CN109446431A - For the method, apparatus of information recommendation, medium and calculate equipment - Google Patents
For the method, apparatus of information recommendation, medium and calculate equipment Download PDFInfo
- Publication number
- CN109446431A CN109446431A CN201811507489.6A CN201811507489A CN109446431A CN 109446431 A CN109446431 A CN 109446431A CN 201811507489 A CN201811507489 A CN 201811507489A CN 109446431 A CN109446431 A CN 109446431A
- Authority
- CN
- China
- Prior art keywords
- information
- user
- model
- probability
- behavior
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Embodiments of the present invention provide a kind of method for information recommendation.The method includes the probability of the first information is clicked by first the first user of model prediction, it obtains clicking probability, the probability of the first information is not clicked by the first user described in the second model prediction, do not clicked probability, and when the click probability does not click probability greater than described in, using the first information as the alternate information recommended to first user.Wherein, first model is different from second model.The alternate information for information recommendation that method of the invention obtains screening is more targeted, can reduce the uncontrollability of empirically determined recommendation results, effectively improves the conversion ratio of recommendation results.In addition, embodiments of the present invention provide it is a kind of for the device of information recommendation, medium and calculate equipment.
Description
Technical field
Embodiments of the present invention are related to internet area, are used for more specifically, embodiments of the present invention are related to one kind
Method, apparatus, medium and the calculating equipment of information recommendation.
Background technique
Background that this section is intended to provide an explanation of the embodiments of the present invention set forth in the claims or context.Herein
Description recognizes it is the prior art not because not being included in this section.
It the appearance of internet and popularizes and brings a large amount of information to user, meet user in the information age to information
Demand, but the increasing substantially for bring network information amount with the rapid development of network, so that user is facing bulk information
Shi Wufa therefrom obtains the part information actually useful to oneself, reduces instead to the service efficiency of information, here it is institutes
Information overload (information overload) problem of meaning.Currently, for a kind of very potential of problem of information overload
Method be that information recommendation is carried out by personalized recommendation system.Personalized recommendation system can according to the information requirement of user,
Interest etc., by recommendation of personalized information such as the interested information of user, products to user.One good recommender system can not only be
User provides personalized service, moreover it is possible to establish substantial connection between user, user is allowed to generate dependence to recommendation.
The process that recommender system carries out information recommendation generally comprises two stages of recalling and sort.Recall main completion target
The preliminary screening of set limits recommendable information aggregate in a certain range.Sequence is main to be completed to the essence for recalling result
Refinement screening, for exporting more accurate recommendation results.
In the prior art, information sieve either is carried out in the stage of recalling or phase sorting (especially in the stage of recalling)
A kind of scheme when selecting is the click probability for predicting a certain information, and the click probability compares with fixed threshold then, obtains
To the selection result.The shortcomings that this scheme is that this fixed threshold is determined often by artificial priori, lacks rationale,
Final computational accuracy is caused to be very easy to be affected by this.Another scheme is that probability progress is clicked in the prediction to bulk information
Sequence, N information before then choosing.However, in this scheme the determination of N value be still by priori formulate, there is no from
Fundamentally solve the problems, such as that the recommendation results as caused by priori are uncontrollable.
Summary of the invention
Therefore in the prior art, when carrying out information recommendation, filter information, is to make us very much how more scientific and reasonablely
Worried process.
Thus, it is also very desirable to a kind of more scientific reasonable Information Recommendation Mechanism, so that having more for information recommendation is directed to
Property.
In the present context, embodiments of the present invention are intended to provide a kind of method, apparatus for information recommendation, medium
With calculating equipment.
In the first aspect of embodiment of the present invention, a kind of method for information recommendation is provided.The method packet
It includes: clicking the probability of the first information by first the first user of model prediction, obtain clicking probability；Pass through the second model prediction institute
The probability that the first user does not click the first information is stated, probability is not clicked；And it is greater than in the click probability described
When not clicking probability, using the first information as the alternate information recommended to first user.Wherein, first model
It is different from second model.
In one embodiment of the invention, first model is the mould for predicting the behavior of user's click information
Type and second model are the model for predicting user's not behavior of click information.
In another embodiment of the present invention, the method also includes by machine learning obtain first model and
Second model.It specifically includes through first kind characteristic training first model, and passes through the second category feature number
According to training second model.Wherein, the first kind characteristic includes the interest tags of each user and characterizes this often
The data of the browsing behavior of a user, the second category feature data include the label and described of loseing interest in of each user
Characterize the data of the browsing behavior of each user.In one embodiment of the invention, the browsing behavior includes following
Anticipate one or more behaviors: the behavior of click information, the behavior for browsing title but not clicking on information, takes the behavior of evaluation information
Disappear the behavior for paying close attention to information and the behavior for forcibly closing information.
In yet another embodiment of the present invention, the method also includes: obtain initial data, the initial data includes
In specific time in predetermined network platform multiple users data relevant to the browsing behavior；And from the initial data
The middle extraction first kind characteristic and the second category feature data.
In the second aspect of embodiment of the present invention, a kind of device for information recommendation is provided.Described device packet
It includes and clicks probabilistic forecasting module, do not click probabilistic forecasting module and decision-making module.Probabilistic forecasting module is clicked to be used for by the
One the first user of model prediction clicks the probability of the first information, obtains clicking probability.Probabilistic forecasting module is not clicked for passing through
First user described in second model prediction does not click the probability of the first information, is not clicked probability.Decision-making module is used for
It is when not clicking probability described in being greater than in the click probability, the first information is alternative as recommending to first user
Information.Wherein, first model is different from second model.
In one embodiment of the invention, first model is the mould for predicting the behavior of user's click information
Type and second model are the model for predicting user's not behavior of click information.
In another embodiment of the present invention, described device further includes that model obtains module.Model obtains module and is used for
First model and second model are obtained by machine learning.The model obtain module include first training submodule with
And second training submodule.First training submodule is used for through first kind characteristic training first model.Second instruction
Practice submodule to be used for through the second category feature data training second model.Wherein, the first kind characteristic includes every
The interest tags of a user and characterize each user browsing behavior data, the second category feature data include every
The data of lose interest in label and the browsing behavior for characterizing each user of a user.In a reality of the invention
Apply in example, the browsing behavior includes any of the following or a variety of behaviors: the behavior of click information, the behavior of evaluation information,
Browsing title but the behavior for not clicking on the behavior of information, cancelling the behavior for paying close attention to information and positive closing information.
In yet another embodiment of the present invention, described device further includes that initial data obtains module and characteristic extraction
Module.Initial data obtains module for obtaining initial data, and the initial data includes predetermined network platform in specific time
In multiple users data characteristics data extraction module relevant to the browsing behavior for being extracted from the initial data
The first kind characteristic and the second category feature data.
In the third party of embodiment of the present invention, a kind of computer readable storage medium is provided, being stored thereon with can
It executes instruction, described instruction makes processor execute the method for being used for information recommendation as described above when being executed by processor.
In the fourth aspect of embodiment of the present invention, a kind of calculating equipment is provided.The calculating equipment includes storage
There are the one or more memories and one or more processors of executable instruction.The processor executes described executable
Instruction, to realize the method as described above for being used for information recommendation.
The method, apparatus of embodiment, medium and calculating equipment according to the present invention, in screening for the alternative of information recommendation
It is more scientific when information, more targetedly, it can reduce the uncontrollability rule of thumb screened in the prior art, Neng Gouyou
Effect improves the conversion ratio of recommendation information.
Detailed description of the invention
The following detailed description is read with reference to the accompanying drawings, above-mentioned and other mesh of exemplary embodiment of the invention
, feature and advantage will become prone to understand.In the accompanying drawings, if showing by way of example rather than limitation of the invention
Dry embodiment, in which:
Fig. 1 schematically show embodiment according to the present invention by the method, apparatus of information recommendation, medium and based on
Calculate the application scenarios of equipment；
Fig. 2 schematically shows the method flow diagrams according to an embodiment of the invention for information recommendation；
Fig. 3 schematically shows the method flow diagram according to another embodiment of the present invention for information recommendation；
First model and the second mould are obtained by machine learning Fig. 4 schematically shows according to an embodiment of the invention
The flow chart of type；
Fig. 5 schematically shows the design schematic diagram according to an embodiment of the present invention for information recommendation method；
Fig. 6 schematically shows the block diagram of the device according to an embodiment of the present invention for information recommendation；
Fig. 7 schematically shows the schematic diagram of the program product according to an embodiment of the present invention for information recommendation；With
And
Fig. 8 schematically shows the schematic diagram of the calculating equipment according to an embodiment of the present invention for information recommendation.
In the accompanying drawings, identical or corresponding label indicates identical or corresponding part.
Specific embodiment
The principle and spirit of the invention are described below with reference to several illustrative embodiments.It should be appreciated that providing this
A little embodiments are used for the purpose of making those skilled in the art can better understand that realizing the present invention in turn, and be not with any
Mode limits the scope of the invention.On the contrary, these embodiments are provided so that this disclosure will be more thorough and complete, and energy
It is enough that the scope of the present disclosure is completely communicated to those skilled in the art.
One skilled in the art will appreciate that embodiments of the present invention can be implemented as a kind of system, device, equipment, method
Or computer program product.Therefore, the present disclosure may be embodied in the following forms, it may be assumed that complete hardware, complete software
The form that (including firmware, resident software, microcode etc.) or hardware and software combine.
Embodiment according to the present invention is proposed and a kind of is set for the method for information recommendation, medium, device and calculating
It is standby.
Herein, any number of elements in attached drawing is used to example rather than limitation and any name are only used for
It distinguishes, without any restrictions meaning.
Below with reference to several representative embodiments of the invention, the principle and spirit of the present invention are explained in detail.
Summary of the invention
The inventors discovered that can just do this part thing when the power that people usually only regard something is greater than resistance
Feelings.Same reason can consider same information to be only greater than in the hope that user clicks the information and not click the information
Hope when, user just can really click this information.Thus inventor associates in information recommendation field, if user clicks
When the probability of one information does not click the probability of the information greater than user, so that it may think that user can click the information.According to this
One logic, for an information, if it is possible to which prediction obtains that user clicks the click probability of the information and user does not click the letter
Breath does not click probability, does not then click according to the click probability and the comparison of probability and is screened, it will be able to so that screening
The specific aim of the information for recommendation out is stronger, conversion ratio is higher.
After introduced the basic principles of the present invention, lower mask body introduces various non-limiting embodiment party of the invention
Formula.
Application scenarios overview
Referring initially to Fig. 1.
Fig. 1 schematically show embodiment according to the present invention by the method, apparatus of information recommendation, medium and based on
Calculate the application scenarios of equipment.
As shown in Figure 1, the application scenarios include user terminal 11 and server 12.User terminal 11 and server 12 pass through
Network connection.The network can be local area network, wide area network, mobile Internet etc., may include various connection types, such as have
Line, wireless communication link etc..
User terminal 11 can be but not limited to portable equipment (such as: mobile phone, plate, laptop etc.), can also be with
For PC (PC, Personal Computer).User can be used user terminal 11 and be handed over by network and server 12
Mutually, to receive or send message etc..Various telecommunication customer end applications can be installed, such as shopping class is answered on user terminal 11
(only shown with, web browser applications, news category application, instant messaging tools, mailbox client, social platform software etc.
Example).
Server 12 can be to provide the server of various services, such as to the net that user utilizes user terminal 11 to be browsed
It stands and the back-stage management server (merely illustrative) of support is provided.Back-stage management server can be to numbers such as the user's requests received
According to carrying out the processing such as analyzing, and processing result is fed back into user terminal.For example, according to an embodiment of the invention, server 12
Recommendation of personalized information can be carried out to user terminal 11.
It should be noted that the method provided by the embodiment of the present invention for information recommendation can be held by server 12
Row.Correspondingly, provided by the embodiment of the present invention for the device of information recommendation, medium or calculate equipment generally can be set in
In server 12.Alternatively, can also be by being different from server 12 for the method for information recommendation provided by the embodiment of the present invention
And other Website servers or Website server cluster that can be communicated with user terminal 11 and/or server 12 execute.Accordingly
Ground, the embodiment of the present invention is provided also to be can be set for the device of information recommendation, medium or calculating equipment in different from clothes
In business device 12 and other Website servers or Website server cluster that can be communicated with user terminal 11 and/or server 12.
It should be understood that the number of user terminal, network and Website server in Fig. 1 is only schematical.According to reality
It now needs, can have any number of user terminal, network and server.
Illustrative methods
Below with reference to the application scenarios of Fig. 1, the use of illustrative embodiments according to the present invention is described with reference to Fig. 2~Fig. 5
In the method for information recommendation.It should be noted that above-mentioned application scenarios are merely for convenience of understanding spirit and principles of the present invention
And show, embodiments of the present invention are not limited in this respect.On the contrary, embodiments of the present invention can be applied to fit
Any scene.
Fig. 2 schematically shows the method flow diagrams according to an embodiment of the invention for information recommendation.
As shown in Fig. 2, the method for being used for information recommendation may include operation S201~operation S203.
In operation S201, the probability of the first information is clicked by first the first user of model prediction, obtains clicking probability.
In operation S202, does not click the probability of the first information by the second model prediction first user, obtain not point
Probability is hit, wherein first model is different from the second model.
Then, in operation S203, when the click probability is greater than this and does not click probability, using the first information as to this
The alternate information that one user recommends.
According to an embodiment of the invention, first model is different from the second model, the first model and second specifically can be
Model is two different models, wherein the two can be to be trained by different training datas.According to this hair
Bright embodiment, first model be for predict the behavior of user's click information model and second model be for
Predict the model of user's not behavior of click information.
According to the method for the embodiment of the present invention, in the prior art (such as empirically determined consolidated according to priori
Determine the top n of threshold value or sequence) method to screen recommendation information compares, can reduce using priori knowledge bring result
Uncontrollability, improve the specific aim and validity of information recommendation.
Fig. 3 schematically shows the method flow diagram according to another embodiment of the present invention for information recommendation.
As shown in figure 3, according to another embodiment of the present invention, this is used for the method for information recommendation in addition to operating S201~behaviour
Make other than S203, can also include operation S301~S303.
Firstly, obtaining initial data in operation S301.The initial data includes more in predetermined network platform in specific time
The data relevant to browsing behavior of a user.
According to embodiments of the present invention, which includes any of the following or a variety of behaviors: the behavior of click information,
The behavior of evaluation information, the behavior for browsing title but not clicking on information, the behavior for cancelling concern information and positive closing information
Behavior etc..
Table 1 diagrammatically illustrates an example of initial data according to an embodiment of the present invention.
Table 1
As shown in table 1, which for example can be collects arrangement in specific time from predetermined network platform
The relevant data of browsing behavior of a large number of users (for example, the users such as U1, U2, U3, U4).The initial data may include user's
Essential information and the browsing behavior data of user.
Reference table 1, the essential information of user for example may include User ID, age, the interest tags of user and do not feel emerging
Interesting label.In some embodiments, the essential information of user for example may include the gender of user, log in the predetermined network platform
When the information such as geographical location, temporal information, and/or the equipment that uses.
According to an embodiment of the invention, the interest tags of user, such as can be user (for example, user U1) in the network
The interest tags for selecting or filling in the personal information of platform are also possible to count the emerging of acquisition according to the browsing behavior of user U1
Interesting label.Specifically, according to the interest tags that the browsing behavior of user U1 statistics obtains, such as can be to user U1 at this
The frequency of occurrence or frequency that label included in favorable comment or the information of collection is clicked or carried out in specific period are counted,
Using several highest labels of frequency of occurrence or frequency as the interest tags of user U1.
According to an embodiment of the invention, the label of loseing interest in of user, such as can be user (for example, user U1) at this
The label of loseing interest in for selecting or filling in the personal information of the network platform is also possible to be counted according to the browsing behavior of user U1
The label of loseing interest in obtained.Specifically, according to the label of loseing interest in that the browsing behavior of user U1 statistics obtains, such as can
To be the information of complaint to user U1 within the specific period or give the information of information or positive closing that difference is commented
Included in label frequency of occurrence counted, and will wherein frequency of occurrence or several highest labels of frequency if with
The label of loseing interest in of family U1.
With continued reference to table 1, the browsing behavior data of user may include such as user to which article carried out click,
And the article of the browser interface of user is exposed to which and is not clicked on.In some embodiments, user's as shown in table 1 is clear
Behavioral data of looking at can also include other data, such as user carried out evaluation and evaluation content to which article, to which
Article has carried out collection, to which article carries out cancelling concern, and/or forcibly closes etc. to which article.It is every in table 1
A article ID, which is used for the corresponding article content of unique identification, each article content in the predetermined network platform, can have at least
The label of one or more reflection this article content.
Then in operation S302, the first kind characteristic and the second category feature data are extracted from the initial data.
Such as specifically can be first by initial data standardization and structuring handle, then therefrom extract the first kind characteristic and
The second category feature data.
According to an embodiment of the invention, the first kind characteristic includes that the interest tags of each user and characterization are somebody's turn to do
The data of the browsing behavior of each user.The second category feature data include lose interest in label and table of each user
Levy the data of the browsing behavior of each user.Other embodiments according to the present invention, the browsing row of each user of the characterization
For data, such as can be the corresponding point of the included label information of information involved in the browsing behavior of user, each label
Hit number, complain number, ignore number, favorable comment number etc. and/or geographical location information when user browses each information, when
Between information etc..
Specifically, by taking the initial data of table 1 as an example.The first kind characteristic can be the initial data shown in the table 1
In extract the data of " User ID ", " age ", " interest tags ", " article ID " and corresponding " browsing behavior ", and carry out
Structuring and standardization processing obtain.Feature in the first kind characteristic, it is intended to describe what user clicked information
Reason.
The second category feature data can be extracts " User ID ", " age ", " no from initial data shown in table 1
Label interested ", " article ID " and corresponding " browsing behavior " data, and carry out structuring and standardization processing obtains.It should
Feature in second category feature data, it is intended to describe user to information without clicking the reason of.
Then, in operation S303, first model and second model are obtained by machine learning.Wherein, with the present invention
Embodiment, operate S303 specific implementation can refer to Fig. 4 signal.
Fig. 4 schematically shows obtain the first mould by machine learning in operation S303 according to an embodiment of the invention
The flow chart of type and the second model.
As shown in figure 4, according to an embodiment of the invention, operation S303 may include operation S413 and operation S423.
Wherein, in operation S413, pass through first kind characteristic training first model.And in operation S423, pass through
Second category feature data train second model.First model obtained after training can be used for predicting user's click information
Behavior, the second model can be used for predicting the behavior of user's not click information.Hereafter, so that it may use first model and second
Model carries out user and clicks behavior or do not click the prediction of behavior.
By this method, according to an embodiment of the invention, predicting first user's point respectively by the first model and the second model
What the click probability and the first user for hitting the first information did not clicked the first information does not click probability, then in the click probability
When greater than not clicking probability, it is believed that user can click the first information, thus using the first information as recommend alternate information,
Effectively improve the specific aim and validity of information recommendation.
The present invention is not limited the first model and the second model, any to predict that the first user clicks the first information
Click probability and the first user do not click the model for not clicking probability of the first information and be suitable for the present invention.For example, the
One model and the second model can be logistics regression model, but the characteristic used is different.
Fig. 5 schematically shows the design schematic diagram according to an embodiment of the present invention for information recommendation method.
As shown in figure 5, the method according to an embodiment of the present invention for information recommendation, it is intended to for same information, prediction
Obtain that user clicks the click probability of the information and what user did not clicked the information do not click probability, then according to the click probability
It does not click the comparison of probability to decide whether to retain the information to be recommended, avoids leading by empirical decision making in the prior art
Write the uncontrollability of breath recommendation results.
Assuming that there is bulk information in database, to determine for different users to each user and recommend those information, no
Which information recommended.An arbitrary user is referred to the first user, and appointing in the bulk information is referred to the first information
It anticipates an information.
With reference to the signal of Fig. 5, the learning objective of first model be the click behavior of each user is modeled, thus
It can be used to predict the click probability that each user (for example, first user) clicks the first information.
First model mainly constructs process, and reading initial data, the initial data are gone through from a large number of users first
History browsing behavior data.Then, first kind characteristic is extracted.The feature of first kind characteristic is intended to describe the network platform
In each user the reason of information is clicked, wherein the particular content of the first kind characteristic is as previously described.Finally,
By the first kind characteristic training the first model, so as to so that the first model learning to each user click behavior.
Then in forecast period, for the first user, it can predict to obtain first user by the first model and click the
The click probability of one information.
With continued reference to Fig. 5, the learning objective of second model be the behavior of not clicking of each user is modeled, thus
It can be used to predict that the first user do not clicked the first information does not click probability.
The main building process of second model is to read the initial data first.Then, is always extracted from initial data
Two category feature data.The feature of the second category feature data is intended to describe what each user in the network platform did not clicked information
Reason, wherein the particular content of the second category feature data is as previously described.Finally, passing through the second category feature data training second
Model, so as to so that the second model learning does not click behavior to each user.
Then in forecast period, for the first user, it can predict that obtaining first user does not click by the second model
The first information does not click probability.
With continued reference to Fig. 5, finally, first user compares the click probability of the first information with probability is not clicked
Compared with when click probability, which is greater than, does not click probability using the first information as recommendation alternate information.Theoretical foundation herein is, when
When the probability that first user clicks the first information is greater than its probability without click, it is believed that user just can be to one
A sample is really clicked.
And so on, for each of information to be screened a large amount of in database, all in accordance with Fig. 2~Fig. 5 signal
Method, which obtains, to be clicked probability and does not click probability, and retains the information when clicking probability and being greater than and do not click probability, so as to
It is effectively reduced and carries out screening bring uncontrollability using priori knowledge, improve the specific aim and validity of information recommendation.
Exemplary means
After describing the method for exemplary embodiment of the invention, next, with reference to Fig. 6 to the exemplary reality of the present invention
The device for information recommendation for applying mode is introduced.
Fig. 6 schematically shows the block diagram of the device 600 according to an embodiment of the present invention for information recommendation.
As shown in fig. 6, the device 600 for being used for information recommendation includes clicking that probabilistic forecasting module 610, not click probability pre-
Survey module 620 and decision-making module 630.
The probability that probabilistic forecasting module 610 is used to click the first information by first the first user of model prediction is clicked, is obtained
To click probability (operation S201)).Probabilistic forecasting module 620 is not clicked for by the second model prediction first user not point
The probability for hitting the first information, is not clicked probability, wherein first model is different from the second model (operation S202)).Certainly
Plan module is used for when the click probability is greater than this and does not click probability, and the first information is standby as recommending to first user
Select information (operation S203)).
According to one embodiment of present invention, which is the model for predicting the behavior of user's click information,
And second model is the model for predicting user's not behavior of click information.
According to another embodiment of the invention, which further includes that model obtains module 640.Model obtains module
640 for obtaining first model and second model (operation S303) by machine learning.
The model, which obtains module 640, can specifically include the first training submodule 641 and the second training submodule 642.
First training submodule 641 is used for through first kind characteristic training first model (operation S413).Second training submodule
Block 642 is used for through the second category feature data training second model (operation S423).Wherein, which includes
The interest tags of each user and characterize each user browsing behavior data, which includes every
The data of the browsing behavior of the label and the characterization each user of loseing interest in of a user.A reality according to the present invention
Example is applied, which includes any of the following or a variety of behaviors: the behavior of click information, the behavior of evaluation information, browsing
Title but the behavior for not clicking on the behavior of information, cancelling the behavior for paying close attention to information and positive closing information.
According to still another embodiment of the invention, which further includes that initial data obtains module 650 and characteristic
Extraction module 660.Initial data obtains module 650 for obtaining initial data, which includes making a reservation in specific time
The data relevant to the browsing behavior (operation S301) of multiple users in the network platform.Characteristic extraction module 660 is used for
The first kind characteristic and the second category feature data (operation S302) are extracted from the initial data.
Implementation according to the present invention, the device 600 for being used for information recommendation can be used to implement as described above according to this hair
The method for information recommendation of bright embodiment, improves the specific aim and validity of information recommendation, and detailed content can refer to Fig. 2
The description of~Fig. 5, details are not described herein.
Exemplary media
After describing the method and apparatus of exemplary embodiment of the invention, next, showing with reference to Fig. 7 the present invention
The program product (can also be referred to as computer readable storage medium) for information recommendation of example property embodiment is illustrated.
According to an embodiment of the invention, additionally providing a kind of computer readable storage medium, it is stored thereon with executable finger
It enables, described instruction makes processor execute the method according to an embodiment of the present invention for information recommendation when being executed by processor.
In some possible embodiments, various aspects of the invention are also implemented as a kind of shape of program product
Formula comprising program code, when described program product is run on the computing device, said program code is for making the calculating
Equipment executes described in above-mentioned " illustrative methods " part of this specification the use of various illustrative embodiments according to the present invention
Step in the method for information recommendation, for example, the calculating equipment can execute operation S201 as shown in Figure 2: passing through
First the first user of model prediction clicks the probability of the first information, obtains clicking probability；Operation S202: pass through the second model prediction
First user does not click the probability of the first information, is not clicked probability；And operation S203: it is clicked generally described
When rate does not click probability greater than described in, using the first information as the alternate information recommended to first user；Wherein, institute
It is different from second model to state the first model.
Described program product can be using any combination of one or more readable mediums.Readable medium can be readable letter
Number medium or readable storage medium storing program for executing.Readable storage medium storing program for executing for example may be-but not limited to-electricity, magnetic, optical, electromagnetic, red
The system of outside line or semiconductor, device or device, or any above combination.The more specific example of readable storage medium storing program for executing
(non exhaustive list) includes: the electrical connection with one or more conducting wires, portable disc, hard disk, random access memory
(RAM), read-only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc
Read memory (CD-ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
As shown in fig. 7, the program product 700 for information recommendation of embodiment according to the present invention is described, it can
To use portable compact disc read only memory (CD-ROM) and including program code, and equipment can be being calculated, such as personal
It is run on computer.However, program product of the invention is without being limited thereto, in this document, readable storage medium storing program for executing can be any packet
Contain or store the tangible medium of program, which can be commanded execution system, device or device use or in connection
It uses.
Readable signal medium may include in a base band or as the data-signal that carrier wave a part is propagated, wherein carrying
Readable program code.The data-signal of this propagation can take various forms, including --- but being not limited to --- electromagnetism letter
Number, optical signal or above-mentioned any appropriate combination.Readable signal medium can also be other than readable storage medium storing program for executing it is any can
Read medium, the readable medium can send, propagate or transmit for by instruction execution system, device or device use or
Program in connection.
The program code for including on readable medium can transmit with any suitable medium, including --- but being not limited to ---
Wirelessly, wired, optical cable, RF etc. or above-mentioned any appropriate combination.
The program for executing operation of the present invention can be write with any combination of one or more programming languages
Code, described program design language include object oriented program language --- and such as Java, C++ etc. further include routine
Procedural programming language --- such as " C ", language or similar programming language.Program code can fully exist
It is executed in user calculating equipment, part executes on a remote computing or completely remote on the user computing device for part
Journey calculates to be executed on equipment or server.In the situation for being related to remote computing device, remote computing device can be by any
The network of type --- it is connected to user calculating equipment including local area network (LAN) or wide area network (WAN) one, alternatively, can connect
To external computing device (such as being connected using ISP by internet).
Exemplary computer device
After the method, apparatus and medium for describing exemplary embodiment of the invention, next, with reference to Fig. 8 to this
The calculating equipment for information recommendation of invention illustrative embodiments is illustrated.
According to an embodiment of the invention, additionally providing a kind of calculating equipment.The calculating equipment is executable including being stored with
The one or more memories and one or more processors of instruction.The processor executes the executable instruction, to
Realize the method according to an embodiment of the present invention for information recommendation.
The embodiment of the invention also provides a kind of calculating equipment.Person of ordinary skill in the field is it is understood that this hair
Bright various aspects can be implemented as system, method or program product.Therefore, various aspects of the invention can be implemented as
Following form, it may be assumed that complete hardware embodiment, complete Software Implementation (including firmware, microcode etc.) or hardware and
The embodiment that software aspects combine, may be collectively referred to as circuit, " module " or " system " here.
In some possible embodiments, it is single can to include at least at least one processing for calculating equipment according to the present invention
Member and at least one storage unit.Wherein, the storage unit is stored with program code, when said program code is described
When processing unit executes, so that the processing unit executes described in above-mentioned " illustrative methods " part of this specification according to this
Invent the operation for information recommendation of various illustrative embodiments.For example, the processing unit can be executed such as institute in Fig. 2
The operation S201 shown: clicking the probability of the first information by first the first user of model prediction, obtains clicking probability；Operation
S202: it does not click the probability of the first information by the first user described in the second model prediction, is not clicked probability；And
Operation S203: when not clicking probability described in being greater than in the clicks probability, using the first information as to first user
The alternate information of recommendation；Wherein, first model is different from second model.
The calculating equipment 800 for information recommendation of this embodiment according to the present invention is described referring to Fig. 8.
Calculating equipment 800 as shown in Figure 8 is only an example, should not function to the embodiment of the present invention and use scope bring and appoint
What is limited.
It is showed in the form of universal computing device as shown in figure 8, calculating equipment 800.The component for calculating equipment 80 can wrap
It includes but is not limited to: at least one above-mentioned processing unit 810, at least one above-mentioned storage unit 820, the different system components of connection
The bus 830 of (including storage unit 820 and processing unit 810).
Bus 830 includes data/address bus, control bus and address bus.
Storage unit 820 may include volatile memory, such as random access memory (RAM) 821 and/or high speed are delayed
Memory 822 is deposited, can further include read-only memory (ROM) 823.
Storage unit 820 can also include program/utility 825 with one group of (at least one) program module 824,
Such program module 824 includes but is not limited to: operating system, one or more application program, other program modules and
It may include the realization of network environment in program data, each of these examples or certain combination.
Calculating equipment 800 can also be with one or more external equipments 840 (such as keyboard, sensing equipment, blue lance equipment
Deng) communicate, this communication can be carried out by input/output (I/0) interface 850.Also, calculating equipment 800 can also pass through
Network adapter 860 and one or more network (such as local area network (LAN), wide area network (WAN) and/or public network, such as
Internet) communication.As shown, network adapter 860 is communicated by bus 830 with the other modules for calculating equipment 800.It should
Understand, although not shown in the drawings, other hardware and/or software module can be used in conjunction with equipment 800 is calculated, including but unlimited
In: microcode, device driver, redundant processing unit, external disk drive array, RAID system, tape drive and number
According to backup storage system etc..
It should be noted that although being referred to several units/modules or subelement/module of device in the above detailed description,
But it is this division be only exemplary it is not enforceable.In fact, embodiment according to the present invention, above-described
The feature and function of two or more units/modules can embody in a units/modules.Conversely, above-described one
The feature and function of a units/modules can be to be embodied by multiple units/modules with further division.
In addition, although describing the operation of the method for the present invention in the accompanying drawings with particular order, this do not require that or
Hint must execute these operations in this particular order, or have to carry out shown in whole operation be just able to achieve it is desired
As a result.Additionally or alternatively, it is convenient to omit multiple steps are merged into a step and executed by certain steps, and/or by one
Step is decomposed into execution of multiple steps.
Although detailed description of the preferred embodimentsthe spirit and principles of the present invention are described by reference to several, it should be appreciated that, this
It is not limited to the specific embodiments disclosed for invention, does not also mean that the feature in these aspects cannot to the division of various aspects
Combination is benefited to carry out, this to divide the convenience merely to statement.The present invention is directed to cover appended claims spirit and
Included various modifications and equivalent arrangements in range.
Claims (10)
1. a kind of method for information recommendation, comprising:
The probability that the first information is clicked by first the first user of model prediction obtains clicking probability；
The probability for not clicking the first information by the first user described in the second model prediction, is not clicked probability；And
The click probability be greater than it is described do not click probability when, using the first information as recommending to first user
Alternate information；
Wherein:
First model is different from second model.
2. according to the method described in claim 1, wherein:
First model is the model for predicting the behavior of user's click information；And
Second model is the model for predicting user's not behavior of click information.
3. according to the method described in claim 1, further include:
First model and second model are obtained by machine learning, comprising:
Pass through first kind characteristic training first model；And
Pass through the second category feature data training second model；
Wherein:
The first kind characteristic includes the number of the interest tags of each user and the browsing behavior for characterizing each user
According to；
The second category feature data include lose interest in label and the browsing for characterizing each user of each user
The data of behavior.
4. according to the method described in claim 3, wherein, the browsing behavior includes any of the following or a variety of behaviors:
The behavior of click information, the behavior of evaluation information, the behavior for browsing title but not clicking on information, the row for cancelling concern information
For and forcibly close information behavior.
5. according to the method described in claim 4, further include:
Obtain initial data, the initial data include in specific time in predetermined network platform multiple users with the browsing
The relevant data of behavior；And
The first kind characteristic and the second category feature data are extracted from the initial data.
6. a kind of device for information recommendation, comprising:
Probabilistic forecasting module is clicked, for clicking the probability of the first information by first the first user of model prediction, is clicked
Probability；
Probabilistic forecasting module is not clicked, for not clicking the general of the first information by the first user described in the second model prediction
Rate is not clicked probability；And
Decision-making module, for the click probability be greater than it is described do not click probability when, using the first information as to described
The alternate information that first user recommends；
Wherein:
First model is different from second model.
7. device according to claim 6, further includes:
Model obtains module, for obtaining first model and second model by machine learning, comprising:
First training submodule, for passing through first kind characteristic training first model；And
Second training submodule, for passing through the second category feature data training second model；
Wherein:
The first kind characteristic includes the number of the interest tags of each user and the browsing behavior for characterizing each user
According to；
The second category feature data include lose interest in label and the browsing for characterizing each user of each user
The data of behavior.
8. device according to claim 7, further includes:
Initial data obtains module, and for obtaining initial data, the initial data includes predetermined network platform in specific time
In multiple users data relevant to the browsing behavior；And
Characteristic extraction module, it is special for extracting the first kind characteristic and second class from the initial data
Levy data.
9. a kind of computer readable storage medium, is stored thereon with executable instruction, described instruction makes when being executed by processor
Processor executes method described in any one according to claim 1~5.
10. a kind of calculating equipment, comprising:
One or more memories, are stored with executable instruction；
One or more processors execute the executable instruction, to realize described in any one according to claim 1~5
Method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811507489.6A CN109446431A (en) | 2018-12-10 | 2018-12-10 | For the method, apparatus of information recommendation, medium and calculate equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811507489.6A CN109446431A (en) | 2018-12-10 | 2018-12-10 | For the method, apparatus of information recommendation, medium and calculate equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109446431A true CN109446431A (en) | 2019-03-08 |
Family
ID=65556602
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811507489.6A Pending CN109446431A (en) | 2018-12-10 | 2018-12-10 | For the method, apparatus of information recommendation, medium and calculate equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446431A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353825A (en) * | 2020-03-10 | 2020-06-30 | 北京爱笔科技有限公司 | Message transmission method and device |
| CN111898019A (en) * | 2019-05-06 | 2020-11-06 | 北京达佳互联信息技术有限公司 | Information pushing method and device |
| CN112446736A (en) * | 2020-12-02 | 2021-03-05 | 平安科技（深圳）有限公司 | Click through rate CTR prediction method and device |
| CN112784151A (en) * | 2019-11-08 | 2021-05-11 | 北京搜狗科技发展有限公司 | Method and related device for determining recommendation information |
| CN113643073A (en) * | 2021-09-01 | 2021-11-12 | 深圳市聚商鼎力网络技术有限公司 | Automatic information delivery system based on big data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090313286A1 (en) * | 2008-06-17 | 2009-12-17 | Microsoft Corporation | Generating training data from click logs |
| CN104090919A (en) * | 2014-06-16 | 2014-10-08 | 华为技术有限公司 | Advertisement recommending method and advertisement recommending server |
| US8938463B1 (en) * | 2007-03-12 | 2015-01-20 | Google Inc. | Modifying search result ranking based on implicit user feedback and a model of presentation bias |
| CN104965890A (en) * | 2015-06-17 | 2015-10-07 | 深圳市腾讯计算机***有限公司 | Advertisement recommendation method and apparatus |
| CN105335491A (en) * | 2015-10-20 | 2016-02-17 | 杭州东信北邮信息技术有限公司 | Method and system for recommending books to users on basis of clicking behavior of users |
-
2018
- 2018-12-10 CN CN201811507489.6A patent/CN109446431A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8938463B1 (en) * | 2007-03-12 | 2015-01-20 | Google Inc. | Modifying search result ranking based on implicit user feedback and a model of presentation bias |
| US20090313286A1 (en) * | 2008-06-17 | 2009-12-17 | Microsoft Corporation | Generating training data from click logs |
| CN104090919A (en) * | 2014-06-16 | 2014-10-08 | 华为技术有限公司 | Advertisement recommending method and advertisement recommending server |
| CN104965890A (en) * | 2015-06-17 | 2015-10-07 | 深圳市腾讯计算机***有限公司 | Advertisement recommendation method and apparatus |
| CN105335491A (en) * | 2015-10-20 | 2016-02-17 | 杭州东信北邮信息技术有限公司 | Method and system for recommending books to users on basis of clicking behavior of users |
Non-Patent Citations (1)
| Title |
|---|
| 黄冬梅等: "《案例驱动的大数据原理技术及应用》", 30 November 2018 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898019A (en) * | 2019-05-06 | 2020-11-06 | 北京达佳互联信息技术有限公司 | Information pushing method and device |
| CN111898019B (en) * | 2019-05-06 | 2024-04-16 | 北京达佳互联信息技术有限公司 | Information pushing method and device |
| CN112784151A (en) * | 2019-11-08 | 2021-05-11 | 北京搜狗科技发展有限公司 | Method and related device for determining recommendation information |
| CN112784151B (en) * | 2019-11-08 | 2024-02-06 | 北京搜狗科技发展有限公司 | Method and related device for determining recommended information |
| CN111353825A (en) * | 2020-03-10 | 2020-06-30 | 北京爱笔科技有限公司 | Message transmission method and device |
| CN111353825B (en) * | 2020-03-10 | 2024-01-12 | 北京爱笔科技有限公司 | Message transmission method and device |
| CN112446736A (en) * | 2020-12-02 | 2021-03-05 | 平安科技（深圳）有限公司 | Click through rate CTR prediction method and device |
| WO2022116431A1 (en) * | 2020-12-02 | 2022-06-09 | 平安科技（深圳）有限公司 | Click through rate (ctr) prediction method and apparatus |
| CN113643073A (en) * | 2021-09-01 | 2021-11-12 | 深圳市聚商鼎力网络技术有限公司 | Automatic information delivery system based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110825957B (en) | Deep learning-based information recommendation method, device, equipment and storage medium | |
| US11531867B2 (en) | User behavior prediction method and apparatus, and behavior prediction model training method and apparatus | |
| CN109446431A (en) | For the method, apparatus of information recommendation, medium and calculate equipment | |
| CN109460514B (en) | Method and device for pushing information | |
| JP6967612B2 (en) | Information retrieval methods, devices and systems | |
| CN101517967B (en) | Traffic prediction for web sites | |
| CN108885624B (en) | Information recommendation system and method | |
| CN112669096B (en) | Object recommendation model training method and device | |
| CN111815415A (en) | Commodity recommendation method, system and equipment | |
| JP6911603B2 (en) | How to generate predictive models for the categories of facilities visited by users, programs, server equipment, and processing equipment | |
| CN104899315A (en) | Method and device for pushing user information | |
| US11257019B2 (en) | Method and system for search provider selection based on performance scores with respect to each search query | |
| CN109189931A (en) | A kind of screening technique and device of object statement | |
| CN108632311A (en) | Information-pushing method and device | |
| CN110033294A (en) | A kind of determination method of business score value, business score value determining device and medium | |
| CN107977678A (en) | Method and apparatus for output information | |
| CN112017059A (en) | Hierarchical optimization risk control method and device and electronic equipment | |
| CN107368499B (en) | Client label modeling and recommending method and device | |
| CN110209921B (en) | Method and device for pushing media resource, storage medium and electronic device | |
| CN112100221A (en) | Information recommendation method and device, recommendation server and storage medium | |
| CN112115710B (en) | Industry information identification method and device | |
| CN112819528A (en) | Crowd pack online method and device and electronic equipment | |
| KR101931624B1 (en) | Trend Analyzing Method for Fassion Field and Storage Medium Having the Same | |
| CN115935185A (en) | Training method and device for recommendation model | |
| CN113378067A (en) | Message recommendation method, device, medium, and program product based on user mining |
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 |