CN110069714A - A kind of video recommendation system - Google Patents
A kind of video recommendation system Download PDFInfo
- Publication number
- CN110069714A CN110069714A CN201910341302.8A CN201910341302A CN110069714A CN 110069714 A CN110069714 A CN 110069714A CN 201910341302 A CN201910341302 A CN 201910341302A CN 110069714 A CN110069714 A CN 110069714A
- Authority
- CN
- China
- Prior art keywords
- data
- server
- bucket
- user
- recommendation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/68—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/686—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title or artist information, time, location or usage information, user ratings
-
- 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
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/7867—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
-
- 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
Abstract
The present invention relates to a kind of video recommendation systems, at least two HTTP server http-server, as system access entrance, the recommendation server cluster being made of recommendation server, each recommendation server internal logic is divided into: data distributor, data recall device, data filter, two history servers, using active-standby mode, history server is divided to two class data: one kind is the data in model, one kind is real time data, two are recalled pool server, using active-standby mode, play data buffer storage function, data recall device and pool server are recalled in the data recalled deposit, data filter obtains data from recalling in pool server.The present invention, system architecture is clear, and treatment effeciency is high, and stability is good, can effectively solve the problem of system resource deficiency, it can be achieved that video recommendations and audio recommendation function, promotion user experience are avoided that repetition pushes, promote the affinity of push experience.
Description
Technical field
The present invention relates to carry out multimedia content recommended technology field, specifically a kind of video recommendations according to user behavior
System.
Background technique
In the epoch of multimedia technology rapid development, the new trend of music and short-sighted frequency merged as development, in " music
On the basis of the mode of+social activity ", more and more music APP begin trying and provide short video capability, pass through " music+short-sighted
Frequently mode ", further the entertainment of abundant music APP, further enriches the content quality of friend's APP dynamic circle and enriches
Degree, the user activity of enhancing friend's APP dynamic circle.
Music+short-sighted frequency mode, has been related to B2C(Business-to-Customer) business, it is characterized in that stream
Amount is big, data are more, has the access that a large amount of user carries out high concurrent to system in the same period, this causes largely system
Access pressure, easily there is problem of system resource deficiency, such as memory, cpu, the network bandwidth deficiency of server end etc. and ask
Topic.
When system resource deficiency, how equipment (addition is dynamically added in the case where not influencing current system operation
Server), it needs to be optimized accordingly, when flow is reduced, when the situation of system resource deficiency is eased, how in not shadow
Equipment (reducing server) is dynamically reduced in the case where ringing current system operation, it is also desirable to further consider and design.
Since system is huge, need to consider and realize fully automated management (the automation generation of such as model data, pattern number
According to verification, data are by the synchronization on line, system is restarted under line), and when in cluster certain server exception exit
Afterwards, Realtime Alerts are carried out to it, to handle in time, it is ensured that system safety.
In addition, since scheme is personalized recommendation, how safety is carried out to system and restarts upgrading, it is subsequent be also required to give examine
Consider.
Summary of the invention
In view of the deficiencies in the prior art, the purpose of the present invention is to provide a kind of video recommendation system, system trays
Structure is clear, and treatment effeciency is high, and stability is good, can effectively solve the problem of system resource deficiency, it can be achieved that video recommendations and audio
Recommendation function promotes user experience, is avoided that repetition pushes, promotes the affinity of push experience.
To achieve the above objectives, the technical solution adopted by the present invention is that:
Appended claims are replicated at this place, please don't modify.
A kind of video recommendation system characterized by comprising
At least two HTTP server http-server, as system access entrance, for receiving and processing recommendation request, and
Corresponding recommending data is returned,
HTTP server adds recommendation server by building bucket on demand in bucket,
The recommendation server cluster being made of recommendation server, each recommendation server internal logic are divided into: data distributor
Deliver Server, data recall device Recall Server, data filter Filter Server,
Two history servers, using active-standby mode, history server is divided to two class data:
One kind is the data in model, such data is from the data for playing training statistics in record, which mainly includes going through
History plays record, the user's score counted from record etc.,
One kind is real time data, there are no the data for being trained to model,
Two are recalled pool server, using active-standby mode, play data buffer storage function, data recall device and call the data recalled deposit together
Pool server is returned, data filter obtains data from recalling in pool server.
Based on the above technical solution, HTTP server is responsible for data distribution, and total flow is divided into every by 100%
A bucket, each barrel of flow are any.
Based on the above technical solution, HTTP server is responsible for the test of AB strategy, bucket is arranged different strategies, bucket
In recommendation server then execute these strategy, then carry out the data comparison in different buckets between data.
Based on the above technical solution, data distributor: receiving the solicited message that http-server is sended over,
According to user's score that history server provides, the quantity to be recalled of device is recalled to calculate each data, is recalled then to data
Device sends data recall request,
Data recall device: receiving the request of data that data distributor sends over, data are recalled device and are divided into: using personalized model
Personalization recall device, device is recalled using the prevalence of Epidemic model and recall device using the cold data of cold data model,
Data filter: after data distributor recalls device transmission data to data, it is notified that data filter there are data needs to call together
It returns, then obtains all data in pool server and summarize from recalling, after all data summarizations, returned to by redis queue
Http-server, and history server is written into the data recalled.
Based on the above technical solution, the data distributor Deliver Server, data recall device Recall
Far call is completed using redis between Server, data filter Filter Server.
Based on the above technical solution, the data distributor Deliver Server, data recall device Recall
Parallel processing is used between Server, data filter Filter Server, to realize the low-response time.
Based on the above technical solution, the data distributor reads HTTP server from redis queue and sends
The solicited message to come over obtains the marking of the user according to solicited message to history server；
The various numbers for recalling device data are pushed according to the marking of user, are specifically included:
Judge user type, if marking < 1 of user, otherwise it is other users that user type, which is colod-application family,
For colod-application family:
Wpop=1, it represents hot data and recommends accounting, for hitting height, full dose is pushed away,
Wcf=0, it represents personalization and recalls recommendation accounting, for realizing personalized long-tail, improve user experience,
Wcold=0, it represents cold data and recommends accounting, for improving diversity, warmable data, when customized information deficiency, is pushed away,
For other users:
Wpop=1- Wcold - Wcf,
Wcf=min(∑ score/160,1- Wcold - Wpop), score is the marking of user,
Wcold=0.125。
Based on the above technical solution, on recommendation server, using index file technology, direct broadcasting user
History is put, the similarity information storage of song is to disk, to realize the occupancy for reducing a large amount of memories.
Based on the above technical solution, it builds bucket and specifically uses following strategy:
A point bucket is carried out to server cluster, adds server in bucket, traffic partition is carried out to bucket, calculates 1 to 100 according to User ID
Cryptographic Hash, so that user be made to fall in specified bucket, service or random selection one in bucket, or according to different strategies come
Selection one.
Based on the above technical solution, when inadequate resource, using following strategy:
New bucket is added, adds server in bucket, then divides and takes flow；
The IP and port and password of configuration recommendation system server redis in bucket then recommends clothes by information connection
Business device.
Video recommendation system of the present invention, system architecture is clear, and treatment effeciency is high, and stability is good, can effectively solve
The problem of system resource deficiency promotes user experience, it can be achieved that video recommendations and audio recommendation function, is avoided that repetition pushes,
Promote the affinity of push experience.
Video recommendation system of the present invention can be used as extremely my music APP video recommendations module, according to different user row
To recommend different individualized video data, so that a large amount of user is absorbed, to improve platform popularity.It can also be further
The amusement function of my abundant cruel music APP, expansion system business make the APP software of a multifunctional entertainment, meet more use
The individual demand at family.
The user behavior includes: the use habit of user, hobby etc..
Detailed description of the invention
The present invention has following attached drawing:
System network topology figure Fig. 1 of the invention.
System logic architecture figure Fig. 2 of the invention.
Specific embodiment
Below in conjunction with attached drawing, invention is further described in detail.
As shown in Figure 1, 2, video recommendation system of the present invention, comprising:
At least two HTTP server http-server, as system access entrance, for receiving and processing recommendation request, and
Corresponding recommending data is returned,
HTTP server http-server major function includes:
Data receiver: as the access entrance of recommender system, receiving user and request (recommendation request),
Build bucket: bucket is the concept of logic, and http-server can create multiple buckets, add recommendation server in bucket,
Addition recommendation server: adding recommendation server on demand in bucket,
Data distribution: total flow is divided into each bucket by 100%, and each barrel of flow can be any, such as: 5 buckets of creation, each
Containing two recommendation servers (quantity is merely illustrative, can be any) in bucket, each barrel of flow be 20% (each barrel of flow can be any,
But 100) summation is necessary for,
The test of AB strategy: bucket can be arranged different strategies, and the recommendation server in bucket then executes these strategies, such as: root
Recommend different types of popular song and newest test data etc. according to different districts and cities, then carry out in different buckets data it
Between data comparison,
Using two or more HTTP server http-server, load balancing can be better achieved, when certain HTTP server
After http-server is exited extremely, system is operated normally without influence, it can be ensured that system is reliable and stable,
When it is implemented, HTTP server is advisable with three or more,
The recommendation server cluster being made of recommendation server, each recommendation server internal logic are divided into: data distributor
Deliver Server, data recall device Recall Server, data filter Filter Server, in which:
Data distributor: the solicited message that http-server is sended over is received, is obtained according to the user that history server provides
Point, the quantity to be recalled of device is recalled to calculate each data, device is recalled then to data and sends data recall request,
Data recall device: receiving the request of data that data distributor sends over, data are recalled device and are divided into: using personalized model
Personalization recall device, device is recalled using the prevalence of Epidemic model and recall device using the cold data of cold data model,
Recalling device is exactly a kind of strategy, by calculated result of recalling by history server duplicate removal, is subsequently written and recalls pond clothes
Business device is summarized so that data filter obtains call back data,
Data filter: after data distributor recalls device transmission data to data, it is notified that data filter there are data needs to call together
It returns, then obtains all data in pool server and summarize from recalling, after all data summarizations, returned to by redis queue
Http-server, and history server is written into the data recalled,
By the reasonable decomposition of the above task module (referring to that data distributor, data recall device, data filter), height can be provided
Concurrent treatment effeciency,
The data distributor Deliver Server, data recall device Recall Server, data filter Filter
Far call is completed using redis between Server,
The data distributor Deliver Server, data recall device Recall Server, data filter Filter
Parallel processing is used between Server, to realize the low-response time,
Two history servers, using active-standby mode, history server is divided to two class data:
One kind is the data in model, such data is that (training of model is from the data for playing training statistics in record with day
Unit, the model as yesterday can be used today), which mainly includes the user that history plays record, counts from record
Score etc.,
One kind is real time data, and there are no the data for being trained to model (such as the broadcasting record of same day user, not to be counted on also
In model),
Two are recalled pool server, using active-standby mode, play data buffer storage function, data recall device and call the data recalled deposit together
Pool server is returned, data filter obtains data from recalling in pool server.
Based on the above technical solution, the data distributor reads HTTP server from redis queue and sends
The solicited message to come over obtains the marking of the user according to solicited message to history server.
The marking (user's score) of the user is recorded come the marking of counting user according to the broadcasting of the user, is specifically adopted
Obtain with the following methods: the broadcasting total duration of song obtains the score of the user song, then to institute divided by playout length
There is song score to be added the marking for just obtaining the user.
Based on the above technical solution, the various numbers for recalling device data are pushed according to the marking of user, specifically
Include:
Judge user type, if marking < 1 of user, otherwise it is other users that user type, which is colod-application family,
For colod-application family:
Wpop=1, it represents hot data and recommends accounting, for hitting height, full dose is pushed away,
Wcf=0, it represents personalization and recalls recommendation accounting, for realizing personalized long-tail, improve user experience,
Wcold=0, it represents cold data and recommends accounting, for improving diversity, warmable data, when customized information deficiency, is pushed away,
For other users:
Wpop=1- Wcold - Wcf,
Wcf=min(∑ score/160,1- Wcold - Wpop), score is the marking of user,
Wcold=0.125。
Based on the above technical solution, the data recall device, specifically include:
Device cf Recall Server is recalled in personalization, is played song records according to user and is extrapolated by algorithm turicreate
The individuation data of the user is filtered by History server and had pushed to obtain data, to avoid repeating to push；
The data recalled are stored in Recalling pool and recall pool server；
The algorithm turicreate refers to the Processing Algorithm based on open source machine learning frame turicreate, presses " proposed algorithm
Title, function name, content " divides as follows:
Based on the similar recommendation of item, item_similarity_recommender has a forecast function, favorite between item
Similarity degree.When being useful in unknown crowd recommendation, the similar right of item can be searched out；
Factorization, ranking_factorization_recommender and factorization_recommender,
It is the most frequently used, support the common progressive die type of additional information；
Similar recommendation based on content, item_content_recommender do not have user concept, Item oneself content
(various dimensions) determine, similar recommendation, and do not have comment data that can apply when can extracting；
Project popularity is recommended, and item popularity is recommended based on project popularity degree, and user does not enter model, lacks
Point: it can not vary with each individual, be affected by exceptional value；
Prevalence recalls device pop Recall Server, is sorted from high to low according to all playback of songs situations, selects the use
Family has not been seen and the higher data of score, is filtered by History server, to avoid repeating to push；
The data recalled are stored in Recalling pool and recall pool server；
Cold data recalls device cold Recall Server, and cold data is current latest data, i.e., new online data, it is unclear that
Whether liked by user, be likely to become prevalence data in later playing process, was carried out by History server
Filter, to avoid repeating to push
The data recalled are stored in Recalling pool and recall pool server.
Based on the above technical solution, user is by subscribing to source server Feed Server or video server
Video Channel Server issues the recommendation request.
Source server Feed Server is subscribed to, video server Video Channel Server is referred to as external docking
Server.
Based on the above technical solution, further include following one of any or all:
Log server Log Server, the service are connected with each task module, for recording system operation state；
Monitoring server Monitor Server, for monitoring the operation state of current system, if redis caches exception, currently
Recommender system processing pressure etc..
As one of optional embodiment, can configuration monitoring program on a recommendation server wherein, the monitoring
Program includes at least switch alarm.
Based on the above technical solution, on recommendation server, using index file technology, direct broadcasting user
History is put, the similarity information storage of song is to disk, to realize the occupancy for reducing a large amount of memories.
By disk storage, the case where can effectively solving low memory appearance, the stability of lifting system.
Based on the above technical solution, it builds bucket and specifically uses following strategy:
A point bucket is carried out to server cluster, adds server in bucket, traffic partition is carried out to bucket, calculates 1 to 100 according to User ID
Cryptographic Hash, so that user be made to fall in specified bucket, service or random selection one in bucket, or according to different strategies come
Selection one.
Bucket is the concept of logic, one or more recommendation server is arranged in bucket, and then HTTP server is according to shunting plan
Bucket is slightly selected first, is then randomly choosed certain recommendation server in bucket, is handled related service.
It needs to access big flow user as one, needs high concurrent and the distributed system supported, default server collection
Recommendation server in group is no less than 10.Minimalist configuration are as follows: every configuration memory minimum 126G, CPU need 32 cores, hard disk
Need 5T.
The following are the examples recommended on recommender system line:
If system recalls pond comprising three http-server services, ten recommendation services, two desk calendar history services (active and standby), two
Server (active and standby).
Three buckets are created in http-sever administration page, three recommendation services are added in first bucket, and second bucket is added
Four recommendation services are added in three recommendation services, third bucket, and flow is respectively 30%, 30%, 40%, i.e. the hash value range of bucket 1
For [1-30], the hash value range of bucket 2 is [31-60], and the hash value range of bucket 3 is [61-100].
Http-server receives request, and http-sever can calculate hash value corresponding to the user according to User ID
[1-100], by taking account 197076680 as an example, it is 50 that hash value, which is obtained by calculation, in http-server, then can be by the user point
It is fitted in second bucket, one recommendation service is selected according to load balancing from bucket, send user's request in this service
It goes.
Recommendation service receives solicited message, can be sent and be requested to history server according to User ID, obtain broadcasting for the user
Scoring information is put, specifically:
Data distributor is given a mark according to user, calculate the user respectively recall device recall ratio (such as the user request 10 videos,
Active user marking be 2.5, then by be calculated individualized video recall 3, popular video recall 6, cold data recall 1
It is a), then by the number recalled be sent to it is each recall device, and be sent to data filter, all data waited to recall.
Data recall device and calculate the data recalled, and are serviced by history and carry out data deduplication, prevent from repeating to push, call together
It returns device and sends the data recalled to and recall pool server and keep in.
Data filter is summarized from acquisition data in pool server are recalled, the data write-in history clothes after summarizing
Business, so as to subsequent training pattern and data deduplication, then sends http-server by redis queue for data.
Http-sever returns to away data.
Based on the above technical solution, when inadequate resource, using following strategy:
New bucket is added, adds server in bucket, then divides and takes flow.
As one of optional embodiment, can in bucket configuration recommendation system server redis IP and port, and
Password then connects recommendation server by the information.
The content being not described in detail in this specification belongs to the prior art well known to professional and technical personnel in the field.
Claims (10)
1. a kind of video recommendation system characterized by comprising
At least two HTTP server http-server, as system access entrance, for receiving and processing recommendation request, and
Corresponding recommending data is returned,
HTTP server adds recommendation server by building bucket on demand in bucket,
The recommendation server cluster being made of recommendation server, each recommendation server internal logic are divided into: data distributor
Deliver Server, data recall device Recall Server, data filter Filter Server,
Two history servers, using active-standby mode, history server is divided to two class data:
One kind is the data in model, such data is from the data for playing training statistics in record, which mainly includes going through
History plays record, the user's score counted from record etc.,
One kind is real time data, there are no the data for being trained to model,
Two are recalled pool server, using active-standby mode, play data buffer storage function, data recall device and call the data recalled deposit together
Pool server is returned, data filter obtains data from recalling in pool server.
2. video recommendation system as described in claim 1, it is characterised in that: HTTP server is responsible for data distribution, total stream
Amount is divided into each bucket by 100%, and each barrel of flow is any.
3. video recommendation system as described in claim 1, it is characterised in that: HTTP server is responsible for the test of AB strategy, to bucket
Different strategies is set, and the recommendation server in bucket then executes these strategies, then carries out the data in different buckets between data
Comparison.
4. video recommendation system as described in claim 1, it is characterised in that: data distributor: receiving http-server and send
The solicited message to come over recalls the quantity to be recalled of device according to user's score that history server provides to calculate each data,
Device, which is recalled, then to data sends data recall request,
Data recall device: receiving the request of data that data distributor sends over, data are recalled device and are divided into: using personalized model
Personalization recall device, device is recalled using the prevalence of Epidemic model and recall device using the cold data of cold data model,
Data filter: after data distributor recalls device transmission data to data, it is notified that data filter there are data needs to call together
It returns, then obtains all data in pool server and summarize from recalling, after all data summarizations, returned to by redis queue
Http-server, and history server is written into the data recalled.
5. video recommendation system as described in claim 1, it is characterised in that: the data distributor Deliver Server,
Data recall device Recall Server, complete far call using redis between data filter Filter Server.
6. video recommendation system as described in claim 1, it is characterised in that: the data distributor Deliver Server,
Data recall device Recall Server, parallel processing are used between data filter Filter Server, when realizing low-response
Between.
7. video recommendation system as claimed in claim 4, it is characterised in that: the data distributor is read from redis queue
The solicited message that HTTP server sends over is taken, obtains the marking of the user to history server according to solicited message；
The various numbers for recalling device data are pushed according to the marking of user, are specifically included:
Judge user type, if marking < 1 of user, otherwise it is other users that user type, which is colod-application family,
For colod-application family:
Wpop=1, it represents hot data and recommends accounting, for hitting height, full dose is pushed away,
Wcf=0, it represents personalization and recalls recommendation accounting, for realizing personalized long-tail, improve user experience,
Wcold=0, it represents cold data and recommends accounting, for improving diversity, warmable data, when customized information deficiency, is pushed away,
For other users:
Wpop=1- Wcold - Wcf,
Wcf=min(∑ score/160,1- Wcold - Wpop), score is the marking of user,
Wcold=0.125。
8. video recommendation system as described in claim 1, it is characterised in that: on recommendation server, using index file skill
Art, directly by the play history of user, the similarity information storage of song is to disk, to realize the occupancy for reducing a large amount of memories.
9. video recommendation system as described in claim 1, it is characterised in that: build bucket and specifically use following strategy:
A point bucket is carried out to server cluster, adds server in bucket, traffic partition is carried out to bucket, calculates 1 to 100 according to User ID
Cryptographic Hash, so that user be made to fall in specified bucket, service or random selection one in bucket, or according to different strategies come
Selection one.
10. video recommendation system as claimed in claim 9, it is characterised in that: when inadequate resource, using following strategy:
New bucket is added, adds server in bucket, then divides and takes flow；
The IP and port and password of configuration recommendation system server redis in bucket then recommends clothes by information connection
Business device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910341302.8A CN110069714B (en) | 2019-04-25 | 2019-04-25 | Video recommendation system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910341302.8A CN110069714B (en) | 2019-04-25 | 2019-04-25 | Video recommendation system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110069714A true CN110069714A (en) | 2019-07-30 |
| CN110069714B CN110069714B (en) | 2021-06-22 |
Family
ID=67368991
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910341302.8A Active CN110069714B (en) | 2019-04-25 | 2019-04-25 | Video recommendation system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110069714B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457582A (en) * | 2019-08-10 | 2019-11-15 | 北京酷我科技有限公司 | A kind of data distributing method and recommender system |
| CN110532468A (en) * | 2019-08-26 | 2019-12-03 | 北京齐尔布莱特科技有限公司 | A kind of recommended method of site resource, device and calculate equipment |
| CN110633760A (en) * | 2019-09-25 | 2019-12-31 | 北京酷我科技有限公司 | Recommendation system integration strategy and recommendation system |
| CN110704488A (en) * | 2019-09-29 | 2020-01-17 | 北京元年科技股份有限公司 | Method for managing data and corresponding system, computer device and medium |
| CN110809177A (en) * | 2019-10-22 | 2020-02-18 | 腾讯科技（深圳）有限公司 | Content processing method, content processing device, server and storage medium |
| CN110851647A (en) * | 2019-09-29 | 2020-02-28 | 广州荔支网络技术有限公司 | Intelligent distribution method, device and equipment for audio content flow and readable storage medium |
| CN111190815A (en) * | 2019-12-19 | 2020-05-22 | 广州华多网络科技有限公司 | Recall test method, apparatus, electronic device and storage medium |
| CN111400546A (en) * | 2020-03-18 | 2020-07-10 | 腾讯科技（深圳）有限公司 | Video recall method and video recommendation method and device |
| CN111914175A (en) * | 2020-08-07 | 2020-11-10 | 平安科技（深圳）有限公司 | Recommendation process optimization method, device, equipment and medium |
| CN112347355A (en) * | 2020-11-11 | 2021-02-09 | 广州酷狗计算机科技有限公司 | Data processing method, device, server and storage medium |
| CN112435091A (en) * | 2020-11-23 | 2021-03-02 | 百果园技术(新加坡)有限公司 | Recommended content selection method, device, equipment and storage medium |
| CN114039919A (en) * | 2021-11-08 | 2022-02-11 | 杭州网易云音乐科技有限公司 | Traffic scheduling method, medium, device and computing equipment |
| CN114079826A (en) * | 2020-08-14 | 2022-02-22 | 北京达佳互联信息技术有限公司 | Video recommendation list generation method and device, server and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103870972A (en) * | 2012-12-07 | 2014-06-18 | 盛乐信息技术（上海）有限公司 | Data recommendation method and data recommendation system |
| CN103974096A (en) * | 2013-02-01 | 2014-08-06 | 腾讯科技（北京）有限公司 | Video file assessment method, device and system |
| CN106331147A (en) * | 2016-09-09 | 2017-01-11 | 深圳市彬讯科技有限公司 | REDIS distributed type invoking method and system thereof |
| CN106682004A (en) * | 2015-11-06 | 2017-05-17 | 网宿科技股份有限公司 | Redis Key management method and system |
| US20180121549A1 (en) * | 2016-10-27 | 2018-05-03 | Conduent Business Services, Llc | Method and system for processing social media data for content recommendation |
| CN108205537A (en) * | 2016-12-16 | 2018-06-26 | 北京酷我科技有限公司 | A kind of video recommendation method and system |
| CN109284417A (en) * | 2018-08-27 | 2019-01-29 | 广州飞磨科技有限公司 | Video pushing method, device, computer equipment and storage medium |
-
2019
- 2019-04-25 CN CN201910341302.8A patent/CN110069714B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103870972A (en) * | 2012-12-07 | 2014-06-18 | 盛乐信息技术（上海）有限公司 | Data recommendation method and data recommendation system |
| CN103974096A (en) * | 2013-02-01 | 2014-08-06 | 腾讯科技（北京）有限公司 | Video file assessment method, device and system |
| CN106682004A (en) * | 2015-11-06 | 2017-05-17 | 网宿科技股份有限公司 | Redis Key management method and system |
| CN106331147A (en) * | 2016-09-09 | 2017-01-11 | 深圳市彬讯科技有限公司 | REDIS distributed type invoking method and system thereof |
| US20180121549A1 (en) * | 2016-10-27 | 2018-05-03 | Conduent Business Services, Llc | Method and system for processing social media data for content recommendation |
| CN108205537A (en) * | 2016-12-16 | 2018-06-26 | 北京酷我科技有限公司 | A kind of video recommendation method and system |
| CN109284417A (en) * | 2018-08-27 | 2019-01-29 | 广州飞磨科技有限公司 | Video pushing method, device, computer equipment and storage medium |
Non-Patent Citations (4)
| Title |
|---|
| ORK DE ROOIJ 等: "Active Bucket Categorization for High Recall Video Retrieval", 《IEEE》 * |
| PENGCHAO SUN 等: "Research on Personalized Recommendation Algorithm Based on Time Weighted and Sparse Space Clustering", 《IEEE》 * |
| 曹丹丹 等: "Ｒedis 数据库在视频推荐服务系统中的应用", 《计算机与现代化》 * |
| 钟前建: "基于Hadoop和Mahout的视频推荐技术研究及应用", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457582A (en) * | 2019-08-10 | 2019-11-15 | 北京酷我科技有限公司 | A kind of data distributing method and recommender system |
| CN110532468A (en) * | 2019-08-26 | 2019-12-03 | 北京齐尔布莱特科技有限公司 | A kind of recommended method of site resource, device and calculate equipment |
| CN110633760B (en) * | 2019-09-25 | 2023-01-17 | 北京酷我科技有限公司 | Recommendation system integration method and recommendation system |
| CN110633760A (en) * | 2019-09-25 | 2019-12-31 | 北京酷我科技有限公司 | Recommendation system integration strategy and recommendation system |
| CN110704488A (en) * | 2019-09-29 | 2020-01-17 | 北京元年科技股份有限公司 | Method for managing data and corresponding system, computer device and medium |
| CN110851647A (en) * | 2019-09-29 | 2020-02-28 | 广州荔支网络技术有限公司 | Intelligent distribution method, device and equipment for audio content flow and readable storage medium |
| CN110704488B (en) * | 2019-09-29 | 2022-02-08 | 北京元年科技股份有限公司 | Method for managing data and corresponding system, computer device and medium |
| CN110809177A (en) * | 2019-10-22 | 2020-02-18 | 腾讯科技（深圳）有限公司 | Content processing method, content processing device, server and storage medium |
| CN110809177B (en) * | 2019-10-22 | 2021-11-05 | 腾讯科技（深圳）有限公司 | Content processing method, content processing device, server and storage medium |
| CN111190815A (en) * | 2019-12-19 | 2020-05-22 | 广州华多网络科技有限公司 | Recall test method, apparatus, electronic device and storage medium |
| CN111190815B (en) * | 2019-12-19 | 2023-03-10 | 广州方硅信息技术有限公司 | Recall test method, apparatus, electronic device and storage medium |
| CN111400546B (en) * | 2020-03-18 | 2020-12-01 | 腾讯科技（深圳）有限公司 | Video recall method and video recommendation method and device |
| CN111400546A (en) * | 2020-03-18 | 2020-07-10 | 腾讯科技（深圳）有限公司 | Video recall method and video recommendation method and device |
| CN111914175A (en) * | 2020-08-07 | 2020-11-10 | 平安科技（深圳）有限公司 | Recommendation process optimization method, device, equipment and medium |
| CN114079826A (en) * | 2020-08-14 | 2022-02-22 | 北京达佳互联信息技术有限公司 | Video recommendation list generation method and device, server and storage medium |
| CN114079826B (en) * | 2020-08-14 | 2023-11-28 | 北京达佳互联信息技术有限公司 | Video recommendation list generation method, device, server and storage medium |
| CN112347355A (en) * | 2020-11-11 | 2021-02-09 | 广州酷狗计算机科技有限公司 | Data processing method, device, server and storage medium |
| CN112435091A (en) * | 2020-11-23 | 2021-03-02 | 百果园技术(新加坡)有限公司 | Recommended content selection method, device, equipment and storage medium |
| CN112435091B (en) * | 2020-11-23 | 2024-03-29 | 百果园技术(新加坡)有限公司 | Recommended content selection method, device, equipment and storage medium |
| CN114039919A (en) * | 2021-11-08 | 2022-02-11 | 杭州网易云音乐科技有限公司 | Traffic scheduling method, medium, device and computing equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110069714B (en) | 2021-06-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110069714A (en) | A kind of video recommendation system | |
| US9734507B2 (en) | Method and system for simulating recommendations in a social network for an offline user | |
| CN104967861B (en) | Video caching system and method in CDN network | |
| US9199165B2 (en) | Game clip popularity based control | |
| US20130191857A1 (en) | Method and System for a Vote Based Media System | |
| US20010044851A1 (en) | System and method for reducing the resources required to deliver streaming media | |
| US10574747B2 (en) | Improving peer-to-peer network performance through coordination of digital content distribution | |
| US20100268821A1 (en) | Sequenced transmission of digital content items | |
| US10523722B1 (en) | System and method for shared media session creation | |
| CN100454815C (en) | Method for realizing individualized advertsing managed by stream media | |
| CN104303205A (en) | Methods and systems for ordering and voting on shared media playlists | |
| CN109766200A (en) | A kind of message queue processing method, device, equipment and storage medium | |
| US20140067827A1 (en) | Automatically generating music playlists based on an implicitly selected seed | |
| CN101425317A (en) | System and method for reducing the multiple listing of a media item in a playlist | |
| US20120210383A1 (en) | Presenting streaming media for an event | |
| CN109257631A (en) | Video carousel playback method, device, computer equipment and storage medium | |
| CN108601980A (en) | Information processing system, information processing method, program, server and the information processing terminal | |
| CN105868325A (en) | Music playing method and system | |
| CN101114316A (en) | Network gaming system capable of ordering song and song ordering method in network game | |
| CN105897854A (en) | Mobile terminal alarm clock response method, device and system | |
| CN103841001A (en) | Social network information processing method and system | |
| CN106888406A (en) | A kind of time-based interaction content method for pushing and system | |
| CN104166820B (en) | A kind of data managing method and system | |
| US20080256176A1 (en) | Internet radio system and the broadcasting method thereof | |
| CN110457582A (en) | A kind of data distributing method and recommender system |
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 |