CN106886524A - A kind of community network community division method based on random walk - Google Patents
A kind of community network community division method based on random walk Download PDFInfo
- Publication number
- CN106886524A CN106886524A CN201510938101.8A CN201510938101A CN106886524A CN 106886524 A CN106886524 A CN 106886524A CN 201510938101 A CN201510938101 A CN 201510938101A CN 106886524 A CN106886524 A CN 106886524A
- Authority
- CN
- China
- Prior art keywords
- node
- community
- label
- network
- similarity
- 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 57
- 238000005295 random walk Methods 0.000 title claims abstract description 30
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 44
- 239000011159 matrix material Substances 0.000 claims abstract description 22
- 238000010586 diagram Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 8
- 230000033228 biological regulation Effects 0.000 claims description 4
- 238000013508 migration Methods 0.000 claims description 4
- 230000005012 migration Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 230000006872 improvement Effects 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 2
- 238000001027 hydrothermal synthesis Methods 0.000 claims 1
- 238000012544 monitoring process Methods 0.000 abstract description 3
- 230000000644 propagated effect Effects 0.000 abstract description 3
- 238000009412 basement excavation Methods 0.000 abstract description 2
- 230000006854 communication Effects 0.000 abstract description 2
- 230000008859 change Effects 0.000 description 9
- 235000005156 Brassica carinata Nutrition 0.000 description 6
- 244000257790 Brassica carinata Species 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 238000005192 partition Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005354 coacervation Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000009474 immediate action Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000012795 verification 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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Primary Health Care (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of community network community division method based on random walk, it is related to community network field.Invention solves the problems, such as that conventional labels algorithm randomness during tag update is excessive.First, introduce random walk thought, be calculated it is a kind of weigh network node between similarity matrix；Secondly, it is not random selection when the frequency of occurrences is signed in neighbor node acceptance of the bid there is multiple highest in label communication process, but the tag update that selection similarity highest neighbor node is possessed, it is to avoid label is arbitrarily propagated between community；Finally, tested with live network and artificial baseline network, as a result show that the algorithm obtains more preferable performance than primal algorithm in community discovery.Community network group dividing method according to embodiments of the present invention, according to personal relationship similarity properties, by improving label propagation algorithm, user group is divided, division result all has preferable application value to network public-opinion monitoring, commercial user's excavation etc..
Description
Technical field
It is a kind of community network community division method based on random walk the present invention relates to community network field.
Background technology
Annexation according to network node can be divided into some communities, and internal node connection in community's is relatively tight
It is close, connect then more sparse between community.Community discovery has very heavy for network public-opinion monitoring, safe early warning, ecommerce etc.
The application value wanted.As the good friend that chat software is recommended belongs to same community, shopping website is recommended to the user of different communities
The commodity of different-style, public security system is monitored when the term frequencies such as heresy community " parade " are raised and taken an immediate action.To community
It was found that research, a large amount of reliable valuable information can be obtained.
The research of community discovery achieves sizable progress in recent years, and many scholars propose new theory and new method.
These methods can be largely classified into four classes：Figure dividing method, W-H algorithms, hierarchical clustering method and label propagation algorithm.Figure segmentation
Method is commonly used to computer realm, and it is based on iteration to a point technology：Divide every time and network is all divided into two optimal sons
Figure, subgraph is further continued for iteration to point, until quantity reaches requirement.Figure split plot design can substantially be divided into two classes：Based on Laplce
Spectral radius bisection method and Kerninghan-Lin algorithms.It has the disadvantage to be needed to obtain result by network to point every time
Want continuous iteration.To solve this problem, Wu and Huberman propose W-H algorithms：Two nodes of different communities are chosen, point
The initial point that voltage is 1 and the destination node that voltage is 0 not being set to, each edge resistance being set to 1, other nodes can obtain different
Magnitude of voltage.By the similar node division of magnitude of voltage to same community.W-H algorithm shortcomings are that community structure is must be known by before division
Part prior information, to ensure initial point and destination node not in same community.Hierarchical clustering method is according to the connection between node
Relation and similarity degree divide community, and the method can be divided into coacervation and disintegrating method again.Represent algorithm respectively G-N algorithms
With Newman fast algorithms, but the point extremely low due to there are many similarities in community, hierarchical clustering method often ignore these section
Point, final result is unsatisfactory.Label propagation algorithm (Label Propagation Algorithm, LPA) and former classes
Method is compared, it is not necessary to is known network structure or priori community structure, the propagation characteristic of network is only relied upon, with linear
Time complexity, it is very high that community divides efficiency.Cause the extensive concern of domestic and foreign scholars.Its time complexity is close to linear
O (m) (m is the number on side), detects, for fairly large community's (106-109 node) by starting to receive after 5 iteration
Hold back.In addition, label propagation algorithm need not both optimize predefined object function, it is not required that quantity and scale on community
Etc. prior information, the size to community is not also limited, thus label propagation algorithm to have become current application relatively broad
One of community discovery algorithm, is widely applied in the fields such as multimedia messages classification, virtual community excavation.
Label propagation algorithm precise and high efficiency, but in communication process, sign the frequency of occurrences when nodes neighbors acceptance of the bid and there is multiple most
Gao Shi, understand equality treats each node, randomly selects a highest label, and this randomness causes label in different communities
Between propagation, community division result is unstable, and randomness is stronger, and robustness has much room for improvement.In sum, existing community's hair
All there is very big room for promotion in the degree of accuracy and time complexity in existing method.
The content of the invention
It is an object of the invention to provide a kind of community network community division method based on random walk, the method is favourable
In the accuracy and stability that improve Web Community's division.
To achieve the above object, the technical scheme is that：A kind of community network community based on random walk divides
Method, comprises the following steps：
Step A：Read community network data, construction with network individuality as node, individuality between society of the correlation as side
Hand over network；
Step B：Random Walk Algorithm is improved, according to the superposition algorithm calculate node similarity matrix after improvement；
Step C：Initialization community：Upset node sequence, be that each user node distributes a label value, label value mark
Know the affiliated community of node；
Step D：Tag update, the tag update of x is by each label frequency of occurrences in the adjacent node of calculate node x：Go out
Existing frequency highest label, if label frequency has multiple highests, compares the node institute in the similarity matrix of adjacent node
The similar value being expert at, the label that the maximum node of selection similarity is possessed is updated, if in the presence of multiple maximum similar values,
Randomly select one；
Step E：Judge whether to meet stop condition：Reach regulation iterations or several times after iteration label value reaches
To stabilization.Otherwise, return to step 2 continues to update label；
Step F：All nodes with same label are classified as a community；
Further, in above-mentioned steps B, as follows is improved to Random Walk Algorithm：In traditional Random Walk Algorithm, only exist
1 random walk person walker, due to the Markov property of random walk, the final similarity for obtaining can exist it is very big not
Certainty, to eliminate error, the t walker of release of innovatory algorithm in given time period t with Δ t=1 in rapid succession.
Further, in above-mentioned steps B, the calculating to random walk similarity proposes new computing formula：
By social network diagram it is abstract be a simple non-directed graph, G (N, E), wherein, N represents the set of node, and E is represented
The set on side；
Use PxyWalker is represented from node x, trip makes a move and reaches the probability of y：
kxIt is the degree of node x, axyIt is respective value in adjacency matrix；
Use πxyT () represents that walker reaches the probability of y within node x migration t steps；πxyT () can be by a step
Transfer matrix recursion is obtained, πxT () is the column matrix of π (t) matrixes xth row；πxyT () is πxThe value of (t) matrix y rows.πx(0)
It is a vector for n*1 dimensions, x-th value is 1,
πx(t)=PTπx(t-1)
PTIt is the transposition of matrix P；
WithRepresent the random walk similarity between node x and y.Computing formula is as follows：
Connection between wherein | E | is nodes is total；
For a fixed network, its total side number | E | is fixed, therefore in calculating process, and 2 | E | can be with
It is ignored.The OLRW similarities of omission are proposed in step B similarity solution procedurees for this：
The calculating of above similarity is all based on what single walker was carried out, for improve rear stability superposition higher with
Machine migration algorithm, the OSRW calculating formula of similarity after superposition is as follows：
Further, in above-mentioned steps C, to ensure label equality of each node and updated when original state is distributed
Randomness in journey, upsets node sequence first, then for each user node distributes a label value, i.e. Cn=Ln, CnTable
Show the affiliated communities of node n, LnRepresent the label value of node n.
Further, in above-mentioned steps D, node label value updates detailed process and is：All summit iteration in figure are updated,
The label value of node is updated to the most label value of quantity in the label of its adjacent node；If the label frequency of occurrences exists multiple
Highest, the label that the maximum node of selection similarity is possessed is updated, if in the presence of multiple maximum similar values, randomly selecting
One.
Further, in above-mentioned steps E, stopping criterion for iteration is each node label of iterations or network for reaching regulation
No longer change.
Brief description of the drawings
Fig. 1 realizes flow chart for the inventive method.
Fig. 2 is the influence that different step number t divide the degree of accuracy to community.Using the inventive method (being represented with RWLPA) in section
It is 500 to count, and under conditions of hybrid parameter μ=0.6, when taking different value to t, the degree of accuracy that community divides is verified.μ is represented
Have the part on even side in different intercommunal nodes) value change comparison diagram.
Fig. 3 is the influence that different step number t divide the degree of accuracy to community.Using the inventive method (being represented with RWLPA) in section
It is 500 to count, and under conditions of hybrid parameter μ=0.65, when taking different value to t, the degree of accuracy that community divides is verified.μ tables
Show there is the part on even side in different intercommunal nodes) value change comparison diagram.
Fig. 4 is that Zachary ' s karate club data sets are divided using original tag propagation algorithm (being represented by LPA)
Result schematic diagram.
Fig. 5 is to Zachary ' s karate club data set division results using the inventive method (being represented by RWLPA)
Schematic diagram.
Fig. 6 is in 250 nodes using the inventive method (being represented by RWLPA) and label propagation algorithm (being represented by LPA)
Baseline network in, the average number of degrees of node in a network<k>=15, under the conditions of the maximum maxk=50 of the number of degrees, two kinds of calculations
The average NMI values that method runs 100 times change comparison diagram with the value of hybrid parameter μ (μ ∈ [0,0.90]).
Fig. 7 is in 500 nodes using the inventive method (being represented by RWLPA) and label propagation algorithm (being represented by LPA)
Baseline network in, the average number of degrees of node in a network<k>=15, under the conditions of the maximum maxk=50 of the number of degrees, two kinds of calculations
The average NMI values that method runs 100 times change comparison diagram with the value of hybrid parameter μ (μ ∈ [0,0.90]).
Fig. 8 is in 750 nodes using the inventive method (being represented by RWLPA) and label propagation algorithm (being represented by LPA)
Baseline network in, the average number of degrees of node in a network<k>=15, under the conditions of the maximum maxk=50 of the number of degrees, two kinds of calculations
The average NMI values that method runs 100 times change comparison diagram with the value of hybrid parameter μ (μ ∈ [0,0.90]).
Fig. 9 is in 1000 nodes using the inventive method (being represented by RWLPA) and label propagation algorithm (being represented by LPA)
Baseline network in, the average number of degrees of node in a network<k>=15, under the conditions of the maximum maxk=50 of the number of degrees, two kinds of calculations
The average NMI values that method runs 100 times change comparison diagram with the value of hybrid parameter μ (μ ∈ [0,0.90]).
Specific embodiment
Below in conjunction with accompanying drawing, the feature and advantage above-mentioned to the present invention make more detailed explanation.
Fig. 1 is a kind of community network community division method based on random walk of the present invention.As shown in figure 1, methods described
Comprise the following steps：
Step A：Community network data are read, is constructed with individuality as node, relation is the social network diagram on side between individuality.
It is identical to have in user using each user an as node in network such as in microblogging social networks
The a line as network of feature or viewpoint.Then, many communities with same characteristic features are formd, this is supervised to network public-opinion
Survey has great importance；In WWW, if it is known that a small amount of information of some webpages, it is possible to relevant with other webpage compositions
The company side of system, this is highly useful to search engine；In scientist's coorporative network, using each author an as node, two
Author's cooperation article is to produce a company side, forms huge collaborative network.
In the present embodiment, using 4 kinds of baseline networks being made up of different nodes, the node number of degrees and hybrid parameter.
Wherein, hybrid parameter μ represents the obvious degree of the community structure of community network, and the smaller community structure of μ values is more obvious.
Step B：Similarity matrix between calculate node.
Such as in microblogging social networks, if the walker of random walk is when walking within the specified time period, from
The probability that family A reaches user B is higher, and similarity is bigger, and the possibility that two users finally belong to same community is bigger.
Step C：Initialization community, flag node；It is one label for representing its affiliated community of each node distribution, iteration
Number of times t=1.
It is label detailed process that each node distribution one represents its affiliated community specifically, in the step C
For i.e. Cn=Ln, CnRepresent the affiliated communities of node n, LnRepresent the label value of node n.
Step D：Label starts iteration renewal, when the label frequency of occurrences has multiple highests, one is entered according to similarity matrix
Step is divided.
Step E：Judge whether to meet stop condition.
Specifically, in the step D, node label updates and specifically includes following steps：
Step D1：For each node x in sequence node X, with frequency of occurrences highest in its neighbor node label, that is marked
Sign to update the label of the node, it is assumed that the k neighbor node of node x is respectively x1, x2..., xk, the label of iteration node x at the t times
It is dependent in its neighbor node all labels that have passed through t iteration posterior nodal point and other have passed through institute after the t-1 times iteration node
The label for obtaining, the tag update formula of node x is：
Wherein function g return be node x neighbor node label in frequency highest label.
Step D2：If during the frequency highest that multiple labels occur, lookup possesses the node of highest label and is updated section
Similarity between point, the label for selecting most like node to possess is updated.
Step E：If the label of all nodes no longer changes, algorithm stops；Otherwise, t=t+1, and return to step E.
Specifically, in the step F, when the iterations specified is reached, algorithm will also stop.
Step F：All summits with same label are classified as a community.
In the present embodiment first by live network data set come the validity of verification algorithm.In Zachary ' s
Karate club data sets carry out verifying that Karate data sets include 34 members in karate club of one, the U.S., 78
Individual member contact.Contradiction of this 34 members due to two leaders each other generates division, as Liang Ge groups.Use module
Degree Q used as evaluation index, with average value tried one's best elimination error by experiment operation 100 times.Q is defined as follows：
| E | represents the total side number of non-directed graph, A in formulaijIt is adjacency matrix, kiIt is the number of degrees of node i, node i is with j same
δ=1 during community, otherwise δ=0.The modularity of community after computation partition, the Q values of RWLPA algorithm partitions are 0.3916, LPA algorithms
The Q values of division are 0.3896.Effect of the invention is better than traditional algorithm.
Secondly, in the investigation present invention and baseline network of the label propagation algorithm under the conditions of 4 kinds of different parameters, with mixing
The value change of parameter μ, the average NMI values comparison diagram that each operation is 100 times.Wherein, NMI (normalized mutual
Information) value is the judging basis of judgment experiment division result and the degree of closeness of actual legitimate reading.Specific rules are detailed
It is thin as follows：
Wherein, a confusion matrix N is defined, line number represents community content, and columns is represented and has found community, NijRepresent true
The shared node number in community j is had found of node in community i, cARepresent the quantity of community content, cBRepresentative has found society
The quantity in area, NiRepresent matrix NijThe summation of middle i rows, NjRepresent the summation of j row.
A kind of community network community division method based on random walk of the present invention, is reading by community's partition process
Community network data, calculate node similarity matrix, initialization community, tag update, division five stages of community are taken, is read first
Take community network data, construction is with individuality as node, social network diagram of the correlation as side between individuality；Phase between structure node
Like degree matrix；The all of node of random alignment, is one label for representing its affiliated community of each node distribution in network,
Iterations t=1；Node label updates, and for each node x in sequence node, occurs frequently with its neighbor node label
That label of rate highest updates the label of the node, if multiple labels occur frequency highest when, search similarity moment
Battle array, finds and is updated the label that the most like node of node possesses and be updated.If the label of all nodes no longer becomes
Change or reach the iterations of regulation, then algorithm stops；Finally, all summits with same label are classified as a society
Area.There is very big randomness in label propagation algorithm, have impact on the accuracy of community division result during tag update.
The present invention starts with from the randomness for suppressing label propagation, introduces random walk thought, the range formula definition based on random walk
A kind of new similarity calculating method, builds the similarity matrix between node.During label is propagated, work as nodes neighbors
It is no longer random selected when multiple highests occurs in middle label frequency, but the label that the most like node of selection is possessed is carried out more
Newly, node is effectively prevent in intercommunal any propagation, improves the degree of accuracy of community's division.In the present invention, exist
One parameter t, the selection of the parameter produces different influences to experimental result, under the conditions of have chosen 2 different parameters for this
Baseline network verifies that experiment proves that community division result is more accurate at 3≤t≤8 to the value of t.In order to prove this
The advantage of inventive method, embodiment have chosen live network and artificial baseline network is verified respectively, to Zachary ' s
The result that karate club data sets are divided shows that the modularity after division is higher than tradition LPA algorithms, and division result is more excellent
It is elegant.Baseline network under the conditions of 4 different parameters is tested, network parameter is different, as a result such as Fig. 6, Fig. 7, Fig. 8, Fig. 9 institute
Show.When hybrid parameter μ values are in the range of 0 to 0.9, the NMI values result of the inventive method is overall to be propagated better than conventional labels
Algorithm.Above-described embodiment shows RWLPA from the randomness for largely limiting label propagation, substantially increases community discovery
The accuracy rate of algorithm.To sum up, the inventive method substantially increases the accuracy rate of original community discovery algorithm, can effectively excavate society
Community structure in network, can apply to the field of the different scales such as network public-opinion monitoring, search engine.
Presently preferred embodiments of the present invention is the foregoing is only, it is carried out in invention claim limited range
It is changes, modifications, even equivalent, will all belong in protection scope of the present invention.
Claims (6)
1. a kind of community network community division method based on random walk, it is characterised in that the described method comprises the following steps：
Step A：Read community network data, construction with network individuality as node, individuality between social network of the correlation as side
Network figure；
Step B：Random Walk Algorithm is improved, according to the superposition algorithm calculate node similarity matrix after improvement；
Step C：Initialization community：Upset node sequence, be that each user node distributes a label value, label value mark section
Community belonging to point；
Step D：Tag update, the tag update of x is by each label frequency of occurrences in the adjacent node of calculate node x：There is frequency
Rate highest label, if label frequency has multiple highests, the node is expert in comparing the similarity matrix of adjacent node
Similar value, choose the label that the maximum node of similarity possessed and be updated, if in the presence of multiple maximum similar values, at random
Choose one；
Step E：Judge whether to meet stop condition：Reach regulation iterations or several times after iteration label value reaches surely
It is fixed；Otherwise, return to step 2 continues to update label；
Step F：All nodes with same label are classified as a community.
2. a kind of community network community division method based on random walk according to claim 1, it is characterised in that：
In above-mentioned steps B, a walker for migration, unstable result are only existed in original Random Walk Algorithm.To eliminate
The error that randomness in random walk process is caused, the release in given time period t with Δ t=1 in rapid succession
Walker, until first walker step number being released is t, similarity between two nodes is tried to achieve according to modified hydrothermal process.
3. a kind of community network community division method based on random walk according to claim 1, it is characterised in that：
In above-mentioned steps B, the calculating to random walk similarity proposes new criterion；Similarity matrix xth row y
The value a of rowxyThe similarity degree of node x and node y is represented, its value is bigger, represent the individuality representated by the two nodes same
The possibility of individual community is bigger, for the social network diagram constructed in step A, is abstracted into simple non-directed graph G (N, E),
Wherein, N represents the set of node, and E represents the set on side；The specific solution procedure of similarity is as follows between node：
Use PxyWalker is represented from node x, trip makes a move and reaches the probability of y：
kxIt is the degree of node x, axyIt is respective value in adjacency matrix；Use πxyT () represents that walker walks it from node x migration t
The probability of interior arrival y；πxyT () can be obtained by Matrix of shifting of a step recursion, πxT () is the column matrix of π (t) matrixes xth row；
πxyT () is πxThe value of (t) matrix y rows.πx(0) it is a vector for n*1 dimensions, x-th value is 1, PTIt is the transposition of matrix P；
πx(t)=PTπx(t-1)
WithThe random walk similarity between node x and y is represented, computing formula is as follows：
Connection between wherein | E | is nodes is total；
For a fixed network, its total side number | E | is fixed, therefore in calculating process, 2 | E | can be neglected
Slightly.The OLRW similarities of omission are proposed in step B similarity solution procedurees for this：
The calculating of above similarity is all based on what single walker was carried out, is swum at random for improving rear stability superposition higher
Algorithm is walked, the OSRW calculating formula of similarity after superposition is as follows：
4. a kind of community network community division method based on random walk according to claim 1, it is characterised in that：
It is random in the equality and renewal process of each node when original state is distributed to ensure label in above-mentioned steps C
Property, node sequence is upset first, then for each user node distributes a label value, i.e. Cn=Ln, CnRepresent node n institutes
Category community, LnRepresent the label value of node n.
5. a kind of community network community division method based on random walk according to claim 1, it is characterised in that：
In above-mentioned steps D, node label value updates detailed process and is：All summit iteration in figure are updated, by the mark of node
Label value is updated to the most label value of quantity in the label of its adjacent node；If the label frequency of occurrences has multiple highests, choose
The label that the maximum node of similarity is possessed is updated, if in the presence of multiple maximum similar values, randomly selecting one.
6. a kind of community network community division method based on random walk according to claim 1, it is characterised in that：
In above-mentioned steps E, stopping criterion for iteration is that each node label of iterations or social networks for reaching regulation is no longer sent out
Changing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510938101.8A CN106886524A (en) | 2015-12-15 | 2015-12-15 | A kind of community network community division method based on random walk |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510938101.8A CN106886524A (en) | 2015-12-15 | 2015-12-15 | A kind of community network community division method based on random walk |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106886524A true CN106886524A (en) | 2017-06-23 |
Family
ID=59173836
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510938101.8A Pending CN106886524A (en) | 2015-12-15 | 2015-12-15 | A kind of community network community division method based on random walk |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106886524A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107609063A (en) * | 2017-08-29 | 2018-01-19 | 重庆邮电大学 | A kind of the mobile phone application commending system and its method of multi-tag classification |
| CN107862618A (en) * | 2017-11-06 | 2018-03-30 | 郑州云海信息技术有限公司 | A kind of community discovery method and device based on label propagation algorithm |
| CN107870978A (en) * | 2017-09-30 | 2018-04-03 | 上海壹账通金融科技有限公司 | Relational network topological method and application server |
| CN108133234A (en) * | 2017-12-21 | 2018-06-08 | 广东工业大学 | Community detection method, device and equipment based on sparse subset selection algorithm |
| CN109657160A (en) * | 2018-12-29 | 2019-04-19 | 中国人民解放军国防科技大学 | Method and system for estimating incoming degree information based on random walk access frequency number |
| CN109784636A (en) * | 2018-12-13 | 2019-05-21 | 中国平安财产保险股份有限公司 | Fraudulent user recognition methods, device, computer equipment and storage medium |
| CN110033036A (en) * | 2019-04-04 | 2019-07-19 | 厦门小圈网络科技有限公司 | A kind of social networks classification method based on circle |
| CN110110154A (en) * | 2018-02-01 | 2019-08-09 | 腾讯科技（深圳）有限公司 | A kind of processing method of map file, device and storage medium |
| CN110377822A (en) * | 2019-06-25 | 2019-10-25 | 腾讯科技（深圳）有限公司 | Method, apparatus and electronic equipment for network characterisation study |
| CN110399595A (en) * | 2019-07-31 | 2019-11-01 | 腾讯科技(成都)有限公司 | A kind of method and relevant apparatus of text information mark |
| CN111523918A (en) * | 2019-02-02 | 2020-08-11 | 北京极智嘉科技有限公司 | Commodity clustering method, commodity clustering device, commodity clustering equipment and storage medium |
| CN111597665A (en) * | 2020-05-15 | 2020-08-28 | 天津科技大学 | Hierarchical network embedding method based on network partition |
| CN111667373A (en) * | 2020-06-08 | 2020-09-15 | 上海大学 | Evolution community discovery method based on neighbor subgraph social network dynamic increment |
| CN112465301A (en) * | 2020-11-06 | 2021-03-09 | 山东大学 | Edge smart power grid cooperation decision method based on differential privacy mechanism |
| CN112926999A (en) * | 2021-03-26 | 2021-06-08 | 东南大学 | Self-adaptive recommendation method and system for associated supervision range of bulk commodity transaction platform |
| CN112989209A (en) * | 2021-05-10 | 2021-06-18 | 腾讯科技（深圳）有限公司 | Content recommendation method, device and storage medium |
| CN113011471A (en) * | 2021-02-26 | 2021-06-22 | 山东英信计算机技术有限公司 | Social group dividing method, social group dividing system and related devices |
| CN113609880A (en) * | 2021-07-06 | 2021-11-05 | 燕山大学 | Method for identifying outlier misplaced RFID (radio frequency identification) tags and system for implementing method |
| CN113723583A (en) * | 2021-08-28 | 2021-11-30 | 重庆理工大学 | Multi-domain network community discovery method based on discrete time quantum migration |
| CN114513426A (en) * | 2022-03-02 | 2022-05-17 | 郑州轻工业大学 | CCN community division method based on node similarity and influence |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102857525A (en) * | 2011-06-28 | 2013-01-02 | 安徽大学 | Community discovery method on basis of random walk strategy |
| CN103699617A (en) * | 2013-12-16 | 2014-04-02 | 西安交通大学 | Community discovery method based on random walk |
| CN104657901A (en) * | 2015-01-14 | 2015-05-27 | 重庆邮电大学 | Community discovery method based on label propagation in random walk |
-
2015
- 2015-12-15 CN CN201510938101.8A patent/CN106886524A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102857525A (en) * | 2011-06-28 | 2013-01-02 | 安徽大学 | Community discovery method on basis of random walk strategy |
| CN103699617A (en) * | 2013-12-16 | 2014-04-02 | 西安交通大学 | Community discovery method based on random walk |
| CN104657901A (en) * | 2015-01-14 | 2015-05-27 | 重庆邮电大学 | Community discovery method based on label propagation in random walk |
Non-Patent Citations (4)
| Title |
|---|
| WEIPING LIU 等: "Link prediction based on local random walk", 《EUROPHYSICS LETTERS》 * |
| 刘栋 等: "基于隐马尔可夫随机场的社区结构发现算法", 《计算机工程与设计》 * |
| 康旭彬 等: "一种改进的标签传播快速社区发现方法", 《合肥工业大学学报（自然科学版）》 * |
| 张贤坤 等: "一种基于局部回路的标签传播改进算法", 《计算机应用与软件》 * |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107609063A (en) * | 2017-08-29 | 2018-01-19 | 重庆邮电大学 | A kind of the mobile phone application commending system and its method of multi-tag classification |
| CN107609063B (en) * | 2017-08-29 | 2020-03-17 | 重庆邮电大学 | Multi-label classified mobile phone application recommendation system and method thereof |
| CN107870978A (en) * | 2017-09-30 | 2018-04-03 | 上海壹账通金融科技有限公司 | Relational network topological method and application server |
| CN107862618A (en) * | 2017-11-06 | 2018-03-30 | 郑州云海信息技术有限公司 | A kind of community discovery method and device based on label propagation algorithm |
| CN108133234A (en) * | 2017-12-21 | 2018-06-08 | 广东工业大学 | Community detection method, device and equipment based on sparse subset selection algorithm |
| CN108133234B (en) * | 2017-12-21 | 2021-03-16 | 广东工业大学 | Sparse subset selection algorithm-based community detection method, device and equipment |
| CN110110154B (en) * | 2018-02-01 | 2023-07-11 | 腾讯科技（深圳）有限公司 | Graph file processing method, device and storage medium |
| CN110110154A (en) * | 2018-02-01 | 2019-08-09 | 腾讯科技（深圳）有限公司 | A kind of processing method of map file, device and storage medium |
| CN109784636A (en) * | 2018-12-13 | 2019-05-21 | 中国平安财产保险股份有限公司 | Fraudulent user recognition methods, device, computer equipment and storage medium |
| CN109657160A (en) * | 2018-12-29 | 2019-04-19 | 中国人民解放军国防科技大学 | Method and system for estimating incoming degree information based on random walk access frequency number |
| CN109657160B (en) * | 2018-12-29 | 2023-01-06 | 中国人民解放军国防科技大学 | Method and system for estimating incoming degree information based on random walk access frequency number |
| CN111523918B (en) * | 2019-02-02 | 2023-09-19 | 北京极智嘉科技股份有限公司 | Commodity clustering method, device, equipment and storage medium |
| CN111523918A (en) * | 2019-02-02 | 2020-08-11 | 北京极智嘉科技有限公司 | Commodity clustering method, commodity clustering device, commodity clustering equipment and storage medium |
| CN110033036A (en) * | 2019-04-04 | 2019-07-19 | 厦门小圈网络科技有限公司 | A kind of social networks classification method based on circle |
| CN110033036B (en) * | 2019-04-04 | 2022-05-17 | 厦门小圈网络科技有限公司 | Circle-based social relationship classification method |
| CN110377822A (en) * | 2019-06-25 | 2019-10-25 | 腾讯科技（深圳）有限公司 | Method, apparatus and electronic equipment for network characterisation study |
| CN110377822B (en) * | 2019-06-25 | 2024-04-16 | 腾讯科技（深圳）有限公司 | Method and device for network characterization learning and electronic equipment |
| CN110399595A (en) * | 2019-07-31 | 2019-11-01 | 腾讯科技(成都)有限公司 | A kind of method and relevant apparatus of text information mark |
| CN110399595B (en) * | 2019-07-31 | 2024-04-05 | 腾讯科技(成都)有限公司 | Text information labeling method and related device |
| CN111597665A (en) * | 2020-05-15 | 2020-08-28 | 天津科技大学 | Hierarchical network embedding method based on network partition |
| CN111667373A (en) * | 2020-06-08 | 2020-09-15 | 上海大学 | Evolution community discovery method based on neighbor subgraph social network dynamic increment |
| CN111667373B (en) * | 2020-06-08 | 2023-08-11 | 上海大学 | Evolution community discovery method based on dynamic increment of neighbor subgraph social network |
| CN112465301A (en) * | 2020-11-06 | 2021-03-09 | 山东大学 | Edge smart power grid cooperation decision method based on differential privacy mechanism |
| CN112465301B (en) * | 2020-11-06 | 2022-12-13 | 山东大学 | Edge smart power grid cooperation decision method based on differential privacy mechanism |
| CN113011471A (en) * | 2021-02-26 | 2021-06-22 | 山东英信计算机技术有限公司 | Social group dividing method, social group dividing system and related devices |
| CN112926999B (en) * | 2021-03-26 | 2023-10-10 | 东南大学 | Self-adaptive recommendation method and system for association supervision range of bulk commodity transaction platform |
| CN112926999A (en) * | 2021-03-26 | 2021-06-08 | 东南大学 | Self-adaptive recommendation method and system for associated supervision range of bulk commodity transaction platform |
| CN112989209A (en) * | 2021-05-10 | 2021-06-18 | 腾讯科技（深圳）有限公司 | Content recommendation method, device and storage medium |
| CN113609880B (en) * | 2021-07-06 | 2024-04-19 | 燕山大学 | Method for identifying outlier misplaced RFID (radio frequency identification) tag and system for realizing method |
| CN113609880A (en) * | 2021-07-06 | 2021-11-05 | 燕山大学 | Method for identifying outlier misplaced RFID (radio frequency identification) tags and system for implementing method |
| CN113723583A (en) * | 2021-08-28 | 2021-11-30 | 重庆理工大学 | Multi-domain network community discovery method based on discrete time quantum migration |
| CN113723583B (en) * | 2021-08-28 | 2023-07-21 | 重庆理工大学 | Multi-domain network community discovery method based on discrete time quantum migration |
| CN114513426B (en) * | 2022-03-02 | 2023-09-15 | 郑州轻工业大学 | CCN community division method based on node similarity and influence |
| CN114513426A (en) * | 2022-03-02 | 2022-05-17 | 郑州轻工业大学 | CCN community division method based on node similarity and influence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106886524A (en) | A kind of community network community division method based on random walk | |
| CN110532436B (en) | Cross-social network user identity recognition method based on community structure | |
| Kumar et al. | IM-ELPR: Influence maximization in social networks using label propagation based community structure | |
| Cho et al. | Authority-shift clustering: Hierarchical clustering by authority seeking on graphs | |
| CN105893382A (en) | Priori knowledge based microblog user group division method | |
| CN105893381A (en) | Semi-supervised label propagation based microblog user group division method | |
| CN111428323A (en) | Method for identifying group of key nodes by using generalized discount degree and k-shell in complex network | |
| CN105335438A (en) | Local shortest loop based social network group division method | |
| Cai et al. | An improved random walk based clustering algorithm for community detection in complex networks | |
| Chang et al. | Relative centrality and local community detection | |
| Wang et al. | Integrating clustering and ranking on hybrid heterogeneous information network | |
| Rossi et al. | Hone: Higher-order network embeddings | |
| Zhou et al. | A similarity-based community detection method with multiple prototype representation | |
| Tang et al. | An adaptive discrete particle swarm optimization for influence maximization based on network community structure | |
| Su et al. | A new random-walk based label propagation community detection algorithm | |
| Sahu et al. | A neighbour-similarity based community discovery algorithm | |
| Emadi Kouchak et al. | Graph entropies-graph energies indices for quantifying network structural irregularity | |
| Lu et al. | From Comparing Clusterings to Combining Clusterings. | |
| Qiang et al. | A method of personalized recommendation based on multi-label propagation for overlapping community detection | |
| Aghaalizadeh et al. | Improving the stability of label propagation algorithm by propagating from low-significance nodes for community detection in social networks | |
| Zhou et al. | A graph clustering algorithm using attraction-force similarity for community detection | |
| Cetin et al. | A new network-based community detection algorithm for disjoint communities | |
| Sangamuang | A graph-based algorithm for interpersonal ties clustering in signed networks | |
| Baofang | A cultural algorithm based on artificial bee colony optimization for community detection in signed social networks | |
| Dong et al. | Bi-scale link prediction on networks |
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 |
Application publication date: 20170623 |
|
| RJ01 | Rejection of invention patent application after publication |