CN110442750A - Sea area network video recommended method based on time and space sequence information - Google Patents
Sea area network video recommended method based on time and space sequence information Download PDFInfo
- Publication number
- CN110442750A CN110442750A CN201910740690.7A CN201910740690A CN110442750A CN 110442750 A CN110442750 A CN 110442750A CN 201910740690 A CN201910740690 A CN 201910740690A CN 110442750 A CN110442750 A CN 110442750A
- Authority
- CN
- China
- Prior art keywords
- sea area
- data
- network video
- recommended method
- etl
- 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
- 238000000034 method Methods 0.000 title claims abstract description 99
- 230000008569 process Effects 0.000 claims abstract description 17
- 238000011084 recovery Methods 0.000 claims abstract description 17
- 238000006243 chemical reaction Methods 0.000 claims abstract description 13
- 238000013075 data extraction Methods 0.000 claims abstract description 3
- 238000004422 calculation algorithm Methods 0.000 claims description 37
- 239000011159 matrix material Substances 0.000 claims description 26
- 230000006399 behavior Effects 0.000 claims description 19
- 230000008859 change Effects 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 230000003542 behavioural effect Effects 0.000 claims description 4
- 230000010354 integration Effects 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 2
- 238000001914 filtration Methods 0.000 description 10
- 238000011156 evaluation Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 1
- 244000046052 Phaseolus vulgaris Species 0.000 description 1
- 238000012550 audit Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/23—Updating
- G06F16/2365—Ensuring data consistency and integrity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/25—Integrating or interfacing systems involving database management systems
- G06F16/254—Extract, transform and load [ETL] procedures, e.g. ETL data flows in data warehouses
-
- 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
-
- 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
- G06F16/787—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using geographical or spatial information, e.g. location
-
- 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
- G06F16/9536—Search customisation based on social or collaborative filtering
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Library & Information Science (AREA)
- Computational Linguistics (AREA)
- Computer Security & Cryptography (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The sea area network video recommended method based on time and space sequence information that this application discloses a kind of, it is updated including data warehouse ETL: during time point is recommended in setting, using incremental updating strategy, using sea area network as specifying hot recommendation method website to do user behavior data synchronization association in receptor and land network, to follow the temperature and popularity of standard recommendation method website;Further include ETL Fault recovery: being labeled as the module being skipped in the error model that only resumes operation, only re-executes the failure procedure of error in recovery process and TL changes table;The error model includes: that generation is converted and the mistake in loading procedure, the mistake of mistake and generation in factual data extraction process occurred in dimension data conversion and loading procedure in factual data.The application can evade cold start-up problem, can promote fault-tolerant efficiency.
Description
Technical field
This application involves recommended method technical field, in particular to a kind of sea area net based on time and space sequence information
Network video recommendation method.
Background technique
With the continuous development of Internet technology, the state of information content and knowledge quantity all in huge explosion.In order to cope with " letter
The challenge of breath overload ", it is current mainly to use two sets of feasible methods of search engine and recommended method.Search engine technique more at
It is ripe, but there is certain limitation, then recommender system technology is come into being.Recommender system is one of information filtering system
Special shape determines the item that user is present or may like future by analyzing the historical interest and preference information of user
Mesh, and then actively provide a user corresponding project recommendation service.Currently used proposed algorithm is content-based filtering calculation
Method and collaborative filtering.
Content-based recommendation algorithm is based primarily upon the description of recommended project and user characteristics give user's recommendation corresponding item
Mesh.The shortcomings that this algorithm, is: first, since this algorithm is based only upon the user characteristics of historical behavior, can not act on
The increased new user of machine;Second, the user characteristics with complex properties can not be handled.
Probability matrix decompose Collaborative Filtering Recommendation Algorithm usually assume that each user interest only by a few because
User's (project), is then mapped in the feature space of low-dimensional, is learned with the score information of each user (project) by the influence of element
The feature vector of family (project) is commonly used, to reconstruct rating matrix, and then using the low-dimensional Matrix prediction user of reconstruct to project
Scoring, recommended accordingly.The shortcomings that this algorithm, is: first, cold start-up problem;Due to that can not obtain new user's
Historical behavior record, recommending it will be extremely difficult;Similar, new projects are due to seldom by the number of selection evaluation, equally very
Difficulty is recommended suitable user;Second, interest drift and accuracy problem;The project popular, score is high is with the time
Can occur the variation of non-regularity, unexpected winner, new projects cannot recommend chance, also just can not effective renewal item library, can not
Bring the accurate recommendation for meeting user demand.
Sea area network is due to the uniqueness of its own application scenarios, and with communication standard, incompatible, coverage area exists
The features such as blind area.Therefore, simply above two scheme is applied to be difficult to see better effects in sea area net scene, it is also difficult to
Meet the application demand of special scenes and target minority user.Since type of merchandize and quantity are indefinite, and over time,
User demand will also update, and occlude area similar to information such as complicated application scenarios such as sea area, remote mountains, user may be only to few
The several commodity of number had evaluation information, or even no any evaluation information occurred to lead to the problem of cold start-up.And such use
Family very likely only represents minority's hobby, cannot provide reference to update recommendation information next time.
The disclosure of background above technology contents is only used for auxiliary and understands present invention design and technical solution, not
The prior art for necessarily belonging to the application shows that above content has disclosed in the applying date of the application in no tangible proof
In the case where, above-mentioned background technique should not be taken to the novelty and creativeness of evaluation the application.
Summary of the invention
The application proposes a kind of sea area network video recommended method based on time and space sequence information, can evade cold open
Dynamic problem.
In a first aspect, the application provides a kind of sea area network video recommendation side based on time and space sequence information
Method, including data warehouse ETL update: during time point is recommended in setting, using incremental updating strategy, using sea area network as by
Hot recommendation method website is specified to do user behavior data synchronization association in body and land network, to follow standard recommendation method net
The temperature and popularity stood.
In some preferred embodiments, the data warehouse ETL updates: only updating within the scope of setting time every time
Dimension and true behavioral data.
In some preferred embodiments, further include ETL Fault recovery: be labeled as in the error model that only resumes operation by
The module skipped only re-executes the failure procedure of error in recovery process and TL changes table;The error model includes: to occur
It converts in factual data and exists with mistake and generation of the mistake, generation in loading procedure in dimension data conversion and loading procedure
Mistake in factual data extraction process.
In some preferred embodiments, further includes:
It establishes and extracts log and conversion and load log;
If a certain data module failure, when ETL process proceeds to next data module, obtains from data module allocation list
The data module that next data module relies on, and check described in the extraction log and the conversion and load log
Whether the data module that next data module relies on correctly executes, if so, the operation of next data module is carried out, if
It is no, then skip next data module.
In some preferred embodiments, further includes:
The timing information of receptor user in the sea area network video recommended method and land network are specified into recommendation side
Similar structural relation between the timing information building user of subject user in method or project;
The regional impact factor is obtained in conjunction with regional context feature;
By the similar structural relation and the regional impact indexes integration of factors to the sea area network video recommended method
In, form new recommendation frame.
In some preferred embodiments, further includes: introduce time series impact factor, the time relationship that will be excavated
Information application is into matrix decomposition model, to establish the collaborative filtering recommending model based on timing behavior.
In some preferred embodiments, further includes: the geographical space in specified sea areas network range is done into node volume
Number, the characteristics of according to each geographical space, construct respective score in predicting matrix;Prior clearly all regional impact degree and corresponding
Score in predicting matrix, one domain nodes of every arrival, by corresponding score in predicting matrix update to the sea area network
In the recommended models of video recommendation method.
In some preferred embodiments, further includes: measurement ship between bank base at a distance from, then switch less than setting value
It for spatial sequence algorithm, is greater than the set value, is switched to time series algorithm.
In second aspect, the application provides a kind of calculating equipment, and the calculating equipment includes:
One or more processors;
Memory, for storing one or more programs;
One or more of programs can be executed by one or more of processors, to realize the above method.
In the third aspect, the application provides a kind of computer readable storage medium, in the computer readable storage medium
It is stored with program instruction, described program, which instructs, makes the processor execute the above method when being executed by the processor of computer.
Compared with prior art, the beneficial effect of the application has:
It can unify the content standard of sea area network recommendation method with land Network Synchronization, it can be achieved that being associated between heterogeneous network,
So as to evade cold start-up problem.
Detailed description of the invention
Fig. 1 schematically shows the first kind ETL error model of the application first embodiment;
Fig. 2 schematically shows the second class ETL error model of the application first embodiment;
Fig. 3 schematically shows the third class ETL error model of the application first embodiment;
Fig. 4 schematically shows the recovery process of the first kind ETL mistake of the application first embodiment;
Fig. 5 is the main receptor user behavior associated diagram based on timing in the application first embodiment;
Fig. 6 is that algorithm renewal process block diagram after timing impact factor is introduced in the application first embodiment;
Fig. 7 is that predictive equation renewal process block diagram after the spacial influence factor is introduced in the application first embodiment;
Fig. 8 is the structural schematic diagram of the calculating equipment of the application second embodiment.
Specific embodiment
In order to which the embodiment of the present application technical problem to be solved, technical solution and beneficial effect is more clearly understood,
Below in conjunction with Fig. 1 to Fig. 8 and embodiment, the application is further elaborated.It should be appreciated that described herein specific
Embodiment only to explain the application, is not used to limit the application.
First embodiment
The sea area network video recommended method based on time and space sequence information that the present embodiment provides a kind of, is to be exclusively used in
The video recommendation method of the remote overlay network of sea area scene.
The sea area network video recommended method of the present embodiment includes that data warehouse ETL updates, ETL error model models, ETL
Fault recovery, timing behavior modeling are commented with proposed algorithm output, space attribute modeling and reconstruct score in predicting, proposed algorithm is updated
Price card standard and performance evaluation.Wherein, ETL (Extract, Transform, Load) refers to extraction, conversion and load.
Data warehouse ETL, which is updated, updates the realization of strategy program module by data warehouse ETL.ETL error model is modeled by ETL
Error model modeling program module is realized.ETL Fault recovery is realized by ETL error recovery procedure module.
The present embodiment proposes the ETL method of Data Integration normalizing under the conditions of the magnanimity multidimensional data that recommended method relies on,
It is updated including data warehouse ETL, ETL error model models and ETL Fault recovery.
Data warehouse ETL update specifically includes: during time point is recommended in setting, passing through corresponding data pick-up and update
Mode does use for sea area network as hot recommendation method website specified in receptor and land network using incremental updating strategy
Family behavioral data synchronization association, to follow the temperature and popularity of standard recommendation method website.The user of sea area network is only capable of generation
Table extremely minority group, and the content standard of sea area network recommendation method can be unified with land Network Synchronization, with sea area network
Rather than its user carries out data synchronization association, it can be achieved that being associated between heterogeneous network, so as to evade cold start-up problem as receptor.
Data warehouse ETL updates dimension and the true behavioral data only updated within the scope of setting time every time, specifically
It is the new record of insertion while the attribute for updating already existing record variation.
ETL occurs needing to rerun when run-time error or output data are more than certain limit.Therefore, the view of the present embodiment
Frequency recommended method includes ETL Fault recovery.The present embodiment is divided ETL error model by ETL error model modeling program module
For following three classes:
1) occur factual data (Fact) convert with load (TL) during mistake, it only influence the fact module with
Last changes table step;
2) mistake during dimension data (Dim) is converted and loads (TL) occurs, it is influenced dependent on the dimension
Factual data TL and last change table step;
3) occur factual data extract (E) during mistake, it can skip extraction change table, factual data TL with
And TL it is last change table.
Above-mentioned three classes mistake, the data module that the follow-up link and needs influenced is skipped are as shown in Figure 1 to Figure 3.
ETL Fault recovery only resumes operation and is labeled as being skipped the module of (skip) in error model, in recovery process
Failure (fail) process and TL for only re-executing error change table.By taking first kind ETL mistake as an example, no longer by reruning
There is impacted data module to restore data, and is labeled as being skipped the module of (skip) in the Fig. 1 that only resumes operation.Fig. 4
It is the recovery process of Error type I, the data module fact1TL process and TL of error has only been re-executed in recovery process
Change table.
As described above, the present embodiment establishes the model of each data module weak coupling, low ETL mistake, again
Resume operation erroneous part when can reduce influence to other modules, promote fault-tolerant efficiency.
In the case where the error of ETL data module, the sea area network video recommended method of the present embodiment passes through log strategy
Program module, which is established, extracts whether log, conversion and load log, dependence mapping table jump come subsequent module of determining to stagger the time
It crosses.Text specific as follows.
Log and conversion and load log are extracted firstly, establishing.
Secondly, if a certain data module fails, when ETL process proceeds to next data module, from data module allocation list
The middle data module for obtaining next data module and relying on, and check and extract next data mould in log and conversion and load log
Whether the data module that block relies on correctly executes;If so, carrying out the operation of next data module;If it is not, then skipping next number
According to module.
Above content is illustrated so that ETL is related to two dimension tables and two true tables as an example.Record extract, conversion and
Each part implementing result is loaded into log, respectively corresponds table one and table two.
One ETL extraction process log of table
Two ETL of table conversion and loading procedure log
Log crawl during all data modules implementing result and to data module fact1TL execute failure before
All data modules execute record.After data module fact1TL failure, ETL process proceeds to data module
Fact2TL first checks for which number data module fact2TL in allocation list relies on when judging whether to execute the data module
According to module, then whether audit log table corresponding data module is correctly executed, as all carried out data module if successful execution
Otherwise the data module is skipped in the operation of fact2TL.
As described above, the present embodiment is in the case where ETL data module malfunctions, it is quickly extensive according to log and allocation list
It is multiple, the conjunction coupling degree between different dimensions data is effectively reduced, ensure that the data validity that proposed algorithm is based on, after being
Effective application of continuous proposed algorithm provides necessary accurate data.
The interest preference and demand of user can change with time change and sea area spatial information namely interest drift
It moves.Therefore, the sea area network video recommended method of the present embodiment includes timing behavior modeling and update proposed algorithm output.Specifically
It is as follows.
The timing information of receptor user in the network video recommended method of sea area and land network are specified in recommended method
Subject user timing information building user or project between similar structural relation;
The regional impact factor is obtained in conjunction with regional context feature;
By similar structural relation and regional impact indexes integration of factors into sea area network video recommended method, so that recommendation side
The model of method can learn the dynamic change to data, form new recommendation frame.
Wherein, timing behavior modeling process description is as follows.
The present embodiment excavates the partial rules and connection between user or project, introduces the main receptor user row based on timing
For relating heading, as shown in Figure 5.In this user behavior relating heading, U is the set of user.If W indicates behavior weight
Number, T is responsibility coefficient.For example, the U within the period of settingi(land network specifies the subject user in recommended method) and Uj
(the receptor user in the network recommendation method of sea area) successively has viewed the same project, then weight Wi-jNumerical value increases by 1.Subsequent time
Go through the project situation in all coextensives.It is described according to above procedure, the responsibility coefficient for defining user is as follows:
Ti-j=Wi-j/f(Ui,Uj)
Wherein, f (Ui,Uj) it is user UiAnd UjWatch the union number of video., it is specified that only focusing on U under this application sceneiIt is right
UjIt is unidirectional to influence behavior.
Similarly, project-based passive behavior association is established, corresponding responsibility coefficient calculates are as follows:
Si-j=Wi-j/f(Vi,Vj)
Wherein, ViRefer to that land network specifies the principal Item in recommended method, VjRefer to the receptor in sea area network recommendation method
User.
The influence power calculation method that the present embodiment proposes is simple and effective, can search out for specific user or project to it
Maximum Neighbourhood set is influenced, further these nearly Neighbourhood sets are fused in the method filter algorithm based on matrix decomposition,
Traditional collaborative filtering is improved, recommendation accuracy can be improved.Improvement of the present embodiment to traditional collaborative filtering
Specifically: time series impact factor is introduced, by the time relationship Information application excavated into matrix decomposition model, to build
The collaborative filtering recommending model of the timing that is based on behavior.
It is as follows to update proposed algorithm process:
After excavating similar structural relation between corresponding user and project and finding nearest-neighbor, probability is applied it to
In matrix decomposition model, the feature vector of user's (project) should be influenced by its neighborhood user (project) at this time, i.e., similar
User's (project) should have similar feature vector:
Uu *=Tv-uUu
Vi *=Sj-iVi
Wherein, Uu *、Vi *Indicate approximate feature vector, u, v are neighborhood user, and i, j are neighborhood commodity.Therefore, UuIt indicates
The feature vector of user, ViThe feature vector of expression project.
In this way, the feature vector of each user (project) can be obtained on the basis of above, new scoring square is finally obtained
Battle array, uses RijRepresent the rating matrix of user.It is as shown in Figure 6 that algorithm updates reasoning flow.
As described above, the present embodiment improves traditional collaborative filtering, time series impact factor is introduced,
It is decomposed in common probability matrix and increases by one group of main receptor timing behavior association impact factor in collaborative filtering, utilize two classes
The sequential correlation Behavior mining user of the different user of network and commodity neighbor relationships, to line lower network difficulty scene such as sea area office
Information and resource in the net of domain do the accurate update deployment in accordance with trend, update rating matrix, can cope with interest drift.
For for the sea area network video recommended method of special application scenarios sea area network, the present embodiment include space belong to
Property modeling with reconstruct score in predicting.The sea area network video recommended method of the present embodiment introduces the probability matrix of regional context information
Decompose proposed algorithm, abbreviation spatial sequence algorithm;The standard set in its last score in predicting Xiang Weiben LAN subscriber is pre-
Survey the linear combination of item with the prediction term containing the specified regional context ingredient with disturbance degree, its predictive equation are as follows:
Rij=α Ui TVi+(1-α)∑TijUj TVj
Fig. 7 indicates the renewal process block diagram of predictive equation after the introducing spacial influence factor.Wherein α is control spatial information shadow
Ring power Uj1Parameter (respective weights influence value is set according to specific region location number), i is the present embodiment application scenarios sea
User behavior domain in the network of domain, j are the user behavior domain of several specified regional contexts with disturbance degree, TijIt is specified
To the impact factor of target user in the network of sea area, it often to the reversed relevant coefficient of similarity, is being predicted regional context
The sum of similarity can be normalized to 1 in formula.
As described above, the present embodiment is directed to sea area application scenarios, each particular space region sequence letter in sea area is introduced
Breath proposes that Optimization-type probability matrix decomposes proposed algorithm on the basis of tradition coordinates filter algorithm, constructs line in conjunction with space attribute
Property combination score in predicting matrix, be acting exclusively on the video recommendations application of the sea area network user and offshore applications scene.Specifically
, the geographical space in specified sea areas network range is done into node serial number, the characteristics of according to each geographical space, is constructed respective
Score in predicting matrix;Prior clearly all regional impact degree and corresponding score in predicting matrix, one domain nodes of every arrival will
Corresponding score in predicting matrix update is into the recommended models of the sea area network video recommended method of the present embodiment.
The present embodiment is using RSME (Root Mean Squared Error, root-mean-square error) and Riming time of algorithm etc.
As the algorithm evaluation criterion of update, to evaluate the algorithm of update.That is, the sea area network video of the present embodiment is recommended
Method includes proposed algorithm evaluation criterion and performance evaluation.
Wherein the calculation formula of RSME is as follows:
RMSE=(∑I=1 C(pi-ri)2/C)1/2
Assuming that algorithm is expressed as { p to the score data of C commodity projection1, p2... ..., pc, corresponding user's scoring number
According to for { r1, r2... ..., rc}.The present embodiment uses the data grabbed from bean cotyledon website as experimental data set.It grabs altogether
Two group data sets, one group of data are score information, user social contact relationship and label information of the user to books, another group of data
Score information and corresponding other information for user to film.Data cases are as shown in Table 3.
Data information of three user of table to books and film
The present embodiment will introduce the probability matrix decomposition algorithm of timing information and introduce the probability matrix of regional context information
The index that index coordinates filter algorithm with tradition compares, and result is as shown in Table 4.
Four timing of table and Spatial Probability matrix and tradition coordinate the comparison of filter algorithm index
According to table four it is found that index comparing result shows time and space modified proposed algorithm runing time expense more
Small, resource overhead is less, recommends for individual character scene more acurrate.
In the present embodiment, there are a quantization Rule of judgment with time series in space.Measure between ship and bank base away from
From being switched to spatial sequence algorithm less than setting value, be greater than the set value and be switched to time series algorithm, to guarantee the suitable of algorithm
With property and updating survey.
The sea area network video recommended method of the present embodiment can remote overlay network, can effectively evade conventional recommendation calculation
Cold start-up problem in method.By introducing time and spatial sequence information, the applicability of recommendation can be promoted.The present embodiment is directed to and pushes away
The building process ETL for recommending part and parcel data warehouse in method, propose it is a kind of from each data source extract data scheduling more
New method, and different dimensional can effectively reduce according to log and the fast quick-recovery of allocation list in the case where the error of ETL data module
Conjunction coupling degree of the degree between, it is ensured that the data validity that proposed algorithm is based on, it being capable of having for subsequent recommendation algorithm
Effect provides necessary accurate data.
Second embodiment
With reference to Fig. 8, the present embodiment provides a kind of calculating equipment, including one or more processors 5 and memory 6.Wherein,
Memory 5 is for storing one or more programs.One or more programs can be executed by one or more processors 6, to realize
The above method.
It will be appreciated by those skilled in the art that all or part of the process in embodiment method can be by computer program
Carry out the relevant hardware of order to complete, program can be stored in computer-readable storage medium, and program is when being executed, it may include such as
The process of each method embodiment.And storage medium above-mentioned includes: ROM or random access memory RAM, magnetic or disk etc.
The medium of various program storage codes.
The above content is combining specific/preferred embodiment to be further described to made by the application, cannot recognize
The specific implementation for determining the application is only limited to these instructions.For those of ordinary skill in the art to which this application belongs,
Without departing from the concept of this application, some replacements or modifications can also be made to the embodiment that these have been described,
And these substitutions or variant all shall be regarded as belonging to the protection scope of the application.
Claims (10)
1. a kind of sea area network video recommended method based on time and space sequence information, it is characterised in that including data warehouse
ETL updates:
During time point is recommended in setting, using incremental updating strategy, using sea area network as specified in receptor and land network
User behavior data synchronization association is done in hot recommendation method website, to follow the temperature and popularity of standard recommendation method website.
2. network video recommended method in sea area according to claim 1, it is characterised in that the data warehouse ETL updates:
The dimension within the scope of setting time and true behavioral data are only updated every time.
3. network video recommended method in sea area according to claim 1, it is characterised in that:
Further include ETL Fault recovery: being labeled as the module being skipped in the error model that only resumes operation, it is only heavy in recovery process
The new failure procedure for executing error and TL change table;
The error model includes: the mistake occurred in factual data conversion and loading procedure, occurs to convert in dimension data
With the mistake of mistake and generation in factual data extraction process in loading procedure.
4. network video recommended method in sea area according to claim 3, it is characterised in that further include:
It establishes and extracts log and conversion and load log;
If a certain data module failure, when ETL process proceeds to next data module, from data module allocation list described in acquisition
The data module that next data module relies on, and check next described in the extraction log and the conversion and load log
Whether the data module that data module relies on correctly executes, if so, the operation of next data module is carried out, if it is not, then
Skip next data module.
5. network video recommended method in sea area according to claim 1, it is characterised in that further include:
The timing information of receptor user in the sea area network video recommended method and land network are specified in recommended method
Subject user timing information building user or project between similar structural relation;
The regional impact factor is obtained in conjunction with regional context feature;
By the similar structural relation and the regional impact indexes integration of factors into the sea area network video recommended method, shape
The recommendation frame of Cheng Xin.
6. network video recommended method in sea area according to claim 1, it is characterised in that further include: introduce time series shadow
The factor is rung, by the time relationship Information application excavated into matrix decomposition model, to establish the collaboration based on timing behavior
Filtered recommendation model.
7. network video recommended method in sea area according to claim 1, it is characterised in that further include: by specified sea areas network
Geographical space in range does node serial number, the characteristics of according to each geographical space, constructs respective score in predicting matrix;In advance
All regional impact degree and corresponding score in predicting matrix are specified, one domain nodes of every arrival are pre- by corresponding scoring
Matrix update is surveyed into the recommended models of the sea area network video recommended method.
8. network video recommended method in sea area according to claim 1, it is characterised in that further include: measurement ship and bank base
Between distance, spatial sequence algorithm is then switched to less than setting value, is greater than the set value, time series algorithm is switched to.
9. a kind of calculating equipment, it is characterised in that the calculating equipment includes:
One or more processors;
Memory, for storing one or more programs;
One or more of programs can be executed by one or more of processors, any to 8 according to claim 1 to realize
Item the method.
10. a kind of computer readable storage medium, it is characterised in that: be stored with program in the computer readable storage medium and refer to
It enables, described program, which instructs, executes the processor according to claim 1 to any one of 8 institutes
State method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910740690.7A CN110442750A (en) | 2019-08-12 | 2019-08-12 | Sea area network video recommended method based on time and space sequence information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910740690.7A CN110442750A (en) | 2019-08-12 | 2019-08-12 | Sea area network video recommended method based on time and space sequence information |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110442750A true CN110442750A (en) | 2019-11-12 |
Family
ID=68434743
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910740690.7A Pending CN110442750A (en) | 2019-08-12 | 2019-08-12 | Sea area network video recommended method based on time and space sequence information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110442750A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115017400A (en) * | 2021-11-30 | 2022-09-06 | 荣耀终端有限公司 | Application APP recommendation method and electronic equipment |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7596549B1 (en) * | 2006-04-03 | 2009-09-29 | Qurio Holdings, Inc. | Methods, systems, and products for analyzing annotations for related content |
CN103209342A (en) * | 2013-04-01 | 2013-07-17 | 电子科技大学 | Collaborative filtering recommendation method introducing video popularity and user interest change |
CN105872790A (en) * | 2015-12-02 | 2016-08-17 | 乐视网信息技术(北京)股份有限公司 | Method and system for recommending audio/video program |
US20180039910A1 (en) * | 2016-08-02 | 2018-02-08 | Telefonaktiebolaget Lm Ericsson (Publ) | System and method for recommending semantically similar items |
CN108920503A (en) * | 2018-05-28 | 2018-11-30 | 哈尔滨工程大学 | A kind of micro- video personalized recommendation algorithm based on social networks degree of belief |
CN109408734A (en) * | 2018-09-28 | 2019-03-01 | 嘉兴学院 | A kind of collaborative filtering recommending method of fuse information Entropy conformability degree and dynamic trust |
-
2019
- 2019-08-12 CN CN201910740690.7A patent/CN110442750A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7596549B1 (en) * | 2006-04-03 | 2009-09-29 | Qurio Holdings, Inc. | Methods, systems, and products for analyzing annotations for related content |
CN103209342A (en) * | 2013-04-01 | 2013-07-17 | 电子科技大学 | Collaborative filtering recommendation method introducing video popularity and user interest change |
CN105872790A (en) * | 2015-12-02 | 2016-08-17 | 乐视网信息技术(北京)股份有限公司 | Method and system for recommending audio/video program |
US20180039910A1 (en) * | 2016-08-02 | 2018-02-08 | Telefonaktiebolaget Lm Ericsson (Publ) | System and method for recommending semantically similar items |
CN108920503A (en) * | 2018-05-28 | 2018-11-30 | 哈尔滨工程大学 | A kind of micro- video personalized recommendation algorithm based on social networks degree of belief |
CN109408734A (en) * | 2018-09-28 | 2019-03-01 | 嘉兴学院 | A kind of collaborative filtering recommending method of fuse information Entropy conformability degree and dynamic trust |
Non-Patent Citations (2)
Title |
---|
刘畅等: "推荐系统冷启动问题分析", 《电信网技术》 * |
孙光福等: "基于时序行为的协同过滤推荐算法", 《软件学报》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115017400A (en) * | 2021-11-30 | 2022-09-06 | 荣耀终端有限公司 | Application APP recommendation method and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112528035B (en) | Knowledge graph reasoning method and device based on relational attention and computer equipment | |
CN103118132B (en) | A kind of distributed cache system towards space-time data and method | |
CN109960722B (en) | Information processing method and device | |
Botta et al. | Finding network communities using modularity density | |
CN113761219A (en) | Knowledge graph-based retrieval method and device, electronic equipment and storage medium | |
CN113641797A (en) | Data processing method, device, equipment, storage medium and computer program product | |
CN113806537A (en) | Commodity category classification method and device, equipment, medium and product thereof | |
CN110442750A (en) | Sea area network video recommended method based on time and space sequence information | |
CN114549845A (en) | Logo image detection method and system based on feature fusion | |
Afshani et al. | A Fuzzy DEMATEL‐ANP‐Based Approach to Prioritize Activities in Enterprise Architecture | |
CN114491236A (en) | Intelligent supply chain matching technology based on knowledge learning | |
Gu et al. | Feaco: Reaching robust feature-level consensus in noisy pose conditions | |
KR101273646B1 (en) | Method and system for indexing and searching in multi-modality data | |
Liu et al. | Wl-align: Weisfeiler-lehman relabeling for aligning users across networks via regularized representation learning | |
CN113822346A (en) | Bipartite graph classification method based on capsule graph neural network | |
CN114529325A (en) | Automatic click rate estimation model generation method based on Darts | |
Martins et al. | Athena: A knowledge fusion algorithm for the internet of things | |
CN110197305B (en) | Relay protection data model searching and optimizing method and system based on shortest path algorithm | |
Fang et al. | Multi-task metric learning on network data | |
Shahzad et al. | Automated Generation of Graphs from Relational Sources to Optimise Queries for Collaborative Filtering | |
Qin et al. | [Retracted] A Hierarchical Children’s Dance Movement Pose Estimation Method Based on Sequence Multiscale Feature Fusion Representation | |
Song et al. | A unified geolocation framework for web videos | |
CN116628247B (en) | Image recommendation method based on reinforcement learning and knowledge graph | |
Yang et al. | Transfer learning by reusing structured knowledge | |
CN114511024B (en) | Node classification method, apparatus, device, medium and computer program product |
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 | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20221018 |
|
AD01 | Patent right deemed abandoned |