CN105631748A - Parallel label propagation-based heterogeneous network community discovery method - Google Patents
Parallel label propagation-based heterogeneous network community discovery method Download PDFInfo
- Publication number
- CN105631748A CN105631748A CN201510968384.0A CN201510968384A CN105631748A CN 105631748 A CN105631748 A CN 105631748A CN 201510968384 A CN201510968384 A CN 201510968384A CN 105631748 A CN105631748 A CN 105631748A
- Authority
- CN
- China
- Prior art keywords
- label
- heterogeneous network
- propagated
- community
- node
- 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 75
- 230000000644 propagated effect Effects 0.000 claims abstract description 36
- 230000004927 fusion Effects 0.000 claims description 16
- 238000010276 construction Methods 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 5
- 230000002452 interceptive effect Effects 0.000 abstract description 3
- 125000004122 cyclic group Chemical group 0.000 abstract 2
- 230000007246 mechanism Effects 0.000 description 38
- 235000013162 Cocos nucifera Nutrition 0.000 description 17
- 244000060011 Cocos nucifera Species 0.000 description 17
- 230000005540 biological transmission Effects 0.000 description 17
- 230000006854 communication Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 230000003997 social interaction Effects 0.000 description 7
- 230000001360 synchronised effect Effects 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 5
- 230000010354 integration Effects 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 238000007500 overflow downdraw method Methods 0.000 description 4
- 238000012217 deletion Methods 0.000 description 3
- 230000037430 deletion Effects 0.000 description 3
- 230000006855 networking Effects 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000004132 cross linking Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013401 experimental design Methods 0.000 description 1
- 230000003447 ipsilateral effect Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
Classifications
-
- 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
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a parallel label propagation-based heterogeneous network community discovery method. The method is characterized by comprising a label initialization step, a label cyclic refreshing step and a community building step. According to the label cyclic refreshing step, based on parallel label propagation, node labels are allowed to be propagated in multiple sub networks of the heterogeneous network in a relatively independent and parallel mode; and the node labels are refreshed through combining parallel propagation results of the multiple sub networks. Compared with a linear combination method LinearComb, the parallel label propagation-based combination method can use heterogeneous interactive information between nodes more effectively, and HLPA is more suitable for heterogeneous network community discovery.
Description
Technical field
The present invention relates to Crosslinking Structural technical field; Particularly relate to the community discovery method of heterogeneous network.
Background technology
The isomerism of Social Interaction behavior is one of distinguishing feature of mobile social networking, and it shows as on the line both existing in Virtual Space between user mutual, exists again in physical space under the line between user mutual. Specifically, the online interaction between user forms social networks on line on the one hand, forms social networks under line on the other hand under the line between user alternately. Therefore, how to realize Social Interaction and the effective integration of Social Interaction under line on line, accurately to portray the isomery interbehavior between user, become the overriding challenge of the required reply of isomery mobile social networking community discovery method research.
Label propagation algorithm (LabelPropagationAlgorithm, LPA) is that a kind of fast community in recent years proposed finds method, has linear time complexity. But, the community discovery method based on tradition LPA cannot find eclipsed form community structure. LPA has been extended by COPRA, but still there is the deficiency of three aspects: one, although COPRA is it can be found that eclipsed form community structure, but cannot be used for the community discovery towards the isomery Internet; Its two, it is comparatively harsh that the iteration of COPRA terminates decision condition, slower for its convergence rate of some network structure; Its three, COPRA can cause the appearance of " huge community ", affects the overall performance of found community structure. SLPA is the up-to-date extension to LPA, although the method avoids in a large number the appearance of " mini community ", but its overall performance is heavily dependent on iterations and degree of membership threshold value, its Key Performance Indicator (such as modularity) significantly changes along with the difference of above-mentioned parameter value, causes that availability declines.
Summary of the invention
For disadvantages described above, the present invention provides one effectively to realize on line Social Interaction effective integration under Social Interaction and line, can accurately portray the heterogeneous network community discovery method of isomery interbehavior between user.
A kind of technical scheme concrete based on the heterogeneous network community discovery method that also row label is propagated of the present invention is:
A kind of based on the heterogeneous network community discovery method that also row label is propagated, including label initial phase, mark cycle more new stage and community's construction phase; The mark cycle more new stage is based on and row label is propagated, it is allowed to node label is relatively independent ground propagated in parallel in multiple subnets of heterogeneous network, updates node label by merging the propagated in parallel result of multiple subnet.
Preferably, a kind of mark cycle more new stage based on the also heterogeneous network community discovery method that row label is propagated comprises the following steps:
Step one: label and label under line on the line of computing node;
Step 2: label and label under line on the line of aggregators;
Step 3: judge whether the node label after merging meets the end condition of label propagation stage; If so, final label is obtained; If not; Then continue from step whole tag update step at the beginning.
Preferably, a kind of based on and the end condition of label propagation stage of heterogeneous network community discovery method propagated of row label be: mark cycle updates and reaches predefined maximum iteration time, or the minimum node that the tag set that occurs in twice circulation of continuous print is identical and arbitrary label identifies is equal.
Preferably, a kind of community's construction phase based on the heterogeneous network community discovery method that also row label is propagated is divided among the community of overlap according to label two tuple of node.
Preferably, a kind of propagation also introducing decay factor constraint label based on the heterogeneous network community discovery method that also row label is propagated, to control the appearance of huge community.
Preferably, when a kind of heterogeneous network community discovery method based on also row label propagation is for weighting heterogeneous network, according to the weight of link between node, label assignment function need to be adjusted.
Preferably, when a kind of heterogeneous network community discovery method based on also row label propagation is for oriented heterogeneous network, adjacency matrix assignment need to be adjusted.
Preferably, a kind of based on and when the heterogeneous network community discovery method propagated of row label is suitable for inclusion in the heterogeneous network converged of three and above subnet with community discovery, tag fusion subfunction need to be revised.
The present invention is centered by heterogeneous network converged, it is proposed to a kind of heterogeneous network community discovery method (HybridLabelPropagationAlgorithm, HLPA) based on also row label mechanism of transmission. Wherein, and the core of row label mechanism of transmission allows for node label relatively independent ground propagated in parallel in multiple subnets of heterogeneous network, updates node label by merging the propagated in parallel result of multiple subnet. The present invention can effectively realize Social Interaction and the effective integration of Social Interaction under line on line, accurately to portray the isomery interbehavior between user.
Accompanying drawing explanation
Figure-1 network community discovery method schematic flow sheet;
Figure-2 mark cycle more new stage schematic diagram;
Figure-3 location-based social networks signal;
Figure-4 isomery Internet signal;
Figure-5 is based on the heterogeneous network converged mechanism that also row label is propagated;
Figure-6HLPA method running signal;
Figure-7 isomery Internet linear fusion result signal;
Figure-8COPRA method running signal;
HLPA method performance signal under the different degree of membership threshold value of figure-9;
HLPA method performance signal under the different tag update mechanism of figure-10;
HLPA method performance signal when figure-11 different label decay factor;
The different label transmission method performance signal of figure-12.
Detailed description of the invention
Below in conjunction with embodiment and accompanying drawing, the present invention is described in detail.
On the one hand, the present invention proposes and row label mechanism of transmission, and its core allows for node label relatively independent ground propagated in parallel in multiple subnets of heterogeneous network, updates node label by merging the propagated in parallel result of multiple subnet. Below in conjunction with example to and row label mechanism of transmission make and being expanded on further.
The particular content of this mechanism is as follows: for any one node vi, it is assumed that it is c at the label of moment ti(t),WithRepresent v respectivelyiIn network AonWith AoffIn adjacent node set, then viLabel at subsequent time t+1 is:
Wherein fmergeFor tag fusion function,WithThen represent v during moment t+1 respectivelyiIn network AonWith AoffIn label, adopt asynchronous refresh mechanism, formalization representation is:
Based on above-mentioned and row label mechanism of transmission, the present invention proposes the HLPA method towards heterogeneous network community discovery, by the seamless communication process being embedded in label of fusion of heterogeneous network, is the effective ways of large scale scale heterogeneous network overlapped formula community discovery. Specifically, for realizing eclipsed form community discovery, HLPA allows a node to have multiple label, and method comprises three below step:
1) label initial phase, each node viIt is associated with two tuple < ci,bi>, wherein ciFor community's label, biFor node viTo community ciDegree of membership, all b time initialiValue be 1;
2) the mark cycle more new stage, node label on line/line under relatively independent ground propagated in parallel in the Internet, namely at a renewal process interior joint viLabel on line will be given simultaneouslyWith label under lineWhereinBy viLine on adjacent node setCalculate according to formula (2), andThen by viLine under adjacent node setCalculate according to formula (3); Subsequently according to label on formula (1) fusion lineWith label under lineObtain this circulation interior joint viFinal label. Update mechanism is as follows:
Computing node viLine on label under label and line: withFor example, it is first v by its assignmentiThe set of wired upper adjacent node label, and the degree of membership of cumulative same label; Then label and degree of membership normalization thereof are obtained two tuple-setsMaking degree of membership sum is 1; Being simultaneously introduced degree of membership threshold ��=1/v, wherein v is the maximum community number that any node can be subordinate to, from setThe all degrees of membership of middle deletion two tuples less than ��; IfValue be smaller than threshold ��, then only retain two tuples corresponding to maximum membership degree; After deletion action, again rightIt is normalized, obtains label on lineEnd valueSimilarly, label under line is tried to achieve in calculatingEnd value
Aggregators viLine on label under label and line: will according to formula (1)WithLinear superposition normalization so that degree of membership sum is 1; Delete degree of membership less than two tuples of threshold �� normalization again, obtain node viFinal label
The end condition of label propagation stage is: mark cycle renewal reaches predefined maximum iteration time or following condition is satisfied: in twice circulation of continuous print, the tag set occurredWithIdentical and arbitrary labelThe minimum node quantity m identifiedt-1(x) and mtX () is equal, formalization representation is:
Wherein:
3) community's construction phase, is divided among the community of overlap according to label two tuple of node. Owing to being likely to occur identical community, it is necessary to carry out duplicate detection and the deletion of community.
Similar with other label transmission methods, HLPA method also can cause the appearance of " huge community ", and some reasons are in that the propagation that label is unrestricted in a network. For this, present invention introduces decay factor �� (DecayCoefficient) and retrain the propagation of label, to control the appearance of " huge community ", it is achieved the optimization of method. Label decay factor �� is defined as the real number that span is [0,1]. Corresponding tlv triple (c, b, inf) the parameter inf of each node label represents the power of influence of label c, and its value is together decided on by the propagation distance of decay factor �� Yu label. The propagation distance assuming label c is d, then have inf=(1-��)d, it is clear that decay factor �� can retrain the propagation distance of label. In the more big then communication process of decay factor, the decrease speed of label power of influence is more fast, and its effective propagation distance is more little, thus controlling the appearance of " huge community " to a certain extent.
For weighting heterogeneous network, only need to carry out label assignment function adjusting as follows: b=bj��wij, wherein wijRepresent node viWith vjBetween link weight.
For oriented heterogeneous network, only adjacency matrix assignment need to be adjusted, it may be assumed that if there is directed edge (vi,vj), then Aij=1. Obviously, generally matrix A is unsymmetrical matrix.
For plurality of subnets heterogeneous network, only need to revise tag fusion subfunction so that it is suitable in a node multiple sub-networks and the fusion of row label.
Assume to have location-based social networks as shown in Figure 3, its line contained can be expressed as two separate subnet networks as shown in Figure 4 under mutual and line alternately, for convenience, seven users in Fig. 3 are respectively labeled as A-G from left to right.
First, the label of isomery Internet interior joint is initialized as a-g, in Fig. 5 shown in ground floor; Then, the node label difference Internet on line (completes the node label within subnet based on asynchronous refresh mechanism to propagate with relatively independent ground propagated in parallel in the Internet under line herein, update sequence is A-G), in the such as Fig. 5 of the result after an iteration in the second layer shown in red/blue label; Afterwards, according to fusion function, the propagation result of different sub-network network is merged, in result such as figure shown in second layer Green label.
For the above-mentioned isomery Internet, Fig. 5 describes the once execution process of also row label mechanism of transmission. For node B, on its line adjacent node set be A, C, D}, under line, adjacent node set is { A, C}; For subnet mutual on line, process the label of node A during node B and be updated to b, so from { tri-labels of b, c, d} randomly selecting c as label on the line of B; Similarly, for subnet mutual under line, under the line of node B, label is updated to b; To the whole isomery Internet, the label obtaining node B according to fusion function is c.
As from the foregoing: and under row label mechanism of transmission, label is relatively independent ground propagated in parallel in multiple sub-networks of heterogeneous network, the new label of node depend on simultaneously on its line with the Internet under line, and the fusion of interactive information dynamically betides among the process of community discovery, rather than once complete before being typically in community discovery process such as linear fusion mechanism. Compare other heterogeneous network converged mechanism, both the information that different society type of interaction contains had been remained based on the syncretizing mechanism that also row label is propagated, achieve again the effective integration that isomery is mutual, inherit label simultaneously and propagate the high efficiency having, be a kind of syncretizing mechanism being suitable for the discovery of large scale scale heterogeneous Web Community.
On the other hand, the present invention is based on also row label mechanism of transmission, it is proposed to towards the HLPA method of heterogeneous network community discovery, by the seamless communication process being embedded in label of fusion of heterogeneous network, it is achieved towards the efficient community discovery of heterogeneous network. Below for the isomery Internet shown in Fig. 4, set forth the execution process of HLPA method and analyze the difference of itself and existing label transmission method, as shown in Figure 6.
Concrete process is as follows: first initialize the label of all nodes in heterogeneous network, and selected tag update order is A-G, as shown in Figure 6; Then arranging v=2, i.e. ��=0.5, circulation updates node label, until end condition is satisfied. For current goal network, its third time iteration result is identical with second time iteration result, after namely label communication process converges on two-wheeled circulation renewal. According to iteration result, objective network is divided into Liang Ge community, i.e. { A, B, C, D} and { D, E, F, G}.
Select COPRA method to compare with HLPA method as an example below, first the isomery Internet in Fig. 4 is carried out linear fusion for this, obtain weighted network as shown in Figure 7.
Similarly, same selected tag update order is A-G, and arranges v=2, i.e. ��=0.5; Node label is updated, until end condition is satisfied according to the circulation of COPRA tag update mechanism. Because the network merging gained is weighted network, so needing to be multiplied by the weight of corresponding sides in label communication process, it may be assumed that
COPRA method operation result is as shown in Figure 8. Objective network is restrained after four iteration, it has been found that community { A, B, C} and { D, E, F, G}. Obviously, the heterogeneous network through linear fusion lost the information about type of interaction completely, causes mutual equivalent under mutual on label communication process center line and line. So, compare linear fusion, the discovery of community can be carried out based on the heterogeneous network converged mechanism that also row label is propagated more accurately according to the feature that heterogeneous network is intrinsic.
In order to assess the effectiveness of HLPA method, the present invention carries out verification experimental verification in conjunction with multiple embodiments, specifically includes that the comparison of different tag update mechanism: the impact for HLPA method performance of checking synchronized update and asynchronous refresh mechanism; The comparison of end condition propagated by different labels: the end condition impact for HLPA method performance propagated by the different label of test; The comparison of different degree of membership threshold values: the different values of assessment degree of membership threshold value are for the impact of HLPA method performance; The comparison of different label decay factors: the different values of the checking label decay factor impact on HLPA method performance; The comparison of different label transmission methods: the quality of test COPRA, SPLA and proposed HLPA method.
Embodiment 1: the comparison of different degree of membership threshold values
For the community discovery method propagated based on label, the community number v, i.e. ��=1/v that degree of membership threshold �� major effect node can be subordinate to.
Experiment parameter is provided that the span of parameter v is [1,20], and selects asynchronous refresh mechanism. Experimental result is as shown in the figure.
According to experimental result just like drawing a conclusion:
The value of parameter v is more big, and namely the value of degree of membership threshold �� is more little, and respective community number is more few. The degree of membership threshold value that its reason is less makes the transmission capacity of label higher, is more readily formed bigger community, causes that the quantity of community declines in iterative process.
Modularity is had considerable influence by parameter v value. When v is constantly increased by 1, the value of modularity progressivelyes reach maximum (about v=5); Its value presents the trend being gradually reduced along with the increase of v afterwards.
Along with the increase of parameter v value, community's degree of overlapping presents the trend continuing to rise, and meets the physical significance of this parameter. The reason causing these results is apparent from, and namely less degree of membership threshold value 1/v makes node be easier to be identified by multiple labels simultaneously.
When parameter v is less, community's coverage rate is lower than 100%, and namely part of nodes is not belonging to any community and is in isolated state; Along with the increase of v value, community's coverage rate progressivelyes reach 100%. Reason be v less time corresponding degree of membership threshold value relatively big, constrain the propagation of label, cause that part labels only covers a node.
Embodiment 2: the comparison of different tag update mechanism
Experiment parameter is provided that the span of parameter v is [1,20], uses two kinds of tag update mechanism of synchronous versus asynchronous to test respectively. Experimental result is as shown in Figure 10.
According to experimental result just like drawing a conclusion:
Under two kinds of tag update mechanism of synchronous versus asynchronous, HLPA without significant difference, only has nuance when the value of parameter v is less in community's quantity with two performance indications of community's coverage rate.
Under two kinds of tag update mechanism of synchronous versus asynchronous, there were significant differences in modularity with two performance indications of community's degree of overlapping for HLPA. When the value of parameter v is less, the performance of two kinds of tag update mechanism is closer to; Along with the increase of v, the difference of the two performance is gradually increased. On the one hand, modularity corresponding to synchronized update mechanism presents the trend of rapid decrease after reaching a maximum value, with the performance gap rapid expansion of asynchronous refresh mechanism; On the other hand, gathering way of community's degree of overlapping that synchronized update mechanism is corresponding is significantly faster than that asynchronous refresh mechanism. Such as, when v value is 20, the value of the two modularity respectively 0.67 and 0.76, the value of community's degree of overlapping then respectively 2.06 and 1.29.
Comprehensive above experimental result is it can be seen that compare synchronized update mechanism, and the performance of asynchronous refresh mechanism is more stable, therefore will mainly adopt asynchronous tag update mechanism in subsequent experimental.
Embodiment 3: the comparison of different label decay factors
It is (0,0.10,0.15,0.20,0.25,0.30,0.35,0.40) that experiment parameter is provided that v takes the span of 3 and 5, �� respectively, adopts asynchronous tag update mechanism. Relevant experimental result is as shown in the figure.
According to experimental result just like drawing a conclusion:
Along with the increase of label decay factor ��, the maximum community that HLPA finds presents the trend being gradually reduced. When �� is increased to about 0.25 by 0.00, maximum community reduces continuously and healthily; And after the value of �� exceedes certain threshold value (about 0.25), the maximum community kept stable found. This conclusion keeps when the value of parameter v changes setting up.
Along with the increase of label decay factor ��, the modularity that HLPA reaches presents the trend being gradually reduced equally. Specifically, when �� is increased to about 0.20 by 0.00, the reduction speed of respective mode lumpiness is slower; And after the value of �� exceedes certain threshold value (about 0.20), the modularity reached then quickly reduces. This conclusion keeps when the value of parameter v changes setting up equally.
Comprehensive conclusions is it can be seen that label decay factor can effectively control the appearance of " huge community ", but it can bring certain negative effect equally, for instance cause that modularity declines. Accordingly, it would be desirable to combining target data set and practical application request, it is determined by experiment the label decay factor of optimum.
Embodiment 4: the comparison of different label transmission methods
In order to compare different label transmission methods, experimental design is as follows: process isomery mobile social networking based on linear fusion method LinearComb, run two kinds of label transmission methods of COPRA and SLPA above, and the result of corresponding experimental result Yu HLPA is compared.
Experiment parameter is provided that on linear fusion method center line and the value of the weight factor �� of type of interaction under line is 0.5, in COPRA and HLPA, the span of parameter v is [1,20], in SLPA, the span of threshold value r is [0.05,0.5], asynchronous tag update mechanism is adopted to perform 100 iteration. Relevant experimental result is as shown in the figure. For ensureing comparability and the intuitive of experimental result, community's quantity that in figure, when abscissa selection different parameters, each method finds.
According to experimental result just like drawing a conclusion:
When the value of parameter v changes, the modularity fluctuation of COPRA is relatively big, and its maximum 0.8246 obtains when v is 4, and community's number of correspondence is 675; Similarly, the modularity of HLPA presents more significant fluctuation equally along with the change of parameter v value, and its maximum 0.8311 obtains when v is 6, and corresponding community's number is 928; Being different from above-mentioned two kinds of methods, during the change of threshold value r value, the modularity of SLPA is held essentially constant.
When being equally based on linear fusion method LinearComb, modularity produced by COPRA is higher than SLPA, and community's degree of overlapping of the latter is then higher. Reason is in that SLPA have recorded the historical information in label communication process so that the coverage of label increases, and declining occurs in modularity. It follows that use communication process record might not bring performance boost in label transmission method, garbage may be introduced on the contrary so that declining occurs in performance. Additionally, compare COPRA and SLPA, HLPA to be obtained in that higher modularity, and community's degree of overlapping produced by it is close with COPRA.
Increase along with community's number, on the one hand, community's degree of overlapping value of COPRA and HLPA presents the trend being gradually reduced, and the value of SLPA then presents the trend of rising; On the other hand, the value of the maximum community of COPRA and HLPA constantly reduces, and the analog value of SLPA then slowly increases. Above-mentioned observation never ipsilateral reflects the different qualities of COPRA/HLPA and SLPA, and its reason is then in that the difference with update mechanism propagated by label.
In summary, comparing linear fusion method LinearComb, can more effectively utilize internodal isomery interactive information based on the fusion method that also row label is propagated, HLPA is also more suitable for heterogeneous network community discovery.
Claims (8)
1. the heterogeneous network community discovery method based on also row label propagation, it is characterised in that: described method includes label initial phase, mark cycle more new stage and community's construction phase; The described mark cycle more new stage is based on and row label is propagated, it is allowed to node label is relatively independent ground propagated in parallel in multiple subnets of heterogeneous network, updates node label by merging the propagated in parallel result of multiple subnet.
2. according to claim 1 based on the heterogeneous network community discovery method that also row label is propagated, it is characterised in that: the described mark cycle more new stage comprises the following steps:
Step one: label and label under line on the line of computing node;
Step 2: label and label under line on the line of aggregators;
Step 3: judge whether the node label after merging meets the end condition of label propagation stage; If so, final label is obtained; If not; Then continue from step whole tag update step at the beginning.
3. according to the arbitrary described heterogeneous network community discovery method based on also row label propagation of claim 2, it is characterized in that: the end condition of described label propagation stage is: mark cycle updates and reaches predefined maximum iteration time, or the minimum node that the tag set that occurs in twice circulation of continuous print is identical and arbitrary label identifies is equal.
4. according to claim 1 based on the heterogeneous network community discovery method that also row label is propagated, it is characterised in that: described community construction phase is divided among the community of overlap according to label two tuple of node.
5. according to claim 1 based on the heterogeneous network community discovery method that also row label is propagated, it is characterised in that: described method also introduces the propagation of decay factor constraint label, to control the appearance of huge community.
6. according to claim 1 based on the heterogeneous network community discovery method that also row label is propagated, it is characterised in that: when described method is for weighting heterogeneous network, according to the weight of link between node, label assignment function need to be adjusted.
7. according to claim 1 based on the heterogeneous network community discovery method that also row label is propagated, it is characterised in that: when described method is for oriented heterogeneous network, adjacency matrix assignment need to be adjusted.
8. according to claim 1 based on and the heterogeneous network community discovery method propagated of row label, it is characterised in that: when described method is suitable for inclusion in the heterogeneous network converged of three and above subnet with community discovery, tag fusion subfunction need to be revised.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510968384.0A CN105631748A (en) | 2015-12-21 | 2015-12-21 | Parallel label propagation-based heterogeneous network community discovery method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510968384.0A CN105631748A (en) | 2015-12-21 | 2015-12-21 | Parallel label propagation-based heterogeneous network community discovery method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105631748A true CN105631748A (en) | 2016-06-01 |
Family
ID=56046639
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510968384.0A Pending CN105631748A (en) | 2015-12-21 | 2015-12-21 | Parallel label propagation-based heterogeneous network community discovery method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105631748A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106295702A (en) * | 2016-08-15 | 2017-01-04 | 西北工业大学 | A kind of social platform user classification method analyzed based on individual affective behavior |
CN108446862A (en) * | 2018-03-29 | 2018-08-24 | 山东科技大学 | The three-stage policy algorithm of overlapping community detection in a kind of community network |
CN108898264A (en) * | 2018-04-26 | 2018-11-27 | 深圳大学 | A kind of calculation method and device being overlapped community's set quality Measure Indexes |
CN110321492A (en) * | 2019-06-13 | 2019-10-11 | 华中科技大学 | A kind of item recommendation method and system based on community information |
CN111597396A (en) * | 2020-05-13 | 2020-08-28 | 深圳计算科学研究院 | Heterogeneous network community detection method and device, computer equipment and storage medium |
CN112052404A (en) * | 2020-09-23 | 2020-12-08 | 西安交通大学 | Group discovery method, system, device and medium for multi-source heterogeneous relation network |
CN113053540A (en) * | 2021-04-01 | 2021-06-29 | 电子科技大学 | Community discovery method for traditional Chinese medicine core medicine identification |
-
2015
- 2015-12-21 CN CN201510968384.0A patent/CN105631748A/en active Pending
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106295702A (en) * | 2016-08-15 | 2017-01-04 | 西北工业大学 | A kind of social platform user classification method analyzed based on individual affective behavior |
CN106295702B (en) * | 2016-08-15 | 2019-10-25 | 西北工业大学 | A kind of social platform user classification method based on the analysis of individual affective behavior |
CN108446862A (en) * | 2018-03-29 | 2018-08-24 | 山东科技大学 | The three-stage policy algorithm of overlapping community detection in a kind of community network |
CN108898264A (en) * | 2018-04-26 | 2018-11-27 | 深圳大学 | A kind of calculation method and device being overlapped community's set quality Measure Indexes |
CN108898264B (en) * | 2018-04-26 | 2021-10-29 | 深圳大学 | Method and device for calculating quality metric index of overlapping community set |
CN110321492A (en) * | 2019-06-13 | 2019-10-11 | 华中科技大学 | A kind of item recommendation method and system based on community information |
CN111597396A (en) * | 2020-05-13 | 2020-08-28 | 深圳计算科学研究院 | Heterogeneous network community detection method and device, computer equipment and storage medium |
CN111597396B (en) * | 2020-05-13 | 2021-05-28 | 深圳计算科学研究院 | Heterogeneous network community detection method and device, computer equipment and storage medium |
CN112052404A (en) * | 2020-09-23 | 2020-12-08 | 西安交通大学 | Group discovery method, system, device and medium for multi-source heterogeneous relation network |
CN112052404B (en) * | 2020-09-23 | 2023-08-15 | 西安交通大学 | Group discovery method, system, equipment and medium of multi-source heterogeneous relation network |
CN113053540A (en) * | 2021-04-01 | 2021-06-29 | 电子科技大学 | Community discovery method for traditional Chinese medicine core medicine identification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105631748A (en) | Parallel label propagation-based heterogeneous network community discovery method | |
Papadopoulos et al. | Network mapping by replaying hyperbolic growth | |
CN104836711B (en) | A kind of command and control network generates the construction method of model | |
CN110674869A (en) | Classification processing and graph convolution neural network model training method and device | |
Assadian et al. | GCPSO in cooperation with graph theory to distribution network reconfiguration for energy saving | |
CN105050095B (en) | A kind of topological construction method of the heterogeneous wireless sensor net based on energy predicting | |
CN109451459B (en) | Mobile fog node-based sensing cloud bottom node trust evaluation method | |
CN108600022B (en) | Dynamic network layout accelerating method | |
CN113573333B (en) | Particle swarm heterogeneous WSNs coverage optimization algorithm based on virtual force | |
CN108289285A (en) | A kind of ocean wireless sensor network is lost data and is restored and reconstructing method | |
Tang et al. | An adaptive discrete particle swarm optimization for influence maximization based on network community structure | |
Darwish et al. | An adaptive cellular automata scheme for diagnosis of fault tolerance and connectivity preserving in wireless sensor networks | |
CN106953801B (en) | Random shortest path realization method based on hierarchical learning automaton | |
CN106127595A (en) | A kind of community structure detection method based on positive and negative side information | |
Hsiao et al. | Static and dynamic minimum energy broadcast problem in wireless ad-hoc networks: A PSO-based approach and analysis | |
CN103701647A (en) | Wireless network topology optimizing and generating method | |
CN104680263B (en) | Electric power transportation network Topology Structure Design method based on particle cluster algorithm | |
Hollocou et al. | Improving PageRank for local community detection | |
CN115396322A (en) | Command control hyper-network modeling method based on layer betweenness edge-connecting strategy | |
Zhang et al. | An adaptive honeypot deployment algorithm based on learning automata | |
CN105682170A (en) | Clustering-type wireless sensor network data collection method based on energy consumption balance | |
CN103281697B (en) | A kind of wireless sensor network topology reconstructing method of center type | |
Umar et al. | Designing of dynamic re-clustering leach protocol for calculating total residual time and performance | |
Hasson et al. | Developed clustering approaches to enhance the data transmissions in WSNs | |
CN110086654B (en) | Master and slave deployment method based on reliability under software defined data center |
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: 20160601 |
|
RJ01 | Rejection of invention patent application after publication |