CN103457800A - Network community detection method based on M elite coevolution strategy - Google Patents
Network community detection method based on M elite coevolution strategy Download PDFInfo
- Publication number
- CN103457800A CN103457800A CN2013104049101A CN201310404910A CN103457800A CN 103457800 A CN103457800 A CN 103457800A CN 2013104049101 A CN2013104049101 A CN 2013104049101A CN 201310404910 A CN201310404910 A CN 201310404910A CN 103457800 A CN103457800 A CN 103457800A
- Authority
- CN
- China
- Prior art keywords
- community
- network
- web community
- web
- divides
- 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
Images
Abstract
The invention discloses a network community detection method based on an M elite coevolution strategy, wherein the network community detection method solves the problems that in the prior art, the convergence rate is low and easily lapses into the local optimum, multiresolution analysis of a network structure cannot be achieved. The implementation steps include that (1) network data are loaded; (2) network community populations are initialized; (3) the network community populations are divided; (4) a network community team is organized; (5) candidate network community division is detected; (6) the network community populations are updated; (7) local network communities are detected; (8) the network community populations are updated; (9) whether iteration is terminated or not is judged; (10) a network community detection result is output. When the network community detection method is used for detecting community structures in a network, expanded module density functions serve as fitness functions, a network structure is analyzed with different resolutions, and the convergence rate is quickened through leading-in of local detection and does not easily lapse into the local optimum.
Description
Technical field
The invention belongs to technical field of the computer network, further relate to the community network detection method based on M elite coevolution strategy of field of artificial intelligence.The present invention, by using the expansion module density function as fitness function, introduces M elite Cooperative Evolutionary Algorithm, finds the community structure of the different resolution in real network, has higher convergence rate and stability.The present invention can be used for solving the community structure test problems in network.
Background technology
A lot of complication systems of real world can be expressed as network, as World Wide Web (WWW), and power network, bio-networks and social networks etc.Except the worldlet effect, outside the network attributes such as scale, community structure is another one important property in complex network structures.Community refers in network that similarity is higher or interconnects the set of node closely.This community structure analysis to network in real world has important directive significance: in social networks, community structure can help to analyze interpersonal interpersonal relationships; In large scale integrated circuit, community structure can help the function of our analysis circuit, the one-step optimization circuit layout of going forward side by side; In molecular biology, community structure can help our analysing protein structure and predict its function.The complex network community structure detects the various fields such as various bio-networks analyses such as being widely used in metabolic network analysis, the analysis of the protein Internet and Web community mining at present.
At present, there is Various Complex Web Community detection method, according to taked basic solution strategies, can be summarized as two large classes: method and heuristic based on optimizing.The former is converted into optimization problem by complex network community test problems, carrys out the community structure of calculation of complex network by the predefined target function of optimization, and the latter is converted into complex network community test problems the design problem of predefine heuristic rule.
Chongqing Mail and Telephones Unvi is at patent " a kind of distributed opportunistic network community division method " (number of patent application 201210178330.0 of its application, a kind of division methods of distributed opportunistic network community structure is disclosed publication number CN102685255A), the method is by the period of motion characteristic of Web Community's node and the historical information of meeting, improve the accuracy that the opportunistic network community divides, but can not analyze network configuration with different resolution.
Xian Electronics Science and Technology University discloses a kind of method of utilizing close female algorithm solution complex network structures to detect in the patent " based on close female community structure detection method of calculating " (number of patent application 201110366154.9, publication number CN102521649A) of its application.The method is using the modularity density of expansion as fitness function, thereby analyze network configuration with different resolution, and, by introducing Local Search convergence speedup speed, although the deficiency of the method is accelerated convergence rate but easily be absorbed in local optimum, cause the method robustness poor.
Summary of the invention
The object of the invention is to overcome the deficiency of above-mentioned prior art, propose a kind of based on M elite coevolution strategy community network structure detection method.The present invention passes through using the expansion module density function as fitness function, introduce simulated annealing and detect strategy as part, to solve resolution limit in existing network community structure detection method, poor robustness, easily to be absorbed in the shortcoming such as local optimum, improved the accuracy that Web Community is detected.
The specific embodiment of the invention step is as follows:
(1) be written into network data:
The adjacency matrix A (N*N) of tectonic network, the number that N is nodes, if Web Community's node i has while connection with the node j of Web Community, the element a in adjacency matrix
ij=1; If Web Community's node i with the node j of Web Community without being connected, a
ij=0.
(2) initialization Web Community population:
Adopt the direct coding mode, generate at random N integer numerical value that is no more than Web Community's interstitial content, by these integer numerical value difference marks, give S gene position on every chromosome; Repeat above operation, until obtain W bar chromosome, every chromosome represents that a kind of Web Community divides, and forms the population θ of Web Community by W bar chromosome.
(3) divide the Web Community population:
3a) by fitness function, the fitness value that in the population θ of computing network community, each Web Community divides;
The fitness value of 3b) all-network community in the Web Community population being divided is sorted from high to low, divides front 40% Web Community corresponding to fitness value in sequence into elite Web Community population, and remainder divides general network community population into.
(4) building network community team:
Each Web Community in elite Web Community population is divided and is set as the division of basic network community, respectively each basic network community division is divided to building network community team with the Web Community in general network community population, the number that in each team of Web Community, Web Community divides is as follows:
Wherein, G means the number that in each team of Web Community, Web Community divides,
mean to round up several Web Communities division operation, W means the total scale of elite Web Community population and general network community population, and M means the number of team of Web Community.
(5) detecting the candidate network community divides:
5a) to the random chance p of random generation in (0,1) interval of each team of Web Community, when random chance p is greater than 0.5, in elite Web Community population, Web Community of random selection divides B, execution step 5b); When random chance p is less than 0.5, a random division C of Web Community, the execution step 5c of selecting in general network community population).
5b) by the division E of the basic network community to current network community team and Web Community, divide B and carry out cooperative operation, detect and obtain the division δ of Web Community;
5c) by the division E of the basic network community to current network community team and Web Community, divide C and guide operation, detect and obtain the division φ of Web Community;
(6) upgrade the Web Community population:
The Web Community that detection is obtained divides δ or Web Community divides φ and the former population θ of Web Community is merged, obtain transition Web Community population, all-network community in transition Web Community population is divided and presses the fitness value sequence, and in sequence, the corresponding Web Community of front 450 fitness values divides the population α of network consisting community.
(7) detect the localized network community:
7a) select Web Community's division of fitness value maximum in the population α of Web Community to carry out part detection operation;
7b) adopt simulated annealing, Web Community's division of fitness value maximum in the population α of Web Community is carried out to part and detect, detect and obtain an optimal network community division.
(8) upgrade the Web Community population:
The Web Community that divides fitness value maximum in the population α of alternate network community with the optimal network community divides.
(9) judge whether termination of iterations:
Judge whether iterations reaches 40 times, if reach, perform step (10); Otherwise, execution step (3).
(10) output network community testing result:
Web Community's division result that output finally detects.
The present invention compared with prior art has the following advantages:
The first, because the present invention has adopted the expansion module density function, overcome the shortcoming that prior art can't the multiresolution analysis network configuration, the present invention is had advantages of and can detect the community structure of multiple resolution.
The second, because the present invention has adopted cooperation and guided two kinds of operations to guarantee the diversity of population, overcome prior art and easily be absorbed in the shortcoming of local optimum, the present invention has been had advantages of and can detect community structure more accurately.
The accompanying drawing explanation
Fig. 1 is flow chart of the present invention;
Fig. 2 is emulation of the present invention karate club network data model schematic diagram used;
Fig. 3 is emulation of the present invention dolphin social network data model schematic diagram used;
Fig. 4 is for adopting the present invention to detect the result schematic diagram of karate club network;
Fig. 5 is for adopting the present invention to detect the result schematic diagram of dolphin social networks.
Embodiment
Below in conjunction with accompanying drawing, the present invention will be further described.
With reference to accompanying drawing 1, the concrete implementation step of the present invention is as follows:
Step 1. is written into network data.
The adjacency matrix A (N*N) of tectonic network, the number that N is nodes, if Web Community's node i has while connection with the node j of Web Community, the element a in adjacency matrix
ij=1; If Web Community's node i with the node j of Web Community without being connected, a
ij=0.
Adopt the direct coding mode, generate at random N integer numerical value that is no more than Web Community's interstitial content, by these integer numerical value difference marks, give S gene position on every chromosome; Repeat above operation, until obtain W bar chromosome, every chromosome represents that a kind of Web Community divides, and forms the population θ of Web Community by W bar chromosome.
Pass through fitness function, the fitness value that in the population θ of computing network community, each Web Community divides, the fitness value that all-network community in the Web Community population is divided is sorted from high to low, divide front 40% Web Community corresponding to fitness value in sequence into elite Web Community population, remainder is divided into general network community population.Fitness function is as follows:
Wherein, D
λmean the fitness function value, λ means the resolution adjustment parameter of Web Community, and m means the number of Web Community, V
imean i the set formed by Web Community's node,
mean except V
ioutside the set that forms of other Web Community node, L (V
i, V
i) mean the interior degree of all-network community node in the i of Web Community,
mean in the i of Web Community outside all-network community node and i the outer degree of Web Community's node in other Web Community, | V
i| mean the number of the Web Community's node in the i of Web Community.
Each Web Community in elite Web Community population is divided and is set as the division of basic network community, respectively each basic network community is divided with the Web Community's division in general network community population and set up into team of Web Community, the number that in each team of Web Community, Web Community divides is as follows:
Wherein, G means the number that in each team of Web Community, Web Community divides,
mean to round up several Web Communities division operation, W means the total scale of elite Web Community population and general network community population, and M means the number of team of Web Community.
A random chance p of random generation in interval in (0,1) to each team of Web Community when random chance p is greater than 0.5, selects at random a Web Community to divide B in elite Web Community population, carries out cooperative operation; When random chance p is less than 0.5, in general network community population, Web Community of random selection divides C, carries out the guiding operation.
The cooperative operation process is as follows:
The 1st step, carry out first via interlace operation by following formula:
Wherein, E
kmean that the basic network community divides the classification mark of upper k the Web Community's node of E, ← expression assign operation, B
jmean that elite Web Community divides the classification mark of upper j the Web Community's node of B,
mean " to arbitrary " symbol, j means that classification mark and elite Web Community divide the classification mark B of upper j the node of B
jidentical Web Community's node, ∈ means " belonging to " symbol, | mean conditional code, B
kmean that elite Web Community divides the classification mark of upper k the Web Community's node of B.
The 2nd step, carry out the second tunnel interlace operation by following formula:
Wherein, B
kmean that elite Web Community divides the classification mark of upper k the Web Community's node of B, ← expression assign operation, E
jmean that the basic network community divides the classification mark of upper j the Web Community's node of E,
mean " to arbitrary " symbol, k means that classification mark and basic network community divide the classification mark E of the upper the j node of E
jidentical Web Community's node, ∈ means " belonging to " symbol, | mean the conditional code in probability theory, E
kmean that the basic network community divides the classification mark of upper k the Web Community's node of E.
The guiding operating process is as follows:
The 1st step, arrange and select probability β=0.3;
The 2nd step, in (0,1) the interval interior random random number that generates, if β is greater than this random number, carry out the 3rd step; Otherwise, carry out the 4th step;
The 3rd step, carry out interlace operation by following formula:
Wherein, C
kmean that the general network community divides the classification mark of upper k the Web Community's node of C, ← expression assign operation, E
jmean that the basic network community divides the classification mark of upper j the Web Community's node of E,
mean " to arbitrary " symbol, k means that classification mark and basic network community divide upper j the node classification mark E of Web Community of E
jidentical Web Community's node, ∈ means " belonging to " symbol, | mean the conditional code in probability theory, E
kmean that the basic network community divides the classification mark of upper k the Web Community's node of E.
The 4th step, dividing Web Community's Node configuration of random selection in C in the general network community is change point, the category label of change point is changed to another one and this node to be had and is connected but the category label of the node in same Web Community not, obtains a new Web Community and divides.
The Web Community that detection is obtained divides δ or Web Community divides φ and the former population θ of Web Community is merged, obtain transition Web Community population, transition Web Community population is pressed to the fitness value sequence, and in sequence, the corresponding Web Community of front 450 fitness values divides the population α of network consisting community.
Select the Web Community of the fitness value maximum in the population α of Web Community to divide, by simulated annealing, this Web Community is divided and carries out the part detection, obtain the optimal network community and divide.Concrete steps are as follows:
The 1st step, in the current network community divides, Web Community's Node configuration of random selection is change point, the category label of change point is changed to another one and this node to be had and is connected but the category label of the node in same Web Community not, obtains an adjacent networks community and divides; Repeat above step, divide until obtain 180 adjacent networks communities.
The 2nd step, by fitness function, calculate the fitness value that all adjacent networks community divides.
The 3rd step, the fitness value that relatively each adjacent networks community divides and the current network community divides, if the former is greater than the latter, retains this adjacent networks community and divide; If the former is less than the latter, in (0,1) the interval random random number that generates, judge the size of this random number and acceptance probability P, if P is greater than this random number, retains this adjacent networks community and divide; Otherwise do not retain.
The expression formula of acceptance probability P is as follows:
P=exp(-(X-Y)/T)
Wherein, exp () means the operation of fetching number, and X means that the current network community divides fitness value, and Y means the fitness value that the adjacent networks community divides, and T means annealing temperature.
The 4th step, upgrade annealing temperature T by the temperature renewal function, and the temperature renewal function is as follows:
T=ξ*T
0
Wherein, ξ means annealing coefficient, wherein ξ=0.85; T
0mean initial temperature, T
0get the variance that in the population α of Web Community, the all-network community divides fitness value.
The 5th step, judge whether annealing temperature T reaches stopping criterion for iteration, if T<0.005, the optimal network community that output detections obtains divides; Otherwise return to the 3rd step.
The Web Community that divides fitness value maximum in the population α of alternate network community with the optimal network community divides.
Judge whether iterations reaches 40 times, if reach, perform step 10; Otherwise, perform step 3.
Web Community's division result that output finally detects.
Effect of the present invention can further illustrate by following emulation:
1. simulated conditions:
The present invention is to be Intel Core22.3GHz at CPU, and internal memory 2G is used Matlab2010 to carry out emulation on the Windows7 system.
2. emulation content:
Choose karate network and dolphin social networks as simulation object.
With reference to accompanying drawing 2, network model in Fig. 2 means karate club network of the U.S., have 34 Web Community's nodes, the line between Web Community's node means between node to exist contact, and the numeral in circle and Web Community's node be label one to one.
With reference to accompanying drawing 3, network model in Fig. 3 means the bottle-nosed dolphin network of New Zealand Dao Erfu fyord, have 62 Web Community's nodes, the line between Web Community's node means between node to exist contact, and the numeral in circle and Web Community's node be label one to one.
With reference to accompanying drawing 4, when the network model in Fig. 4 means resolution adjustable parameter λ is got to different value, the present invention is to karate club network of network community testing result.
With reference to accompanying drawing 4 (a), network model in Fig. 4 (a) means when resolution adjustable parameter λ=0.3, the present invention is to karate club network of network community testing result, the represented Web Community's node { 1 of circle in Fig. 4 (a), 2, 3, 4, 5, 6, 7, 8, 11, 12, 13, 14, 17, 18, 20, 22 } form a Web Community, numeral in circle and Web Community's node be label one to one, the represented Web Community's node { 10 of square in Fig. 4 (a), 15, 16, 19, 21, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34 } form a Web Community, numeral in square and Web Community's node be label one to one, different shapes represents different Web Communities.
With reference to accompanying drawing 4 (b), network model in Fig. 4 (b) means when resolution adjustable parameter λ=0.4, the present invention is to karate club network of network community testing result, the represented Web Community's node { 1 of circle in Fig. 4 (b), 2, 3, 4, 8, 12, 13, 14, 18, 20, 22} is a Web Community, numeral in circle and Web Community's node be label one to one, the represented Web Community's node { 5 of rhombus in Fig. 4 (b), 6, 7, 11, 17} forms a Web Community, numeral in rhombus and Web Community's node be label one to one, the represented Web Community's node { 10 of square in Fig. 4 (b), 15, 16, 19, 21, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34} forms a Web Community, numeral in square and Web Community's node be label one to one, different shapes represents different Web Communities.
With reference to accompanying drawing 4 (c), network model in Fig. 4 (c) means when resolution adjustable parameter λ=0.6, the present invention is to karate club network of network community testing result, the represented Web Community's node { 1 of circle in Fig. 4 (c), 2, 3, 4, 8, 12, 13, 14, 18, 20, 22} forms a Web Community, numeral in circle and Web Community's node be label one to one, the represented Web Community's node { 5 of rhombus in Fig. 4 (c), 6, 7, 11, 17} forms a Web Community, numeral in rhombus and Web Community's node be label one to one, the represented Web Community's node { 10 of square in Fig. 4 (c), 15, 16, 19, 21, 23, 24, 27, 28, 30, 31, 33, 34} forms a Web Community, numeral in square and Web Community's node be label one to one, the represented Web Community's node { 25 of Fig. 4 (c) intermediate cam shape, 26, 29, 32} forms a Web Community, numeral in triangle and Web Community's node be label one to one, different shapes represents different communities.
With reference to accompanying drawing 5, when the network model in Fig. 5 means resolution adjustable parameter λ is got to different value, the Web Community division result of the present invention to the dolphin social networks.
With reference to accompanying drawing 5 (a), network model in Fig. 5 (a) means when resolution adjustable parameter λ=0.4, the Web Community testing result of the present invention to the dolphin social networks, the represented Web Community's node { 1 of circle in Fig. 5 (a), 3, 4, 5, 9, 11, 12, 13, 15, 16, 17, 19, 21, 22, 24, 25, 29, 30, 31, 34, 35, 36, 37, 38, 39, 40, 41, 43, 44, 45, 46, 47, 48, 50, 51, 52, 53, 54, 56, 59, 60, 62} forms a Web Community, numeral in circle and Web Community's node be label one to one, the represented Web Community's node { 2 of square in Fig. 5 (a), 6, 7, 8, 10, 14, 18, 20, 23, 26, 27, 28, 32, 33, 42, 49, 55, 57, 58, 61} forms a Web Community, numeral in square and Web Community's node be label one to one, different shapes represents different communities.
With reference to accompanying drawing 5 (b), network model in Fig. 5 (b) means when resolution adjustable parameter λ=0.6, the Web Community testing result of the present invention to the dolphin social networks, the represented Web Community's node { 1 of circle in Fig. 5 (b), 3, 4, 9, 11, 13, 15, 17, 21, 29, 31, 34, 35, 37, 38, 39, 40, 41, 43, 44, 45, 47, 48, 50, 51, 53, 54, 59, 60, 62} forms a Web Community, numeral in circle and Web Community's node be label one to one, the represented Web Community's node { 5 of Fig. 5 (b) intermediate cam shape, 12, 16, 19, 22, 24, 25, 30, 36, 46, 52, 56} forms a Web Community, numeral in triangle and Web Community's node be label one to one, the represented Web Community's node { 2 of square in Fig. 5 (b), 6, 7, 8, 10, 14, 18, 20, 23, 26, 27, 28, 32, 33, 42, 49, 55, 57, 58, 61} forms a Web Community, numeral in square and Web Community's node be label one to one, different shapes represents different communities.
For further analyzing advantage of the present invention, the present invention and the close female algorithm MA of prior art are contrasted, to the simulation result of accompanying drawing 4 and accompanying drawing 5 respectively as following table 1 and table 2:
In table 1 and table 2, λ means resolution adjustable parameter, D
λmean 20 gained fitness evaluation values of emulation, VAR means the variance of 20 operation results of emulation.
Table 1 the present invention and the close female algorithm MA contrast simulation result to the karate network
Table 2 the present invention and the close female algorithm MA contrast simulation result to the dolphin social networks
Can find out the fitness function value D that the present invention obtains from table 1 and table 2
λbe better than close female algorithm MA, i.e. the present invention can detect community structure effectively; From the comparing result of operation result variance yields, variance of the present invention will be significantly less than close female algorithm MA, and operation result is obviously stable than close female algorithm MA.
Comprehensive above simulation result can find out, when the controlled resolution parameter lambda is got different value, for same network, the present invention not only can effectively detect the community structure of different resolution, and reliable and stablely is difficult for being absorbed in local optimum.Therefore, the Web Community's detection method based on M elite coevolution strategy can effectively detect the community structure of different resolution.
Claims (5)
1. the Web Community's detection method based on M elite coevolution strategy, its specific implementation step is as follows:
(1) be written into network data:
The adjacency matrix A (N*N) of tectonic network, the number that N is nodes, if network node i has while connection with network node j, the element a in adjacency matrix
ij=1; If network node i with network node j without being connected, a
ij=0;
(2) initialization Web Community population:
Adopt the direct coding mode, generate at random N and be no more than number of network node purpose integer numerical value, by these integer numerical value difference marks, give S gene position on every chromosome; Repeat above operation, until obtain W bar chromosome, every chromosome represents that a kind of Web Community divides, and forms the population θ of Web Community by W bar chromosome;
(3) divide the Web Community population:
3a) by fitness function, the fitness value that in the population θ of computing network community, each Web Community divides;
The fitness value of 3b) all-network community in the population θ of Web Community being divided is sorted from high to low, front 40% Web Community corresponding to fitness value in sequence is divided into to elite Web Community population, and remainder is divided into general network community population;
(4) building network community team:
Each Web Community in elite Web Community population is divided and is set as the division of basic network community, respectively each basic network community is divided with the Web Community's division in general network community population and set up into team of Web Community, the number that in each team of Web Community, Web Community divides is as follows:
Wherein, G means the number that in each team of Web Community, Web Community divides,
mean to round up several Web Communities division operation, W means the total scale of elite Web Community population and general network community population, and M means the number of team of Web Community;
(5) detecting the candidate network community divides:
5a) to the random chance p of random generation in (0,1) interval of each team of Web Community, when random chance p is greater than 0.5, in elite Web Community population, Web Community of random selection divides B, execution step 5b); When random chance p is less than 0.5, a random division C of Web Community, the execution step 5c of selecting in general network community population);
5b) by the division E of the basic network community to current network community team and Web Community, divide B and carry out cooperative operation, detect the Web Community that obtains the candidate and divide δ;
5c) by the division E of the basic network community to current network community team and Web Community, divide C and guide operation, detect the Web Community that obtains the candidate and divide φ;
(6) upgrade the Web Community population:
The Web Community that detection is obtained divides δ or Web Community divides φ and the former population θ of Web Community is merged, obtain transition Web Community population, all-network community in transition Web Community population is divided and presses the fitness value sequence, and in sequence, the corresponding Web Community of front 450 fitness values divides the population α of network consisting community;
(7) detect the localized network community:
7a) select Web Community's division of fitness value maximum in the population α of Web Community to carry out part detection operation;
7b) adopt simulated annealing, Web Community's division of fitness value maximum in the population α of Web Community is carried out to part and detect, detect and obtain an optimal network community division;
(8) upgrade the Web Community population:
The Web Community that divides fitness value maximum in the population α of alternate network community with the optimal network community divides;
(9) judge whether termination of iterations:
Judge whether iterations reaches 40 times, if reach, perform step (10); Otherwise, execution step (3);
(10) output network community testing result:
Web Community's division result that output finally detects.
2. the Web Community's detection method based on M elite coevolution strategy according to claim 1, is characterized in that step 3a) described fitness function is as follows:
Wherein, D
λmean the fitness function value, λ means the resolution adjustment parameter of Web Community, and m means the number of Web Community, V
imean i the set formed by network node,
mean except V
ioutside the set that forms of other network node, L (V
i, V
i) mean the interior degree of all-network node in the i of Web Community,
mean in the i of Web Community outside all-network node and i the outer degree of network node in other Web Community, | V
i| mean the number of the network node in the i of Web Community.
3. the Web Community's detection method based on M elite coevolution strategy according to claim 1, is characterized in that step 5b) described cooperative operation, concrete operation step is as follows:
The 1st step, carry out first via interlace operation by following formula:
Wherein, E
kmean that the basic network community divides the classification mark of upper k the network node of E, ← expression assign operation, B
jmean that Web Community divides the classification mark of upper j the network node of B,
mean " to arbitrary " symbol, j means that classification mark and Web Community divide the classification mark B of upper j the node of B
jidentical network node, ∈ means " belonging to " symbol, | mean conditional code, B
kmean that Web Community divides the classification mark of upper k the network node of B;
The 2nd step, carry out the second tunnel interlace operation by following formula:
Wherein, B
kmean that Web Community divides the classification mark of upper k the network node of B, ← expression assign operation, E
jmean that the basic network community divides the classification mark of upper j the network node of E,
mean " to arbitrary " symbol, k means that classification mark and basic network community divide the classification mark E of the upper the j node of E
jidentical network node, ∈ means " belonging to " symbol, | mean the conditional code in probability theory, E
kmean that the basic network community divides the classification mark of upper k the network node of E.
4. the Web Community's detection method based on M elite coevolution strategy according to claim 1, is characterized in that step 5c) described guiding operation, concrete steps are as follows:
The 1st step, arrange and select probability β=0.3;
The 2nd step, in (0,1) the interval interior random random number that generates, if β is greater than this random number, carry out the 3rd step; Otherwise, carry out the 4th step;
The 3rd step, carry out a road interlace operation by following formula:
Wherein, C
kmean that the general network community divides the classification mark of upper k the network node of C, ← expression assign operation, E
jmean that the basic network community divides the classification mark of upper j the network node of E,
mean " to arbitrary " symbol, k means that classification mark and basic network community divide upper j the network node classification mark E of E
jidentical network node, ∈ means " belonging to " symbol, | mean the conditional code in probability theory, E
kmean that the basic network community divides the classification mark of upper k the network node of E;
The 4th step, divide a network node of random selection in C in the general network community and be set to change point, the category label of change point is changed to another one and this node to be had and is connected but the category label of the node in same Web Community not, obtains a new Web Community and divides.
5. the Web Community's detection method based on M elite coevolution strategy according to claim 1, is characterized in that step 7b) concrete steps of described simulated annealing are as follows:
The 1st step, in the current network community divides, a network node of random selection is set to change point, the category label of change point is changed to another one and this node to be had and is connected but the category label of the node in same Web Community not, obtains an adjacent networks community and divides; Repeat above step, divide until obtain 180 adjacent networks communities;
The 2nd step, by fitness function, calculate the fitness value that all adjacent networks community divides;
The 3rd step, the fitness value that relatively each adjacent networks community divides and the current network community divides, if the former is greater than the latter, retains this adjacent networks community and divide; If the former is less than the latter, in (0,1) the interval random random number that generates, judge the size of this random number and acceptance probability P, if P is greater than this random number, retains this adjacent networks community and divide; Otherwise do not retain; Acceptance probability P is as follows:
P=exp(-(X-Y)/T)
Wherein, P means acceptance probability, and exp () means the operation of fetching number, and X means that the current network community divides fitness value, and Y means the fitness value that the adjacent networks community divides, and T means annealing temperature;
The 4th step, upgrade annealing temperature T by the temperature renewal function, and the temperature renewal function is as follows:
T=ξ*T
0
Wherein, ξ means annealing coefficient, wherein ξ=0.85; T
0mean initial temperature, T
0get the variance that in the population α of Web Community, the all-network community divides fitness value;
The 5th step, judge whether annealing temperature T reaches stopping criterion for iteration, if T<0.005, the optimal network community that output detections obtains divides; Otherwise, return to the 3rd step.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013104049101A CN103457800A (en) | 2013-09-08 | 2013-09-08 | Network community detection method based on M elite coevolution strategy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013104049101A CN103457800A (en) | 2013-09-08 | 2013-09-08 | Network community detection method based on M elite coevolution strategy |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103457800A true CN103457800A (en) | 2013-12-18 |
Family
ID=49739785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013104049101A Pending CN103457800A (en) | 2013-09-08 | 2013-09-08 | Network community detection method based on M elite coevolution strategy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103457800A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104268270A (en) * | 2014-10-13 | 2015-01-07 | 北京建筑大学 | Map Reduce based method for mining triangles in massive social network data |
| CN107480213A (en) * | 2017-07-27 | 2017-12-15 | 上海交通大学 | Community's detection and customer relationship Forecasting Methodology based on sequential text network |
| CN108154233A (en) * | 2018-01-15 | 2018-06-12 | 大连民族大学 | The complex network community detection method of global convergence performance can be improved |
| CN109218304A (en) * | 2018-09-12 | 2019-01-15 | 北京理工大学 | A kind of network risks blocking-up method based on attack graph and coevolution |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521649A (en) * | 2011-11-17 | 2012-06-27 | 西安电子科技大学 | Network community structure detection method based on memetic computation |
| CN103455612A (en) * | 2013-09-07 | 2013-12-18 | 西安电子科技大学 | Method for detecting non-overlapping network communities and overlapping network communities based on two-stage strategy |
-
2013
- 2013-09-08 CN CN2013104049101A patent/CN103457800A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521649A (en) * | 2011-11-17 | 2012-06-27 | 西安电子科技大学 | Network community structure detection method based on memetic computation |
| CN103455612A (en) * | 2013-09-07 | 2013-12-18 | 西安电子科技大学 | Method for detecting non-overlapping network communities and overlapping network communities based on two-stage strategy |
Non-Patent Citations (4)
| Title |
|---|
| 慕彩红,: ""协同进化数值优化算法及其应用研究"", 《中国博士学位论文全文数据库-信息科技辑》, 15 October 2010 (2010-10-15), pages 140 - 4 * |
| 慕彩红,焦李成,刘逸,: ""M-精英协同进化数值优化算法"", 《软件学报》, vol. 20, no. 11, 18 January 2010 (2010-01-18) * |
| 慕彩红,焦李成,刘逸,: ""M-精英进化算法及其在V-BLAST系统中的应用"", 《电子与信息学报》, vol. 31, no. 10, 18 December 2009 (2009-12-18) * |
| 慕彩红,焦李成,刘逸,: ""求解约束优化问题M-精英协同进化算法"", 《西安电子科技大学学报（自然科学版）》, vol. 37, no. 5, 31 May 2010 (2010-05-31) * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104268270A (en) * | 2014-10-13 | 2015-01-07 | 北京建筑大学 | Map Reduce based method for mining triangles in massive social network data |
| CN107480213A (en) * | 2017-07-27 | 2017-12-15 | 上海交通大学 | Community's detection and customer relationship Forecasting Methodology based on sequential text network |
| CN107480213B (en) * | 2017-07-27 | 2021-12-24 | 上海交通大学 | Community detection and user relation prediction method based on time sequence text network |
| CN108154233A (en) * | 2018-01-15 | 2018-06-12 | 大连民族大学 | The complex network community detection method of global convergence performance can be improved |
| CN109218304A (en) * | 2018-09-12 | 2019-01-15 | 北京理工大学 | A kind of network risks blocking-up method based on attack graph and coevolution |
| CN109218304B (en) * | 2018-09-12 | 2020-09-25 | 北京理工大学 | Network risk blocking method based on attack graph and co-evolution |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Corizzo et al. | Anomaly detection and repair for accurate predictions in geo-distributed big data | |
| Chen et al. | Ensemble correlation-based low-rank matrix completion with applications to traffic data imputation | |
| Gu et al. | Semi-supervised deep extreme learning machine for Wi-Fi based localization | |
| CN103455612A (en) | Method for detecting non-overlapping network communities and overlapping network communities based on two-stage strategy | |
| CN104200087B (en) | For the parameter optimization of machine learning and the method and system of feature tuning | |
| CN103810288B (en) | Method for carrying out community detection on heterogeneous social network on basis of clustering algorithm | |
| CN103455610B (en) | Network community detecting method based on multi-objective memetic computation | |
| Ji et al. | Graph model-based salient object detection using objectness and multiple saliency cues | |
| CN106934722A (en) | Multi-objective community detection method based on k node updates Yu similarity matrix | |
| CN111628494B (en) | Low-voltage distribution network topology identification method and system based on logistic regression method | |
| Pan et al. | Towards better accuracy for missing value estimation of epistatic miniarray profiling data by a novel ensemble approach | |
| CN103457800A (en) | Network community detection method based on M elite coevolution strategy | |
| CN104881735A (en) | System and method of smart power grid big data mining for supporting smart city operation management | |
| CN104268629A (en) | Complex network community detecting method based on prior information and network inherent information | |
| CN103905246A (en) | Link prediction method based on grouping genetic algorithm | |
| CN103885867B (en) | Online evaluation method of performance of analog circuit | |
| Herbster et al. | Fast prediction on a tree | |
| Jia et al. | TTSL: An indoor localization method based on Temporal Convolutional Network using time-series RSSI | |
| CN102760209A (en) | Transmembrane helix predicting method for nonparametric membrane protein | |
| CN102521655A (en) | Method for detecting dynamic network community on basis of non-dominated neighbor immune algorithm | |
| Wang et al. | A bluetooth location method based on kNN algorithm | |
| CN102855462A (en) | Method and device for determining human body action cycle and recognizing human body action | |
| CN102521649A (en) | Network community structure detection method based on memetic computation | |
| CN104156462A (en) | Complex network community mining method based on cellular automatic learning machine | |
| Jazizadeh et al. | Unsupervised clustering of residential electricity consumption measurements for facilitated user-centric non-intrusive load monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20131218 |