CN105337759B - It is a kind of based on inside and outside community structure than measure and community discovery method - Google Patents
It is a kind of based on inside and outside community structure than measure and community discovery method Download PDFInfo
- Publication number
- CN105337759B CN105337759B CN201510526277.2A CN201510526277A CN105337759B CN 105337759 B CN105337759 B CN 105337759B CN 201510526277 A CN201510526277 A CN 201510526277A CN 105337759 B CN105337759 B CN 105337759B
- Authority
- CN
- China
- Prior art keywords
- sub
- community
- network structure
- outside
- network
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
Abstract
It is inside and outside than module based on, than measure and community discovery method, first defining inside and outside community structure that the invention discloses a kind of, for judge sub-network structure whether community's tightness degree of community and the structure.Then it proposes neighbour bidirectional iteration algorithm, using one group of initial subnet network structure after optimization, by putting both direction in increasing neighbors or reducing, based on inside and outside than module, comes iteration discovery community.The present invention community structure high for tightness degree in quickly discovery network, it can faster, more fully find better community, 39.64% is averagely improved than upper in the time for finding best community earliest, 12.67% is averagely improved on search covering surface, and have the characteristics that calculating data dependence degree is low, it is suitble to Distributed Parallel Computing.
Description
Technical field
The present invention relates to community discovery method in community network, especially a kind of inside and outside ratio measurement side based on community structure
Method and community discovery method.
Background technique
Community network (Social Network) can be abstracted as, the set being made of node and line segment, and node is society
Individual in network, line segment be then individual and individual between certain contact relationship.Community network has a kind of community characteristics, i.e.,
A sub-network structure in network, the connection tightness degree of the inside configuration between points is higher, and the structure and its outside
The connection of portion's neighbors is more open.Community discovery (Community Detection) is exactly to find out the internal connection of these in network
It is close sub-network structure.Community discovery technology has many important roles, than as being used in recommender system, finding out has
The people of same interest hobby, and recommend them may interested product.
Traditional network community discovery method needs the global information of whole network, i.e., all points and side in network, then
Community's tightness degree of the sub-network structure is obtained by more complex calculation formula.And judgement to community structure or
Lay particular emphasis on while while between degree of contact or lay particular emphasis on degree of contact between points, considered without balancing simultaneously
Two aspects in side and point.
One feature of community network is that network can be in dynamic change, and the quantity of nodes and line segment can not stop
Change, i.e., the global information of network can dynamic change, if it is determined that need global network information when community structure, that just can only be
It is calculated in the static network structure at a certain moment, and is not applied for dynamic social network.
In recent years, graphics processing unit GPU is as a kind of parallel-processing hardware platform for being applicable to scientific algorithm, simultaneously
Row calculates aspect and has obtained more application.It is multithreading, the multi-core processing of a highly-parallel relative to CPU, GPU
Device.CPU processor more hardware resource for controlling, storage unit, be good at the calculating of complex control logic, and GPU
More hardware resources are then used for parallel computation unit, parallel computation can be carried out to large-scale data with a large amount of threads, be suitble to
Processing calculates density height, logic branch simply calculates.So GPU be more suitable for for do data scale is big, data type is unified,
The low parallel computation of degree of dependence between data.
If needing first to know the global information of network using traditional community discovery method, then to carry out logic more complex
Calculating, such that the calculating data scale of entire community discovery process is very big, it is long to lead to calculate the time, the efficiency of method
It is not high.But because degree of dependence is high between data in conventional method, be not suitable for being used to be counted on GPU parallel processing platform
It calculates.
Traditional community's module, has intermediate degree, modularization, conductibility several, but these community's modules exist
Network global information is needed when calculating, so that the communication cost in calculating is higher.Moreover, some standards are only laid particular emphasis on to network
The measurement of midpoint or line one aspect, it is thus possible to cause to measure not accurate enough.
In traditional community's module, intermediate degree is first to calculate in network for indirect measurement community by certain side
Shortest path quantity, then count the intermediate degree on each side in sub-network structure, if the side centre degree of a sub-network structure compared with
Height then illustrating that the sub-network structure and extraneous degree of contact are also relatively high, therefore is not community.It can be seen that the measurement
Standard needs global information, primarily focuses on the calculating on side, and calculating logic is more complex.
Modularization, conductibility are then can direct community module.Modularization module calculating process such as 2 institute of formula
Show:
M is the quantity on all sides in network, mcIt is the quantity on side in community inside, dcIt is the degree of all the points in community.It can
To find out, which needs global information, and calculation amount is larger.
Conductibility module calculating process is as shown in Equation 3:
The module principle is the side in network to be divided into three parts, eAAIt is the side inside community, eABIt is community
Adjacent side, eBBBe then in whole network it is remaining not with the associated side in community.And define eA=eAA+eBB, eB=eBB+eAB.It can
To find out, which needs global information, and primarily focuses on the calculating on side.
Traditional community discovery method is constantly split there are two types of type one is since big network, is formed small
Sub-network, and judge whether it is community structure.Another kind is constantly polymerize since point or small sub-network, is formed
New sub-network, and judge whether it is community structure.
Current community discovery method is all based on traditional community's module, because traditional module calculating is patrolled
It is volume more complex, so that the computational efficiency of method is not high, while not being suitable for fast parallel computing platform yet.
Summary of the invention
The technical problem to be solved by the present invention is in view of the shortcomings of the prior art, provide a kind of based in community structure
It is outer than measure and community discovery method.In order to solve the above technical problems, the technical scheme adopted by the invention is that: Yi Zhongji
Compare measure inside and outside community structure, which comprises the following steps:
1) input sub-network structure G=(V, E), V and E are respectively the set of point and side in sub-network；Define the subnet
Interior point, inner edge, the exterior point, outside of network structure；Wherein, interior quantity >=4；The quantity of inner edge is greater than 0；Outside, exterior point quantity
It is all larger than 0；
2) the inside and outside ratio of definition: internal ratio InnerRatio=inner edge/interior point, it is outer than outside OuterRatio=/compare outside, it is inside and outside
Than IOR=internal ratio/outer ratio；
3) according to inside and outside than judging whether the sub-network structure is community: because in sub-network structure inner edge with it is interior
The ratio of point can reflect the tightness degree of the sub-network structure internal connection, and another height is regarded as in adjacent outside and exterior point
Network structure, if inside and outside be greater than 1 than IOR value, i.e. internal ratio is greater than outer ratio, and the connection that can be released inside the sub-network structure is close
Degree is greater than the connection tightness degree of another sub-network structure, show that the sub-network structure internal connection tightness degree is greater than phase
Adjacent outside, it can be determined that going out the sub-network structure is community.IOR ratio is bigger, illustrates that the tightness degree inside community is higher.
The present invention also provides a kind of community discovery methods based on above-mentioned measure, comprising the following steps:
1) it inputs whole network structure N=(V ', E '), V ' and E ' are respectively the set of point and side in network；
2) for all nodes in the network structure N=(V ', E '), neighbors comparative approach is used to generate at the beginning of one group
Beginning sub-network structure；
3) the first sub-network structure G=(V, E) generated for step 2) is calculated using claim 1 the method
The sub-network structure it is inside and outside than IOR value, if inside and outside ratio≤1, judges that the sub-network structure is not community, records
Into the set of inspected sub-network structure, next sub-network structure is detected；
If 4) inside and outside ratio > 1, judges to be community, first it is recorded in the community's set having found, then use bidirectional iteration
Method, a direction are that new sub-network structure is constituted by successively increasing exterior point, i.e., for the community, if there is n is a outer
Point then generates n new sub-network structures, enters back into step 5).Another direction is to put in successively reducing to generate new son
Network structure successively removes each interior point, if there are points in m in the community, generate m new sub-network structures, then into
Enter step 5)；
5) for above-mentioned one group of new sub-network structure, first remove and duplicate portion in inspected sub-network structure set
Point, then judge whether there is new community and generate；
6) step 4) and step 5) are repeated, until not new sub-network structure and community generate；
7) Bit-reversed is carried out than IOR value size by inside and outside to all communities of generation, most internal connection tightness degree
High community comes foremost.
Include: using the detailed process that neighbors comparative approach generates one group of initial subnet network structure
1) the neighbors quantity for calculating each node in network structure N=(V ', E '), in network structure N=(V ', E ')
All nodes, according to its neighbors quantity from more to less carry out inverted order arrangement；
2) successively to each node, its neighbors is all added, forms a sub-network structure, i.e., each is saved in network
Point, which can all correspond to, generates a sub-network structure, forms one group of sub-network structure；This group of sub-network knot that initialization is generated
Structure carries out duplicate removal, then to the interior point in each sub-network structure, carries out Bit-reversed by neighbors quantity.
Compared with prior art, the advantageous effect of present invention is that: relative to traditional community's structure module, this
Invention does not need global network information, only can calculate the tightness degree inside community, energy with adjacent partial network information
The degree of dependence calculated between data is reduced, graphics processing unit parallel computation hardware is applicable to；Neighbors of the invention compares
Bidirectional iteration method has wider array of covering surface, can quickly find the high community of inner tight degree.
Detailed description of the invention
Fig. 1 is to compare module inside and outside community；
Fig. 2 is neighbour bidirectional iteration method flow diagram；
Fig. 3 is bidirectional iteration schematic diagram；
Fig. 4 (a) is dolphin community network topological diagram；Fig. 4 (b) is community discovery schematic diagram；
Fig. 5 (a) is karate club community network data；Fig. 5 (b) dolphin community network data；
Fig. 6 is to find that the time of best community accounts for the ratio of total time earliest；
Fig. 7 is community discovery search coverage rate.
Specific embodiment
Community network is made of the relationship between node and node, this scene can be abstracted as to a non-directed graph, use G
=(V, E) expression, V and E are respectively the set of point and side in network.Community discovery is exactly to find out some internal connections in network
Compare close sub-network structure.
Traditional community's module and community discovery method need the global information of network, that is, judge a sub-network knot
When whether structure is community, the structural information of whole network is used, causes computationally intensive, calculating logic is complicated, is not suitable for fast
Fast parallel computation.
Therefore, we to solve the problems, such as be provide a kind of community's module for not needing global network information so that
Degree of dependence between data reduces, while proposing a kind of community discovery method suitable for parallel computation based on the standard, mentions
The efficiency of high community discovery.
We solve this problem in terms of two, are design community's module first, then design community
It was found that method.
1. community's module
The challenge of module of the present invention is the community's measure designed, and the degree of dependence calculated between data is wanted
It is low, it should no longer to need global network information, while the relationship for considering the two kinds of factors in side and point in network can be balanced again, and can body
The characteristics of revealing community's internal connection relative close, external relation relative loose.
For the inside and outside principle than community module proposed by the present invention as referring to Fig.1, interior point is all in community structure
The interior point of point, minimum community should be greater than 4.Inner edge is all sides in community network structure, and feature is two ends on the side
Point is all in community.Exterior point is all neighbors put in community.Outside is all adjacent sides of community, and feature is that the side only has one
A endpoint is in community.Internal ratio (InnerRatio)=inner edge/interior point, it is outer than (OuterRatio)=outside/outer ratio, inside and outside ratio
(IOR)=internal ratio/outer ratio.
If inside and outside ratio > 1, illustrate internal ratio > compare outside, can reflect out the internal connection of the sub-network structure than external connection
Fasten close, and inside and outside ratio is bigger, illustrates that internal connection is closer.
The characteristics of inside and outside module than community, is: partial network information only needed, the relationship for considering point and side can be balanced,
Calculation amount is small, and degree of dependence is low between data, is suitble to parallel computation.
Performance compares, it is inside and outside than community module compared with other modules such as table 1:
Table 1 referring to Fig.1 in community several module results
Because centre degree is indirect measurement, not accuracy result relatively in.As can be seen that the society in referring to Fig.1
There are 12 points, 16 sides in area, and (A, B, C, D) should be a community structure, but be calculated with modularization and conductibility module,
As a result all it is negative value, is not community, because both judgment criterias can be by global network information or weighting side or the shadow of point
It rings, and the problem of accuracy occurs.But it is inside and outside than module result be 1.5, still maintained to community judge
Correctness.
2. community discovery method
The present invention is to design the method for being suitable for fast parallel calculating in the challenge of community discovery method, and needs
Improve the search coverage rate of sub-network structure.
For the principle of this community discovery method invention as referring to Fig. 2, process is the continuous iteration since one group of sub-network structure
It generates new sub-network structure and finds community.The present invention is neighbour sequence initialization and two-way in the characteristics of community discovery method
Iterative search.
The community network data of the method for the present invention are all disclosed data sources, there is karate club data collection, are come
From " An information flow model for conflict and fission in small groups "；There is dolphin
Data set comes from " The bottlenose dolphin community of doubtful sound featuresa
large proportion of long-lasting associations"；There are also lexical relation data sets, come from
“Finding community structure in networks using the eigenvectors of matrices”。
The initialization section of the method for the present invention, it is main that the high sub-network structure of internal connection degree is considered to come scouting team
Before arranging, the community structure that can more early find in this way.It completes to generate the initial of one group of sub-network structure using neighbour sort method
Steps are as follows for change:
Step 1, the neighbors quantity for calculating each node in network, all nodes in network, according to its neighbors
Quantity from more to less carry out inverted order arrangement.
Step 2, successively to each node, one sub-network structure of composition all is added in its neighbors.
Step 3, duplicate removal is carried out to this group of sub-network structure, to the node in each sub-network, is carried out by neighbors quantity
Bit-reversed, initialization are completed.
As referring to Fig. 6, using neighbour sort method relative to traditional random device, find earliest best community when
Between account in the ratio of total run time, can averagely improve 39.64%.
The community discovery part of method, it is main to consider to use bidirectional iteration method, it is single compared to traditional segmentation or polymerization
To search, the covering surface of search can be improved, Fig. 3 is such as referred to.Because the sub-network structure combination variety of a network is too many, such as
100 nodes appoint and 50 nodes are taken to constitute networks, then its quantity can reach 1.00891E+29, and in computer 64 without symbol
The expression range of integer just arrives 1.84467E+19, has exceeded 10 orders of magnitude.It will be generated using complete searching method huge
Operand, and most of is invalid search, so the discovery method of community network is typically all to cannot be used up all direction search method, such as
Fruit can improve search covering surface as far as possible, just be more likely to the community found out.
The module of inside and outside ratio of this method based on proposition simultaneously, can be effectively performed parallel computation, utilize hardware
The high efficiency of platform quickly finds out community structure.Steps are as follows:
Step 4, to this group of sub-network structure after initialization, parallel thread can be used and execute calculating.To each subnet
Network structure calculates inside and outside ratio, judges whether the structure is community, the set of inspected structure is then recorded if not community
In, it goes to detect next sub-network structure；
Step 5, it if it is community, is first recorded in the community's set having found, then use bidirectional iteration method, a side
To being to constitute new sub-network structure by successively increasing exterior point, another direction is to put in successively reducing to constitute new son
Network structure generates one group of new sub-network structure.
Step 6, for newly generated one group of sub-network structure, first remove and have duplicate portion in inspected structured set
Point, then judge whether there is new community and generate, the process of iteration judgement, until not new sub-network structure and community produce
It is raw.
Step 7, Bit-reversed finally is carried out by inside and outside ratio size to all communities of generation, defines and compares threshold inside and outside one
Value, finds out the high community of internal connection tightness degree.
The performance of the method for the present invention compares:
Fig. 4 (a) is dolphin community network topological diagram；Fig. 4 (b) is found out after community discovery method of the invention
It is inside and outside than highest community.
Fig. 5 (a) and Fig. 5 (b) are the different community discovery methods of comparison, and whether see can find that community is inside and outside than highest society
Area, can also verify whether higher search coverage rate from side, and Fig. 5 (a) is karate club community network data, Fig. 5
(b) dolphin community network data, it can be seen that neighbour bidirectional iteration method compares conventional method, it can be found that having inside and outside highest
The community of ratio.
Fig. 7 is more several searching methods, and in different society network data, community discovery searches for coverage rate, coverage rate
It is higher, indicate that the sub-network structure of search is more, it is as a result also more accurate, the search covering surface of bidirectional iteration method is calculated
Highest can averagely improve 12.67%.
It is proposed by the present invention inside and outside than module, it realizes in the case where only needing part neighbouring network information, more
Accurately, the contiguity characteristic of community is more evenly measured, and degree of dependence is low between calculating data, is suitble to parallel computation；Together
When the neighbour bidirectional iteration community discovery method that proposes, can faster, more fully find better community, earliest discovery most
The time of good community accounts in the ratio of total run time, can averagely improve 39.64%；On sub-network structure search covering surface,
12.67% can averagely be improved.
Claims (2)
1. a kind of based on comparing measure inside and outside community structure, which comprises the following steps:
1) input sub-network structure G=(V, E), V and E are respectively the set of point and side in sub-network；Define the sub-network knot
Interior point, inner edge, the exterior point, outside of structure；Wherein, interior quantity >=4；Inner edge, outside, exterior point quantity be all larger than 0；Interior point is
All the points in sub-network structure；Inner edge is all sides in sub-network structure, and two endpoints of inner edge are all in sub-network structure
It is interior；Exterior point is all neighbors of interior point in sub-network structure；Outside is all adjacent sides of sub-network structure, and outside only has one
Endpoint is in sub-network structure；
2) definition is inside and outside than the quantity of internal ratio InnerRatio=inner edge/interior point quantity, it is outer than OuterRatio=outside
Quantity/exterior point quantity, it is inside and outside than IOR=internal ratio/outer ratio；
3) according to inside and outside than judging whether the sub-network structure is community: if inside and outside be greater than 1 than IOR value, i.e. internal ratio is greater than
Outer ratio then judges that the sub-network structure is community.
2. a kind of community discovery method based on measure described in claim 1, which comprises the following steps:
1) it inputs whole network structure N=(V ', E '), V ' and E ' are respectively the set of point and side in network；
2) for all nodes in the network structure N=(V ', E '), using neighbors comparative approach generate one group it is initially sub
Network structure；Include: using the detailed process that neighbors comparative approach generates one group of initial subnet network structure
A) the neighbors quantity for calculating each node in network structure N=(V ', E '), the institute in network structure N=(V ', E ')
Have node, according to its neighbors quantity from more to less carry out inverted order arrangement；
B) successively to each node, its neighbors is all added, forms a sub-network structure, i.e., each node in network
It is corresponding to generate a sub-network structure, form one group of sub-network structure；
C) sub-network structure that step b) is obtained carries out duplicate removal, then to the interior point in each sub-network structure, by neighbors quantity
Carry out Bit-reversed；
3) the first sub-network structure G=(V, E) generated for step 2), calculates the son using claim 1 the method
Network structure it is inside and outside than IOR value, if it is described it is inside and outside judge that the sub-network structure is not community than IOR value≤1, record
The sub-network structure detects next sub-network structure into the set of inspected sub-network structure；
If 4) in step 3) sub-network structure G=(V, E) it is inside and outside than IOR value > 1, judge that the sub-network structure is community,
First by the sub-network structure be recorded it has been found that community set in, then use bidirectional iteration method, a direction be by according to
Secondary increase exterior point constitutes new sub-network structure, i.e., for the community, if there is n exterior point, then generates n new sub-networks
Structure enters back into step 5)；Another direction is that in successively reducing point successively removes each to generate new sub-network structure
It is interior, if there is point in m in the community, m new sub-network structures are generated, step 5) is entered back into；
5) for above-mentioned new sub-network structure, first remove with duplicate part in inspected sub-network structure set, then judge
Whether there is new community to generate；
6) step 4) and step 5) are repeated, until not new sub-network structure and community generate；
7) Bit-reversed is carried out than IOR value size by inside and outside to all communities of generation, internal connection tightness degree is highest
Community comes foremost；IOR value is bigger, illustrates that the tightness degree inside community is higher.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510526277.2A CN105337759B (en) | 2015-08-25 | 2015-08-25 | It is a kind of based on inside and outside community structure than measure and community discovery method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510526277.2A CN105337759B (en) | 2015-08-25 | 2015-08-25 | It is a kind of based on inside and outside community structure than measure and community discovery method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105337759A CN105337759A (en) | 2016-02-17 |
| CN105337759B true CN105337759B (en) | 2018-12-25 |
Family
ID=55288071
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510526277.2A Active CN105337759B (en) | 2015-08-25 | 2015-08-25 | It is a kind of based on inside and outside community structure than measure and community discovery method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105337759B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106453096B (en) * | 2016-09-05 | 2019-06-14 | 北京邮电大学 | A kind of dynamic network community discovery method and device |
| CN109842642B (en) * | 2017-11-24 | 2022-11-01 | 中兴通讯股份有限公司 | Deployment method and device of containerized cluster system and containerized cluster system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009217535A (en) * | 2008-03-11 | 2009-09-24 | Dainippon Printing Co Ltd | Community detection system in virtual world |
| CN101877711A (en) * | 2009-04-28 | 2010-11-03 | 华为技术有限公司 | Social network establishment method and device, and community discovery method and device |
| CN101916256A (en) * | 2010-07-13 | 2010-12-15 | 北京大学 | Community discovery method for synthesizing actor interests and network topology |
| US20130132442A1 (en) * | 2011-11-21 | 2013-05-23 | Motorola Mobility, Inc. | Ontology construction |
-
2015
- 2015-08-25 CN CN201510526277.2A patent/CN105337759B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009217535A (en) * | 2008-03-11 | 2009-09-24 | Dainippon Printing Co Ltd | Community detection system in virtual world |
| CN101877711A (en) * | 2009-04-28 | 2010-11-03 | 华为技术有限公司 | Social network establishment method and device, and community discovery method and device |
| CN101916256A (en) * | 2010-07-13 | 2010-12-15 | 北京大学 | Community discovery method for synthesizing actor interests and network topology |
| US20130132442A1 (en) * | 2011-11-21 | 2013-05-23 | Motorola Mobility, Inc. | Ontology construction |
Non-Patent Citations (1)
| Title |
|---|
| 动态网络社区发现算法研究;王玲;《中国优秀硕士学位论文全文数据库》;20140731;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105337759A (en) | 2016-02-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107291847A (en) | A kind of large-scale data Distributed Cluster processing method based on MapReduce | |
| Deng et al. | An efficient online direction-preserving compression approach for trajectory streaming data | |
| CN110222029A (en) | A kind of big data multidimensional analysis computational efficiency method for improving and system | |
| CN107623639A (en) | Data flow distribution similarity join method based on EMD distances | |
| CN110135565A (en) | Realize the assessment system of performance on the integrated for neural network algorithm | |
| CN105337759B (en) | It is a kind of based on inside and outside community structure than measure and community discovery method | |
| Hu et al. | Parallel clustering of big data of spatio-temporal trajectory | |
| CN116662412B (en) | Data mining method for big data of power grid distribution and utilization | |
| WO2016112618A1 (en) | Distance-based algorithm for solving representative node set in two dimensional space | |
| Pan et al. | Design and application of real-time network abnormal traffic detection system based on Spark Streaming | |
| Yao et al. | Community detection based on variable vertex influence | |
| CN111711530A (en) | Link prediction algorithm based on community topological structure information | |
| CN107332687B (en) | Link prediction method based on Bayesian estimation and common neighbor | |
| CN107018027B (en) | Link prediction method based on Bayesian estimation and common neighbor node degree | |
| CN105354243B (en) | The frequent probability subgraph search method of parallelization based on merger cluster | |
| Chen et al. | A multi-label propagation algorithm for community detection based on average mutual information | |
| Xia et al. | An improved PSO with detecting and local-learning strategy | |
| Wang et al. | Clustering algorithm based on grid and density for data stream | |
| Shi et al. | Community detection using robust label propagation algorithm | |
| Du et al. | FASTBEE: A fast and self-adaptive clustering algorithm towards to edge computing | |
| Fang et al. | HuGE: An entropy-driven approach to efficient and scalable graph embeddings | |
| Ruan et al. | The community detection of complex networks based on Markov matrix spectrum optimization | |
| Liu et al. | Social Network Community-Discovery Algorithm Based on a Balance Factor | |
| Pirouz et al. | Optimized label propagation community detection on big data networks | |
| Cai et al. | Efficient algorithms for distance-based representative skyline computation in 2D space |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |