CN106294497A - Information recommendation method and device - Google Patents
Information recommendation method and device Download PDFInfo
- Publication number
- CN106294497A CN106294497A CN201510312846.3A CN201510312846A CN106294497A CN 106294497 A CN106294497 A CN 106294497A CN 201510312846 A CN201510312846 A CN 201510312846A CN 106294497 A CN106294497 A CN 106294497A
- Authority
- CN
- China
- Prior art keywords
- algorithm
- attention rate
- user
- targeted customer
- recommendation information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention is about a kind of information recommendation method and device, belongs to networking technology area.Described method includes: obtain the user characteristics of targeted customer；User characteristics according to targeted customer is inquired about the attention rate curve of at least two algorithm respectively and is obtained at least two attention rate；Size based at least two attention rate, determines that at least two algorithm at least one algorithm is as target algorithm；The recommendation information of targeted customer is determined according to target algorithm；The recommendation information of targeted customer is sent to the terminal of targeted customer.By the size of attention rate, the present invention determines that the recommendation information generated according to this target algorithm, as target algorithm, and is recommended user by an algorithm；Solve what the algorithm that recommendation information in correlation technique used was typically to pre-set, it is recommended that the motility of information is relatively low, the problem that specific aim is poor；Reach the size according to attention rate and determined target algorithm, it is recommended that the effect that the specific aim of information is stronger.
Description
Technical field
The present invention relates to networking technology area, particularly to a kind of information recommendation method and device.
Background technology
At present, user is protruded day by day for the individual demand of internet product or service, various information recommendations
Method (to the method for the recommendation information that user recommends) is widely used in meeting the individual demand of user.
Information recommendation method in correlation technique generally obtains recommendation information by polyalgorithm, and by acquisition
Recommendation information is sent to the terminal of user.Example, the plurality of algorithm can include collaborative filtering
(Collaborative Filtering) and content-based filtering algorithm, wherein, collaborative filtering can according to
The content (such as history viewing record etc.) of the historical record of the user higher with targeted customer's similarity is come for mesh
Mark user recommends the content similar with the content of this historical record；Content-based filtering algorithm is according to target
The content (such as history viewing record etc.) of the historical record of user is recommended and this historical record for targeted customer
The similar content of content.
Inventor, during realizing the present invention, finds that aforesaid way at least has following defects that relevant skill
In art, it is recommended that the algorithm that information is used is typically to pre-set, the kind of algorithm during information recommendation
Class is single, therefore, it is recommended that the motility of information is relatively low, specific aim is poor.
Summary of the invention
In order to solve the problem in correlation technique, embodiments provide a kind of information recommendation method and dress
Put.Described technical scheme is as follows:
On the one hand, it is provided that a kind of information recommendation method, described method includes:
Obtaining the user characteristics of targeted customer, described user characteristics is for characterizing the characteristic of user；
User characteristics according to described targeted customer inquire about respectively the attention rate curve of at least two algorithm obtain to
Few two attention rates, wherein, the attention rate curve of arbitrary algorithm is for recording the use using described algorithm to obtain
Corresponding relation between family feature and attention rate, described attention rate is paid close attention to by user for characterizing preset content
Degree；
Size based on described at least two attention rate, determines that at least one is calculated in described at least two algorithm
Method is as target algorithm；
The recommendation information of described targeted customer is determined according to described target algorithm；
The recommendation information of described targeted customer is sent to the terminal of described targeted customer.
On the other hand, it is provided that a kind of information recommending apparatus, described device includes:
Acquiring unit, for obtaining the user characteristics of targeted customer, described user characteristics is for characterizing user's
Characteristic；
Query unit, for inquiring about the pass of at least two algorithm respectively according to the user characteristics of described targeted customer
Noting line of writing music and obtain at least two attention rate, wherein, the attention rate curve of arbitrary algorithm is used for recording employing institute
State the corresponding relation between user characteristics and the attention rate that algorithm obtains, in described attention rate is used for characterizing and presets
Hold the degree of concern of user；
Algorithm determines unit, for size based on described at least two attention rate, calculates in described at least two
Method determining, at least one algorithm is as target algorithm；
Information determination unit, for determining the recommendation information of described targeted customer according to described target algorithm；
Transmitting element, for sending the recommendation information of described targeted customer to the terminal of described targeted customer.
The technical scheme that the embodiment of the present invention provides has the benefit that
Determine that an algorithm, and will be raw according to this target algorithm as target algorithm by the size of attention rate
The recommendation information become recommends user；Solve the algorithm that recommendation information in correlation technique used and be typically pre-
First arrange, it is recommended that the motility of information is relatively low, the problem that specific aim is poor；Reached according to attention rate is big
Little determine target algorithm, it is recommended that the effect that the specific aim of information is stronger.
It should be appreciated that it is only exemplary and explanatory that above general description and details hereinafter describe,
The present invention can not be limited.
Accompanying drawing explanation
Accompanying drawing herein is merged in description and constitutes the part of this specification, it is shown that meet the present invention
Embodiment, and for explaining the principle of the present invention together with description.
Fig. 1 is the signal of the implementation environment involved by information recommendation method provided in section Example of the present invention
Figure；
Fig. 2 is the schematic flow sheet of a kind of information recommendation method that the embodiment of the present invention provides；
Fig. 3-1 is the schematic flow sheet of the another kind of information recommendation method that the embodiment of the present invention provides；
Fig. 3-2 is the attention rate curve chart of two kinds of algorithms in Fig. 3-1 illustrated embodiment；
Fig. 4-1 is the schematic flow sheet of the another kind of information recommendation method that the embodiment of the present invention provides；
Fig. 4-2 is the flow chart arranging weights in Fig. 4-1 illustrated embodiment；
Fig. 5 is the block diagram of a kind of information recommending apparatus shown in the embodiment of the present invention；
Fig. 6-1 is the block diagram of the another kind of information recommending apparatus shown in the embodiment of the present invention；
Fig. 6-2 is the block diagram that in Fig. 6-1 illustrated embodiment, algorithm determines unit；
Fig. 7 is the block diagram of a kind of device shown in the embodiment of the present invention.
By above-mentioned accompanying drawing, it has been shown that the embodiment that the present invention is clear and definite, hereinafter will be described in more detail.
These accompanying drawings and word are described and are not intended to be limited by any mode the scope of present inventive concept, but logical
Crossing with reference to specific embodiment is that those skilled in the art illustrate idea of the invention.
Detailed description of the invention
Here will illustrate exemplary embodiment in detail, its example represents in the accompanying drawings.Following retouches
Stating when relating to accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represents same or analogous key element.
Embodiment described in following exemplary embodiment does not represent all embodiment party consistent with the present invention
Formula.On the contrary, they only with describe in detail in appended claims, the present invention some in terms of mutually one
The example of the apparatus and method caused.
Refer to Fig. 1, the reality involved by information recommendation method provided in section Example of the present invention is provided
Execute the schematic diagram of environment.This implementation environment may include that server 110 and at least one terminal 120.
Server 110 can be a station server, or the server cluster being made up of some station servers,
Or a cloud computing service center.Terminal 120 can be smart mobile phone, computer, multimedia player,
Electronic reader, Wearable device etc..
Can be set up by cable network or wireless network between server 110 and terminal 120 and connect.
Fig. 2 is the schematic flow sheet of a kind of information recommendation method that the embodiment of the present invention provides, the present embodiment with
The server 110 that this information recommendation is applied in implementation environment shown in Fig. 1 illustrates.This information recommendation
Method may include that
Step 201, the user characteristics of acquisition targeted customer, this user characteristics is for characterizing the characteristic of user.
Step 202, user characteristics according to targeted customer are inquired about the attention rate curve of at least two algorithm respectively and are obtained
To at least two attention rate, wherein, the attention rate curve of arbitrary algorithm uses this algorithm to obtain for recording
Corresponding relation between user characteristics and attention rate, this attention rate is paid close attention to by user for characterizing preset content
Degree.
Step 203, size based at least two attention rate, determine that at least one is calculated at least two algorithm
Method is as target algorithm.
Step 204, determine the recommendation information of targeted customer according to target algorithm.
Step 205, to targeted customer terminal send targeted customer recommendation information.
In sum, the information recommendation method that the embodiment of the present invention provides, determined by the size of attention rate
One algorithm is as target algorithm, and the recommendation information generated according to this target algorithm is recommended user；Solve
The algorithm that in correlation technique of having determined, recommendation information is used is typically to pre-set, the motility of recommendation information
Relatively low, that specific aim is poor problem；Reach the size according to attention rate and determined target algorithm, it is recommended that letter
The stronger effect of specific aim of breath.
Optionally, size based at least two attention rate, at least two algorithm, determine that at least one is calculated
Method as target algorithm, including:
Determine attention rate maximum at least two attention rate.
Algorithm corresponding for maximum attention rate is defined as target algorithm.
Optionally, pay close attention to the size of angle value based at least two, at least two algorithm, determine at least one
Algorithm as target algorithm, including:
For each algorithm at least two algorithm, weights are set, wherein, the weights of arbitrary algorithm with according to mesh
The size positive correlation of the concern angle value that the attention rate curve of the user characteristics arbitrary algorithm of inquiry of mark user obtains,
Weights represent that the recommendation information obtained according to the algorithm with weights is shared in the recommendation information of targeted customer
Proportion.
Hybrid algorithm at least two algorithm combination with respective weights obtained is as target algorithm.
Optionally, weights are set for each algorithm at least two algorithm, including:
At least two attention rate is normalized, obtains the normalized value that each attention rate is corresponding, often
Individual normalized value belongs to [0,1], and normalized value sum corresponding to this at least two attention rate is 1.
Using normalized value corresponding for each attention rate as the weights of algorithm corresponding to each attention rate.
Optionally, inquire about the attention rate curve of at least two algorithm respectively according to the user characteristics of targeted customer to obtain
Before at least two attention rate, method also includes:
Obtaining the data acquisition system of at least two algorithm in historical time section, data acquisition system record has that to meet user special
The attention rate of the user the levied recommendation information to being obtained by each algorithm.
The attention rate curve of at least two algorithm is obtained according to data acquisition system.
Optionally, user characteristics includes that user's liveness, user's liveness are user's point in preset time period
Hit the number of times of preset content.
Optionally, at least two algorithm includes: collaborative filtering and social proposed algorithm, collaborative filtering
The content of the historical record of user higher with targeted customer's similarity according to algorithm is recommended for targeted customer
The algorithm of the content similar with the content of historical record, social proposed algorithm be by social platform with mesh
The content of the historical record of the user of mark user-association determines the algorithm of recommendation information.
Optionally, attention rate includes: any one in conversion ratio, clicking rate and thousand advertising incomes ECPM
Kind, conversion ratio is the visit capacity ratio with total visit capacity of preset content, and clicking rate is that preset content is clicked
The ratio of number of times and shown number of times, ECPM is that preset content shows the advertising income obtained each thousand times.
In sum, the information recommendation method that the embodiment of the present invention provides, determined by the size of attention rate
One algorithm is as target algorithm, and the recommendation information generated according to this target algorithm is recommended user；Solve
The algorithm that in correlation technique of having determined, recommendation information is used is typically to pre-set, the motility of recommendation information
Relatively low, that specific aim is poor problem；Reach the size according to attention rate and determined target algorithm, it is recommended that letter
The stronger effect of specific aim of breath.
Fig. 3-1 is the schematic flow sheet of the another kind of information recommendation method that the embodiment of the present invention provides, the present embodiment
Illustrate with the server 110 that this information recommendation is applied in implementation environment shown in Fig. 1.This information pushes away
The method of recommending may include that
Step 301, the user characteristics of acquisition targeted customer, this user characteristics is for characterizing the characteristic of user.
When needs carry out information recommendation to targeted customer, can first obtain the user characteristics of targeted customer.
This user characteristics can include that user's liveness, user's liveness are that user clicks in preset time period
The number of times of preset content, preset content can be film, books, Shopping directery etc..Additionally, user is special
Levy at least one that can also include the age of user, occupation, sex and region.
The data acquisition system of at least two algorithm in step 302, acquisition historical time section, data acquisition system record has symbol
The attention rate of the user of the conjunction user characteristics recommendation information to being obtained by each algorithm.
The data acquisition system of at least two algorithm, the arbitrary algorithm in the embodiment of the present invention in acquisition historical time section
For having the algorithm of time stability, i.e. meet the user of the user characteristics recommendation to being obtained by this algorithm
The change over time of the curve of the attention rate of breath is less, and curve is more steady, and this at least two algorithm is permissible
Including: collaborative filtering and social proposed algorithm.
The content of the historical record of user higher with targeted customer's similarity according to collaborative filtering come for
Targeted customer recommends the algorithm of the content similar with the content of historical record, higher with targeted customer's similarity
User can include identity information (such as showing the log-on message etc. of the user interest) similarity with targeted customer
Higher user.
Additionally, can also include and the similarity of targeted customer's behavior over the similarity of targeted customer.
As a example by viewing video, (first user is except targeted customer with first user can to pass through comparison object user
Outer any user) within the past period, the similarity of viewing record determines that targeted customer uses with first
The similarity at family.Concrete comparative approach can be: if targeted customer and first user have all seen a certain portion/class
Film, then can judge targeted customer and the according to targeted customer and first user to the scoring of this video
Whether one user likes this portion/class film, and if marking is 1～5 point, 5 points of expressions like best, and 1 point represents least
Liking, the similarity of the marking of the video jointly seen them using targeted customer and first user is as target
User and the similarity of first user, can obtain the user higher with targeted customer's similarity according to the method.
In social proposed algorithm is the historical record by the user associated in social platform with targeted customer
Hold the algorithm determining recommendation information.
It should be noted that at least two algorithm can also include other algorithm more, the present invention implements
Example does not restricts.
Wherein, attention rate is for characterizing the preset content degree of concern by user, and this attention rate may include that
Conversion ratio, clicking rate and thousand advertising incomes are (English: effective cost per mille；It is called for short: ECPM)
In any one, this conversion ratio is the ratio of visit capacity and total visit capacity of preset content, such as, with one
Portion's video is preset content, then this conversion ratio is the viewing number of times ratio with total visit capacity of video website of video
Value, this ratio is the biggest, and to represent this video the highest by the attention rate of user.This clicking rate is that preset content is by point
Hitting the ratio of number of times and shown number of times, such as, with a video link as preset content, then this clicking rate is
The clicked number of times of this video link and this video link are shown the ratio of number of times, and the biggest representative of this ratio should
Video link is the highest by the degree of concern of user.This ECPM is that preset content shows the wide of acquisition each thousand times
Accusing income, such as preset content is a video, then ECPM represents that this video is shown for each thousand times, and this regards
The expense that the advertiser of the most subsidiary advertisement is paid, this expense can also reflect that from side this video is subject to
The degree of concern of user.
Additionally, the acquisition process of data acquisition system can be to test the customer group including a large number of users
Arriving, customer group can also be made up of a part of user in historical record.Exemplary, with clicking rate table
Levy attention rate, be the recommendation that customer group carries out information according to collaborative filtering, obtain in customer group afterwards and use
Family is for the clicking rate of recommendation information, after this, or at this simultaneously, according to social proposed algorithm come for
In customer group, user carries out the recommendation of information, and obtains in customer group user for the clicking rate of recommendation information,
Method obtains the data acquisition system of all algorithms according to this.Concrete, can be by A/B's test (A/B test)
Mode obtains data acquisition system, and A/B test is a kind of gray scale published method, gray scale issue refer to B&W it
Between, it is possible to a kind of published method seamlessly transitted.A/B test is to allow a part of user continue by A, a part
User starts with B, if what opposing views user does not has to B, then progressively expanded scope, all
User moves to come above B.Gray scale issue can ensure that stablizing of total system, initial gray time
Wait just it appeared that, adjust problem, to ensure its disturbance degree.In embodiments of the present invention, survey can be set up
The examination page, and user recommends to include the survey of the recommendation information obtained according to algorithms of different at random in customer group
The examination page, and record the attention rate of user.When the mode using A/B test obtains data acquisition system, need
Want the number of user in customer group more.It should be noted that when carrying out test and obtaining data acquisition system, and
Need not carry out completely the recommendation of information to same multiple users, it is only necessary to the user characteristics of user is identical i.e.
Can.
It addition, the user in customer group can be equally distributed according to user characteristics, in such as customer group
User has 1000, and when characterizing user characteristics with user's liveness, these 1000 users can include 200
Individual liveness is the user of 20, and 200 liveness are the user of 30, and 200 liveness are the user of 40,
200 liveness are the user of 50, and 200 liveness are the user of 60, wherein, when preset content is electricity
During shadow, user's liveness can be defined as over one month in the viewing quantity of user, one month the most in the past
One user watched 20 films (generally repeating to watch same portion film to be denoted as have viewed a film), then
User's liveness of this user is 20, and user's liveness can also otherwise be defined, and such as passes by
The viewing duration etc. of user in one month, the embodiment of the present invention does not restricts.
Step 303, according to data acquisition system obtain at least two algorithm attention rate curve.
After obtaining data acquisition system, the attention rate curve of at least two algorithm can be obtained according to this data acquisition system,
The attention rate curve of arbitrary algorithm may be used for record and uses between user characteristics and the attention rate that this algorithm obtains
Corresponding relation.Exemplary, as shown in figure 3-2, it is clicking rate for attention rate, and user characteristics is user
During liveness, the attention rate curve of collaborative filtering and social proposed algorithm.In Fig. 3-2, worked in coordination with
The attention rate curve q2 of attention rate curve q1 and the social proposed algorithm of filter algorithm intersects at an A, with
When family liveness is less than user's liveness of some A, the clicking rate of the recommendation information that collaborative filtering obtains is little
Clicking rate in the recommendation information that social proposed algorithm obtains；At user's liveness, the user more than some A lives
During jerk, the clicking rate of the recommendation information that collaborative filtering obtains is more than pushing away that social proposed algorithm obtains
Recommend the clicking rate of information it can be understood as, when the historical record relatively horn of plenty of user, collaborative filtering
Performance be better than social proposed algorithm, and when the historical record of user is more rare, collaborative filtering
Performance be weaker than social proposed algorithm.Attention rate curve can be clear and definite be shown with user during each algorithm
Feature and the corresponding relation of attention rate, this corresponding relation is typically relatively stable but it also may every in advance
The attention rate curve of at least two algorithm of interval acquiring of fixing time, the embodiment of the present invention does not restricts.
It should be noted that step 302 can also perform to step 303 before step 301, the present invention
Embodiment does not restricts.
Step 304, user characteristics according to targeted customer are inquired about the attention rate curve of at least two algorithm respectively and are obtained
To at least two attention rate.
After obtaining the attention rate curve of at least two algorithm, can be according to the user of the targeted customer obtained
Feature obtains this user characteristics at least two corresponding in the attention rate curve of at least two algorithm and pays close attention to
Degree.As a example by the attention rate curve of two algorithms represented by Fig. 3-2, if the user characterizing user characteristics enlivens
Degree is 20, then the clicking rate inquiring about the sign attention rate that the attention rate curve q1 of collaborative filtering obtains is
0.16, the clicking rate of the sign attention rate that the attention rate curve q2 of inquiry social proposed algorithm obtains is 0.19.
Step 305, determine at least two attention rate maximum attention rate.
After having obtained at least two attention rate, it may be determined that go out attention rate maximum at least two attention rate.
It should be noted that maximum concern can be all defined as by differing a range of multiple attention rates
Degree, the attention rate such as obtained according to user characteristics five attention rate curves of inquiry of targeted customer is 100,98,
50,99,80 these five, actual maximum attention rate is 100, but within can differing 5 with 100
98,99 and maximum 100 attention rates being together defined as maximum of reality.
Step 306, algorithm corresponding for maximum attention rate is defined as target algorithm.
After the attention rate determining maximum, algorithm corresponding for maximum attention rate can be defined as target and calculate
Method, when maximum attention rate has two or more, the algorithm that plural maximum attention degree is corresponding is the most corresponding
Have two or more, at this moment can be by two or more algorithm corresponding for plural maximum attention degree
One is defined as target algorithm.
Additionally, as shown in figure 3-2, when the attention rate curve of the algorithm obtained is two, the two attention rate
Curve can intersect at a some A, and the abscissa of some A can represent the user's liveness characterizing user characteristics,
Vertical coordinate can represent the clicking rate characterizing attention rate, and the user's liveness targeted customer is more than the use of some A
During the liveness of family, can be little at user's liveness of targeted customer using collaborative filtering as target algorithm
When user's liveness of an A, can be using social proposed algorithm as purpose algorithm, targeted customer's
When user's liveness is equal to user's liveness of some A, can be in collaborative filtering and social proposed algorithm
In choose any one kind of them algorithm as target algorithm.
Step 307, determine the recommendation information of targeted customer according to target algorithm.
After determining target algorithm, the recommendation information of targeted customer can be determined according to target algorithm.
Exemplary, when the target algorithm determined is collaborative filtering, determine according to collaborative filtering
The process of the recommendation information of targeted customer can be:
Customer data base (customer data base may be located in home server) is searched with targeted customer
For similar k (k is preset value) individual user, and obtain the similarity of this k user and targeted customer,
Such as during k=3, the content that 3 users the highest with targeted customer's similarity like is respectively as follows:
User A likes war film, comedy and action movie；
User B likes war film, romance movie and ethical film；
User C likes war film, action movie and romance movie.
And the similarity of targeted customer and user A is 0.25, it is 0.35 with the similarity of user B, with user
The similarity of C is 0.45, then can obtain the recommendation degree of various types of film:
War film=0.25+0.35+0.45=1.05；
Comedy=0.25；
Action movie=0.25+0.45=0.7；
Ethical film=0.35；
Romance movie=0.35+0.45=0.8.
The recommendation degree ranking of the most various types of films is: 1, war film, 2, romance movie, 3, action movie,
4, ethical film, 5, comedy.Can be using the film of type higher for recommendation degree as the recommendation of targeted customer
Information, or using ranking before forward several type movie as the recommendation information of targeted customer, it is assumed that will row
Forward front 3 type movie of name are as the recommendation information of targeted customer, then the recommendation information bag of targeted customer
Include war film, romance movie and action movie.
And when the target algorithm determined is social proposed algorithm, determine target according to social proposed algorithm
The process of the recommendation information of user can be:
In social proposed algorithm is the historical record by the user associated in social platform with targeted customer
Holding the algorithm determining recommendation information, wherein, social platform can include QQ good friend, QQ group, wechat friend
At least one in circle and microblogging.As a example by recommending film to targeted customer, social platform is QQ group, mesh
Mark user may add multiple QQ group, thus first can be according in the viewing record of targeted customer and group
Liveness in multiple QQ groups of the registration of overall viewing record or targeted customer and other data are come really
One or several candidate population fixed, generates according to the recommendation information in candidate population afterwards and includes multiple recommendation information
Recommendation list, finally according to recommendation list obtain targeted customer recommendation information.Such as recommendation list is permissible
As shown in table 1:
Table 1
| Group's title | Marking | Recommendation information |
| Classmate | 0.2 | War film, comedy and action movie |
| Work | 0.3 | War film, romance movie and ethical film |
| XX forum | 0.5 | War film, action movie and romance movie |
In Table 1, group's name column represents the title of several candidate population, and marking row represent beating of several candidate population
Point, it is recommended that information row represent the recommendation information of several candidate population.Each type of electricity can be obtained according to table 1
The marking of shadow, such as war film=0.2+0.3+0.5=1, comedy=0.2, action movie=0.2+0.5=0.7, love
Sheet=0.3+0.5=0.8, ethical film=0.3, recommendation information can be determined according to these marking afterwards, such as will
The war film of highest scoring is as the recommendation information of targeted customer, or by several types electricity before forward for ranking
Shadow is as the recommendation information of targeted customer, it is assumed that using front 3 type movie forward for ranking as targeted customer
Recommendation information, then the recommendation information of targeted customer includes war film, romance movie and action movie.In Table 1
Film types can also represent a film in this film types, such as war film represents " tunnel warfare ",
Comedy represents " having a narrow escape from death " etc., and the embodiment of the present invention does not restricts.
It should be noted that the embodiment of the present invention is when using social proposed algorithm, it is also possible to multiple societies
Handing over platform to carry out comprehensive consideration to obtain recommendation information, the embodiment of the present invention does not restricts.
Step 308, to targeted customer terminal send targeted customer recommendation information.
After the recommendation information determining targeted customer, the recommendation information of this targeted customer can be sent to
The terminal of user.
It should be added that, the information recommendation method that the embodiment of the present invention provides, by inquiry at least two
The attention rate curve of individual algorithm determines the algorithm corresponding to attention rate of maximum, and will determine according to this algorithm
Recommendation information is sent to the terminal of user, has reached to be determined by the algorithm that maximum attention rate is corresponding
The effect of recommendation information.
In sum, the information recommendation method that the embodiment of the present invention provides, determined by the size of attention rate
One algorithm is as target algorithm, and the recommendation information generated according to this target algorithm is recommended user；Solve
The algorithm that in correlation technique of having determined, recommendation information is used is typically to pre-set, the motility of recommendation information
Relatively low, that specific aim is poor problem；Reach the size according to attention rate and determined target algorithm, it is recommended that letter
The stronger effect of specific aim of breath.
Fig. 4-1 is the schematic flow sheet of the another kind of information recommendation method that the embodiment of the present invention provides, the present embodiment
Illustrate with the server 110 that this information recommendation is applied in implementation environment shown in Fig. 1.This information pushes away
The method of recommending may include that
Step 401, the user characteristics of acquisition targeted customer, this user characteristics is for characterizing the characteristic of user.
When needs carry out information recommendation to targeted customer, can first obtain the user characteristics of targeted customer
User characteristics include user's liveness, user's liveness be user click in preset time period preset in
The number of times held, preset content can be film, books, Shopping directery etc..Additionally, user characteristics also may be used
To include at least one in the age of user, occupation, sex and region.
The data acquisition system of at least two algorithm in step 402, acquisition historical time section, data acquisition system record has symbol
The attention rate of the user of the conjunction user characteristics recommendation information to being obtained by each algorithm.
The data acquisition system of at least two algorithm, the arbitrary algorithm in the embodiment of the present invention in acquisition historical time section
For having the algorithm of time stability, i.e. meet the user of the user characteristics recommendation to being obtained by this algorithm
The change over time of the curve of the attention rate of breath is less, and curve is more steady, and this at least two algorithm is permissible
Including: collaborative filtering and social proposed algorithm, similar to targeted customer according to collaborative filtering
The content of the historical record spending higher user come for targeted customer recommend similar with the content of historical record in
The algorithm held, social proposed algorithm is the history note of the user by associating with targeted customer in social platform
The content of record determines the algorithm of recommendation information.Wherein, attention rate is paid close attention to by user for characterizing preset content
Degree, and this attention rate may include that in conversion ratio, clicking rate and thousand advertising incomes ECPM any
One, this conversion ratio is the visit capacity ratio with total visit capacity of preset content, and this clicking rate is preset content
The ratio of clicked number of times and shown number of times, this ECPM is that preset content shows the advertisement obtained each thousand times
Income.
The user higher with targeted customer's similarity can include that the identity information with targeted customer (such as shows to use
The log-on message etc. of family interest) user that similarity is higher.Additionally, it is all right with the similarity of targeted customer
Including the similarity with targeted customer's behavior over.
Additionally, the acquisition process of data acquisition system can be to test the customer group including a large number of users
Arriving, customer group can also be made up of a part of user in historical record.It should be noted that carry out
When test obtains data acquisition system, it is not required that carry out the recommendation of information completely to same multiple users, only need
The user characteristics wanting user is identical.
Step 403, according to data acquisition system obtain at least two algorithm attention rate curve.
After obtaining data acquisition system, the attention rate curve of at least two algorithm can be obtained according to this data acquisition system,
The attention rate curve of arbitrary algorithm may be used for record and uses between user characteristics and the attention rate that algorithm obtains
Corresponding relation, this corresponding relation is typically relatively stable but it also may interval acquiring at predetermined time intervals
The attention rate curve of at least two algorithm, the embodiment of the present invention does not restricts.When algorithm has two kinds,
The attention rate curve that this step obtains can be as shown in figure 3-2.
It should be noted that step 402 can also perform to step 403 before step 401, the present invention
Embodiment does not restricts.
Step 404, user characteristics according to targeted customer are inquired about the attention rate curve of at least two algorithm respectively and are obtained
To at least two attention rate.
After obtaining the attention rate curve of at least two algorithm, can be according to the user of the targeted customer obtained
Feature obtains this user characteristics at least two corresponding in the attention rate curve of at least two algorithm and pays close attention to
Degree.As a example by the attention rate curve of two algorithms represented by Fig. 3-2, if the user characterizing user characteristics enlivens
Degree is 20, then the clicking rate inquiring about the sign attention rate that the attention rate curve q1 of collaborative filtering obtains is
0.16, the clicking rate of the sign attention rate that the attention rate curve q2 of inquiry social proposed algorithm obtains is 0.19.
Step 405, weights are set for each algorithm at least two algorithm.
Wherein, the weights of arbitrary algorithm and the user characteristics according to targeted customer inquire about the attention rate of arbitrary algorithm
The size positive correlation of the concern angle value that curve obtains, the attention rate i.e. obtained is the biggest, then this attention rate is corresponding
The weights of algorithm also can be the biggest.And weights can represent that the recommendation information obtained according to the algorithm with weights exists
Proportion shared in the recommendation information of targeted customer.
As shown in the Fig. 4-2, step 405 can include following two steps:
Step 4051, at least two attention rate is normalized, obtains corresponding the returning of each attention rate
One change value, each normalized value belongs to [0,1], and normalized value sum corresponding to this at least two attention rate is
1。
Wherein the process of normalized can be: setting attention rate and have n, n attention rate is A1、
A2···An, then in n attention rate, i-th (i belongs to [1, n]) individual attention rate is corresponding normalized value Gi=[Ai/
(A1+A2···An-1+An)], wherein 0≤Gi≤ 1, and G1+G2+······+Gn-1+Gn=1.
Exemplary, it is two with attention rate, and as a example by attention rate is clicking rate, two clicking rates are 0.16
With 0.19, the normalized value of 0.16 correspondence is 0.16/ (0.16+0.19)=0.457, and the normalized value of 0.19 correspondence is
0.19/ (0.16+0.19)=0.543.
Step 4052, using normalized value corresponding for each attention rate as the power of algorithm corresponding to each attention rate
Value.
After obtaining the normalized value that each attention rate is corresponding, can be by normalized value corresponding for each attention rate
Weights as algorithm corresponding to each attention rate.
Additionally, when the attention rate curve of the algorithm obtained is two, for each algorithm in the two algorithm
The mode arranging weights can also be: as shown in figure 3-2, when the attention rate curve of the algorithm obtained is two,
The two attention rate curve can intersect at an A, and the abscissa of some A can represent the use characterizing user characteristics
Family liveness, vertical coordinate can represent the clicking rate characterizing attention rate, can live according to the user of targeted customer
Jerk arranges power with the difference of the user's liveness putting A for collaborative filtering and social proposed algorithm
Value, when user's liveness of such as targeted customer is more than user's liveness of some A, the user of targeted customer is special
The difference of the user characteristics levied and put A is the biggest, then the weights of collaborative filtering are the biggest, and social recommends to calculate
The weights of method are the least, and the situation of the user's liveness of targeted customer user's liveness less than an A can push away with this
Go out.
Step 406, the hybrid algorithm at least two algorithm combination with respective weights obtained are calculated as target
Method.
After the weights obtaining algorithm corresponding to each attention rate, the algorithm with weights can be combined
Obtain hybrid algorithm, and using hybrid algorithm as target algorithm.
By when being combined with the algorithm of weights, can be using weights as the marking of each algorithm, such as
Shown in table 2:
Table 2
| Algorithm | Marking (weights) |
| First algorithm | 0.1 |
| Second algorithm | 0.5 |
| Third algorithm | 0.4 |
The marking of the first algorithm is 0.1 in table 2, and the marking of the second algorithm is 0.5, the marking of third algorithm
It is 0.4, can be combined obtaining hybrid algorithm by several algorithms of higher for marking (such as more than 0.3),
In hybrid algorithm, the weights of each algorithm can represent that the recommendation information determined according to this algorithm is used in target
Proportion shared in the recommendation information at family.
Exemplary, the weights of collaborative filtering are 0.4, and the weights of social proposed algorithm are 0.6, then
Hybrid algorithm=(0.4* collaborative filtering+0.6* social proposed algorithm), wherein 0.4* collaborative filtering
Can represent that the recommendation information determined by collaborative filtering accounts for the 40% of the recommendation information of targeted customer, such as
The recommendation degree being obtained various types of film by collaborative filtering is:
War film=1.05；Comedy=0.25；Action movie=0.7；Ethical film=0.35；Romance movie=0.8.
And by the recommendation degree of the social proposed algorithm various types of films of acquisition be:
War film=1；Comedy=0.2；Action movie=0.7；Romance movie=0.8；Ethical film=0.3.
Then can obtain obtaining the recommendation degree of various types of film by target algorithm is:
War film=0.4*1.05+0.6*1=1.02；
Comedy=0.4*0.25+0.6*0.2=0.22；
Action movie=0.4*0.7+0.6*0.7=0.7；
Romance movie=0.4*0.8+0.6*0.8=0.8；
Ethical film=0.4*0.35+0.6*0.3=0.32.
Recommendation information can be determined, such as using the war film of highest scoring as mesh afterwards according to these marking
Mark user recommendation information, or using ranking before forward several type movie as the recommendation of targeted customer
Breath, it is assumed that using front 3 type movie forward for ranking as the recommendation information of targeted customer, then targeted customer
Recommendation information include war film, comedy and action movie.
Step 407, determine the recommendation information of targeted customer according to target algorithm.
After determining target algorithm, the recommendation information of targeted customer can be determined according to target algorithm.
Step 408, to targeted customer terminal send targeted customer recommendation information.
After the recommendation information determining targeted customer, can send targeted customer's to the terminal of targeted customer
Recommendation information.
It should be added that, the information recommendation method that the embodiment of the present invention provides, by inquiry at least two
The attention rate curve of individual algorithm determines the weights of each algorithm, and by least two algorithm groups with weights
Conjunction obtains hybrid algorithm, determines recommendation information finally by hybrid algorithm and sends to the terminal of targeted customer,
Not only ensure that the attention rate of recommendation information, it is ensured that the multiformity of recommendation information.
In sum, the information recommendation method that the embodiment of the present invention provides, determined by the size of attention rate
One algorithm is as target algorithm, and the recommendation information generated according to this target algorithm is recommended user；Solve
The algorithm that in correlation technique of having determined, recommendation information is used is typically to pre-set, the motility of recommendation information
Relatively low, that specific aim is poor problem；Reach the size according to attention rate and determined target algorithm, it is recommended that letter
The stronger effect of specific aim of breath.
Following for apparatus of the present invention embodiment, may be used for performing the inventive method embodiment.For the present invention
The details not disclosed in device embodiment, refer to the inventive method embodiment.
Fig. 5 is the block diagram of a kind of information recommending apparatus shown in the embodiment of the present invention, and this information recommending apparatus can
With by software, hardware or both be implemented in combination with becoming server 110 in implementation environment shown in Fig. 1
Some or all of.This information recommending apparatus may include that
Acquiring unit 510, for obtaining the user characteristics of targeted customer, this user characteristics is for characterizing user's
Characteristic.
Query unit 520, for inquiring about the concern of at least two algorithm respectively according to the user characteristics of targeted customer
Line of writing music obtains at least two attention rate, and wherein, the attention rate curve of arbitrary algorithm is used for recording employing algorithm
Corresponding relation between the user characteristics and the attention rate that obtain, this attention rate is used for characterizing preset content by user
Degree of concern.
Algorithm determines unit 530, for size based at least two attention rate, at least two algorithm really
At least one algorithm fixed is as target algorithm.
Information determination unit 540, for determining the recommendation information of targeted customer according to target algorithm.
Transmitting element 550, for sending the recommendation information of targeted customer to the terminal of targeted customer.
In sum, the information recommending apparatus that the embodiment of the present invention provides, determined by the size of attention rate
One algorithm is as target algorithm, and the recommendation information generated according to this target algorithm is recommended user；Solve
The algorithm that in correlation technique of having determined, recommendation information is used is typically to pre-set, the motility of recommendation information
Relatively low, that specific aim is poor problem；Reach the size according to attention rate and determined target algorithm, it is recommended that letter
The stronger effect of specific aim of breath.
Fig. 6-1 is the block diagram of the another kind of information recommending apparatus shown in the embodiment of the present invention, this information recommending apparatus
Can pass through software, hardware or both be implemented in combination with becoming server 110 in implementation environment shown in Fig. 1
Some or all of.This information recommending apparatus may include that
Acquiring unit 510, for obtaining the user characteristics of targeted customer, this user characteristics is for characterizing user's
Characteristic.
Query unit 520, for inquiring about the concern of at least two algorithm respectively according to the user characteristics of targeted customer
Line of writing music obtains at least two attention rate, and wherein, the attention rate curve of arbitrary algorithm is used for recording employing algorithm
Corresponding relation between the user characteristics and the attention rate that obtain, this attention rate is used for characterizing preset content by user
Degree of concern.
Algorithm determines unit 530, for size based at least two attention rate, at least two algorithm really
At least one algorithm fixed is as target algorithm.
Information determination unit 540, for determining the recommendation information of targeted customer according to target algorithm.
Transmitting element 550, for sending the recommendation information of targeted customer to the terminal of targeted customer.
Optionally, this device also includes:
Set acquiring unit 560, the data acquisition system of at least two algorithm, data in obtaining historical time section
Set record has the attention rate of the user the meeting user characteristics recommendation information to being obtained by each algorithm.
Curve acquisition unit 570, for obtaining the attention rate curve of at least two algorithm according to data acquisition system.
Optionally, algorithm determines unit 530, for determining attention rate maximum at least two attention rate；Will
The algorithm that maximum attention rate is corresponding is defined as target algorithm.
Optionally, as in fig. 6-2, algorithm determines unit 530, including:
Weights module 531, for arranging weights for each algorithm at least two algorithm, wherein, arbitrary calculation
The weights of method inquire about, with the user characteristics according to targeted customer, the attention rate that the attention rate curve of arbitrary algorithm obtains
The size positive correlation of value, weights represent that the recommendation information according to the algorithm acquisition with weights is targeted customer's
Proportion shared in recommendation information.
Composite module 532, makees for the hybrid algorithm at least two algorithm combination with respective weights obtained
For target algorithm.
Optionally, weights module 531, at least two attention rate is normalized, obtain each
The normalized value that attention rate is corresponding, each normalized value belongs to [0,1], and this at least two attention rate is corresponding
Normalized value sum is 1；Using normalized value corresponding for each attention rate as algorithm corresponding to each attention rate
Weights.
Optionally, user characteristics includes that user's liveness, user's liveness are user's point in preset time period
Hit the number of times of preset content.
Optionally, at least two algorithm includes: collaborative filtering and social proposed algorithm, collaborative filtering
The content of the historical record of user higher with targeted customer's similarity according to algorithm is recommended for targeted customer
The algorithm of the content similar with the content of historical record, social proposed algorithm be by social platform with mesh
The content of the historical record of the user of mark user-association determines the algorithm of recommendation information.
Optionally, attention rate includes: any one in conversion ratio, clicking rate and thousand advertising incomes ECPM
Kind, conversion ratio is the visit capacity ratio with total visit capacity of preset content, and clicking rate is that preset content is clicked
The ratio of number of times and shown number of times, ECPM is that preset content shows the advertising income obtained each thousand times.
It should be added that, the information recommending apparatus that the embodiment of the present invention provides, by inquiry at least two
The attention rate curve of individual algorithm determines the weights of each algorithm, and by least two algorithm groups with weights
Conjunction obtains hybrid algorithm, determines recommendation information finally by hybrid algorithm and sends to the terminal of targeted customer,
Not only ensure that the attention rate of recommendation information, it is ensured that the multiformity of recommendation information.
In sum, the information recommending apparatus that the embodiment of the present invention provides, determined by the size of attention rate
One algorithm is as target algorithm, and the recommendation information generated according to this target algorithm is recommended user；Solve
The algorithm that in correlation technique of having determined, recommendation information is used is typically to pre-set, the motility of recommendation information
Relatively low, that specific aim is poor problem；Reach the size according to attention rate and determined target algorithm, it is recommended that letter
The stronger effect of specific aim of breath.
About the device in above-described embodiment, wherein modules performs the concrete mode of operation relevant
The embodiment of the method is described in detail, explanation will be not set forth in detail herein.
Fig. 7 shows the structural representation of the server in the information recommendation method that the embodiment of the present invention provides.
Described server 700 includes CPU (CPU) 701, includes random access memory (RAM)
702 and the system storage 704 of read only memory (ROM) 703, and connection system memorizer 704 He
The system bus 705 of CPU 701.Described server 700 also includes helping each in computer
Transmit the basic input/output (I/O system) 706 of information between device, and be used for storing operating system
713, the mass-memory unit 707 of application program 714 and other program modules 715.
Described basic input/output 706 includes the display 708 for showing information and for user
The input equipment 709 of such as mouse, keyboard etc of input information.Wherein said display 708 and input set
Standby 709 are all connected to CPU by being connected to the IOC 710 of system bus 705
701.Described basic input/output 706 can also include IOC 710 for receive and
Process the input from other equipment multiple such as keyboard, mouse or electronic touch pens.Similarly, input defeated
Go out controller 710 and also provide output to display screen, printer or other kinds of outut device.
Described mass-memory unit 707 is by being connected to the bulk memory controller of system bus 705 (not
Illustrate) it is connected to CPU 701.Described mass-memory unit 707 and the computer being associated thereof
Computer-readable recording medium provides non-volatile memories for server 700.It is to say, described mass-memory unit 707
The computer-readable medium (not shown) of such as hard disk or CD-ROM drive etc can be included.
Without loss of generality, described computer-readable medium can include computer-readable storage medium and communication media.
Computer-readable storage medium include for store such as computer-readable instruction, data structure, program module or
Volatibility that any method of the information such as other data or technology realize and non-volatile, removable and can not move
Moving medium.Computer-readable storage medium include RAM, ROM, EPROM, EEPROM, flash memory or other
Its technology of solid-state storage, CD-ROM, DVD or other optical storage, cartridge, tape, disk storage
Or other magnetic storage apparatus.Certainly, skilled person will appreciate that described computer-readable storage medium does not limits to
In above-mentioned several.Above-mentioned system storage 704 and mass-memory unit 707 may be collectively referred to as memorizer.
According to various embodiments of the present invention, described server 700 can also be by networks such as such as the Internets
The remote computer being connected on network runs.Namely server 700 can be total by being connected to described system
NIU 711 on line 705 is connected to network 712, in other words, it is possible to use network interface list
Unit 711 is connected to other kinds of network or remote computer system (not shown).
Described memorizer also includes one or more than one program, one or more than one program
Being stored in memorizer, one or more than one program comprise for carrying out embodiment of the present invention offer
The instruction of information recommendation method.
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all the present invention's
Within spirit and principle, any modification, equivalent substitution and improvement etc. made, should be included in the present invention's
Within protection domain.
Claims (16)
1. an information recommendation method, it is characterised in that described method includes:
Obtaining the user characteristics of targeted customer, described user characteristics is for characterizing the characteristic of user；
User characteristics according to described targeted customer inquire about respectively the attention rate curve of at least two algorithm obtain to
Few two attention rates, wherein, the attention rate curve of arbitrary algorithm is for recording the use using described algorithm to obtain
Corresponding relation between family feature and attention rate, described attention rate is paid close attention to by user for characterizing preset content
Degree；
Size based on described at least two attention rate, determines that at least one is calculated in described at least two algorithm
Method is as target algorithm；
The recommendation information of described targeted customer is determined according to described target algorithm；
The recommendation information of described targeted customer is sent to the terminal of described targeted customer.
Method the most according to claim 1, it is characterised in that described pay close attention to based on described at least two
The size of degree, determine in described at least two algorithm at least one algorithm as target algorithm, including:
Determine attention rate maximum in described at least two attention rate；
Algorithm corresponding for the attention rate of described maximum is defined as described target algorithm.
Method the most according to claim 1, it is characterised in that described pay close attention to based on described at least two
The size of angle value, determine in described at least two algorithm at least one algorithm as target algorithm, including:
For each algorithm in described at least two algorithm, weights are set, wherein, the weights of arbitrary algorithm and root
The concern angle value that obtains of attention rate curve of described arbitrary algorithm is inquired about according to the user characteristics of described targeted customer
Size positive correlation, described weights represent that the recommendation information according to the algorithm acquisition with described weights is at described mesh
Proportion shared in the recommendation information of mark user；
The hybrid algorithm that described at least two algorithm combination with respective weights obtains is calculated as described target
Method.
Method the most according to claim 3, it is characterised in that described in described at least two algorithm
Each algorithm weights are set, including:
Described at least two attention rate is normalized, obtains the normalized value that each attention rate is corresponding,
Each described normalized value belongs to [0,1], and normalized value sum corresponding to described at least two attention rate is 1；
Using normalized value corresponding for described each attention rate as the power of algorithm corresponding to described each attention rate
Value.
Method the most according to claim 1, it is characterised in that the described use according to described targeted customer
Family feature is inquired about before the attention rate curve of at least two algorithm obtains at least two attention rate respectively, described side
Method also includes:
The data acquisition system of described at least two algorithm in acquisition historical time section, described data acquisition system record has symbol
Close the attention rate of the user of the described user characteristics recommendation information to being obtained by each described algorithm；
The attention rate curve of described at least two algorithm is obtained according to described data acquisition system.
6. according to the arbitrary described method of claim 1 to 5, it is characterised in that
Described user characteristics includes that user's liveness, described user's liveness are user's point in preset time period
Hit the number of times of preset content.
7. according to the arbitrary described method of claim 1 to 5, it is characterised in that
Described at least two algorithm includes: collaborative filtering and social proposed algorithm, described collaborative filtering
The content of the historical record of user higher with targeted customer's similarity according to algorithm is come for described targeted customer
Recommending the algorithm of the content similar with the content of described historical record, described social proposed algorithm is for passing through society
The content handing over the historical record of the user associated in platform with targeted customer determines the algorithm of recommendation information.
8. according to the arbitrary described method of claim 1 to 5, it is characterised in that
Described attention rate includes: any one in conversion ratio, clicking rate and thousand advertising incomes ECPM,
Described conversion ratio is the visit capacity ratio with total visit capacity of preset content, and described clicking rate is preset content quilt
The ratio of number of clicks and shown number of times, described ECPM is that preset content shows the advertisement obtained each thousand times
Income.
9. an information recommending apparatus, it is characterised in that described device includes:
Acquiring unit, for obtaining the user characteristics of targeted customer, described user characteristics is for characterizing user's
Characteristic；
Query unit, for inquiring about the pass of at least two algorithm respectively according to the user characteristics of described targeted customer
Noting line of writing music and obtain at least two attention rate, wherein, the attention rate curve of arbitrary algorithm is used for recording employing institute
State the corresponding relation between user characteristics and the attention rate that algorithm obtains, in described attention rate is used for characterizing and presets
Hold the degree of concern of user；
Algorithm determines unit, for size based on described at least two attention rate, calculates in described at least two
Method determining, at least one algorithm is as target algorithm；
Information determination unit, for determining the recommendation information of described targeted customer according to described target algorithm；
Transmitting element, for sending the recommendation information of described targeted customer to the terminal of described targeted customer.
Device the most according to claim 9, it is characterised in that described algorithm determines unit, is used for:
Determine attention rate maximum in described at least two attention rate；
Algorithm corresponding for the attention rate of described maximum is defined as described target algorithm.
11. devices according to claim 9, it is characterised in that described algorithm determines unit, including:
Weights module, for arranging weights for each algorithm in described at least two algorithm, wherein, arbitrary
The attention rate curve that the weights of algorithm inquire about described arbitrary algorithm with the user characteristics according to described targeted customer obtains
The size positive correlation of the concern angle value arrived, described weights represent according to pushing away that the algorithm with described weights obtains
The proportion that information of recommending is shared in the recommendation information of described targeted customer；
Composite module, for the hybrid algorithm described at least two algorithm combination with respective weights obtained
As described target algorithm.
12. devices according to claim 11, it is characterised in that described weights module, are used for:
Described at least two attention rate is normalized, obtains the normalized value that each attention rate is corresponding,
Each described normalized value belongs to [0,1], and normalized value sum corresponding to described at least two attention rate is 1；
Using normalized value corresponding for described each attention rate as the power of algorithm corresponding to described each attention rate
Value.
13. devices according to claim 9, it is characterised in that described device also includes:
Set acquiring unit, the data acquisition system of described at least two algorithm, institute in obtaining historical time section
State data acquisition system record and have the user the meeting described user characteristics recommendation to being obtained by each described algorithm
The attention rate of breath；
Curve acquisition unit, writes music for obtaining the concern of described at least two algorithm according to described data acquisition system
Line.
14. according to the arbitrary described device of claim 9 to 13, it is characterised in that
Described user characteristics includes that user's liveness, described user's liveness are user's point in preset time period
Hit the number of times of preset content.
15. according to the arbitrary described device of claim 9 to 13, it is characterised in that
Described at least two algorithm includes: collaborative filtering and social proposed algorithm, described collaborative filtering
The content of the historical record of user higher with targeted customer's similarity according to algorithm is come for described targeted customer
Recommending the algorithm of the content similar with the content of described historical record, described social proposed algorithm is for passing through society
The content handing over the historical record of the user associated in platform with targeted customer determines the algorithm of recommendation information.
16. according to the arbitrary described device of claim 9 to 13, it is characterised in that
Described attention rate includes: any one in conversion ratio, clicking rate and thousand advertising incomes ECPM,
Described conversion ratio is the visit capacity ratio with total visit capacity of preset content, and described clicking rate is preset content quilt
The ratio of number of clicks and shown number of times, described ECPM is that preset content shows the advertisement obtained each thousand times
Income.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510312846.3A CN106294497B (en) | 2015-06-09 | 2015-06-09 | Information recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510312846.3A CN106294497B (en) | 2015-06-09 | 2015-06-09 | Information recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106294497A true CN106294497A (en) | 2017-01-04 |
| CN106294497B CN106294497B (en) | 2020-05-12 |
Family
ID=57659972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510312846.3A Active CN106294497B (en) | 2015-06-09 | 2015-06-09 | Information recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106294497B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107623863A (en) * | 2017-09-21 | 2018-01-23 | 广州华多网络科技有限公司 | Test of heuristics method, apparatus and server |
| CN107665135A (en) * | 2017-10-20 | 2018-02-06 | 维沃移动通信有限公司 | The method for pushing and device of a kind of ROMPaq |
| CN108011941A (en) * | 2017-11-29 | 2018-05-08 | 广东欧珀移动通信有限公司 | Content delivery method, device, server and storage medium |
| CN109753601A (en) * | 2018-11-28 | 2019-05-14 | 北京奇艺世纪科技有限公司 | Recommendation information clicking rate determines method, apparatus and electronic equipment |
| CN109934605A (en) * | 2017-12-15 | 2019-06-25 | 北京京东尚科信息技术有限公司 | Method, apparatus, medium and the electronic equipment that conversion ratio based on big data is promoted |
| CN110245301A (en) * | 2018-11-29 | 2019-09-17 | 腾讯科技（深圳）有限公司 | A recommended method, device and storage medium |
| CN110413881A (en) * | 2019-07-11 | 2019-11-05 | 腾讯科技（深圳）有限公司 | A kind of method, apparatus, the network equipment and storage medium identifying label accuracy |
| CN110415023A (en) * | 2019-07-09 | 2019-11-05 | 日立楼宇技术（广州）有限公司 | Elevator advertisement recommendation method, device, equipment and storage medium |
| CN111177062A (en) * | 2019-12-02 | 2020-05-19 | 上海连尚网络科技有限公司 | A method and apparatus for providing reading presentation information |
| CN112417263A (en) * | 2019-08-23 | 2021-02-26 | 北京达佳互联信息技术有限公司 | Data recommendation method and device and storage medium |
| CN112667892A (en) * | 2020-12-25 | 2021-04-16 | 北京达佳互联信息技术有限公司 | Information recommendation method, device, server and storage medium |
| WO2021196470A1 (en) * | 2020-03-31 | 2021-10-07 | 百度在线网络技术（北京）有限公司 | Information pushing method and apparatus, device, and storage medium |
| CN113722601A (en) * | 2021-09-07 | 2021-11-30 | 南方电网数字电网研究院有限公司 | Power measurement information recommendation method and device, computer equipment and storage medium |
| CN114969519A (en) * | 2022-05-13 | 2022-08-30 | 海知见信息技术(北京)有限公司 | Attention-degree-based customer recommendation method and device |
| US11669530B2 (en) | 2020-03-31 | 2023-06-06 | Baidu Online Network Technology (Beijing) Co., Ltd. | Information push method and apparatus, device, and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4951404B2 (en) * | 2007-05-08 | 2012-06-13 | 楽天株式会社 | Product recommendation system |
| CN102611785A (en) * | 2011-01-20 | 2012-07-25 | 北京邮电大学 | Personalized active news recommending service system and method for mobile phone user |
| CN103106285A (en) * | 2013-03-04 | 2013-05-15 | 中国信息安全测评中心 | Recommendation algorithm based on information security professional social network platform |
| CN103678672A (en) * | 2013-12-25 | 2014-03-26 | 北京中兴通软件科技股份有限公司 | Method for recommending information |
-
2015
- 2015-06-09 CN CN201510312846.3A patent/CN106294497B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4951404B2 (en) * | 2007-05-08 | 2012-06-13 | 楽天株式会社 | Product recommendation system |
| CN102611785A (en) * | 2011-01-20 | 2012-07-25 | 北京邮电大学 | Personalized active news recommending service system and method for mobile phone user |
| CN103106285A (en) * | 2013-03-04 | 2013-05-15 | 中国信息安全测评中心 | Recommendation algorithm based on information security professional social network platform |
| CN103678672A (en) * | 2013-12-25 | 2014-03-26 | 北京中兴通软件科技股份有限公司 | Method for recommending information |
Non-Patent Citations (2)
| Title |
|---|
| 贺栋博: "视频推荐系统数据分析决策平台", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 陈诺言: "基于个性化推荐引擎组合的推荐系统的设计与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107623863B (en) * | 2017-09-21 | 2020-11-06 | 广州华多网络科技有限公司 | Algorithm testing method and device and server |
| CN107623863A (en) * | 2017-09-21 | 2018-01-23 | 广州华多网络科技有限公司 | Test of heuristics method, apparatus and server |
| CN107665135A (en) * | 2017-10-20 | 2018-02-06 | 维沃移动通信有限公司 | The method for pushing and device of a kind of ROMPaq |
| CN108011941A (en) * | 2017-11-29 | 2018-05-08 | 广东欧珀移动通信有限公司 | Content delivery method, device, server and storage medium |
| CN109934605A (en) * | 2017-12-15 | 2019-06-25 | 北京京东尚科信息技术有限公司 | Method, apparatus, medium and the electronic equipment that conversion ratio based on big data is promoted |
| CN109753601B (en) * | 2018-11-28 | 2021-10-22 | 北京奇艺世纪科技有限公司 | Method and device for determining click rate of recommended information and electronic equipment |
| CN109753601A (en) * | 2018-11-28 | 2019-05-14 | 北京奇艺世纪科技有限公司 | Recommendation information clicking rate determines method, apparatus and electronic equipment |
| CN110245301A (en) * | 2018-11-29 | 2019-09-17 | 腾讯科技（深圳）有限公司 | A recommended method, device and storage medium |
| CN110415023A (en) * | 2019-07-09 | 2019-11-05 | 日立楼宇技术（广州）有限公司 | Elevator advertisement recommendation method, device, equipment and storage medium |
| CN110413881A (en) * | 2019-07-11 | 2019-11-05 | 腾讯科技（深圳）有限公司 | A kind of method, apparatus, the network equipment and storage medium identifying label accuracy |
| CN110413881B (en) * | 2019-07-11 | 2023-10-20 | 腾讯科技（深圳）有限公司 | Method, device, network equipment and storage medium for identifying label accuracy |
| CN112417263A (en) * | 2019-08-23 | 2021-02-26 | 北京达佳互联信息技术有限公司 | Data recommendation method and device and storage medium |
| CN112417263B (en) * | 2019-08-23 | 2024-01-09 | 北京达佳互联信息技术有限公司 | Data recommendation method, device and storage medium |
| CN111177062A (en) * | 2019-12-02 | 2020-05-19 | 上海连尚网络科技有限公司 | A method and apparatus for providing reading presentation information |
| CN111177062B (en) * | 2019-12-02 | 2024-04-05 | 上海连尚网络科技有限公司 | Method and device for providing reading presentation information |
| WO2021196470A1 (en) * | 2020-03-31 | 2021-10-07 | 百度在线网络技术（北京）有限公司 | Information pushing method and apparatus, device, and storage medium |
| US11669530B2 (en) | 2020-03-31 | 2023-06-06 | Baidu Online Network Technology (Beijing) Co., Ltd. | Information push method and apparatus, device, and storage medium |
| CN112667892A (en) * | 2020-12-25 | 2021-04-16 | 北京达佳互联信息技术有限公司 | Information recommendation method, device, server and storage medium |
| CN112667892B (en) * | 2020-12-25 | 2024-01-19 | 北京达佳互联信息技术有限公司 | Information recommendation method, device, server and storage medium |
| CN113722601A (en) * | 2021-09-07 | 2021-11-30 | 南方电网数字电网研究院有限公司 | Power measurement information recommendation method and device, computer equipment and storage medium |
| CN113722601B (en) * | 2021-09-07 | 2024-04-16 | 南方电网数字电网研究院股份有限公司 | Power measurement information recommendation method, device, computer equipment and storage medium |
| CN114969519A (en) * | 2022-05-13 | 2022-08-30 | 海知见信息技术(北京)有限公司 | Attention-degree-based customer recommendation method and device |
| CN114969519B (en) * | 2022-05-13 | 2023-05-12 | 海知见信息技术(北京)有限公司 | Attention-based client recommendation method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106294497B (en) | 2020-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106294497A (en) | Information recommendation method and device | |
| US9576016B2 (en) | Targeting stories based on influencer scores | |
| JP6440781B2 (en) | Generation of a sponsor article unit containing related posts and input elements | |
| US9928280B2 (en) | Suggesting connections to users with low activity in a social networking system | |
| US9990652B2 (en) | Targeting social advertising to friends of users who have interacted with an object associated with the advertising | |
| KR102104256B1 (en) | Sponsored advertisement ranking and pricing in a social networking system | |
| CA2857526C (en) | Grouping and ordering advertising units based on user activity | |
| US20020161838A1 (en) | Method and apparatus for targeting information | |
| WO2009023070A1 (en) | Systems and methods for keyword selection in a web-based social network | |
| CN110489663B (en) | Social content control method and device and computer equipment | |
| US11562014B1 (en) | Generating visual media collections for a dynamic social networking account | |
| CN103891245A (en) | Location-aware content detection | |
| CN105956086B (en) | Multimedia resource recommendation method and device | |
| CN112514403A (en) | Distribution of embedded content items by an online system | |
| JP6416108B2 (en) | Generate metrics based on client device ownership | |
| US20190122259A1 (en) | Managing a frequency of presentation of a content item associated with a category within a hierarchical taxonomy to a user of an online system | |
| US9959258B2 (en) | Generating characteristics of users of an online system presented with content in a context relative to other content | |
| CN106156166A (en) | Relation chain inquiry system, document retrieval method, index establishing method and device | |
| HK1179726A (en) | User interfaces for personalized recommendations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |