CN105260477A - Information pushing method and device - Google Patents

Abstract

The embodiment of the application discloses an information pushing method and device, relating to the technical field of internet, wherein the information pushing method comprises the following steps: receiving an information push request aiming at a target user; obtaining the probability that the target user is interested in each piece of information recorded in a preset information base from a preset user information database, wherein the preset user information database is used for recording the probability that each user is interested in each piece of information recorded in the preset information base; determining information to be pushed from all pieces of information recorded in the preset information database according to the sequence of the obtained probabilities from high to low; and pushing the information to be pushed. By applying the scheme provided by the embodiment of the application, information can be pushed in a targeted manner.

Description

A kind of information-pushing method and device

Technical field

The application relates to Internet technical field, particularly a kind of information-pushing method and device.

Background technology

In recent years along with the fast development of network technology; network user's cumulative year after year; businessman is in order to promote its product with larger dynamics; more and more tend to throw in advertisement by network; consider this demand of businessman; operator usually can the demand of comprehensive multiple businessman, to user's advertisement information.

In practical application, when operator is to user's advertisement, often mainly consider the demand of businessman, such as, advertisement pushing frequency that the advertisement exposure rate that businessman requires, businessman require etc.Concrete, in prior art, when carrying out advertisement pushing, server comprehensively analyzes the demand of existing businessman, selects advertisement to be pushed, then push advertisement above-mentioned to be pushed according to analysis result from the advertisement that Current ad storehouse comprises.

Application aforesaid way can successfully to user's advertisement information, but in fact different user is different to the interest level of different advertisement, such as, some users are interested in toiletries advertisement, some users are interested in etc. automotive-type advertisement, so it is strong to apply specific aim when mode of the prior art carries out advertisement pushing, poor user experience.

Summary of the invention

The embodiment of the present application discloses a kind of information-pushing method and device, to carry out information pushing targetedly based on user, improves Consumer's Experience.

For achieving the above object, the embodiment of the present application discloses a kind of information-pushing method, and described method comprises:

Receive the information pushing request for targeted customer;

Described targeted customer is obtained to the interested probability of each bar information recorded in the information bank preset from the User Information Database preset, wherein, described default User Information Database is for recording each user to the interested probability of each bar information recorded in described default information bank;

According to obtained probability order from high to low, in each bar information recorded from described default information database, determine information to be pushed;

Push described information to be pushed.

In a kind of specific implementation of the application, generate described default User Information Database in such a way:

Obtain current existing information classification;

Obtain current all users to the interested probability of each information classification above-mentioned, generating probability matrix A;

Obtain in described presupposed information storehouse the probability that each bar information recorded belongs to each information classification above-mentioned, generating probability matrix B;

According to described probability matrix A and described probability matrix B, predict that each user is to the interested probability of each information recorded in described default information bank, obtain probability matrix D;

Judge whether described probability matrix D meets the convergence Rule of judgment preset, if do not meet, each element in described probability matrix D is adjusted according to the regulation rule preset, and upgrade described probability matrix D according to the element after adjustment, return the described step whether described probability matrix D meets the convergence Rule of judgment preset that judges, until described probability matrix D meets described default convergence Rule of judgment;

Described default User Information Database is generated according to described probability matrix D.

In a kind of specific implementation of the application, describedly judge that whether described probability matrix D meets the convergence Rule of judgment preset, comprising:

According to the matrix decomposition algorithm preset, described probability matrix D is decomposed into probability matrix A ' and probability matrix B ', wherein, described probability matrix A ' is the matrix corresponding with described probability matrix A, and described probability matrix B ' is the matrix corresponding with described probability matrix B;

According to described probability matrix A ', described probability matrix B and described probability matrix D, judge whether described probability matrix D meets the convergence Rule of judgment preset.

In a kind of specific implementation of the application, described according to described probability matrix A ', described probability matrix B ' and described probability matrix D, judge whether described probability matrix D meets the convergence Rule of judgment preset, and comprising:

According to the error argmin between following formula computational prediction probability and true probability _{α, β}l _(D),

${\mathrm{argmin}}_{\mathrm{\α},\mathrm{\β}}{L}_{\left(D\right)}=\underset{ij}{\Σ}{(d_{ij}-\stackrel{\→}{{\mathrm{\α}}_i}\·\stackrel{\→}{{\mathrm{\β}}_j})}^2+\mathrm{\λ}\left(|\stackrel{\→}{{\mathrm{\α}}_i}{|}^2+|\stackrel{\→}{{\mathrm{\β}}_j}{|}^2\right),$

Wherein, represent the vector of the element composition of described probability matrix A ' i-th row, represent the vector of the element composition of described probability matrix B ' jth row, λ represents regulation coefficient, d _ijrepresent the element of described probability matrix D;

According to described error argmin _{α, β}l _(D)judge whether described probability matrix D meets the convergence Rule of judgment preset.

In a kind of specific implementation of the application, described according to described error argmin _{α, β}l _(D)judge whether described probability matrix D meets the convergence Rule of judgment preset, and comprising:

Judge described error argmin _{α, β}l _(D)whether be less than the first default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment; Or

When described probability matrix D be according to adjustment after element upgrade after probability matrix, judge described error argmin _{α, β}l _(D)with error argmin _{α, β}l _(D)' between absolute difference whether be less than the second default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment, wherein, described error argmin _{α, β}l _(D)' represent based on the error that calculates of described probability matrix D before upgrading.

In a kind of specific implementation of the application, described information-pushing method also comprises:

When judge described probability matrix D meet default convergence Rule of judgment, upgrade described probability matrix A and described probability matrix B according to described probability matrix A ' and described probability matrix B '.

In a kind of specific implementation of the application, described according to described probability matrix A and described probability matrix B, predict that each user is to the interested probability of each information recorded in described default information bank, obtain probability matrix D, comprising:

According to following formula, predict that each user is to the interested probability of each information recorded in described default information bank, obtain probability matrix D,

$d_{ij}=\underset{k=1}{\overset{K}{\Σ}}{\mathrm{\α}}_{ik}{\mathrm{\β}}_{jk}$

Wherein, d _ijrepresent the element of described probability matrix D, α _ikrepresent the element that described probability matrix A i-th row kth arranges, β _jkrepresent the element that described probability matrix B jth row kth arranges, K represents the quantity of obtained information classification.

For achieving the above object, the embodiment of the present application discloses a kind of information push-delivery apparatus, and described device comprises:

Push request receiving module, for receiving the information pushing request for targeted customer;

Probability obtains module, for obtaining described targeted customer to the interested probability of each bar information recorded in the information bank preset from the User Information Database preset, wherein, described default User Information Database is for recording each user to the interested probability of each bar information recorded in described default information bank;

Information to be pushed determination module, for according to obtained probability order from high to low, determines information to be pushed in each bar information recorded from described default information database;

Info push module, for pushing described information to be pushed.

In a kind of specific implementation of the application, described information push-delivery apparatus also comprises:

Database generation module, for generating described default User Information Database;

Wherein, described database generation module, comprising:

Information classification obtains submodule, for obtaining current existing information classification;

First probability matrix obtains submodule, for obtaining current all users to the interested probability of each information classification above-mentioned, generating probability matrix A;

Second probability matrix obtains submodule, belongs to the probability of each information classification above-mentioned, generating probability matrix B for obtaining in described presupposed information storehouse each bar information recorded;

3rd probability matrix obtains submodule, for according to described probability matrix A and described probability matrix B, predicts that each user is to the interested probability of each information recorded in described default information bank, obtains probability matrix D;

Convergence judges submodule, for judging whether described probability matrix D meets the convergence Rule of judgment preset;

Probability matrix upgrades submodule, for judging that the judged result of submodule is no in described convergence, each element in described probability matrix D is adjusted according to the regulation rule preset, and upgrade described probability matrix D according to the element after adjustment, trigger described convergence and judge that submodule judges, until described probability matrix D meets described default convergence Rule of judgment;

Database generates submodule, for generating described default User Information Database according to described probability matrix D.

In a kind of specific implementation of the application, described convergence judges submodule, comprising:

Matrix decomposition unit, for according to the matrix decomposition algorithm preset, described probability matrix D is decomposed into probability matrix A ' and probability matrix B ', wherein, described probability matrix A ' is the matrix corresponding with described probability matrix A, and described probability matrix B ' is the matrix corresponding with described probability matrix B;

Convergence judging unit, for according to described probability matrix A ', described probability matrix B and described probability matrix D, judges whether described probability matrix D meets the convergence Rule of judgment preset.

In a kind of specific implementation of the application, described convergence judging unit, comprising:

Error calculation subelement, for according to the error argmin between following formula computational prediction probability and true probability _{α, β}l _(D),

${\mathrm{argmin}}_{\mathrm{\α},\mathrm{\β}}{L}_{\left(D\right)}=\underset{ij}{\Σ}{(d_{ij}-\stackrel{\→}{{\mathrm{\α}}_i}\·\stackrel{\→}{{\mathrm{\β}}_j})}^2+\mathrm{\λ}\left(|\stackrel{\→}{{\mathrm{\α}}_i}{|}^2+|\stackrel{\→}{{\mathrm{\β}}_j}{|}^2\right),$

Wherein, represent the vector of the element composition of described probability matrix A ' i-th row, represent the vector of the element composition of described probability matrix B ' jth row, λ represents regulation coefficient, d _ijrepresent the element of described probability matrix D;

Convergence judgment sub-unit, for according to described error argmin _{α, β}l _(D)judge whether described probability matrix D meets the convergence Rule of judgment preset.

In a kind of specific implementation of the application, described convergence judgment sub-unit,

Specifically for judging described error argmin _{α, β}l _(D)whether be less than the first default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment; Or

Specifically at described probability matrix D be according to adjustment after element upgrade after probability matrix, judge described error argmin _{α, β}l _(D)with error argmin _{α, β}l _(D)' between absolute difference whether be less than the second default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment, wherein, described error argmin _{α, β}l _(D)' represent based on the error that calculates of described probability matrix D before upgrading.

In a kind of specific implementation of the application, described database generation module, also comprises:

Matrix update submodule, for when judge described probability matrix D meet default convergence Rule of judgment, upgrade described probability matrix A and described probability matrix B according to described probability matrix A ' and described probability matrix B '.

In a kind of specific implementation of the application, described 3rd probability matrix obtains submodule, specifically for according to following formula, predicts that each user is to the interested probability of each information recorded in described default information bank, obtains probability matrix D,

$d_{ij}=\underset{k=1}{\overset{K}{\Σ}}{\mathrm{\α}}_{ik}{\mathrm{\β}}_{jk}$

Wherein, d _ijrepresent the element of described probability matrix D, α _ikrepresent the element that described probability matrix A i-th row kth arranges, β _jkrepresent the element that described probability matrix B jth row kth arranges, K represents the quantity of obtained information classification.

As seen from the above, in the scheme that the embodiment of the present application provides, according to current all users, each information recorded in the interested probability of existing information classification and default information bank is belonged to the probability of each information classification, predict that each user is to the interested probability of each information recorded in the information bank preset, and determine information to be pushed according to predicting the outcome, and then carry out information pushing.Visible, during the scheme determination information to be pushed that application the embodiment of the present application provides, consider user to the interested probability of each bar information, therefore, it is possible to carry out information pushing targetedly based on user, improve Consumer's Experience.

Accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present application or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only some embodiments of the application, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.

The schematic flow sheet of a kind of information-pushing method that Fig. 1 provides for the embodiment of the present application;

The schematic flow sheet of a kind of data library generating method that Fig. 2 provides for the embodiment of the present application;

The structural representation of a kind of information push-delivery apparatus that Fig. 3 provides for the embodiment of the present application;

The structural representation of a kind of database generating apparatus that Fig. 4 provides for the embodiment of the present application.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present application, be clearly and completely described the technical scheme in the embodiment of the present application, obviously, described embodiment is only some embodiments of the present application, instead of whole embodiments.Based on the embodiment in the application, those of ordinary skill in the art are not making the every other embodiment obtained under creative work prerequisite, all belong to the scope of the application's protection.

The schematic flow sheet of a kind of information-pushing method that Fig. 1 provides for the embodiment of the present application, the method comprises:

S101: receive the information pushing request for targeted customer.

At least need the mark of carrying targeted customer in above-mentioned information pushing request, certainly, can also carry other information in this request, the application does not limit this.

It should be noted that, information involved in the embodiment of the present application can be advertising message, news information, weather forecast information etc., and the application does not limit this.

S102: obtain targeted customer to the interested probability of each bar information recorded in the information bank preset from the User Information Database preset.

Wherein, the User Information Database preset is for recording each user to the interested probability of each bar information recorded in the information bank preset.

Concrete, can consider when generating above-mentioned default User Information Database in advance that current all users belong to the information such as the probability of each information classification to bar information each in the interested probability of existing information classification, default information bank, concrete generative process can embodiment shown in Figure 2, wouldn't describe in detail here.

Understandable, because customer demand etc. constantly changes, so the information recorded in the information bank preset also is constantly change, in addition, the hobby of user is not unalterable yet, based on above-mentioned several reason, need constantly update above-mentioned default User Information Database, such as, can be according to the above-mentioned default User Information Database of fixed time interval, as, within one day, upgrade once, renewal in a week once etc.

S103: according to obtained probability order from high to low, determines information to be pushed in each bar information recorded from the information database preset.

According to obtained probability, when determining information to be pushed from each bar information of the information database record preset, information to be pushed can be determined according to obtained probability order from high to low, preferentially can push its interested information to user like this.

In addition, when determining information to be pushed except the probability needing consideration to obtain, it is also conceivable to the number of times pushing information in default information bank in preset period of time to user, such as, today had pushed the information that in the probability obtained, probability is the highest to targeted customer, then information the highest for this probability can not be defined as information to be pushed, cause user to dislike to prevent repeating to push same information to targeted customer in a short time.

Certainly, it is also conceivable to the factors such as the information exposure rate of customer requirement, information pushing frequency when determining information to be pushed, the application does not limit this.

S104: push described information to be pushed.

The User Information Database preset related to is introduced in detail above below by specific embodiment.

The schematic flow sheet of a kind of data library generating method that Fig. 2 provides for the embodiment of the present application, the method comprises:

S201: obtain current existing information classification.

Concrete, current existing information classification can be that in actual application, operation and maintenance personnel artificially divides.

S202: obtain current all users to the interested probability of each information classification above-mentioned, generating probability matrix A.

Because probability obtained here is just as the probability of follow-up generating user information database, so in a kind of specific implementation of the application, current all users can be stochastic generation to the interested probability of each information classification above-mentioned.

All user current relative to above-mentioned acquisition is to the mode of the interested probability of each information classification above-mentioned, in the another kind of specific implementation of the application, can also within a period of time real-time statistics user for the number of operations of various information, such as, the number of times of operation reserve game, the number of times of installation leisure game, click number of times of strategy game advertisement etc., then obtain machine learning model corresponding to each information classification by existing machine learning algorithm, thus obtain current all users to the interested probability of each information classification above-mentioned.

It should be noted that, if the current all users of acquisition in the process of the interested probability of each information classification above-mentioned, cannot obtain a certain user to the interested probability of a certain information classification, can arrange this user to the interested probability of this information classification is zero.

It should be noted that, the application is just described for above-mentioned, obtains the mode of current all users to the interested probability of each information classification above-mentioned and be not limited in this in practical application.

S203: obtain the probability that each bar information recorded in presupposed information storehouse belongs to each information classification above-mentioned, generating probability matrix B.

Which information classification is each bar information recorded in presupposed information storehouse belong to can be artificial setting, also can according to feedack setting in user's browsing information process, in addition, each information recorded in presupposed information storehouse only can belong to an information classification, also can belong to multiple information classification.

It should be noted that, the each bar information recorded in acquisition presupposed information storehouse belongs in the process of the probability of each information classification above-mentioned, likely can find that a certain bar record does not belong to any one information classification, can arrange the probability that this record belongs to each information classification is in this case zero.

S204: according to probability matrix A and probability matrix B, predicts that each user is to the interested probability of each information recorded in the information bank preset, and obtains probability matrix D.

In a kind of specific implementation of the application, according to probability matrix A and probability matrix B, predict that each user is to the interested probability of each information recorded in the information bank preset, when obtaining probability matrix D, can according to following formula, predict that each user is to the interested probability of each information recorded in the information bank preset, and obtains probability matrix D

$d_{ij}=\underset{k=1}{\overset{K}{\Σ}}{\mathrm{\α}}_{ik}{\mathrm{\β}}_{jk}$

Wherein, d _ijrepresent the element of probability matrix D, concrete, d _ijfor the element of the i-th row jth row in probability matrix D, α _ikrepresent the element of probability matrix A i-th row kth row, β _jkrepresent the element of probability matrix B jth row kth row, K represents the quantity of obtained information classification.

S205: judge whether probability matrix D meets the convergence Rule of judgment preset, if do not meet, meets if perform S206, performs S207.

Because probability matrix D records in advance according to probability matrix A and probability matrix B, so, can consider to judge whether probability matrix D meets the convergence Rule of judgment preset by matrix decomposition mode.Concrete, when judging whether probability matrix D meets the convergence Rule of judgment preset, can according to the matrix decomposition algorithm preset, probability matrix D is decomposed into probability matrix A ' and probability matrix B ', then according to probability matrix A ', probability matrix B and probability matrix D, judge whether probability matrix D meets the convergence Rule of judgment preset.

Wherein, probability matrix A ' is the matrix corresponding with probability matrix A, and probability matrix B ' is the matrix corresponding with probability matrix B.

Concrete, according to probability matrix A ', probability matrix B ' and probability matrix D, when judging whether probability matrix D meets the convergence Rule of judgment preset, can first according to the error argmin between following formula computational prediction probability and true probability _{α, β}l _(D), then according to error argmin _{α, β}l _(D)judge whether probability matrix D meets the convergence Rule of judgment preset.

Wherein, the expression formula related to above is:

${\mathrm{argmin}}_{\mathrm{\α},\mathrm{\β}}{L}_{\left(D\right)}=\underset{ij}{\Σ}{(d_{ij}-\stackrel{\→}{{\mathrm{\α}}_i}\·\stackrel{\→}{{\mathrm{\β}}_j})}^2+\mathrm{\λ}\left(|\stackrel{\→}{{\mathrm{\α}}_i}{|}^2+|\stackrel{\→}{{\mathrm{\β}}_j}{|}^2\right),$

represent the vector of the element composition of probability matrix A ' i-th row, represent the vector of the element composition of probability matrix B ' jth row, λ represents regulation coefficient, d _ijrepresent the element of probability matrix D, concrete, d _ijrepresent the element of probability matrix D i-th row jth row.

In a kind of specific implementation of the application, according to error argmin _{α, β}l _(D)can judge according to following several situation when judging whether probability matrix D meets the convergence Rule of judgment preset:

The first situation: error in judgement argmin _{α, β}l _(D)whether be less than the first default error threshold, if be less than, decision probability matrix D meets default convergence Rule of judgment;

The second situation: when probability matrix D be according to adjustment after element upgrade after probability matrix, error in judgement argmin _{α, β}l _(D)with error argmin _{α, β}l _(D)' between absolute difference whether be less than the second default error threshold, if be less than, decision probability matrix D meets default convergence Rule of judgment, wherein, error argmin _{α, β}l _(D)' represent based on the error that calculates of probability matrix D before upgrading.

S206: according to each element in the regulation rule adjustment probability matrix D preset, and according to the element update probability matrix D after adjustment, return S205.

S207: generate the User Information Database preset according to probability matrix D.

It should be noted that, the probability matrix D related in this step can be pass through the probability matrix D upgraded, the probability matrix D after also can being through once or repeatedly adjusting.

In the optional implementation of the one of the application, when judge probability matrix D meet default convergence Rule of judgment, can also according to probability matrix A ' and probability matrix B ' update probability matrix A and probability matrix B, be equivalent to so oppositely have updated each user current to the interested probability of each information classification, the information that simultaneously have updated in default information bank belongs to the probability of each information classification, especially for belonging to the information of any one information classification before, by the renewal of probability matrix B, the classification to this information can be realized, in addition, upgrade above-mentioned probability matrix A and above-mentioned probability matrix B and can be conducive to the default User Information Database of follow-up renewal.

Further, although can find out the process generating the User Information Database preset from foregoing description needs a certain amount of each bar information to belong to the probabilistic information of each information classification as initial value, but to these data volumes as the probability of initial value and accuracy not requirement, so, when information being classified in practical application, even if there is the phenomenon of grouped data scarcity and classification error, the User Information Database generating and preset also substantially can not be affected.

As seen from the above, in the scheme that each embodiment above-mentioned provides, according to current all users, each information recorded in the interested probability of existing information classification and default information bank is belonged to the probability of each information classification, predict that each user is to the interested probability of each information recorded in the information bank preset, and determine information to be pushed according to predicting the outcome, and then carry out information pushing.Visible, when applying the scheme determination information to be pushed that each embodiment above-mentioned provides, consider user to the interested probability of each bar information, therefore, it is possible to carry out information pushing targetedly based on user, improve Consumer's Experience.

Corresponding with above-mentioned information-pushing method, the embodiment of the present application additionally provides a kind of information push-delivery apparatus.

The structural representation of a kind of information push-delivery apparatus that Fig. 3 provides for the embodiment of the present application, this device comprises:

Push request receiving module 301, for receiving the information pushing request for targeted customer;

Probability obtains module 302, for obtaining described targeted customer to the interested probability of each bar information recorded in the information bank preset from the User Information Database preset, wherein, described default User Information Database is for recording each user to the interested probability of each bar information recorded in described default information bank;

Information to be pushed determination module 303, for according to obtained probability order from high to low, determines information to be pushed in each bar information recorded from described default information database;

Info push module 304, for pushing described information to be pushed.

In a kind of specific implementation of the application, above-mentioned information push-delivery apparatus can also comprise:

Database generation module, for generating described default User Information Database.

The User Information Database preset how generating and relate to is introduced in detail above below by specific embodiment.

The structural representation of a kind of database generating apparatus that Fig. 4 provides for the embodiment of the present application, this device is the concrete device of one of database generation module, comprising:

Information classification obtains submodule 401, for obtaining current existing information classification;

First probability matrix obtains submodule 402, for obtaining current all users to the interested probability of each information classification above-mentioned, generating probability matrix A;

Second probability matrix obtains submodule 403, belongs to the probability of each information classification above-mentioned, generating probability matrix B for obtaining in described presupposed information storehouse each bar information recorded;

3rd probability matrix obtains submodule 404, for according to described probability matrix A and described probability matrix B, predicts that each user is to the interested probability of each information recorded in described default information bank, obtains probability matrix D;

Convergence judges submodule 405, for judging whether described probability matrix D meets the convergence Rule of judgment preset;

Probability matrix upgrades submodule 406, for judging that the judged result of submodule 405 is no in described convergence, each element in described probability matrix D is adjusted according to the regulation rule preset, and upgrade described probability matrix D according to the element after adjustment, trigger described convergence and judge that submodule 405 judges, until described probability matrix D meets described default convergence Rule of judgment;

Database generates submodule 407, for generating described default User Information Database according to described probability matrix D.

Concrete, described convergence judges that submodule 405 can comprise:

Matrix decomposition unit, for according to the matrix decomposition algorithm preset, described probability matrix D is decomposed into probability matrix A ' and probability matrix B ', wherein, described probability matrix A ' is the matrix corresponding with described probability matrix A, and described probability matrix B ' is the matrix corresponding with described probability matrix B;

Convergence judging unit, for according to described probability matrix A ', described probability matrix B and described probability matrix D, judges whether described probability matrix D meets the convergence Rule of judgment preset.

Concrete, described convergence judging unit can comprise:

Error calculation subelement, for according to the error argmin between following formula computational prediction probability and true probability _{α, β}l _(D),

${\mathrm{argmin}}_{\mathrm{\α},\mathrm{\β}}{L}_{\left(D\right)}=\underset{ij}{\Σ}{(d_{ij}-\stackrel{\→}{{\mathrm{\α}}_i}\·\stackrel{\→}{{\mathrm{\β}}_j})}^2+\mathrm{\λ}\left(|\stackrel{\→}{{\mathrm{\α}}_i}{|}^2+|\stackrel{\→}{{\mathrm{\β}}_j}{|}^2\right),$

Wherein, represent the vector of the element composition of described probability matrix A ' i-th row, represent the vector of the element composition of described probability matrix B ' jth row, λ represents regulation coefficient, d _ijrepresent the element of described probability matrix D;

Convergence judgment sub-unit, for according to described error argmin _{α, β}l _(D)judge whether described probability matrix D meets the convergence Rule of judgment preset.

Concrete, described convergence judgment sub-unit,

Can specifically for judging described error argmin _{α, β}l _(D)whether be less than the first default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment; Or

Can specifically at described probability matrix D be according to adjustment after element upgrade after probability matrix, judge described error argmin _{α, β}l _(D)with error argmin _{α, β}l _(D)' between absolute difference whether be less than the second default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment, wherein, described error argmin _{α, β}l _(D)' represent based on the error that calculates of described probability matrix D before upgrading.

In the better implementation of the one of the application, described database generation module can also comprise:

Matrix update submodule, for when judge described probability matrix D meet default convergence Rule of judgment, upgrade described probability matrix A and described probability matrix B according to described probability matrix A ' and described probability matrix B '.

Concrete, described 3rd probability matrix obtains submodule, specifically for according to following formula, can predict that each user is to the interested probability of each information recorded in described default information bank, obtains probability matrix D,

$d_{ij}=\underset{k=1}{\overset{K}{\Σ}}{\mathrm{\α}}_{ik}{\mathrm{\β}}_{jk}$

Wherein, d _ijrepresent the element of described probability matrix D, α _ikrepresent the element that described probability matrix A i-th row kth arranges, β _jkrepresent the element that described probability matrix B jth row kth arranges, K represents the quantity of obtained information classification.

As seen from the above, in the scheme that each embodiment above-mentioned provides, according to current all users, each information recorded in the interested probability of existing information classification and default information bank is belonged to the probability of each information classification, predict that each user is to the interested probability of each information recorded in the information bank preset, and determine information to be pushed according to predicting the outcome, and then carry out information pushing.Visible, when applying the scheme determination information to be pushed that each embodiment above-mentioned provides, consider user to the interested probability of each bar information, therefore, it is possible to carry out information pushing targetedly based on user, improve Consumer's Experience.

For device embodiment, because it is substantially similar to embodiment of the method, so description is fairly simple, relevant part illustrates see the part of embodiment of the method.

It should be noted that, in this article, the such as relational terms of first and second grades and so on is only used for an entity or operation to separate with another entity or operational zone, and not necessarily requires or imply the relation that there is any this reality between these entities or operation or sequentially.And, term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability, thus make to comprise the process of a series of key element, method, article or equipment and not only comprise those key elements, but also comprise other key elements clearly do not listed, or also comprise by the intrinsic key element of this process, method, article or equipment.When not more restrictions, the key element limited by statement " comprising ... ", and be not precluded within process, method, article or the equipment comprising described key element and also there is other identical element.

One of ordinary skill in the art will appreciate that all or part of step realized in said method embodiment is that the hardware that can carry out instruction relevant by program has come, described program can be stored in computer read/write memory medium, here the alleged storage medium obtained, as: ROM/RAM, magnetic disc, CD etc.

The foregoing is only the preferred embodiment of the application, be not intended to limit the protection domain of the application.Any amendment done within all spirit in the application and principle, equivalent replacement, improvement etc., be all included in the protection domain of the application.

Claims (10)

1. an information-pushing method, is characterized in that, described method comprises:

Receive the information pushing request for targeted customer;

Described targeted customer is obtained to the interested probability of each bar information recorded in the information bank preset from the User Information Database preset, wherein, described default User Information Database is for recording each user to the interested probability of each bar information recorded in described default information bank;

According to obtained probability order from high to low, in each bar information recorded from described default information database, determine information to be pushed;

Push described information to be pushed.

2. method according to claim 1, is characterized in that, generates described default User Information Database in such a way:

Obtain current existing information classification;

Obtain current all users to the interested probability of each information classification above-mentioned, generating probability matrix A;

Obtain in described presupposed information storehouse the probability that each bar information recorded belongs to each information classification above-mentioned, generating probability matrix B;

According to described probability matrix A and described probability matrix B, predict that each user is to the interested probability of each information recorded in described default information bank, obtain probability matrix D;

Judge whether described probability matrix D meets the convergence Rule of judgment preset, if do not meet, each element in described probability matrix D is adjusted according to the regulation rule preset, and upgrade described probability matrix D according to the element after adjustment, return the described step whether described probability matrix D meets the convergence Rule of judgment preset that judges, until described probability matrix D meets described default convergence Rule of judgment;

Described default User Information Database is generated according to described probability matrix D.

3. method according to claim 2, is characterized in that, describedly judges that whether described probability matrix D meets the convergence Rule of judgment preset, and comprising:

According to the matrix decomposition algorithm preset, described probability matrix D is decomposed into probability matrix A ' and probability matrix B ', wherein, described probability matrix A ' is the matrix corresponding with described probability matrix A, and described probability matrix B ' is the matrix corresponding with described probability matrix B;

According to described probability matrix A ', described probability matrix B and described probability matrix D, judge whether described probability matrix D meets the convergence Rule of judgment preset.

4. method according to claim 3, is characterized in that, described according to described probability matrix A ', described probability matrix B ' and described probability matrix D, judges whether described probability matrix D meets the convergence Rule of judgment preset, and comprising:

According to the error argmin between following formula computational prediction probability and true probability _{α, β}l _(D),

${\mathrm{argmin}}_{\mathrm{\α},\mathrm{\β}}{L}_{\left(D\right)}=\underset{ij}{\Σ}{(d_{ij}-\stackrel{\→}{{\mathrm{\α}}_i}\·\stackrel{\→}{{\mathrm{\β}}_j})}^2+\mathrm{\λ}\left(\right|\stackrel{\→}{{\mathrm{\α}}_i}|\·|\stackrel{\→}{{\mathrm{\β}}_j}{|}^2),$

Wherein, represent the vector of the element composition of described probability matrix A ' i-th row, represent the vector of the element composition of described probability matrix B ' jth row, λ represents regulation coefficient, d _ijrepresent the element of described probability matrix D;

According to described error argmin _{α, β}l _(D)judge whether described probability matrix D meets the convergence Rule of judgment preset.

5. method according to claim 4, is characterized in that, described according to described error argmin _{α, β}l _(D)judge whether described probability matrix D meets the convergence Rule of judgment preset, and comprising:

Judge described error argmin _{α, β}l _(D)whether be less than the first default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment; Or

When described probability matrix D be according to adjustment after element upgrade after probability matrix, judge described error argmin _{α, β}l _(D)with error argmin _{α, β}l _(D)' between absolute difference whether be less than the second default error threshold, if be less than, judge that described probability matrix D meets default convergence Rule of judgment, wherein, described error argmin _{α, β}l _(D)' represent based on the error that calculates of described probability matrix D before upgrading.

6. the method according to any one of claim 3-5, is characterized in that, described method also comprises:

When judge described probability matrix D meet default convergence Rule of judgment, upgrade described probability matrix A and described probability matrix B according to described probability matrix A ' and described probability matrix B '.

7. method according to claim 2, is characterized in that, described according to described probability matrix A and described probability matrix B, predicts that each user is to the interested probability of each information recorded in described default information bank, obtains probability matrix D, comprising:

According to following formula, predict that each user is to the interested probability of each information recorded in described default information bank, obtain probability matrix D,

$d_{ij}=\underset{k=1}{\overset{K}{\Σ}}{\mathrm{\α}}_{ik}{\mathrm{\β}}_{jk}$

Wherein, d _ijrepresent the element of described probability matrix D, α _ikrepresent the element that described probability matrix A i-th row kth arranges, β _jkrepresent the element that described probability matrix B jth row kth arranges, K represents the quantity of obtained information classification.

8. an information push-delivery apparatus, is characterized in that, described device comprises:

Push request receiving module, for receiving the information pushing request for targeted customer;

Probability obtains module, for obtaining described targeted customer to the interested probability of each bar information recorded in the information bank preset from the User Information Database preset, wherein, described default User Information Database is for recording each user to the interested probability of each bar information recorded in described default information bank;

Information to be pushed determination module, for according to obtained probability order from high to low, determines information to be pushed in each bar information recorded from described default information database;

Info push module, for pushing described information to be pushed.

9. device according to claim 8, is characterized in that, described device also comprises:

Database generation module, for generating described default User Information Database;

Wherein, described database generation module, comprising:

Information classification obtains submodule, for obtaining current existing information classification;

First probability matrix obtains submodule, for obtaining current all users to the interested probability of each information classification above-mentioned, generating probability matrix A;

Second probability matrix obtains submodule, belongs to the probability of each information classification above-mentioned, generating probability matrix B for obtaining in described presupposed information storehouse each bar information recorded;

3rd probability matrix obtains submodule, for according to described probability matrix A and described probability matrix B, predicts that each user is to the interested probability of each information recorded in described default information bank, obtains probability matrix D;

Convergence judges submodule, for judging whether described probability matrix D meets the convergence Rule of judgment preset;

Probability matrix upgrades submodule, for judging that the judged result of submodule is no in described convergence, each element in described probability matrix D is adjusted according to the regulation rule preset, and upgrade described probability matrix D according to the element after adjustment, trigger described convergence and judge that submodule judges, until described probability matrix D meets described default convergence Rule of judgment;

Database generates submodule, for generating described default User Information Database according to described probability matrix D.

10. device according to claim 9, is characterized in that, described convergence judges submodule, comprising:

Matrix decomposition unit, for according to the matrix decomposition algorithm preset, described probability matrix D is decomposed into probability matrix A ' and probability matrix B ', wherein, described probability matrix A ' is the matrix corresponding with described probability matrix A, and described probability matrix B ' is the matrix corresponding with described probability matrix B;

Convergence judging unit, for according to described probability matrix A ', described probability matrix B and described probability matrix D, judges whether described probability matrix D meets the convergence Rule of judgment preset.