CN107454474A - A kind of television terminal program personalized recommendation method based on collaborative filtering - Google Patents
A kind of television terminal program personalized recommendation method based on collaborative filtering Download PDFInfo
- Publication number
- CN107454474A CN107454474A CN201710708711.8A CN201710708711A CN107454474A CN 107454474 A CN107454474 A CN 107454474A CN 201710708711 A CN201710708711 A CN 201710708711A CN 107454474 A CN107454474 A CN 107454474A
- Authority
- CN
- China
- Prior art keywords
- mrow
- user
- msub
- program
- mtd
- 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]
- H04N21/40—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
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
-
- 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
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4667—Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The invention discloses a kind of television terminal program personalized recommendation method based on collaborative filtering, collaborative filtering based on article, mainly it is carried out in two steps, the similitude between article and article is calculated by consumer articles rating matrix first, then preference of the user to other articles is calculated using user's history score data and article article similarity matrix, article collaborative filtering mainly make use of the group wisdom between user to calculate the similitude between article, and then calculate preference of the user to article.The present invention directly obtains first against television user's history score data is more difficult, constructs a kind of marking strategy and the viewing behavior of user is converted into score data, then calculate article and article similitude using the score data of structure, finally realize and recommend.
Description
Technical field
The present invention relates to big data applied technical field, more particularly to a kind of television terminal program based on collaborative filtering
Propertyization recommends method.
Background technology
With the development of Internet technology, the information content that people get from network is increasing, how to believe from magnanimity
Quick obtaining is always the study hotspot of computer application field to effective information in breath.Wherein, proposed algorithm has become solution
Certainly in internet environment " information overload " problem important means.At present, recommended technology is widely used in ecommerce, social activity
In the Internet, applications such as network, such as recommended technology is introduced campus books by Patent No. CN201410212505.4 patent
In personalized recommendation.But the cases that such recommended technology is applied to television terminal few at present.
Nowadays TV programme are numerous, and user is difficult to quickly find oneself real TV programme interested.In magnanimity
In face of TV programme, television terminal user how is helped to be quickly found out the program for wanting to see, already as the research heat of TV industry
Point.Therefore, wide variety of recommended technology in internet is introduced into TV, user can be helped more to efficiently find viewing section
Mesh.This patent is based on the actual demand of TV applications, investigates the information resources that television terminal is readily available, reasonable in design
Recommendation method, wisdom is cooperateed with using the collective of television terminal user, recommends the program that they may need to user, helps user
Select to want the program seen personalizedly, so as to improve the utilization rate of television terminal, terminal user's retention ratio.
The content of the invention
Part in view of the shortcomings of the prior art, it is an object of the invention to provide a kind of TV based on collaborative filtering
Terminal program personalized recommendation method, the present invention directly obtain first against television user's history score data is more difficult, structure
The viewing behavior of user is converted to score data by a kind of marking strategy, then calculates article using the score data of structure
With article similitude, finally realize and recommend.
The purpose of the present invention is achieved through the following technical solutions：
A kind of television terminal program personalized recommendation method based on collaborative filtering, its method and step are as follows：
A, user-program-rating matrix R is built according to the history viewing behavior of user；
User's viewing behavior of nearest T_day days is extracted from the viewing behavior of collection as algorithm source data, to algorithm
Source data carries out following two steps cleaning treatment：
A1, remove excessively active user：The viewing feelings of all users are counted from user's viewing behavior of T_day days
Condition, takes the relatively inactive user of certain proportion to participate in subsequent step B, and certain proportion is whole user institute accounting X1；Wherein,
The liveness of any active ues refers to the number of programs of the unique user viewing counted in user's viewing behavior of T_day days；
A2, reject excessively popular program：The feelings that all programs are watched are counted from user's viewing behavior of T_day days
Condition, the relatively non-popular program of certain proportion is taken to participate in subsequent step B；Wherein, the popular degree of popular program referred at T_day days
User's viewing behavior in single program how many user for counting watched；
Data after step A cleanings have m user, and user is expressed as user, n program, and program is expressed as item,
Then R is expressed as：
In formula, ruiRepresent user useruTo program itemiScoring；
Wherein ruiBuild in the following way：
In formula, null represents user useruProgram item is not watchedi, C is user useruTo program itemiThat watches is total
Number, T are program itemiTotal duration, tcRepresent user useruViewing program item every timeiDuration, α is contraction factor,The company of expression multiplies symbol；
In formula,It is allSet, above formula represent to after structure score data carry out max-min normalization at
Reason；
B, program-program-similarity matrix W, program-program-similarity matrix are calculated according to consumer articles rating matrix R
W is expressed as：
In formula, wijRepresent program itemiWith program itemjSimilitude；
Scoring of all users that two programs of taking-up are each watched to it, and take wherein publicly-owned user to comment them
It is divided to calculate the similitude of two programs；The expression of specific mathematicization is as follows：
Wherein, UijExpression both have viewed itemiItem is have viewed againjUser, | Uij| expression both have viewed itemiSee again
Item is seenjNumber of users,Represent all and have viewed program itemiUser to program itemiAverage score；λ is flat
Sliding parameter,For correction term；
C, user's program preferences matrix is calculated according to user-program-rating matrix R and program-program-similarity matrix W
P, preference matrix P represent as follows：
In formula, puiRepresent the user user that algorithm predictsuTo program itemiPreference；
, can be individual to all n to calculate m user according to user's program score data and program-program similarity matrix W
The preference of program；Find user useruScored and itemiMost like K program calculates useruTo itemiIt is inclined
Good, the expression of specific mathematicization is as follows：
Wherein, Nu(i) represent in the program that scored of user with itemiIt is immediate | Nu(i) | individual user, | Nu(i)|≤
K, λ=100 are smoothing parameter, and K is arest neighbors number；
D, the Top-N recommendation list results L to user is formed：According to user preference matrix P, recommendation list result L is obtained,
The expression of specific mathematicization is as follows：
Wherein, Top-N (pui) represent to take family useruPreference puiTop-N；
E, offline evaluation, optimized algorithm model are carried out to algorithm：The quality of prediction effect is weighed using mean square error, its
Mathematicization represents as follows：
Unitary variant analytic approach is used using according to RMSE, carrys out in set-up procedure A to step D the parameter being related to, so as to calculate
Method recommendation effect is optimal offline；
F, the personalized recommendation to user is realized：When model debugging is to offline optimum state, test of reaching the standard grade can be carried out, it is real
Now to the personalized recommendation of user.
The present invention compared with the prior art, has advantages below and beneficial effect：
(1) collaborative filtering based on article of the invention, is mainly carried out in two steps, and is scored square by consumer articles first
Battle array calculates the similitude between article and article, is then calculated using user's history score data and article article similarity matrix
User to the preferences of other articles, article collaborative filtering mainly make use of the group wisdom between user calculate article it
Between similitude, and then calculate user to the preference of article.
(2) present invention directly obtains first against television user's history score data is more difficult, constructs a kind of marking plan
The viewing behavior of user is slightly converted into score data, it is similar to article then to calculate article using the score data of structure
Property, finally realize and recommend.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of the present invention.
Embodiment
The present invention is described in further detail with reference to embodiment：
Embodiment one
As shown in figure 1, a kind of television terminal program personalized recommendation method based on collaborative filtering, its method and step is such as
Under：
A, user-program-rating matrix R is built according to the history viewing behavior of user；
User's viewing behavior of nearest T_day days is extracted from the viewing behavior of collection as algorithm source data, to algorithm
Source data carries out following two steps cleaning treatment：
A1, remove excessively active user：The viewing feelings of all users are counted from user's viewing behavior of T_day days
Condition, takes the relatively inactive user of certain proportion to participate in subsequent step B, and certain proportion is whole user institute accounting X1；Wherein,
The liveness of any active ues refers to the number of programs of the unique user viewing counted in user's viewing behavior of T_day days；
A2, reject excessively popular program：The feelings that all programs are watched are counted from user's viewing behavior of T_day days
Condition, the relatively non-popular program of certain proportion is taken to participate in subsequent step B；Wherein, the popular degree of popular program referred at T_day days
User's viewing behavior in single program how many user for counting watched；
Data after step A cleanings have m user, and user is expressed as user, n program, and program is expressed as item,
Then R is expressed as：
In formula, ruiRepresent user useruTo program itemiScoring；
Wherein ruiBuild in the following way：
In formula, null represents user useruProgram item is not watchedi, C is user useruTo program itemiThat watches is total
Number, T are program itemiTotal duration, tcRepresent user useruViewing program item every timeiDuration, α is contraction factor,The company of expression multiplies symbol；
In formula,It is allSet, above formula represent to after structure score data carry out max-min normalization at
Reason；
B, program-program-similarity matrix W, program-program-similarity matrix are calculated according to consumer articles rating matrix R
W is expressed as：
In formula, wijRepresent program itemiWith program itemjSimilitude；
Scoring of all users that two programs of taking-up are each watched to it, and take wherein publicly-owned user to comment them
It is divided to calculate the similitude of two programs；The expression of specific mathematicization is as follows：
Wherein, UijExpression both have viewed itemiItem is have viewed againjUser, | Uij| expression both have viewed itemiSee again
Item is seenjNumber of users,Represent all and have viewed program itemiUser to program itemiAverage score；λ is flat
Sliding parameter,For correction term；
C, user's program preferences matrix is calculated according to user-program-rating matrix R and program-program-similarity matrix W
P, preference matrix P represent as follows：
In formula, puiRepresent the user user that algorithm predictsuTo program itemiPreference；
, can be individual to all n to calculate m user according to user's program score data and program-program similarity matrix W
The preference of program；Find user useruScored and itemiMost like K program calculates useruTo itemiIt is inclined
Good, the expression of specific mathematicization is as follows：
Wherein, Nu(i) represent in the program that scored of user with itemiIt is immediate | Nu(i) | individual user, | Nu(i)|≤
K, λ=100 are smoothing parameter, and K is arest neighbors number；
D, the Top-N recommendation list results L to user is formed：According to user preference matrix P, recommendation list result L is obtained,
The expression of specific mathematicization is as follows：
Wherein, Top-N (pui) represent to take family useruPreference puiTop-N；
E, offline evaluation, optimized algorithm model are carried out to algorithm：The quality of prediction effect is weighed using mean square error, its
Mathematicization represents as follows：
Unitary variant analytic approach is used using according to RMSE, carrys out in set-up procedure A to step D the parameter being related to, so as to calculate
Method recommendation effect is optimal offline；
F, the personalized recommendation to user is realized：When model debugging is to offline optimum state, test of reaching the standard grade can be carried out, it is real
Now to the personalized recommendation of user.
Embodiment two
As shown in figure 1, a kind of television terminal program personalized recommendation method based on collaborative filtering, its method and step is such as
Under：
Step 1：User-program-rating matrix R is built according to the history viewing behavior of user
In user's viewing historical behavior of the television terminal of collection, it is typically only capable to obtain some implicit scores data, and
True marking of the user to film can not be obtained, it is therefore desirable to which rating matrix R is built by the viewing behavior of user.
It is now assumed that user's viewing behavior of nearest T_day days is extracted from the viewing behavior of collection as algorithm source data.
Effectively to prevent the influence of small part extremely any active ues (being probably potential malicious attack user), and suppressed thermal center
Influence of the mesh to personalized recommendation result, two step cleaning treatments need to be made to source data first：
The first step：Remove excessively active user
The viewing situation of all users is counted from user's viewing behavior of T_day days, takes certain proportion relatively inactive
User (depend on user radix have it is much) participate in subsequent step；Wherein, the liveness of any active ues referred at T_day days
User's viewing behavior in count unique user viewing number of programs, viewing number of programs it is more, this user is more active.
Second step：Reject excessively popular program
The situation that all programs are watched is counted from user's viewing behavior of T_day days, takes certain proportion relatively non-thermal
The program (for example, 90%) of door participates in subsequent step；Wherein, the popular degree of popular program refers to user's viewing at T_day days
Single program how many user counted in behavior watched, and the number of users of viewing is more, and this program is more popular.
It is assumed that the data after cleaning have m user (user), n program (item), then R is represented by：
Wherein, ruiRepresent user useruTo program itemiScoring.Wherein ruiBuild in the following way：
In formula, null represents user useruProgram item is not watchedi, C is user useruTo program itemiThat watches is total
Number, T are program itemiTotal duration, tcRepresent user useruViewing program item every timeiDuration, α is contraction factor,
The control scoring growth rate that repeatedly viewing triggers, such as desirable α=2,The company of expression multiplies symbol, because user is in certain week
May be multiple to same program viewing in phase, therefore above formula Section 2 calculates user to one by the way of tired multiply here
The actual preferences of program, if Section 3 represents that the number of viewing is a lot, it is believed that user enjoys a lot to this program, directly
Give full mark.
In formula,It is allSet, above formula represent to after structure score data carry out max-min normalizeds.
Step 2：Program-program-similarity matrix w is calculated according to consumer articles rating matrix R
Program-program-similarity matrix w is represented by：
In formula, wijRepresent program itemiWith program itemjSimilitude.
After having article score data, the similitude between program-program can be calculated with the thought of " collaboration ".It is counted
Thought is calculated to take out scoring of all users that each watch of two programs to it, takes scoring of the wherein publicly-owned user to them
To calculate the similitude of two programs.The expression of specific mathematicization is as follows：
Wherein, UijExpression both have viewed itemiItem is have viewed againjUser, | Uij| expression both have viewed itemiSee again
Item is seenjNumber of users,Represent all and have viewed program itemiUser to program itemiAverage score.λ is flat
Sliding parameter,Correction term introduces the concept of " anticaustic family frequency "[2], can effectively lift the confidence water of Similarity Measure result
It is flat.
Step 3：User's program preferences are calculated according to user-program-rating matrix R and program-program-similarity matrix w
Matrix P, preference matrix P are represented by：
In formula, puiRepresent the user user that algorithm predictsuTo program itemiPreference.
According to user's program score data and program-program similarity matrix, all n can be saved with calculating m user
Purpose preference.Its Computation schema is to find user useruScored and itemiMost like K program calculates useru
To itemiPreference.The expression of specific mathematicization is as follows：
Wherein, Nu(i) represent in the program that scored of user with itemiIt is immediate | Nu(i) | individual user, | Nu(i)|≤
K, λ=100 are smoothing parameter, and K is arest neighbors number.
Step 4：The Top-N recommendation list result L to user are formed, according to user preference matrix P, obtain recommendation list
As a result L.Basic thought is：Take family useruTop-N puiRecommended as recommendation results to user.Specific mathematicization
Expression it is as follows：
Wherein, Top-N (pui) represent to take family useruPreference puiTop-N.
Step 5：Offline evaluation, optimized algorithm model are carried out to algorithm；There are preference matrix P and the user's scoring of prediction
Matrix R, can design evaluatio index evaluation prediction effect quality.For example, prediction effect can be weighed using mean square error
It is good and bad.Mathematicization represents as follows：
It can use and unitary variant analytic approach is used according to RMSE, carry out the parameter that set-up procedure one is related into step 4, with
Make algorithm recommendation effect optimal offline.
Step 6：Realize the personalized recommendation to user；When model debugging is to offline optimum state, survey of reaching the standard grade can be carried out
Examination, realizes the personalized recommendation to user.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement made within refreshing and principle etc., should be included in the scope of the protection.
Claims (1)
- A kind of 1. television terminal program personalized recommendation method based on collaborative filtering, it is characterised in that：Its method and step is as follows：A, user-program-rating matrix R is built according to the history viewing behavior of user；User's viewing behavior of nearest T_day days is extracted from the viewing behavior of collection as algorithm source data, to algorithm source number According to the following two steps cleaning treatment of progress：A1, remove excessively active user：The viewing situation of all users is counted from user's viewing behavior of T_day days, is taken The relatively inactive user of certain proportion participates in subsequent step B, and certain proportion is whole user institute accounting X1；Wherein, it is active The liveness of user refers to the number of programs of the unique user viewing counted in user's viewing behavior of T_day days；A2, reject excessively popular program：The situation that all programs are watched is counted from user's viewing behavior of T_day days, The relatively non-popular program of certain proportion is taken to participate in subsequent step B；Wherein, the popular degree of popular program referred at T_day days Single program how many user counted in user's viewing behavior watched；Data after step A cleanings have m user, and user is expressed as user, n program, and program is expressed as item, then R It is expressed as：In formula, ruiRepresent user useruTo program itemiScoring；Wherein ruiBuild in the following way：<mrow> <mover> <msub> <mi>r</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>&OverBar;</mo> </mover> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>n</mi> <mi>u</mi> <mi>l</mi> <mi>l</mi> </mrow> </mtd> <mtd> <mrow> <mi>i</mi> <mi>f</mi> <mi> </mi> <msub> <mi>user</mi> <mi>u</mi> </msub> <msup> <mi>don</mi> <mo>&prime;</mo> </msup> <mi>t</mi> <mi> </mi> <mi>p</mi> <mi>l</mi> <mi>a</mi> <mi>y</mi> <mi> </mi> <msub> <mi>item</mi> <mi>i</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <munderover> <mi>&Pi;</mi> <mrow> <mi>c</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>C</mi> </munderover> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>&alpha;</mi> <mo>&CenterDot;</mo> <mfrac> <msub> <mi>t</mi> <mi>c</mi> </msub> <mi>T</mi> </mfrac> <mo>)</mo> </mrow> </mrow> </mtd> <mtd> <mrow> <mi>i</mi> <mi>f</mi> <munderover> <mi>&Pi;</mi> <mrow> <mi>c</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>C</mi> </munderover> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>&alpha;</mi> <mo>&CenterDot;</mo> <mfrac> <msub> <mi>t</mi> <mi>c</mi> </msub> <mi>T</mi> </mfrac> <mo>)</mo> </mrow> <mo>&le;</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>&alpha;</mi> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>1</mn> <mo>+</mo> <mi>&alpha;</mi> </mrow> </mtd> <mtd> <mrow> <mi>i</mi> <mi>f</mi> <munderover> <mi>&Pi;</mi> <mrow> <mi>c</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>C</mi> </munderover> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>&alpha;</mi> <mo>&CenterDot;</mo> <mfrac> <msub> <mi>t</mi> <mi>c</mi> </msub> <mi>T</mi> </mfrac> <mo>)</mo> </mrow> <mo>></mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>&alpha;</mi> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>In formula, null represents user useruProgram item is not watchedi, C is user useruTo program itemiThe total degree of viewing, T For program itemiTotal duration, tcRepresent user useruViewing program item every timeiDuration, α is contraction factor,Represent Company multiplies symbol；<mrow> <msub> <mi>r</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>=</mo> <mfrac> <mrow> <mover> <msub> <mi>r</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>&OverBar;</mo> </mover> <mo>-</mo> <mi>min</mi> <mrow> <mo>(</mo> <mover> <mi>r</mi> <mo>&OverBar;</mo> </mover> <mo>)</mo> </mrow> </mrow> <mrow> <mi>max</mi> <mrow> <mo>(</mo> <mover> <mi>r</mi> <mo>&OverBar;</mo> </mover> <mo>)</mo> </mrow> <mo>-</mo> <mi>min</mi> <mrow> <mo>(</mo> <mover> <mi>r</mi> <mo>&OverBar;</mo> </mover> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>,</mo> <mover> <mi>r</mi> <mo>&OverBar;</mo> </mover> <mo>=</mo> <mo>&cup;</mo> <mover> <msub> <mi>r</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>&OverBar;</mo> </mover> </mrow>In formula,It is allSet, above formula represent to after structure score data carry out max-min normalizeds；B, program-program-similarity matrix W, program-program-similarity matrix W are calculated according to consumer articles rating matrix R It is expressed as：In formula, wijRepresent program itemiWith program itemjSimilitude；C, user program preferences matrix P is calculated according to user-program-rating matrix R and program-program-similarity matrix W, partially Good matrix P represents as follows：In formula, puiRepresent the user user that algorithm predictsuTo program itemiPreference；D, the Top-N recommendation list results L to user is formed：According to user preference matrix P, recommendation list result L is obtained, specifically The expression of mathematicization is as follows：<mrow> <mi>L</mi> <mo>=</mo> <msub> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mrow> <mi>T</mi> <mi>o</mi> <mi>p</mi> <mo>-</mo> <mi>N</mi> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mrow> <mn>1</mn> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>T</mi> <mi>o</mi> <mi>p</mi> <mo>-</mo> <mi>N</mi> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mrow> <mn>2</mn> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mn>...</mn> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>T</mi> <mi>o</mi> <mi>p</mi> <mo>-</mo> <mi>N</mi> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mn>...</mn> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>T</mi> <mi>o</mi> <mi>p</mi> <mo>-</mo> <mi>N</mi> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mrow> <mi>m</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> </mtable> </mfenced> <mrow> <mi>m</mi> <mo>*</mo> <mi>N</mi> </mrow> </msub> </mrow>Wherein, Top-N (pui) represent to take family useruPreference puiTop-N；E, offline evaluation, optimized algorithm model are carried out to algorithm：The quality of prediction effect is weighed using mean square error, its mathematics Change and represent as follows：<mfenced open = "" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>R</mi> <mi>M</mi> <mi>S</mi> <mi>E</mi> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>u</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </munderover> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>r</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>-</mo> <msub> <mi>p</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </mtd> <mtd> <mrow> <mi>s</mi> <mo>.</mo> <mi>t</mi> <mo>.</mo> </mrow> </mtd> <mtd> <mrow> <msub> <mi>r</mi> <mrow> <mi>u</mi> <mi>i</mi> </mrow> </msub> <mo>&NotEqual;</mo> <mi>n</mi> <mi>u</mi> <mi>l</mi> <mi>l</mi> </mrow> </mtd> </mtr> </mtable> </mfenced>Unitary variant analytic approach is used using according to RMSE, carrys out in set-up procedure A to step D the parameter being related to, so that algorithm pushes away It is optimal offline to recommend effect；F, the personalized recommendation to user is realized：When model debugging is to offline optimum state, test of reaching the standard grade, realization pair can be carried out The personalized recommendation of user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710708711.8A CN107454474B (en) | 2017-08-17 | 2017-08-17 | A kind of television terminal program personalized recommendation method based on collaborative filtering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710708711.8A CN107454474B (en) | 2017-08-17 | 2017-08-17 | A kind of television terminal program personalized recommendation method based on collaborative filtering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107454474A true CN107454474A (en) | 2017-12-08 |
| CN107454474B CN107454474B (en) | 2019-11-05 |
Family
ID=60492424
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710708711.8A Active CN107454474B (en) | 2017-08-17 | 2017-08-17 | A kind of television terminal program personalized recommendation method based on collaborative filtering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107454474B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108287870A (en) * | 2017-12-27 | 2018-07-17 | 优地网络有限公司 | Intelligent method for running and device |
| CN108920577A (en) * | 2018-06-25 | 2018-11-30 | 西北工业大学 | Television set intelligently recommended method |
| CN109246450A (en) * | 2018-08-06 | 2019-01-18 | 上海大学 | A kind of video display preferentially recommender system and method based on implicit information scoring |
| CN109672938A (en) * | 2019-01-07 | 2019-04-23 | 河北工业大学 | A kind of IPTV program commending method |
| CN109862431A (en) * | 2019-01-23 | 2019-06-07 | 重庆第二师范学院 | A kind of TV programme mixed recommendation method based on MCL-HCF algorithm |
| CN110430471A (en) * | 2019-07-24 | 2019-11-08 | 山东海看新媒体研究院有限公司 | It is a kind of based on the television recommendations method and system instantaneously calculated |
| CN110737800A (en) * | 2019-10-14 | 2020-01-31 | 北京弘远博学科技有限公司 | similarity recommendation method based on video watched by students |
| CN112767085A (en) * | 2021-01-22 | 2021-05-07 | 武汉蔚来能源有限公司 | Commodity similarity analysis and commodity recommendation method, commodity similarity analysis and commodity recommendation device, and computer storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101489107A (en) * | 2009-01-21 | 2009-07-22 | 华东师范大学 | Collaborative filtering recommendation method based on population attribute keyword vector |
| US20130132982A1 (en) * | 2009-12-02 | 2013-05-23 | Nbcuniversal Media, Llc | Methods and systems for online recommendation |
| CN105338408A (en) * | 2015-12-02 | 2016-02-17 | 南京理工大学 | Video recommending method based on time factor |
-
2017
- 2017-08-17 CN CN201710708711.8A patent/CN107454474B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101489107A (en) * | 2009-01-21 | 2009-07-22 | 华东师范大学 | Collaborative filtering recommendation method based on population attribute keyword vector |
| US20130132982A1 (en) * | 2009-12-02 | 2013-05-23 | Nbcuniversal Media, Llc | Methods and systems for online recommendation |
| CN105338408A (en) * | 2015-12-02 | 2016-02-17 | 南京理工大学 | Video recommending method based on time factor |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108287870A (en) * | 2017-12-27 | 2018-07-17 | 优地网络有限公司 | Intelligent method for running and device |
| CN108287870B (en) * | 2017-12-27 | 2020-09-01 | 优地网络有限公司 | Intelligent operation method and device |
| CN108920577A (en) * | 2018-06-25 | 2018-11-30 | 西北工业大学 | Television set intelligently recommended method |
| CN109246450A (en) * | 2018-08-06 | 2019-01-18 | 上海大学 | A kind of video display preferentially recommender system and method based on implicit information scoring |
| CN109246450B (en) * | 2018-08-06 | 2021-07-06 | 上海大学 | Movie and television preferred recommendation method based on implicit information scoring |
| CN109672938A (en) * | 2019-01-07 | 2019-04-23 | 河北工业大学 | A kind of IPTV program commending method |
| CN109862431A (en) * | 2019-01-23 | 2019-06-07 | 重庆第二师范学院 | A kind of TV programme mixed recommendation method based on MCL-HCF algorithm |
| CN109862431B (en) * | 2019-01-23 | 2023-09-29 | 重庆第二师范学院 | MCL-HCF algorithm-based television program mixed recommendation method |
| CN110430471A (en) * | 2019-07-24 | 2019-11-08 | 山东海看新媒体研究院有限公司 | It is a kind of based on the television recommendations method and system instantaneously calculated |
| CN110737800A (en) * | 2019-10-14 | 2020-01-31 | 北京弘远博学科技有限公司 | similarity recommendation method based on video watched by students |
| CN112767085A (en) * | 2021-01-22 | 2021-05-07 | 武汉蔚来能源有限公司 | Commodity similarity analysis and commodity recommendation method, commodity similarity analysis and commodity recommendation device, and computer storage medium |
| CN112767085B (en) * | 2021-01-22 | 2024-05-24 | 武汉蔚来能源有限公司 | Commodity similarity analysis and commodity recommendation method and device and computer storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107454474B (en) | 2019-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107454474B (en) | A kind of television terminal program personalized recommendation method based on collaborative filtering | |
| US10664509B1 (en) | Processing non-uniform datasets | |
| CN104063481B (en) | A kind of film personalized recommendation method based on the real-time interest vector of user | |
| CN102332006B (en) | A kind of information push control method and device | |
| Liu et al. | Information filtering via weighted heat conduction algorithm | |
| Powell et al. | Broadening the focus of public service consumerism | |
| CN104361062B (en) | A kind of recommendation method and device of related information | |
| Ren et al. | Avoiding congestion in recommender systems | |
| CN102184364A (en) | Semi-supervised learning-based recommendation system shilling attack detection method | |
| CN107657034A (en) | A kind of event social networks proposed algorithm of social information enhancing | |
| CN103631949B (en) | A kind of social network data acquisition method and system | |
| CN108629010A (en) | It is a kind of to recommend method based on the web services of theme and service combination information | |
| CN108230169B (en) | Information propagation model based on social influence and situation perception system and method | |
| Chen et al. | Enhancing network cluster synchronization capability based on artificial immune algorithm | |
| US20130031093A1 (en) | Information processing system, information processing method, program, and non-transitory information storage medium | |
| Liu et al. | Personal recommendation via unequal resource allocation on bipartite networks | |
| CN103530428A (en) | Same-occupation type recommendation method based on developer practical skill similarity | |
| Wu et al. | Optimal computational power allocation in multi-access mobile edge computing for blockchain | |
| CN104933595A (en) | Collaborative filtering recommendation method based on Markov prediction model | |
| Zhou et al. | Statistical study of view preferences for online videos with cross-platform information | |
| Wang et al. | Heterogenous scaling in the inter-event time of on-line bookmarking | |
| Checco et al. | Blc: Private matrix factorization recommenders via automatic group learning | |
| WO2013034027A1 (en) | Method and system for recommending media information post | |
| Wang et al. | Latency-optimal computational offloading strategy for sensitive tasks in smart homes | |
| CN109640128A (en) | A kind of TV user watching behavior feature extracting method and system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |