CN101431694B - Digital television program recommending method and system based on Bayesian algorithm - Google Patents

Abstract

The invention relates to television program recommendation. The invention provides a method and a system for recommending digital television program based on bayesian algorithm aiming that existing television program recommendation method only statistics user's viewing behavior but ignores playing behavior of television station. The method includes: collecting user's viewing recorder; analysing user's viewing recorder and obtaining user's viewing action parameter. The method also includes: obtaining television station playing action parameter according with viewing behavior parameter; predicting user's viewing probability to waiting-played television program according with viewing action parameter and playing action, and recommending television program to user thereof. The invention also provides a recommendation system corresponding to the method. Because combining the user's viewing action parameter and television playing action parameter when predicting user viewing special program, a recommendation result obtained by the television program recommendation is more accurately and reflects real user's viewing taste.

Description

A kind of digital television program recommending method and system based on bayesian algorithm

Technical field

The present invention relates to the television program recommendations technology, more particularly, relate to a kind of digital television program recommending method and system based on bayesian algorithm.

Background technology

The world today is among the wave of digitalization, and radio and television also are like this.Before and after American-European main developed country all fixed on time of radio and television total digitalization 2010, China also planned in round Realization digitlization in 2015.In the end of the year 2006, China Digital TV user has reached 1,200 ten thousand families, and according to the prediction of CCID Consulting, by 2007, global digital cable customers will reach 6.3 hundred million families.

One of change that television digitization brings is exactly the greatly abundant of TV programme.According to the video coding mode of current MPEG2, cable television system can be transmitted the digital television program of 500 cover single-definitions.If use and H.264 wait advanced coded format, the digital television program of transmission will reach 1500 covers, under this trend, TV user is being faced the colorful TV programme that becomes increasingly abundant very happily on the one hand, and they for how to select their interested content in so numerous TV programme are worrying on the other hand, and TV user will face similar " information overload " problem with the Internet user.Traditional printing television program listing and channel surfing mode can not be offered help to them this moment.Because for 500 channels, if 1 day program inventory of 10 channels is printed on one page paper, the television program listing in so whole 500 one weeks of channel will be one 350 pages a thick book, and in the face of such book, the user is difficult to read patiently and search his needed program; In addition, if 10 seconds of each channel browsing, the user adopts content that the channel surfing method browsed whole 500 channels with 82.5 minutes consuming time, and such time user is difficult to accept.Present electronic program guides adopts the mode display program inventory based on channel or classification (for example physical culture, finance and economics, film etc.), though this kind mode can partly address the above problem, but still does not thoroughly deal with problems.

To solve the problem of TV information " overload " completely, just need research user's viewing behavior, judge user's rating hobby and other hobbies, according to user's interest, hobby and rule automatically to user's recommending television and service.

The prior art automatic recommendations of adopting following method to realize TV programme more.At first, from user watched record, extract the viewing behavior parameter, be used to characterize user's rating hobby.Subsequently, be TV programme definition programs feature, be used to characterize each attribute of program.For realizing the automatic coupling of programs feature and user watched hobby, need to use identical component to describe viewing behavior parameter and programs feature, for example, these components can be to play period, place channel etc.After the viewing behavior component and programs feature that obtain using same components to describe, just can be by relatively being about to play the programs feature of TV programme and the similarity between the user watched behavior parameter, TV programme and user preferences are mated, then recommend its TV programme of liking to the user, i.e. the higher program of similarity between programs feature and user's the viewing behavior parameter.

Yet, be not difficult to find out that by mentioned earlier existing television program recommendations scheme is only added up user's viewing behavior parameter, and ignored the broadcast behavior parameter of TV station, be i.e. the shared ratio of every class program in the various types of programs of being play.So, mate the recommendation results that obtains and obviously can't accurately reflect the real rating hobby of user.For example rating record explicit user has been watched 130 TV programme altogether in nearest one month, wherein 50 of news category programs, TV play class program 40, sport category program 20, amusement and recreation class program 20.And TV station has play 10000 TV programme altogether in the same period, wherein 5000 of news category programs, 3000 of TV play class programs, 300 of sport category programs, 1700 of amusement and recreation class programs.If according to existing program commending method, the news category program is liked by this user most beyond doubt, secondly is TV play class program, is sport category program and amusement and recreation class program at last, and the favorable rating of the two equates.And if according to the broadcast/audience ratings of various types of programs, then the sport category program is most popular undoubtedly, secondly be TV play class program, be amusement class program once more, be only the news category program at last.Be not difficult to find out that thus the broadcast behavior meeting of TV station produces bigger influence to recommendation results, and prior art has been ignored the broadcast behavior, so its recommendation results is accurate inadequately.

Therefore, need a kind of television program recommendations scheme, can overcome the defective that prior art exists.

Summary of the invention

The technical problem to be solved in the present invention is, the broadcast behavior of having ignored TV station owing to the viewing behavior of only adding up the user at existing TV programme suggesting method causes the accurate inadequately defective of recommendation results, and a kind of digital television program recommending method and system based on bayesian algorithm is provided.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of digital television program recommending method based on bayesian algorithm comprises:

S1, gather user watched record;

S2, analysis user rating record obtain the viewing behavior parameter, and this viewing behavior parameter comprises:

The probability that the user televiews;

The probability that the user does not teleview;

The probability of every class program in the program that the user watches;

The probability of every class program in the program that the user does not watch;

Described method also comprises:

S3, according to the viewing behavior parameter, obtain playing the behavior parameter, this broadcast behavior parameter comprises the broadcast probability of every class program;

S4, based on bayesian algorithm, by the viewing behavior parameter with play behavior parameter predictive user and treat the probability of watching of broadcasting TV programme, in view of the above to user's recommending television.

In the digital television program recommending method based on bayesian algorithm of the present invention, establish:

Each program X is expressed as { x ₁f ₁..., x _if _i..., x _nf _n, f wherein _iRepresent each attribute of program X, x _iRepresent the weight of each attribute of program X, wherein, 1≤i≤n, n represent the number of attributes of program, and n 〉=1;

C+ and C-represent the behavior of watching respectively and do not watch behavior;

The TV programme sum that TV station in a period of time is broadcasted is expressed as k (C+)+k (C-), and wherein, k (C+) and k (C-) distinguish the quantity of representative of consumer TV reception and the quantity of TV reception not;

K (f _i| C+) and k (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the quantity of every class program in the program that the quantity of every class program and user do not watch in the program that the user watches;

P (C+) and p (C-) be the probability that do not teleview of the probability that televiews of representative of consumer and user respectively;

P (f _i| C+) and p (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the probability of every class program in the TV programme that the probability of every class program and user do not watch in the program that the user watches;

P (X|C+) and p (X|C-) represent respectively at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the probability of every class program in the program that the probability of every class program and user do not watch in the program that the user watches;

Then in described S2, calculate described viewing behavior parameter according to following steps:

K (C+), k (C-), k (f added up respectively in S21, the TV programme quantity and the user watched record that broadcast according to TV station _i| C+) and k (f _i| C-);

S22, calculating:

$p(C+)=\frac{k(C+)}{k(C+)+k(C-)};$

$p(C-)=\frac{k(C-)}{k(C+)+k(C-)};$

$p\left(f_i\right|C+)=\frac{k\left(f_i\right|C+)}{k(C+)};$

$p\left(f_i\right|C-)=\frac{k\left(f_i\right|C-)}{k(C-)};$

$p\left(X\right|C+)=\underset{i=1}{\overset{n}{\mathrm{\Π}}}p{\left(f_i\right|C+)}^{x_i}{(1-p\left(f_i\right|C+))}^{1-x_i};$

$p\left(X\right|C-)=\underset{i=1}{\overset{n}{\mathrm{\Π}}}p{\left(f_i\right|C-)}^{x_i}{(1-p\left(f_i\right|C-))}^{1-x_i}.$

In the digital television program recommending method based on bayesian algorithm of the present invention, establish:

P (X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the broadcast probability of every class program;

Then in described S3, calculate described broadcast behavior parameter according to following steps:

S31, calculating:

p(X)＝p(X|C+)p(C+)+p(X|C-)p(C-)。

In the digital television program recommending method based on bayesian algorithm of the present invention, establish:

P (C+|X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, and the user is to the probability of watching of every class program;

Then in described S4, predict the probability of watching of every class program according to following steps:

S41, calculate based on Bayesian formula:

$p(C+|X)=\frac{p\left(X\right|C+)p(C+)}{p\left(X\right)}.$

In the digital television program recommending method based on bayesian algorithm of the present invention, at least one under each program comprises among the Column Properties:

Play the period;

The place channel;

Affiliated big class;

Affiliated group.

The present invention also provides a kind of digital television program recommendation system based on bayesian algorithm, comprising:

Collecting unit, the rating record and the storage that are used to gather digital TV terminal;

Viewing behavior parameter extraction module with the collecting unit communication linkage, is used for analysis user rating record, output viewing behavior parameter, and this viewing behavior parameter comprises:

The probability that the user televiews;

The probability that the user does not teleview;

The probability of every class program in the program that the user watches;

The probability of every class program in the program that the user does not watch;

Also comprise:

Broadcast behavior parameter extraction module communicates to connect with viewing behavior parameter extraction module, is used to analyze the viewing behavior parameter, output broadcast behavior parameter, and this broadcast behavior parameter comprises the broadcast probability of every class program;

Watch the probability extraction module, communicate to connect, be used to receive viewing behavior parameter and the behavior of broadcast parameter, treat the probability of watching of broadcasting TV programme based on bayesian algorithm output user with viewing behavior parameter extraction module and the behavior of broadcast parameter extraction module;

Recommending module communicates to connect with watching the probability extraction module, is used to receive the probability of watching of all kinds of TV programme, in view of the above to user's recommending television.

In the digital television program recommendation system based on bayesian algorithm of the present invention, establish:

Each program X is expressed as { x ₁f ₁..., x _if _i..., x _nf _n, f wherein _iRepresent each attribute of program X, x _iRepresent the weight of each attribute of program X, wherein, 1≤i≤n, n represent the number of attributes of program, and n 〉=1;

C+ and C-represent the behavior of watching respectively and do not watch behavior;

The TV programme sum that TV station in a period of time is broadcasted is expressed as k (C+)+k (C-), and wherein, k (C+) and k (C-) distinguish the quantity of representative of consumer TV reception and the quantity of TV reception not;

K (f _i| C+) and k (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the quantity of every class program in the program that the quantity of every class program and user do not watch in the program that the user watches;

P (C+) and p (C-) be the probability that do not teleview of the probability that televiews of representative of consumer and user respectively;

P (f _i| C+) and p (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the probability of every class program in the TV programme that the probability of every class program and user do not watch in the program that the user watches;

P (X|C+) and p (X|C-) represent respectively at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the probability of every class program in the program that the probability of every class program and user do not watch in the program that the user watches;

P (X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the broadcast probability of every class program;

P (C+|X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, and the user is to the probability of watching of every class program;

Then described viewing behavior parameter extraction module is used for calculating described viewing behavior parameter according to following steps:

K (C+), k (C-), k (f added up respectively in the quantity and the user watched record of S1, the TV programme broadcasted according to TV station _i| C+) and k (f _i| C-);

S2, calculating:

$p(C+)=\frac{k(C+)}{k(C+)+k(C-)};$

$p(C-)=\frac{k(C-)}{k(C+)+k(C-)};$

$p\left(f_i\right|C+)=\frac{k\left(f_i\right|C+)}{k(C+)};$

$p\left(f_i\right|C-)=\frac{k\left(f_i\right|C-)}{k(C-)};$

$p\left(X\right|C+)=\underset{i=1}{\overset{n}{\mathrm{\Π}}}p{\left(f_i\right|C+)}^{x_i}{(1-p\left(f_i\right|C+))}^{1-x_i};$

$p\left(X\right|C-)=\underset{i=1}{\overset{n}{\mathrm{\Π}}}p{\left(f_i\right|C-)}^{x_i}{(1-p\left(f_i\right|C-))}^{1-x_i}.$

Described broadcast behavior parameter extraction module is used for calculating described broadcast behavior parameter according to following steps:

S3, calculating:

p(X)＝p(X|C+)p(C+)+p(X|C-)p(C-)；

The described probability extraction module of watching is used for predicting according to following steps the probability of watching of every class program:

S4, calculate based on Bayesian formula:

$p(C+|X)=\frac{p\left(X\right|C+)p(C+)}{p\left(X\right)}.$

Implement technical scheme of the present invention, has following beneficial effect, owing to when calculating the broadcast probability, be used in combination user's the viewing behavior parameter and the broadcast behavior parameter of TV station, therefore adopt the resulting recommendation results of television program recommendations scheme provided by the invention more accurate, more can reflect the real rating hobby of user.

Description of drawings

The invention will be further described below in conjunction with drawings and Examples, in the accompanying drawing:

Fig. 1 is the structural representation according to a preferred embodiment of the present invention digital TV network;

Fig. 2 is the flow chart according to the digital television program recommending method of a preferred embodiment of the present invention.

Embodiment

In order to make purpose of the present invention, technical scheme and advantage clearer,, the present invention is further elaborated below in conjunction with drawings and Examples.Should be appreciated that specific embodiment described herein only in order to explanation the present invention, and be not used in qualification the present invention.

Fig. 1 is the structural representation according to a preferred embodiment of the present invention digital TV network 100.As shown in Figure 1, digital TV network 100 comprises digital television program recommending system 102, a plurality of digital TV terminal 104,106 and 108, and broadband metropolitan area network 110, wherein, a plurality of digital TV terminals 104,106 and 108 communicate to connect with digital television program recommending system 102 by broadband metropolitan area network 110.

Digital TV terminal 104 and 106 is connected to broadband metropolitan area network 110 by bi-directional set-top box.Bi-directional set-top box both can be used for by broadband metropolitan area network 110 received television programs, and the user watched record that statistics can be obtained sends to digital television program recommending system 102 by broadband metropolitan area network 110 again.Be different from digital TV terminal 104 and 106, digital TV terminal 108 need not can be connected to broadband metropolitan area network 110 by set-top box, and received television program, and sends user watched record to digital television program recommending system 102.

Digital television program recommending system 102 comprises that acquisition server (collecting unit) 112, viewing behavior parameter extraction server (viewing behavior parameter extraction unit) 114, the broadcast behavior parameter that sequenced communications connects extracts server (playing behavior parameter extraction unit) 116, the rating probability extracts server (rating probability extraction unit) 118 and recommendation server (recommendation unit) 120, and rating probability extraction server 118 communicates to connect viewing behavior parameter extraction server 114.

Acquisition server 112 communicates to connect broadband metropolitan area network 110, is used to receive the user watched record that a plurality of digital TV terminals 104,106 and 108 are sent.Record the relevant recorded information of TV programme of watching in the recent period in the user watched record, comprise title, broadcasting channel, reproduction time, affiliated big class, affiliated group of the TV programme of watching in the recent period or the like such as but not limited to the user with the user.

The viewing behavior parameter extracts server 114 and is used to read the user watched record that acquisition server 112 is gathered, and therefrom extracts user's viewing behavior parameter.Record user's viewing behavior in the viewing behavior parameter, comprise the broadcast probability of every class program in the broadcast probability of every class program in probability that the user televiews, probability that the user does not teleview, the program that the user watches and the program that the user does not watch.The viewing behavior parameter extracts server 114 and extract the viewing behavior parameter from user watched record, and it is mail to broadcast behavior parameter extraction server 116 and rating probability extraction server 118.The extracting method of relevant viewing behavior parameter will be described in detail hereinafter.

Broadcast behavior parameter extracts server 116 and is used to receive the viewing behavior parameter that viewing behavior parameter extraction server 114 sends, and therefrom extracts broadcast behavior parameter.The record television platform is to the broadcast probability of every class program in the broadcast behavior parameter.The broadcast behavior parameter that extracts will mail to the rating probability and extract server 118.The extracting method of relevant broadcast behavior parameter will be described in detail hereinafter.

The rating probability extracts server 118 and is used to receive the viewing behavior parameter that viewing behavior parameter extraction server 114 sends, extract the broadcast behavior parameter that server 116 is sent with the behavior of broadcast parameter, and calculate the probability of watching of all kinds of TV programme based on bayesian algorithm, mail to recommendation server 120.The extracting method of relevant rating probability will be described in detail hereinafter.

Recommendation server 120 is used to receive the probability of watching that the rating probability extracts all kinds of TV programme that server 118 sends, search the probability of watching of its correspondence according to the type of each program to be broadcast, and all programs to be broadcast are sorted according to watching that probability is descending, recommend the forward program to be broadcast that sorts to the user.

The present invention also provides a kind of digital television program recommending method, below just be described in detail in conjunction with Fig. 2.

Fig. 2 is the flow chart according to the digital television program recommending method 200 of a preferred embodiment of the present invention.As shown in Figure 2, method 200 starts from step 202.

Subsequently, at next step 204, gather user watched record.

Subsequently, at next step 206, obtain the viewing behavior parameter according to user watched record.As indicated above, the viewing behavior parameter comprises the broadcast probability of every class program in the broadcast probability of every class program in probability that the user televiews, probability that the user does not teleview, the program that the user watches and the program that the user does not watch, and its computational methods are as follows:

If:

Each program X is expressed as { x ₁f ₁..., x _if _i..., x _nf _n, wherein, f _i(1≤i≤n) represents each attribute of program X, for example broadcast slot of program X, place channel, affiliated big class and affiliated group etc.; x _i(1≤i≤n) represents the weight of each attribute of program X, and n represents the number of attributes of program, and n 〉=1.

The behavior of C representative of consumer, wherein C+ and C-represent the behavior of watching respectively and do not watch behavior;

The TV programme sum that TV station in a period of time is broadcasted is expressed as k (C+)+k (C-), and wherein, k (C+) and k (C-) represent the quantity of described user's TV reception and the quantity of TV reception not respectively.K (C+) and k (C-) can directly add up from user watched record and draw.

K (f _i| C+) and k (f _i| C-) represent respectively at foundation attribute f _iWhen (for example place channel) classifies (for example that the place channel is identical program is included into a class) to all TV programme, the quantity of every class program in the program that the quantity of every class program and user do not watch in the program that the user watches.K (f _i| C+) and k (f _i| C-) can draw from the direct statistics of user watched record.

P (C+) and p (C-) be the probability that do not teleview of the probability that televiews of representative of consumer and user respectively, wherein,

$p(C+)=\frac{k(C+)}{k(C+)+k(C-)}$ (formula 1)

$p(C-)=\frac{k(C-)}{k(C+)+k(C-)}$ (formula 2)

P (f _i| C+) and p (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the broadcast probability of every class program in the TV programme that the probability of every class program and user do not watch in the program that the user watches, wherein,

$p\left(f_i\right|C+)=\frac{k\left(f_i\right|C+)}{k(C+)}$ (formula 3)

$p\left(f_i\right|C-)=\frac{k\left(f_i\right|C-)}{k(C-)}$ (formula 4)

P (X|C+) and p (X|C-) represent respectively at foundation attribute f ₁～f _nWhen (broadcast slot, place channel, affiliated big class and affiliated group) classifies (for example broadcast slot, place channel, affiliated big class and all identical program of affiliated group being included into a class) to all TV programme, the probability of every class program in the program that the probability of every class program and user do not watch in the program that the user watches, wherein

$p\left(X\right|C+)=\underset{i=1}{\overset{n}{\mathrm{\Π}}}p{\left(f_i\right|C+)}^{x_i}{(1-p\left(f_i\right|C+))}^{1-x_i}$ (formula 5)

$p\left(X\right|C-)=\underset{i=1}{\overset{n}{\mathrm{\Π}}}p{\left(f_i\right|C-)}^{x_i}{(1-p\left(f_i\right|C-))}^{1-x_i}$ (formula 6)

Thus, we just can be according to add up k (C+), k (C-), the k (f that obtains from user watched record _i| C+) and k (f _i| C-), use formula 1～6 to calculate the viewing behavior parameter, i.e. the broadcast probability of every class program in the broadcast probability of every class program and the program that the user does not watch in the probability that televiews of user, probability that the user does not teleview, the program that the user watches.

Subsequently, at next step 208, play the behavior parameter according to the viewing behavior Parameters Calculation that calculates.As indicated above, play the broadcast probability that the behavior parameter comprises every class program, its computational methods are as follows:

If:

P (X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the broadcast probability of every class program, wherein,

P (X)=p (X|C+) p (C+)+p (X|C-) p (C-) (formula 7)

Thus, we just can use the viewing behavior parameter of trying to achieve in the step 206 to calculate the broadcast probability of every class program according to formula 7.

At next step 210, according to the viewing behavior parameter that calculates and the behavior of broadcast parameter, calculate the probability of watching of every class program based on Bayesian formula, its computational methods are as follows:

If:

P (C+|X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, and the user is to the probability of watching of every class program, wherein,

$p(C+|X)=\frac{p\left(X\right|C+)p(C+)}{p\left(X\right)}$ (formula 8)

Thus, we just can use the broadcast behavior parameter that calculates in the viewing behavior parameter that calculates in the step 206 and the step 208 to calculate based on Bayesian formula the probability of watching of every class program.

Subsequently, at next step 212, obtain every class program watch probability after, just can be in view of the above to user's recommending television, promptly at first determine to wait to broadcast the classification of program, search it then and watch probability, will respectively wait to broadcast program at last and sort by watching that probability is descending, the program commending that ordering is forward is given the user.

At last, method 200 ends at step 214.

The above only is preferred embodiment of the present invention, not in order to restriction the present invention, all any modifications of being done within the spirit and principles in the present invention, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.

Claims (7)

1. digital television program recommending method based on bayesian algorithm comprises:

S1, gather user watched record;

S2, analysis user rating record obtain the viewing behavior parameter, and this viewing behavior parameter comprises:

The probability that the user televiews;

The probability that the user does not teleview;

The probability of every class program in the program that the user watches;

The probability of every class program in the program that the user does not watch;

It is characterized in that described method also comprises:

S3, according to the viewing behavior parameter, obtain playing the behavior parameter, this broadcast behavior parameter comprises the broadcast probability of every class program;

S4, based on bayesian algorithm, by the viewing behavior parameter with play behavior parameter predictive user and treat the probability of watching of broadcasting TV programme, in view of the above to user's recommending television.

2. the digital television program recommending method based on bayesian algorithm according to claim 1 is characterized in that, establishes:

Each program X is expressed as { x ₁f ₁..., x _if _i..., x _nf _n, f wherein _iRepresent each attribute of program X, x _iRepresent the weight of each attribute of program X, wherein, 1≤i≤n, n represent the number of attributes of program, and n 〉=1;

C+ and C-represent the behavior of watching respectively and do not watch behavior;

The TV programme sum that TV station in a period of time is broadcasted is expressed as k (C+)+k (C-), and wherein, k (C+) and k (C-) distinguish the quantity of representative of consumer TV reception and the quantity of TV reception not;

K (f _i| C+) and k (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the quantity of every class program in the program that the quantity of every class program and user do not watch in the program that the user watches;

P (C+) and p (C-) be the probability that do not teleview of the probability that televiews of representative of consumer and user respectively;

P (f _i| C+) and p (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the probability of every class program in the TV programme that the probability of every class program and user do not watch in the program that the user watches;

P (X|C+) and p (X|C-) represent respectively at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the probability of every class program in the program that the probability of every class program and user do not watch in the program that the user watches;

Then in described S2, calculate described viewing behavior parameter according to following steps:

S21, the TV programme quantity and the user watched record that broadcast according to TV station are added up respectively

K (C+), k (C-), k (f _i| C+) and k (f _i| C-);

S22, calculating:

3. the digital television program recommending method based on bayesian algorithm according to claim 2 is characterized in that, establishes:

P (X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the broadcast probability of every class program;

Then in described S3, calculate described broadcast behavior parameter according to following steps:

S31, calculating:

p(X)＝p(X|C+)p(C+)+p(X|C-)p(C-)。

4. the digital television program recommending method based on bayesian algorithm according to claim 3 is characterized in that, establishes:

P (C+|X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, and the user is to the probability of watching of every class program;

Then in described S4, predict the probability of watching of every class program according to following steps:

S41, calculate based on Bayesian formula:

5. the digital television program recommending method based on bayesian algorithm according to claim 4 is characterized in that, at least one under each program comprises among the Column Properties:

Play the period;

The place channel;

Affiliated big class;

Affiliated group.

6. digital television program recommending system based on bayesian algorithm comprises:

Collecting unit, the rating record and the storage that are used to gather digital TV terminal;

Viewing behavior parameter extraction module with the collecting unit communication linkage, is used for analysis user rating record, output viewing behavior parameter, and this viewing behavior parameter comprises:

The probability that the user televiews;

The probability that the user does not teleview;

The probability of every class program in the program that the user watches;

The probability of every class program in the program that the user does not watch;

It is characterized in that, also comprise:

Broadcast behavior parameter extraction module communicates to connect with viewing behavior parameter extraction module, is used to analyze the viewing behavior parameter, output broadcast behavior parameter, and this broadcast behavior parameter comprises the broadcast probability of every class program;

Watch the probability extraction module, communicate to connect, be used to receive viewing behavior parameter and the behavior of broadcast parameter, treat the probability of watching of broadcasting TV programme based on bayesian algorithm output user with viewing behavior parameter extraction module and the behavior of broadcast parameter extraction module;

Recommending module communicates to connect with watching the probability extraction module, is used to receive the probability of watching of all kinds of TV programme, in view of the above to user's recommending television.

7. the digital television program recommending system based on bayesian algorithm according to claim 6 is characterized in that, establishes:

Each program X is expressed as { x ₁f ₁..., x _if _i..., x _nf _n, f wherein _iRepresent each attribute of program X, x _iRepresent the weight of each attribute of program X, wherein, 1≤i≤n, n represent the number of attributes of program, and n 〉=1;

C+ and C-represent the behavior of watching respectively and do not watch behavior;

The TV programme sum that TV station in a period of time is broadcasted is expressed as k (C+)+k (C-), and wherein, k (C+) and k (C-) distinguish the quantity of representative of consumer TV reception and the quantity of TV reception not;

K (f _i| C+) and k (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the quantity of every class program in the program that the quantity of every class program and user do not watch in the program that the user watches;

P (C+) and p (C-) be the probability that do not teleview of the probability that televiews of representative of consumer and user respectively;

P (f _i| C+) and p (f _i| C-) represent respectively at foundation attribute f _iAll TV programme are carried out the branch time-like, the probability of every class program in the TV programme that the probability of every class program and user do not watch in the program that the user watches;

P (X|C+) and p (X|C-) represent respectively at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the probability of every class program in the program that the probability of every class program and user do not watch in the program that the user watches;

P (X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, the broadcast probability of every class program;

P (C+|X) representative is at foundation attribute f ₁～f _nAll TV programme are carried out the branch time-like, and the user is to the probability of watching of every class program;

Then described viewing behavior parameter extraction module is used for calculating described viewing behavior parameter according to following steps:

K (C+), k (C-), k (f added up respectively in S1, the TV programme quantity and the user watched record that broadcast according to TV station _i| C+) and k (f _i| C-);

S2, calculating:

Described broadcast behavior parameter extraction module is used for calculating described broadcast behavior parameter according to following steps:

S3, calculating:

p(X)＝p(X|C+)p(C+)+p(X|C-)p(C-)；

The described probability extraction module of watching is used for predicting according to following steps the probability of watching of every class program:

S4, calculate based on Bayesian formula:

Images