CN105868791A - Multi-resolution community discovering method based on fuzzy clustering - Google Patents
Multi-resolution community discovering method based on fuzzy clustering Download PDFInfo
- Publication number
- CN105868791A CN105868791A CN201610237640.3A CN201610237640A CN105868791A CN 105868791 A CN105868791 A CN 105868791A CN 201610237640 A CN201610237640 A CN 201610237640A CN 105868791 A CN105868791 A CN 105868791A
- Authority
- CN
- China
- Prior art keywords
- fuzzy
- node
- community
- network
- matrix
- 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 67
- 239000011159 matrix material Substances 0.000 claims abstract description 60
- 230000008569 process Effects 0.000 claims abstract description 8
- 230000009466 transformation Effects 0.000 claims abstract description 4
- 230000005540 biological transmission Effects 0.000 claims description 14
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 230000000694 effects Effects 0.000 claims description 8
- 230000033228 biological regulation Effects 0.000 claims description 7
- 230000015572 biosynthetic process Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 238000012546 transfer Methods 0.000 claims description 4
- 238000003786 synthesis reaction Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 abstract description 9
- 238000001514 detection method Methods 0.000 abstract description 6
- 238000005457 optimization Methods 0.000 abstract description 2
- 230000003993 interaction Effects 0.000 abstract 1
- 238000012916 structural analysis Methods 0.000 abstract 1
- 238000012800 visualization Methods 0.000 abstract 1
- 238000011160 research Methods 0.000 description 6
- 238000005192 partition Methods 0.000 description 5
- 239000000654 additive Substances 0.000 description 3
- 230000000996 additive effect Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000005295 random walk Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 101710134784 Agnoprotein Proteins 0.000 description 1
- 235000005156 Brassica carinata Nutrition 0.000 description 1
- 244000257790 Brassica carinata Species 0.000 description 1
- 101710124584 Probable DNA-binding protein Proteins 0.000 description 1
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000003997 social interaction Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 210000004885 white matter Anatomy 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering 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)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Computing Systems (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention provides a multi-resolution community discovering method based on fuzzy clustering. According to local interaction information of adjacent nodes, the structural similarity is introduced for measuring the fuzzy relation between the nodes, fuzzy transitivity of the fuzzy similarity between the nodes in a network topology is partially considered, fuzzy parameters are used for performing set cutting on a fuzzy transitivity matrix to obtain community structures under different resolutions, and therefore network communities can be discovered. Matrix transformation operation is adopted, a network community detection model based on fuzzy clustering is built, iterative optimization processes in a traditional method are reduced, the time complexity is lowered, a large number of experiments prove that the community structures in a network can be effectively revealed, the universality is strong, and the high application value is achieved; network structural analysis and community structural visualization can be effectively achieved.
Description
Technical field
The present invention relates to Complex Networks Analysis technical field of research, in particular it relates to a kind of based on fuzzy clustering many points
Resolution community discovery method.
Background technology
Complex network is as a cross discipline widely, and it relates to computer, physics, mathematics, information science, system
The ambit such as science, Network Science, is increasingly becoming the powerful tool solving challenge, and in various fields
Have a wide range of applications, such as social network analysis, biological engineering, economy and finance, electric power and traffic, behavior of men analysis,
Big data analysis etc..Complex network is researched and analysed, is greatly expanded it is appreciated that the breadth and depth in the world, has
Major and immediate significance.In substantial amounts of complex network is studied, the research to complex network community structure, is a weight
Big research focus.Dividing generally speaking, for given network structure, community's internal node has relative between community
And connect more closely.Community structure characteristic in complex network is proved by substantial amounts of research.Such as, protein is handed over
The research of community structure in network mutually, can analyze different known or agnoprotein matter functional modules, further appreciate that egg
The complex characteristics of white matter structure.In social networks, due to the enhancing of Social Interaction, based on different interest, theme,
A large amount of colonies that the features such as occupation, region are formed, community structure feature is particularly evident.Therefore, excavation network closely joins
The community structure of system has important to understanding and analyzing network structure attribute, regularity of information dissemination, human society organizational structure etc.
Theory significance and be widely applied value.
Community structure discovery is intended to the community structure in detection network with certain natural quality, i.e. according to certain rule,
It is some modules by node division interconnective in network so that the contact of each inside modules is the densest, intermodule
Connect the most sparse.Although the concept of community structure readily appreciates, but due to the multiformity of network structure and complexity, society
District finds that method is the most complicated various.In the community structure research of complex network, substantial amounts of community discovery algorithm is suggested,
As based on figure segmentation, hierarchical clustering, factions' filtration, centrad measurement, spectral clustering, based on modularity optimization, company frontier inspection
Survey.While it is true, major part method needs priori to instruct could realize effectively division, the most do not propose one
Unified measurement criterion, therefore has its limitation.According to the definition of community structure, a network can be regarded as multiple society
District combines.Community structure is divided, has two key issues to need to solve: one determines that the quantity of community, right
In unknown network structure, its community divides and architectural feature is unknown;Another is exactly that each community member determines,
Avoid the unreasonable division of community structure.The segmentation of traditional community division method such as figure is by rigid for each node in network
Be divided into a specific community, and have ignored its internal relation.In reality network, due to the complexity of network mechanism
Property and multiformity, a node may belong simultaneously to multiple community, i.e. in partition process, there is the uncertainty of node
Or ambiguity, the ambiguity of community structure divides closer to real network structure.
Secondly, for community structure itself, it is based on certain similarity or common trait such as social activity between its internal node
The relationships such as different interest in network, hobby, theme together, this be people study starting point that community divides it
One.But, the most existing community structure detection method using this similarity relationships as a kind of definitiveness or rigid degree
Amount, division so can cause the unreasonable division of community. it practice, entity in live network structure is such as social networks
Between similarity relation be fuzzy or probabilistic, and divide ignoring in network other heavy with a kind of certainty measure
Want information.Present invention thought based on fuzzy theory proposes a kind of fuzzy division method of community structure.Solve fuzzy
The basis of partition problem is exactly Fuzzy Set Theory.In the network architecture, between node, fuzzy relation emphasizes that network node is not with
Same degree is under the jurisdiction of multiple community class, and non-critical is divided into a certain particular community.Meanwhile, the similarity of ambiguity
The one that relation is considered as deterministic dependence is extensive.
In detecting for community structure, community effectively divides and divides two basic problems of number, and the present invention proposes a kind of base
Community structure in Fuzzy clustering techniques finds method, discloses the Web Community under the conditions of different resolution simultaneously
Hierarchical structure.The thought of proposition fuzzy clustering of the present invention finds the community structure in complex network, it is achieved fuzzy stroke
Point, the method is based on a kind of fuzzy relational model rather than conventional graph model solves community discovery problem.
Summary of the invention
For defect of the prior art, it is an object of the invention to provide a kind of multiresolution community based on fuzzy clustering
Discovery method.
The multiresolution community discovery method based on fuzzy clustering provided according to the present invention, comprises the steps:
Fuzzy switch process: set up adjacency matrix A according to network topology structure, and calculate adjacent node based on adjacency matrix A
Between fuzzy relation, the fuzzy relation matrix obtained is carried out fuzzy transmitting and converting, it is thus achieved that fuzzy equivalent matrix;
Fuzzy intercepting step: be mapped in network structure by fuzzy equivalent matrix, obtain fuzzy equivalence relation class, utilizes fuzzy threshold
Value intercepts this fuzzy equivalence relation class, obtains the fuzzy community structure cluster in corresponding Fuzzy Threshold level, and comes by modularity
Evaluate the Clustering Effect of fuzzy community.
Preferably, described fuzzy switch process includes:
Step M1: network to be analyzed is set up the adjacency matrix A of network, is numbered node according to node sequence, compiles
Number from the beginning of 1, building the N rank square formation that element is 0 or 1, wherein N is the total number of network node;
Step M2: obtain the similarity between adjacent node, is converted into fuzzy resembling relation by deterministic syntopy,
Realization will abut against matrix A and is converted to fuzzy relationship matrix r;
Step M3: for the similarity between tolerance nonneighbor node further, fuzzy relationship matrix r is carried out fuzzy transmission and becomes
Change, make internodal similarity reach consistent stability;
Step M4: the uniformly convergent fuzzy relation obtained by fuzzy transmission function, sets up fuzzy transfer matrix, and
Obtain fuzzy equivalent matrix.
Preferably, in described step M2, according to the structural similarity metric algorithm of definition obtain between adjacent node similar
Degree, the computing formula of described structural similarity metric algorithm is as follows:
In formula, u, v are respectively the arbitrary node in set of network nodes, and Γ () represents the adjacent node set of certain node, Γ (u)
Representing the adjacent node set of node u, Γ (v) represents the adjacent node set of node v, and w () represents and connects between certain two node
The weight on limit, (u, x) connects the weight on limit to w between expression node u and node x, (v x) represents and connects between node v and node x w
The weighting structure similarity on limit, s (u, v) ∈ [0,1];For connecting the weight on limit between undirected arbitrary node without weighting network
W ()=1, then the formula of structural similarity metric algorithm is further simplified as following form:
Preferably, the fuzzy transmission transforming function transformation function in described step M3 is as follows:
In formula, the dimension of n representing matrix, n=| V |, V represent network node sequence vector, and R represents fuzzy relation matrix,
T (R) represents fuzzy equivalence relation, and ∪ represents that fuzzy relation synthesizes computing, and k represents the number of times of fuzzy relation synthesis computing, its
MeetFuzzy equivalence relation meets
Preferably, described fuzzy equivalent matrix, have the property that
Character 1: symmetry;For Undirected networks structure, internodal fuzzy relation meets symmetry;I.e. node i is to joint
Point j fuzzy relation r (i, j) be equivalent to node j to node i fuzzy relation r (j, i);
Character 2: reflexivity;In fuzzy relation matrix, any node is 1 to the fuzzy relation perseverance of self, i.e. and r (i, i)=1;
Character 3: transitivity;For the fuzzy equivalent matrix through fuzzy transmission conversion, meet
Preferably, described fuzzy intercepting step includes:
Step N1: fuzzy equivalence relation matrix and the network node one_to_one corresponding that will obtain, obtains fuzzy equivalence relation class, institute
State fuzzy equivalence relation class and be equivalent to the hierarchical clustering tree construction of network;
Step N2: selected arbitrary Fuzzy Threshold ε ∈ [0,1], intercepts fuzzy equivalence relation class, obtains the mould of correspondence
Stick with paste community division result;
Step N3: calculate the module angle value of corresponding fuzzy community division result;
Step N4: regulation Fuzzy Threshold ε, obtains the community structure under the conditions of different resolution.
Preferably, in described step N3, the computing formula of modularity value Q is as follows:
In formula, i represents the community's quantity in network division, eiiRepresent that the quantity connecting limit between i-th community internal node accounts for whole
The ratio of individual network edge number, aiRepresent the ratio of one section of company's limit quantity being connected with i-th community interior joint.
Compared with prior art, the present invention has a following beneficial effect:
1, the present invention is on the basis of legacy network community discovery, uses fuzzy clustering method to realize community structure
Effectively divide;Unlike existing algorithm, similarity relation between its Fuzzy processing node and community rather than one determine
The hard plot of property, thus avoid the unreasonable division of network structure.
2, the present invention achieves the community structure division under different resolution in the control to Fuzzy strategy, enters one
Step has excavated the network structure characteristic of complex network.
3, the present invention is based on fuzzy behaviour, in conjunction with the partial structurtes information of node, it is proposed that effective structural similarity degree
Metering method, improves reasonability and reliability that community structure divides.
Accompanying drawing explanation
By the detailed description non-limiting example made with reference to the following drawings of reading, the further feature of the present invention,
Purpose and advantage will become more apparent upon:
Fig. 1 is method provided by the present invention and additive method performance comparison figure on GN network.
Fig. 2 is method provided by the present invention and additive method performance comparison figure on LFR network.
Fig. 3 is the present invention scattergram to the Fuzzy Threshold value on above-mentioned GN network.
Fig. 4 is the present invention scattergram to the Fuzzy Threshold value on above-mentioned LFR network.
Fig. 5 (a) is a kind of division result schematic diagram that the present invention analyzes a live network, and Fig. 5 (b) is another kind of
Division result schematic diagram.
Fig. 6 is that the present invention carries out Web Community's scattergram that fuzzy clustering obtains under different resolution to above-mentioned live network.
Detailed description of the invention
Below in conjunction with specific embodiment, the present invention is described in detail.Following example will assist in those skilled in the art
Member is further appreciated by the present invention, but limits the present invention the most in any form.It should be pointed out that, the common skill to this area
For art personnel, without departing from the inventive concept of the premise, it is also possible to make some changes and improvements.These broadly fall into
Protection scope of the present invention.
The multiresolution community discovery method based on fuzzy clustering provided according to the present invention, comprises the steps:
Fuzzy switch process: set up adjacency matrix A according to network topology structure, and calculate adjacent node based on adjacency matrix A
Between fuzzy relation, the fuzzy relation matrix obtained is carried out fuzzy transmitting and converting, it is thus achieved that fuzzy equivalent matrix;
Fuzzy intercepting step: be mapped in network structure by fuzzy equivalent matrix, obtain fuzzy equivalence relation class, utilizes fuzzy threshold
Value intercepts this fuzzy equivalence relation class, obtains the fuzzy community structure cluster in this Fuzzy Threshold level, and comments by modularity
Valency obscures the Clustering Effect of community.
Described fuzzy switch process includes:
Step M1: network to be analyzed is set up the adjacency matrix A of network, is numbered node according to node sequence, compiles
Number from the beginning of 1, building the N rank square formation that element is 0 or 1, wherein N is the total number of node;
Step M2: obtain the similarity between adjacent node, by deterministic according to the structural similarity metric algorithm of definition
Syntopy is converted into fuzzy resembling relation, it is achieved will abut against matrix A and is converted to fuzzy relationship matrix r;
Step M3: for the similarity between tolerance nonneighbor node further, fuzzy relationship matrix r is carried out fuzzy transmission and becomes
Change, make internodal similarity reach consistent stability;
Step M4: the uniformly convergent fuzzy relation obtained by fuzzy transmission function, sets up fuzzy transfer matrix, and
Obtain fuzzy equivalent matrix.
Structural similarity metric algorithm in described step M2 is as follows:
In formula, u, v are respectively the arbitrary node in set of network nodes, and Γ () represents the adjacent node set of certain node, Γ (u)
Representing the adjacent node set of node u, Γ (v) represents the adjacent node set of node v, and w () represents and connects limit between certain node
Weight, w (u, x) represents and connects the weight on limit between node u, w (v, x) represents and connects the weighting structure similarity on limit between node v,
s(u,v)∈[0,1];For connecting weight w ()=1 on limit, then structural similarity degree between undirected arbitrary node without weighting network
The formula of quantity algorithm is further simplified as following form:
Fuzzy transmission transforming function transformation function in described step M3 is as follows:
In formula, the dimension of n representing matrix, its satisfied=| V | (V represents network node sequence), R represents fuzzy relation square
Battle array, t (R) represents fuzzy equivalence relation, and ∪ represents that fuzzy relation synthesizes computing, and k represents the number of times of fuzzy relation synthesis computing,
It meetsFuzzy equivalence relation meets
Described fuzzy equivalent matrix, has the property that
Character 1: symmetry;For Undirected networks structure, internodal fuzzy relation meets symmetry;I.e. node i is to joint
Point j fuzzy relation r (i, j) be equivalent to node j to node i fuzzy relation r (j, i);
Character 2: reflexivity;In fuzzy relation matrix, any node is 1 to the fuzzy relation perseverance of self, i.e. and r (i, i)=1.
Character 3: transitivity;For the fuzzy equivalent matrix through fuzzy transmission conversion, meet
Described fuzzy intercepting step includes:
Step N1: fuzzy equivalence relation matrix and the network node one_to_one corresponding that will obtain, obtains fuzzy equivalence relation class, institute
State fuzzy equivalence relation class and be equivalent to the hierarchical clustering tree construction of network;
Step N2: selected arbitrary Fuzzy Threshold ε ∈ [0,1], intercepts fuzzy equivalence relation class, obtains the mould of correspondence
Stick with paste community division result;
Step N3: calculate the module angle value of corresponding fuzzy community division result;
Step N4: regulation Fuzzy Threshold ε, obtains the community structure under the conditions of different resolution.
In described step N3, the computing formula of modularity value Q is as follows:
In formula, i represents the community's quantity in network division, eiiRepresent that the quantity connecting limit between community's i internal node accounts for whole net
The ratio of network limit number, aiRepresent the ratio of one section of company's limit quantity being connected with community's i interior joint.Usually, module angle value is more
Greatly, represent that the effect that Web Community divides is the best.
Specifically, comprise the steps:
Step S1: set up the adjacency matrix A of network according to network structure information, according to node sequence, node is numbered,
Numbering, from the beginning of 1, builds N rank square formation, and wherein N is the total number of node, directly connects limit with 1 if had between two nodes
Represent, be otherwise 0;
Step S2: conversion fuzzy relation.The similarity between structural similarity metric calculation adjacent node according to definition,
Deterministic linking relationship is converted into fuzzy resembling relation, it is achieved will abut against matrix A and be converted to fuzzy relationship matrix r;
Step S3: fuzzy transmitting and converting.For the similarity between tolerance nonneighbor node further, fuzzy relationship matrix r is entered
Row transmission conversion, makes internodal similarity reach consistent stability, to R matrix by fuzzy transfer function by fuzzy
Transmission conversion;
Step S4: fuzzy equivalent matrix;According to the uniformly convergent fuzzy relation obtained by fuzzy transferometer
Set up fuzzy equivalent matrix;
Step S5: fuzzy equivalence relation class, the fuzzy equivalence relation matrix that will obtain, with network node one_to_one corresponding, obtain
Fuzzy equivalence relation class, this equivalence class is equivalent to the hierarchical clustering tree construction of network;
Step S6: selected arbitrary Fuzzy Threshold ε ∈ [0,1], intercepts fuzzy equivalence relation class, obtains the fuzzy of correspondence
Colony divides;
Step S7: define according to above-mentioned modularity, calculates the module angle value that corresponding fuzzy community divides, module angle value
The biggest, illustrate that Web Community divides effect the best;
Step S8: regulation Fuzzy Threshold ε, obtains the community structure under the conditions of different resolution.
By describe in detail the effectiveness of offer method and extensibility, the present invention has carried out the following examples by experiment.
1) experiment condition: CPU Intel Pentium Dual-Core 2.0-GHz, RAM 4.00GB, Windows 7
Operating system, simulation software RStudio.
2) experimental subject:
Synthetic network and real world network are chosen in experiment respectively.
Synthetic network
Further, Girvan and Newman in 2002 at paper " Community structure in social and
biological networks”Girvan M,Newman M.E.Proceedings of the National Academy of
Sciences of the United States of America.2002,99 (12): 7821-6. (manually generated network uses GN respectively
Baseline network and LFR baseline network) middle proposition.This network is formed community's knot of four equal scales by 128 nodes
Structure, node has identical degree and is distributed, and the out-degree of node and in-degree scalable, it represents node with hybrid parameter μ
Going out the in-degree ratio fog-level with adjustment network, μ value is the biggest, and the boundary between Web Community is the fuzzyyest.
Lancichinetti et al. proposed (Lancichinetti A, Fortunato S, Radicchi F.Benchmark in 2008
graphs for testing community detection algorithms.Physical Review E.2008,78(4):046110.)
LFR baseline network;It practice, LFR baseline network is the extension of GN baseline network, that reflects node degree distribution and
The homogeneity of community structure scale distribution, i.e. node degree distribution and community's scale meet power-law distribution so that web results is more
Close to live network.In order to evaluate the performance of inventive method, standard mutual information (NMI) is used to measure division result.
If NMI value is closer to 1, illustrate that the community structure found is closer to real community structure.
Presently disclosed method represents with Strsim in an experiment.Meanwhile, for embodying its impact of performance, this method and its
He compares by the community discovery method of several classics.These methods include: Pons et al. in 2005 at " 20th
International Symposium on Computer and Information Sciences " on " Computing that delivers
Communities in large networks using random walks " the middle Walktrap method proposed, Vincent et al.
" the Fast unfolding of communities in delivered on " Journal of Statistical Mechanics " in 2008
Large networks " in propose BGLL method, and Rosvall and Bergstrom in 2008 at " Proceedings
Of the National Academy of Sciences of the United States of America " on " the Maps of that delivers
Random walks on complex networks reveal community structure " the middle Infomap method proposed.
The simulation experiment result based on GN baseline network is as shown in Figure 1.When mixed coefficint is less than 0.4, all methods
Can effectively detect real community structure, i.e. NMI=1.Along with mixed coefficint is gradually increased, the community of each method
Structure recognition ability presents and declines in various degree.As it can be seen, when mixed coefficint is more than 0.When 5, Infomap method
Community detection ability drastically drops to 0, and the performance of Walktrap and BGLL method also gradually reduces, and the present invention is carried
Performance for method but presents stable.Reason is that this method takes the phase that a kind of local searching strategy comes between node metric
Like degree, in the case of Web Community's obscurity boundary, still can identify that the node of arest neighbors keeps certain identification ability.
It is clear that method provided by the present invention has clear superiority for the network structure that community boundary is fuzzy.
The simulation experiment result based on LFR baseline network is as shown in Figure 2.In LFR network, node degree distribution, society
District's quantity and community's scale all change.When mixed coefficint is not more than 0.5, all methods are to the network society in LFR network
Plot structure has good identification ability.Along with mixed coefficint constantly increases, the performance of each method also presents in various degree
Decline.When mixed coefficint is more than 0.6, Infomap method None-identified goes out community structure therein (NMI=0),
And the performance of additive method is gradually lowered.As it can be seen, when mixed coefficint is 0.6, BGLL method can obtain relatively
Good effect, but, when Web Community border more obscures, its performance is poor compared with Walktrap and Strsim method.And
Presently disclosed method still can keep certain identification ability when mixed coefficint constantly reduces.The further body of this experiment
Effectiveness and the stability of this method are showed.
For the fuzzy clustering of network structure, Fuzzy Threshold is chosen and is evaluated by this method, and obtains Fuzzy Threshold
Optimum experience interval, as shown in Figure 3 and Figure 4.Fig. 3 is that this method obscures in difference on GN baseline network
NMI value distribution under coefficient condition.This distribution presents stepped distribution, and under different mixed coefficints, presents
Similar trend, is finally reached steady statue, and this is consistent with above-mentioned experimental result.As it can be seen, when Fuzzy Threshold value
When scope is [0.2,0.4], this method can obtain optimal result.Meanwhile, by analysis chart 4 it can be seen that work as mixed stocker
When number is less than 0.5, Fuzzy Threshold value distribution of results difference is bigger.When mixed coefficint is more than 0.5, present similar
Trend, and reach stable result.Relatively it is found that when Fuzzy Threshold span is [0.2,0.4], this method
Optimal result can be obtained.
Real world network
Real world network in this experiment selects Zachary karate club social networks.This network is widely used in survey
The validity and reliability of examination community detection algorithm.This network is to the society's friendship between this clubbite based on Zachery
Carry out the investigation up to two years and observation mutually, and construct a friendship with 78 limits of 34 clubbites
Network mutually.During observing, owing to the suggestion between Club Management person (node 1) and coach's (node 33) is divided
Discrimination, final club splits into two less community structures.This method is applied to the result of this network as shown in Figure 5.
Figure only gives two of which division result.When Fuzzy Threshold is adjusted, can obtain under different resolution
The module angle value of community structure, its community's number and partition structure is as shown in Figure 6.When this network obtains 4 community structure,
Modularity obtains maximum.
In an experiment, institute of the present invention extracting method can effectively detect the separating phenomenon in network structure.Test result indicate that,
Two groups with obvious community structure are detected, and as shown in Fig. 5 (a), different Web Communities is with different joints
Point shape representation.By regulation fuzzy parameter, the community structure of higher resolution can be obtained.As shown in Fig. 5 (b),
This network is divided into 3 communities, respectively with different shape representations in figure.In this division, due to tight between node
Similarity, node 25,26,29 and 32 is extracted as a single community.Meanwhile, by regulation clustering parameter ε,
The community of the smaller particle size being closely connected within community is found.When choosing bigger blur parameter value, this network quilt
Being divided into 4 community structures, it is consistent with the result of the CNM algorithm partition that Clauset et al. proposes.Closed by regulation
Suitable parameter value, the community structure of varying number is extracted, and corresponding different module angle value, as shown in Figure 6.Work as quilt
A modularity of maximum is had when being divided into 4 community structures, and the division of 2 communities corresponding to real network non-optimal,
Therefore, when modularity is on close level, suitably should select according to more effective criterion.And in this experiment, according to being carried
Going out method, its community's partition structure is accurate and effective.
Above the specific embodiment of the present invention is described.It is to be appreciated that the invention is not limited in
Stating particular implementation, those skilled in the art can make a variety of changes within the scope of the claims or revise,
This has no effect on the flesh and blood of the present invention.In the case of not conflicting, in embodiments herein and embodiment
Feature can arbitrarily be mutually combined.
Claims (7)
1. a multiresolution community discovery method based on fuzzy clustering, it is characterised in that comprise the steps:
Fuzzy switch process: set up adjacency matrix A according to network topology structure, and calculate adjacent node based on adjacency matrix A
Between fuzzy relation, the fuzzy relation matrix obtained is carried out fuzzy transmitting and converting, it is thus achieved that fuzzy equivalent matrix;
Fuzzy intercepting step: be mapped in network structure by fuzzy equivalent matrix, obtain fuzzy equivalence relation class, utilizes fuzzy threshold
Value intercepts this fuzzy equivalence relation class, obtains the fuzzy community structure cluster in corresponding Fuzzy Threshold level, and comes by modularity
Evaluate the Clustering Effect of fuzzy community.
Multiresolution community discovery method based on fuzzy clustering the most according to claim 1, it is characterised in that institute
State fuzzy switch process to include:
Step M1: network to be analyzed is set up the adjacency matrix A of network, is numbered node according to node sequence, compiles
Number from the beginning of 1, building the N rank square formation that element is 0 or 1, wherein N is the total number of network node;
Step M2: obtain the similarity between adjacent node, is converted into fuzzy resembling relation by deterministic syntopy,
Realization will abut against matrix A and is converted to fuzzy relationship matrix r;
Step M3: for the similarity between tolerance nonneighbor node further, fuzzy relationship matrix r is carried out fuzzy transmission and becomes
Change, make internodal similarity reach consistent stability;
Step M4: the uniformly convergent fuzzy relation obtained by fuzzy transmission function, sets up fuzzy transfer matrix, and
Obtain fuzzy equivalent matrix.
Multiresolution community discovery method based on fuzzy clustering the most according to claim 2, it is characterised in that institute
State in step M2, obtain the similarity between adjacent node, described structure phase according to the structural similarity metric algorithm of definition
As follows like the computing formula of property metric algorithm:
In formula, u, v are respectively the arbitrary node in set of network nodes, and Γ () represents the adjacent node set of certain node, Γ (u)
Representing the adjacent node set of node u, Γ (v) represents the adjacent node set of node v, and w () represents and connects between certain two node
The weight on limit, (u, x) connects the weight on limit to w between expression node u and node x, (v x) represents and connects between node v and node x w
The weighting structure similarity on limit, s (u, v) ∈ [0,1];For connecting the weight on limit between undirected arbitrary node without weighting network
W ()=1, then the formula of structural similarity metric algorithm is further simplified as following form:
Multiresolution community discovery method based on fuzzy clustering the most according to claim 2, it is characterised in that institute
State the fuzzy transmission transforming function transformation function in step M3 as follows:
In formula, the dimension of n representing matrix, n=| V |, V represent network node sequence vector, and R represents fuzzy relation matrix,
T (R) represents fuzzy equivalence relation, and U represents that fuzzy relation synthesizes computing, and k represents the number of times of fuzzy relation synthesis computing, its
MeetFuzzy equivalence relation meets
Multiresolution community discovery method based on fuzzy clustering the most according to claim 2, it is characterised in that institute
State fuzzy equivalent matrix, have the property that
Character 1: symmetry;For Undirected networks structure, internodal fuzzy relation meets symmetry;I.e. node i is to joint
Point j fuzzy relation r (i, j) be equivalent to node j to node i fuzzy relation r (j, i);
Character 2: reflexivity;In fuzzy relation matrix, any node is 1 to the fuzzy relation perseverance of self, i.e. and r (i, i)=1;
Character 3: transitivity;For the fuzzy equivalent matrix through fuzzy transmission conversion, meet
Multiresolution community discovery method based on fuzzy clustering the most according to claim 1, it is characterised in that
Described fuzzy intercepting step includes:
Step N1: fuzzy equivalence relation matrix and the network node one_to_one corresponding that will obtain, obtains fuzzy equivalence relation class, institute
State fuzzy equivalence relation class and be equivalent to the hierarchical clustering tree construction of network;
Step N2: selected arbitrary Fuzzy Threshold ε ∈ [0,1], intercepts fuzzy equivalence relation class, obtains the mould of correspondence
Stick with paste community division result;
Step N3: calculate the module angle value of corresponding fuzzy community division result;
Step N4: regulation Fuzzy Threshold ε, obtains the community structure under the conditions of different resolution.
Multiresolution community discovery method based on fuzzy clustering the most according to claim 6, it is characterised in that
In described step N3, the computing formula of modularity value Q is as follows:
In formula, i represents the community's quantity in network division, eiiRepresent that the quantity connecting limit between i-th community internal node accounts for whole
The ratio of individual network edge number, aiRepresent the ratio of one section of company's limit quantity being connected with i-th community interior joint.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610237640.3A CN105868791A (en) | 2016-04-15 | 2016-04-15 | Multi-resolution community discovering method based on fuzzy clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610237640.3A CN105868791A (en) | 2016-04-15 | 2016-04-15 | Multi-resolution community discovering method based on fuzzy clustering |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105868791A true CN105868791A (en) | 2016-08-17 |
Family
ID=56633254
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610237640.3A Pending CN105868791A (en) | 2016-04-15 | 2016-04-15 | Multi-resolution community discovering method based on fuzzy clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105868791A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106599187A (en) * | 2016-12-14 | 2017-04-26 | 北京航空航天大学 | Edge instability-based community discovery system and method |
CN107357858A (en) * | 2017-06-30 | 2017-11-17 | 中山大学 | A kind of network reconstruction method based on geographical position |
CN107609982A (en) * | 2017-09-14 | 2018-01-19 | 福州大学 | Consider community structure stability and the method that increment interdependent node carries out community discovery |
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 |
CN110097072A (en) * | 2019-03-19 | 2019-08-06 | 河南理工大学 | A kind of fuzzy clustering evaluation method based on two sub-module degree |
CN110163379A (en) * | 2019-05-30 | 2019-08-23 | 长沙医学院 | A kind of improved network module recognition methods based on Surprise optimization |
CN111241362A (en) * | 2020-01-13 | 2020-06-05 | 广州工程技术职业学院 | Fuzzy data clustering method, device, equipment and storage medium |
CN111353551A (en) * | 2020-03-13 | 2020-06-30 | 广东工业大学 | Non-uniform sample equalization method and system for product assembly process |
CN115527610A (en) * | 2022-11-09 | 2022-12-27 | 上海交通大学 | Cluster analysis method of unicellular omics data |
CN115630772A (en) * | 2022-12-19 | 2023-01-20 | 国网浙江省电力有限公司宁波供电公司 | Integrated energy detection and distribution method, system, equipment and storage medium |
-
2016
- 2016-04-15 CN CN201610237640.3A patent/CN105868791A/en active Pending
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106599187B (en) * | 2016-12-14 | 2020-06-16 | 北京航空航天大学 | Edge instability based community discovery system and method |
CN106599187A (en) * | 2016-12-14 | 2017-04-26 | 北京航空航天大学 | Edge instability-based community discovery system and method |
CN107357858A (en) * | 2017-06-30 | 2017-11-17 | 中山大学 | A kind of network reconstruction method based on geographical position |
CN107357858B (en) * | 2017-06-30 | 2020-09-08 | 中山大学 | Network reconstruction method based on geographic position |
CN107609982A (en) * | 2017-09-14 | 2018-01-19 | 福州大学 | Consider community structure stability and the method that increment interdependent node carries out community discovery |
CN107609982B (en) * | 2017-09-14 | 2020-11-24 | 福州大学 | Method for carrying out community discovery by considering community structure stability and increment related nodes |
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 |
CN110097072A (en) * | 2019-03-19 | 2019-08-06 | 河南理工大学 | A kind of fuzzy clustering evaluation method based on two sub-module degree |
CN110097072B (en) * | 2019-03-19 | 2022-10-04 | 河南理工大学 | Fuzzy clustering evaluation method based on two-degree-of-modularity |
CN110163379A (en) * | 2019-05-30 | 2019-08-23 | 长沙医学院 | A kind of improved network module recognition methods based on Surprise optimization |
CN110163379B (en) * | 2019-05-30 | 2023-04-07 | 长沙医学院 | Improved network module identification method based on Surrise optimization |
CN111241362A (en) * | 2020-01-13 | 2020-06-05 | 广州工程技术职业学院 | Fuzzy data clustering method, device, equipment and storage medium |
CN111353551A (en) * | 2020-03-13 | 2020-06-30 | 广东工业大学 | Non-uniform sample equalization method and system for product assembly process |
US11429070B2 (en) | 2020-03-13 | 2022-08-30 | Guangdong University Of Technology | Inhomogeneous sample equalization method and system for product assembly process |
CN115527610A (en) * | 2022-11-09 | 2022-12-27 | 上海交通大学 | Cluster analysis method of unicellular omics data |
CN115527610B (en) * | 2022-11-09 | 2023-11-24 | 上海交通大学 | Cluster analysis method for single-cell histology data |
CN115630772A (en) * | 2022-12-19 | 2023-01-20 | 国网浙江省电力有限公司宁波供电公司 | Integrated energy detection and distribution method, system, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105868791A (en) | Multi-resolution community discovering method based on fuzzy clustering | |
Wang et al. | Urban expansion patterns and their driving forces based on the center of gravity-GTWR model: A case study of the Beijing-Tianjin-Hebei urban agglomeration | |
Zhu et al. | Construction of membership functions for predictive soil mapping under fuzzy logic | |
CN103500344B (en) | Method and module for extracting and interpreting information of remote-sensing image | |
CN109272170B (en) | A kind of traffic zone dividing system based on Louvain algorithm | |
CN106599090A (en) | Structure centrality-based network community discovery method | |
CN102750333B (en) | Method for extracting semiconductor nano-structure feature size | |
CN103903012A (en) | Polarimetric SAR data classifying method based on orientation object and support vector machine | |
CN104966075B (en) | A kind of face identification method and system differentiating feature based on two dimension | |
Chen et al. | Agricultural remote sensing image cultivated land extraction technology based on deep learning | |
CN109614520A (en) | One kind is towards the matched parallel acceleration method of multi-mode figure | |
CN111783879A (en) | Hierarchical compression map matching method and system based on orthogonal attention mechanism | |
Li | Revealing network communities with a nonlinear programming method | |
CN108764541A (en) | A kind of wind energy prediction technique of combination space-time characteristic and Error processing | |
CN109472712A (en) | A kind of efficient Markov random field Combo discovering method strengthened based on structure feature | |
Zhao et al. | Anyscan: An efficient anytime framework with active learning for large-scale network clustering | |
Dahal | Effect of different distance measures in result of cluster analysis | |
Sathik et al. | Comparative analysis of community discovery methods in social networks | |
CN107240028A (en) | The overlapping community discovery and Forecasting Methodology of asymmetric corporations extension based on node liveness in complex network | |
Fanoiki et al. | Case-based reasoning retrieval and reuse using case resemblance hypergraphs | |
Ye et al. | A study of destination selection model based on link flows | |
CN109523143B (en) | Land evaluation method based on multi-granularity calculation | |
Zhou et al. | Evidential communities for complex networks | |
Zhang et al. | Comparative Variance and Multiple Imputation Used for Missing Values in Land Price DataSet. | |
Ferdowsi et al. | Generating high-quality synthetic graphs for community detection in social networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160817 |
|
RJ01 | Rejection of invention patent application after publication |