CN106777067A - Information recommendation method and system - Google Patents
Information recommendation method and system Download PDFInfo
- Publication number
- CN106777067A CN106777067A CN201611140338.2A CN201611140338A CN106777067A CN 106777067 A CN106777067 A CN 106777067A CN 201611140338 A CN201611140338 A CN 201611140338A CN 106777067 A CN106777067 A CN 106777067A
- Authority
- CN
- China
- Prior art keywords
- scene
- user behavior
- recommendation
- information
- scenario simulation
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides information recommendation method and system, including:The multiple sampled data streams read according to prefixed time interval cutting are cut into slices with obtaining multiple sampled datas;User behavior scene is defined, the user behavior scene includes at least one and/or at least one class user behavior event;The part for meeting user behavior scene is selected out to constitute scene snapshot from the data of each sampled data section respectively;Using scene snapshot as scenario simulation model input, so as to obtain scenario simulation model output scene portrait;Scene portrait is matched with default recommendation results, if the match is successful, default recommendation results is defined as recommendation information.The present invention realizes the dynamically interactive type personalized recommendation of stream data reading, realizes reading in sampled data, instant personalized recommendation immediately based on user behavior data, while ensureing ageing, the individual individual demand of different user can be taken into account again.
Description
Technical field
The present invention relates to information recommendation field, more particularly to information recommendation method and system.
Background technology
The streamed information services that social coverage rate is extremely wide, online any active ues quantity is huge are referred to as public character streaming information clothes
Business, the Typical Representative of public character streamed information services is all kinds of medium living broadcast services.The recommendation face of public character streamed information services
Face problems with:1) must recommend within the online cycle of content, so as to avoid offline commending contents to user;2) must be by
Commending contents are individual to suitable user, so as to avoid that other users colony is interfered.
It can be seen that, for the recommendation of public character streamed information services, it is desirable to be able to while ensureing ageing and taking into account different use
The new technology of the individual individual demand in family.
The content of the invention
The shortcoming of prior art, it is an object of the invention to provide information recommendation method and system, uses in view of the above
Cannot ensure ageing in the recommendation for solving public character streamed information services in the prior art, and different user cannot be taken into account
The problems such as individual demand of body.
In order to achieve the above objects and other related objects, the present invention provides a kind of information recommendation method, including:According to default
Multiple sampled data streams that time interval cutting is read in are cut into slices with obtaining multiple sampled datas;Define one or more user behaviors
Scene, described each user behavior scene includes at least one and/or at least one class user behavior event;Respectively from described in each
The part for meeting the user behavior scene is selected out to constitute scene snapshot in the data of sampled data section;By the scene
Snapshot as scenario simulation model input, so as to obtain the scene portrait of scenario simulation model output;By the scene
Portrait is matched with default recommendation results, if the match is successful, the default recommendation results is defined as into recommendation information.
In one embodiment of the invention, when one or more user behavior scenes are defined, one or more of use
Family behavior scene constituting action pattern, certain the user behavior scene in the behavior pattern scene portrait be by it is described certain
The scene of the scene snapshot of user behavior scene and previous user behavior scene is drawn a portrait collectively as the scenario simulation model
Input, and output through the scenario simulation model obtains, and methods described also includes:To finally give scene portrait with
The default recommendation results are matched, if the match is successful, the default recommendation results are defined as into recommendation information.
In one embodiment of the invention, the scenario simulation model includes multiple, and each is used to be directed to a kind of characteristic information
Carry out scenario simulation.
In one embodiment of the invention, the scenario simulation model includes:Supervised learning model, and/or unsupervised
Practise model.
In one embodiment of the invention, the unsupervised learning model includes:Deep learning model.
In one embodiment of the invention, the deep learning model is set up according to back-propagation algorithm.
In one embodiment of the invention, each described sampled data stream includes a class user behavior event data, described many
Individual sampled data section includes the portion with the one-to-one multiple user behavior events of the multiple sampled data stream difference
Divided data.
In one embodiment of the invention, user behavior event data includes multiple user behavior event numbers described in per class
According to each described user behavior event data includes:Timestamp, the user for identifying the object for producing corresponding behavior event
The sampled data of mark and corresponding behavior event.
In one embodiment of the invention, before the scene snapshot is input into the scenario simulation model, methods described
Also include:Each user behavior event data that the scene snapshot is included is sorted sequentially in time, and is sequentially input into institute
State scenario simulation model.
In one embodiment of the invention, the default recommendation results include multiple, and the matching includes:Calculate respectively described
The degree of correlation that scene is drawn a portrait with each default recommendation results, the maximum default recommendation results of the degree of correlation are defined as pushing away
Recommend information.
In order to achieve the above objects and other related objects, the present invention provides a kind of information recommendation system, including:Data slicer
Module, the multiple sampled data streams for being read according to prefixed time interval cutting are cut into slices with obtaining multiple sampled datas;Scene
Definition module, for defining one or more user behavior scenes, each user behavior scene includes at least one and/or at least
One class user behavior event;Snapshot selects module, for selecting out symbol in the data of sampled data section from each described respectively
The part of the user behavior scene is closed to constitute scene snapshot;Scenario simulation module, for using the scene snapshot as field
The input of scape simulation model, so as to obtain the scene portrait of the scenario simulation model output;Matching module, for by the field
Scape portrait is matched with default recommendation results, if the match is successful, the default recommendation results is defined as into recommendation information.
In one embodiment of the invention, when one or more user behavior scenes are defined, one or more of use
Family behavior scene constituting action pattern, certain the user behavior scene in the behavior pattern scene portrait be by it is described certain
The scene of the scene snapshot of user behavior scene and previous user behavior scene is drawn a portrait collectively as the scenario simulation model
Input, and output through the scenario simulation model obtains, and the matching module is additionally operable to:The scene that will be finally given is drawn
As being matched with the default recommendation results, if the match is successful, the default recommendation results are defined as recommendation information.
In one embodiment of the invention, the scenario simulation model includes multiple, and each is used to be directed to a kind of characteristic information
Carry out scenario simulation.
In one embodiment of the invention, the scenario simulation model includes:Supervised learning model, and/or unsupervised
Practise model.
In one embodiment of the invention, the unsupervised learning model includes:Deep learning model.
In one embodiment of the invention, the deep learning model is set up according to back-propagation algorithm.
In one embodiment of the invention, each described sampled data stream includes a class user behavior event data, described many
Individual sampled data section includes the portion with the one-to-one multiple user behavior events of the multiple sampled data stream difference
Divided data.
In one embodiment of the invention, user behavior event data includes multiple user behavior event numbers described in per class
According to each described user behavior event data includes:Timestamp, the user for identifying the object for producing corresponding behavior event
The sampled data of mark and corresponding behavior event.
In one embodiment of the invention, the system also includes:Order module, for the scene snapshot to be input into institute
Before stating scenario simulation model, each user behavior event data that the scene snapshot is included is sorted sequentially in time,
And sequentially it is input into the scenario simulation model.
In one embodiment of the invention, the default recommendation results include multiple, and the matching includes:Calculate respectively described
The degree of correlation that scene is drawn a portrait with each default recommendation results, the maximum default recommendation results of the degree of correlation are defined as pushing away
Recommend information.
As described above, information recommendation method of the invention and system, are realized to be read in immediately based on user behavior data and adopted
The technical scheme of sample data, instant personalized recommendation, has the advantages that:
1) time cycle that recommendation results update is short, is calculated with the chronomere of minute or second or smaller;
2) after reading in sampled data, influence of the sampled data to recommendation results in a few minutes/display afterwards for several seconds,
A kind of interactive recommendation of " behavior sampling-recommend feedback " can be formed;
3) recommendation results are sampled based on user behavior data, are the probability results of user's individual behavior successive changes, are had
Individuality.
Brief description of the drawings
Fig. 1 is shown as the information recommendation method flow chart of one embodiment of the invention.
Fig. 2 is shown as the cutting sampled data stream of one embodiment of the invention to generate the schematic diagram that sampled data is cut into slices.
Fig. 3 is shown as the inside composition schematic diagram of the A class behavior sampled data streams of one embodiment of the invention.
Fig. 4 is shown as the schematic diagram of the user behavior scene of the definition of one embodiment of the invention.
Fig. 5 A~5B is shown as the snapshot selection process schematic of one embodiment of the invention.
Fig. 6 A~6B is shown as the scenario simulation process schematic of one embodiment of the invention.
Multiple user behavior scenes that Fig. 7 is shown as one embodiment of the invention connect to form the simulation process of behavior pattern and show
It is intended to.
Fig. 8 is shown as the process schematic of the consequently recommended result of selection of one embodiment of the invention.
Fig. 9 is shown as the information recommendation system module map of one embodiment of the invention.
Specific embodiment
Embodiments of the present invention are illustrated below by way of specific instantiation, those skilled in the art can be by this specification
Disclosed content understands other advantages of the invention and effect easily.The present invention can also be by specific realities different in addition
The mode of applying is embodied or practiced, the various details in this specification can also based on different viewpoints with application, without departing from
Various modifications or alterations are carried out under spirit of the invention.It should be noted that, in the case where not conflicting, following examples and implementation
Feature in example can be mutually combined.
It should be noted that the diagram provided in following examples only illustrates basic structure of the invention in a schematic way
Think, component count, shape and size when only display is with relevant component in the present invention rather than according to actual implementation in schema then
Draw, it is actual when the implementing kenel of each component, quantity and ratio can be a kind of random change, and its assembly layout kenel
It is likely more complexity.
With the arrival in big data epoch, following behavior of user is predicted by the historical behavior of digging user, set up
The relation of user and content is played, the accuracy of information recommendation could be improved.Fig. 1 is referred to, the present invention provides a kind of information recommendation
Method, on the basis of guarantee is ageing, takes into account the individual demand of different user, realizes the dynamic friendship of stream data reading
Mutual formula personalized recommendation, specifically includes following steps:
Step S101:The multiple sampled data streams read according to prefixed time interval cutting are cut with obtaining multiple sampled datas
Piece, Fig. 2 shows the process once cut, and the time point of cutting is recorded as time point 1, and each described sampled data stream includes one
Class user behavior event data, the section of sampled data that each cuts out include in the user behavior event data of correspondence classification and
The corresponding part of prefixed time interval.As shown in figure 3, each sampled data stream is made up of multiple user behavior event datas
, wherein, each user behavior event data includes:Record the timestamp of the data generation time, identify data generation
The ID of source objects, and the data body portion.
It should be noted that in each sampled data stream, each user behavior event data can be it is unordered, with
When prefixed time interval cuts, the user behavior event data corresponding to the timestamp in the range of prefixed time interval is split
Out, so as to generate the sampled data section of such user behavior event.After repeatedly cutting, the priority according to clipping time is suitable
Be ranked up for each batch of sampled data section by sequence.
For example, user behavior event is:Spectators watch certain channel, and corresponding sampled data stream is:All online spectators see
The data of the channel are seen, these data will be obtained in the form of data flow in real time, corresponding sampled data section is:In the time
After carrying out multiple slicing treatment to these data on axle, according to the section that clipping time sorts between a group of acquisition.
Step S102:One or more user behavior scenes are defined, as shown in figure 4, the user behavior scene of each definition
Can include one, multiple, a class, catergories of user behavior event.
For example:A class user behavior events are:Spectators watch certain channel, and B class user behavior events are:Spectators are switched to certain
Channel, C class user behavior events are:Channel plays certain program, and when the scene a of definition includes the class of A, C two, then scene a is:Spectators
Certain channel program is watched, when the scene b of definition includes the class of B, C two, then scene b is:Spectators are switched to certain channel program.
Step S103:For cutting every time, selected out from the data of sampled data section each described meet institute respectively
The part of user behavior scene is stated to constitute scene snapshot corresponding with each clipping time point.
For example, for Fig. 5 A, the section of A class user behavior events sampled data is:All online spectators are in different time sections
Watch the data of certain channel, after carrying out multiple slicing treatment to these data on a timeline, between a group of acquisition according to
The section of time-sequencing.C class user behavior events sampled data is cut into slices:All direct broadcast bands play program within certain time
Playbill data, after carrying out 3 slicing treatments to these data on a timeline, between a group of acquisition according to cutting when
Between sort section.The scene a snapshots 1 selected out are:Time point 1, spectators watch the sampled data composition of certain channel program
Snapshot, the scene a snapshots 2 selected out are:Time point 2, spectators watch the snapshot of the sampled data composition of certain channel program,
The scene a snapshots 3 selected out are:Time point 3, spectators watch the snapshot of the sampled data composition of certain channel program.For figure
5B, B class user behavior event sampled data are cut into slices:All online spectators are switched to the data of certain channel in different time sections,
After carrying out 3 slicing treatments to these data on a timeline, according to the section of time-sequencing between a group of acquisition.C classes
User behavior event sampled data is cut into slices:All direct broadcast bands play the playbill data of program within certain time, when
After carrying out multiple slicing treatment to these data on countershaft, according to the section of time-sequencing between a group of acquisition.Select out
Scene b snapshots 1 be:Time point 1, the snapshot of the sampled data composition that certain channel program audience is switched in is selected out
Scene b snapshots 2 are:Time point 2, the snapshot of the sampled data composition that certain channel program audience is switched in, the field selected out
Scape b snapshots 3 are:Time point 3, the snapshot of the sampled data composition that certain channel program audience is switched in.
Step S104:According to clipping time order, respectively using each scene snapshot as scenario simulation model input,
So as to obtain each scene portrait of the scenario simulation model output.It should be noted that scenario simulation model is customizing model,
Can be one, or multiple, each model is used to carry out scenario simulation for a kind of characteristic information.The life of customizing model
The need for according to portrait, by all kinds of deep learning algorithms that select (such as:Back-propagation algorithm) and/or unsupervised-learning algorithm
It is formed by combining.
For example, for Fig. 6 A, simulation model simulates the general of this feature for the special characteristic contained in the snapshot of scene a
Rate is drawn a portrait.Specifically, scene a portraits 1 are:The probability of this feature simulated for the special characteristic contained in scene a snapshots 1
Portrait, i.e. time point 1 and its neighbouring time domain, audience may watch the probabilistic forecasting of certain channel program, scene a portraits 1 with
Scene a snapshots 1 are compared and adjusting parameter so that the difference of the two reduces;Scene a portraits 2 are:For in scene a snapshots 2
The probability portrait of this feature that the special characteristic for containing is simulated, i.e. time point 2 and its neighbouring time domain, audience may watch certain
The probabilistic forecasting of channel program, scene a portraits 2 are compared and adjusting parameter with scene a snapshots 2 so that the difference of the two
Reduce;Scene a portraits 3 are:The probability portrait of this feature simulated for the special characteristic contained in scene a snapshots 3, immediately
Between point 3 and its neighbouring time domain, audience may watch the probabilistic forecasting of certain channel program, scene a portraits 3 and scene a snapshots 3
Compare and adjusting parameter so that the difference of the two reduces.By that analogy, scene a portraits n is:For in scene a snapshots n
The probability portrait of this feature that the special characteristic for containing is simulated, i.e. time point n and its neighbouring time domain, audience may watch certain
The probabilistic forecasting of channel program, scene a portrait n and scene a snapshots n compare and adjusting parameter so that the difference of the two
Reduce.
Again for example, for Fig. 6 B, simulation model simulates the general of this feature to the special characteristic contained in the snapshot of scene b
Rate is drawn a portrait.Specifically, scene b portraits 1 are:The probability of this feature simulated for the special characteristic contained in scene b snapshots 1
Portrait, i.e. time point 1 and its neighbouring time domain, audience may be switched to the probabilistic forecasting of certain channel program, scene b portraits 1
Compare with scene b snapshots 1 and adjusting parameter so that the difference of the two reduces;Scene b portraits 2 are:For scene b snapshots 2
In the probability portrait of this feature that simulates of the special characteristic that contains, i.e. time point 2 and its neighbouring time domain, audience may switch
To the probabilistic forecasting of certain channel program, scene b portraits 1 and scene b snapshots 1 are compared and adjusting parameter so that the two
Difference reduces;Scene b portraits 3 are:The probability portrait of this feature simulated for the special characteristic contained in scene b snapshots 3,
That is time point 3 and its neighbouring time domain, audience may be switched to the probabilistic forecasting of certain channel program, and scene b draws a portrait 1 and scene
B snapshots 1 are compared and adjusting parameter so that the difference of the two reduces.By that analogy, scene b portraits n is:It is fast for scene b
The probability portrait of this feature simulated according to the special characteristic contained in n, i.e. time point n and its neighbouring time domain, audience may
The probabilistic forecasting of certain channel program is watched, scene b portrait n and scene b snapshots n compare and adjusting parameter so that the two
Difference reduce.
It should be noted that in another embodiment, each scene snapshot is input into corresponding scenario simulation carries out model
Before, according to the timestamp of user behavior event, each user behavior event data that will first be included in scene snapshot according to when
Between sequencing sequence, then be sequentially input into corresponding scenario simulation model.
Particularly, when one or more user behavior scenes are defined, one or more of user behavior scene structures
Into behavior pattern, specifically, an independent user behavior scene may be constructed behavior pattern, multiple user behavior scene compositions
Getting up (such as in the form of connecting) can also constituting action pattern.Fig. 7 shows multiple user behavior scene rows in series
It is the situation of pattern, wherein, the scene portrait of certain the user behavior scene in the behavior pattern is by described certain user
The scene of the scene snapshot of behavior scene and previous user behavior scene is drawn a portrait collectively as the defeated of the scenario simulation model
Enter, and output through the scenario simulation model is obtained.
For example, considering 2 factors simultaneously:1) in time point 1 and its neighbouring time domain, audience may watch certain channel section
Purpose probability, from scene a portraits 1;2) at time point 1, spectators are switched into the actual samples data of certain channel program, come from
Scene b snapshots 1.Simulation:Time point 1 and its neighbouring time domain, audience may be switched to the probabilistic forecasting of certain channel program.
Step S105:Scene portrait is matched with default recommendation results, if the match is successful, will be described default
Recommendation results are defined as recommendation information.Optionally, the default recommendation results can include multiple, and the matching is specifically included:
The degree of correlation of each scene portrait and each default recommendation results is calculated respectively, by the default recommendation results that the degree of correlation is maximum
It is defined as final recommendation information.
For example shown in Fig. 8, the scene b of the behavior pattern simulation output being made up of all of scenario simulation portrait as
Final recommendation results, then, it is possible to certain channel may be switched to according to audience in certain time point and its neighbouring time domain
The probability of program, to do recommendation of the channels to spectators.
Fig. 9 is referred to, with above method embodiment principle similarly, the present invention provides a kind of information recommendation system 900,
Realized as a kind of software, can be equipped on and be performed with input, output, on the electronic equipment of data processing function.By
Each technical characteristic in previous embodiment can apply to the system embodiment, thus it is no longer repeated.
System 900 includes:Data slicer module 901, scene definition module 902, snapshot selection module 903, scenario simulation
Module 904 and matching module 905.Specifically:Data slicer module 901 is adopted according to the multiple that prefixed time interval cutting is read in
Sample data flow is cut into slices with obtaining multiple sampled datas;Scene definition module 902 defines one or more user behavior scenes, each
User behavior scene includes at least one and/or at least one class user behavior event;Snapshot selects module 903 respectively from each institute
The part for meeting the user behavior scene is selected out to constitute scene snapshot in the data for stating sampled data section;Scenario simulation
Module 904 using the scene snapshot as scenario simulation model input, so as to obtain the field of scenario simulation model output
Scape is drawn a portrait;Matching module 905 is matched scene portrait with default recommendation results, if the match is successful, will be described pre-
If recommendation results are defined as recommendation information.
In one embodiment, the system also includes:Order module, for the scene snapshot to be input into the scene
Before simulation model, each user behavior event data that the scene snapshot is included is sorted sequentially in time, and sequentially
It is input into the scenario simulation model.
In sum, the present invention effectively overcomes various shortcoming of the prior art and has high industrial utilization.
The above-described embodiments merely illustrate the principles and effects of the present invention, not for the limitation present invention.It is any ripe
The personage for knowing this technology all can carry out modifications and changes under without prejudice to spirit and scope of the invention to above-described embodiment.Cause
This, those of ordinary skill in the art is complete with institute under technological thought without departing from disclosed spirit such as
Into all equivalent modifications or change, should be covered by claim of the invention.
Claims (20)
1. a kind of information recommendation method, it is characterised in that including:
The multiple sampled data streams read according to prefixed time interval cutting are cut into slices with obtaining multiple sampled datas;
One or more user behavior scenes are defined, each user behavior scene includes at least one and/or at least one class user
Behavior event;
The part for meeting the user behavior scene is selected out to constitute in the data of sampled data section from each described respectively
Scene snapshot;
Using the scene snapshot as scenario simulation model input, so as to the scene for obtaining scenario simulation model output is drawn
Picture;
Scene portrait is matched with default recommendation results, if the match is successful, the default recommendation results is determined
It is recommendation information.
2. information recommendation method according to claim 1, it is characterised in that when defining one or more user behaviors
Jing Shi, one or more of user behavior scene constituting action patterns, certain the user behavior scene in the behavior pattern
Scene portrait be by certain user behavior scene scene snapshot and previous user behavior scene scene portrait altogether
With the input as the scenario simulation model, and output through the scenario simulation model is obtained, and methods described also includes:
The scene portrait that will be finally given is matched with the default recommendation results, if the match is successful, by the default recommendation knot
Fruit is defined as recommendation information.
3. information recommendation method according to claim 1, it is characterised in that the scenario simulation model includes multiple, often
It is individual for carrying out scenario simulation for a kind of characteristic information.
4. information recommendation method according to claim 1, it is characterised in that the scenario simulation model includes:There is supervision
Learning model, and/or unsupervised learning model.
5. information recommendation method according to claim 4, it is characterised in that the unsupervised learning model includes:Depth
Learning model.
6. information recommendation method according to claim 5, it is characterised in that the deep learning model is according to reversely biography
Broadcast algorithm foundation.
7. information recommendation method according to claim 1, it is characterised in that each described sampled data stream includes that a class is used
Family behavior event data, the multiple sampled data section includes one-to-one multiple respectively with the multiple sampled data stream
The partial data of the user behavior event.
8. information recommendation method according to claim 7, it is characterised in that user behavior event data described in per class includes
Multiple user behavior event datas, each described user behavior event data includes:Timestamp, for identify produce it is corresponding
Behavior event object ID and the sampled data of corresponding behavior event.
9. information recommendation method according to claim 8, it is characterised in that the scene snapshot is input into the scene
Before simulation model, also include:Each user behavior event data that the scene snapshot is included is sorted sequentially in time,
And sequentially it is input into the scenario simulation model.
10. information recommendation method according to claim 1, it is characterised in that the default recommendation results include multiple, institute
Stating matching includes:The degree of correlation of the scene portrait and each default recommendation results is calculated respectively, by the degree of correlation most
Big default recommendation results are defined as recommendation information.
A kind of 11. information recommendation systems, it is characterised in that including:
Data slicer module, for cutting the multiple sampled data streams for reading in obtain multiple hits according to prefixed time interval
According to section;
Scene definition module, for defining one or more user behavior scenes, each user behavior scene includes at least one
And/or an at least class user behavior event;
Snapshot selects module, and the user behavior is met for being selected out from the data of sampled data section each described respectively
The part of scene is constituting scene snapshot;
Scenario simulation module, for using the scene snapshot as scenario simulation model input, so as to obtain the scene mould
The scene portrait of analog model output;
Matching module, for scene portrait to be matched with default recommendation results, if the match is successful, will be described default
Recommendation results are defined as recommendation information.
12. information recommendation systems according to claim 11, it is characterised in that when defining one or more user behaviors
During scene, one or more of user behavior scene constituting action patterns, certain user behavior in the behavior pattern
The scene portrait of scape is drawn a portrait by the scene snapshot of certain user behavior scene and the scene of previous user behavior scene
Collectively as the input of the scenario simulation model, and output through the scenario simulation model is obtained, the matching module
It is additionally operable to:The scene portrait that will be finally given is matched with the default recommendation results, if the match is successful, will be described default
Recommendation results are defined as recommendation information.
13. information recommendation systems according to claim 11, it is characterised in that the scenario simulation model includes multiple,
Each is used to carry out scenario simulation for a kind of characteristic information.
14. information recommendation systems according to claim 11, it is characterised in that the scenario simulation model includes:There is prison
Superintend and direct learning model, and/or unsupervised learning model.
15. information recommendation systems according to claim 14, it is characterised in that the unsupervised learning model includes:It is deep
Degree learning model.
16. information recommendation systems according to claim 15, it is characterised in that the deep learning model is according to reversely
What propagation algorithm was set up.
17. information recommendation systems according to claim 11, it is characterised in that each described sampled data stream includes a class
User behavior event data, the multiple sampled data section includes one-to-one many respectively with the multiple sampled data stream
The partial data of the individual user behavior event.
18. information recommendation systems according to claim 17, it is characterised in that user behavior event packet described in per class
Multiple user behavior event datas are included, each described user behavior event data includes:Timestamp, for identify produce phase
The sampled data of the ID of the object of the behavior event answered and corresponding behavior event.
19. information recommendation systems according to claim 18, it is characterised in that also include:Order module, for by institute
Before stating the scene snapshot input scenario simulation model, each user behavior event data that the scene snapshot is included is pressed
According to time sequencing sequence, and sequentially it is input into the scenario simulation model.
20. information recommendation systems according to claim 11, it is characterised in that the default recommendation results include multiple,
The matching includes:The degree of correlation of the scene portrait and each default recommendation results is calculated respectively, by the degree of correlation
Maximum default recommendation results are defined as recommendation information.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2016110308323 | 2016-11-16 | ||
CN201611030832 | 2016-11-16 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106777067A true CN106777067A (en) | 2017-05-31 |
Family
ID=58876062
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611140338.2A Pending CN106777067A (en) | 2016-11-16 | 2016-12-12 | Information recommendation method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106777067A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109165347A (en) * | 2018-08-20 | 2019-01-08 | 腾讯科技(深圳)有限公司 | Data push method and device, storage medium and electronic device |
CN109670106A (en) * | 2018-12-06 | 2019-04-23 | 百度在线网络技术(北京)有限公司 | Things recommended method and device based on scene |
CN111797309A (en) * | 2020-06-19 | 2020-10-20 | 一汽奔腾轿车有限公司 | Vehicle-mounted intelligent recommendation device and method based on travel mode |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102841922A (en) * | 2012-07-04 | 2012-12-26 | 北京国双科技有限公司 | Data collection method and data collection device |
US20130325858A1 (en) * | 2012-03-07 | 2013-12-05 | Ut-Battelle, Llc | Personalized professional content recommendation |
CN104036002A (en) * | 2014-06-16 | 2014-09-10 | 深圳市英威诺科技有限公司 | Technical method for intelligently recommending data |
CN104731861A (en) * | 2015-02-05 | 2015-06-24 | 腾讯科技(深圳)有限公司 | Method and device for pushing multimedia data |
-
2016
- 2016-12-12 CN CN201611140338.2A patent/CN106777067A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130325858A1 (en) * | 2012-03-07 | 2013-12-05 | Ut-Battelle, Llc | Personalized professional content recommendation |
CN102841922A (en) * | 2012-07-04 | 2012-12-26 | 北京国双科技有限公司 | Data collection method and data collection device |
CN104036002A (en) * | 2014-06-16 | 2014-09-10 | 深圳市英威诺科技有限公司 | Technical method for intelligently recommending data |
CN104731861A (en) * | 2015-02-05 | 2015-06-24 | 腾讯科技(深圳)有限公司 | Method and device for pushing multimedia data |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109165347A (en) * | 2018-08-20 | 2019-01-08 | 腾讯科技(深圳)有限公司 | Data push method and device, storage medium and electronic device |
CN109165347B (en) * | 2018-08-20 | 2021-03-26 | 腾讯科技(深圳)有限公司 | Data pushing method and device, storage medium and electronic device |
CN109670106A (en) * | 2018-12-06 | 2019-04-23 | 百度在线网络技术(北京)有限公司 | Things recommended method and device based on scene |
CN111797309A (en) * | 2020-06-19 | 2020-10-20 | 一汽奔腾轿车有限公司 | Vehicle-mounted intelligent recommendation device and method based on travel mode |
CN111797309B (en) * | 2020-06-19 | 2024-04-16 | 一汽奔腾轿车有限公司 | Vehicle-mounted intelligent recommendation device and method based on travel mode |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103718166B (en) | Messaging device, information processing method | |
CN107566907A (en) | video clipping method, device, storage medium and terminal | |
Ejaz et al. | Feature aggregation based visual attention model for video summarization | |
CN106658200A (en) | Live video sharing and obtaining methods and devices, and terminal equipment thereof | |
CN105744292A (en) | Video data processing method and device | |
US20140207797A1 (en) | Methods, systems, and media for generating sentimental information associated with media content | |
CN106777067A (en) | Information recommendation method and system | |
CN107172452A (en) | Direct broadcasting room recommends method and device | |
CN106202269A (en) | A kind of obtain the method for augmented reality Operating Guideline, device and mobile terminal | |
CN106162223A (en) | A kind of news video cutting method and device | |
CN105872786B (en) | A kind of method and device for launching advertisement by barrage in a program | |
US20150199995A1 (en) | Modular content generation, modification, and delivery system | |
Lin et al. | The quest for omnioculars: Embedded visualization for augmenting basketball game viewing experiences | |
CN111277910A (en) | Bullet screen display method and device, electronic equipment and storage medium | |
CN104809218B (en) | A kind of UGC video classification methods and device | |
CN109996122A (en) | A kind of video recommendation method, device, server and storage medium | |
CN109325516A (en) | A kind of integrated learning approach and device towards image classification | |
Maity et al. | Opinion formation in time-varying social networks: The case of the naming game | |
Coutrot et al. | An efficient audiovisual saliency model to predict eye positions when looking at conversations | |
CN114339362A (en) | Video bullet screen matching method and device, computer equipment and storage medium | |
CN112218159A (en) | Multimedia information playing method and device, storage medium and electronic device | |
CN115222858A (en) | Method and equipment for training animation reconstruction network and image reconstruction and video reconstruction thereof | |
CN109670623A (en) | Neural net prediction method and device | |
CN108428234A (en) | Interactive Segmentation performance optimization method based on image segmentation result evaluation | |
CN108960130A (en) | Video file intelligent processing method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170531 |