CN102087730B - A kind of product user network establishing method and device - Google Patents
A kind of product user network establishing method and device Download PDFInfo
- Publication number
- CN102087730B CN102087730B CN200910188682.2A CN200910188682A CN102087730B CN 102087730 B CN102087730 B CN 102087730B CN 200910188682 A CN200910188682 A CN 200910188682A CN 102087730 B CN102087730 B CN 102087730B
- Authority
- CN
- China
- Prior art keywords
- link
- product
- network
- user
- information
- 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
- 238000004458 analytical method Methods 0.000 claims abstract description 35
- 238000004140 cleaning Methods 0.000 claims description 14
- 238000009795 derivation Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 5
- 238000000034 method Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 241000406668 Loxodonta cyclotis Species 0.000 description 1
- 230000003542 behavioural Effects 0.000 description 1
- 230000000875 corresponding Effects 0.000 description 1
- 238000001802 infusion Methods 0.000 description 1
- 230000002452 interceptive Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000005201 scrubbing Methods 0.000 description 1
Abstract
The invention discloses a kind of product user network establishing method and device, the method includes: obtain relation information, and described relation information includes user and product relation information, user and customer relationship information, product and product relation information；Generate according to described relation information with described user and the product relational network as node；Described relational network is carried out link prediction, it is thus achieved that relational network after prediction；Relational network after described prediction carrying out link analysis and obtains link analysis information, described link analysis information includes the one or more relevant information in described link and user's recommendation, Products Show or bundle sale.In this programme, relational network is carried out link prediction and analyzes acquisition the most accurately completely relational network.
Description
Technical field
The present invention relates to data processing field, particularly relate to a kind of product user network establishing method and device.
Background technology
In various products under current internet technology, a lot of Web is had to apply, such as instant messaging (IM,
Instant Messaging) and social network services (SNS, Social Networking Services)
Etc..And these Web application relates to the user of magnanimity, and may between user in such applications
There is also the strongest association.It addition, these application often extend some with apply relevant virtual
Product or service.
Summary of the invention
Embodiment of the present invention technical problem to be solved is, it is provided that a kind of product user network struction side
Method and device, can embody the pass between user, between product and between user and product to realize obtaining
The product user network of system.
To this end, embodiments provide a kind of product user network establishing method, including: obtain and close
Being information, described relation information includes user and product relation information, user and customer relationship information, product
Product and product relation information；Generate according to described relation information with described user and the product relation as node
Network；Described relational network is carried out link prediction, it is thus achieved that relational network after prediction；After described prediction
Relational network carries out link analysis and obtains link analysis information, and described link analysis information includes described chain
Connect recommend with user, one or more relevant information in Products Show or bundle sale.
Wherein, relational network carries out link prediction include: according to prediction rule, described relational network is entered
Row link prediction；Wherein, one or more during described prediction rule includes following rule: judge described
The dependency between two nodes in relational network, if described dependency is more than relevance threshold, then exists
A limit is added between said two node；Excavate the topological structure of described relational network, and revise institute accordingly
State the link of relational network；The structure of the known node being present in described relational network is carried out abstract,
And Structure learning one probabilistic model abstract to these, model derive and reasonably there is not structure, so
After further according to derive structure add link or leave out link.
Accordingly, the embodiment of the present invention additionally provides a kind of product user network struction device, including: close
Be acquisition module, be used for obtaining relation information, described relation information include user and product relation information,
User and customer relationship information, product and product relation information；Network generation module, for according to described
Relation information generates with described user and the product relational network as node；Link prediction module, for right
Described relational network carries out link prediction, it is thus achieved that relational network after prediction；Link analysis module, for right
After described prediction, relational network carries out link analysis acquisition link analysis information, in described link analysis information
Including the one or more relevant letter in described link and user's recommendation, Products Show or bundle sale
Breath.
Wherein, link prediction module is additionally operable to carry out linking prediction to described relational network according to prediction rule；
Wherein, one or more during described prediction rule includes following rule: judge in described relational network
Dependency between two nodes, if described dependency is more than relevance threshold, then at said two node
Between add a limit；Excavate the topological structure of described relational network, and revise described relational network accordingly
Link；The structure of the known node being present in described relational network is carried out abstract, and abstract to these
One probabilistic model of Structure learning, by model derive reasonably there is not structure, then further according to derivation
Structure add link or leave out link.
In the scheme described by the embodiment of the present invention, relational network is carried out link prediction and analyzes acquisition
The most accurately completely relational network, includes the user after screening in the network and closes with product
It is information, user and customer relationship information, product and product relation information.
Accompanying drawing explanation
Fig. 1 is a concrete schematic diagram of the product user network in the inventive method embodiment；
Fig. 2 is that an idiographic flow of the product user network establishing method in the inventive method embodiment shows
It is intended to；
Fig. 3 is that a concrete structure of the product user network struction device in present system embodiment shows
It is intended to；
Fig. 4 is a concrete structure schematic diagram of the network generation module in Fig. 3；
Fig. 5 is another concrete structure of the product user network struction device in present system embodiment
Schematic diagram；
Fig. 6 is another concrete structure of the product user network struction device in present system embodiment
Schematic diagram.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out
Clearly and completely describe, it is clear that described embodiment is only a part of embodiment of the present invention, and
It is not all, of embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art are not making
Go out the every other embodiment obtained under creative work premise, broadly fall into the scope of protection of the invention.
Product user network in embodiments of the present invention considers multiple relation, including: deposit between user
Linking relationship, such as be probably IM contact person, SNS community good friend between user, be friend, or
It is in same group etc；The linking relationship existed between user and product, such as user is likely to purchase
A kind of product, employs a kind of product, or has browsed a kind of product；The link existed between product
Relation, such as one product is probably the sub-product of another part product, or two pieces product belongs to one greatly
Classification.As it is shown in figure 1, be a schematic diagram of this network, it is positioned in figure aobvious in the square frame of the first half
Show is that (small circle representative products, the real line between circle represents and produces the relation between product and product
Linking relationship between product), the pass being shown that between user and user in the square frame of lower half in figure
System's (small circle represents user, and the real line between circle represents the linking relationship between user), and two
The line that virtually connects between square frame then represents the linking relationship between user and product.By the node of network is entered
Row link prediction, it is thus achieved that the relational network described in the embodiment of the present invention.
To this end, the embodiment of the present invention proposes a kind of product user network establishing method, as in figure 2 it is shown,
The method includes:
201, obtaining relation information, described relation information includes user and product relation information, user and use
Family relation information, product and product relation information.As included, user profile and product information are entered
Row process extract user use product information (such as user to the use time length of certain part product,
The number of infusion of financial resources), relation information between user (such as user be IM contact person, alumnus, with
Individual group), information (for example whether selling together) between product.
202, generate according to described relation information with described user and the product relational network as node.Such as root
According to the information architecture relational network extracted in back.By abstract to user and product for relational network
In node, and build the link in relational network according to the relation information that extracts.Meanwhile, one
Can increase cleaning and optimization process in a little embodiments, to obtaining preferably network, then this step can include
Following process:
A, according to expert info, described relation information is cleared up, it is thus achieved that the relation information after cleaning；I.e.
Unnecessary information, the manager etc. of such as product is cleared up according to expert info, experience；It addition, pass through
The information that expert info cleaning causes due to error in data is inaccurate.
B, build according to the relation information after cleaning with described user and the product relational network as node.
C, according to expert info or rule base, described relational network is optimized the network of personal connections obtained after optimizing
Network.Wherein, rule base may be empty when modeling (i.e. opening relationships network) for the first time, coming of information
Source is the frequent mode that modeling finds in the past.
203, described relational network is carried out link prediction, it is thus achieved that relational network after prediction.As, can basis
Prediction rule carries out link prediction to described relational network, and wherein link prediction can include deleting described relation
Link in network or interpolation link in described relational network；Described prediction rule includes in following rule
One or more:
1, the dependency between two nodes in described relational network, the shortest path of such as two nodes are judged
Distance, the common number etc. of consecutive points, if described dependency is more than relevance threshold, then in said two
A limit is added between node.
2, excavate the topological structure of described relational network, and revise the link of described relational network accordingly.Example
As can be first the Frequent tree mining of this network excavated, the topological structure of extraction subgraph, expanded to
In whole network, if adding a link to meet this topological structure, then by this link of interpolation.
3, the structure of the known node being present in described relational network is carried out abstract, and abstract to these
One probabilistic model of Structure learning, by model derive reasonably there is not structure, then further according to derivation
Structure add link or leave out link.
204, relational network after described prediction is carried out link analysis and obtains link analysis information, described link
Analysis information includes described link and or many in user's recommendation, Products Show or bundle sale
Individual relevant information.This link analysis includes, the link to the link deleted in back and interpolation is carried out
Classifying, judge the product user relation that these links represent, such as user is to buy product or browse product
Product, user and user become good friend or can join same colony etc..
Meanwhile, in order to improve rule base etc. further, can after step 204, according to described link analysis
Information and relation information obtain new Rule Information；And add described new Rule Information to described rule
Storehouse.So, network can be the most perfect in continuous renewal.
So, by introducing the relational network between user and product, according to the pass between user and product
System's (buy, use etc.), relation between user with user (communicate, friend relation etc.), utilization
Link Predicting Technique obtains relational network.
Accordingly, the embodiment of the present invention additionally provides a kind of product user network struction device, as it is shown on figure 3,
This product user network struction device 3 includes:
Relation acquisition module 31, is used for obtaining relation information, and described relation information includes that user is closed with product
It is information, user and customer relationship information, product and product relation information；
Network generation module 32, for generating with described user and product as node according to described relation information
Relational network；
Link prediction module 33, for carrying out link prediction, it is thus achieved that relation after prediction to described relational network
Network；
Link analysis module 34, obtains link point for relational network after described prediction carries out link analysis
Analysis information, described link analysis information includes described link and user's recommendation, Products Show or binding
One or more relevant information in sale.
Wherein, this link prediction module 33 can be additionally used in, according to prediction rule, described relational network carried out chain
Connect prediction.Described prediction rule includes one or more in following rule: judge in described relational network
Two nodes between dependency, if described dependency is more than relevance threshold, then save in said two
A limit is added between point；Excavate the topological structure of described relational network, and revise described relational network accordingly
Link；The structure of the known node being present in described relational network is carried out abstract, and these are taken out
One probabilistic model of the Structure learning of elephant, is derived by model and reasonably there is not structure, then further according to leading
The structure gone out is added link or leaves out link.
Network generation module 32 includes: cleaning submodule 321, is used for according to expert info described relation
Information is cleared up, it is thus achieved that the relation information after cleaning；Build submodule 322, after according to cleaning
Relation information builds with described user and the product relational network as node；Optimize submodule 323, for root
According to expert info or rule base, described relational network is optimized the relational network obtained after optimizing.Such as Fig. 4
Shown in.
As it is shown in figure 5, this product user network struction device 3 also can farther include rule acquisition module 35,
For obtaining new Rule Information according to described link analysis information and relation information；Rule adds module 36,
For adding described new Rule Information to described rule base.Wherein, rule acquisition module 35 can also
Corresponding relation information is obtained at Relation acquisition module 31.
In actual applications, this product user network struction device 3 can also be according to function forming by other
Mode, such as the said apparatus according to software programming custom realization as shown in Figure 6, this device includes such as lower module:
Product information storehouse: store the essential information of product, such as the feature of product, classification etc.；
User information database: store the basic document of user, behavioural information, product interactive information, such as IM
Subscriber data, good friend's situation, the purchase situation of product, service condition etc.；
Information extraction/cleaning module: clear up unnecessary letter according to expert info (including expertise etc.)
Breath, such as can be by manager's information removal etc. of product；It addition, cleared up due to data by expert info
The information that mistake causes is inaccurate；
Network struction device: according to the user that obtains of cleaning and product information and expert info (e.g., some
Product is bundle sale, certain form of people to be little interest to some kinds of product
Etc.) building product user network, there is linking relationship, such as user between part of nodes may use
Spend a product a period of time, be that a product pays certain amount of money etc., and the quantization given by expert
Standard and weight are by these factor superpositions, thus are formed with the weights on linking relationship node limit；
Link prediction: be modeled the network constructed according to relevant algorithm, is adding expert's letter
On the basis of breath and in the past institute's extracting rule, it was predicted that the link that link that may be present maybe may disappear；
Link analysis: be analyzed link newly added, that delete, specifically can be carried out point these links
Class, it is judged which kind of contact these links represent.Can derive which product from the result of link analysis can
With bundle sale, which user may use product maybe may run off, and which user can recommended become
The good friend of other user.Such as, link prediction may be predicted and will produce one between user A and user B
Limit, link analysis is then according to the information that the two user is concrete, such as age, sex etc., it is judged that A and B
Between possible existence relation etc..On the other hand, by importing user information database, link analysis can be led
Go out rule base, then which link is likely that there are, and such as which kind user is with the biggest probability
Use certain product.And these rule bases will play directive function to may model again later.
Expert info: be primarily referred to as the experience of business personnel, adds one for data scrubbing and model testing
A little constraints and guidance, such as, determine which data can be used, and which link needs to set up, and which link can not
Energy appearance etc..
Rule base: may be empty when first time modeling, the source of information be the frequent of modeling discovery in the past
Pattern, the most certain type of user often buys the product of certain kind, has bought the use of some category
Other particular kind of product is often bought at family.
Wherein, the relation of above-mentioned each module as shown in Figure 6, does not repeats.
Implement the embodiment of the present invention, owing to generating product user network according to multiple relation information, and to this
Network carries out link prediction and analyzes, it is possible to obtain accurately completely relation information.
Meanwhile, network model is obtained in embodiments of the present invention, it is only necessary to use this model (i.e. to close
It is network) solve the problem in multiple different business field simultaneously, reduce modeling cost.On the other hand, originally
The means that effectively model provided in inventive embodiments, set up more on the basis of integrated multi-source information
Accurate model.
Through the above description of the embodiments, those skilled in the art is it can be understood that arrive each reality
The mode of executing can add the mode of required general hardware platform by software and realize, naturally it is also possible to by firmly
Part.Based on such understanding, the portion that prior art is contributed by technique scheme the most in other words
Dividing and can embody with the form of software product, this computer software product can be stored in computer can
Read in storage medium, such as ROM/RAM, magnetic disc, CD etc., including some instructions with so that one
Computer equipment (can be personal computer, server, or the network equipment etc.) performs each to be implemented
The method described in some part of example or embodiment.
Embodiments described above, is not intended that the restriction to this technical scheme protection domain.Any
Amendment, equivalent and the improvement etc. made within the spirit of above-mentioned embodiment and principle, all should comprise
Within the protection domain of this technical scheme.
Claims (8)
1. a product user network establishing method, it is characterised in that the method includes:
Obtaining relation information, described relation information includes that user is closed with user with product relation information, user
It is information, product and product relation information；
Generate according to described relation information with described user and the product relational network as node；
Described relational network is carried out link prediction, it is thus achieved that relational network after prediction, described link prediction bag
Include, delete the link in described relational network or in described relational network, add link；
Relational network after described prediction is carried out link analysis and obtains link analysis information, described link analysis
Including, the link to the link deleted and interpolation is classified, is judged the product user that these links represent
Relation；Described link analysis information includes described link and user's recommendation, Products Show or binding pin
Sell the one or more relevant information in three.
2. the method for claim 1, it is characterised in that described described relational network is carried out chain
Connect prediction to include:
According to prediction rule, described relational network is carried out link to predict；
Wherein, one or more during described prediction rule includes following rule:
Judge the dependency between two nodes in described relational network, if described dependency is more than relevant
Property threshold value, then between said two node add a limit；
Excavate the topological structure of described relational network, and revise the link of described relational network accordingly；
The structure of the known node being present in described relational network is carried out abstract, and abstract to these
One probabilistic model of Structure learning, is derived by model and reasonably there is not structure, then further according to derivation
Structure is added link or leaves out link.
3. the method for claim 1, it is characterised in that described generate according to described relation information
Include with described user with the product relational network as node:
According to expert info, described relation information is cleared up, it is thus achieved that the relation information after cleaning；
Build according to the relation information after cleaning with described user and the product relational network as node；
According to expert info or rule base, described relational network is optimized the network of personal connections obtained after optimizing
Network.
4. method as claimed in claim 3, it is characterised in that described method also includes:
New Rule Information is obtained according to described link analysis information and relation information；
Add described new Rule Information to described rule base.
5. a product user network struction device, it is characterised in that described device includes:
Relation acquisition module, is used for obtaining relation information, and described relation information includes user and product relation
Information, user and customer relationship information, product and product relation information；
Network generation module, for generating with described user and product as node according to described relation information
Relational network；
Link prediction module, for carrying out link prediction, it is thus achieved that network of personal connections after prediction to described relational network
Network, wherein, described link prediction includes, deletes the link in described relational network or in described network of personal connections
Network adds link；
Link analysis module, obtains link analysis for relational network after described prediction carries out link analysis
Information, wherein, described link analysis includes, the link to the link deleted and interpolation is classified, sentenced
The product user relation that these links disconnected represent；Described link analysis information includes described link and user
One or more relevant information in recommendation, Products Show or bundle sale three.
6. device as claimed in claim 5, it is characterised in that described link prediction module is additionally operable to root
It is predicted that rule carries out link prediction to described relational network；
Wherein, one or more during described prediction rule includes following rule:
Judge the dependency between two nodes in described relational network, if described dependency is more than relevant
Property threshold value, then between said two node add a limit；
Excavate the topological structure of described relational network, and revise the link of described relational network accordingly；
The structure of the known node being present in described relational network is carried out abstract, and abstract to these
One probabilistic model of Structure learning, is derived by model and reasonably there is not structure, then further according to derivation
Structure is added link or leaves out link.
7. device as claimed in claim 6, it is characterised in that described network generation module includes:
Cleaning submodule, for clearing up described relation information according to expert info, it is thus achieved that after cleaning
Relation information；
Build submodule, for building with described user and product as node according to the relation information after cleaning
Relational network；
Optimize submodule, for described relational network being optimized acquisition according to expert info or rule base
Relational network after optimization.
8. device as claimed in claim 7, it is characterised in that described device also includes:
Rule acquisition module, for obtaining new rule letter according to described link analysis information and relation information
Breath；
Rule adds module, for adding described new Rule Information to described rule base.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910188682.2A CN102087730B (en) | 2009-12-08 | A kind of product user network establishing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910188682.2A CN102087730B (en) | 2009-12-08 | A kind of product user network establishing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102087730A CN102087730A (en) | 2011-06-08 |
| CN102087730B true CN102087730B (en) | 2016-12-14 |
Family
ID=
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1439995A (en) * | 2002-02-20 | 2003-09-03 | e－制造有限公司 | Management in providing chain product management based on stream management |
| CN101084496A (en) * | 2004-05-04 | 2007-12-05 | 波士顿咨询集团公司 | Method and apparatus for selecting, analyzing, and visualizing related database records as a network |
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1439995A (en) * | 2002-02-20 | 2003-09-03 | e－制造有限公司 | Management in providing chain product management based on stream management |
| CN101084496A (en) * | 2004-05-04 | 2007-12-05 | 波士顿咨询集团公司 | Method and apparatus for selecting, analyzing, and visualizing related database records as a network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106355506B (en) | Influence maximization initial node selection method in online social network | |
| US20130198191A1 (en) | Method for detecting communities in massive social networks by means of an agglomerative approach | |
| CN107145527A (en) | Link prediction method based on first path in alignment isomery social networks | |
| CN109598563A (en) | Brush single detection method, device, storage medium and electronic equipment | |
| CN107330445A (en) | The Forecasting Methodology and device of user property | |
| CN103310353B (en) | The data filtering of a kind of attack resistance optimizes system and method | |
| CN108765042A (en) | A kind of retail shop's proposed algorithm based on user Yu retail shop's network-physical space interbehavior | |
| Yeshwanth et al. | Evolutionary churn prediction in mobile networks using hybrid learning | |
| Jin et al. | Toward understanding and evaluating structural node embeddings | |
| CN107220745A (en) | A kind of recognition methods, system and equipment for being intended to behavioral data | |
| CN107832421A (en) | A kind of adaptive data splitting mining model system and method | |
| CN105590237A (en) | Application of high utility sequential pattern with negative-profit items in electronic commerce business decision making | |
| Deylami et al. | Link prediction in social networks using hierarchical community detection | |
| CN110443265A (en) | A kind of behavioral value method and apparatus based on corporations | |
| CN113706263A (en) | Electronic commerce system based on cloud platform | |
| Goldstein et al. | Group-based Yule model for bipartite author-paper networks | |
| Sharma et al. | A study of tree based machine learning techniques for restaurant reviews | |
| CN113868537B (en) | Recommendation method based on multi-behavior session graph fusion | |
| Bakirli et al. | DTreeSim: A new approach to compute decision tree similarity using re-mining | |
| Yu et al. | Predicting nft classification with gnn: A recommender system for web3 assets | |
| Xu et al. | A comprehensive review on recent developments in quality function deployment | |
| CN110413900A (en) | More social networks account matching process based on viterbi algorithm | |
| CN118071400A (en) | Application method and system based on graph computing technology in information consumption field | |
| Zaim et al. | A novel approach to dynamic profiling of E-customers considering clickstream data and online reviews. | |
| CN108345620A (en) | Brand message processing method, device, storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| CP02 | Change in the address of a patent holder |
Address after: 518000 Tencent Building, No. 1 High-tech Zone, Nanshan District, Shenzhen City, Guangdong Province, 35 Floors Patentee after: Shenzhen Tencent Computer System Co.,Ltd. Address before: 518057 5-10 building, Fiyta building, Gaoxin Nan Road, Nanshan District hi tech Zone, Shenzhen, Guangdong Patentee before: Shenzhen Tencent Computer System Co.,Ltd. |