CN108419134A - The recommendation of the channels method merged with group current behavior based on individual history - Google Patents
The recommendation of the channels method merged with group current behavior based on individual history Download PDFInfo
- Publication number
- CN108419134A CN108419134A CN201810110304.1A CN201810110304A CN108419134A CN 108419134 A CN108419134 A CN 108419134A CN 201810110304 A CN201810110304 A CN 201810110304A CN 108419134 A CN108419134 A CN 108419134A
- Authority
- CN
- China
- Prior art keywords
- channel
- matrix
- user
- state matrix
- viewing
- 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
-
- 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
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
-
- 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
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44213—Monitoring of end-user related data
-
- 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
- H04N21/4508—Management of client data or end-user data
- H04N21/4532—Management of client data or end-user data involving end-user characteristics, e.g. viewer profile, 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]
- H04N21/40—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
Abstract
The invention discloses a kind of recommendation of the channels methods merged with group current behavior based on individual history, the currently viewing channel state matrix of building group first, feature for describing current each channel, feature includes the current viewing number of each channel and the online viewing number rate of rise of each channel；It is directed to the history viewing channel state matrix that individual consumer builds its people, the viewing channels feature for describing the user in time in the past section, including the user again and watches the number of each channel and each by the weight of viewing channel；Finally group's current channel state matrix and individual history are watched channel state matrix and be sent into and recommends fusion calculation module, and choose method using intersection or calculate the channel that each user's current time may watch apart from selection method, the N number of channel most possibly watched is pushed to the user.
Description
Technical field
The present invention relates to the fields IPTV, more particularly to a kind of channel merged with group current behavior based on individual history is pushed away
Recommend method.
Background technology
IPTV be it is a kind of can interactive Web TV, before the television channel commending based on IPTV has tempting application
Scape.With the development of IPTV and internet television direct seeding technique, user can watch more and more television channels, therefore tradition
TV Guide (EPG) cannot meet the use demand that people have found its interested channel in time.In recent years, people
The personalized recommendation system that begins one's study for IPTV, however most of commending systems are not just in request program
Live streaming recommend because the following characteristics of direct broadcast band lead to that it recommend it is increasingly complex：
Timeliness：Live telecast content can only play in specific time window, if user wants to watch some program,
It needs to watch the direct broadcast band in specific time zone, and request program does not limit time zone, user can be arbitrary
Time is watched.
Complexity：One IPTV account is typically that family is shared, i.e., more people share the same account, this makes when different
Between in section, the hobby of user is different, to be more difficult to capture the behavior of user.
Noise：The viewing record of direct broadcast band, which compares program request, more noise datas.In live streaming, user has
The behavior and advertisement being much switched fast are skipped, these noise datas can interfere the recommendation results of user.
The characteristics of based on the above IPTV direct broadcast bands, excavated in a direct broadcast bands up to a hundred the channel liked of user seem compared with
For complexity.Therefore the commending system user that wants help finds its direct broadcast band not a duck soup liked in a channels up to a hundred.It pushes away
The system of recommending needs to carry out depth analysis to the history viewing behavior of user, and establishes suitable model, is carried out to user corresponding
Recommendation, current proposed algorithm mainly recommended using collaborative filtering or some other correlation machine learning methods.
The channel that collaborative filtering can be liked according to other users recommends target user, solves the problems, such as cold start-up.Engineering
The method of habit is also begun in the program recommendation applied to IPTV, such as is first gathered to user using the methods of k-means now
Class, and the method for merging collaborative filtering recommends user.But or these methods need additional user's scoring letter
It ceases or needs a large amount of machine to calculate to take, be found at present directly against effective proposed algorithm of direct broadcast band is still very few.
Invention content
The shortcomings that it is an object of the invention to overcome the prior art and deficiency are provided one kind and are worked as with group based on individual history
The recommendation of the channels method of preceding action amalgamation is come in conjunction with the history viewing behavior of user and group's viewing behavior of user for each use
Family carries out personalized channels recommendation.
The purpose of the present invention is realized by the following technical solution：One kind is merged based on individual history with group current behavior
Recommendation of the channels method, including：
For overall user, a current channel state matrix generator (101) is designed, is being watched for describing user
The current state of each channel when IPTV；For personal user, property historic state matrix generator (102) one by one is designed,
For describing each user in certain historical time, the state of each history viewing channel, and it is current by going out constructed by 101
Personalized historic state matrix constructed by channel state matrix and 102 is sent into user and recommends computing module (103), the module institute
Proposed algorithm can be that each user generates the rendition list recommended, and it is that user's progress is independent to be sent into pushing module (104)
Push.
One, it for the structure of current channel state matrix generator (101), comprises the steps of：
1. access time window delta t.
2. under pair total user viewing, statistics actual time window [t- Δs t, t) in, the channel temperature of each channel
pi。piThe viewing number for being each channel in time window.
3. under pair total user viewing, statistics actual time window [t- Δs t, t) in, each online number of channel
Rate of rise ri, riCalculation formula it is as follows：
4. couple piIt is normalized,PiFor to piValue after normalization, to riIt is normalized
Processing,RiFor to riValue after normalization.Build the currently viewing channel state square of group of current t moment
Battle array Wherein first row CiFor channel number, secondary series indicates the channel temperature of each channel after normalization,
Third row indicate that the instantaneous rate of increase of each channel after normalization, n are channel number.
Two, it for the structure of personalized historic state matrix generator (102), comprises the steps of：
1. choosing history sliding window Δ T, here Δ T ＞ ＞ Δs t.
2. for each user in IPTV, the time watched of the user each channel within the period [t- Δs T, t] is counted
oi,
3. for each user in IPTV, the weighted value w of the user each channel within the period [t- Δs T, t] is calculatedi,
wiCalculation formula iswiFor the channel C of user viewing in the period [t- Δs T, t]iWeight, Nci be in [t-
Δ T, t] the watched channel C of the user U in sectioniNumber, τkFor user's kth in user's history window time viewing channel Ci's
Weight, τkA kind of representation of calculation formula be：tkIt is that the user watches frequency in [t- Δs T, t] interior kth time
Road CiAt the time of residing, ts is the initial time of sliding window, i.e. ts=t- Δs T；T is current time, and Δ T is the sliding of history
Window size.In addition, τkCan also be expressed as with another calculation formula：T is current time, tkFor the use
At the time of family is residing for time window [t- Δs T, t] interior kth time viewing channel i, Δ T is sliding window size.
4. couple oiIt is normalized,OiFor to oiValue after normalization, to wiIt is normalized
Processing,WiFor to wiValue after normalization.Build the personalized historic state matrix H of each user, example
As the case history viewing channel state matrix of certain user User isMatrix each column indicates successively
The weight of channel number, channel time watched and channel, m are the channel numbers that the user is watched in the time window, and m is
Channel number.
Three, it is of the structures of the current channel state matrix C generated according to 101 and 102 that user, which recommends computing module (103),
Property historic state matrix H select suitable algorithm to be recommended, recommend the structure of computing module (103) for user, including
Following methods：
1. choosing method using intersection, comprise the following steps：
The 1.1 current channel state matrix C generated according to the step 4 of structure 101, by channel number CiAccording to CiFeature
It is ranked up, such as according to channel viewing temperature PiDescending arrangement, select popular TOP x channels, composition current candidate
Collection:V1=(C1,C2……Cx)；
1.2 are directed to the personalized historic state matrix that each user is generated according to the step 4 of structure 102, by channel number
Ci is according to CiFeature be ranked up, such as by each channel time watched OiDescending arrangement, select y popular channel, structure
At history Candidate Set:V2=(C1,C2……Cy)；
1.3 take V12=V1∩V2It, should if the maximum number K of channel is recommended to be greater than or equal to as the recommendation channel of user
The number of intersection element then recommends the channel in whole intersections to user；If the maximum number K for the channel recommended is less than the friendship
The number of the element of collection then presses the Top-K that the element of the intersection is recommended in the arrangement of channel temperature size descending.
2. distance chooses method, comprise the following steps：
2.1 according in the current channel state matrix C that is generated in 101 and 102 and personalized historic state matrix H, each
The characteristic value of channel is come each channel C in calculating 101iWith C in 102jDistance, constitute a distance matrix (201) Wherein DijIndicate 101 mid band Ci102 mid band C of distancejDistance；(each line index in matrix D
Each channel number being worth in homography C, each channel number in matrix D in each column index value homography H；)Dij
It is calculated using one of following formula：
Euclidean distance：Wherein, C (Fi)n101 mid band C of representing matrixi's
N-th of feature train value, H (Fi)n102 mid band C of representing matrixjN-th of feature train value, N be matrix characteristic series sum.
Manhatton distance：
COS distance：
2.2 according to matrix D, the index value arranged belonging to the minimum value in finding out matrix per a line, i.e., in 102 H-matrix
Channel number obtains channel vector V3=(C1,C2…..Cn), which indicates that each channel in matrix H distance matrix C is nearest
Channel；
2.3 by vectorial V3It is arranged according to 101 rate of rise descending of mid band, n fastest-rising channels are made before taking
To recommend.
Compared with prior art, the present invention having the following advantages that and advantageous effect：
The present invention is using the viewing behavior of the group of active user as one of the foundation recommended, in conjunction with the history of each user
Viewing behavior is recommended.It is every to watch channel state matrix by the case history of current channel state matrix and each user
A user carries out real-time recommendation, and method proposed by the invention can preferably capture the viewing behavior of each user.
It is that user recommends that proposed algorithm proposed by the invention, which can use less user behavior information,.Traditional
Television program recommendations need to obtain the relevant information of program or the rating matrix of user, but algorithm provided by the present invention can
To use less user behavior characteristics to be recommended, data tuple structure used in the present invention is { User ID, channel viewing
Initial time, channel number, channel watch duration, which is easier to obtain relative to programme information.
Intersection proposed by the invention chooses method and can more save computing resource when calculating apart from selection method.It is different
It is trained in using machine learning or deep learning, recommendation method proposed by the invention does not have a large amount of of training time to open
Pin, therefore current channel state matrix and personal viewing channel history state matrix can be generated within the short time, and be
User carries out real-time recommendation.
Description of the drawings
Fig. 1 is commending system basic framework schematic diagram.
Fig. 2 is the recommendation of the channels method flow diagram based on intersection that embodiment 2 provides.
Fig. 3 is 2 weight computing schematic diagram of embodiment.
Fig. 4 is the recommendation of the channels method flow diagram based on distance that embodiment 3 provides.
Specific implementation mode
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
Embodiment 1
A kind of recommendation of the channels method merged with group current behavior based on individual history, such as Fig. 1, including：
For overall user, a current channel state matrix generator (101) is designed, is being watched for describing user
The current state of each channel when IPTV；For personal user, property historic state matrix generator (102) one by one is designed,
For describing each user in certain historical time, the state of each history viewing channel, and it is current by going out constructed by 101
Personalized historic state matrix constructed by channel state matrix and 102 is sent into user and recommends computing module (103), the module institute
Proposed algorithm can be that each user generates the rendition list recommended, and it is that user's progress is independent to be sent into pushing module (104)
Push.
One, it for the structure of current channel state matrix generator (101), comprises the steps of：
1. access time window delta t.
2. under pair total user viewing, statistics actual time window [t- Δs t, t) in, the channel temperature of each channel
pi。piThe viewing number for being each channel in time window.
3. under pair total user viewing, statistics actual time window [t- Δs t, t) in, each online number of channel
Rate of rise ri, riCalculation formula it is as follows：
4. couple piIt is normalized,PiFor to piValue after normalization, to riIt is normalized
Processing,RiFor to riValue after normalization.Build the currently viewing channel state square of group of current t moment
Battle array Wherein first row CiFor channel number, secondary series indicates the channel temperature of each channel after normalization,
Third row indicate that the instantaneous rate of increase of each channel after normalization, n are channel number.
Two, it for the structure of personalized historic state matrix generator (102), comprises the steps of：
1. choosing history sliding window Δ T, here Δ T ＞ ＞ Δs t.
2. for each user in IPTV, the time watched of the user each channel within the period [t- Δs T, t] is counted
oi,
3. for each user in IPTV, the weighted value w of the user each channel within the period [t- Δs T, t] is calculatedi,
wiCalculation formula iswiFor the channel C of user viewing in the period [t- Δs T, t]iWeight, Nci be in [t-
Δ T, t] the watched channel C of the user U in sectioniNumber, τkFor user's kth in user's history window time viewing channel Ci's
Weight, τkA kind of representation of calculation formula be：tkIt is that the user watches frequency in [t- Δs T, t] interior kth time
Road CiAt the time of residing, ts is the initial time of sliding window, i.e. ts=t- Δs T；T is current time, and Δ T is the sliding of history
Window size.In addition, τkCan also be expressed as with another calculation formula：T is current time, tkFor the use
At the time of family is residing for time window [t- Δs T, t] interior kth time viewing channel i, Δ T is sliding window size.
4. couple oiIt is normalized,OiFor to oiValue after normalization, to wiIt is normalized
Processing,WiFor to wiValue after normalization.Build the personalized historic state matrix H of each user, example
As the case history viewing channel state matrix of certain user User isMatrix each column indicates successively
The weight of channel number, channel time watched and channel, m are the channel numbers that the user is watched in the time window, and m is
Channel number.
Three, it is of the structures of the current channel state matrix C generated according to 101 and 102 that user, which recommends computing module (103),
Property historic state matrix H select suitable algorithm to be recommended, recommend the structure of computing module (103) for user, including
Following methods：
1. method is chosen using intersection, including as follows：
The 1.1 current channel state matrix C generated according to the step 4 of structure 101, by channel number CiAccording to CiFeature
It is ranked up, such as according to channel viewing temperature PiDescending arrangement, select popular TOP x channels, composition current candidate
Collection:V1=(C1,C2……Cx)；
1.2 are directed to the personalized historic state matrix that each user is generated according to the step 4 of structure 102, by channel number
CiAccording to CiFeature be ranked up, such as by each channel time watched OiDescending arrangement, select y popular channel, structure
At history Candidate Set:V2=(C1,C2……Cy)；
1.3 take V12=V1∩V2It, should if the maximum number K of channel is recommended to be greater than or equal to as the recommendation channel of user
The number of intersection element then recommends the channel in whole intersections to user；If the maximum number K for the channel recommended is less than the friendship
The number of the element of collection then presses the Top-K that the element of the intersection is recommended in the arrangement of channel temperature size descending.
Method rule is chosen in other intersections：
Regulation then 1 is chosen in intersection：According to currently viewing channel state matrix, by channel number according to the drop of channel rate of rise
Sequence arranges, and selects x most fast channel of rate of rise in the period, constitutes current candidate collection V1=(C1,C2……Cx)；According to
Case history watches channel state matrix, and channel number is arranged according to the descending of the time watched of channel, selects user's U viewings time
Y most channel of number, constitutes history Candidate Set V2=(C1,C2……Cy)；Take V1∩V2Recommendation channel as the user.
Regulation then 2 is chosen in intersection：According to currently viewing channel state matrix, by channel number according to the drop of channel rate of rise
Sequence arranges, and selects x most fast channel of rate of rise in the period, constitutes current candidate collection V1=(C1,C2……Cx)；According to
Case history watches channel state matrix, and channel number is arranged according to the weights descending of channel, selects y larger frequency of weights
Road constitutes history Candidate Set V2=(C1,C2……Cy)；Take V1∩V2Recommendation channel as the user.
Regulation then 3 is chosen in intersection：According to currently viewing channel state matrix, channel number is watched to the drop of number according to channel
Sequence arranges, and selects x most channel of viewing number, constitutes current candidate collection V1=(C1,C2……Cx)；According to case history
Channel state matrix is watched, channel number is subjected to descending arrangement according to the weights of channel, is selected maximum relative to user's weight
Y channel, constitute history Candidate Set V2=(C1,C2……Cy)；Take V1∩V2Recommendation channel as the user.
2. distance chooses method, comprise the following steps：
2.1 according in the current channel state matrix C that is generated in 101 and 102 and personalized historic state matrix H, each
The characteristic value of channel is come each channel C in calculating 101iWith C in 102jDistance, constitute a distance matrix (201) Wherein DijIndicate 101 mid band Ci102 mid band C of distancejDistance；(each line index in matrix D
Each channel number being worth in homography C, each channel number in matrix D in each column index value homography H；)Dij
It is calculated using one of following formula：
Euclidean distance：Wherein, C (Fi)n101 mid band C of representing matrixi's
N-th of feature train value, H (Fi)n102 mid band C of representing matrixjN-th of feature train value, N be matrix characteristic series sum.
Manhatton distance：
COS distance：
2.2 according to matrix D, the index value arranged belonging to the minimum value in finding out matrix per a line, i.e., in 102 H-matrix
Channel number obtains channel vector V3=(C1,C2…..Cn), which indicates that each channel in matrix H distance matrix C is nearest
Channel；
2.3 by vectorial V3It is arranged according to 101 rate of rise descending of mid band, n fastest-rising channels are made before taking
To recommend.
Embodiment 2
A kind of instantaneous channel Generalization bounds for being chosen method based on intersection, Fig. 2 are described one and are using intersection selection method
IPTV user carries out the example of instantaneous channel recommendation, in this example, chooses method by intersection and carries out recommending to comprise the steps of：
1) access time window delta t, and timing statistics window [t- Δs t, t) in, the viewing number p of each channeli。
2) calculate each channel time window [t- Δs t, t) in, the rate of rise of the online number of each channel
3) to piAnd riIt is normalized,PiFor piValue after normalization,
RiFor to riValue after normalization.Build the current channel state matrix of t momentMatrix it is each
Row indicate time window [t- Δs t, t) in, channel CiNumber, viewing number watches the rate of rise of number online；
4) it is directed to user U, the case history viewing channel state matrix H of the user is built, chooses history sliding window Δ
In historical time [t- Δs T, t], the time watched of each channel is denoted as o by T, here Δ T ＞ ＞ Δ t, counting user Ui；
5) weights that user U in [t- Δs T, t] time window watched each channel are calculated, weights formula is described as： wiIndicate that user U watches channel C in time windowiWeight, τkFor in [t- Δs T, t], user U kth is secondary
Watch channel CiWeight, N be channel CiThe watched channel C of the user U in the section [t- Δs T, t]iNumber.Formula describes
FortkFor user U kth time viewing channels CiAt the time of, at the beginning of ts is historical time window, i.e. t- Δs T,
Δ T is the width of the sliding window of history, to oiAnd wiIt is normalized,OiFor to oiNormalization
Value later,WiFor to wiValue after normalization.Obtain the case history viewing channel state square of user U
Battle array
Fig. 3 describes user U in [t- Δs T, t] interior time window, channel CiThe calculating process of weights, in the time window
In mouthful, user's U viewing channels CiAltogether twice, therefore weights are wi=τ1+τ2,Ts is historical time window
At the beginning of mouthful, i.e. t- Δs T；
6) it to the channel number in Matrix C, is ranked up according to the feature of channel, the present embodiment is with the viewing people after normalizing
Number PiDescending channel is arranged, select x popular channel, obtain current candidate collection V1=(C1,C2……Cx), intermediate frequency
Road number x is less than or equal to the channel number of current channel state matrix, for the matrix H of user U, is carried out according to the feature of channel
Sequence, this example arrange channel according to the descending of the channel time watched after normalization, select user U in historical period
Y channel of interior hot topic, y are less than or equal to the channel number of case history viewing channel state matrix, and the history for constituting user U is waited
Selected works V2=(C1,C2……Cy)。
7) V is selected1∩V2As the recommendation channel to user U, the result of intersection is arranged according to the descending of channel rate of rise
Row select the top n recommendation of the channels in the intersection to the user.
Embodiment 3
A kind of instantaneous channel Generalization bounds for being chosen method based on distance, Fig. 4 are described one and are using apart from selection method
IPTV user carries out the example of instantaneous channel recommendation, by carrying out recommending to comprise the steps of apart from selection method：
Step 1：For each row vector in matrix H at a distance from each row vector, row vector does not include frequency in calculating matrix C
A variety of calculations can be used in the calculating of Taoist monastic name, distance, this example is by taking Euclidean distance as an example, such as row vector vec in Matrix C1
=(P1,R1) with matrix H in row vector vec2=(O1,W1) calculation is as follows：
Here P1Representing matrix C mid bands C1Viewing number, O1It indicates
Matrix H mid band C1Time watched, R1Representing matrix C mid bands C1Rate of rise, W1Representing matrix H mid bands C1Power
Weight.In the same way,Here PiRepresenting matrix C mid bands CiViewing number, OjTable
Show matrix H mid band CiTime watched, RiRepresenting matrix C mid bands CiOnline number rate of rise, WjIn representing matrix H
Channel CjWeight.
In calculating matrix C the distance of a n × m is obtained after the distance of each row vector in each row vector distance matrix H
Matrix D, n are the channel number of Matrix C, and m is the channel number of matrix H, the index value homography C's in matrix D per a line
Channel number, the channel number of the index value homography H of each row in matrix D.
Step 2：The index value arranged belonging to the minimum value in matrix D per a line, i.e. channel number in matrix H are selected, finally
Obtain the Candidate Set S=(CH of a n channel1, CH2... ..., CHn), the nearest frequency of representing matrix H distance matrix C mid bands
Road；
Step 3：Candidate Set S is ranked up according to the feature of the channel in matrix H, this example is with the sight of each channel
See number RiDescending arrangement is carried out, and selects preceding K channel as the recommendation to user U.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, it is other it is any without departing from the spirit and principles of the present invention made by changes, modifications, substitutions, combinations, simplifications,
Equivalent substitute mode is should be, is included within the scope of the present invention.
Claims (6)
1. the recommendation of the channels method merged with group current behavior based on individual history, which is characterized in that comprise the steps of：
The currently viewing channel state matrix of 1.1 building groups；Acknowledging time window delta t and timing statistics window [t- Δs t, t) in,
The viewing number p of each channel under all user's viewingsi, and to piIt is normalized,PiFor pi
Value after normalized, and calculate the growth rate r of each channeli, channel growth rate is：
And to riIt is normalized,According to the currently viewing number of channel and growth rate structure after normalization
Make the currently viewing channel state matrix of groupCiFor channel number, matrix indicates each channel per a line
Channel number, the current instantaneous rate of increase of currently viewing number and the channel；
1.2 structure case histories watch channel state matrix；History sliding window Δ T is chosen, and should for each user statistics
The time watched o of user's each channel in the section time window [t- Δs T, t]i, and to oiIt is normalized,OiFor oiValue after normalization；Calculate power of each channel of the user in time window [t- Δs T, t- Δ t]
Value wi, and to wiIt is normalized,WiFor wiValue after normalization, according to the viewing people of each channel
The weights of number and each channel build the case history viewing channel state matrix of each userSquare
Battle array indicates the time watched of each channel of the user and the weights of each channel per a line；
The 1.1 current channel state matrix C and 1.2 personalized historic state matrix Hs are sent into recommendation computing module by 1.3
And recommend the selection of channel.
2. recommendation of the channels method according to claim 1, which is characterized in that in step 1.2, each channel is in time window
Weight w in [t- Δs T, t- Δ t]iComputational methods are taken one of following：
2.1 methods one：The weights of each channel are interior in time window [t- Δs T, t] by the user, what which was watched every time
The sum of weight forms, and formula is described as：wiFor channel CiWeights, Nci be channel CiIn the area [t- Δs T, t]
The interior user watches channel CiNumber, τkFor user's kth time viewing channel CiWeight, τkFormula be described as：At the beginning of ts is time window, i.e. t- Δs T, tkIt is that the user is secondary in time window [t- Δs T, t] interior kth
At the time of viewing residing for channel i, Δ T is sliding window size；
2.2 methods two：The weights of each channel are interior in time window [t- Δs T, t] by the user, what which was watched every time
The sum of weight forms, and formula is described as：wiFor channel CiWeights, Nci be channel CiIn the area [t- Δs T, t]
The interior user watches channel CiNumber, τkFor user's kth time viewing channel CiWeight, τkFormula be described as：T is current time, tkAt the time of for the user residing for time window [t- Δs T, t] interior kth time viewing channel i,
Δ T is sliding window size.
3. recommendation of the channels method according to claim 1, which is characterized in that in step 1.3, according to currently viewing channel shape
State Matrix C and case history viewing channel state matrix H carry out that channel is recommended to choose, and method is chosen using intersection, use following rule
One of then：
Method rule one is chosen in 3.1 intersections：According to currently viewing channel state matrix, channel number is watched to the drop of number according to channel
Sequence arranges, and selects x most channel of viewing number, constitutes current candidate collection V1=(C1,C2……Cx)；According to case history
Channel state matrix is watched, channel number is arranged according to the descending of the time watched of channel, it is most to select user U time watcheds
Y channel constitutes history Candidate Set V2=(C1,C2……Cy)；Take V1∩V2Recommendation channel as the user U；
Method rule two is chosen in 3.2 intersections：According to currently viewing channel state matrix, by channel number according to the drop of channel rate of rise
Sequence arranges, and selects x most fast channel of rate of rise in the period, constitutes current candidate collection V1=(C1,C2……Cx)；According to
Case history watches channel state matrix, and channel number is arranged according to the descending of the time watched of channel, selects user's U viewings time
Y most channel of number, constitutes history Candidate Set V2=(C1,C2……Cy)；Take V1∩V2Recommendation channel as the user；
Method rule three is chosen in 3.3 intersections：According to currently viewing channel state matrix, by channel number according to the drop of channel rate of rise
Sequence arranges, and selects x most fast channel of rate of rise in the period, constitutes current candidate collection V1=(C1,C2……Cx)；According to
Case history watches channel state matrix, and channel number is arranged according to the weights descending of channel, selects y larger frequency of weights
Road constitutes history Candidate Set V2=(C1,C2……Cy)；Take V1∩V2Recommendation channel as the user；
Method rule four is chosen in 3.4 intersections：According to currently viewing channel state matrix, channel number is watched to the drop of number according to channel
Sequence arranges, and selects x most channel of viewing number, constitutes current candidate collection V1=(C1,C2……Cx)；According to case history
Channel state matrix is watched, channel number is subjected to descending arrangement according to the weights of channel, is selected maximum relative to user's weight
Y channel, constitute history Candidate Set V2=(C1,C2……Cy)；Take V1∩V2Recommendation channel as the user.
4. recommendation of the channels method according to claim 1, which is characterized in that in step 1.3, according to currently viewing channel shape
State matrix and case history viewing channel state matrix carry out that channel is recommended to choose, and method is chosen using Euclidean distance：
4.1 use Euclidean distance computational methods, calculate each row vector in currently viewing channel state Matrix C and are seen with case history
It sees that the distance of each row vector in channel state matrix H, the channel number of row vector are not put into calculation formula, builds a n × m
The Distance matrix D of dimension, n is the channel number of Matrix C, i.e. the row vector number of C in matrix D；M is the channel number in matrix H,
That is the row vector number of H；Index value in matrix D per a line corresponds to each channel number of currently viewing channel state Matrix C, i.e.,
First row in C；The index value of each row in matrix D corresponds to each channel number of the individual subscriber history matrix H, i.e. in H
One row；
4.2 find out in matrix D per the index value arranged belonging to a line minimum value, that is, are exactly case history viewing channel state matrix H
In channel number, total n candidate channel, n is the number of current channel state matrix C channels, the i.e. line number of C；
4.3 arrange the n channel obtained in 4.2 according to the characteristic series in C or H, according to the history time watched of channel
Or the weights of each channel are ranked up, and choose preceding K channel and recommend as relative users.
5. recommendation of the channels method according to claim 1, which is characterized in that in step 1.3, according to current channel state square
Battle array and personalized historic state matrix carry out that channel is recommended to choose, and method is chosen using manhatton distance：
5.1 use manhatton distance computational methods, calculate each row vector and case history in currently viewing channel state Matrix C
The distance of each row vector in channel state matrix H is watched, the channel number of row vector is not put into calculation formula, builds a n
The Distance matrix D of × m dimensions, n is the channel number of Matrix C, i.e. the row vector number of C in matrix D；M is the channel in matrix H
Number, i.e. the row vector number of H；Index value in matrix D per a line corresponds to each channel of currently viewing channel state Matrix C
Number, i.e. first row in C；The index value of each row in matrix D corresponds to each channel number of the individual subscriber history matrix H, i.e. H
Middle first row；
5.2 find out in matrix D per the index value arranged belonging to a line minimum value, that is, are exactly case history viewing channel state matrix H
In channel number, total n candidate channel, n is the number of current channel state matrix C channels, the i.e. line number of C；
5.3 arrange the n channel obtained in 5.2 according to the characteristic series in C or H, according to the history time watched of channel
Or the weights of each channel are ranked up, and choose preceding K channel and recommend as relative users.
6. recommendation of the channels method according to claim 1, which is characterized in that in step 1.3, according to current channel state square
Battle array and personalized historic state matrix carry out that channel is recommended to choose, and method is chosen using COS distance：
6.1 use manhatton distance computational methods, calculate each row vector and case history in currently viewing channel state Matrix C
The distance of each row vector in channel state matrix H is watched, the channel number of row vector is not put into calculation formula, builds a n
The Distance matrix D of × m dimensions, n is the channel number of Matrix C, i.e. the row vector number of C in matrix D；M is the channel in matrix H
Number, i.e. the row vector number of H；Index value in matrix D per a line corresponds to each channel of currently viewing channel state Matrix C
Number, i.e. first row in C；The index value of each row in matrix D corresponds to each channel number of the individual subscriber history matrix H, i.e. H
Middle first row；
6.2 find out in matrix D per the index value arranged belonging to a line minimum value, that is, are exactly case history viewing channel state matrix H
In channel number, total n candidate channel, n is the number of current channel state matrix C channels, the i.e. line number of C；
6.3 arrange the n channel obtained in 6.2 according to the characteristic series in C or H, according to the history time watched of channel
Or the weights of each channel are ranked up, and choose preceding K channel and recommend as relative users.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810110304.1A CN108419134B (en) | 2018-02-05 | 2018-02-05 | Channel recommendation method based on fusion of individual history and group current behaviors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810110304.1A CN108419134B (en) | 2018-02-05 | 2018-02-05 | Channel recommendation method based on fusion of individual history and group current behaviors |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108419134A true CN108419134A (en) | 2018-08-17 |
| CN108419134B CN108419134B (en) | 2020-02-18 |
Family
ID=63127699
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810110304.1A Active CN108419134B (en) | 2018-02-05 | 2018-02-05 | Channel recommendation method based on fusion of individual history and group current behaviors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108419134B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109600641A (en) * | 2018-12-18 | 2019-04-09 | 北京字节跳动网络技术有限公司 | Method and apparatus for handling information |
| CN109889864A (en) * | 2019-01-03 | 2019-06-14 | 百度在线网络技术（北京）有限公司 | Video pushing method, device, equipment and storage medium |
| CN110087119A (en) * | 2019-04-26 | 2019-08-02 | 广州酷狗计算机科技有限公司 | Homepage display methods, device and computer readable storage medium is broadcast live |
| CN110351580A (en) * | 2019-07-12 | 2019-10-18 | 四川长虹电器股份有限公司 | TV programme special recommendation method and system based on Non-negative Matrix Factorization |
| CN111461516A (en) * | 2020-03-26 | 2020-07-28 | 深圳市百纳九洲科技有限公司 | Comprehensive grid management system platform for smart city |
| CN116320511A (en) * | 2023-02-03 | 2023-06-23 | 华南理工大学 | Cross-domain fusion recommendation method based on graph convolution network |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1756336A (en) * | 2004-09-29 | 2006-04-05 | 松下电器产业株式会社 | Television channel commending system and commending method |
| WO2012077000A1 (en) * | 2010-12-10 | 2012-06-14 | Nokia Corporation | Method and apparatus for registering a content provider channel for recommendation of content segments |
| CN102769781A (en) * | 2012-07-17 | 2012-11-07 | 青岛海信传媒网络技术有限公司 | Method and device for recommending television program |
| CN102999634A (en) * | 2012-12-18 | 2013-03-27 | 百度在线网络技术（北京）有限公司 | User navigation recommending method and system based on browser data as well as cloud server |
| CN103491441A (en) * | 2013-09-09 | 2014-01-01 | 东软集团股份有限公司 | Recommendation method and system of live television programs |
| CN103546773A (en) * | 2013-08-15 | 2014-01-29 | Tcl集团股份有限公司 | Television program recommendation method and system |
| CN103888852A (en) * | 2014-03-24 | 2014-06-25 | 清华大学 | Video recommendation method and device for social television |
| CN104123284A (en) * | 2013-04-24 | 2014-10-29 | 华为技术有限公司 | Recommendation method and server |
| US20140344840A1 (en) * | 2012-02-09 | 2014-11-20 | Panasonic Corporation | Information recommendation device, information recommendation system, and information recommendation method |
| CN104486339A (en) * | 2014-12-16 | 2015-04-01 | 广州华多网络科技有限公司 | Method and device for displaying recommendation data in social application |
| KR101552952B1 (en) * | 2014-12-02 | 2015-09-14 | 성균관대학교산학협력단 | Tv channel recomendation method, apparatus and system by using tv watching time record |
| CN105915949A (en) * | 2015-12-23 | 2016-08-31 | 乐视网信息技术（北京）股份有限公司 | Video content recommending method, device and system |
| CN106791901A (en) * | 2016-12-15 | 2017-05-31 | Tcl集团股份有限公司 | A kind of direct broadcast band sort method and its device |
| KR20170078464A (en) * | 2015-12-29 | 2017-07-07 | 성균관대학교산학협력단 | Personalized tv channel recommendation considering time dependent propensity of viewer using constrained optimization technique |
-
2018
- 2018-02-05 CN CN201810110304.1A patent/CN108419134B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1756336A (en) * | 2004-09-29 | 2006-04-05 | 松下电器产业株式会社 | Television channel commending system and commending method |
| WO2012077000A1 (en) * | 2010-12-10 | 2012-06-14 | Nokia Corporation | Method and apparatus for registering a content provider channel for recommendation of content segments |
| US20140344840A1 (en) * | 2012-02-09 | 2014-11-20 | Panasonic Corporation | Information recommendation device, information recommendation system, and information recommendation method |
| CN102769781A (en) * | 2012-07-17 | 2012-11-07 | 青岛海信传媒网络技术有限公司 | Method and device for recommending television program |
| CN102999634A (en) * | 2012-12-18 | 2013-03-27 | 百度在线网络技术（北京）有限公司 | User navigation recommending method and system based on browser data as well as cloud server |
| CN104123284A (en) * | 2013-04-24 | 2014-10-29 | 华为技术有限公司 | Recommendation method and server |
| CN103546773A (en) * | 2013-08-15 | 2014-01-29 | Tcl集团股份有限公司 | Television program recommendation method and system |
| CN103491441A (en) * | 2013-09-09 | 2014-01-01 | 东软集团股份有限公司 | Recommendation method and system of live television programs |
| CN103888852A (en) * | 2014-03-24 | 2014-06-25 | 清华大学 | Video recommendation method and device for social television |
| KR101552952B1 (en) * | 2014-12-02 | 2015-09-14 | 성균관대학교산학협력단 | Tv channel recomendation method, apparatus and system by using tv watching time record |
| CN104486339A (en) * | 2014-12-16 | 2015-04-01 | 广州华多网络科技有限公司 | Method and device for displaying recommendation data in social application |
| CN105915949A (en) * | 2015-12-23 | 2016-08-31 | 乐视网信息技术（北京）股份有限公司 | Video content recommending method, device and system |
| KR20170078464A (en) * | 2015-12-29 | 2017-07-07 | 성균관대학교산학협력단 | Personalized tv channel recommendation considering time dependent propensity of viewer using constrained optimization technique |
| CN106791901A (en) * | 2016-12-15 | 2017-05-31 | Tcl集团股份有限公司 | A kind of direct broadcast band sort method and its device |
Non-Patent Citations (4)
| Title |
|---|
| HYOKYUNG BAHN ETAL.: "An Intelligent Channel Navigation Scheme for DTV Channel Selectors", 《IEEE TRANSACTIONS ON CONSUMER ELECTRONICS》 * |
| JINOH OH ETAL.: "When to recommend: A new issue on TV show recommendation", 《INFORMATION SCIENCES》 * |
| LI NING ET AL.: "Realtime Channel Recommendation: Switch Smartly While Watching TV", 《INTERNATIONAL WORKSHOP ON FRONTIERS IN ALGORITHMICS》 * |
| 姜源璐: "基于用户行为分析的网络电视推荐系统", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109600641A (en) * | 2018-12-18 | 2019-04-09 | 北京字节跳动网络技术有限公司 | Method and apparatus for handling information |
| CN109600641B (en) * | 2018-12-18 | 2021-08-10 | 北京字节跳动网络技术有限公司 | Method and apparatus for processing information |
| CN109889864A (en) * | 2019-01-03 | 2019-06-14 | 百度在线网络技术（北京）有限公司 | Video pushing method, device, equipment and storage medium |
| CN109889864B (en) * | 2019-01-03 | 2021-08-31 | 百度在线网络技术（北京）有限公司 | Video pushing method, device, equipment and storage medium |
| US11272253B2 (en) | 2019-01-03 | 2022-03-08 | Baidu Online Network Technology (Beijing) Co., Ltd. | Video pushing method, apparatus, device and storage medium |
| CN110087119A (en) * | 2019-04-26 | 2019-08-02 | 广州酷狗计算机科技有限公司 | Homepage display methods, device and computer readable storage medium is broadcast live |
| CN110087119B (en) * | 2019-04-26 | 2022-02-22 | 广州酷狗计算机科技有限公司 | Live broadcast home page display method and device and computer readable storage medium |
| CN110351580A (en) * | 2019-07-12 | 2019-10-18 | 四川长虹电器股份有限公司 | TV programme special recommendation method and system based on Non-negative Matrix Factorization |
| CN110351580B (en) * | 2019-07-12 | 2021-07-13 | 四川长虹电器股份有限公司 | Television program topic recommendation method and system based on non-negative matrix factorization |
| CN111461516A (en) * | 2020-03-26 | 2020-07-28 | 深圳市百纳九洲科技有限公司 | Comprehensive grid management system platform for smart city |
| CN111461516B (en) * | 2020-03-26 | 2021-04-13 | 深圳市百纳九洲科技有限公司 | Comprehensive grid management system platform for smart city |
| CN116320511A (en) * | 2023-02-03 | 2023-06-23 | 华南理工大学 | Cross-domain fusion recommendation method based on graph convolution network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108419134B (en) | 2020-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108419134A (en) | The recommendation of the channels method merged with group current behavior based on individual history | |
| US11938393B2 (en) | Devices, systems, and their methods of use for desktop evaluation | |
| CN105404700B (en) | A kind of video column recommendation system and recommended method based on collaborative filtering | |
| CN101489107B (en) | Collaborative filtering recommendation method based on population attribute keyword vector | |
| CN103546773B (en) | The recommendation method and its system of TV programme | |
| CN103649981B (en) | Method and apparatus for delivering targeted content | |
| CN103888852B (en) | For the video recommendation method and device of social television | |
| US8799936B2 (en) | Cooperative filtering algorithm-based personal preference program recommendation system for IPTV | |
| CN103491441B (en) | Recommendation method and system of live television programs | |
| CN102917269B (en) | A kind of television program recommendation system and method | |
| CN106791966B (en) | A kind of direct broadcasting room recommended method and system based on modified correlation rule | |
| CN103260061B (en) | A kind of IPTV program commending method of context-aware | |
| CN103546778A (en) | Television program recommendation method and system, and implementation method of system | |
| CN106844446A (en) | Video methods of marking, device and video system based on user's viewing behavior | |
| CN102630052A (en) | Real time streaming-oriented television program recommendation system | |
| CN105574198B (en) | column recommendation method and device | |
| CN106168980A (en) | Multimedia resource recommends sort method and device | |
| Kui et al. | TVseer: A visual analytics system for television ratings | |
| Aharon et al. | Watch-it-next: a contextual TV recommendation system | |
| CN107368584A (en) | A kind of individualized video recommends method and system | |
| CN109391829A (en) | Video gets position analysis system, analysis method and storage media ready | |
| CN109640128A (en) | A kind of TV user watching behavior feature extracting method and system | |
| CN108647724A (en) | A kind of user's recommendation method and device based on simulated annealing | |
| CN109819326A (en) | A kind of TV station's competition situation appraisal procedure and device | |
| Zhang et al. | GroupBox: A generative model for group recommendation |
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 |