CN110110233A

CN110110233A - Information processing method, device, medium and calculating equipment

Info

Publication number: CN110110233A
Application number: CN201910388205.4A
Authority: CN
Inventors: 谢鹏; 刘洪彬; 魏望; 高畅; 任重起
Original assignee: Netease Media Technology Beijing Co Ltd
Current assignee: Netease Media Technology Beijing Co Ltd
Priority date: 2019-05-09
Filing date: 2019-05-09
Publication date: 2019-08-09
Anticipated expiration: 2039-05-09
Also published as: CN110110233B

Abstract

Embodiments of the present invention provide a kind of information processing method.This method comprises: obtaining the user information of user；According to the user information, multiple information to be recommended are obtained, wherein the multiple information to be recommended includes having the first information to be recommended of pre- sequencing information；And according to the pre- sequencing information of the described first information to be recommended, the user information and the multiple information to be recommended, obtained and the multiple information to be recommended multiple prediction clicking rates correspondingly using clicking rate prediction model.Method of the invention is due to when determining the prediction clicking rate of the first information to be recommended, the pre- sequencing information of first model to be recommended is considered simultaneously, therefore the accuracy for the prediction clicking rate that can be improved, in favor of improving the accuracy and recommendation effect of information recommendation.In addition, embodiments of the present invention additionally provide a kind of information processing unit, medium and calculate equipment.

Description

Information processing method, device, medium and calculating equipment

Technical field

Embodiments of the present invention are related to information recommendation field, more specifically, embodiments of the present invention are related to a kind of letter It ceases processing method, device, medium and calculates equipment.

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.

The groundwork of information recommendation is to solve the problems, such as information overload, i.e., a small amount of user is filtered out from a large amount of information Interested information.Common information recommendation generally comprises two stages of recalling and sort.Wherein, recalling the stage specifically is certainly The interested partial information of user is picked out in the massive information of server storage.Phase sorting is to pick out to the stage of recalling Partial information be ranked up.

Wherein the stage of recalling can recall method using rule and model recalls method.Rule is recalled for example to can be and is based on Artificial rule carries out selecting for information；And it is to carry out selecting for information using computation model that model, which is recalled,.In general, using model Recalling the multiple information to be recommended selected has pre- sequencing information.And phase sorting use proposed algorithm model often The pre- sequencing information of multiple information to be recommended cannot be absorbed.I.e. phase sorting can not often consider pre- sequencing information, this can make The recommendation performance for obtaining the recommender system based on proposed algorithm model foundation has certain lose.

Summary of the invention

Therefore in the prior art, when using existing information recommendation method come to user's recommendation information, due to that can not examine Consider the pre- sequencing information of the information to be recommended obtained, therefore the clicking rate accuracy rate that there is the information to be recommended of prediction is low, information The defect of recommendation effect difference.

Thus, it is also very desirable to which a kind of improved information processing method is mentioned with improving the accuracy rate of determining prediction clicking rate High information recommendation effect.

In the present context, the embodiment of the present invention is desirable to provide a kind of information processing method, can believe obtaining The pre- sequencing information of information to be recommended is considered when the prediction clicking rate of breath, to improve the accuracy rate of determining prediction clicking rate.

In the first aspect of embodiment of the present invention, a kind of information processing method is provided, comprising: obtain the user of user Information；According to the user information, multiple information to be recommended are obtained, wherein the multiple information to be recommended includes having pre- sequence The information to be recommended of the first of information；And according to the pre- sequencing information of the described first information to be recommended, the user information and institute Multiple information to be recommended are stated, are obtained and the multiple predictions correspondingly of the multiple information to be recommended using clicking rate prediction model Clicking rate.

In one embodiment of the invention, above-mentioned multiple information to be recommended include the multiple first information to be recommended, above-mentioned Being obtained using clicking rate prediction model with the one-to-one multiple prediction clicking rates of the multiple information to be recommended includes: according to institute The multiple first information to be recommended is divided at least one information area by the pre- sequencing information for stating the multiple first information to be recommended Between, it obtains being used for table with the multiple first information to be recommended multiple block informations correspondingly, the multiple block information Levy information section belonging to the multiple first information to be recommended；According to the multiple first information to be recommended and the multiple area Between information, obtain and the one-to-one multiple first inputs information of the multiple first information to be recommended, the first input letter Breath is spliced to obtain by first information to be recommended and block information corresponding with the one first information to be recommended；And it will Other information to be recommended in the user information, the multiple information to be recommended in addition to the described first information to be recommended and described Clicking rate prediction model described in multiple first input information inputs, obtains multiple correspondingly with the multiple information to be recommended Predict clicking rate.

In another embodiment of the present invention, the multiple first information to be recommended is divided at least one information area Between include: according to the pre- sequencing information of the multiple first information to be recommended, will be described more using the discretization method based on entropy A first information to be recommended is divided at least one information section.

In yet another embodiment of the present invention, above- mentioned information processing method further include: multiple sample datas are obtained, it is described At least one sample data in multiple sample datas include it has been recommended that information, with described it has been recommended that the corresponding section letter of information It is breath, described it has been recommended that information is clicked information and the user information, it is described it has been recommended that information is clicked information for table Sign is described it has been recommended that whether information is clicked by the user；And mould is predicted using the multiple sample data as the clicking rate The input of type, using the predetermined optimization algorithm optimization training clicking rate prediction model.Wherein, the clicking rate prediction model packet It includes Logic Regression Models, decision-tree model or gradient and promotes tree-model.

In yet another embodiment of the present invention, according to the user information, obtaining multiple information to be pushed includes: basis The user information obtains the multiple first information to be recommended using model is recalled；And the first input information is by one A first information to be recommended, with the one first corresponding block information of information to be recommended and one first letter to be recommended The source-information of breath splices to obtain, the source-information be used to characterize obtain the first information to be recommended use recall model.Its In, the model of recalling includes that matrix decomposition recalls model, collaborative filtering recalls model and neural network is recalled in model extremely It is one few.

In yet another embodiment of the present invention, above-mentioned multiple information to be recommended further include the second information to be recommended, described According to the user information, multiple information to be pushed are obtained further include: recall rule according to predetermined, it is to be recommended to obtain described second Information.The predetermined rule of recalling includes: that hot spot recalls rule, rule is recalled in region and emergency event is recalled in rule at least One.

In yet another embodiment of the present invention, before obtaining the multiple prediction clicking rate, the information processing side Method further include: according to the user information and the multiple information to be recommended, determine the user information with the multiple wait push away Recommend the intersection information of information；And obtaining the multiple prediction clicking rate includes: the walkthrough according to the described first information to be recommended Sequence information, the user information, the multiple information to be recommended and the intersection information, using clicking rate prediction model obtain with The multiple information to be recommended multiple prediction clicking rates correspondingly.

In yet another embodiment of the present invention, above- mentioned information processing method further include: clicked according to the multiple prediction Rate, Xiang Suoshu user recommend information to be recommended, comprising: by the multiple information to be recommended according to one-to-one prediction clicking rate Size successively sort；And recommend to come the information to be recommended in predetermined position to the user.

In the second aspect of embodiment of the present invention, a kind of information processing unit is provided, comprising: user information obtains Module, for obtaining the user information of user；Recommendation information obtains module, for according to the user information, obtain it is multiple to Recommendation information, wherein the multiple information to be recommended includes having the first information to be recommended of pre- sequencing information；And clicking rate Module is obtained, for according to the pre- sequencing information of the described first information to be recommended, the user information and the multiple to be recommended Information is obtained and the multiple information to be recommended multiple prediction clicking rates correspondingly using clicking rate prediction model.

In one embodiment of the invention, above-mentioned multiple information to be recommended include the multiple first information to be recommended, described It includes: information interval division submodule that clicking rate, which obtains module, for the pre- sequence according to the multiple first information to be recommended The multiple first information to be recommended is divided at least one information section by information, obtain with it is the multiple first to be recommended Multiple block informations, the multiple block information are used to characterize belonging to the multiple first information to be recommended information correspondingly Information section；First input acquisition of information submodule, for according to the multiple first information to be recommended and the multiple area Between information, obtain and the one-to-one multiple first inputs information of the multiple first information to be recommended, the first input letter Breath is spliced to obtain by first information to be recommended and block information corresponding with the one first information to be recommended；And it is pre- Clicking rate acquisition submodule is surveyed, for the described first letter to be recommended will to be removed in the user information, the multiple information to be recommended Clicking rate prediction model described in other information to be recommended and the multiple first input information input outside breath, obtains and described more A information to be recommended multiple prediction clicking rates correspondingly.

In another embodiment of the present invention, above- mentioned information interval division submodule is specifically used for: according to the multiple The pre- sequencing information of first information to be recommended is divided the multiple first information to be recommended using the discretization method based on entropy To at least one information section.

In yet another embodiment of the present invention, above- mentioned information processing unit further include: sample data obtains module, is used for Obtain multiple sample datas, at least one sample data in the multiple sample data include it has been recommended that information, with it is described It is the corresponding block information of recommendation information, described it has been recommended that information is clicked information and the user information, it is described it has been recommended that letter Breath to be clicked information described it has been recommended that whether information is clicked by the user for characterizing；Prediction model optimization module, is used for Using the multiple sample data as the input of the clicking rate prediction model, using the predetermined optimization algorithm optimization training point Hit rate prediction model.Wherein, the clicking rate prediction model includes Logic Regression Models, decision-tree model or gradient boosted tree mould Type.

In yet another embodiment of the present invention, it includes first information acquisition submodule that the recommendation information, which obtains module: For obtaining the multiple first information to be recommended using model is recalled according to the user information.The first input information By first information to be recommended, block information corresponding with the one first information to be recommended and one first wait push away The source-information for recommending information splices to obtain, the source-information be used to characterize obtain the first information to be recommended use recall mould Type.Wherein, the model of recalling includes that matrix decomposition recalls model, collaborative filtering recalls model and neural network is recalled in model At least one.

In yet another embodiment of the present invention, the multiple information to be recommended further includes the second information to be recommended, described It further includes the second acquisition of information submodule that recommendation information, which obtains module, for recalling rule according to predetermined, obtain described second to Recommendation information.Wherein, the predetermined rule of recalling includes: that hot spot recalls rule, rule is recalled in region and rule are recalled in emergency event At least one of then.

In yet another embodiment of the present invention, the information processing unit further include: intersection information determining module is used for It is obtained before module obtains the multiple prediction clicking rate in the clicking rate, according to the user information and the multiple wait push away Information is recommended, determines the intersection information of the user information Yu the multiple information to be recommended.It is specific that the clicking rate obtains module For: according to the pre- sequencing information of the described first information to be recommended, the user information, the multiple information to be recommended and described Intersection information is obtained and the multiple information to be recommended multiple prediction clicking rates correspondingly using clicking rate prediction model.

In yet another embodiment of the present invention, above- mentioned information processing unit further includes information recommendation module, is used for basis The multiple prediction clicking rate, Xiang Suoshu user recommend information to be recommended.Specifically, the information recommendation module includes: information Sorting sub-module, for the multiple information to be recommended successively to sort according to the size of one-to-one prediction clicking rate；With And information recommendation submodule, for coming the information to be recommended in predetermined position to user recommendation.

In the third aspect of embodiment of the present invention, a kind of computer readable storage medium is provided, is stored thereon with Executable instruction, the first aspect which makes processor execute embodiment according to the present invention when being executed by processor are mentioned The information processing method of confession.

In the fourth aspect of embodiment of the present invention, a kind of calculating equipment is provided.The calculating equipment includes being stored with The one or more memories and one or more processors of executable instruction.The processor executes the executable instruction, uses To realize information processing method provided by the first aspect of embodiment according to the present invention.

The information processing method, device, medium of embodiment and calculating equipment according to the present invention, are predicted using clicking rate When the clicking rate of model prediction information to be recommended, it can be considered that the pre- sequencing information of the first information to be recommended, so as to abundant Clicking rate predict when the considerations of factor, the accuracy of the prediction clicking rate improved.Therefore it can be improved and clicked according to prediction Rate carries out the recommendation effect of information recommendation, improves user experience.

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 diagrammatically illustrates the information processing method, device, medium of embodiment according to the present invention and calculates equipment Application scenarios；

Fig. 2 diagrammatically illustrates the flow chart of information processing method according to a first embodiment of the present invention；

Fig. 3 A diagrammatically illustrates the flow chart of information processing method according to a second embodiment of the present invention；

Fig. 3 B is diagrammatically illustrated according to multiple prediction flow charts of the clicking rate to user's recommendation information；

Fig. 4 diagrammatically illustrates the stream according to an embodiment of the present invention for obtaining prediction clicking rate corresponding with information to be recommended Cheng Tu；

Fig. 5 diagrammatically illustrates the flow chart of information processing method according to a third embodiment of the present invention；

Fig. 6 diagrammatically illustrates the flow chart of information processing method according to a fourth embodiment of the present invention；

Fig. 7 diagrammatically illustrates the process structure figure of information processing method according to an embodiment of the invention；

Fig. 8 diagrammatically illustrates the block diagram of information processing unit according to an embodiment of the invention；

Fig. 9 diagrammatically illustrates showing for the program product according to an embodiment of the invention for being adapted for carrying out information processing method It is intended to；And

Figure 10 diagrammatically illustrates the calculating equipment according to an embodiment of the invention for being adapted for carrying out information processing method Block diagram.

In the accompanying drawings, the identical or identical or corresponding part of corresponding label expression.

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 proposes a kind of information processing method, device, medium and calculates equipment.

Herein, it is to be understood that related term is explained as follows:

Logistic regression (Logistic Regression) is a kind of for solving the machine learning method of classification problem, is used In estimate certain things a possibility that.Such as certain advertisement a possibility that being clicked by user etc..Here be " possibility ", rather than " probability " mathematically.The result of logistic regression is not the probability value in mathematical definition, it is not possible to directly be used as probability value. The result is often used for and other characteristic value weighted sums, and indirect multiplication.Logistic regression and linear regression (Linear It Regression) is all a kind of generalized linear model (generalized linear model).Logistic regression assumes dependent variable v Bernoulli Jacob's distribution is obeyed, and dependent variable y Gaussian distributed is assumed in linear regression.Therefore have with linear regression it is many it is identical it Place, if removing Sigmoid mapping function, logistic regression algorithm is exactly a linear regression.It can be said that logistic regression is with line Property to return be theories integration, but logistic regression introduces non-linear factor by Sigmoid function, therefore can easily locate Manage classification problem.

Clicking rate (Click-through Rate, CTR) refers to and specifies content to be clicked number in website or application program The ratio between with exposure frequency, clicking rate is usually that the important indicator of recommendation effect is measured in recommender system.

Matrix decomposition (Matrix Factorization, MF), disassembles the product for several matrixes for matrix, the matrix Decomposition may include triangle decomposition, full-rank factorization, QR (ORTHOGONAL TRIANGULAR) is decomposed, Jordan is decomposed and singular value decomposition Decomposition methods such as (Singular Value Decomposition, SVD), wherein common matrix disassembling method includes: triangle It decomposes, QR is decomposed and singular value decomposition.

Collaborative filtering (Collaborative Filtering, CF) is common using having similar tastes and interests, possessing in simple terms The hobby of the group of experience is recommended interested information to user, and it is considerable degree of that the personal mechanism by cooperation gives information It responds (as scored) and records to achieve the purpose that filtering, and then help others' filter information.Wherein, different foregone conclusion is responded It is limited to of special interest, the response of information of especially loseing interest in record is also quite important.Collaborative filtering can be divided into comparation and assessment again (rating) and (social filtering) filters in group.

Neural network, a kind of neural network that simulating human brain is to can be realized the machine learning techniques of class artificial intelligence. Neural network includes: input layer, hidden layer, output layer, and when planned network, input layer and output layer number of nodes are fixed, hidden layer It can freely specify.Every layer is made of neuron, and neuron is the model comprising input, output and computing function.

MDLP (Minimal description length principle, most short description length principle) feature is discrete Change: applied to the supervision discretization method of continuous feature, data cut-point is found by the way of information gain.

Additionally, it should be appreciated that any number of elements in attached drawing is used to example rather than limitation and any name It is only used for distinguishing, 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

In using the information to be recommended for recalling model acquisition, pre- sequencing information can characterize user couple to a certain extent The interest level of information to be recommended.And in the prior art when phase sorting determines the prediction clicking rate of information to be recommended, it is past It is past to fail to consider pre- sequencing information.It therefore, undoubtedly can be due to row when carrying out information recommendation according to the ranking results of phase sorting Lose pre- sequencing information when sequence, so that recommend efficiency undesirable, the degraded performance of recommender system.The inventors discovered that if logical It crosses fusion method and will recall model and merged with clicking rate prediction model, then can effectively consider to be recommended when predicting clicking rate The pre- sequencing information of information, and clicking rate predictablity rate is therefore improved, improve information recommendation effect.

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 diagrammatically illustrates the information processing method, device, medium of embodiment according to the present invention and calculates equipment Application scenarios.It should be noted that being only the example that can apply the application scenarios of the embodiment of the present invention shown in Fig. 1, to help Those skilled in the art understand that technology contents of the invention, but it is not meant to that the embodiment of the present invention may not be usable for other and set Standby, system, environment or scene.

As shown in Figure 1, the application scenarios 100 include terminal device 111,112,113, network 120 and database 130. Network 120 between terminal device 111,112,113 and database 130 to provide the medium of communication link.Network 120 can be with Including various connection types, such as wireless communication link or fiber optic cables etc..

Terminal device 111,112,113 can be interacted by network 120 with database 130 in response to the operation of user, with Obtain information to be recommended recommended to the user.Various client applications can be installed on terminal device 111,112,113, such as (merely illustrative) such as web browser applications, the application of news browsing class, searching class application, social platform softwares.

Terminal device 111,112,113 can be the various electronic equipments with display screen and supported web page browsing, with Recommendation information is shown to user.The terminal device 111,112,113 includes but is not limited to smart phone, tablet computer, on knee Portable computer and desktop computer etc..

Terminal device 111,112,113 can also for example have processing function, for obtaining from database 130 Information to be recommended is handled, and the prediction clicking rate of information to be recommended is obtained；And according to prediction clicking rate treat recommendation information into Row sequence, with according to ranking results to the interested information of user recommended user.

According to an embodiment of the invention, as shown in Figure 1, the application scenarios 100 for example can also include server 140.It should Server 140 can be to provide the server of various services, such as the information of user is recommended to terminal device 111,112,113 The back-stage management server (merely illustrative) supported is provided.Correspondingly, database 130 for example can be to be integrated in server 140 In database.

According to an embodiment of the invention, the server 140 for example may also respond to asking for terminal device 111,112,113 It asks, treats recommendation information and handled, obtain the prediction clicking rate of information to be recommended.And recommendation is treated according to prediction clicking rate Breath is ranked up, to determine the information for needing to recommend according to ranking results.The information for eventually needing to recommend feeds back to terminal and sets Standby 111,112,113 show user for terminal device.

It should be noted that information processing method provided by the embodiment of the present disclosure generally can by terminal device 111, 112,113 or server 140 execute.Correspondingly, information processing unit provided by the embodiment of the present invention generally can be set in In terminal device 111,112,113 or server 140.Information processing method provided by the embodiment of the present invention can also be by difference In server 140 and the server or server cluster that can be communicated with terminal device 111,112,113 and/or server 140 It executes.Correspondingly, information processing unit provided by the embodiment of the present invention also can be set in being different from server 140 and can In the server or server cluster communicated with terminal device 111,112,113 and/or server 140.

It should be understood that the terminal device, network, server, the number of database and type in Fig. 1 are only schematic 's.According to needs are realized, terminal device, network, server and the database of arbitrary number and type can have.

Illustrative methods

Below with reference to the application scenarios of Fig. 1, the information of illustrative embodiments according to the present invention is described with reference to Fig. 2~7 Processing method.It should be noted which is shown only for the purpose of facilitating an understanding of the spirit and principles of the present invention for above-mentioned application scenarios, this The embodiment of invention is unrestricted in this regard.On the contrary, embodiments of the present invention can be applied to it is applicable any Scene.

Fig. 2 diagrammatically illustrates the flow chart of information processing method according to a first embodiment of the present invention.

As shown in Fig. 2, information processing method according to a first embodiment of the present invention includes operation S210~operation S230.It should Information processing method for example can by reference Fig. 1 terminal device 111,112,113 or server 140 execute.The operation The information that S210~operation S230 specifically can be used to implement information recommendation system recalls the rank of recalling and sort of information in the stage The prediction of clicking rate in section.

In operation S210, the user information of user is obtained.

Pacify according to an embodiment of the invention, operation S210 specifically for example may is that according in user's registration terminal equipment The account information inputted when the application program of dress obtains user information corresponding with account information from server or database. The user information for example may include that user basic information (age, gender) and/or user preference information (can be and pre-select Information type: sport, finance and economics and/or amusement etc.) etc..

Multiple information to be recommended are obtained according to the user information in operation S220, multiple information to be recommended includes tool There is the first information to be recommended of pre- sequencing information.

According to an embodiment of the invention, the operation for obtaining the first information to be recommended in operation S220 for example can be with are as follows: according to User information obtains the first information to be recommended using model is recalled.Wherein, recalling model for example may include that matrix decomposition is recalled Model, collaborative filtering recall model and/or neural network recalls model etc..The first information tool to be recommended is obtained using model is recalled Body may include: using user information as the input for recalling model, using matrix disassembling method, collaborative filtering method and/or mind Recommendation information corresponding with user information in magnanimity recommendation information is obtained from database 130 or server 140 through network, will acquire Recommendation information as the described first information to be recommended.

Wherein, due to recalling model when obtaining information to be recommended, user information is considered, therefore use each mould The multiple first information to be recommended that type is got all have accurately ranking score corresponding with each model.Wherein, sequence point Number can be used for characterizing the matching degree of the first information and user information to be recommended obtained using each model, and matching degree is higher, Then ranking score is higher.Above-mentioned pre- sequencing information for example can be the ranking score and/or according to multiple first wait push away Recommend the sorting position that the accurate ranking score of information is obtained from high to low arrangement.For example, being arranged in multiple first information to be recommended The corresponding sorting position of the highest information of sequence score is position 1, then the sorting position in pre- sequencing information can be for example expressed as Order=1.

According to an embodiment of the invention, the multiple information to be recommended obtained for example can also include recalling method by rule Second got information to be recommended.Then operating the method that S220 obtains the second information to be recommended specifically can also include: basis It is preset to recall rule, from obtaining and recall rule match in the magnanimity recommendation information of database 130 or server 140 Recommendation information, and will acquire with recall the recommendation information of rule match as the described second information to be recommended.Wherein, scheduled Rule is recalled such as may include: hot spot recalls rule, rule is recalled in region and/or emergency event recalls rule.

Wherein, according to hot spot recall Rule to recommendation information for example can be clicking rate in the given time and be higher than The information etc. of predetermined clicking rate.According to region recall Rule to recommendation information for example can be record event occur exist Information etc. in predetermined region.According to emergency event recall Rule to recommendation information for example to can be include " Shake ", " explosion " or " mud-rock flow " etc. can characterize information of keyword of emergency event etc..It is understood that above-mentioned basis The type for the recommendation information that the rule method of recalling is got is used as example only in favor of understanding that the present invention, the present invention do not limit this It is fixed.

It is adopted in operation S230 according to the pre- sequencing information of the first information to be recommended, user information and multiple information to be recommended It is obtained and multiple information to be recommended multiple prediction clicking rates correspondingly with clicking rate prediction model.

According to an embodiment of the invention, operation S230 specifically for example may is that the walkthrough of first information to be recommended Sequence information, user information and multiple information to be recommended are input in clicking rate prediction model simultaneously, via clicking rate prediction model The prediction clicking rate of each information to be recommended in multiple information to be recommended is calculated.Wherein, the clicking rate prediction model It such as may include that Logic Regression Models, decision-tree model or gradient promote tree-model etc..

According to an embodiment of the invention, for the ease of when clicking rate prediction model obtains prediction clicking rate, it can be by the One information to be recommended and pre- sequencing information and the model of recalling for obtaining first information to be recommended correspond.Operation S230 can specifically include: first by the first information to be recommended, the first information to be recommended pre- sequencing information and obtain first wait push away The identification information for recalling model for recommending information is spliced to form an input information；Then again by the input information being spliced to form, more Other information and user information to be recommended in a information to be recommended in addition to the first information to be recommended input clicking rate prediction model, It is calculated and multiple information to be recommended multiple prediction clicking rates correspondingly.

According to an embodiment of the invention, operation S431~operation that operation S230 specifically can for example be described by Fig. 4 S433 come determine with multiple information to be recommended multiple prediction clicking rates correspondingly, this will not be detailed here.

In summary, the information processing method of the embodiment of the present invention, when predicting the clicking rate of information to be recommended, Ke Yitong When consider pre- sequencing information by recalling the first information to be recommended that model obtains.Therefore clicking rate prediction model can be improved Predict the accuracy rate of the prediction clicking rate of the first obtained information to be recommended.Recommended according to the prediction clicking rate to user with improving Recommendation effect when information improves user experience.

Fig. 3 A diagrammatically illustrates the flow chart of information processing method according to a second embodiment of the present invention, and Fig. 3 B is schematic It shows according to multiple prediction flow charts of the clicking rate to user's recommendation information.

According to an embodiment of the invention, obtaining the multiple pre- of multiple information to be recommended by operating S210~operation S230 After surveying clicking rate, multiple information to be recommended can be carried out selected according to multiple prediction clicking rates, select to obtain and recommend to user Information.Therefore, as shown in Figure 3A, the information processing method of second embodiment of the invention is in addition to operating S210~operation S230 It outside, can also include operation S340.Operation S340 is executed after operating S230.

In operation S340, according to multiple prediction clicking rates, recommend information to be recommended to user.

According to an embodiment of the invention, operation S340 specifically for example may include: first according to multiple prediction clicking rates, really The information to be recommended that directional user recommends；Then by terminal device 111,112,113 that this is recommended to the user to be recommended again Information shows user.Wherein, operation S340 specifically can be will be greater than predetermined clicking rate prediction clicking rate it is corresponding to Recommendation information is determined as information to be recommended recommended to the user.

According to an embodiment of the invention, as shown in Figure 3B, operation S340 specifically can also include operation S341~operation S342.In operation S341, multiple information to be recommended are successively sorted according to the size of one-to-one prediction clicking rate；It is operating S342 recommends the information to be recommended for coming predetermined position to user.

According to an embodiment of the invention, since clicking rate prediction model is obtaining the prediction clicking rate of multiple information to be recommended When, user information is considered, therefore the information to be recommended of obtained prediction clicking rate greatly is usually to match with user information High information is spent, i.e. the big information to be recommended of prediction clicking rate is the interested information of user.Therefore pass through operation S341 The information to be recommended for predicting clicking rate big (i.e. user is interested) is come into forward position, will predict small (the i.e. user of clicking rate Lose interest in) information to be recommended come rearward position.Then operate S342 specifically and can be before coming n position wait push away Information is recommended as information to be recommended recommended to the user, user is showed by terminal device 111,112,113.Wherein, make a reservation for Position is preceding n position, and n can be for example the arbitrary positive integer values such as 5,10,12, and the value of the n specifically for example can root It is set according to user demand.

According to an embodiment of the invention, in order to reduce resource consume, aforesaid operations S210~operation S230 as far as possible Such as it can periodically be executed using first time period as the period.And it operates S341~S342 and can be acquisition in response to user Request is performed.Wherein, first time period for example can be one day, 12 hours or 6 hours etc..The acquisition request of the user Such as it can be when the application program in user's using terminal equipment 111,112,113 browses recommendation information, in response to user What the operation of " slide downward " page or the operation of click " refreshing " control generated.

According to an embodiment of the invention, when the execution frequency of operation S341~S342 is higher than operation S210~operation S230 Execute frequency when, execute operate S210~operation S230 after for the first time execute operation S341 when, such as can be to pass through behaviour All information to be recommended for making S220 acquisition are ranked up.And subsequent execution operate S341 when, then be to pass through operation In all information to be recommended that S220 is obtained other information to be recommended in addition to the information to be recommended recommended to user into Row sequence.

Fig. 4 diagrammatically illustrates the stream according to an embodiment of the present invention for obtaining prediction clicking rate corresponding with information to be recommended Cheng Tu.

According to an embodiment of the invention, in view of the first information to be recommended obtained for different user using model is recalled Ranking score be likely to be at different sections.It, can not if therefore directly inputting clicking rate prediction model using ranking score The accuracy of prediction clicking rate is effectively improved, performance is recommended also just not to be obviously improved.Therefore, the embodiment of the present invention can be preferred Ground is using the sorting position obtained according to ranking score as pre- sequencing information.

Furthermore in order to reduce influence of the abnormal ranking score to clicking rate prediction model, promoting fitting effect and increase base In the robustness of the recommender system of clicking rate prediction model.It, can also basis in the case where the first information to be recommended is multiple The sorting position and/or ranking score of multiple first information to be recommended, by the multiple first information discretizations to be recommended to several Information section.Therefore, as shown in figure 4, the operation S230 in Fig. 2 can specifically include operation S431~operation S433.

Multiple first information to be recommended are drawn according to the pre- sequencing information of the multiple first information to be recommended in operation S431 At least one information section is assigned to, is obtained and the multiple first information to be recommended multiple block informations correspondingly.Wherein, multiple Block information for characterizing information section belonging to the multiple first information to be recommended respectively.

According to an embodiment of the invention, operation S431 specifically for example may is that according to the pre- of the multiple first information to be recommended The sorting position of sequencing information characterization, is divided to same information section for the first close information to be recommended of sorting position.For example, The first information to be recommended that sorting position is 1~5 can be divided to same information according to the principle that sorting position equalization divides The first information to be recommended that sorting position is 6~10 is divided to same information section by section, and so on, obtain at least one A information section.Alternatively, operation S431 can also first to the pre- sequence information representation of the multiple first information to be recommended sequence Score carries out the division in score section, is then again divided to the first information to be recommended that ranking score belongs to same score section Same information section.Alternatively, operation S431 can also comprehensively consider sorting position and ranking score comes to the first letter to be recommended Breath carries out the division in information section.

According to an embodiment of the invention, in order to preferably react the authenticity of the data of pre- sequencing information, operation S431 specifically for example may is that the pre- sequencing information according to the multiple first information to be recommended, using based on entropy (or based on information increase Benefit) discretization method the multiple first information to be recommended are divided at least one information section.Wherein, based on the discretization of entropy Method can specifically use the thinking similar with decision-tree model, be calculated using synthetic method or splitting method according to entropy and preset Determine to determine to synthesize or classify.According to an embodiment of the invention, the discretization method based on entropy specifically for example can be MDLP Discretization method finds the cut-point in each information section in a manner of using information gain.

According to an embodiment of the invention, operation S431 obtain with the multiple first information to be recommended multiple areas correspondingly Between information specifically for example can be characterization the first information to be recommended belonging to information section section number.Correspondingly, it is operating It, can also be at least one information area while multiple first information to be recommended are divided at least one information section by S431 Between distribution section number.

In operation S432, according to the multiple first information to be recommended and multiple block informations, obtain with it is multiple first to be recommended The one-to-one multiple first input information of information.

According to an embodiment of the invention, for the ease of when clicking rate prediction model obtains prediction clicking rate, it can be by the The block information of one information to be recommended and the first information to be recommended corresponds, above-mentioned and each first information pair to be recommended The the first input information answered specifically can be by first information to be recommended and block information corresponding with first information to be recommended Splicing obtains.

According to an embodiment of the invention, in view of using different record identical content first recalling model and obtaining The pre- sequencing information of information to be recommended may be different, therefore difference recalls record identical content first that model obtains Information to be recommended might have different block informations.Therefore, in order to completely express each first information to be recommended, Mei Ge First input information of one information to be recommended specifically can be by each first information and each first information pair to be recommended to be recommended The source-information of the block information and each first information to be recommended answered splices to obtain.Wherein, the source-information is used for table What is used when sign each first information to be recommended of acquisition recalls model, and specifically, which, which for example can be, recalls model Identification information etc..Therefore, by two it is different recall model and obtain two record the first to be recommended of identical content When information, the source-information for the first information to be recommended for recording identical content due to this two is different, this two record phase The first information to be recommended with content is two information to be recommended of different first, and therefore available two different first Input information.

According to an embodiment of the invention, after getting the first input information, i.e., executable operation S433 believes user Other information to be recommended and multiple first input information input points in breath, multiple information to be recommended in addition to the first information to be recommended It hits rate prediction model, obtains and multiple information to be recommended multiple prediction clicking rates correspondingly.

In summary, the embodiment of the present invention is when determining the prediction clicking rate of the first information to be recommended, by according to walkthrough First information to be recommended is divided to multiple information sections by sequence information, and with the area in information section belonging to the first information to be recommended Between information as feature input clicking rate prediction model, can be avoided influence of the abnormal pre- sequencing information to prediction result, and because This further increases the accuracy of the prediction clicking rate of the information to be recommended of determining first, further increases based on prediction clicking rate Carry out the recommendation effect and user experience of information recommendation.

Fig. 5 diagrammatically illustrates the flow chart of information processing method according to a third embodiment of the present invention.

According to an embodiment of the invention, needing before treating recommendation information and carrying out clicking rate prediction to initial clicking rate Prediction model is trained.Further, clicking rate prediction model can also be carried out excellent after carrying out clicking rate prediction Change.Therefore, as shown in figure 5, the information processing method of third embodiment of the invention may be used also other than operating S210~operation S230 To include operation S550~operation S560.Operation S550~operation S560 can be held before operating S210~operation S230 Row, or executed after operating S210~operation S230.

In operation S550, multiple sample datas are obtained.

Wherein, multiple sample data is specially the input data of clicking rate prediction model.Multiple sample data should wrap Included by operating S230 information to be recommended recommended to the user, i.e., it has been recommended that information, or by existing information at It is that reason method was recommended to user it has been recommended that information.It is understood that each sample data in multiple sample data is also It should include to be clicked information, this is clicked information for characterizing that sample data includes it has been recommended that whether information is by user's point It hits.Specifically, multiple sample data can be using the corresponding information that is clicked as label.For example, when being clicked information Characterize that sample data includes it has been recommended that the label of the sample data can be 1 when information is clicked by user；And works as and be clicked letter Cease that characterization sample data includes it has been recommended that the label of the sample data can be -1 when information is not clicked by user.

According to an embodiment of the invention, in order to enable the clicking rate prediction model of training optimization is it can be considered that first is to be recommended The pre- sequencing information of information, then at least one sample data in above-mentioned multiple sample datas should include: it has been recommended that information, with It has been recommended that the corresponding block information of information, it has been recommended that information is clicked information and user information.Wherein, at least one sample It is that data include it has been recommended that information should for by recall model acquisition recommendation information, as operation S220 described in first Information to be recommended.Wherein, and it has been recommended that the corresponding block information of information can be is obtained by the operation S431 determination of Fig. 4 description , details are not described herein.

It is excellent using predetermined optimization algorithm using multiple sample datas as the input of clicking rate prediction model in operation S560 Change training clicking rate prediction model.

According to an embodiment of the invention, operation S560 specifically may is that the input clicking rate prediction of multiple sample datas Model, by clicking rate prediction model respectively obtain it is that multiple sample data includes it has been recommended that information prediction clicking rate.And The prediction clicking rate is compared with the information that is clicked that multiple sample datas include, which is calculated by loss function Hit the penalty values of rate prediction model；Then the parameters in clicking rate prediction model are adjusted according to the penalty values excellent Change.Then the predetermined optimization algorithm is the loss function, which is specifically as follows cross entropy loss function Deng.

According to an embodiment of the invention, operation S560 specifically can also be by using forward-backward algorithm cutting (Forward- Backward Splitting, FOBOS) algorithm or FTRL (Follow The Regularized Leader) algorithm etc. carry out root According to multiple sample datas include it has been recommended that the prediction clicking rate of information, optimizes clicking rate prediction model.It is preferred that using FTRL algorithm, loss can be passed through under the premise of guaranteeing the obtained clicking rate prediction model of optimization precision with higher Certain precision improves the sparsity of clicking rate prediction model.

Fig. 6 diagrammatically illustrates the flow chart of information processing method according to a fourth embodiment of the present invention.

According to an embodiment of the invention, in view of being that there are relevant passes between the information and user information to be recommended of acquisition System.Then in order to further embody the incidence relation, better data characteristics is obtained, it can also be to clicking rate prediction model When inputting information to be recommended, pre- sequencing information and user information, while inputting the intersection information of user information Yu information to be recommended. To further increase the accuracy for the prediction clicking rate that clicking rate prediction model determines.Therefore, as shown in fig. 6, the present invention the The information processing method of four embodiments can also include operation S670 other than operating S210~operation S230.Operation S670 It should be executed between operation S220 and operation S230.

User information and multiple information to be recommended are determined according to user information and multiple information to be recommended in operation S670 Intersection information.

According to an embodiment of the invention, the intersection information specifically can for example be determined by following operation: using One- The mode of Hot vector carries out characteristic crossover to user information and multiple information to be recommended.Specifically, multiple information to be recommended are for example It can have the information to be recommended for recording competitive sports, in user information when with the information that user preferences are sport, lead to The information that user preferences are sport and the information mixing together to be recommended for recording competitive sports can will be characterized by crossing operation S670 For a cross feature.

Correspondingly, operation S230 can then be realized by operation S680 shown in fig. 6.In operation S680, according to first Pre- sequencing information, user information, multiple information to be recommended and the intersection information of information to be recommended, are obtained using clicking rate prediction model It takes and multiple information to be recommended multiple prediction clicking rates correspondingly.It is specific to be are as follows: by the pre- sequence of the first information to be recommended The input of information, user information, multiple information to be recommended and intersection information as clicking rate prediction model, be calculated it is multiple to The prediction clicking rate of recommendation information.

In summary, the information processing method of the embodiment of the present invention is determining letter to be recommended using clicking rate prediction model When the prediction clicking rate of breath, capable of considering the intersection information of user information and information to be recommended simultaneously, (what is intersected intersects Feature).Therefore, the accuracy rate that clicking rate prediction model can be further improved further increases information recommendation effect and use Family experience.

Fig. 7 diagrammatically illustrates the process structure figure of information processing method according to an embodiment of the invention.

As shown in fig. 7, in an embodiment of the present invention, the overall flow of information processing method may include:

Model is recalled in first use and predetermined rule of recalling recalls information to be recommended from million information of database respectively, Obtain recommendation information Candidate Set 1 and recommendation information Candidate Set 2.Wherein, the information to be recommended in recommendation information Candidate Set 1 is to pass through Recall what model was recalled.Difference recalls the information to be recommended that model is recalled and belongs to different recommendation information candidate subsets, and this is pushed away The information to be recommended recommended in information candidate collection 1 has pre- sequencing information.Information to be recommended in recommendation information Candidate Set 2 is to pass through It is predetermined to recall what rule was recalled.Wherein, recalling model includes that matrix decomposition recalls model, collaborative filtering recalls model and nerve net Network recalls model.Predetermined rule of recalling includes that hot spot recalls rule, rule is recalled in region and rule is recalled in emergency event.It considers Model is recalled when recalling information to be recommended, needs to use user information, it therefore, can be with before recalling information to be recommended First obtain user information.The user information can for example be obtained by the operation S210 that Fig. 2 is described, and details are not described herein；

Then to the information to be recommended for including in the different recommendation information candidate subsets in recommendation information Candidate Set 1 respectively into Row recommendation information is merged with pre- sequencing information, obtains the input information of clicking rate prediction model.Recommendation information and pre- sequence are believed The fusion of breath obtains input information can specifically be realized by operation S431~operation S432 that Fig. 4 is described, herein no longer in detail It states；

Finally the information input to be recommended that obtained input information, user information and recommendation information Candidate Set 2 include is arrived Clicking rate prediction model obtains the prediction clicking rate of each information to be recommended after handling via clicking rate prediction model.For example, The prediction clicking rate of the information to be recommended 1 arrived is 0.15, and the prediction clicking rate of information 2 to be recommended is 0.12 ... ..., letter to be recommended The prediction clicking rate for ceasing N is 0.04.

In summary, the information processing method of the embodiment of the present invention is due to it can be considered that by recalling that model recalls wait push away The pre- sequencing information of information is recommended, it can be using the available information for recalling model generation, so as to improve determining future position The accuracy of rate is hit, and therefore improves the recommendation performance of the information recommendation system based on information processing method building.Compared to The recommender system for not considering pre- sequencing information in the prior art, can make it is recommendatory can improve 2.6% so that recommend information Line on clicking rate promoted 3%.

Exemplary means

After describing the method for exemplary embodiment of the invention, next, with reference to Fig. 8 to the exemplary reality of the present invention The information processing unit for applying mode is illustrated.

Fig. 8 diagrammatically illustrates the block diagram of information processing unit according to an embodiment of the invention.

As shown in figure 8, according to embodiments of the present invention, which may include User profile acquisition module 810, recommendation information obtains module 820 and clicking rate obtains module 830.The information processing unit 800 can be used to implement basis The information processing method of the embodiment of the present invention.

User profile acquisition module 810 is used to obtain the user information (operation S210) of user.

Recommendation information obtains module 820 and is used to obtain multiple information to be recommended according to user information, plurality of to be recommended Information includes having the first information (operation S220) to be recommended of pre- sequencing information.

Clicking rate obtain module 830 be used for according to the pre- sequencing information of the first information to be recommended, user information and it is multiple to Recommendation information is obtained and the one-to-one multiple prediction clicking rate (operations of multiple information to be recommended using clicking rate prediction model S230)。

According to an embodiment of the invention, above-mentioned multiple information to be recommended include the multiple first information to be recommended.Such as Fig. 8 institute Show, it includes that information interval division submodule 831, first inputs acquisition of information submodule 832 and prediction that clicking rate, which obtains module 830, Clicking rate acquisition submodule 833.Information interval division submodule 831 is used to be believed according to the pre- sequence of the multiple first information to be recommended Multiple first information to be recommended are divided at least one information section by breath, are obtained a pair of with the multiple first information one to be recommended The multiple block informations (operation S431) answered.Wherein, multiple block informations are for characterizing belonging to the multiple first information to be recommended Information section.First input acquisition of information submodule 832 according to the multiple first information to be recommended and multiple block informations for obtaining Information (operation S432) is inputted to the multiple first information to be recommended one-to-one multiple first.Wherein, the first input information Splice to obtain by first information to be recommended and block information corresponding with first information to be recommended.Prediction clicking rate obtains Submodule 833 is taken to be used for other information to be recommended in user information, multiple information to be recommended in addition to the first information to be recommended And multiple first inputs information input clicking rate prediction models, it obtains and multiple information to be recommended multiple future positions correspondingly Hit rate (operation S433).

According to an embodiment of the invention, above- mentioned information interval division submodule 831 is specifically used for: according to multiple first wait push away Multiple first information to be recommended are divided at least one letter using the discretization method based on entropy by the pre- sequencing information for recommending information Cease section.

According to an embodiment of the invention, as shown in figure 8, above- mentioned information processing unit 800 further includes that sample data obtains mould Block 840 and prediction model optimization module 850.Sample data obtains module 840 for obtaining multiple sample datas (operation S550). At least one sample data in multiple sample data include it has been recommended that information, with it has been recommended that the corresponding block information of information, It has been recommended that information is clicked information and user information.Wherein, it has been recommended that information is clicked information for characterizing it has been recommended that believing Whether breath is clicked by user.Prediction model optimization module 850 is used for using multiple sample datas as the defeated of clicking rate prediction model Enter, training clicking rate prediction model (operation S560) is optimized using predetermined optimization algorithm.Wherein, clicking rate prediction model includes patrolling It collects regression model, decision-tree model or gradient and promotes tree-model.

According to an embodiment of the invention, as shown in figure 8, above-mentioned recommendation information, which obtains module 820, includes first information acquisition Submodule 821.The first information acquisition submodule 821 is used for according to user information, using recall model obtain multiple first to Recommendation information.Above-mentioned first input information is by first information to be recommended, section corresponding with first information to be recommended The source-information of information and first information to be recommended splices to obtain, and source-information obtains the first information to be recommended for characterizing What is used recalls model.Wherein, recalling model includes that matrix decomposition recalls model, collaborative filtering recalls model and neural network is called together Return at least one of model.

According to an embodiment of the invention, above-mentioned multiple information to be recommended further include the second information to be recommended.As shown in figure 8, It further includes the second acquisition of information submodule 822 that above-mentioned recommendation information, which obtains module 820,.The second acquisition of information submodule 822 is used In recalling rule according to predetermined, the second information to be recommended is obtained.Wherein, make a reservation for recall rule to include: that hot spot recalls rule, region It recalls rule and at least one of rule is recalled in emergency event.

According to an embodiment of the invention, as shown in figure 8, above- mentioned information processing unit 800 further includes that intersection information determines mould Block 860.The intersection information determining module 860 is used for before clicking rate acquisition module 830 obtains multiple prediction clicking rates, according to User information and multiple information to be recommended determine the intersection information (operation S670) of user information and multiple information to be recommended.On It states clicking rate and obtains module 830 and be specifically used for: according to the pre- sequencing information of the first information to be recommended, user information, multiple wait push away Information and intersection information are recommended, is obtained using clicking rate prediction model and is clicked with the one-to-one multiple predictions of multiple information to be recommended Rate (operation S680).

According to an embodiment of the invention, as shown in figure 8, above- mentioned information processing unit 800 further includes information recommendation module 870.The information recommendation module 870 is used for according to multiple prediction clicking rates, recommends information to be recommended (operation S340) to user.Tool Body, which may include information sorting submodule 871 and information recommendation submodule 872.Information sorting Module 871 is used to multiple information to be recommended successively sorting (operation S341) according to the size of one-to-one prediction clicking rate. Information recommendation submodule 872 is used to recommend to come to user the information to be recommended (operation S342) in predetermined position.

Exemplary media

After describing the method for exemplary embodiment of the invention, next, with reference to Fig. 9 to the exemplary reality of the present invention The computer readable storage medium for being adapted for carrying out information processing method for applying mode is introduced.

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 information processing method according to an embodiment of the present invention 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 execution information processing method, for example, the calculating equipment can execute step S210 as shown in Figure 2: obtaining Take the user information at family；Step S220: according to user information, multiple information to be recommended, multiple packet to be recommended are obtained Include the first information to be recommended with pre- sequencing information；Step S230: according to the pre- sequencing information of the first information to be recommended, user Information and multiple information to be recommended are obtained and the multiple predictions correspondingly of multiple information to be recommended using clicking rate prediction model Clicking rate.

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 figure 9, describing the program product for being adapted for carrying out information processing method of embodiment according to the present invention 900, can be using portable compact disc read only memory (CD-ROM) and including program code, and equipment can be being calculated, Such as it is run on PC.However, program product of the invention is without being limited thereto, in this document, readable storage medium storing program for executing can be with To be any include or the tangible medium of storage program, the program can be commanded execution system, device or device use or It is in connection.

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)-, or, it may be connected to External computing device (such as being connected using ISP by internet).

Exemplary computer device

After method, medium and the device for describing exemplary embodiment of the invention, next, with reference to Figure 10 to this The calculating equipment for being adapted for carrying out information processing method of invention illustrative embodiments is illustrated.

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, calculating equipment according to the present invention can include at least at least one processing Device and at least one processor.Wherein, the memory is stored with program code, when said program code is by the processing Device execute when so that the processor execute it is various according to the present invention described in above-mentioned " illustrative methods " part of this specification Step in the information processing method of illustrative embodiments.For example, the processor can execute step as shown in Figure 2 S210: the user information of user is obtained；Step S220: according to user information, obtaining multiple information to be recommended, multiple to be recommended Information includes having the first information to be recommended of pre- sequencing information；Step S230: believed according to the pre- sequence of the first information to be recommended Breath, user information and multiple information to be recommended are obtained one-to-one with multiple information to be recommended using clicking rate prediction model Multiple prediction clicking rates.

The meter for being adapted for carrying out information processing method of this embodiment according to the present invention is described referring to Figure 10 Calculate equipment 1000.Calculating equipment 1000 as shown in Figure 10 is only an example, should not function to the embodiment of the present invention and Use scope brings any restrictions.

As shown in Figure 10, equipment 1000 is calculated to show in the form of universal computing device.The component for calculating equipment 1000 can To include but is not limited to: at least one above-mentioned processor 1001, above-mentioned at least one processor 1002, the different system components of connection The bus 1003 of (including memory 1002 and processor 1001).

Bus 1003 may include data/address bus, address bus and control bus.

Memory 1002 may include volatile memory, such as random access memory (RAM) 10021 and/or high speed Buffer memory 10022 can further include read-only memory (ROM) 1023.

Memory 1002 can also include program/utility with one group of (at least one) program module 10024 10025, such program module 10024 includes but is not limited to: operating system, one or more application program, other programs It may include the realization of network environment in module and program data, each of these examples or certain combination.

Calculating equipment 1000 can also be with one or more external equipments 1004 (such as keyboard, sensing equipment, bluetooth equipment Deng) communicate, this communication can be carried out by input/output (I/O) interface 1005.Also, calculating equipment 1000 can also lead to Cross network adapter 1006 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 1006 is logical by other modules of bus 1003 and calculating equipment 1000 Letter.It should be understood that other hardware and/or software module are used although not shown in the drawings, can combine and calculate equipment 1000, including But it is not limited to: microcode, device driver, redundant processing unit, external disk drive array, RAID system, tape drive And data backup storage system etc..

It should be noted that although being referred to several units/modules or subelement/submodule of device in the above detailed description Block, but it is this division be only exemplary it is not enforceable.In fact, embodiment according to the present invention, is retouched above The feature and function for two or more units/modules stated can embody in a units/modules.Conversely, above description A units/modules feature and function can with further division be embodied by multiple units/modules.

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

1. a kind of information processing method, comprising:

Obtain the user information of user；

According to the user information, multiple information to be recommended are obtained, wherein the multiple information to be recommended includes having pre- sequence The information to be recommended of the first of information；And

According to the pre- sequencing information of the described first information to be recommended, the user information and the multiple information to be recommended, use Clicking rate prediction model obtains and the multiple information to be recommended multiple prediction clicking rates correspondingly.

2. according to the method described in claim 1, wherein, the multiple information to be recommended includes the multiple first information to be recommended, It is described to include: using the acquisition of clicking rate prediction model and the one-to-one multiple prediction clicking rates of the multiple information to be recommended

According to the pre- sequencing information of the multiple first information to be recommended, the multiple first information to be recommended is divided at least One information section obtains and the multiple first information to be recommended multiple block informations correspondingly, the multiple section Information is for characterizing information section belonging to the multiple first information to be recommended；

According to the multiple first information to be recommended and the multiple block information, obtain and the multiple first information to be recommended One-to-one multiple first inputs information, the first input information is by first information to be recommended and with one the The corresponding block information of one information to be recommended splices to obtain；And

By other information to be recommended in the user information, the multiple information to be recommended in addition to the described first information to be recommended And clicking rate prediction model described in the multiple first input information input, it obtains and is corresponded with the multiple information to be recommended Multiple prediction clicking rates.

3. according to the method described in claim 2, wherein, the multiple first information to be recommended is divided at least one information Section includes:

According to the pre- sequencing information of the multiple first information to be recommended, using the discretization method based on entropy by the multiple One information to be recommended is divided at least one information section.

4. according to the method described in claim 2, further include:

Multiple sample datas are obtained, at least one sample data in the multiple sample data includes it has been recommended that information and institute It states it has been recommended that the corresponding block information of information, described it has been recommended that information is clicked information and the user information, it is described to have pushed away Recommend information to be clicked information described it has been recommended that whether information is clicked by the user for characterizing；And

Using the multiple sample data as the input of the clicking rate prediction model, training institute is optimized using predetermined optimization algorithm Clicking rate prediction model is stated,

Wherein, the clicking rate prediction model includes that Logic Regression Models, decision-tree model or gradient promote tree-model.

5. according to the method described in claim 2, wherein:

According to the user information, obtaining multiple information to be pushed includes: to be obtained according to the user information using model is recalled The multiple first information to be recommended；And

The first input information is by first information to be recommended, section letter corresponding with the one first information to be recommended Breath and the source-information of one first information to be recommended splice to obtain, and the source-information obtains first wait push away for characterizing That recommends information use recalls model,

Wherein, the model of recalling includes that matrix decomposition recalls model, collaborative filtering recalls model and neural network recalls model At least one of.

6. according to the method described in claim 5, wherein, the multiple information to be recommended further includes the second information to be recommended, institute It states according to the user information, obtains multiple information to be pushed further include:

Rule is recalled according to predetermined, obtains second information to be recommended,

The predetermined rule of recalling includes: that hot spot recalls rule, rule is recalled in region and emergency event is recalled in rule at least One.

7. according to the method described in claim 1, wherein:

Before obtaining the multiple prediction clicking rate, the method also includes: according to the user information and it is the multiple to Recommendation information determines the intersection information of the user information Yu the multiple information to be recommended；And

Obtaining the multiple prediction clicking rate includes: to be believed according to the pre- sequencing information of the described first information to be recommended, the user Breath, the multiple information to be recommended and the intersection information are obtained and the multiple letter to be recommended using clicking rate prediction model Cease one-to-one multiple prediction clicking rates.

8. according to the method described in claim 1, further include: according to the multiple prediction clicking rate, Xiang Suoshu user recommend to Recommendation information, comprising:

The multiple information to be recommended is successively sorted according to the size of one-to-one prediction clicking rate；And

The information to be recommended in predetermined position is come to user recommendation.

9. a kind of information processing unit, comprising:

User profile acquisition module, for obtaining the user information of user；

Recommendation information obtains module, for multiple information to be recommended being obtained, wherein the multiple wait push away according to the user information Recommending information includes having the first information to be recommended of pre- sequencing information；And

Clicking rate obtains module, for according to the pre- sequencing information of the described first information to be recommended, the user information and described Multiple information to be recommended are obtained and the multiple information to be recommended multiple future positions correspondingly using clicking rate prediction model Hit rate.

10. a kind of computer readable storage medium, is stored thereon with executable instruction, described instruction is real when being executed by processor Now according to claim 1~any one of 8 described in method.

11. a kind of calculating equipment, comprising:

One or more memories, are stored with executable instruction；And

One or more processors execute the executable instruction, described in realization according to claim 1~any one of 8 Method.