Summary of the invention
In order to solve the above technical problems, present disclose provides one kind for carrying out Object Push from server to user terminal
Method, comprising:
Multiple correlation behavior data of the user terminal about multiple objects to be recommended are obtained from database;
For each of multiple objects to be recommended recommended, according to the user terminal about the object to be recommended
The first correlation behavior data determine the score range of the object to be recommended, the score range includes the score value upper bound and lower bound;
For each described object to be recommended, according in its score range score value and the user terminal about this
Second correlation behavior data of object to be recommended determine the corresponding recommendation of the object to be recommended;And
Maximum recommended is worth corresponding object to be recommended as optimal Object Push to be recommended to user by the server
Terminal.
Optionally, this method further comprises:
Determine the ranking factor of the multiple object to be recommended;
For each in the multiple object to be recommended in addition to the optimal object to be recommended, based on described optimal
The ranking factor of object to be recommended and the object to be recommended adjusts the score range of the object to be recommended, and to be recommended at this
A score value is selected in the adjusted score range of object;
According to the second correlation behavior number of the selected score value of the multiple object to be recommended and the multiple object to be recommended
It is ranked up according to the multiple object to be recommended;And
The multiple object to be recommended is pushed according to the sequence.
Optionally, the ranking factor is determined based on the score value upper bound and third correlation behavior data.
Optionally, the ranking factor of the object to be recommended includes the third correlation behavior data of the object to be recommended
With the product in the score value upper bound, and described adjusted based on the ranking factor of the optimal object to be recommended and the object to be recommended
The score value of the object to be recommended includes:
The ranking factor of the object to be recommended is adjusted so that the ranking factor of the object to be recommended is no more than institute
State the ranking factor of optimal object to be recommended;And
Adjusted score value is obtained using the adjusted ranking factor of the object to be recommended.
Optionally, the multiple correlation behavior data are to be obtained by model prediction and/or are the user terminals
Historical behavior data.
Optionally, the multiple correlation behavior data include the user terminal about object to be recommended clicking rate and
Conversion ratio and combinations thereof.
Optionally, the score value upper bound of the determination object to be recommended and lower bound include:
If the first correlation behavior data of the object to be recommended are higher than threshold value, point of the object to be recommended is improved
It is worth the upper bound;And
If the first correlation behavior data of the object to be recommended are lower than the threshold value, the object to be recommended is reduced
Score value lower bound.
Optionally, the corresponding recommendation of the determination object to be recommended comprises determining that the score value of the object to be recommended
The maximum value of objective function in range, the objective function are the functions of the second correlation behavior data and score value.
Another aspect provides a kind of for making server to the device of user terminal progress Object Push, packet
It includes:
For obtaining mould of the user terminal about multiple correlation behavior data of multiple objects to be recommended from database
Block;
It is to be recommended about this according to the user terminal for being directed to each of multiple objects to be recommended recommended
First correlation behavior data of object determine the module of the score range of the object to be recommended, and the score range includes on score value
Boundary and lower bound;
For for each described object to be recommended, according in its score range score value and the user terminal close
In the second correlation behavior data of the object to be recommended, the module of the corresponding recommendation of the object to be recommended is determined;And
For make the server using maximum recommended be worth corresponding object to be recommended as optimal Object Push to be recommended to
The module of user terminal.
Optionally, which further comprises:
For determining the module of the ranking factor of the multiple object to be recommended;
For being based on described for each in the multiple object to be recommended in addition to the optimal object to be recommended
The ranking factor of optimal object to be recommended and the object to be recommended adjusts the score range of the object to be recommended, and waits at this
The module of one score value of selection in the adjusted score range of recommended;
For according to the selected score value of the multiple object to be recommended and the second associated line of the multiple object to be recommended
The module that the multiple object to be recommended is ranked up for data;And
For according to the module for sorting and being pushed to the multiple object to be recommended.
Optionally, the ranking factor is determined based on the score value upper bound and the third correlation behavior data
's.
Optionally, the ranking factor of the object to be recommended includes the third correlation behavior data of the object to be recommended
With the product in the score value upper bound, and it is described for the ranking factor based on the optimal object to be recommended and the object to be recommended come
The module for adjusting the score value of the object to be recommended includes:
It is adjusted for the ranking factor to the object to be recommended so that the ranking factor of the object to be recommended is little
In the module of the ranking factor of the optimal object to be recommended;And
The module of adjusted score value is obtained for using the adjusted ranking factor of the object to be recommended.
Optionally, the multiple correlation behavior data are to be obtained by model prediction and/or are the user terminals
Historical behavior data.
Optionally, the multiple correlation behavior data include the user terminal about object to be recommended clicking rate and
Conversion ratio and combinations thereof.
Optionally, described for determining the score value upper bound of the object to be recommended and the module of lower bound includes:
If the first correlation behavior data for the object to be recommended are higher than threshold value, the object to be recommended is improved
The score value upper bound module;And
If the first correlation behavior data for the object to be recommended are lower than the threshold value, reduce described to be recommended
The module of the score value lower bound of object.
Optionally, the module for determining the corresponding recommendation of the object to be recommended includes: for determine should be to
The module of the maximum value of objective function in the score range of recommended, the objective function are the second correlation behavior data
With the function of score value.
Further aspect of the invention provides a kind of computer equipment, comprising:
Processor;And
It is arranged to the memory of storage computer executable instructions, the executable instruction makes the place when executed
It manages device and executes following operation:
Multiple correlation behavior data of the user terminal about multiple objects to be recommended are obtained from database;
For each of multiple objects to be recommended recommended, according to the user terminal about the object to be recommended
The first correlation behavior data determine the score range of the object to be recommended, the score range includes the score value upper bound and lower bound;
For each described object to be recommended, according in its score range score value and the user terminal about this
Second correlation behavior data of object to be recommended determine the corresponding recommendation of the object to be recommended;And
Maximum recommended is worth corresponding object to be recommended as optimal Object Push to be recommended to user terminal.
Optionally, the executable instruction makes the processor further execute following operation when executed:
Determine the ranking factor of the multiple object to be recommended;
For each in the multiple object to be recommended in addition to the optimal object to be recommended, based on described optimal
The ranking factor of object to be recommended and the object to be recommended adjusts the score range of the object to be recommended, and to be recommended at this
A score value is selected in the adjusted score range of object;
According to the second correlation behavior number of the selected score value of the multiple object to be recommended and the multiple object to be recommended
It is ranked up according to the multiple object to be recommended;And
The multiple object to be recommended is pushed according to the sequence.
Compared with prior art, the disclosure has the advantage that
The disclosure both considers preference of the user about object to be recommended in the push of object to be recommended, it is also contemplated that
Recommended is advocated peace the interests of subject table side to be recommended, using user about object to be recommended behavioral data (for example, association
The historical behavior data of behavioral data, user about object to be recommended) and the score value of object to be recommended itself come to pushing
Object to be recommended be ranked up, thus, it is possible to reach efficient Object Push to be recommended.
Specific embodiment
For the above objects, features, and advantages of the disclosure can be clearer and more comprehensible, below in conjunction with attached drawing to the tool of the disclosure
Body embodiment elaborates.
Many details are explained in the following description in order to fully understand the disclosure, but the disclosure can be with
It is different from other way described herein using other and implements, therefore the disclosure is by the limit of following public specific embodiment
System.
In the popularization of object to be recommended (for example, advertisement), object score value to be recommended can embody the object pair to be recommended
In the importance or value of ob-ject provider to be recommended.Ob-ject provider to be recommended wishes that the higher object to be recommended of score value can
It is preferential to show for client's selection.
For example, the score value can be the bid of advertisement (that is, advertising platform side in the case where object to be recommended is advertisement
The expense for the displaying advertisement collected to advertiser).It is generally desirable to higher advertisements of bidding preferentially to show to user for advertising platform side,
Thus bigger income is obtained.
The disclosure had both considered the preference of user in the push sequence of object to be recommended, it is also contemplated that object to be recommended
Score value, so that the optimum combination of user preference and object score value to be recommended is obtained, to improve the marketing objectives of object to be recommended.Tool
For body, the disclosure determines correlation behavior data (for example, prediction clicking rate and conversion ratio) of the user about object to be recommended, makes
The score range (for example, the score value upper bound and lower bound) for adjusting object to be recommended with the correlation behavior data, further use to
The variation range of recommended score value is ranked up to be shown to the user multiple objects to be recommended, so that wait push away
The push for recommending object is highly efficient.
The behavioral data of object to be recommended may include clicking rate and conversion ratio.Clicking rate refers in a measurement period
The ratio between sum interior, that the number and object to be recommended that object to be recommended is clicked are demonstrated.Conversion ratio referred to a statistics week
In phase, the ratio between number that number that object to be recommended is selected and object to be recommended are clicked.
For example, in the case where object to be recommended is Insurance Advertisement, the behavioral data of advertisement may include that insurance is wide
The clicking rate (click-through rate, CTR) and conversion ratio (conversion rate, CVR) of announcement, wherein clicking rate is
Refer in a measurement period, the sum that specific insurance advertisement is demonstrated by the number that user clicks divided by the advertisement, that is, click
Rate=(click volume/displaying amount) × 100%;Conversion ratio refers to that in a measurement period, specific insurance advertisement is insured secondary
The sum that number is clicked divided by the advertisement, that is, click conversion ratio=(amount of insuring/click volume) × 100%.
Fig. 1 is the information transmission system figure according to all aspects of this disclosure.
As shown in Figure 1, information transmission system includes server 101, database 102 and user terminal 103.
Database 102 can store user characteristics (for example, age, gender, constellation, region, membership information etc.), each user
Correlation behavior data about object to be recommended are (for example, prediction/historic click-through rate of object to be recommended, conversion ratio or its group
Close), the score information (for example, the bid of advertisement, bid ranges etc.) of object to be recommended, etc..
The information stored can be supplied to server 101 so that server 101 is handled by database 102.
For example, server 101 can determine correlation behavior data of the user about object to be recommended;Determine and/or adjust to
The score range of recommended;Determine the optimal object to be recommended about user;Multiple objects to be recommended are ranked up;And
One or more objects to be recommended are pushed to user terminal 103.Server 101 can be by identified various data transmissions to data
Library 102 is stored.
Server 101 can be the cluster of one or more computers.
Alternatively, information transmission system may also include third party's processor (not shown), it can be achieved that above-mentioned about server
Some functions of 101 descriptions, for example, determining correlation behavior data of the user about object to be recommended;It determines and/or adjusts wait push away
Recommend the score range of object;Determine the optimal object to be recommended about user;And multiple objects to be recommended are ranked up, etc.
Deng.
The method that Object Push to be recommended is carried out to user is explained in detailed below.
Fig. 2 is the flow chart according to the method for carrying out Object Push to be recommended to user of the various aspects of the application.
In step 202, for each of multiple objects to be recommended object to be recommended, according to user terminal about wait push away
The first correlation behavior data of object are recommended to determine score range of the object to be recommended about the user terminal.
User terminal about the first correlation behavior data of object to be recommended may include the user that is predicted about to
The clicking rate (pctr) of recommended and conversion ratio (pcvr) or combinations thereof.
First correlation behavior data can be obtained by model prediction.For example, user characteristics, object to be recommended can be used
Feature carries out model prediction as the input of model to obtain user about the prediction clicking rate (pctr) of object to be recommended and turn
Rate (pcvr).The model can be Logic Regression Models, deep neural network etc..
Other first correlation behavior data are also in the conception of the disclosure, as long as the data are able to reflect user about business
Preference.For example, user about object to be recommended historical behavior data (for example, the history row in a cycle
For data), etc..
The score range of object to be recommended can be indicated by the score value upper bound and lower bound.
The behavioral data (for example, clicking rate and/or conversion ratio) of object to be recommended is not constant, but at one section
Between on have fluctuation.Correspondingly, specific user terminal also has fluctuation about the first correlation behavior data of object to be recommended.This
Open the fluctuating come to the object to be recommended according to the first correlation behavior data of object to be recommended for specific user terminal
Score range is adjusted.
For example, if user terminal is higher than about the current first correlation behavior data of object to be recommended at one
Between the first correlation behavior data of all user terminals in set of user terminals (for example, the target user of advertisement gathers) in section
The score value upper bound of the object to be recommended about the user terminal then can be improved in average value.If current behavior data are average lower than this
Value, then can reduce score value lower bound of the object to be recommended about the user terminal.The score value model of object to be recommended is enabled in this way
It encloses according to the fluctuation of the first correlation behavior data of user terminal and changes, the consideration when determining the score range of object to be recommended
To user terminal factor.The score value upper bound of determination object to be recommended described below and a specific example of lower bound.
Wherein btIndicate object score value (for example, advertisement bid) to be recommended,Indicate the lower bound of object score value to be recommended,Indicating the upper bound of object score value to be recommended, u indicates user, and t indicates object to be recommended, pcvr (c | u, t) indicate that user is whole
Prediction conversion ratio (that is, first correlation behavior data) of the end about the object to be recommended, avg (pcvr (c | u, t)) it is certain
In period in set of user terminals the pcvr (c | u, t) of all user terminals average value.rtIt is the constraint to score range
The factor, so that score value will not be adjusted too high or too low, 0 < rt< 1.
In formula (2), it is greater than or equal to avg (pcvr (c | u t) in pcvr (c | u, t)) in the case where,Value
ForIn other words, existThe case where
Under, it can be used lesserTo calculate btThe upper bound, thus make in the range that user is subjected to
Score value is lower, being capable of save the cost.
Note that and be greater than or equal to avg (pcvr (c | u t) in pcvr (c | u, t)) in the case where,Value beIt is preferred embodiment of the present disclosure.It can also be used other increases to be recommended right
As the mode in the score value upper bound, for example,Also in the conception of the disclosure.
As shown in formula (1) and (2), if the pcvr (c | u, t) of user is lower than in the set of user within a certain period of time
The average value of the pcvr (c | u, t) of all users, then can be by btLower bound lower.If the pcvr (c | u, t) of user is higher than
Pcvr (c | u, t) average value, then it can be by btThe upper bound up-regulation.
Specifically, the lower bound of object score value to be recommended can be reduced if the prediction conversion ratio of user is lower;If with
The prediction conversion ratio at family is higher, then the upper bound of object score value to be recommended can be improved.By taking Insurance Advertisement as an example, if the prediction of user
Conversion ratio is lower, reduces advertisement bid under low-quality flow scene (that is, in the lower situation of probability of user's selection advertisement)
Lower bound can save certain cost for advertiser.If the prediction conversion ratio of user is higher, illustrate that user selects the general of the advertisement
Rate is higher, and the income of advertising platform side can be improved in the upper bound for improving advertisement bid.
The model for adjusting object score value to be recommended is illustrated so that user is about the prediction conversion ratio of object to be recommended as an example above
The example enclosed.In other examples, it is possible to use other correlation behavior data are (for example, prediction clicking rate, prediction conversion ratio and pre-
Survey the combination etc. of clicking rate) adjust the range of object score value to be recommended.
Can optional step 204, for each of multiple objects to be recommended object to be recommended, according to its score value upper bound
Ranking factor is determined with the second correlation behavior data.
Ranking factorCan by by user about object to be recommended the second correlation behavior data (for example, clicking rate
Pctr (c | u, t)) and score value upper bound u (bt* it) is multiplied to obtain:
In step 206, optimal object to be recommended is determined.
Specifically, the recommendation of each object to be recommended can be determined, and according to the recommendation of multiple objects to be recommended
Value determines optimal recommended.
It can be for each of multiple objects to be recommended object to be recommended, according in its score range and user is whole
The second correlation behavior data about the object to be recommended are held, determine the corresponding recommendation of the object to be recommended, and will be maximum
The corresponding object to be recommended of recommendation is determined as optimal object to be recommended.
As an example, the corresponding recommendation of object to be recommended can be determined by objective function.In this example,
Can determine in multiple object to be recommended under the constraint of the score range of multiple objects to be recommended maximizes objective function
Optimal object to be recommended.
Score value restriction range is the score value upper bound determined in step 202 and lower bound.Optimal object to be recommended is to user
Displaying in the object to be recommended that can be made number one.
Objective function can be the function of the second correlation behavior data and score value.
For example, objective function can be arranged as follows:
Wherein k indicates object indexing to be recommended, bkIndicate the score value of corresponding object to be recommended, T indicate to include multiple wait push away
Recommend the object set to be recommended of object, | T | indicate in the object set to be recommended object to be recommended sum, ω be normalization because
Son, rtIt is Dynamic gene as described above.Wherein ω and rtIt can be selected according to different characteristics of objects to be recommended.
Further, identical Dynamic gene r has been used in formula (1) and (2)t, but can also be aboutWith
Use different Dynamic gene rtlAnd rtu.In this case, the r in formula (4)tIt can be rtlAnd rtuWeighted sum.
The recommendation C of each object k to be recommendedkIt is the maximum value of objective function in its score range.
It then can be by recommendation CkMaximum object is determined as optimal object to be recommended.
WhereinWithIt is the score value upper bound determined in step 202 and lower bound respectively.Objective function f (k, b* k) be
pctrk·pcvrk·bkMajorized function (for example, joined normalization factor, Dynamic gene).
The example that note that objective function illustrated above, other examples are also possible, as long as objective function is examined
Consider pcvr, pctr and bk?.In other words, objective function in view of user terminal selecting object to be recommended probability (pcvr,
Pctr and/or pctrpcvr (number of insuring/displaying number)) and score value (for example, advertisement bid).For example, objective function can
To be the probability of user terminal selecting object to be recommended and the product of score value, such as, pcvr*bk、pctr*bk、pcvr*pctr*bk
Etc. or product optimization (for example, being included in Dynamic gene as described above, normalization factor).
Can optional step 208, adjusted using the ranking factor of optimal object to be recommended in multiple objects to be recommended
The score range (for example, adjustment score value upper bound) of other objects to be recommended.
Specifically, making the maximum optimal object t to be recommended of objective function f to allowkIt is forward as far as possible, it can update other
Object t to be recommendediThe score value upper boundSo that its ranking factorIt is to be recommended no more than optimal
Object tkRanking factorAnd use the score value upper bound by updatingTo determine the adjusted score value upper bound.
The score value upper bound of object to be recommended is adjusted it is, for example, possible to use following formula:
As shown in formula (6), make other object t to be recommended firstiRanking factorIt is to be recommended right no more than optimal
As tkRanking factorIt then will be initial(it is determined in step 202) withDivided by the second ranking factor
It (is in this example pctri) result in minimum value as final
Can optional step 210, these other objects to be recommended are ranked up using adjusted score range.
Specifically, can determined after the adjusted score range of object to be recommended has been determined in a step 208
In the range of be one score value of Object Selection to be recommended.It may then use that selected score value is ranked up multiple objects to be recommended.
It is, for example, possible to use the simple versions (product as described above) of objective function or objective function to calculate wait push away
The ranking value for recommending object, according to the ranking value of multiple objects to be recommended come to multiple object order to be recommended.
For example, can be object t to be recommended in the case where objective function is formula (4)iCalculate ranking value pctrk·
pcvrk·bk, it is ranked up according to calculated result.
It is related to pctr in objective functionkAnd bkIn the case where, it can be object t to be recommendediCalculate ranking value pctrk·bk,
It is ranked up according to calculated result.
It is related to pcvr in objective functionkAnd bkIn the case where, it can be object t to be recommendediCalculate ranking value pcvrk·bk,
It is ranked up according to calculated result.
In some embodiments, only push optimal object to be recommended to user, therefore without to other objects to be recommended into
Row price adjustment and sequence, therefore step 204,208 and 210 are optional.
In step 212, Object Push to be recommended is carried out to user.
For example, optimal object to be recommended can be pushed to user, to be recommended object of the sequence at former can also be chosen,
It is pushed according to the sequence of step 210.
Note that it is above calculate ranking factor using pctr (c | u, t) in step 204, and correspondingly make in step 208
WithIt determines the score value upper bound of each object to be recommended, thus seeks higher object clicking rate to be recommended.
But other predictive behavior data can also be used to execute step 204 and 208.For example, ranking factorIt can also be with
It is by pcvr (c | u, t) by user about object to be recommended and the score value upper boundWhat multiplication obtained.That is,And correspondingly following formula can be used to determine point of object to be recommended in step 210
It is worth the upper bound:
Thus it can seek higher object conversion ratio to be recommended.
In other examples, it is possible to use the combination of pcvr (c | u, t) and pctr (c | u, t) calculates ranking factor, with
Seek suitable object conversion ratio to be recommended and clicking rate.
Fig. 3 is according to the to be recommended right to determine about the correlation behavior data of the object to be recommended according to user of the disclosure
The flow chart of the score range of elephant.
Fig. 3 corresponds to step 202 described in Fig. 2.
As shown in figure 3, obtaining multiple correlation behavior data of the user about multiple objects to be recommended in step 302.
Multiple correlation behavior data can be clicking rate and conversion ratio and its group of the user terminal about object to be recommended
It closes (for example, product of clicking rate and conversion ratio).
It is for each of multiple objects to be recommended object to be recommended, user is to be recommended right about this in step 304
The first correlation behavior data (for example, pcvr (c | u, t) in step 202) of elephant are compared with a threshold value.
The threshold value can be average correlation behavior data, that is, in a period user set in all users first
The average value of correlation behavior data.
If the first correlation behavior data are greater than threshold value, the score value upper bound of object to be recommended is improved in step 306.
If the first correlation behavior data are less than threshold value, the score value upper bound of object to be recommended is reduced in step 306.
By taking Insurance Advertisement as an example, if the prediction conversion ratio of user is lower, it is to be recommended right to reduce under low-quality flow scene
As the lower bound of score value can save certain cost for advertiser.If the prediction conversion ratio of user is higher, illustrate that user's selection should
The probability of object to be recommended is higher, and the income of advertising platform side can be improved in the upper bound for improving object score value to be recommended.
Fig. 4 is that object to be recommended is determined according to the correlation behavior data of user terminal according to all aspects of this disclosure
The process schematic of score range.
Fig. 4 schematically shows the optimal objects to be recommended determined from 3 objects to be recommended about user terminal
Example.Those skilled in the art will appreciate that the number of object to be recommended can be set according to actual needs.
In frame 401, the first correlation behavior data of object to be recommended are obtained.
For example, the first correlation behavior data of object to be recommended can be obtained from database.
The first correlation behavior data can be predictive behavior data of the user terminal about object to be recommended, be also possible to
Historical behavior data of the user terminal about object to be recommended.For example, the first correlation behavior data can be in step 202
pcvr(c|u,t)。
As an example, the first correlation behavior data can be user terminal in the case where object to be recommended is advertisement
Prediction clicking rate and conversion ratio or historic click-through rate and conversion ratio or combinations thereof about advertisement.
In frame 402, the score range of each object to be recommended is obtained.The score range may include the score value upper bound and lower bound.
Specifically, if user terminal is higher than a threshold value about the current first correlation behavior data of object to be recommended,
The score value upper bound of the object to be recommended about the user terminal then can be improved.If current behavior data are lower than the threshold value, can drop
Score value lower bound of the low object to be recommended about the user terminal.Enable to the score range of object to be recommended according to user in this way
The fluctuation of the correlation behavior data of terminal and change, when determining the score range of object to be recommended in view of user terminal because
Element.
The threshold value can be of all user terminals object to be recommended in set of user terminals in a period of time
The average value of one correlation behavior data.
In the case where the first correlation behavior data are predictive behavior data, which can be uses in a period of time
Average value of all user terminals about prediction the first correlation behavior data of the object to be recommended in the terminal set of family.
It is that (the historical behavior data can be going through for front a cycle to historical behavior data in the first correlation behavior data
History behavioral data) in the case where, which can be a period (period may include one or more periods)
Average value of all user terminals about history the first correlation behavior data of the object to be recommended in interior set of user terminals.
In frame 403, the recommendation of each object to be recommended is determined.
The recommendation of object to be recommended can be determined by objective function as described above.Objective function can be second
The function (for example, described in step 206) of correlation behavior data and score value.Objective function is considered that user terminal selecting waits pushing away
The probability (pcvr, pctr and/or pctrpcvr (number of insuring/displaying number)) and score value for recommending object are (for example, advertisement goes out
Valence).
It can determine recommendation of the objective function maximum value as the object in the score range of each object to be recommended.
In frame 404, optimal object to be recommended is determined.
Specifically, being pushed away object to be recommended corresponding with highest recommendation in multiple objects to be recommended as most preferential treatment
Object Push is recommended to user terminal.
Fig. 5 is the process schematic to multiple object adjustment score values and sequence to be recommended according to all aspects of this disclosure.
As shown in figure 5, determining the ranking factor of each object to be recommended in frame 501.
The ranking factor can be based on user terminal about the second correlation behavior data of object to be recommended and the score value upper bound
To determine.For example, the ranking factor can be the product of the second correlation behavior data and the score value upper bound.In the example of formula (3),
Second correlation behavior data are pctr.
In frame 502, determine this its based on the ranking factor of optimal object to be recommended and each other object to be recommended
The score range of its object to be recommended.
For example, other object t to be recommended can be updatediThe score value upper boundSo that its ranking factorIt is no more than
Optimal object t to be recommendedkThe score value upper bound
A score value is selected in its score range for each object to be recommended in frame 503.
It can be according to actual needs a score value of its score range of Object Selection to be recommended as final score value.For example,
It is a price in advertisement selection bid ranges as its final bid in the case where object to be recommended is advertisement.
In frame 504, the score value of multiple objects to be recommended is ranked up.
Specifically, the score value of multiple objects to be recommended can be sorted according to sequence from big to small, and according to this
It sorts and pushes multiple object to be recommended to user terminal.
Fig. 6 is the block diagram according to the device for carrying out Object Push to be recommended to user of all aspects of this disclosure.
As shown in fig. 6, for including score range determination/adjustment module to the device that user carries out Object Push to be recommended
601, ranking factor determining module 602, optimal object determining module 603, sorting module 604 and pushing module 605 to be recommended.
Score range determination/adjustment module 601 can be directed to each of this multiple object to be recommended object to be recommended,
The score range of object to be recommended is determined about the first correlation behavior data of the object to be recommended according to user terminal.Into one
Step, it is multiple to be recommended right to adjust that the score value upper bound of optimal object to be recommended can be used in score range determination/adjustment module 601
The score range (for example, adjustment score value upper bound) of other objects to be recommended as in.As retouched above with respect to step 202 and 208
It states.
Ranking factor determining module 602 can be for each of multiple objects to be recommended object to be recommended, according to it
The score value upper bound determines ranking factor.Ranking factor can be the product of third correlation behavior data and score value.Such as above with respect to
Described in step 204.
Optimal object determining module 603 to be recommended can determine that most preferential treatment is pushed away according to the recommendation of multiple objects to be recommended
Recommend object.For example, can be for each of multiple objects to be recommended object to be recommended, according to the score value in its score range
And second correlation behavior data of the user terminal about the object to be recommended, determine the corresponding recommendation of the object to be recommended,
And maximum recommended is worth corresponding object to be recommended and is determined as optimal object to be recommended.As above with respect to described by step 206
's.
The score value upper bound of optimal object to be recommended can be used to adjust its in multiple objects to be recommended in sorting module 604
The score range of its object to be recommended is the value in each its score range of Object Selection to be recommended as end value, and
Multiple objects to be recommended are ranked up using selected score value.As above with respect to described in step 210.
Pushing module 605 can carry out Object Push to be recommended to user.For example, can only push it is optimal to be recommended right
As.Alternatively, the object of top N can also be pushed to user terminal according to the sequence of object.As retouched above with respect to step 212
It states.
Claim can be implemented or fall in without representing by describing example arrangement herein in conjunction with the explanation that attached drawing illustrates
In the range of all examples.Term as used herein " exemplary " means " being used as example, example or explanation ", and simultaneously unexpectedly
Refer to " being better than " or " surpassing other examples ".This detailed description includes detail to provide the understanding to described technology.So
And these technologies can be practiced without these specific details.In some instances, it well-known structure and sets
It is standby to be shown in block diagram form to avoid fuzzy described exemplary concept.
In the accompanying drawings, similar assembly or feature can appended drawing references having the same.In addition, the various components of same type can
It is distinguish by the second label distinguished followed by dash line and between similar assembly in appended drawing reference.If
The first appended drawing reference is used only in the description, then the description can be applied to the similar assembly of the first appended drawing reference having the same
Any one of component regardless of the second appended drawing reference how.
It can be described herein with being designed to carry out in conjunction with the various illustrative frames and module of open description herein
The general processor of function, DSP, ASIC, FPGA or other programmable logic device, discrete door or transistor logic, point
Vertical hardware component, or any combination thereof realize or execute.General processor can be microprocessor, but in alternative
In, processor can be any conventional processor, controller, microcontroller or state machine.Processor can also be implemented as counting
The combination of equipment is calculated (for example, DSP and the combination of microprocessor, multi-microprocessor, the one or more cooperateed with DSP core
Microprocessor or any other such configuration).
Function described herein can hardware, the software executed by processor, firmware, or any combination thereof in it is real
It is existing.If realized in the software executed by processor, each function can be used as one or more instruction or code is stored in
It is transmitted on computer-readable medium or by it.Other examples and realization fall in the disclosure and scope of the appended claims
It is interior.For example, function described above can be used the software executed by processor, hardware, firmware, connect firmly due to the essence of software
Line or any combination thereof is realized.It realizes that the feature of function can also be physically located in various positions, including is distributed so that function
Each section of energy is realized in different physical locations.In addition, being arranged as used in (including in claim) herein in project
It lifts and is used in (for example, being enumerated with the project with the wording of such as one or more of at least one of " " or " " etc)
"or" instruction inclusive enumerate so that such as at least one of A, B or C enumerate mean A or B or C or AB or AC or
BC or ABC (that is, A and B and C).Equally, as it is used herein, phrase " being based on " is not to be read as citation sealing condition collection.
Illustrative steps for example, be described as " based on condition A " can model based on both condition A and condition B without departing from the disclosure
It encloses.In other words, as it is used herein, phrase " being based on " should be solved in a manner of identical with phrase " being based at least partially on "
It reads.
Computer-readable medium includes both non-transitory, computer storage medium and communication media comprising facilitates computer
Any medium that program shifts from one place to another.Non-transitory storage media, which can be, to be accessed by a general purpose or special purpose computer
Any usable medium.Non-limiting as example, non-transient computer-readable media may include that RAM, ROM, electric erasable can
Program read-only memory (EEPROM), compact disk (CD) ROM or other optical disc storages, disk storage or other magnetic storage apparatus,
Or it can be used to carry or store instruction or the expectation program code means of data structure form and can be by general or specialized calculating
Machine or any other non-transitory media of general or specialized processor access.Any connection is also properly termed computer
Readable medium.For example, if software is using coaxial cable, fiber optic cables, twisted pair, digital subscriber line (DSL) or such as red
Outside, the wireless technology of radio and microwave etc is transmitted from web site, server or other remote sources, then should
Coaxial cable, fiber optic cables, twisted pair, digital subscriber line (DSL) or such as infrared, radio and microwave etc it is wireless
Technology is just included among the definition of medium.As used herein disk (disk) and dish (disc) include CD, laser disc, light
Dish, digital universal dish (DVD), floppy disk and blu-ray disc, which disk usually magnetically reproduce data and dish with laser come optically again
Existing data.Combination of the above media is also included in the range of computer-readable medium.
There is provided description herein is in order to enable those skilled in the art can make or use the disclosure.To the disclosure
Various modifications will be apparent those skilled in the art, and the generic principles being defined herein can be applied to it
He deforms without departing from the scope of the present disclosure.The disclosure is not defined to examples described herein and design as a result, and
It is that the widest scope consistent with principles disclosed herein and novel feature should be awarded.