CN107609984A - The method for digging of social networks deep structure - Google Patents
The method for digging of social networks deep structure Download PDFInfo
- Publication number
- CN107609984A CN107609984A CN201711050005.5A CN201711050005A CN107609984A CN 107609984 A CN107609984 A CN 107609984A CN 201711050005 A CN201711050005 A CN 201711050005A CN 107609984 A CN107609984 A CN 107609984A
- Authority
- CN
- China
- Prior art keywords
- user
- circle
- social
- social networks
- incidence relation
- 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
Abstract
The present invention provides a kind of method for digging of social networks deep structure, belongs to social networks studying technological domain.This method builds the network speech propagation model based on thermonuclear first, according to the personal information of user, the state delivered, thumbs up and the social network data such as comments on and obtain the incidence relation between user.Then structure introduces the LDA models of incidence relation, traditional topic model Smooth LDA are transformed, it is set to make full use of the incidence relation between user, obtain the circle distribution of user, not only include the explicit circle and its probability belonging to user, also can be according to its potential circle of the interest digging of user.This method is particularly suitable for this kind of scene of social networks, is adapted to parallelization and incremental update, and be easily controlled the relative stability of solution afterwards before the update.
Description
Technical field
The present invention relates to social networks studying technological domain, particularly relates to a kind of excavation side of social networks deep structure
Method.
Background technology
The basic logic model of social networks can be write as G=(V, E), and wherein G represents network in itself, and V represents all
User node, E are the annexations between node, and they are all changed over time.The present invention analyzes the deep structure of the network, tool
Body includes information flow, user's circle etc..These the deep informations provide the more information on user, such as some user's rows
For root etc..In addition, by the deep structure of social networks, the accurate change of each network parameter can also be observed, for
Family, which provides, suggests and provides recommendation.Most important of which is that the circle model of social networks.In circle mentioned here refers not only to
There is close-connected circle in portion, also has similar tastes and interests comprising those, temporarily also without the user group of many annexations.
At present, the modeling of social networks is mostly based on network topology and the behavioral statisticses of user, by some
The algorithm of graph theory quantifies the basic structure of network.For example, based on the user data in Google+, the structure of social networks is studied
With differentiation [1];Infer various social networks by building the schematic models comprising social tension's Theoretical Considerations
Social networks [2];The information network of different range is analyzed and designed distributed information retrieval experiment, study each network
Between structure and network structure for retrieve performance importance [3].As circle with Fiel, social networks also can
It is divided into inner circle of people one by one, each inner circle of people often there are some general character, and inside is completely embedded, and is connected between circle loose.Cause
This, the excavation of social circle is the emphasis of social network structure research.It is used for excavating the Girvan-Newman of unity structure in graph theory
Algorithm and its mutation are then usually utilized to find the circle [4] in social networks.
Because social networks is a large scale network changed over time, it is desirable to mining algorithm has time stability,
Incremental update, and concurrency etc..In addition, a user may belong to multiple circles simultaneously, it is therefore desirable to provide user and belong to
The circle distribution of the probability of each circle, referred to as user.But the mining algorithm on circle belongs to greatly hard cluster at present
Algorithm, it is difficult to accurately provide the probability that user belongs to some circle.The clustering algorithms such as conventional Kmeans do not guarantee that stability,
Especially in the case of circle structure is not obviously.And spectral clustering scheduling algorithm is gone back in terms of large-scale parallel and incremental update
In the presence of certain difficulty.So the present invention proposes the soft clustering schemes of the increment that can provide probability.In addition, the present invention wishes
Circle can not only reflect the stronger user of those social relationships, for the weaker user of those social relationships, it is desirable to by it certainly
The attribute of body obtains the distributed intelligence of its potential circle, on the other hand, conventional method can not provide unified framework.
Topic model is very effective soft clustering method, but its most suitable operative scenario is document formula model, can only
Clustered by the key element (such as word of composition document) of object in itself, have no idea directly to utilize the pass between object
Connection relation.The present invention builds the production topic model for the support user-association relation excavated for circle, with reference to topic model
Advantage that is loose and explicitly providing probability, makes full use of the incidence relation between user, can not only provide explicit belonging to user
Circle and its probability, also can be in Unified frame according to its potential circle of the interest digging of user.
Bibliography:
[1]Gong N.Z.,Xu W.,Huang L.,Mittal P.,Stefanov E.,Sekar V.,Song D.:
Evolution of Social-Attribute Networks:Measurements,Modeling,and Implications
using Google+.ACM conference on Internet measurement conference,pp.131-144,
2012.
[2]Tang J.,Lou T.,Kleinberg J.:Inferring Social Ties across
Heterogeneous Networks.ACM International Conference on Web Search and Data
Mining.pp.743-752,2012.
[3]Ke W.,Mostafa J.:Scalability of findability:effective and
efficient IR operations in large information networks.SIGIR,pp.74-81,2010.
[4]Girvan M.,Newman M.E.J.:Community structure in social and
biological networks,Proceedings of the National Academy of Science,99(12):
7821–7826,2002.
[5]M Kitsak,LK Gallos,S Havlin,et al:Identification of influential
spreaders in complex networks.Nature Physics 6(11),888-893.
The content of the invention
The technical problem to be solved in the present invention is to provide a kind of method for digging of social networks deep structure, pass through user's
Social network data, particularly user behavior data, excavate user's circle and be distributed this deep layer network structure information.Circle is distributed
Refer to that each user belongs to the probability of each circle.Here circle both includes the circle that actual incidence relation is formed
Son, the implicit circle also determined including user property.
This method comprises the following steps:
(1) the network speech propagation model based on thermonuclear is built:The social network data of user is inputted, using heat propagation pair
The speech of network is propagated and analysed in depth, and obtains the incidence relation between user;In view of propagation and the thing of social networks speech
Heat propagation process is closely similar in Neo-Confucianism, and the latter is that its property has had deep the problem of fully research in physics
Theory is investigated, and is analysed in depth so being propagated using its speech to network, is obtained accurate user-association relation data.
(2) structure introduces the LDA models of incidence relation:Traditional topic model SmoothLDA is transformed, introduces step
Suddenly the user-association variable J obtained in (1) by the incidence relation between user, the circle distribution of user is obtained.
Wherein, in step (1) user social network data include personal information, deliver state, thumb up and comment on.
The incidence relation between user is obtained in step (1), span is [0,1].
Obtained in step (2) circle distribution can not only reflect the stronger user of social relationships, for social relationships compared with
Weak user, the distributed intelligence of its potential circle can be also obtained by the attribute of its own.
The above-mentioned technical proposal of the present invention has the beneficial effect that:
Network speech propagation model based on thermonuclear makes full use of the association between user social contact Network Data Capture user
Relation.Compared [5] with the method based on transmission model (SIR) of classics, thermonuclear model has clearly analytic structure, energy
Enough the interior of grid of accurately portraying accumulate characteristic, and can accurately provide the spread state of any time.The branch excavated for circle
Hold the production topic model of user-association relation, the soft clustering schemes of increment of probability can be provided, with reference to topic model loosely and
The explicit advantage for providing probability, makes full use of the incidence relation between user, obtains the explicit circle of user and the distribution of potential circle.
In addition, the model is particularly suitable for this kind of scene of social networks, it is adapted to parallelization and incremental update, and rear easily control before the update
Make the relative stability of solution.
Brief description of the drawings
Fig. 1 is the method for digging flow chart of the social networks deep structure of the present invention;
Fig. 2 is present invention introduces the LDA model schematics of incidence relation.
Embodiment
To make the technical problem to be solved in the present invention, technical scheme and advantage clearer, below in conjunction with accompanying drawing and tool
Body embodiment is described in detail.
The present invention provides a kind of method for digging of social networks deep structure.As shown in figure 1, this method step is as follows:(1)
Build the network speech propagation model based on thermonuclear:The social network data of user is inputted, using speech of the heat propagation to network
Propagation is analysed in depth, and obtains the incidence relation between user;
(2) structure introduces the LDA models of incidence relation:Traditional topic model SmoothLDA is transformed, introduces step
Suddenly the user-association variable J obtained in (1) by the incidence relation between user, the circle distribution of user is obtained.
In practical operation, input data is social network user information, including the personal information of user, user are delivered
Status information, user such as thumb up and commented at the user behavior data, export the circle distribution for user, i.e., each user belongs to each
The probability of individual circle.Concretely comprise the following steps:
A. the network speech propagation model based on thermonuclear.The speech for analyzing network first is propagated, and the present invention borrows physics
In thermonuclear (Heat kernel) concept, it is described in a heat propagation medium, single-point to one impact heat after, it is whole
Follow-up heat distribution in individual medium.For the angle of discretization, it is described in a cum rights grid, and certain moment is to single
One impact heat of point, propagation of the heat on all mesh points.This process is propagated very similar with message.Every user is
One single-point, being inputted the personal information of user as static vector, every bar state that user delivers is an impact heat, its
The thumbing up of his user, comment on and forwarding is heat propagation process, prevalence is incremented by successively.According to above-mentioned input, thermonuclear is constructed
Analytic solutions, so as to accurately be simulated to the wide-scale distribution of message.The incidence relation between user is obtained using the model, is taken
Value scope is [0,1].Meanwhile tracking is propagated from existing message, the structure of thermonuclear counter can be pushed away, so as to incidence relation
Further amendment.
B. structure introduces the LDA models of incidence relation, and schematic diagram is as shown in Figure 2.Inspired by LDA models, structure of the present invention
The production that with the addition of an incidence relation topic model is built, the incidence relation between user is also served as into model one can
Observational variable.The model improves from Smooth LDA, on original Smooth LDA bases, introduces user-association
Variable J, represent the power of incidence relation between two users, be observable variable, this data from the result of step (a) and
Come.In figure, " word " can refer to user property in a certain respect or user deliver article, provide comment, like
Label etc.;α is a prior distribution of user's circle distribution;β is the prior distribution that word is distributed in each circle;θ
It is the circle distribution of user;φ is the word distribution of certain circle;Z is the circle of certain user;W is some word observed;γ
It is the distribution of user-association, can be obtained from the result of (a).α, β, θ, φ, Z are hidden variable.For convenience of calculation, prior distribution takes
Dirichlet prior.LDA class models are very easy to parallelization, and the iterative process that its classics solves also contains incremental update very
Easily.So the model is particularly suitable for this kind of scene of social networks, it is adapted to parallelization and incremental update, and hold afterwards before the update
The relative stability of solution easy to control.Secondly, the circle provided by this model can not only reflect the stronger user of those social relationships,
For the weaker user of those social relationships, the distributed intelligence of its potential circle can be also obtained by the attribute of its own.
Described above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, on the premise of principle of the present invention is not departed from, some improvements and modifications can also be made, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (4)
- A kind of 1. method for digging of social networks deep structure, it is characterised in that:Comprise the following steps:(1) the network speech propagation model based on thermonuclear is built:The social network data of user is inputted, using heat propagation to network Speech propagate analysed in depth, obtain the incidence relation between user;(2) structure introduces the LDA models of incidence relation:Traditional topic model Smooth LDA are transformed, introduce step (1) the user-association variable J obtained in by the incidence relation between user, the circle distribution of user is obtained.
- 2. the method for digging of social networks deep structure according to claim 1, it is characterised in that:In the step (1) The social network data of user include personal information, deliver state, thumb up and comment on.
- 3. the method for digging of social networks deep structure according to claim 1, it is characterised in that:In the step (1) The incidence relation between user is obtained, span is [0,1].
- 4. the method for digging of social networks deep structure according to claim 1, it is characterised in that:In the step (2) Obtained circle distribution can not only reflect the stronger user of social relationships, for the weaker user of social relationships, can also lead to Cross the distributed intelligence that the attribute of its own obtains its potential circle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711050005.5A CN107609984A (en) | 2017-10-31 | 2017-10-31 | The method for digging of social networks deep structure |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711050005.5A CN107609984A (en) | 2017-10-31 | 2017-10-31 | The method for digging of social networks deep structure |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107609984A true CN107609984A (en) | 2018-01-19 |
Family
ID=61085012
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711050005.5A Pending CN107609984A (en) | 2017-10-31 | 2017-10-31 | The method for digging of social networks deep structure |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107609984A (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102414706A (en) * | 2009-03-03 | 2012-04-11 | 谷歌公司 | Adheat advertisement model for social network |
CN103106616A (en) * | 2013-02-27 | 2013-05-15 | 中国科学院自动化研究所 | Community detection and evolution method based on features of resources integration and information spreading |
CN105373531A (en) * | 2015-12-09 | 2016-03-02 | 微梦创科网络科技(中国)有限公司 | Short topic text identification method and device based on social network |
-
2017
- 2017-10-31 CN CN201711050005.5A patent/CN107609984A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102414706A (en) * | 2009-03-03 | 2012-04-11 | 谷歌公司 | Adheat advertisement model for social network |
CN103106616A (en) * | 2013-02-27 | 2013-05-15 | 中国科学院自动化研究所 | Community detection and evolution method based on features of resources integration and information spreading |
CN105373531A (en) * | 2015-12-09 | 2016-03-02 | 微梦创科网络科技(中国)有限公司 | Short topic text identification method and device based on social network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105260474B (en) | A kind of microblog users influence power computational methods based on information exchange network | |
CN109902203A (en) | The network representation learning method and device of random walk based on side | |
Zhai et al. | Random forest based traffic classification method in SDN | |
CN110297990A (en) | The associated detecting method and system of crowdsourcing marketing microblogging and waterborne troops | |
Wang et al. | Exploring all-author tripartite citation networks: A case study of gene editing | |
Fang et al. | A survey of social network and information dissemination analysis | |
Xu | Cultural communication in double-layer coupling social network based on association rules in big data | |
Jin et al. | Association rules redundancy processing algorithm based on hypergraph in data mining | |
CN104199947A (en) | Important person speech supervision and incidence relation excavating method | |
Li et al. | An epidemic model for correlated information diffusion in crowd intelligence networks | |
Shang | Mean commute time for random walks on hierarchical scale-free networks | |
Nasution | Social network mining: A discussion | |
Li et al. | Enhancing link prediction using gradient boosting features | |
CN107609984A (en) | The method for digging of social networks deep structure | |
CN103336865B (en) | A kind of dynamic communication network construction method and device | |
Wang et al. | DHCF: Dual disentangled-view hierarchical contrastive learning for fake news detection on social media | |
Jin et al. | Fitting social network models using varying truncation stochastic approximation MCMC algorithm | |
Sun et al. | Hasse diagram based algorithm for continuous temporal subgraph query in graph stream | |
Wang | RETRACTED: Application of Keras neural network in the era of big data | |
Yang et al. | Meta path-based information entropy for modeling social influence in heterogeneous information networks | |
Tian | Design and implementation of distributed government audit system based on multidimensional online analysis | |
Yuan et al. | OLAP4R: A top-k recommendation system for OLAP Sessions | |
Wang et al. | Detecting influential nodes incrementally and evolutionarily in online social networks | |
Wan et al. | A community discovering method based on event network for topic detection | |
Xu et al. | Research on the computer informatization in multimedia public opinion monitoring |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180119 |