CN107894998B - Video recommendation method and device - Google Patents
Video recommendation method and device Download PDFInfo
- Publication number
- CN107894998B CN107894998B CN201710997564.0A CN201710997564A CN107894998B CN 107894998 B CN107894998 B CN 107894998B CN 201710997564 A CN201710997564 A CN 201710997564A CN 107894998 B CN107894998 B CN 107894998B
- Authority
- CN
- China
- Prior art keywords
- user
- video
- cluster
- clustering
- influence power
- 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.)
- Active
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
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/71—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides a kind of video recommendation method, collects multiple user behavior datas in the given time, establishes user behavior data sample；Store the user behavior data sample；It is clustered using video content of the clustering algorithm to the user behavior data sample, obtains the similar user's cluster of at least one interest；Count the video that the similar user's cluster user of the interest was watched；It extracts each frame image of the video and extracts feature；It is clustered using feature of the clustering algorithm to each frame image of the video, obtains at least one Video clustering；The interactive relation in user's cluster described in the Video clustering between each user is analyzed, the influence power of user's cluster is obtained according to the analysis result；Recommended according to the ranking of the influence power of user's cluster.A kind of video recommendations device is also provided.
Description
Technical field
The present invention relates to information technology field more particularly to a kind of video recommendation methods and device.
Background technique
With the rapid development of Web2.0 technology and the rapid proliferation of online social networks, online social networks has become
The important carrier that people maintain social relationships, propagate information and innovative knowledge.The universal generation for allowing information of mobile Internet and
Publication becomes very convenient, and user can sharing video frequency anywhere or anytime.A video platform not still content share platform,
It is simultaneously also a social platform.User can share oneself shooting or interested video on platform, while can also be with
It is interacted with the other users on platform.Therefore, in conjunction with the interactive relationship of video content and user building recommender system be by
To the research direction of industry extensive concern.
In the prior art it is usually that user actively selects the video classification liked, is then pushed away according to the record of user's selection
Recommend the video of the same category；It is identical come the video recommended with user watched there are also being exactly viewing historical record according to user
Other videos of classification.In today of social networks prosperity, user but is predicted without the social spy of full utilization social networks
The potential video classification that may like.Therefore existing video recommendations accuracy is not high, poor user experience.
Summary of the invention
In consideration of it, the embodiment of the present invention provides a kind of video recommendation method and device, the video recommendation method and device
The people that there is high-impact by finding out a part, and suitable video priority is recommended them, wait them to have on video
After some interactions, video recommendations are given their good friend by we again.In the case where given recommendation total quantity, pass through this dispensing
Mode, we will obtain more users to the comment of video or thumb up, and allow video to leave impression more profound to user, are promoted
User experience and frequency of interaction improve user's retention ratio.
A kind of video recommendation method provided in an embodiment of the present invention, comprising the following steps: collect in the given time multiple
User behavior data establishes user behavior data sample；Store the user behavior data sample；Using clustering algorithm to described
The video content of user behavior data sample is clustered, and the similar user's cluster of at least one interest is obtained；Count described emerging
The video that the similar user's cluster user of interest was watched；It extracts each frame image of the video and extracts feature；It is calculated using cluster
Method clusters the feature of each frame image of the video, obtains at least one Video clustering；It analyzes in the Video clustering
Interactive relation in user's cluster between each user obtains the influence power of user's cluster according to the analysis result；
Recommended according to the ranking of the influence power of user's cluster.
Optionally, the clustering algorithm is k-means algorithm.
Optionally, extracting each frame image of the video and extract feature can be by scale invariant feature transfer algorithm
(Scale-invariant feature transform, SIFT), gray level co-occurrence matrixes method, Fourier power spectrum method and depth
One of learning model method is realized.
Optionally, the interactive relation in user's cluster described in the Video clustering between each user is analyzed specifically: root
According between user interactive relation constitute a huge network G, V={ u1, u2 ..., un }, n be user's number, according to
Interactive information of the family on video calculates influence power transfer matrix W ∈ Rn×n:
Wherein, CijComment and/or like time by user j to the user i video issued are commented for user j in user i
By or continue the number commented on or thumbed up, k ∈ N (u after having thumbed up some videoj) paid close attention to for all neighbours good friends of user j
Good friend.α, β are respectively the weights of importance value of different interactions, and W (i, j) describes user i to the influence power of user j.
Optionally, the influence power of user's cluster is obtained according to the analysis result specifically: It+1=bWIt+(1-b)
I0Wherein, It∈R1×nFor the influence value of all users of t moment；When t=0, I0Each element value be equal to 1/n；B is adjustable
Save parameter.
Optionally, the influence power of user's cluster is obtained according to the analysis result specifically: judge ItAnd It+1Difference
Whether do not meet the condition of convergence, exports I if meetingt。
Optionally, the condition of convergence is the threshold value or threshold range of setting.
A kind of video recommendations device provided in an embodiment of the present invention, the video recommendations device includes acquiring unit, is used for
Multiple user behavior datas are collected in the given time；Storage unit, for storing the user behavior data sample；First is poly-
Class unit obtains at least one for clustering using clustering algorithm to the video content of the user behavior data sample
The similar user's cluster of interest；Statistic unit, the video watched for counting the similar user's cluster user of the interest；Figure
As unit, for extracting each frame image of the video and extracting feature；Second cluster cell, using clustering algorithm to the view
Frequently the feature of each frame image is clustered, and obtains at least one Video clustering；Analytical unit, for analyzing the Video clustering
Described in interactive relation in user's cluster between each user, the influence of user's cluster is obtained according to the analysis result
Power；Recommendation unit, the ranking for the influence power according to user's cluster are recommended.
Optionally, the clustering algorithm that first cluster cell and the second cluster cell use is k-means algorithm.
Optionally, described image unit extracts each frame image of the video and extracts feature, can pass through Scale invariant spy
Levy transfer algorithm (Scale-invariant feature transform, SIFT), gray level co-occurrence matrixes method, Fourier's power
One of spectrometry and deep learning model method are realized.
The embodiment of the present invention provides a kind of video recommendation method and device, by according to user behavior and video content cluster knot
Fruit and user force result carry out video recommendations, and the video of high quality that is meeting user interest can be found to user, is mentioned
User experience, user's stickiness, user's retention ratio and the liveness of high video product.
Detailed description of the invention
Fig. 1 is the flow diagram of the video recommendation method of first embodiment of the invention；
Fig. 2 is the application schematic diagram of the video recommendation method of one embodiment of the invention；
Fig. 3 is the user interaction schematic diagram of the video recommendation method of one embodiment of the invention；
Fig. 4 is the video recommendations apparatus module composition schematic diagram of first embodiment of the invention；
Specific embodiment
The embodiment of the present invention provides the flow diagram that a kind of Fig. 1 is the video recommendation method of one embodiment of the invention, institute
The people that video recommendation method and device have high-impact by finding out a part is stated, and suitable video priority is recommended him
, after waiting them to have some interactions on video, video recommendations are given their good friend by we again.In given recommendation total quantity
In the case of, by this putting mode, we will obtain more users to the comment of video or thumb up, and video is allowed to stay to user
Under impression more profound, promote user experience and frequency of interaction, improve user's retention ratio.It the described method comprises the following steps:
S101 collects multiple user behavior datas in the given time, establishes user behavior data sample；
S103 stores the user behavior data sample；
In some embodiments, the primitive behavior data of user are obtained within the predetermined time first, then according in advance
The data standard of foundation formats the arrangement primitive behavior data, forms the new user behavior data sample for meeting specification,
Data storage tag and classified catalogue finally are established for these are complete, meet the user behavior data sample of specification, and will be used
The storage of family behavioral data sample data.
S105 is clustered using video content of the clustering algorithm to the user behavior data sample, obtains at least one
The similar user's cluster of a interest；
In some embodiments, the clustering algorithm is k-means algorithm.For example, on short video platform, we
The record that user watches video in the recent period can be collected.These records reflect user to the preference of video.Such as in video platform
On, it is assumed that there are four short-sighted frequency, three users, users 1 to have viewed video 1,2 and 3 for we, and user 2 has viewed short-sighted frequency 4, uses
Family 3 has viewed video 1,2, we can indicate the historical behavior record of user with table 1.Wherein number 1 indicates user's viewing
The video, 0 indicates no any behavior.If we are quantized with vector, record of the user on all videos, is used
Family 1,2,3 can be expressed as x1=[1110], x2=[0001] and x3=[1100].
User behavior record based on vectorization, we can calculate the similarity between user, then use clustering algorithm
Find out the similar user of hobby.There are many kinds of the functions of similarity selects, for example the Cosine similarity between vector,
Pearson correlation coefficient etc..By clustering algorithm, such as k-means, user 1 can be excavated according to calculation method and user 3 is
Belong to the user of the same type, user 2 is the user of another type.
K-means algorithm is hard clustering algorithm, is the representative of the typically objective function clustering method based on prototype, it is
Data point obtains the tune of interative computation using the method that function seeks extreme value to certain objective function of distance as optimization of prototype
Whole rule.For K-means algorithm using Euclidean distance as similarity measure, it is to seek corresponding a certain initial cluster center vector V most
Optimal sorting class, so that evaluation index J is minimum.Algorithm is using error sum of squares criterion function as clustering criteria function.
S107 counts the video that the similar user's cluster user of the interest was watched；
S109 extracts each frame image of the video and extracts feature；
In some embodiments, extracting each frame image of the video and extracting feature can be converted by scale invariant feature
Algorithm (Scale-invariant feature transform, SIFT), gray level co-occurrence matrixes method, Fourier power spectrum method and
One of deep learning model method is realized.
S111 is clustered using feature of the clustering algorithm to each frame image of the video, obtains at least one video
Cluster；
S113 analyzes the interactive relation in user's cluster described in the Video clustering between each user, according to described point
Analysis result obtains the influence power of user's cluster；
In some embodiments, the interactive relation tool in user's cluster described in the Video clustering between each user is analyzed
Body are as follows: a huge network G is constituted according to the interactive relation between user, V={ u1, u2 ..., un }, n are user's number,
According to interactive information of the user on video, influence power transfer matrix W ∈ R is calculatedn×n:
Wherein, CijComment and/or like time by user j to the user i video issued are commented for user j in user i
By or continue the number commented on or thumbed up, k ∈ N (u after having thumbed up some videoj) paid close attention to for all neighbours good friends of user j
Good friend.α, β are respectively the weights of importance value of different interactions, and W (i, j) describes user i to the influence power of user j.
The influence power of user's cluster is obtained according to the analysis result specifically: It+1=bWIt+(1-b)I0, wherein
It∈R1×nFor the influence value of all users of t moment；When t=0, I0Each element value be equal to 1/n；B is adjustable ginseng
Number.
The influence power of user's cluster is obtained according to the analysis result specifically: judges ItAnd It+1Difference whether
Meet the condition of convergence, exports I if meetingt.The condition of convergence is the threshold value or threshold range of setting.
Referring to Figure 2 together, it is assumed that short video interdynamic network is made of 4 users, interaction situation such as Fig. 1 between them
Shown, node U1 ..., U4 indicate 4 users, and directed edge indicates the interaction between user.Such as directed edge U4- > U1 indicates to use
Behavior of the family U4 to user U1, two numbers on side, which respectively indicate user U4 and have 2 times to the short-sighted frequency that user U1 is issued, to be thumbed up,
There is the comment of 1 follow-up to the short-sighted frequency of user U1 comment.α=0.5 is set, and β=0.5, then being calculated according to formula (1) influences
Power transfer matrix:
Then initialization I0B=0.85 is arranged in=(0.25,0.25,0.25,0.25).By W, I0, b substitution t wheel iteration fortune
It calculates, the last influence value I of each user can be obtainedt=(1.29,1.33,0.87,1.13).It can be seen that user U2
Influence power is maximum, and the influence power of U3 is minimum.
S115 recommends according to the ranking of the influence power of user's cluster.
In some embodiments, also referring to Fig. 3, citing is clustered according to user's watching behavior first, is found out emerging
The similar user's cluster of interest, such as to sport category user cluster and amusement class user cluster, sport category user's cluster is used again
The video that family has been seen is clustered according to content, such as finds out video relevant with football.Finally according to user in every class video
On mutual-action behavior find out the subset of the biggish user of influence power.Finally when being recommended, we preferentially push to user high
Influence power user gives the video of favorable comment and appreciation.
Referring to Fig. 4, the video recommendations apparatus module composition schematic diagram of first embodiment of the invention, the video recommendations dress
Setting 10 can be mobile phone, PDA (Personal Digital Assistant, personal digital assistant or tablet computer), portable
Any terminal having a display function such as communication device, computer, smart television, can also for the wearable devices such as Intelligent bracelet with
And it is embedded in the wearable device in the clothing components such as clothes, jewellery.The video recommendations device includes:
Acquiring unit 101 collects multiple user behavior datas in the given time, establishes user behavior data sample；
Storage unit 103, for storing the user behavior data sample；
In some embodiments, first obtain user primitive behavior data, then according to pre-establish data standard,
It formats and arranges the primitive behavior data, form the new user behavior data sample for meeting specification, it is finally complete for these
, the user behavior data sample for meeting specification establish data storage tag and classified catalogue, and by user behavior data sample
Data storage.
In some embodiments, the storage unit 103 is with flash type, hard disk type, Multimedia Micro Cards class
The storage of any one of type, card type reservoir (for example, SD or XD memory etc.), RAM and ROM (EEPROM etc.) type is situated between
Matter.
First cluster cell 105, for being carried out using video content of the clustering algorithm to the user behavior data sample
Cluster, obtains the similar user's cluster of at least one interest；
In some embodiments, the clustering algorithm is k-means algorithm.
Statistic unit 107, the video watched for counting the similar user's cluster user of the interest；
In some embodiments, statistic unit 107 based on vectorization user behavior record, we can calculate user it
Between similarity, then find out the similar user of hobby with clustering algorithm.There are many kinds of selections for the function of similarity, for example
Cosine similarity between vector, Pearson correlation coefficient etc..
Elementary area 109, for extracting each frame image of the video and extracting feature；
In some embodiments, described image unit 109 extracts each frame image of the video and extracts feature and can pass through
Scale invariant feature transfer algorithm (Scale-invariant feature transform, SIFT), gray level co-occurrence matrixes method,
One of Fourier power spectrum method and deep learning model method are realized.
Second cluster cell 111 is clustered using feature of the clustering algorithm to each frame image of the video, obtain to
A few Video clustering；
Analytical unit 113, for analyzing the interactive relation in user's cluster described in the Video clustering between each user,
The influence power of user's cluster is obtained according to the analysis result；
In some embodiments, the analytical unit 113 analyzes each user in user's cluster described in the Video clustering
Between interactive relation specifically: according between user interactive relation constitute a huge network G, V=u1, u2 ...,
Un }, n is user's number, according to interactive information of the user on video, calculate influence power transfer matrix W ∈ Rn×n:
Wherein, CijComment and/or like time by user j to the user i video issued are commented for user j in user i
By or continue the number commented on or thumbed up, k ∈ N (u after having thumbed up some videoj) paid close attention to for all neighbours good friends of user j
Good friend.α, β are respectively the weights of importance value of different interactions, and W (i, j) describes user i to the influence power of user j.
The influence power of user's cluster is obtained according to the analysis result specifically: It+1=bWIt+(1-b)I0, wherein
It∈R1×nFor the influence value of all users of t moment；When t=0, I0Each element value be equal to 1/n；B is adjustable ginseng
Number.
The influence power of user's cluster is obtained according to the analysis result specifically: judges ItAnd It+1Difference whether
Meet the condition of convergence, exports I if meetingt.The condition of convergence is the threshold value or threshold range of setting.
Referring to Figure 2 together, it is assumed that short video interdynamic network is made of 4 users, interaction situation such as Fig. 1 between them
Shown, node U1 ..., U4 indicate 4 users, and directed edge indicates the interaction between user.Such as directed edge U4- > U1 indicates to use
Behavior of the family U4 to user U1, two numbers on side, which respectively indicate user U4 and have 2 times to the short-sighted frequency that user U1 is issued, to be thumbed up,
There is the comment of 1 follow-up to the short-sighted frequency of user U1 comment.α=0.5 is set, and β=0.5, then being calculated according to formula (1) influences
Power transfer matrix:
After initialize I0B=0.85 is arranged in=(0.25,0.25,0.25,0.25).We are by W, I0, b substitution t wheel iteration
Operation can obtain the last influence value I of each usert=(1.29,1.33,0.87,1.13).It can be seen that user U2
Influence power it is maximum, the influence power of U3 is minimum.
Recommendation unit 115, the ranking for the influence power according to user's cluster are recommended.
Also referring to Fig. 3, citing is clustered according to user's watching behavior first, finds out the similar user's collection of interest
Group, such as to amusement class user cluster and amusement class user cluster, the video root that sport category user's cluster user has been seen again
It is clustered according to content, such as is divided into domestic or American-European relevant entertainment video.It is finally mutual on every class video according to user
The subset of the biggish user of influence power is found out in dynamic behavior.Finally when being recommended, we preferentially push high-impact to user
User gives the video of favorable comment and appreciation.
If the above-mentioned function of the present embodiment be realized in the form of SFU software functional unit and as independent product sale or
In use, can store in a computer readable storage medium, that is, the embodiment of the present invention can be with the shape of software product
Formula embodies comprising some instructions are used so that computer equipment (it can be personal computer, server, or
Network equipment etc.) it performs all or part of the steps of the method described in the various embodiments of the present invention.
On this basis, the above description is only an embodiment of the present invention, is not intended to limit the scope of the invention, all
It is to utilize technology between equivalent structure or equivalent flow shift, such as each embodiment made by description of the invention and accompanying drawing content
Feature be combined with each other, and being applied directly or indirectly in other relevant technical fields, similarly includes in patent of the invention
In protection scope.
Claims (10)
1. a kind of video recommendation method, which comprises the following steps:
Multiple user behavior datas are collected in the given time, establish user behavior data sample；
Store the user behavior data sample；
It is clustered using video content of the clustering algorithm to the user behavior data sample, it is similar to obtain at least one interest
User's cluster；
Count the video that the similar user's cluster user of the interest was watched；
It extracts each frame image of the video and extracts feature；
It is clustered using feature of the clustering algorithm to each frame image of the video, obtains at least one Video clustering；
The interactive relation in user's cluster described in the Video clustering between each user is analyzed, is obtained according to the analysis result
The influence power of user's cluster；
Recommended according to the ranking of the influence power of user's cluster.
2. video recommendation method as described in claim 1, which is characterized in that the clustering algorithm is k-means algorithm.
3. video recommendation method as described in claim 1, which is characterized in that extract each frame image of the video and extract spy
Sign can pass through scale invariant feature transfer algorithm (Scale-invariant feature transform, SIFT), gray scale symbiosis
One of matrix method, Fourier power spectrum method and deep learning model method is realized.
4. video recommendation method as described in claim 1, which is characterized in that analyze user's cluster described in the Video clustering
In interactive relation between each user specifically: a huge network G, V=are constituted according to the interactive relation between user
{ u1, u2 ..., un }；Wherein, u1 is user 1, and u2 is user 2 ..., and un is user n, and n is user's number, and V is user 1, user
2, user n gathers, and according to interactive information of the user on video, calculates influence power transfer matrix W ∈ Rn×n:
Wherein, CijComment and/or like time by user j to the user i video issued, AijIt is commented on for user j in user i
Or continue the number commented on or thumbed up after having thumbed up some video,K∈N(uj)Good friend is paid close attention to for all neighbours good friends of user j,
α, β are respectively the weights of importance value of different interactions, and W (i, j) describes user i to the influence power of user j.
5. video recommendation method as claimed in claim 4, which is characterized in that obtain the user according to the analysis result and collect
The influence power of group specifically: It+1=bWIt+(1-b)I0Wherein, It∈R1×nFor the influence value of all users of t moment；T=0
When, I0Each element value be equal to 1/n；B is customized parameter.
6. video recommendation method as claimed in claim 5, which is characterized in that obtain the user according to the analysis result and collect
The influence power of group specifically: judge ItAnd It+1Difference whether meet the condition of convergence, as meet if export It。
7. video recommendation method as claimed in claim 6, which is characterized in that the condition of convergence is the threshold value or threshold value of setting
Range.
8. a kind of video recommendations device, which is characterized in that the video recommendations device includes:
Acquiring unit establishes user behavior data sample for collecting multiple user behavior datas in the given time；
Storage unit, for storing the user behavior data sample；
First cluster cell is obtained for being clustered using clustering algorithm to the video content of the user behavior data sample
To the similar user's cluster of at least one interest；
Statistic unit, the video watched for counting the similar user's cluster user of the interest；
Elementary area, for extracting each frame image of the video and extracting feature；
Second cluster cell is clustered using feature of the clustering algorithm to each frame image of the video, obtains at least one
Video clustering；
Analytical unit, for analyzing the interactive relation in user's cluster described in the Video clustering between each user, according to institute
It states analysis result and obtains the influence power of user's cluster；
Recommendation unit, the ranking for the influence power according to user's cluster are recommended.
9. video recommendations device as claimed in claim 8, which is characterized in that first cluster cell and the second cluster cell
The clustering algorithm used is k-means algorithm.
10. video recommendations device as claimed in claim 8, which is characterized in that it is each that described image unit extracts the video
Frame image simultaneously extracts feature, can pass through scale invariant feature transfer algorithm (Scale-invariant feature
Transform, SIFT), gray level co-occurrence matrixes method, one of Fourier power spectrum method and deep learning model method realize.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710997564.0A CN107894998B (en) | 2017-10-24 | 2017-10-24 | Video recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710997564.0A CN107894998B (en) | 2017-10-24 | 2017-10-24 | Video recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107894998A CN107894998A (en) | 2018-04-10 |
| CN107894998B true CN107894998B (en) | 2019-04-26 |
Family
ID=61803748
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710997564.0A Active CN107894998B (en) | 2017-10-24 | 2017-10-24 | Video recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107894998B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921221B (en) * | 2018-07-04 | 2022-11-18 | 腾讯科技（深圳）有限公司 | User feature generation method, device, equipment and storage medium |
| CN109242030A (en) * | 2018-09-21 | 2019-01-18 | 京东方科技集团股份有限公司 | Draw single generation method and device, electronic equipment, computer readable storage medium |
| CN109218769B (en) * | 2018-09-30 | 2021-01-01 | 武汉斗鱼网络科技有限公司 | Recommendation method for live broadcast room and related equipment |
| CN109214374B (en) * | 2018-11-06 | 2020-12-18 | 北京达佳互联信息技术有限公司 | Video classification method, device, server and computer-readable storage medium |
| CN109740063A (en) * | 2019-01-17 | 2019-05-10 | 北京奇艺世纪科技有限公司 | Information recalls, information cluster method, device and equipment |
| CN110992137B (en) * | 2019-11-27 | 2023-09-01 | 上海风秩科技有限公司 | Product recommendation method and device, electronic equipment and storage medium |
| CN110996144B (en) * | 2019-12-19 | 2021-10-08 | 哈尔滨工业大学 | Television user viewing preference analysis method |
| CN112287166B (en) * | 2020-09-23 | 2023-03-07 | 山东师范大学 | Movie recommendation method and system based on improved deep belief network |
| CN112449240B (en) * | 2020-11-10 | 2022-12-06 | 深圳市易平方网络科技有限公司 | User loss prediction method and terminal based on Internet television use behaviors |
| CN117708436B (en) * | 2024-02-05 | 2024-04-26 | 福州掌中云科技有限公司 | Network literature short-play recommendation management system based on big data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103945238A (en) * | 2014-05-07 | 2014-07-23 | 北京邮电大学 | Community detection method based on user behaviors |
| CN104408210A (en) * | 2014-12-31 | 2015-03-11 | 合一网络技术(北京)有限公司 | Video recommendation method based on opinion leaders |
| CN106446078A (en) * | 2016-09-08 | 2017-02-22 | 乐视控股（北京）有限公司 | Information recommendation method and recommendation apparatus |
| CN106952166A (en) * | 2016-01-07 | 2017-07-14 | 腾讯科技（深圳）有限公司 | The user force evaluation method and device of a kind of social platform |
| CN107038213A (en) * | 2017-02-28 | 2017-08-11 | 华为技术有限公司 | A kind of method and device of video recommendations |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102334118B (en) * | 2010-11-29 | 2013-08-28 | 华为技术有限公司 | Promoting method and system for personalized advertisement based on interested learning of user |
| CN103377242B (en) * | 2012-04-25 | 2016-06-22 | Tcl集团股份有限公司 | User behavior analysis method, analyzing and predicting method and television program push system |
| JP6046393B2 (en) * | 2012-06-25 | 2016-12-14 | サターン ライセンシング エルエルシーＳａｔｕｒｎ Ｌｉｃｅｎｓｉｎｇ ＬＬＣ | Information processing apparatus, information processing system, information processing method, and recording medium |
-
2017
- 2017-10-24 CN CN201710997564.0A patent/CN107894998B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103945238A (en) * | 2014-05-07 | 2014-07-23 | 北京邮电大学 | Community detection method based on user behaviors |
| CN104408210A (en) * | 2014-12-31 | 2015-03-11 | 合一网络技术(北京)有限公司 | Video recommendation method based on opinion leaders |
| CN106952166A (en) * | 2016-01-07 | 2017-07-14 | 腾讯科技（深圳）有限公司 | The user force evaluation method and device of a kind of social platform |
| CN106446078A (en) * | 2016-09-08 | 2017-02-22 | 乐视控股（北京）有限公司 | Information recommendation method and recommendation apparatus |
| CN107038213A (en) * | 2017-02-28 | 2017-08-11 | 华为技术有限公司 | A kind of method and device of video recommendations |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107894998A (en) | 2018-04-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107894998B (en) | Video recommendation method and device | |
| CN105224699B (en) | News recommendation method and device | |
| CN109697629B (en) | Product data pushing method and device, storage medium and computer equipment | |
| CN104834686B (en) | A kind of video recommendation method based on mixing semantic matrix | |
| CN105677715B (en) | A kind of video recommendation method and device based on multi-user | |
| Yan et al. | Unified youtube video recommendation via cross-network collaboration | |
| CN102591942B (en) | Method and device for automatic application recommendation | |
| CN108665323B (en) | Integration method for financial product recommendation system | |
| WO2017181612A1 (en) | Personalized video recommendation method and device | |
| CN107295361B (en) | A kind of content delivery method | |
| CN106326391A (en) | Method and device for recommending multimedia resources | |
| CN109409928A (en) | A kind of material recommended method, device, storage medium, terminal | |
| CN103365936A (en) | Video recommendation system and method thereof | |
| CN101454771A (en) | System and method of segmenting and tagging entities based on profile matching using a multi-media survey | |
| CN104751354B (en) | A kind of advertisement crowd screening technique | |
| CN101668176A (en) | Multimedia content-on-demand and sharing method based on social interaction graph | |
| CN108959323B (en) | Video classification method and device | |
| CN104572733B (en) | The method and device of user interest labeling | |
| CN108595492A (en) | Method for pushing and device, storage medium, the electronic device of content | |
| CN107590267A (en) | Information-pushing method and device, terminal and readable storage medium storing program for executing based on picture | |
| Dhillon et al. | Modeling dynamic user interests: A neural matrix factorization approach | |
| CN107896153A (en) | A kind of flow package recommendation method and device based on mobile subscriber's internet behavior | |
| CN109597899A (en) | The optimization method of media personalized recommendation system | |
| Ma et al. | Gender estimation for sns user profiling using automatic image annotation | |
| CN107870934A (en) | A kind of app user clusterings 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |