CN103888498A - Information pushing method and apparatus, terminal and server - Google Patents
Information pushing method and apparatus, terminal and server Download PDFInfo
- Publication number
- CN103888498A CN103888498A CN201210562450.0A CN201210562450A CN103888498A CN 103888498 A CN103888498 A CN 103888498A CN 201210562450 A CN201210562450 A CN 201210562450A CN 103888498 A CN103888498 A CN 103888498A
- Authority
- CN
- China
- Prior art keywords
- interest
- probability
- scene
- product
- matrix
- 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
Images
Abstract
The invention discloses an information pushing method and apparatus, a terminal and a server, and belongs to the technical field of a computer. The method comprises: obtaining scene sequences during successive switching from a first scene to a current n-th scene within a preset time period, and interest sequences corresponding to the scene sequences, wherein n>=2; combining the scene sequences with scenes in a preset model to form a scene set, combining the interest sequences with interests in the model to form an interest set, and according to the scene set and the interest set, calculating the model parameters of the model; according to the scene sequences, the interest set and model parameters, calculating the current interest corresponding to a current n-th scene; and according to the current interest, pushing information to a terminal. According to the invention, the problem is solved that because the terminal does not carry out modeling on interests corresponding to windows according to scenes, the accuracy of information push to the terminal is affected, and the effect of improving the information push accuracy is achieved.
Description
Technical field
The present invention relates to field of computer technology, particularly a kind of information-pushing method, device, terminal and server.
Background technology
User produces user data in the process that uses application program, such as, the UGC(User Generated Content that user produces when browsing door information website, microblogging or using player, user-generated content) or status data etc.Along with the interpolation of function in the abundant and application program of application program kind, user uses the user data producing in the process of application program also more and more, abundant user data can the user data in multiple application program merge same user terminal, and according to the data after merging, user's interest is predicted, and then push the interested information of user's possibility to terminal.
In prior art, terminal is in advance for user sets up a model, in this model, the window information of terminal operating and user's interest are relations one to one, in the time that terminal is current window from last windows exchange, can in model, find the interest corresponding with current window according to current window information, push the information relevant to this interest to terminal.Such as, in model, setting in advance interest corresponding to web page windows is news, terminal detects when current window is web page windows, searches the interest corresponding with web page windows in model, obtains current events, pushes the information relevant to current events to terminal.
Realizing in process of the present invention, inventor finds that prior art at least exists following problem:
Move the same window of terminal in different scenes time, the interest that this window is corresponding may be different, such as, the interest corresponding to web page windows of operation in the morning may be that interest corresponding to the web page windows of current events, operation in afternoon may be entertainment news etc., and terminal is not carried out modeling according to scene to interest corresponding to window, affect the accuracy to terminal pushed information.
Summary of the invention
According to scene, interest corresponding to window is not carried out to modeling in order to solve terminal, affected the problem of the accuracy to terminal pushed information, the embodiment of the present invention provides a kind of information-pushing method, device, terminal and server.Described technical scheme is as follows:
On the one hand, provide a kind of information-pushing method, described method comprises:
Obtain the sequence of scenes and interest sequence corresponding to described sequence of scenes that in the default time period, are successively switched to n current scene from the 1st scene, n >=2;
Scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, adds up to according to described scene set and described interest set the model parameter of calculating described model;
Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter;
According to described current interest to terminal pushed information.
On the other hand, provide a kind of information push-delivery apparatus, described device comprises:
The first acquisition module, for obtaining the sequence of scenes and interest sequence corresponding to described sequence of scenes that are successively switched to n current scene in the default time period from the 1st scene, n >=2;
The first computing module, be combined into scene set for described sequence of scenes that described the first acquisition module is obtained and the scene of preset model, the synthetic interest set of interest group in described interest sequence and the described model that described the first acquisition module is obtained, adds up to according to described scene set and described interest set the model parameter of calculating described model;
The second computing module, the described model parameter of calculating for the described sequence of scenes of obtaining according to described the first acquisition module, described interest set and described the first computing module is calculated described current current interest corresponding to n scene;
The first pushing module, for according to described second computing module calculate described current interest to terminal pushed information.
On the one hand, provide a kind of terminal again, described terminal comprises information push-delivery apparatus as above.
Another aspect, provides a kind of server, and described server comprises information push-delivery apparatus as above.
The beneficial effect that the technical scheme that the embodiment of the present invention provides is brought is:
By the scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, add up to according to described scene set and described interest set the model parameter of calculating described model; Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter; , solve terminal and according to scene, interest corresponding to window has not been carried out to modeling to terminal pushed information according to described current interest, affected the problem of the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.
Accompanying drawing explanation
In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing of required use during embodiment is described is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skills, do not paying under the prerequisite of creative work, can also obtain according to these accompanying drawings other accompanying drawing.
Fig. 1 is the method flow diagram of the information-pushing method that provides of the embodiment of the present invention one;
Fig. 2 is the method flow diagram of the information-pushing method that provides of the embodiment of the present invention two;
Fig. 3 is the method flow diagram of the information-pushing method that provides of the embodiment of the present invention three;
Fig. 4 is the structural representation of the information push-delivery apparatus that provides of the embodiment of the present invention four;
Fig. 5 is the structural representation of the information push-delivery apparatus that provides of the embodiment of the present invention five.
Embodiment
For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.
Embodiment mono-
Please refer to Fig. 1, it shows the method flow diagram of the information-pushing method that the embodiment of the present invention one provides, and this information-pushing method can be applied in terminal, and this terminal can be intelligent television, smart mobile phone or panel computer etc.; Or this information-pushing method also can be applied in server.The embodiment of the present invention is applied in terminal as example describes take the method, and the method for this information pushing, comprising:
Step 102: obtain the sequence of scenes and interest sequence corresponding to sequence of scenes that are successively switched to n current scene in the default time period from the 1st scene, n >=2;
Wherein, scene is the operation of terminal to the page or application program.Such as, in the time that the page is operated, scene can include but not limited to: the click behavior producing in content of pages, subject of Web site, short time, page area, the link of current page etc. at current visual angle place; In the time that application programs operates, scene can include but not limited to: application program running time, application category, user's use habit, user generated content (UGC), customer group distribute, use the time period, terminal type, terminal versions number of terminal etc.
Interest is the information that user passes through the page or application program acquisition.Such as, interest can include but not limited to: information technology IT, real estate, dress ornament, personal belongings, industry articles for use, decoration, traffic, education, finance, service, retail, game, consumption, medical treatment, amusement etc.
Step 104: the scene in sequence of scenes and preset model is combined into scene set, by synthetic the interest group in interest sequence and model interest set, the model parameter of closing computation model according to scene set and interest set;
Step 106: calculate current current interest corresponding to n scene according to sequence of scenes, interest set and model parameter;
Step 108: according to current interest to terminal pushed information.
Terminal is carried out to information pushing to be referred in current scene to user and recommends with the relevant information of interest.
In sum, the information-pushing method that the embodiment of the present invention provides, by the scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, add up to according to described scene set and described interest set the model parameter of calculating described model; Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter; , solve terminal and according to scene, interest corresponding to window has not been carried out to modeling to terminal pushed information according to described current interest, affected the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.
Embodiment bis-
Please refer to Fig. 2, it shows the method flow diagram of the information-pushing method that the embodiment of the present invention two provides, and this information-pushing method can be applied in terminal, and this terminal can be intelligent television, smart mobile phone or panel computer etc.; Or this information-pushing method also can be applied in server.The embodiment of the present invention is applied in terminal as example describes take the method, and this information-pushing method, comprising:
Step 202: obtain the sequence of scenes and interest sequence corresponding to this sequence of scenes that are successively switched to n current scene in the default time period from the 1st scene, n >=2;
Wherein, scene is the operation of terminal to the page or application program.Such as, in the time that the page is operated, scene can include but not limited to: the click behavior producing in content of pages, subject of Web site, short time, page area, the link of current page etc. at current visual angle place; In the time that application programs operates, scene can include but not limited to: application program running time, application category, user's use habit, user generated content (UGC), customer group distribute, use the time period, terminal type, terminal versions number of terminal etc.
Particularly, in the time of the page in user's operating terminal or application program, the operation that terminal can be carried out by log recording, can set in advance a time period, within this time period, read daily record in real time in terminal carry out operation, can obtain sequence of scenes.The default time period can arrange voluntarily and adjust, and the present embodiment is not construed as limiting.
Such as, first user has opened news pages, has then opened player, then gets back to again news pages, and the sequence of scenes of obtaining is " news pages, player, news pages ".
Interest is the information that user passes through the page or application program acquisition.Such as, interest can include but not limited to: information technology IT, real estate, dress ornament, personal belongings, industry articles for use, decoration, traffic, education, finance, service, retail, game, consumption, medical treatment, amusement etc.
Further, in the time that user passes through terminal to server requested webpage information, server can carry out record to the information of terminal request, terminal reads the information recording in server, in conjunction with the log information of terminal, can get interest corresponding to each scene, thereby determine the interest sequence corresponding with sequence of scenes.
Such as, user browsed financial finance and economics in news pages, in player, selected folk song, browsed social news, interest sequence that can be corresponding using " financial finance and economics, folk song, social news " as " news pages, player, news pages " while getting back to news pages.
Step 204: the scene in sequence of scenes and preset model is combined into scene set, by synthetic the interest group in interest sequence and model interest set, the model parameter of closing computation model according to scene set and interest set, this model parameter comprises initial probability matrix, transition probability matrix and emission matrix;
Terminal can also be set up model to scene and interest corresponding to this scene in this locality, this model can comprise scene set, interest set and model parameter, be used for predicting the interest that scene is corresponding, or predict next scene, so that server carries out information pushing according to the interest or next scene that dope to terminal, make information pushing hommization and intellectuality more.
Particularly, the model parameter of closing computation model according to scene set and interest set, can comprise:
Set in advance initial probability matrix or obtain previous initial probability matrix;
For i interest in m interest of interest set, calculate successively i interest and transfer to the transition probability of j interest, obtain transition probability matrix;
For i interest in interest set, calculate successively i interest and produce the emission probability of k scene, obtain emission matrix;
Wherein, 1≤i≤m, 1≤j≤m and 1≤k≤n.
Wherein, the initial probability of i interest sets in advance, and the initial probability that each interest is set such as, terminal equates, or the initial probability etc. of each interest is set based on experience value, and the present embodiment is not construed as limiting.Further, terminal can also be adjusted initial probability matrix.
In the time that terminal is calculated transition probability matrix, the number of times that i interest can be transferred to j interest is divided by i the interest total degree shifting of taking up, obtain i interest and transfer to j interest transition probability, certainly, can also calculate by other means transition probability matrix, the present embodiment is not construed as limiting.
In the time that terminal is calculated emission matrix, the number of times that i interest can be produced to k scene produces the total degree of scene divided by i interest, obtain i interest and produce the emission probability of k scene, certain, can also calculate by other means emission matrix, the present embodiment is not construed as limiting.
In addition, the present embodiment does not limit the computation sequence of initial probability matrix, probability transfer matrix and emission matrix.
Step 206: calculate current current interest corresponding to n scene according to sequence of scenes, interest set and model parameter;
Particularly, calculate current current interest corresponding to n scene according to sequence of scenes, interest set and model parameter, can comprise:
In the time calculating the probability of each interest in the 1st scene, the product of emission probability that the initial probability using i interest in initial probability matrix and i interest produce the 1st scene in emission matrix is the probability in the 1st scene as i interest;
In the time calculating the probability of each interest in q scene, first calculate successively the probability of j interest in previous scene and j interest and in transition probability matrix, transfer to the first product of the transition probability of i interest, determine maximum the first product in the first product, calculate again the second product that i interest of maximum the first sum of products produces the emission probability of q scene in emission matrix, using the second product as i interest the probability in q scene, 2≤q≤n-1;
In the time calculating the probability of each interest in n scene, first calculate successively the probability of j interest in previous scene and j interest and in transition probability matrix, transfer to the first product of the transition probability of i interest, determine maximum the first product in the first product, calculate again the second product that i interest of maximum the first sum of products produces the emission probability of n scene in emission matrix, using the second product as i interest the probability in n scene, determine maximum the second product in the second product, using interest corresponding maximum the second product as current interest.
Such as, suppose that interest set is combined into that { a, b, c}, scene set is that { g, h}, the current sequence of scenes of obtaining is ghgh, and initial probability matrix
Transition probability matrix
Emission matrix
The computational process of current interest corresponding to n scene is as follows:
Be incorporated herein intermediate quantity δ
n(j), δ wherein
n(j) be the probability of interest j in n scene;
The probability calculation of each interest in the 1st scene g:
δ
1(a)=π
1·b
1(g)=1·0.7=0.7；
The probability calculation of each interest in the 2nd scene h:
δ
2(a)=δ
1(a)·a
11·b
1(h)=0.7·0.4·0.3=0.084；
δ
2(b)=δ
1(a)·a
12·b
2(h)=0.7·0.6·0.6=0.252；
The probability calculation of each interest in the 3rd scene g:
δ
3(a)=δ
2(a)·a
11·b
1(g)=0.084·0.4·0.7=0.02352；
δ
3(b)=max{δ
2(a)·a
12,δ
2(b)·a
22}·b
2(g)=max{0.084·0.6,0.0252·0.8}·0.4=0.08064；
δ
3(c)=δ
2(b)·a
23·b
3(g)=0.252·0.2·0.8=0.04032；
The probability calculation of each interest in the 4th scene h:
δ
4(a)=δ
3(a)·a
11·b
1(h)=0.02352·0.4·0.3=0.0028224；
δ
4(b)=max{δ
3(a)·a
12,δ
3(b)·a
22}·b
2(h)=max{0.014112,0.064512}·0.6=0.0387072；
δ
4(c)=max{δ
3(b)·a
23,δ
3(c)·a
32}·b
3(h)=max{0.016128,0.04032}·0.2=0.008064；
Known according to result of calculation, the maximum probability of b in the 4th scene h, terminal determines that b is the 4th the current interest that scene h is corresponding.
Further, scene and the interest obtained due to terminal can not be unlimited many, therefore, the interest not getting for physical presence shifts, while calculating transition probability matrix according to above-mentioned computational methods, the transition probability between the interest calculating is zero, such as, in A matrix, to transfer to the transition probability of a be that transition probability that 0, c transfers to b is 0 etc. to c.In the time that transition probability is zero, the probability of the interest calculating in scene must be that zero, one calculating path has just interrupted, and affected the accuracy of model.For fear of this situation, can in the time calculate transition probability matrix, adopt and add a smoothing algorithm, add one by the transfer number between any two interest in interest set.
Such as, interest set be combined into a, b, c}, while calculating the transition probability of a, suppose that the number of times that a transfers to a is 6 times, and the number of times that a forwards b to is 3 times, and the number of times that a transfers to c is 0 time, and the probability that a calculating transfers to c is 0/(6+3)=0.Employing adds after a smoothing algorithm, can obtain the number of times that a transfers to a is 7 times, and the number of times that a forwards b to is 4 times, and the number of times that a transfers to c is 1 time, the probability that a now calculating transfers to c is 1/(7+4+1)=0.08333, the problem of having avoided calculating path to interrupt.
In like manner, in the time calculating emission matrix, also can adopt the above-mentioned smoothing algorithm that adds, not repeat herein.
Step 208: according to current interest to terminal pushed information.
When terminal gets the current interest calculating, obtain the information corresponding with this interest according to this current interest to server, and in terminal, show this information.
In sum, the information-pushing method that the embodiment of the present invention provides, by the scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, add up to according to described scene set and described interest set the model parameter of calculating described model; Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter; , solve terminal and according to scene, interest corresponding to window has not been carried out to modeling to terminal pushed information according to described current interest, affected the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.In addition, by i interest in the interest of the m for described interest set, calculate successively described i interest and transfer to the transition probability of j interest, obtain described transition probability matrix; For i interest in described interest set, calculate successively the emission probability that described i interest produces k scene, obtain described emission matrix, solve and can not calculate the relation of interest in model and scene, affect the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.
Embodiment tri-
Please refer to Fig. 3, it shows the method flow diagram of the information-pushing method that the embodiment of the present invention three provides, and this information-pushing method can be applied in terminal, and this terminal can be intelligent television, smart mobile phone or panel computer etc.; Or this information-pushing method also can be applied in server.The embodiment of the present invention is applied in server as example describes take the method, and this information-pushing method, comprising:
Step 302: obtain the sequence of scenes and interest sequence corresponding to sequence of scenes that are successively switched to n current scene in the default time period from the 1st scene, n >=2;
Wherein, scene is the operation of terminal to the page or application program.Such as, in the time that the page is operated, scene can include but not limited to: the click behavior producing in content of pages, subject of Web site, short time, page area, the link of current page etc. at current visual angle place; In the time that application programs operates, scene can include but not limited to: application program running time, application category, user's use habit, user generated content (UGC), customer group distribute, use the time period, terminal type, terminal versions number of terminal etc.
Wherein, the process that terminal is obtained scene refers to the description in step 202, does not repeat herein.Further, terminal can send to server by getting sequence of scenes, so that server carries out record to scene information.
Interest is the information that user passes through the page or application program acquisition.Such as, interest can include but not limited to: information technology IT, real estate, dress ornament, personal belongings, industry articles for use, decoration, traffic, education, finance, service, retail, game, consumption, medical treatment, amusement etc.
Further, in the time that user passes through terminal to server requested webpage information, server can carry out record to the information of terminal request, server reads the information of record, the scene information sending in conjunction with terminal, can get interest corresponding to each scene, thereby determine the interest sequence corresponding with sequence of scenes.
Step 304: the scene in sequence of scenes and preset model is combined into scene set, by synthetic the interest group in interest sequence and model interest set, the model parameter of closing computation model according to scene set and interest set, this model parameter comprises initial probability matrix, transition probability matrix and emission matrix;
Server can also be set up model to scene and interest corresponding to this scene, this model can comprise scene set, interest set and model parameter, be used for predicting the interest that scene is corresponding, or predict next scene, so that server carries out information pushing according to the interest or next scene that dope to terminal, make information pushing hommization and intellectuality more.
Particularly, the model parameter of closing computation model according to scene set and interest set, can comprise:
Set in advance initial probability matrix or obtain previous initial probability matrix;
For i interest in m interest of interest set, calculate successively i interest and transfer to the transition probability of j interest, obtain transition probability matrix;
For i interest in interest set, calculate successively i interest and produce the emission probability of k scene, obtain emission matrix;
Wherein, 1≤i≤m, 1≤j≤m and 1≤k≤n.
Wherein, server refers to the description in step 204 to the computational process of model parameter, does not repeat herein.
Step 306: obtain at least one forecasting sequence, each forecasting sequence comprises that sequence of scenes adds a prediction scene, and prediction scene is the scene in scene set, and prediction scene difference in each forecasting sequence;
Such as, sequence of scenes is ghg, and scene set is that { forecasting sequence can be ghgg, or ghgh for g, h}.
Step 308: for each forecasting sequence, calculate the probability of forecasting sequence according to forecasting sequence, interest set and model parameter;
Server provides two kinds of methods of calculating the probability of forecasting sequence, particularly, calculates the probability of forecasting sequence according to forecasting sequence, interest set and model parameter, can comprise:
According to the probability of predicting prediction interest corresponding to scene in forecasting sequence, interest set and model parameter forward calculation forecasting sequence, the probability using the probability of prediction interest as forecasting sequence; Or,
According to the probability of the 1st interest that scene is corresponding in forecasting sequence, interest set and model parameter backwards calculation forecasting sequence, the probability using the probability of interest as forecasting sequence.
Further, according to the probability of predicting prediction interest corresponding to scene in forecasting sequence, interest set and model parameter forward calculation forecasting sequence, comprising:
In the time calculating the probability of each interest in the 1st scene, the product of emission probability that the initial probability using i interest in initial probability matrix and i interest produce the 1st scene in emission matrix is the probability in the 1st scene as i interest;
In the time calculating the probability of each interest in q scene, first calculate successively the probability of j interest in previous scene and j interest and in transition probability matrix, transfer to the 3rd product of the transition probability of i interest, calculate the 3rd sum of products after the 3rd product is added, calculate again the 4th product that the 3rd sum of products and i interest produce the emission probability of q scene in emission matrix, using the 4th product as i interest the probability in q scene, 2≤q≤n-1;
In the time calculating the probability of each interest in n scene, first calculate successively the probability of j interest in previous scene and j interest and in transition probability matrix, transfer to the 3rd product of the transition probability of i interest, calculate the 3rd sum of products after the 3rd product is added, calculate again the 4th product that the 3rd sum of products and i interest produce the emission probability of n scene in emission matrix, using the 4th product as i interest the probability in n scene, determine maximum the 4th product in the 4th product, maximum the 4th product is predicted to the probability of prediction interest corresponding to scene in forecasting sequence.
Or, according to the probability of the 1st interest that scene is corresponding in forecasting sequence, interest set and model parameter backwards calculation forecasting sequence, can comprise:
In the time calculating the probability of each interest in n scene, the product of emission probability that the initial probability using i interest in initial probability matrix and i interest produce n scene in emission matrix is the probability in n scene as i interest;
In the time calculating the probability of each interest in q scene, first calculate successively the probability of j interest in a rear scene and j interest and in transition probability matrix, transfer to the 5th product of the transition probability of i interest, calculate the 5th sum of products after the 5th product is added, calculate again the 6th product that the 5th sum of products and i interest produce the emission probability of q scene in emission matrix, using the 6th product as i interest the probability in q scene, 2≤q≤n-l;
In the time calculating the probability of each interest in the 1st scene, first calculate successively the probability of j interest in a rear scene and j interest and in transition probability matrix, transfer to the 5th product of the transition probability of i interest, calculate the 5th sum of products after the 5th product is added, calculate again the 6th product that the 5th sum of products and i interest produce the emission probability of the 1st scene in emission matrix, probability using the 6th product as i interest in the 1st scene, determine maximum the 6th product in the 6th product, using the probability of maximum the 6th product the 1st interest that scene is corresponding in forecasting sequence.
Such as, to describe as example according to the probability of predicting prediction interest corresponding to scene in forecasting sequence, interest set and model parameter forward calculation forecasting sequence, suppose interest set be combined into a, b, c}, scene set is { g, h}, current prediction sequence of scenes is ghgh, and initial probability matrix
Transition probability matrix
Emission matrix
The computational process of current interest corresponding to n scene is as follows:
Be incorporated herein intermediate quantity α
n(j), α wherein
n(j) be the probability of interest j in n scene;
The probability calculation of each interest in the 1st scene g:
α
1(a)=π
1·b
1(g)=1·O.7=O.7，
The probability calculation of each interest in the 2nd scene h:
α
2(a)＝α
1(a)·a
11·b
1(h)＝O.7·O.4·O.3＝0.084；
α
2(b)＝α
1(a)·a
12．b
2(h)＝O.7·O.6·O.6＝0.252；
The probability calculation of each interest in the 3rd scene g:
α
3(a)＝α
2(a)·a
11·b
1(g)＝0.084·O.4·O.7＝0.02352；
α
3(b)＝[α
2(a)·a
12+α
2(b)·a
22]·b
2(g)＝[0.084·0.6+0.0252·0.8]·o.4＝0.1008；
α
3(c)＝α
2(b)·a
23·b
3(g)＝0.252·O.2·O.8＝0.04032；
The probability calculation of each interest in prediction scene (the 4th scene) h:
α
4(a)＝α
3(a)·a
11·b
1(h)＝0.02352·O.4·O.3＝0.0028224；
α
4(b)＝[α
3(a)·a
12+α
3(b)·a
22]·b
2(h)＝[0.02352·0.6+0.1008·0.8]·0.6＝0.0568512
；
α
4(c)＝[α
3(b)·a
23+α
3(c)·a
32J·b
3(h)＝[0.1008·0.2+0.04032·1]0.2＝0.012096；
Known according to result of calculation, the probability of forecasting sequence ghgh is 0.012096, and server can further calculate the probability of ghgg, then performs step 310.
Further, scene and the interest obtained due to terminal can not be unlimited many, therefore, the interest not getting for physical presence shifts, while calculating transition probability matrix according to above-mentioned computational methods, the transition probability between the interest calculating is zero, such as, in A matrix, to transfer to the transition probability of a be that transition probability that 0, c transfers to b is 0 etc. to c.In the time that transition probability is zero, the probability of the interest calculating in scene must be that zero, one calculating path has just interrupted, and affected the accuracy of model.For fear of this situation, can in the time calculate transition probability matrix, adopt and add a smoothing algorithm, add one by the transfer number between any two interest in interest set.In like manner, in the time calculating emission matrix, also can adopt the above-mentioned smoothing algorithm that adds, concrete calculation process please refer to the description in step 206, does not repeat herein.
Step 310: determine forecasting sequence corresponding to maximum probability in probability, and in the prediction scene of the corresponding forecasting sequence of maximum probability to terminal pushed information.
Server determines according to the probability of ghgh and the probability of ghgg that calculate the forecasting sequence that maximum probability is corresponding, if forecasting sequence corresponding to maximum probability is ghgh, server determines that prediction scene is h; If forecasting sequence corresponding to maximum probability is ghgg, server determine prediction scene be g, and in this prediction scene to terminal pushed information.
In sum, the information-pushing method that the embodiment of the present invention provides, by the scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, add up to according to described scene set and described interest set the model parameter of calculating described model; Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter; , solve terminal and according to scene, interest corresponding to window has not been carried out to modeling to terminal pushed information according to described current interest, affected the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.In addition, by i interest in the interest of the m for described interest set, calculate successively described i interest and transfer to the transition probability of j interest, obtain described transition probability matrix; For i interest in described interest set, calculate successively the emission probability that described i interest produces k scene, obtain described emission matrix, solve and can not calculate the relation of interest in model and scene, affect the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.
Embodiment tetra-
Please refer to Fig. 4, it shows the structural framing figure of the information push-delivery apparatus that the embodiment of the present invention four provides, and this information push-delivery apparatus can be applied in terminal, and this terminal can be intelligent television, smart mobile phone or panel computer etc.; Or this information-pushing method also can be applied in server.This information push-delivery apparatus, comprising:
The first acquisition module 410, for obtaining the sequence of scenes and interest sequence corresponding to sequence of scenes that are successively switched to n current scene in the default time period from the 1st scene, n >=2;
The first computing module 420, be combined into scene set for sequence of scenes that the first acquisition module 410 is obtained and the scene of preset model, the synthetic interest set of interest group in interest sequence and the model that the first acquisition module 410 is obtained, the model parameter of closing computation model according to scene set and interest set;
The second computing module 430, the model parameter of calculating for the sequence of scenes, interest set and the first computing module 420 that obtain according to the first acquisition module 410 is calculated current current interest corresponding to n scene;
The first pushing module 440, for the current interest calculated according to the second computing module 430 to terminal pushed information.
In sum, the information push-delivery apparatus that the embodiment of the present invention provides, by the scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, add up to according to described scene set and described interest set the model parameter of calculating described model; Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter; , solve terminal and according to scene, interest corresponding to window has not been carried out to modeling to terminal pushed information according to described current interest, affected the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.
Embodiment five
Please refer to Fig. 5, it shows the structural framing figure of the information push-delivery apparatus that the embodiment of the present invention five provides, and this information-pushing method can be applied in terminal, and this terminal can be intelligent television, smart mobile phone or panel computer etc.; Or this information-pushing method also can be applied in server.This information push-delivery apparatus, comprising: the first acquisition module 410, the first computing module 420, the second computing module 430 and the first pushing module 440.
The first acquisition module 410, for obtaining the sequence of scenes and interest sequence corresponding to sequence of scenes that are successively switched to n current scene in the default time period from the 1st scene, n >=2;
The first computing module 420, be combined into scene set for sequence of scenes that the first acquisition module 410 is obtained and the scene of preset model, the synthetic interest set of interest group in interest sequence and the model that the first acquisition module 410 is obtained, the model parameter of closing computation model according to scene set and interest set;
The second computing module 430, the model parameter of calculating for the sequence of scenes, interest set and the first computing module 420 that obtain according to the first acquisition module 410 is calculated current current interest corresponding to n scene;
The first pushing module 440, for the current interest calculated according to the second computing module 430 to terminal pushed information.
Further, model parameter comprises initial probability matrix, transition probability matrix and emission matrix, and the first computing module 420 can comprise:
Acquiring unit 510, for setting in advance initial probability matrix or obtaining previous initial probability matrix;
The first computing unit 520, for i interest of the interest of the m for interest set, calculates i interest and transfers to the transition probability of j interest successively, obtains transition probability matrix;
The second computing unit 530, for i the interest for interest set, calculates i interest and produces the emission probability of k scene successively, obtains emission matrix;
Wherein, 1≤i≤m, 1≤j≤m and 1≤k≤n.
Further, the second computing module 430 can comprise:
The 3rd computing unit 610, for in the time calculating the probability of the 1st each interest of scene, the product of emission probability that i the interest that the initial probability of i the interest that acquiring unit 510 is obtained in initial probability matrix and the second computing unit 530 calculate produces the 1st scene in emission matrix is the probability in the 1st scene as i interest;
The 4th computing unit 620, for in the time calculating the probability of q each interest of scene, first calculate successively j interest that probability in previous scene of j interest that the 3rd computing unit 610 calculates and the first computing unit 520 calculate and in transition probability matrix, transfer to the first product of the transition probability of i interest, determine maximum the first product in the first product, calculate again the second product that i interest that maximum first sum of products the second computing unit 530 calculates produces the emission probability of q scene in emission matrix, using the second product as i interest the probability in q scene, 2≤q≤n-1,
The 5th computing unit 630, for in the time calculating the probability of n each interest of scene, first calculate successively j interest that probability in previous scene of j interest that the 4th computing unit 620 calculates and the first computing unit 520 calculate and in transition probability matrix, transfer to the first product of the transition probability of i interest, determine maximum the first product in the first product, calculate again the second product that i interest that maximum first sum of products the second computing unit 530 calculates produces the emission probability of n scene in emission matrix, using the second product as i interest the probability in n scene, determine maximum the second product in the second product, using interest corresponding maximum the second product as current interest.
Further, this device can also comprise:
The second acquisition module 450, for obtaining at least one forecasting sequence, each forecasting sequence comprises that sequence of scenes adds a prediction scene, prediction scene is the scene in scene set, and prediction scene difference in each forecasting sequence;
The 3rd computing module 460, for each forecasting sequence obtaining for the second acquisition module 450, calculates the probability of forecasting sequence according to forecasting sequence, interest set and model parameter;
The second pushing module 470, for determining forecasting sequence corresponding to probability maximum probability that calculate of the 3rd computing module 460, and in the prediction scene of the corresponding forecasting sequence of maximum probability to terminal pushed information.
Further, the 3rd computing module 460 can comprise:
The 6th computing unit 710, predicts the probability of prediction interest corresponding to scene for forecasting sequence, interest set and the model parameter forward calculation forecasting sequence obtaining according to the second acquisition module 450, probability that will prediction interest is as the probability of forecasting sequence; Or,
The 7th computing unit 720, for the probability of forecasting sequence, interest set and the 1st interest that scene is corresponding of model parameter backwards calculation forecasting sequence obtained according to the second acquisition module 450, the probability using the probability of interest as forecasting sequence.
Further, the 6th computing unit 710, for in the time calculating the probability of the 1st each interest of scene, the product of emission probability that i the interest that the initial probability of i the interest that acquiring unit 510 is obtained in initial probability matrix and the second computing unit 530 calculate produces the 1st scene in emission matrix is the probability in the 1st scene as i interest;
In the time calculating the probability of each interest in q scene, first calculate successively j interest that probability in previous scene of j interest that the 6th computing unit 710 calculates and the first computing unit 520 calculate and in transition probability matrix, transfer to the 3rd product of the transition probability of i interest, calculate the 3rd sum of products after the 3rd product is added, calculate again the 4th product that i interest that the 3rd sum of products and the second computing unit 530 calculate produces the emission probability of q scene in emission matrix, using the 4th product as i interest the probability in q scene, 2≤q≤n-1,
In the time calculating the probability of each interest in n scene, first calculate successively j interest that probability in previous scene of j interest that the 6th computing unit 710 calculates and the first computing unit 520 calculate and in transition probability matrix, transfer to the 3rd product of the transition probability of i interest, calculate the 3rd sum of products after the 3rd product is added, calculate again the 4th product that i interest that the 3rd sum of products and the second computing unit 530 calculate produces the emission probability of n scene in emission matrix, using the 4th product as i interest the probability in n scene, determine maximum the 4th product in the 4th product, maximum the 4th product is predicted to the probability of prediction interest corresponding to scene in forecasting sequence.
Further, the 7th computing unit 720, for in the time calculating the probability of n each interest of scene, the product of emission probability that i the interest that the initial probability of i the interest that acquiring unit 510 is obtained in initial probability matrix and the second computing unit 530 calculate produces n scene in emission matrix is the probability in n scene as i interest;
In the time calculating the probability of each interest in q scene, first calculate successively j interest that probability in a rear scene of j interest that the 7th computing unit 720 calculates and the first computing unit 520 calculate and in transition probability matrix, transfer to the 5th product of the transition probability of i interest, calculate the 5th sum of products after the 5th product is added, calculate again the 6th product that i interest that the 5th sum of products and the second computing unit 530 calculate produces the emission probability of q scene in emission matrix, using the 6th product as i interest the probability in q scene, 2≤q≤n-1,
In the time calculating the probability of each interest in the 1st scene, first calculate successively j interest that probability in a rear scene of j interest that the 7th computing unit 720 calculates and the first computing unit 520 calculate and in transition probability matrix, transfer to the 5th product of the transition probability of i interest, calculate the 5th sum of products after the 5th product is added, calculate again the 6th product that i interest that the 5th sum of products and the second computing unit 530 calculate produces the emission probability of the 1st scene in emission matrix, probability using the 6th product as i interest in the 1st scene, determine maximum the 6th product in the 6th product, using the probability of maximum the 6th product the 1st interest that scene is corresponding in forecasting sequence.
In sum, the information push-delivery apparatus that the embodiment of the present invention provides, by the scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, add up to according to described scene set and described interest set the model parameter of calculating described model; Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter; , solve terminal and according to scene, interest corresponding to window has not been carried out to modeling to terminal pushed information according to described current interest, affected the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.In addition, by i interest in the interest of the m for described interest set, calculate successively described i interest and transfer to the transition probability of j interest, obtain described transition probability matrix; For i interest in described interest set, calculate successively the emission probability that described i interest produces k scene, obtain described emission matrix, solve and can not calculate the relation of interest in model and scene, affect the problem of the accuracy to terminal pushed information, reached the effect that improves information pushing accuracy.
It should be noted that: the information push-delivery apparatus that above-described embodiment provides is in the time carrying out information pushing, only be illustrated with the division of above-mentioned each functional module, in practical application, can above-mentioned functions be distributed and completed by different functional modules as required, be divided into different functional modules by the internal structure of information push-delivery apparatus, to complete all or part of function described above.In addition, the information push-delivery apparatus that above-described embodiment provides and information-pushing method embodiment belong to same design, and its specific implementation process refers to embodiment of the method, repeats no more here.
The invention described above embodiment sequence number, just to describing, does not represent the quality of embodiment.
One of ordinary skill in the art will appreciate that all or part of step that realizes above-described embodiment can complete by hardware, also can carry out the hardware that instruction is relevant by program completes, described program can be stored in a kind of computer-readable recording medium, the above-mentioned storage medium of mentioning can be read-only memory, disk or CD etc.
The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.
Claims (16)
1. an information-pushing method, is characterized in that, described method comprises:
Obtain the sequence of scenes and interest sequence corresponding to described sequence of scenes that in the default time period, are successively switched to n current scene from the 1st scene, n >=2;
Scene in described sequence of scenes and preset model is combined into scene set, by synthetic the interest group in described interest sequence and described model interest set, adds up to according to described scene set and described interest set the model parameter of calculating described model;
Calculate described current current interest corresponding to n scene according to described sequence of scenes, described interest set and described model parameter;
According to described current interest to terminal pushed information.
2. information-pushing method according to claim 1, it is characterized in that, described model parameter comprises initial probability matrix, transition probability matrix and emission matrix, described according to the model parameter of described scene set and the described model of described interest set total calculation, comprising:
Set in advance initial probability matrix or obtain previous initial probability matrix;
For i interest in m interest of described interest set, calculate successively described i interest and transfer to the transition probability of j interest, obtain described transition probability matrix;
For i interest in described interest set, calculate successively described i interest and produce the emission probability of k scene, obtain described emission matrix;
Wherein, 1≤i≤m, 1≤j≤m and 1≤k≤n.
3. information-pushing method according to claim 2, is characterized in that, described according to described sequence of scenes, described interest set and described current current interest corresponding to n scene of described model parameter calculating, comprising:
In the time calculating the probability of each interest in the 1st scene, the product of emission probability that the initial probability using i interest in described initial probability matrix and described i interest produce described the 1st scene in described emission matrix is the probability in described the 1st scene as described i interest;
In the time calculating the probability of each interest in q scene, first calculate successively the probability of j interest in previous scene and described j interest and in described transition probability matrix, transfer to the first product of the transition probability of i interest, determine maximum the first product in described the first product, calculate again described in described maximum the first sum of products the second product that i interest produces the emission probability of described q scene in described emission matrix, using described the second product as described i interest the probability in described q scene, 2≤q≤n-1;
In the time calculating the probability of each interest in n scene, first calculate successively the probability of j interest in previous scene and described j interest and in described transition probability matrix, transfer to the first product of the transition probability of i interest, determine maximum the first product in described the first product, calculate again described in described maximum the first sum of products the second product that i interest produces the emission probability of described n scene in described emission matrix, using described the second product as described i interest the probability in described n scene, determine maximum the second product in described the second product, using interest corresponding described maximum the second product as described current interest.
4. information-pushing method according to claim 1, is characterized in that, described method, also comprises:
Obtain at least one forecasting sequence, each forecasting sequence comprises that described sequence of scenes adds a prediction scene, and described prediction scene is the scene in described scene set, and described prediction scene difference in each forecasting sequence;
For each forecasting sequence, calculate the probability of described forecasting sequence according to described forecasting sequence, described interest set and described model parameter;
Determine forecasting sequence corresponding to maximum probability in described probability, and in the prediction scene of the corresponding forecasting sequence of described maximum probability to described terminal pushed information.
5. information-pushing method according to claim 4, is characterized in that, the described probability that calculates described forecasting sequence according to described forecasting sequence, described interest set and described model parameter, comprising:
According to the probability of predicting prediction interest corresponding to scene in forecasting sequence described in described forecasting sequence, described interest set and described model parameter forward calculation, the probability using the probability of described prediction interest as described forecasting sequence; Or,
According to the probability of the 1st interest that scene is corresponding in forecasting sequence described in described forecasting sequence, described interest set and described model parameter backwards calculation, the probability using the probability of described interest as described forecasting sequence.
6. information-pushing method according to claim 5, is characterized in that, described according to the probability of predicting prediction interest corresponding to scene in forecasting sequence described in described forecasting sequence, described interest set and described model parameter forward calculation, comprising:
In the time calculating the probability of each interest in the 1st scene, the product of emission probability that the initial probability using i interest in described initial probability matrix and described i interest produce described the 1st scene in described emission matrix is the probability in described the 1st scene as described i interest;
In the time calculating the probability of each interest in q scene, first calculate successively the probability of j interest in previous scene and described j interest and in described transition probability matrix, transfer to the 3rd product of the transition probability of i interest, calculate the 3rd sum of products after described the 3rd product is added, calculate again the 4th product that described the 3rd sum of products and described i interest produce the emission probability of described q scene in described emission matrix, using described the 4th product as described i interest the probability in described q scene, 2≤q≤n-1;
In the time calculating the probability of each interest in n scene, first calculate successively the probability of j interest in previous scene and described j interest and in described transition probability matrix, transfer to the 3rd product of the transition probability of i interest, calculate the 3rd sum of products after described the 3rd product is added, calculate again the 4th product that described the 3rd sum of products and described i interest produce the emission probability of described n scene in described emission matrix, using described the 4th product as described i interest the probability in a described n scene, determine maximum the 4th product in described the 4th product, described maximum the 4th product is predicted to the probability of prediction interest corresponding to scene in described forecasting sequence.
7. information-pushing method according to claim 5, is characterized in that, described according to the probability of the 1st interest that scene is corresponding in forecasting sequence described in described forecasting sequence, described interest set and described model parameter backwards calculation, comprising:
In the time calculating the probability of each interest in n scene, the product of emission probability that the initial probability using i interest in described initial probability matrix and described i interest produce described n scene in described emission matrix is the probability in described n scene as described i interest;
In the time calculating the probability of each interest in q scene, first calculate successively the probability of j interest in a rear scene and described j interest and in described transition probability matrix, transfer to the 5th product of the transition probability of i interest, calculate the 5th sum of products after described the 5th product is added, calculate again the 6th product that described the 5th sum of products and described i interest produce the emission probability of described q scene in described emission matrix, using described the 6th product as described i interest the probability in described q scene, 2≤q≤n-1;
In the time calculating the probability of each interest in the 1st scene, first calculate successively the probability of j interest in a rear scene and described j interest and in described transition probability matrix, transfer to the 5th product of the transition probability of i interest, calculate the 5th sum of products after described the 5th product is added, calculate again the 6th product that described the 5th sum of products and described i interest produce the emission probability of described the 1st scene in described emission matrix, probability using described the 6th product as described i interest in described the 1st scene, determine maximum the 6th product in described the 6th product, using the probability of described maximum the 6th product the 1st interest that scene is corresponding in described forecasting sequence.
8. an information push-delivery apparatus, is characterized in that, described device comprises:
The first acquisition module, for obtaining the sequence of scenes and interest sequence corresponding to described sequence of scenes that are successively switched to n current scene in the default time period from the 1st scene, n >=2;
The first computing module, be combined into scene set for described sequence of scenes that described the first acquisition module is obtained and the scene of preset model, the synthetic interest set of interest group in described interest sequence and the described model that described the first acquisition module is obtained, adds up to according to described scene set and described interest set the model parameter of calculating described model;
The second computing module, the described model parameter of calculating for the described sequence of scenes of obtaining according to described the first acquisition module, described interest set and described the first computing module is calculated described current current interest corresponding to n scene;
The first pushing module, for according to described second computing module calculate described current interest to terminal pushed information.
9. information push-delivery apparatus according to claim 8, is characterized in that, described model parameter comprises initial probability matrix, transition probability matrix and emission matrix, and described the first computing module comprises:
Acquiring unit, for setting in advance initial probability matrix or obtaining previous initial probability matrix;
The first computing unit, for i interest of the interest of the m for described interest set, calculates described i interest and transfers to the transition probability of j interest successively, obtains described transition probability matrix;
The second computing unit, for i the interest for described interest set, calculates described i interest and produces the emission probability of k scene successively, obtains described emission matrix;
Wherein, 1≤i≤m, 1≤j≤m and 1≤k≤n.
10. information push-delivery apparatus according to claim 9, is characterized in that, described the second computing module comprises:
The 3rd computing unit, for in the time calculating the probability of the 1st each interest of scene, the product of emission probability that i interest of the initial probability of i the interest that described acquiring unit is obtained in described initial probability matrix and described the second computing unit calculating produces described the 1st scene in described emission matrix is the probability in described the 1st scene as described i interest;
The 4th computing unit, for in the time calculating the probability of q each interest of scene, j interest first calculating successively probability in previous scene of j interest that described the 3rd computing unit calculates and described the first computing unit calculating is transferred to the first product of the transition probability of i interest in described transition probability matrix, determine maximum the first product in described the first product, calculate again the second product that the second computing unit calculates described in described maximum the first sum of products i interest produces the emission probability of described q scene in described emission matrix, using described the second product as described i interest the probability in described q scene, 2≤q≤n-1,
The 5th computing unit, for in the time calculating the probability of n each interest of scene, j interest first calculating successively probability in previous scene of j interest that described the 4th computing unit calculates and described the first computing unit calculating is transferred to the first product of the transition probability of i interest in described transition probability matrix, determine maximum the first product in described the first product, calculate again the second product that the second computing unit calculates described in described maximum the first sum of products i interest produces the emission probability of described n scene in described emission matrix, using described the second product as described i interest the probability in described n scene, determine maximum the second product in described the second product, using interest corresponding described maximum the second product as described current interest.
11. information push-delivery apparatus according to claim 8, is characterized in that, described device also comprises:
The second acquisition module, for obtaining at least one forecasting sequence, each forecasting sequence comprises that described sequence of scenes adds a prediction scene, described prediction scene is the scene in described scene set, and described prediction scene difference in each forecasting sequence;
The 3rd computing module, for each forecasting sequence obtaining for described the second acquisition module, calculates the probability of described forecasting sequence according to described forecasting sequence, described interest set and described model parameter;
The second pushing module, for forecasting sequence corresponding to probability maximum probability of determining that described the 3rd computing module calculates, and in the prediction scene of the corresponding forecasting sequence of described maximum probability to described terminal pushed information.
12. information push-delivery apparatus according to claim 11, is characterized in that, described the 3rd computing module comprises:
The 6th computing unit, predict the probability of prediction interest corresponding to scene for forecasting sequence described in forecasting sequence, described interest set and the described model parameter forward calculation obtained according to described the second acquisition module, the probability using the probability of described prediction interest as described forecasting sequence; Or,
The 7th computing unit, for the probability of the 1st interest that scene is corresponding of forecasting sequence described in forecasting sequence, described interest set and the described model parameter backwards calculation obtained according to described the second acquisition module, the probability using the probability of described interest as described forecasting sequence.
13. information push-delivery apparatus according to claim 12, it is characterized in that, described the 6th computing unit, for in the time calculating the probability of the 1st each interest of scene, the product of emission probability that i interest of the initial probability of i the interest that described acquiring unit is obtained in described initial probability matrix and described the second computing unit calculating produces described the 1st scene in described emission matrix is the probability in described the 1st scene as described i interest;
In the time calculating the probability of each interest in q scene, j interest first calculating successively probability in previous scene of j interest that described the 6th computing unit calculates and described the first computing unit calculating is transferred to the 3rd product of the transition probability of i interest in described transition probability matrix, calculate the 3rd sum of products after described the 3rd product is added, i interest calculating again described the 3rd sum of products and described the second computing unit calculating produces the 4th product of the emission probability of described q scene in described emission matrix, using described the 4th product as described i interest the probability in described q scene, 2≤q≤n-1,
In the time calculating the probability of each interest in n scene, j interest first calculating successively probability in previous scene of j interest that described the 6th computing unit calculates and described the first computing unit calculating is transferred to the 3rd product of the transition probability of i interest in described transition probability matrix, calculate the 3rd sum of products after described the 3rd product is added, i interest calculating again described the 3rd sum of products and described the second computing unit calculating produces the 4th product of the emission probability of described n scene in described emission matrix, using described the 4th product as described i interest the probability in a described n scene, determine maximum the 4th product in described the 4th product, described maximum the 4th product is predicted to the probability of prediction interest corresponding to scene in described forecasting sequence.
14. information push-delivery apparatus according to claim 12, it is characterized in that, described the 7th computing unit, for in the time calculating the probability of n each interest of scene, the product of emission probability that i interest of the initial probability of i the interest that described acquiring unit is obtained in described initial probability matrix and described the second computing unit calculating produces described n scene in described emission matrix is the probability in described n scene as described i interest;
In the time calculating the probability of each interest in q scene, j interest first calculating successively probability in a rear scene of j interest that described the 7th computing unit calculates and described the first computing unit calculating is transferred to the 5th product of the transition probability of i interest in described transition probability matrix, calculate the 5th sum of products after described the 5th product is added, i interest calculating again described the 5th sum of products and described the second computing unit calculating produces the 6th product of the emission probability of described q scene in described emission matrix, using described the 6th product as described i interest the probability in described q scene, 2≤q≤n-1,
In the time calculating the probability of each interest in the 1st scene, j interest first calculating successively probability in a rear scene of j interest that described the 7th computing unit calculates and described the first computing unit calculating is transferred to the 5th product of the transition probability of i interest in described transition probability matrix, calculate the 5th sum of products after described the 5th product is added, i interest calculating again described the 5th sum of products and described the second computing unit calculating produces the 6th product of the emission probability of described the 1st scene in described emission matrix, probability using described the 6th product as described i interest in described the 1st scene, determine maximum the 6th product in described the 6th product, using the probability of described maximum the 6th product the 1st interest that scene is corresponding in described forecasting sequence.
15. 1 kinds of terminals, is characterized in that, described terminal comprises the information push-delivery apparatus as described in any one in claim 8 to 14.
16. 1 kinds of servers, is characterized in that, described server comprises the information push-delivery apparatus as described in any one in claim 8 to 14.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210562450.0A CN103888498B (en) | 2012-12-21 | 2012-12-21 | Information-pushing method, device, terminal and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210562450.0A CN103888498B (en) | 2012-12-21 | 2012-12-21 | Information-pushing method, device, terminal and server |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103888498A true CN103888498A (en) | 2014-06-25 |
| CN103888498B CN103888498B (en) | 2018-04-13 |
Family
ID=50957216
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210562450.0A Active CN103888498B (en) | 2012-12-21 | 2012-12-21 | Information-pushing method, device, terminal and server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103888498B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104572857A (en) * | 2014-12-17 | 2015-04-29 | 深圳先进技术研究院 | Webpage pushing method and webpage pushing device |
| CN105373608A (en) * | 2015-11-13 | 2016-03-02 | 广州市久邦数码科技有限公司 | Input method based scene-mode content pushing method and system |
| CN113836438A (en) * | 2021-09-14 | 2021-12-24 | 上海任意门科技有限公司 | Method, electronic device, and storage medium for post recommendation |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106339505B (en) * | 2016-09-27 | 2019-09-27 | 电子科技大学 | A kind of music recommended method based on markov chain |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6839680B1 (en) * | 1999-09-30 | 2005-01-04 | Fujitsu Limited | Internet profiling |
| CN101079063A (en) * | 2007-06-25 | 2007-11-28 | 腾讯科技（深圳）有限公司 | Method, system and apparatus for transmitting advertisement based on scene information |
| CN101866341A (en) * | 2009-04-17 | 2010-10-20 | 华为技术有限公司 | Information push method, device and system |
| CN102130933A (en) * | 2010-01-13 | 2011-07-20 | 中国移动通信集团公司 | Recommending method, system and equipment based on mobile Internet |
| CN102681999A (en) * | 2011-03-08 | 2012-09-19 | 阿里巴巴集团控股有限公司 | Method and device for collecting and sending user action information |
-
2012
- 2012-12-21 CN CN201210562450.0A patent/CN103888498B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6839680B1 (en) * | 1999-09-30 | 2005-01-04 | Fujitsu Limited | Internet profiling |
| CN101079063A (en) * | 2007-06-25 | 2007-11-28 | 腾讯科技（深圳）有限公司 | Method, system and apparatus for transmitting advertisement based on scene information |
| CN101866341A (en) * | 2009-04-17 | 2010-10-20 | 华为技术有限公司 | Information push method, device and system |
| CN102130933A (en) * | 2010-01-13 | 2011-07-20 | 中国移动通信集团公司 | Recommending method, system and equipment based on mobile Internet |
| CN102681999A (en) * | 2011-03-08 | 2012-09-19 | 阿里巴巴集团控股有限公司 | Method and device for collecting and sending user action information |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104572857A (en) * | 2014-12-17 | 2015-04-29 | 深圳先进技术研究院 | Webpage pushing method and webpage pushing device |
| CN104572857B (en) * | 2014-12-17 | 2018-03-13 | 深圳先进技术研究院 | A kind of Web page push method and device |
| CN105373608A (en) * | 2015-11-13 | 2016-03-02 | 广州市久邦数码科技有限公司 | Input method based scene-mode content pushing method and system |
| CN113836438A (en) * | 2021-09-14 | 2021-12-24 | 上海任意门科技有限公司 | Method, electronic device, and storage medium for post recommendation |
| CN113836438B (en) * | 2021-09-14 | 2022-09-02 | 上海任意门科技有限公司 | Method, electronic device, and storage medium for post recommendation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103888498B (en) | 2018-04-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110245299B (en) | Sequence recommendation method and system based on dynamic interaction attention mechanism | |
| CN107515909B (en) | Video recommendation method and system | |
| CN105046515B (en) | Method and device for sorting advertisements | |
| CN103329151B (en) | Recommendation based on topic cluster | |
| CN113626719A (en) | Information recommendation method, device, equipment, storage medium and computer program product | |
| CN103500213B (en) | Page hot-spot resource updating method and device based on pre-reading | |
| CN104951518A (en) | Context recommending method based on dynamic incremental updating | |
| CN110197404A (en) | The personalized long-tail Method of Commodity Recommendation and system of popularity deviation can be reduced | |
| CN109993583A (en) | Information-pushing method and device, storage medium and electronic device | |
| CN110175895A (en) | A kind of item recommendation method and device | |
| CN103365842B (en) | A kind of page browsing recommends method and device | |
| CN103888498A (en) | Information pushing method and apparatus, terminal and server | |
| CN110413888A (en) | A kind of books recommended method and device | |
| US8954414B2 (en) | Search model updates | |
| CN111340522B (en) | Resource recommendation method, device, server and storage medium | |
| CN105183878A (en) | Music classification recommending method based on Markov prediction algorithm | |
| Ding et al. | Session-based recommendation with hypergraph convolutional networks and sequential information embeddings | |
| CN111915383A (en) | Window-based article cold start recommendation method and device | |
| CN112269943B (en) | Information recommendation system and method | |
| Ivancevic et al. | Complex dynamics: advanced system dynamics in complex variables | |
| CN113688306A (en) | Recommendation strategy generation method and device based on reinforcement learning | |
| Dong et al. | Improving sequential recommendation with attribute-augmented graph neural networks | |
| CN107615789A (en) | The notice orientation participated in based on downstream user | |
| CN105528247A (en) | Information processing method and client side | |
| CN103885948A (en) | Information showing method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |