CN106878772A - A kind of program commending method and device - Google Patents
A kind of program commending method and device Download PDFInfo
- Publication number
- CN106878772A CN106878772A CN201710107905.2A CN201710107905A CN106878772A CN 106878772 A CN106878772 A CN 106878772A CN 201710107905 A CN201710107905 A CN 201710107905A CN 106878772 A CN106878772 A CN 106878772A
- Authority
- CN
- China
- Prior art keywords
- program
- proposed algorithm
- scoring
- recommended
- weight
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/258—Client or end-user data management, e.g. managing client capabilities, user preferences or demographics, processing of multiple end-users preferences to derive collaborative data
- H04N21/25866—Management of end-user data
- H04N21/25891—Management of end-user data being end-user preferences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/262—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists
- H04N21/26258—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists for generating a list of items to be played back in a given order, e.g. playlist, or scheduling item distribution according to such list
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/262—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists
- H04N21/26291—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists for providing content or additional data updates, e.g. updating software modules, stored at the client
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/458—Scheduling content for creating a personalised stream, e.g. by combining a locally stored advertisement with an incoming stream; Updating operations, e.g. for OS modules ; time-related management operations
- H04N21/4586—Content update operation triggered locally, e.g. by comparing the version of software modules in a DVB carousel to the version stored locally
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/47—End-user applications
- H04N21/482—End-user interface for program selection
- H04N21/4826—End-user interface for program selection using recommendation lists, e.g. of programs or channels sorted out according to their score
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Computer Graphics (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of program commending method and device, wherein, the program commending method includes:Obtain the feedback information that user is based on program recommendation system feedback, the program recommendation system is used to calculate program recommendation list, and the program recommendation system is built based on two or more program recommended models, wherein, described two above program recommended models are built by different program proposed algorithms respectively;Weight of described two above program recommended models when the program recommendation list is calculated is updated based on the feedback information;The weight of described two above program recommended models after according to renewal when the program recommendation list is calculated, updates program recommendation list.The program commending method that the present invention is provided can make full use of the feedback information of user, the weight of each program recommended models in dynamic adjustment program recommendation system, final recommendation results is more conformed to the changes in demand of user and experience result.
Description
Technical field
The present invention relates to the communications field, and in particular to a kind of program commending method and device.
Background technology
With the development of science and technology IPTV (Internet Protocol Television, IPTV) is gradually
Come into the life of people.In order to better meet the demand of people, the program recommendation system of existing IPTV can pass through
Obtain user profile and the magnanimity behavior to user is excavated, be that user recommends related commodity, provide the user with individual character
Change service.Conventional IPTV program recommendation systems include being based on statistical recommendation, based on the recommendation of the content degree of correlation, based on association
With filtered recommendation etc..Above-mentioned these program proposed algorithms are based on user's browsing pages, click on viewing and viewing duration etc. and are
Row feedback information of user is recommended.
But, above-mentioned several program proposed algorithms are also individually present shortcoming.Have that new user's is cold based on statistical recommendation
Starting problem;Although and the recommendation for being based on the content degree of correlation can solve the problems, such as cold start-up, the data of lacking individuality;Base
Personalized recommendation can be carried out according to the behavior of user in collaborative filtering recommending, but also face the problem of cold start-up.
Just because of above-mentioned program proposed algorithm respectively has good and bad point, therefore generally pass through program recommended models in actual use
Comprehensively use.Developer is estimated usually using the secondary feedback behavioural information of user to program recommended models, above-mentioned user
Secondary feedback behavior browses program recommendation list, clicks on viewing IPTV programs and user by program recommendation list including user
Evaluation used program recommendation system etc..Fixed after being designed in the early stage due to existing program recommended models, thus
When the secondary feedback behavior of later stage user changes, can cause correctly be reflected in most the assessment of program recommended models
On the program recommendation list for obtaining eventually.That is, existing program recommendation system is stagnant to the secondary feedback behavior reaction of user
Afterwards, it is necessary to developer it is manual program recommended models are modified could user's be secondary so that program recommendation system is followed
Feedback behavior, this not only inefficiency, but also waste human cost.
The content of the invention
The embodiment of the present invention provides a kind of program commending method and device, it is intended to enable recommendation results quick response user
Behavior, better meets the demand of user.
The first aspect of the embodiment of the present invention, there is provided a kind of program commending method, the program commending method includes:
The feedback information that user is based on program recommendation system feedback is obtained, the program recommendation system is pushed away for calculating program
List is recommended, and the program recommendation system is built based on two or more program recommended models, wherein, described two above programs are pushed away
Model is recommended to be built by different program proposed algorithms respectively;
Based on the feedback information described two above program recommended models of renewal when the program recommendation list is calculated
Weight;
The weight of described two above program recommended models after according to renewal when the program recommendation list is calculated, more
New program recommendation list.
The second aspect of the embodiment of the present invention, there is provided a kind of program recommendation apparatus, the program recommendation apparatus include:
Feedback information acquiring unit, the feedback information of program recommendation system feedback, the program are based on for obtaining user
Commending system is used to calculate program recommendation list, and the program recommendation system is built based on two or more program recommended models,
Wherein, described two above program recommended models are built by different program proposed algorithms respectively;
Weight updating block, the feedback information for being got based on the feedback information acquiring unit updates described two
Weight of the above program recommended models when the program recommendation list is calculated;
Program recommendation list updating block, for after the renewal that is obtained according to the weight updating block it is described two with
Weight of the upper program recommended models when the program recommendation list is calculated, updates program recommendation list.
Therefore, in embodiments of the present invention, the feedback information that user is based on program recommendation system feedback is obtained first,
The program recommendation system is used to calculate program recommendation list, and the program recommendation system is based on two or more programs and recommends mould
Type builds, wherein, described two above program recommended models are built by different program proposed algorithms respectively, are then based on described
Feedback information updates weight of described two above program recommended models when the program recommendation list is calculated, finally according to more
The weight of described two above program recommended models after new when the program recommendation list is calculated, updates program and recommends row
Table.The embodiment of the present invention the automatic weight coefficient to program recommended models can be carried out in the running of program recommendation system
Amendment, enables program recommendation list to carry out real-time update according to factors such as the behaviors of user, better meets the demand of user.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can be with
Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is that program commending method provided in an embodiment of the present invention realizes flow chart;
Fig. 2 is the structured flowchart of program recommendation apparatus provided in an embodiment of the present invention.
Specific embodiment
To enable that goal of the invention of the invention, feature, advantage are more obvious and understandable, below in conjunction with the present invention
Accompanying drawing in embodiment, is clearly and completely described to the technical scheme in the embodiment of the present invention, it is clear that described reality
It is only a part of embodiment of the invention to apply example, and not all embodiments.Based on the embodiment in the present invention, the common skill in this area
The every other embodiment that art personnel are obtained under the premise of creative work is not made, belongs to the model of present invention protection
Enclose.
Realization of the invention is described in detail below in conjunction with specific embodiment:
Embodiment one
Fig. 1 shows that the program commending method that the embodiment of the present invention one is provided realizes flow, and details are as follows:
In step S101, the feedback information that user is based on program recommendation system feedback is obtained.
In embodiments of the present invention, in order to better illustrate the scheme of the embodiment of the present invention, first to program recommendation system
It is briefly described.Above-mentioned program recommendation system is used to calculate program recommendation list, and above-mentioned program recommendation system be based on two kinds with
Upper program recommended models build, wherein, above two above program recommended models are built by different program proposed algorithms respectively,
And each program recommended models has different weights when program recommendation list is calculated.Above-mentioned program recommendation system is with above-mentioned
Two or more program recommended models build the consequently recommended index for calculating each program, and according to above-mentioned consequently recommended index generation section
Mesh recommendation list, program is recommended with to user.In step S101, program recommendation apparatus obtain user first to be recommended based on program
The feedback information of system feedback.
Alternatively, above-mentioned proposed algorithm includes being based on statistical proposed algorithm.Then now, in step S101, it is necessary to
The user for obtaining first is based on the feedback information of program recommendation system feedback for user watches program duration and program category, and presses
According to User IP=1,2 ... ..., i and program ID=1,2 ... ..., j is organized into user watched information matrix, used as based on statistics
Proposed algorithm data source.Above-mentioned user watched information matrix is expressed as follows:
Wherein, t represents the audience information element for getting.In the present embodiment, T1User i is reduced to (for the ease of saying
Bright, user i described in the present embodiment represents the user that User IP is i) the viewing duration accounting of program j is marked as scoring
Standard, if for example, user i have viewed complete program j, setting tijIt is 1;If the discontented program j of duration of user i viewing programs j
The 10% of total duration, then set tijIt is 0.1, by that analogy, by above-mentioned tijValue be limited in the interval range of [0,1].It is above-mentioned
Calculated using above-mentioned user watched information matrix based on statistical proposed algorithm, and used f1(T1) represent.
Alternatively, above-mentioned proposed algorithm also includes the proposed algorithm based on the content degree of correlation.Then now, in step S101
In, the incidence relation between two programs need to be only obtained, the relation between above-mentioned program can be from the angle of programme content
It is described, and the incidence relation between program is organized into programme information matrix, above-mentioned programme information matrix is expressed as follows:
Wherein, p represents the element with programme content as data, and above-mentioned programme content can be reduced to the basic letter of program
Breath, for example programm name, PD program director, program perform personnel etc..The above-mentioned proposed algorithm based on the content degree of correlation is using above-mentioned
User tidies up information matrix and is calculated, and uses f2(T2) represent.
Certainly, above-mentioned proposed algorithm can also be including other algorithms, such as proposed algorithm based on collaborative filtering etc., this
Place is not construed as limiting.In embodiments of the present invention, it is possible to use fn(Tn) any one proposed algorithm is represented, for example, use f3
(T3) represent the proposed algorithm based on collaborative filtering.
In step s 102, above two above program recommended models are updated based on above-mentioned feedback information and computationally states section
Weight during mesh recommendation list.
In embodiments of the present invention, according to the feedback information got in step S101, above two above program is updated
Recommended models computationally state weight during program recommendation list.Specifically, above-mentioned program is calculated by program recommended models to push away
Recommend what list was realized in:Because above-mentioned different program proposed algorithm each constructs a program recommended models, in section
When mesh commending system is initialized, above-mentioned several program recommended models are integrated by proposed algorithm mixed model, it is above-mentioned
Proposed algorithm mixed model can be defined as:
S=H (f1(T1), f2(T2) ... ..., fn(Tn))
Wherein, H represents proposed algorithm mixed model, and S represents the result being calculated by proposed algorithm mixed model, i.e.,
Consequently recommended index.As a rule, the above results S can be expressed as consequently recommended exponential matrix:
Wherein, sijIt is consequently recommended indexes of the program j to user i.
Specifically, in embodiments of the present invention, above-mentioned proposed algorithm mixed model H can be reduced to weighted model, its
It is expressed as follows:
S=H (f1(T1), f2(T2) ... ..., fn(Tn))=a1f1+a2f2+……+anfn
Wherein, above-mentioned anIt is the default initial weight of each proposed algorithm, thus, in step s 102, can be based on
The feedback information obtained in step S101 carries out real-time adjustment to the weight of above-mentioned recommendation program model.Above-mentioned feedback information is main
Express the recommendation degree of accuracy and the reliability of program proposed algorithm used by each program recommended models.When certain program recommended models
When the degree of accuracy of the program proposed algorithm for being used and reliability higher, the program recommended models will be improved in step s 102
Weight;Opposite, when the degree of accuracy and the relatively low reliability of the program proposed algorithm that each program recommended models are used, in step
The weight of the program recommended models will be reduced in rapid S102.
In step s 103, computationally state program according to the above two above program recommended models after renewal and recommend row
Weight during table, updates program recommendation list.
In embodiments of the present invention, according to the above two above program recommended models after the renewal obtained in step S102
Weight during program recommendation list is computationally stated, program recommendation list is updated.Alternatively, in order that program recommendation results faster
Meet the expection and experience of user, quick response user may be incorporated into by control theory to the feedback of program recommendation results
Double-closed-loop control model, then step S103 can show as:Based on first-order transfer function model, according to renewal after above-mentioned two
Kind of above program recommended models computationally state weight during program recommendation list, and this for being calculated each program is consequently recommended
Index;The sequence of this consequently recommended index according to above-mentioned each program, is adjusted to above-mentioned program recommendation list.
Wherein, it is above-mentioned based on first-order transfer function model, existed according to the above two above program recommended models after renewal
Weight during above-mentioned program recommendation list is calculated, this consequently recommended index of each program is calculated, is specifically realized in
's:
gij=y1 ia1f1 ij+y2 ia2f2 ij+...yn ianfn ij
Wherein, gijRepresent the consequently recommended finger to program j by the user i after this renewal that the embodiment of the present invention is obtained
Number, yn iIt is directed to user i personalization difference modification proposed algorithms fnInitial weight an, that is, in step S102 obtain more
Weight after new.It can be seen that, the consequently recommended index S=a obtained when being initialized with program recommendation system1f1+a2f2+……+
anfnCompare, the consequently recommended index for obtaining herein introduces yn i, it is right to realize the trusting degree according to user to proposed algorithm
The weight that the program recommendation list of user carries out each program recommended models of adjustment of personalization is obtained in that real-time adjustment.It is optional
Ground, it would however also be possible to employ other Mathematical Modelings are calculated this consequently recommended index of each program, such as zero pole point residual
Model or zero pole point gain model etc., are not construed as limiting herein.
Alternatively, in order to preferably evaluate the degree of accuracy and the reliability of the program proposed algorithm of each program recommended models, on
Stating the feedback information obtained in step S101 can include user by the click of above-mentioned program recommendation list each program of click frequently
Secondary information, then now, above-mentioned steps S102 is embodied in:
According to the click frequency of above-mentioned each program, scoring of the generation to each program proposed algorithm of each program;
The scoring of each program proposed algorithm according to above-mentioned each program, is weighted to each program proposed algorithm asks respectively
With obtain the scoring of each program proposed algorithm;
Power after the scoring of each program proposed algorithm is updated as the corresponding program recommended models of each program proposed algorithm
Weight;
Wherein, the sequence due to each program in program recommendation list is to recommend to calculate by the program of each program recommended models
What method was determined, therefore, user is reflected by the number of times side that program recommendation list is clicked on to program and calculates the program
The degree of accuracy of each program proposed algorithm during consequently recommended index and reliability.When the click frequency by above-mentioned each program is generated
After for the scoring of each program proposed algorithm of each program, because we are final it is desirable that to each single program here
The scoring of proposed algorithm, so that from the scoring of each program proposed algorithm of each program obtained above, respectively to each section
Mesh proposed algorithm is weighted summation, to obtain the scoring of each program proposed algorithm, and the scoring of each program proposed algorithm is made
Weight after being updated for the corresponding program recommended models of each program proposed algorithm.
Specifically, the above-mentioned click frequency according to above-mentioned each program, generates commenting to each program proposed algorithm of each program
Point, can be achieved in that:
To any program, in acquisition actual program recommendation list, each program being calculated by above-mentioned program recommendation system
Consequently recommended index;
The independent of each program being calculated by each program proposed algorithm is obtained respectively recommends index;
Respectively by the independent recommendation index of above-mentioned each program and the ratio of the consequently recommended index of respective program, saved with corresponding
After purpose clicks on frequency multiplication, the scoring of each program proposed algorithm of respective program is obtained.
Wherein, corresponding independent recommendation index can be calculated to each program due to each program proposed algorithm, thus it is first
The ratio of each independent recommendation index and consequently recommended index is first calculated, then by the above-mentioned click frequency, is calculated respective program
Each program proposed algorithm scoring.In order to represent above-mentioned calculating process, can be first with the click frequency structure of above-mentioned acquisition
User's matrix B is built, above-mentioned user's matrix B can be expressed as:
Then program proposed algorithm fs of the user i to program jnScoring ynFor:
Wherein, above-mentioned is xn ijIn previous consequently recommended index sijIn accounting, xn ijIt is program proposed algorithm fnIndependence
Recommend index, sijIt is previous consequently recommended index, bijFor user i passes through the click frequency of the program recommendation list to program j.Tool
Body ground, clicks on once, then b whenever user i passes through program recommendation list to program jijPlus one.By above-mentioned user's matrix B according to section
The f of mesh proposed algorithmnAfter taking calculating apart, can obtain to program proposed algorithm fnRating matrix Yn, it is expressed as:
Above-mentioned matrix is sued for peace according to the coefficient of row matrix i, user i is obtained to program proposed algorithm fnEvaluation, represent
For:
Above-mentioned yn iAs user i is to program proposed algorithm fnScoring.
Alternatively, in order to avoid the above-mentioned user being calculated is to program proposed algorithm fnAppraisal result over-fitting, also
Following optimization can be carried out to calculate;
The scoring of each program proposed algorithm is added with default regularization coefficient, and result after will add up is divided by default
Normalization coefficient, obtain each program proposed algorithm optimization scoring;
Then now, the above-mentioned result using after above-mentioned weighted sum is used as the corresponding program recommended models of each program proposed algorithm
Weight after renewal, specially:
The optimization of above-mentioned each program proposed algorithm is scored as the corresponding program recommended models of each program proposed algorithm more
Weight after new.
Wherein, the result of calculation of above-mentioned optimization scoring is realized in:
In above formula, m is default normalization coefficient, cnIt is default regularization coefficient.Herein, user i is to being based on
Statistical proposed algorithm f1Be evaluated as y1 i;To the proposed algorithm f based on the content degree of correlation2Be evaluated as y2 i;By that analogy,
User is to other proposed algorithms fnBe evaluated as yn i.By introducing normalization coefficient and regularization coefficient, can effectively prevent from obtaining
The over-fitting of the scoring to each program proposed algorithm for obtaining.
To make above-mentioned steps clearer, below enumerate specific example and elaborate, it is necessary to illustrate, be limited to a piece
Width, following examples may eliminate the optional implementation process of some of above-mentioned steps:
Assuming that there is program 1 in system, program 2, and only one of which user selected program.
Assuming that being based on statistical proposed algorithm f1And the proposed algorithm f based on the content degree of correlation2To independently pushing away for the user
Recommend index as follows;
f1 11(T)=0.7, f1 12(T)=0.2, f2 11(T)=0.1, f2 12(T)=0.8
Assuming that wherein f1And f2Initial weight be respectively:
a1=1, a2=1
Then program 1 and program 2 are to the consequently recommended index of this user:
S11=1 × 0.7+1 × 0.1=0.8
S12=1 × 0.2+1 × 0.8=1
So consequently recommended exponential matrix is:
S=[0.8,1]
Wherein the recommendation scores of program 1 are 0.8, and the recommendation scores of program 2 are 1.
Assuming that the user is by multiple viewed programs, so as to generate influence on commending system.Clicking on program 1 has 4 times, and
Clicking on program 2 has 1 time, so as to obtain user matrix B=[4,1] of the user to program, by splitting, by user to the He of program 1
The scoring of program 2 is converted to program proposed algorithm f1It is scored at:
To program proposed algorithm f2It is scored at:
It is to simplify to calculate in calculating above, normalization coefficient m therein is adjusted to 1, regularization coefficient cnIt is adjusted to 0.
Then program proposed algorithm f is combined1And f2And initial weight, can obtain:
g11=3.7 × 1 × 0.7+1.3 × 1 × 0.1=2.59+0.13=2.72
g12=3.7 × 1 × 0.2+1.3 × 1 × 0.8=0.74+1.04=1.78
Consequently recommended index after the renewal that i.e. this is obtained, program 1 is 2.72, and program 2 is 1.78.Institute is calculated as above
Show, initial program 1 is lower than the consequently recommended index of program 2, the program 1 for repeatedly selecting initial evaluation score relatively low by user
Behavior and transmission function calculate, the new consequently recommended index of program 1 becomes higher than program 2, more meets the hobby of user
And operating experience.
Therefore, the embodiment of the present invention obtains the feedback information that user is based on program recommendation system feedback first, above-mentioned
Program recommendation system is used to calculate program recommendation list, and above-mentioned program recommendation system is based on two or more program recommended models structures
Build, wherein, above two above program recommended models are built by different program proposed algorithms respectively, are then based on above-mentioned feedback
Information updating above two above program recommended models computationally state weight during program recommendation list, after renewal
Weight of above two above program recommended models when computationally stating program recommendation list, update program recommendation list.
During this, the feedback information that the feedback information of user, particularly user use recommendation list is taken full advantage of, form user
Scoring to program proposed algorithm, the weight of each program recommended models, makes final pushing away in dynamic adjustment program recommendation system
Result is recommended to more conform to the changes in demand of user and experience result.And because the embodiment of the present invention also introduces two close cycles control
Simulation so that the dynamic property of program recommendation system is higher, starting performance more preferably, being capable of quick relative users Behavioral change pair
The influence of program proposed algorithm.
One of ordinary skill in the art will appreciate that all or part of step in realizing above-described embodiment method can be
The hardware of correlation is instructed to complete by program, corresponding program can be stored in a computer read/write memory medium,
Above-mentioned storage medium, such as ROM/RAM, disk or CD.
Embodiment two
Fig. 2 shows the concrete structure block diagram of the program recommendation apparatus that the embodiment of the present invention two is provided, for convenience of description,
Illustrate only the part related to the embodiment of the present invention.The program recommendation apparatus 2 include:Feedback information acquiring unit 21, weight
Updating block 22, program recommendation list updating block 23.
Wherein, feedback information acquiring unit 21, the feedback information of program recommendation system feedback is based on for obtaining user, on
Program recommendation system is stated for calculating program recommendation list, and above-mentioned program recommendation system is based on two or more program recommended models
Build, wherein, above two above program recommended models are built by different program proposed algorithms respectively;
Weight updating block 22, the feedback information for being got based on above-mentioned feedback information acquiring unit 21 updates above-mentioned
Two or more program recommended models computationally state weight during program recommendation list;
Program recommendation list updating block 23, for above-mentioned two after the renewal that is obtained according to above-mentioned weight updating block 22
Kind above program recommended models computationally state weight during program recommendation list, update program recommendation list.
Alternatively, above-mentioned feedback information acquiring unit 21, specifically includes:
Click on the frequency and obtain subelement, for obtaining user by the click of above-mentioned program recommendation list each program of click frequently
It is secondary;
Above-mentioned weight updating block 22, specifically includes:
The preliminary generation subelement of scoring, the click for obtaining each program that subelement gets according to the above-mentioned click frequency
The frequency, scoring of the generation to each program proposed algorithm of each program;
Weighted sum computation subunit, each program of each program for being generated according to above-mentioned algorithm scoring generation subelement
The scoring of proposed algorithm, is weighted summation to each program proposed algorithm respectively, obtains the scoring of each program proposed algorithm;
Weight determination subelement, for each program proposed algorithm for being calculated above-mentioned weighted sum computation subunit
The weight scored after being updated as the corresponding program recommended models of each program proposed algorithm.
Alternatively, above-mentioned scoring tentatively generates subelement, specifically includes:
Consequently recommended index obtains subelement, for any program, in acquisition actual program recommendation list, by above-mentioned section
The consequently recommended index of each program that mesh commending system is calculated;
It is independent to recommend index to obtain subelement, for obtaining each program being calculated by each program proposed algorithm respectively
It is independent to recommend index;
Score calculation subelement, for respectively by the only of the above-mentioned independent each program for recommending index acquisition subelement to get
The ratio of the vertical consequently recommended index for recommending index that the respective program that subelement gets is obtained with above-mentioned consequently recommended index, with
After the click frequency of respective program is multiplied, the scoring of each program proposed algorithm of respective program is obtained.
Alternatively, above-mentioned program recommendation apparatus 2 also include:
Optimization score calculation unit, for each program proposed algorithm for being calculated above-mentioned weighted sum computation subunit
Scoring be added with default regularization coefficient, and result after will add up is divided by default normalization coefficient, obtains each program
The optimization scoring of proposed algorithm;
Now, above-mentioned weight determination subelement, specifically for the optimization of above-mentioned each program proposed algorithm is scored as respectively
Weight after the corresponding program recommended models renewal of program proposed algorithm.
Alternatively, above-mentioned program recommendation list updating block 23, specifically includes:
Consequently recommended index computation subunit, for based on first-order transfer function model, according to above-mentioned weight updating block
Above two above program recommended models after 22 renewals for obtaining computationally state weight during program recommendation list, calculate
To this consequently recommended index of each program;
Recommendation list order adjustment subelement, it is each for what is be calculated according to above-mentioned consequently recommended index computation subunit
The sequence of this consequently recommended index of program, is adjusted to above-mentioned program recommendation list.
Therefore, the program recommendation apparatus of the embodiment of the present invention can obtain user and be based on program recommendation system feedback
Feedback information, above-mentioned program recommendation system is used to calculate program recommendation list, and above-mentioned program recommendation system based on two or more
Program recommended models build, wherein, above two above program recommended models are built by different program proposed algorithms respectively, so
Weight during program recommendation list is computationally stated based on above-mentioned feedback information renewal above two above program recommended models afterwards,
Weight during program recommendation list is computationally stated finally according to the above two above program recommended models after renewal, section is updated
Mesh recommendation list.In the process, program recommendation apparatus take full advantage of the feedback information of user, particularly user and use recommendation
The feedback information of list, forms scoring of the user to program proposed algorithm, each program in dynamic adjustment program recommendation system
The weight of recommended models, makes final recommendation results more conform to the changes in demand of user and experience result.
It should be noted that in several embodiments provided herein, it should be understood that disclosed device and side
Method, can realize by another way.For example, device embodiment described above is only schematical, for example, above-mentioned
The division of unit, only a kind of division of logic function, can there is other dividing mode when actually realizing, such as multiple units
Or component can be combined or be desirably integrated into another system, or some features can be ignored, or not perform.It is another, institute
Display or the coupling each other for discussing or direct-coupling or communication connection can be by some interfaces, device or unit
INDIRECT COUPLING or communication connection, can be electrical, mechanical or other forms.
For foregoing each method embodiment, in order to simplicity is described, therefore it is all expressed as a series of combination of actions, but
It is that those skilled in the art should know, the present invention is not limited by described sequence of movement, because according to the present invention, certain
A little steps can sequentially or simultaneously be carried out using other.Secondly, those skilled in the art should also know, be retouched in specification
The embodiment stated belongs to preferred embodiment, necessary to involved action and module might not all be the present invention.
In the above-described embodiments, the description to each embodiment all emphasizes particularly on different fields, and does not have the portion described in detail in certain embodiment
Point, may refer to the associated description of other embodiments.
It is more than to a kind of preferred embodiment provided by the present invention, for those of ordinary skill in the art, according to this
The thought of inventive embodiments, will change in specific embodiments and applications, and to sum up, this specification content is not
It is interpreted as limitation of the present invention.
Claims (10)
1. a kind of program commending method, it is characterised in that the program commending method includes:
The feedback information that user is based on program recommendation system feedback is obtained, the program recommendation system is used to calculate program recommendation row
Table, and the program recommendation system is based on two or more program recommended models structure, wherein, described two above programs recommend moulds
Type is built by different program proposed algorithms respectively;
Power of described two above program recommended models when the program recommendation list is calculated is updated based on the feedback information
Weight;
The weight of described two above program recommended models after according to renewal when the program recommendation list is calculated, updates section
Mesh recommendation list.
2. program commending method as claimed in claim 1, it is characterised in that it is anti-that the acquisition user is based on program recommendation system
The feedback information of feedback, including:
Obtain the click frequency that user clicks on each program by the program recommendation list;
It is described that described two above program recommended models are updated when the program recommendation list is calculated based on the feedback information
Weight, specially:
According to the click frequency of each program, scoring of the generation to each program proposed algorithm of each program;
The scoring of each program proposed algorithm according to each program, is weighted summation to each program proposed algorithm respectively, obtains
To the scoring of each program proposed algorithm;
Weight after the scoring of each program proposed algorithm is updated as the corresponding program recommended models of each program proposed algorithm.
3. program commending method as claimed in claim 2, it is characterised in that the click frequency according to each program,
The scoring to each program proposed algorithm of each program is generated, including:
To any program, in obtaining actual program recommendation list, each program being calculated by the program recommendation system is most
Recommend index eventually;
The independent of each program being calculated by each program proposed algorithm is obtained respectively recommends index;
Respectively by the ratio of the consequently recommended index of the independent recommendation index and respective program of each program, with respective program
After clicking on frequency multiplication, the scoring of each program proposed algorithm of respective program is obtained.
4. program commending method as claimed in claim 2, it is characterised in that the commenting according to each program proposed algorithm
Point, summation is weighted to each program proposed algorithm respectively, the scoring of each program proposed algorithm is obtained, also include afterwards:
The scoring of each program proposed algorithm is added with default regularization coefficient, and result after will add up is returned divided by default
One changes coefficient, obtains the optimization scoring of each program proposed algorithm;
The result using after the weighted sum updated as the corresponding program recommended models of each program proposed algorithm after power
Weight, specially:
After the optimization scoring of each program proposed algorithm is updated as the corresponding program recommended models of each program proposed algorithm
Weight.
5. the program commending method as described in any one of Claims 1-4, it is characterised in that described according to after renewal
Weight of the two or more program recommended models when the program recommendation list is calculated, updates program recommendation list, including:
Based on first-order transfer function model, according to renewal after described two above program recommended models push away calculating the program
Weight during list is recommended, this consequently recommended index of each program is calculated;
The sequence of this consequently recommended index according to each program, is adjusted to the program recommendation list.
6. a kind of program recommendation apparatus, it is characterised in that the program recommendation apparatus include:
Feedback information acquiring unit, the feedback information of program recommendation system feedback is based on for obtaining user, and the program is recommended
System is used to calculate program recommendation list, and the program recommendation system is built based on two or more program recommended models, wherein,
Described two above program recommended models are built by different program proposed algorithms respectively;
Weight updating block, the feedback information for being got based on the feedback information acquiring unit is updated more than described two
Weight of the program recommended models when the program recommendation list is calculated;
Program recommendation list updating block, for more than described two after the renewal that is obtained according to the weight updating block saving
Weight of the mesh recommended models when the program recommendation list is calculated, updates program recommendation list.
7. program recommendation apparatus as claimed in claim 6, it is characterised in that the feedback information acquiring unit, specifically include:
Click on the frequency and obtain subelement, click on the click frequency of each program by the program recommendation list for obtaining user;
The weight updating block, specifically includes:
The preliminary generation subelement of scoring, for each program that subelement gets is obtained according to the click frequency click frequently
It is secondary, scoring of the generation to each program proposed algorithm of each program;
Weighted sum computation subunit, each program of each program for being generated according to algorithm scoring generation subelement is recommended
The scoring of algorithm, is weighted summation to each program proposed algorithm respectively, obtains the scoring of each program proposed algorithm;
Weight determination subelement, the scoring of each program proposed algorithm for the weighted sum computation subunit to be calculated
Weight after being updated as the corresponding program recommended models of each program proposed algorithm.
8. program recommendation apparatus as claimed in claim 7, it is characterised in that the scoring tentatively generates subelement, including:
Consequently recommended index obtains subelement, for any program, obtaining in actual program recommendation list, is pushed away by the program
Recommend the consequently recommended index of each program that system-computed is obtained;
It is independent to recommend index to obtain subelement, the independence for obtaining each program being calculated by each program proposed algorithm respectively
Recommend index;
Score calculation subelement, for independently pushing away the independent each program for recommending index acquisition subelement to get respectively
The ratio that index obtains the consequently recommended index of the respective program that subelement gets with the consequently recommended index is recommended, it is and corresponding
After the click frequency of program is multiplied, the scoring of each program proposed algorithm of respective program is obtained.
9. program recommendation apparatus as claimed in claim 7, it is characterised in that the program recommendation apparatus also include:
Optimization score calculation unit, for commenting for each program proposed algorithm for being calculated the weighted sum computation subunit
Point is added with default regularization coefficient, and result after will add up is divided by default normalization coefficient, obtains each program recommendation
The optimization scoring of algorithm;
The weight determination subelement, specifically for recommending to calculate the optimization scoring of each program proposed algorithm as each program
Weight after the corresponding program recommended models renewal of method.
10. program recommendation apparatus as described in any one of claim 6 to 9, it is characterised in that the program recommendation list updates
Unit, including:
Consequently recommended index computation subunit, for based on first-order transfer function model, being obtained according to the weight updating block
Renewal after described two above program recommended models calculate the program recommendation list when weight, be calculated each section
Purpose this consequently recommended index;
Recommendation list order adjustment subelement, for each program being calculated according to the consequently recommended index computation subunit
This consequently recommended index sequence, the program recommendation list is adjusted.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710107905.2A CN106878772B (en) | 2017-02-27 | 2017-02-27 | Program recommendation method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710107905.2A CN106878772B (en) | 2017-02-27 | 2017-02-27 | Program recommendation method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106878772A true CN106878772A (en) | 2017-06-20 |
CN106878772B CN106878772B (en) | 2020-04-07 |
Family
ID=59168898
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710107905.2A Active CN106878772B (en) | 2017-02-27 | 2017-02-27 | Program recommendation method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106878772B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107948752A (en) * | 2017-11-14 | 2018-04-20 | 广州虎牙信息科技有限公司 | Subscribe to main broadcaster's sort method, device and terminal |
CN110337012A (en) * | 2019-05-08 | 2019-10-15 | 未来电视有限公司 | Intelligent recommendation method and apparatus based on internet television platform |
CN110825971A (en) * | 2019-11-11 | 2020-02-21 | 辽宁师范大学 | Article cold start recommendation algorithm integrating relationship mining and collaborative filtering |
CN112633321A (en) * | 2020-11-26 | 2021-04-09 | 北京瑞友科技股份有限公司 | Artificial intelligence recommendation system and method |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101431694A (en) * | 2008-11-26 | 2009-05-13 | 深圳市天威视讯股份有限公司 | Digital television program recommending method and system based on Bayesian algorithm |
CN102523511A (en) * | 2011-11-09 | 2012-06-27 | 中国传媒大学 | Network program aggregation and recommendation system and network program aggregation and recommendation method |
CN105760544A (en) * | 2016-03-16 | 2016-07-13 | 合网络技术(北京)有限公司 | Video recommendation method and device |
CN106454431A (en) * | 2016-10-14 | 2017-02-22 | 合肥工业大学 | Method and system for recommending television programs |
-
2017
- 2017-02-27 CN CN201710107905.2A patent/CN106878772B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101431694A (en) * | 2008-11-26 | 2009-05-13 | 深圳市天威视讯股份有限公司 | Digital television program recommending method and system based on Bayesian algorithm |
CN102523511A (en) * | 2011-11-09 | 2012-06-27 | 中国传媒大学 | Network program aggregation and recommendation system and network program aggregation and recommendation method |
CN105760544A (en) * | 2016-03-16 | 2016-07-13 | 合网络技术(北京)有限公司 | Video recommendation method and device |
CN106454431A (en) * | 2016-10-14 | 2017-02-22 | 合肥工业大学 | Method and system for recommending television programs |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107948752A (en) * | 2017-11-14 | 2018-04-20 | 广州虎牙信息科技有限公司 | Subscribe to main broadcaster's sort method, device and terminal |
CN107948752B (en) * | 2017-11-14 | 2021-01-08 | 广州虎牙信息科技有限公司 | Ordering method, device and terminal for subscription anchor |
CN110337012A (en) * | 2019-05-08 | 2019-10-15 | 未来电视有限公司 | Intelligent recommendation method and apparatus based on internet television platform |
CN110337012B (en) * | 2019-05-08 | 2021-07-20 | 未来电视有限公司 | Intelligent recommendation method and device based on Internet television platform |
CN110825971A (en) * | 2019-11-11 | 2020-02-21 | 辽宁师范大学 | Article cold start recommendation algorithm integrating relationship mining and collaborative filtering |
CN110825971B (en) * | 2019-11-11 | 2023-04-14 | 辽宁师范大学 | Article cold start recommendation algorithm integrating relationship mining and collaborative filtering |
CN112633321A (en) * | 2020-11-26 | 2021-04-09 | 北京瑞友科技股份有限公司 | Artificial intelligence recommendation system and method |
Also Published As
Publication number | Publication date |
---|---|
CN106878772B (en) | 2020-04-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6873979B2 (en) | Method of building predictive models on transactional data | |
US8924314B2 (en) | Search result ranking using machine learning | |
CN106878772A (en) | A kind of program commending method and device | |
CN106327339A (en) | Second-pass ranker for push notifications in a social network | |
AU2013251195B2 (en) | Program, apparatus, and method for information processing | |
Paulson et al. | Efficient large-scale internet media selection optimization for online display advertising | |
CN108460082A (en) | A kind of recommendation method and device, electronic equipment | |
CN108230058A (en) | Products Show method and system | |
CN105260390A (en) | Group-oriented project recommendation method based on joint probability matrix decomposition | |
US20130132328A1 (en) | Survey Feasibility Estimator | |
CN107194729A (en) | Advertisement price competing method, device, electronic installation and computer-readable medium | |
CN110766427B (en) | Advertisement bidding method and system | |
US11734589B2 (en) | Virtual assistant negotiation system and method thereof | |
CN103177371B (en) | A kind of information displaying method and device | |
EP2573725A1 (en) | Survey prioritization engine | |
CN111079000B (en) | Content processing method, device, equipment and storage medium | |
CN105915957A (en) | Intelligent television playing content display method, device and system | |
CN113688306A (en) | Recommendation strategy generation method and device based on reinforcement learning | |
Park | An adaptive match-making system reflecting the explicit and implicit preferences of users | |
CN116737906A (en) | Information display method, device, electronic equipment and storage medium | |
CN110502639A (en) | Information recommendation method, device and computer equipment based on problem contribution degree | |
US20160055519A1 (en) | Apportioning a media campaign contribution to a media channel in the presence of audience saturation | |
CN114169906B (en) | Electronic coupon pushing method and device | |
CN115345635A (en) | Processing method and device for recommended content, computer equipment and storage medium | |
CN108228833B (en) | Method for solving community project recommendation task by utilizing user tendency learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 518000 Guangdong city of Shenzhen province Nanshan District Guangdong streets Gao Xin Road No. 016 three storey building on the east side of Lenovo Applicant after: Excellent network Co., Ltd. Address before: 518000 Guangdong city of Shenzhen province Nanshan District Guangdong streets Gao Xin Road No. 016 three storey building on the east side of Lenovo Applicant before: UT Starcom (Shenzhen) Technology Co., Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |