of Current ad platform, is not construed as limiting in quantity the present embodiment of preference classification.
S311, it is one month nearly in target user and the advertisement now of preference class interaction data, handled by Feature Engineering.
The interaction data of target user and the advertisement now of preference class in nearly one month, i.e. the second advertisement log data are extracted,
It is handled by Feature Engineering, obtains the second sample data, the training set and test set the second sample data being divided into Fig. 5.
S312, the matching same day effective advertisement.
Effective advertisement on the day of determining advertising platform preference class now according to the real time data on the same day determines advertising platform
The advertisement to be measured of preference class now.
S313, initialization XGB model.
S314, training XGB model.
By in training set input Xgboost (XGB) model in the second sample data, it is trained.
S315, test XGB model.
Trained XGB model is tested by the test set in the second sample data.
Whether S316, AUC are subjected to.
Judge whether the AUC (Area Under the Curve) of XGB model meets the requirements, if meeting the requirements, obtains
Lower list prediction model, executes S317, if being unsatisfactory for requiring, returns and executes S311.
S317, predict target user to lower single probability of the preference class same day effectively advertisement now.
By the related data of target user and preference the class same day effectively advertisement now, by Feature Engineering, treated second
Target input feature vector, the XGB model that input test is completed, available target user is to the same day effectively advertisement now of preference class
Lower list probability.
S318, recommend sequence.
It is effective to the same day determined in S309 according to lower single probability of the target user to the same day effectively advertisement now of preference class
The initial sequence of advertisement is adjusted, and obtains final recommendation sequence.
S319, related service rule-based filtering.
It determines after recommending sequence, can also be filtered by other relevant business rules to recommending to sort, such as
It is filtered by department's ID duplicate removal to recommending to sort.
S320, transmitting advertisement front end.
The recommendation sequence of effective advertisement of the filtered same day is sent to advertisement front end, to show user.
S321, end.
Advertisement recommended method is further explained by a specific example in the present embodiment, the present embodiment
By clicking the initial sequence for the click determine the probability advertisement to be measured that prediction model determines, and determined by lower single prediction model
Lower list probability is modified initial sequence, finally obtains the recommendation sequence of advertisement to be measured, passes through the multi-angle of forward and reverse
Consider, recommends the advertisement that most probable is clicked and most probable places an order to user, can be dropped on the basis of improving click-through-rate
The low probability exposed in vain, and then promote the overall exposing income of ad system.
Example IV
Fig. 6 is the structural schematic diagram of the advertisement recommendation apparatus in the embodiment of the present invention four, and the present embodiment is applicable to realize
The case where advertisement is recommended.Advertisement recommendation apparatus provided by the embodiment of the present invention can be performed provided by any embodiment of the invention
Advertisement recommended method has the corresponding functional module of execution method and beneficial effect.
The device specifically include feature obtain module 410, click prediction module 420, initial sorting module 430, place an order it is pre-
Survey module 440 and sequence correction module 450, in which:
Feature obtains module 410, for obtaining the first object input feature vector and target of target user Yu advertisement to be measured
The second target input feature vector of user and the advertisement to be measured now of emphasis class；
Prediction module 420 is clicked, for first object input feature vector to be inputted the click prediction model constructed in advance, is obtained
Click probability of the target user to each advertisement to be measured；
Initial sorting module 430 determines advertisement to be measured for the click probability according to target user to each advertisement to be measured
Initial sequence；
Lower list prediction module 440, lower single prediction model for constructing the input of the second target input feature vector in advance, determines
Lower single probability of the target user to advertisement to be measured each under emphasis classification；
Sort correction module 450, right for lower single probability according to target user to advertisement to be measured each under emphasis classification
Initial sequence is modified, and determines the recommendation sequence of advertisement to be measured.
Prediction model and lower single prediction model are clicked in building of the embodiment of the present invention, by obtaining target user and advertisement to be measured
First object input feature vector and target user and the advertisement to be measured now of emphasis class the second target input feature vector, by first
Target input feature vector inputs the click prediction model that constructs in advance, obtains target user to the click probability of each advertisement to be measured,
And determine the initial sequence of advertisement to be measured；Lower single prediction model that the input of second target input feature vector is constructed in advance, determines mesh
User is marked to lower single probability of advertisement to be measured each under emphasis classification, and according to target user to each to be measured wide under emphasis classification
The lower single probability accused, is modified initial sequence, determines the recommendation sequence of advertisement to be measured.Technology provided in an embodiment of the present invention
Scheme by clicking the initial sequence for the click determine the probability advertisement to be measured that prediction model determines, and passes through lower single prediction model
Determining lower single probability is modified initial sequence, finally obtains the recommendation sequence of advertisement to be measured, passes through forward and reverse
Multi-angle considers, recommends the advertisement that most probable is clicked and most probable places an order to user, can be in the base for improving click-through-rate
The probability exposed in vain is reduced on plinth, and then promotes the overall exposing income of ad system.
Further, initial sorting module 430 is specifically used for:
According to target user to the click probability of each advertisement to be measured, advertisement to be measured is carried out according to sequence from big to small
Sequence, is initially sorted.
Further, sequence correction module 450 is specifically used for:
For belonging to emphasis classification, and lower single probability of target user is less than the advertisement to be measured of probability threshold value, by its
Position in initial sequence is suppressed according to setting rule, and corrects initial sequence.
Further, the device further include:
Model construction module, for constructing before obtaining the first object input feature vector of target user and advertisement to be measured
Click prediction model and lower single prediction model.
Further, model construction module includes clicking prediction model unit, clicks prediction model unit and is specifically used for:
Obtain the first advertisement log data in the first default historical time；
First sample data are generated according to the first advertisement log data, first sample data include first sample input feature vector
And corresponding click label, first sample input feature vector includes user characteristics, characteristic of advertisement and user and advertisement intersects
Feature；
Higher-order logic regression model is trained according to first sample data, obtains clicking prediction model.
Further, model construction module includes lower single prediction model unit, and lower list prediction model unit is specifically used for:
It obtains in the second default historical time, the second advertisement log data of emphasis class now；
The second sample data is generated according to the second advertisement log data, the second sample data includes the second sample input feature vector
And corresponding lower single label, the second sample input feature vector includes user characteristics, user's air control feature and user and emphasis class
Now the cross feature of advertisement；
Xgboost model is trained according to the second sample data, obtains lower single prediction model.
Further, first object input feature vector include target user's feature, characteristic of advertisement to be measured and target user with
The cross feature of advertisement to be measured, the second target input feature vector include target user's feature, target user's air control feature and target
The cross feature of user and the advertisement to be measured now of emphasis class, target user are the user determined based on hot spot rule, advertisement to be measured
For effective advertisement in the setting object time.
Advertisement recommendation apparatus provided by the embodiment of the present invention can be performed advertisement provided by any embodiment of the invention and push away
Method is recommended, has the corresponding functional module of execution method and beneficial effect.
Embodiment five
Fig. 7 is the structural schematic diagram of the terminal in the embodiment of the present invention five.Fig. 7, which is shown, to be suitable for being used to realizing that the present invention is real
Apply the block diagram of the exemplary terminal 512 of mode.The terminal 512 that Fig. 7 is shown is only an example, should not be to the embodiment of the present invention
Function and use scope bring any restrictions.
As shown in fig. 7, terminal 512 is showed in the form of general purpose terminal.The component of terminal 512 can include but is not limited to:
One or more processor 516, storage device 528 connect different system components (including storage device 528 and processor
516) bus 518.
Bus 518 indicates one of a few class bus structures or a variety of, including storage device bus or storage device control
Device processed, peripheral bus, graphics acceleration port, processor or total using the local of any bus structures in a variety of bus structures
Line.For example, these architectures include but is not limited to industry standard architecture (Industry Subversive
Alliance, ISA) bus, microchannel architecture (Micro Channel Architecture, MAC) bus is enhanced
Isa bus, Video Electronics Standards Association (Video Electronics Standards Association, VESA) local are total
Line and peripheral component interconnection (Peripheral Component Interconnect, PCI) bus.
Terminal 512 typically comprises a variety of computer system readable media.These media can be it is any can be by terminal
The usable medium of 512 access, including volatile and non-volatile media, moveable and immovable medium.
Storage device 528 may include the computer system readable media of form of volatile memory, such as arbitrary access
Memory (Random Access Memory, RAM) 530 and/or cache memory 532.Terminal 512 can be wrapped further
Include other removable/nonremovable, volatile/non-volatile computer system storage mediums.Only as an example, storage system
534 can be used for reading and writing immovable, non-volatile magnetic media (Fig. 7 do not show, commonly referred to as " hard disk drive ").Although
It is not shown in Fig. 7, the disc driver for reading and writing to removable non-volatile magnetic disk (such as " floppy disk ") can be provided, and
To removable anonvolatile optical disk, such as CD-ROM (Compact Disc Read-Only Memory, CD-ROM), number
Optic disk (Digital Video Disc-Read Only Memory, DVD-ROM) or other optical mediums) read-write CD drive
Dynamic device.In these cases, each driver can be connected by one or more data media interfaces with bus 518.It deposits
Storage device 528 may include at least one program product, which has one group of (for example, at least one) program module, this
A little program modules are configured to perform the function of various embodiments of the present invention.
Program/utility 540 with one group of (at least one) program module 542 can store in such as storage dress
It sets in 528, such program module 542 includes but is not limited to operating system, one or more application program, other program moulds
It may include the realization of network environment in block and program data, each of these examples or certain combination.Program module
542 usually execute function and/or method in embodiment described in the invention.
Terminal 512 can also be logical with one or more external equipments 514 (such as keyboard, direction terminal, display 524 etc.)
Letter, can also be enabled a user to one or more terminal interact with the terminal 512 communicate, and/or with make the terminal 512
Any terminal (such as network interface card, modem etc.) communication that can be communicated with one or more of the other computing terminal.This
Kind communication can be carried out by input/output (I/O) interface 522.Also, terminal 512 can also by network adapter 520 with
One or more network (such as local area network (Local Area Network, LAN), wide area network (Wide Area Network,
WAN) and/or public network, for example, internet) communication.As shown in fig. 7, network adapter 520 passes through bus 518 and terminal 512
Other modules communication.It should be understood that although not shown in the drawings, other hardware and/or software mould can be used in conjunction with terminal 512
Block, including but not limited to: microcode, terminal driver, redundant processor, external disk drive array, disk array
(Redundant Arrays of Independent Disks, RAID) system, tape drive and data backup storage system
System etc..
The program that processor 516 is stored in storage device 528 by operation, thereby executing various function application and number
According to processing, such as realize advertisement recommended method provided by the embodiment of the present invention, this method comprises:
The first object input feature vector and target user and emphasis class of acquisition target user and advertisement to be measured are to be measured now
Second target input feature vector of advertisement；
First object input feature vector is inputted to the click prediction model constructed in advance, obtains target user to each to be measured wide
The click probability of announcement；
According to target user to the click probability of each advertisement to be measured, the initial sequence of advertisement to be measured is determined；
Lower single prediction model that the input of second target input feature vector is constructed in advance, determines target user under emphasis classification
Lower single probability of each advertisement to be measured；
According to target user to lower single probability of advertisement to be measured each under emphasis classification, initial sequence is modified, really
The recommendation sequence of fixed advertisement to be measured.
Embodiment six
The embodiment of the present invention six additionally provides a kind of computer readable storage medium, is stored thereon with computer program, should
The advertisement recommended method as provided by the embodiment of the present invention is realized when program is executed by processor, this method comprises:
The first object input feature vector and target user and emphasis class of acquisition target user and advertisement to be measured are to be measured now
Second target input feature vector of advertisement；
First object input feature vector is inputted to the click prediction model constructed in advance, obtains target user to each to be measured wide
The click probability of announcement；
According to target user to the click probability of each advertisement to be measured, the initial sequence of advertisement to be measured is determined；
Lower single prediction model that the input of second target input feature vector is constructed in advance, determines target user under emphasis classification
Lower single probability of each advertisement to be measured；
According to target user to lower single probability of advertisement to be measured each under emphasis classification, initial sequence is modified, really
The recommendation sequence of fixed advertisement to be measured.
The computer storage medium of the embodiment of the present invention, can be using any of one or more computer-readable media
Combination.Computer-readable medium can be computer-readable signal media or computer readable storage medium.It is computer-readable
Storage medium for example may be-but not limited to-the system of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, device or
Device, or any above combination.The more specific example (non exhaustive list) of computer readable storage medium includes: tool
There are electrical connection, the portable computer diskette, hard disk, random access memory (RAM), read-only memory of one or more conducting wires
(ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.In this document, computer-readable storage
Medium can be any tangible medium for including or store program, which can be commanded execution system, device or device
Using or it is in connection.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
The computer for executing operation of the present invention can be write with one or more programming languages or combinations thereof
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
It further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partially executes or executed on remote computer or terminal completely on the remote computer on the user computer.It is relating to
And in the situation of remote computer, remote computer can pass through the network of any kind --- including local area network (LAN) or extensively
Domain net (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as provided using Internet service
Quotient is connected by internet).
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (10)
1. a kind of advertisement recommended method characterized by comprising
The first object input feature vector and the target user and emphasis class of acquisition target user and advertisement to be measured are described now
Second target input feature vector of advertisement to be measured；
The first object input feature vector is inputted to the click prediction model constructed in advance, obtains the target user to each institute
State the click probability of advertisement to be measured；
According to the target user to the click probability of each advertisement to be measured, the initial sequence of the advertisement to be measured is determined；
Lower single prediction model that second target input feature vector input is constructed in advance, determines the target user to emphasis class
Now lower single probability of each advertisement to be measured；
According to the target user to lower single probability of the advertisement to be measured each under emphasis classification, the initial sequence is carried out
Amendment determines the recommendation sequence of the advertisement to be measured.
2. the method according to claim 1, wherein according to the target user to each advertisement to be measured
Probability is clicked, determines the initial sequence of the advertisement to be measured, comprising:
According to the target user to the click probability of each advertisement to be measured, according to sequence from big to small to the advertisement to be measured
It is ranked up, obtains the initial sequence.
3. the method according to claim 1, wherein according to the target user to each described under emphasis classification
Lower single probability of advertisement to be measured, is modified the initial sequence, comprising:
For belonging to the emphasis classification, and lower single probability of the target user is less than the described to be measured wide of probability threshold value
It accuses, its position in the initial sequence is suppressed according to setting rule, and correct the initial sequence.
4. the method according to claim 1, wherein the first object for obtaining target user and advertisement to be measured inputs
Before feature, further includes:
Construct the click prediction model and lower single prediction model.
5. according to the method described in claim 4, it is characterized in that, constructing the click prediction model, comprising:
Obtain the first advertisement log data in the first default historical time；
First sample data are generated according to first advertisement log data, the first sample data include first sample input
Feature and corresponding click label, the first sample input feature vector include user characteristics, characteristic of advertisement and user and wide
The cross feature of announcement；
Higher-order logic regression model is trained according to the first sample data, obtains clicking prediction model.
6. according to the method described in claim 4, it is characterized in that, building lower single prediction model, comprising:
It obtains in the second default historical time, the second advertisement log data of the emphasis class now；
The second sample data is generated according to second advertisement log data, second sample data is inputted including the second sample
Feature and corresponding lower single label, the second sample input feature vector includes user characteristics, user's air control feature and user
With the cross feature of the emphasis class advertisement now；
Xgboost model is trained according to second sample data, obtains lower single prediction model.
7. any method in -6 according to claim 1, which is characterized in that the first object input feature vector includes target
The cross feature of user characteristics, characteristic of advertisement to be measured and the target user and the advertisement to be measured, second target are defeated
Enter feature include target user's feature, target user's air control feature and the target user and the emphasis class it is described now to
The cross feature of advertisement is surveyed, the target user is the user determined based on hot spot rule, and the advertisement to be measured is setting target
Effective advertisement in time.
8. a kind of advertisement recommendation apparatus characterized by comprising
Feature obtains module, and the first object input feature vector and the target for obtaining target user and advertisement to be measured are used
The second target input feature vector at family and the advertisement to be measured described now of emphasis class；
Prediction module is clicked, for the first object input feature vector to be inputted the click prediction model constructed in advance, obtains institute
Target user is stated to the click probability of each advertisement to be measured；
Initial sorting module, for the click probability according to the target user to each advertisement to be measured, determine it is described to
Survey the initial sequence of advertisement；
Lower list prediction module, lower single prediction model for constructing the second target input feature vector input in advance, determines institute
Target user is stated to lower single probability of the advertisement to be measured each under emphasis classification；
Sort correction module, for lower single probability according to the target user to the advertisement to be measured each under emphasis classification,
The initial sequence is modified, determines the recommendation sequence of the advertisement to be measured.
9. a kind of terminal, which is characterized in that the terminal includes:
One or more processors；
Storage device, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now advertisement recommended method as described in any in claim 1-7.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The advertisement recommended method as described in any in claim 1-7 is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910640133.8A CN110363590A (en) | 2019-07-16 | 2019-07-16 | A kind of advertisement recommended method, device, terminal and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910640133.8A CN110363590A (en) | 2019-07-16 | 2019-07-16 | A kind of advertisement recommended method, device, terminal and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110363590A true CN110363590A (en) | 2019-10-22 |
Family
ID=68219550
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910640133.8A Pending CN110363590A (en) | 2019-07-16 | 2019-07-16 | A kind of advertisement recommended method, device, terminal and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110363590A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110782286A (en) * | 2019-10-25 | 2020-02-11 | 上海喜马拉雅科技有限公司 | Advertisement pushing method and device, server and computer readable storage medium |
| CN111028005A (en) * | 2019-11-29 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Advertisement display method and device |
| CN111176965A (en) * | 2019-12-19 | 2020-05-19 | 广州华多网络科技有限公司 | Pre-release test method and device for recommendation system and electronic equipment |
| CN111311328A (en) * | 2020-02-20 | 2020-06-19 | 支付宝(杭州)信息技术有限公司 | Method and device for determining advertisement click rate of product under advertisement channel |
| CN111833096A (en) * | 2020-06-10 | 2020-10-27 | 北京龙云科技有限公司 | Advertisement recommendation method and system based on machine learning |
| CN111861623A (en) * | 2019-12-30 | 2020-10-30 | 北京骑胜科技有限公司 | Information recommendation method, device and equipment |
| CN112036951A (en) * | 2020-09-03 | 2020-12-04 | 猪八戒股份有限公司 | Business opportunity recommendation method, system, electronic device and medium based on CNN model |
| CN112488759A (en) * | 2020-11-30 | 2021-03-12 | 上海酷量信息技术有限公司 | Advertisement sorting system and method |
| CN113065896A (en) * | 2021-03-29 | 2021-07-02 | 上海酷量信息技术有限公司 | Advertisement recommendation method and device |
| CN113344648A (en) * | 2021-08-05 | 2021-09-03 | 北京龙云科技有限公司 | Advertisement recommendation method and system based on machine learning |
| CN113919856A (en) * | 2020-07-09 | 2022-01-11 | 上海钧正网络科技有限公司 | Target user selection method, system, device and storage medium |
| WO2022089020A1 (en) * | 2020-10-30 | 2022-05-05 | 京东方科技集团股份有限公司 | Method and device for event displaying, storage medium, and electronic device |
| CN114862447A (en) * | 2022-04-08 | 2022-08-05 | 无线生活（杭州）信息科技有限公司 | Advertisement reordering method and device |
| CN115049445A (en) * | 2022-08-17 | 2022-09-13 | 铅笔头(深圳)科技有限公司 | Advertisement production evaluation system and method based on influence indexes |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100250335A1 (en) * | 2009-03-31 | 2010-09-30 | Yahoo! Inc | System and method using text features for click prediction of sponsored search advertisements |
| CN104965890A (en) * | 2015-06-17 | 2015-10-07 | 深圳市腾讯计算机系统有限公司 | Advertisement recommendation method and apparatus |
| CN107105031A (en) * | 2017-04-20 | 2017-08-29 | 北京京东尚科信息技术有限公司 | Information-pushing method and device |
| CN108416649A (en) * | 2018-02-05 | 2018-08-17 | 北京三快在线科技有限公司 | Search result ordering method, device, electronic equipment and storage medium |
| CN109558544A (en) * | 2018-12-12 | 2019-04-02 | 拉扎斯网络科技（上海）有限公司 | Sort method and device, server and storage medium |
| CN109615411A (en) * | 2018-10-29 | 2019-04-12 | 中国平安人寿保险股份有限公司 | Advertisement placement method and device, electronic equipment based on algorithm model |
| CN110008399A (en) * | 2019-01-30 | 2019-07-12 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
-
2019
- 2019-07-16 CN CN201910640133.8A patent/CN110363590A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100250335A1 (en) * | 2009-03-31 | 2010-09-30 | Yahoo! Inc | System and method using text features for click prediction of sponsored search advertisements |
| CN104965890A (en) * | 2015-06-17 | 2015-10-07 | 深圳市腾讯计算机系统有限公司 | Advertisement recommendation method and apparatus |
| CN107105031A (en) * | 2017-04-20 | 2017-08-29 | 北京京东尚科信息技术有限公司 | Information-pushing method and device |
| CN108416649A (en) * | 2018-02-05 | 2018-08-17 | 北京三快在线科技有限公司 | Search result ordering method, device, electronic equipment and storage medium |
| CN109615411A (en) * | 2018-10-29 | 2019-04-12 | 中国平安人寿保险股份有限公司 | Advertisement placement method and device, electronic equipment based on algorithm model |
| CN109558544A (en) * | 2018-12-12 | 2019-04-02 | 拉扎斯网络科技（上海）有限公司 | Sort method and device, server and storage medium |
| CN110008399A (en) * | 2019-01-30 | 2019-07-12 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110782286A (en) * | 2019-10-25 | 2020-02-11 | 上海喜马拉雅科技有限公司 | Advertisement pushing method and device, server and computer readable storage medium |
| CN110782286B (en) * | 2019-10-25 | 2023-08-18 | 上海喜马拉雅科技有限公司 | Advertisement pushing method, advertisement pushing device, server and computer readable storage medium |
| CN111028005A (en) * | 2019-11-29 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Advertisement display method and device |
| CN111028005B (en) * | 2019-11-29 | 2021-11-05 | 腾讯科技（深圳）有限公司 | Advertisement display method and device |
| CN111176965A (en) * | 2019-12-19 | 2020-05-19 | 广州华多网络科技有限公司 | Pre-release test method and device for recommendation system and electronic equipment |
| CN111176965B (en) * | 2019-12-19 | 2023-10-13 | 广州方硅信息技术有限公司 | Recommendation system pre-release test method and device and electronic equipment |
| CN111861623A (en) * | 2019-12-30 | 2020-10-30 | 北京骑胜科技有限公司 | Information recommendation method, device and equipment |
| CN111311328A (en) * | 2020-02-20 | 2020-06-19 | 支付宝(杭州)信息技术有限公司 | Method and device for determining advertisement click rate of product under advertisement channel |
| CN111311328B (en) * | 2020-02-20 | 2022-07-01 | 支付宝(杭州)信息技术有限公司 | Method and device for determining advertisement click rate of product under advertisement channel |
| CN111833096A (en) * | 2020-06-10 | 2020-10-27 | 北京龙云科技有限公司 | Advertisement recommendation method and system based on machine learning |
| CN113919856A (en) * | 2020-07-09 | 2022-01-11 | 上海钧正网络科技有限公司 | Target user selection method, system, device and storage medium |
| CN112036951A (en) * | 2020-09-03 | 2020-12-04 | 猪八戒股份有限公司 | Business opportunity recommendation method, system, electronic device and medium based on CNN model |
| WO2022089020A1 (en) * | 2020-10-30 | 2022-05-05 | 京东方科技集团股份有限公司 | Method and device for event displaying, storage medium, and electronic device |
| CN112488759A (en) * | 2020-11-30 | 2021-03-12 | 上海酷量信息技术有限公司 | Advertisement sorting system and method |
| CN113065896A (en) * | 2021-03-29 | 2021-07-02 | 上海酷量信息技术有限公司 | Advertisement recommendation method and device |
| CN113344648A (en) * | 2021-08-05 | 2021-09-03 | 北京龙云科技有限公司 | Advertisement recommendation method and system based on machine learning |
| CN114862447A (en) * | 2022-04-08 | 2022-08-05 | 无线生活（杭州）信息科技有限公司 | Advertisement reordering method and device |
| CN115049445A (en) * | 2022-08-17 | 2022-09-13 | 铅笔头(深圳)科技有限公司 | Advertisement production evaluation system and method based on influence indexes |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110363590A (en) | A kind of advertisement recommended method, device, terminal and storage medium | |
| CN107346496B (en) | Target user orientation method and device | |
| CN102722833B (en) | Open and the persistent metadata of user's control strategy that uses of personal data for online advertisement | |
| EP2242016A1 (en) | Touchpoint customization system | |
| CN106503006A (en) | The sort method and device of application App neutron applications | |
| CN107767172A (en) | Information-pushing method, device, server and medium | |
| Vakulenko et al. | Enriching iTunes App Store Categories via Topic Modeling. | |
| Dang et al. | Consumer attitudes toward facial recognition payment: an examination of antecedents and outcomes | |
| CN108256098A (en) | A kind of method and device of determining user comment Sentiment orientation | |
| CN108921398A (en) | Shop quality evaluating method and device | |
| CN110033294A (en) | A kind of determination method of business score value, business score value determining device and medium | |
| CN109299356A (en) | Activity recommendation method, apparatus, electronic equipment and storage medium based on big data | |
| CN106204115A (en) | Information popularization method, server and client | |
| JP2022015114A (en) | Information processing system | |
| CN102148003A (en) | Text advertisement enabling voice | |
| CN113761348A (en) | Information recommendation method and device, electronic equipment and storage medium | |
| CN112084223A (en) | User portrait generation method, device, equipment and storage medium | |
| CN111191894A (en) | Method and device for processing resource demand based on user classification and electronic equipment | |
| CN113781149A (en) | Information recommendation method and device, computer-readable storage medium and electronic equipment | |
| CN111523914A (en) | User satisfaction evaluation method, device and system and data display platform | |
| CN108229804A (en) | Brand promotion project management method, device, terminal device and storage medium | |
| CN109829593A (en) | The credit rating of target object determines method, apparatus, storage medium and electronic device | |
| CN111192045A (en) | Anti-cheating method, device and system based on transaction record information | |
| CN115809889A (en) | Intelligent passenger group screening method, system, medium and equipment based on marketing effect | |
| CN112015970A (en) | Product recommendation method, related equipment and computer 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191022 |