CN105825430A - Heterogeneous social network-based detection method - Google Patents
Heterogeneous social network-based detection method Download PDFInfo
- Publication number
- CN105825430A CN105825430A CN201610011812.5A CN201610011812A CN105825430A CN 105825430 A CN105825430 A CN 105825430A CN 201610011812 A CN201610011812 A CN 201610011812A CN 105825430 A CN105825430 A CN 105825430A
- Authority
- CN
- China
- Prior art keywords
- relation
- sigma
- node
- centerdot
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Abstract
The invention relates to the field of information science, and provides a heterogeneous social network-based detection method. According to the method, firstly, a heterogeneous social network is mapped into a multi-dimensional matrix, and both the node transition probability and the relationship transition probability of the multi-dimensional matrix are determined. Secondly, based on the random walk algorithm, the equilibrium distribution of nodes and the equilibrium distribution of relationships are realized. Thirdly, based on the equilibrium distribution of nodes and the equilibrium distribution of relationships, a weighted single-relationship social network is obtained. Finally, based on the single-relationship social network detection algorithm, a detection result of the heterogeneous social network is obtained based on the weighted single-relationship social network. According to the technical scheme of the invention, the mutual influence of nodes and relationships in the heterogeneous social network is fully utilized, and the heterogeneous social network is fused into the weighted single-relationship social network. After that, the community detection on the weighted single-relationship social network is conducted based on the community detection method for conventional single-relationship social networks.
Description
Technical field
The present invention relates to information science field, it is provided that a kind of detection method based on isomery community network.
Background technology
This part is intended to introduce the technology of the various aspects of this area that may be relevant with the various aspects of the application to reader, it is believed that this part contributes to reader with background's information, in order to be more fully understood that the various aspects of the application.It will thus be appreciated that should carry out understanding rather than being regarded as being admission of prior art from this angle.
Along with Internet and the fast development of WWW, the research of the community network of Web community and sing on web community is gradually risen, therefore the method for the community structure in searching community network is own through becoming the study hotspot in social network analysis, and there is also the biggest business opportunity, present society analysis of network is applied in a lot of fields.Such as the analysis of public opinion, opinion leader excavation, subject focus, advertisement putting, citation analysis, scientific research cooperative, attack of terrorism analysis, the excavation of crime core, information management, Web Link Analysis, social networks etc..Social network analysis is own through becoming one of important branch of data mining subject, develops especially swift and violent, because social network analysis is to be closely related with actual life and be with a wide range of applications in recent years.From traditional member relation network to Internet era social networks, from large-scale power network to transportation network, business model from real business model to virtual platform, from the cooperative network of researcher to various politics, education, economy, medical treatment, the social relation network etc. of science and technology, it may be said that, various complex networks are flooded with our life, these networks all have the feature of community network, by the analytic learning to these community networks, it is possible to obtain the potential useful information that we need.
Webpage and its chain enter the relation that chain goes out and generally model according to the mode of figure, for example it is known that HITS and PageRank algorithm be used to calculate the authority value of each node, in single relational network, both algorithms can be used as order models.
But when in the face of isomery community network, we are accomplished by the authority value on associating ordering joint and limit.The algorithm of most of social network analysis is only it is considered that the community network of homogeneity relation, i.e. single relational network, such as only exist the relation linked between webpage with webpage, and the community network major part in real world is presented in isomery community network, various relation it is constantly present between most of entities, these relations embody different importances in varied situations, relation is all counted as a kind of single relational network each of in these relations simultaneously, therefore along with the further investigation of community's detection, isomery social network analysis is by increasing focus of attention.At present, existing research generally believes that relations different in heterogeneous network is separate, fair play, and actually this is irrational in actual life.
Summary of the invention
In order to overcome the deficiencies in the prior art, the exemplary embodiment of the present invention takes full advantage of influencing each other between the node in isomery community network and relation, propose a kind of detection method based on isomery community network, this detection method can draw the weight of heterogeneous network interior joint and relation by iterative computation, and heterogeneous network converged can be become the single relational network with weight, then, the community detection method recycling traditional single relational network carries out community's detection to the single relational network with weight merged through the present invention.
According to an aspect of the present invention, it is provided that a kind of detection method based on isomery community network, including:
Isomery community network is mapped to multi-dimensional matrix;
Determine transition probability and the transition probability of relation of described multi-dimensional matrix interior joint;
Utilize Random Walk Algorithm, it is thus achieved that the equiblibrium mass distribution of node and the equiblibrium mass distribution of relation;
Equiblibrium mass distribution according to node and the equiblibrium mass distribution of relation, it is thus achieved that single tie society network of Weight;And
Utilize single tie society Web Community detection algorithm, single tie society network based on described Weight and obtain community's testing result of isomery community network.
In the exemplary embodiment, described multi-dimensional matrix is the matrix of n × n × m, and the value in multi-dimensional matrix represents node i and the weight of node j synthesis under the d relation, and wherein 1≤i, j≤n, 1≤d≤m, m and n is the positive integer more than or equal to 2.
In the exemplary embodiment, determine that the described transition probability of multi-dimensional matrix interior joint and the transition probability of relation include:
Tensor S=[the s that definition is three-dimensionali,j,d], it represents relation and the synthesis of node;And
The transition probability determining node is O=[oi,j,d] and the transition probability of relation be R=[ri,j,d], wherein
In the exemplary embodiment, Random Walk Algorithm is utilized, it is thus achieved that according to the equiblibrium mass distribution of node and the equiblibrium mass distribution of relation
Determine equiblibrium mass distribution and the equiblibrium mass distribution of described relation of described node, whereinWithBeing the prior distribution of node and relation respectively, α and β is Dynamic gene.
In the exemplary embodiment, the weight of described synthesis is the sum of products of node i and node j tensor under d relation and relation weight.
In the exemplary embodiment, described single tie society Web Community detection algorithm includes at least one in Kmeans algorithm, GMM algorithm and GMM-NK algorithm.
More specifically, the exemplary embodiment of the present invention provides a kind of associating sort algorithm, it takes full advantage of influencing each other between the node in isomery community network and relation, heterogeneous network converged is become the single relational network with weight, then, the community detection method recycling traditional single relational network carries out community's detection to the single relational network with weight merged through the present invention.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, the accompanying drawing used required in embodiment or description of the prior art will be briefly described below, apparently, accompanying drawing in describing below is only some embodiments of the present invention, for those of ordinary skill in the art, on the premise of not paying creative work, it is also possible to obtain other accompanying drawing according to these accompanying drawings, wherein
Fig. 1 is the transition probability R of the transition probability O of node and relation coordinate diagram in hyperspace;
Fig. 2 is conditional probability coordinate diagram in hyperspace;
Fig. 3 combines the false code of sort algorithm;
Fig. 4 is the coordinate diagram closely spending matrix of 4 relations on Iris data set;
Fig. 5 is the iterative computation by combining sort algorithm in invention, the block diagram of the equiblibrium mass distribution of Iris and the Breast data set co-relation finally given;
Fig. 6 is on Iris data set, and the synthesis network obtained by associating sort algorithm iteration is carried out community's detection and singly closes, with each, the block diagram that the network fastened carries out the performance comparison of community's detection;And
Fig. 7 is the curve chart of the convergence situation combining sort algorithm on the generated data collection of Iris.
Detailed description of the invention
For making the purpose of the embodiment of the present invention, technical scheme and advantage clearer, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is carried out clear, complete description, obviously, described embodiment is only a part of embodiment of the present invention rather than whole embodiments.Based on embodiments of the invention, the every other embodiment that those of ordinary skill in the art are obtained under not making creative work premise, broadly fall into the scope of protection of the invention.
Additionally, in describing the invention, except as otherwise noted, " multiple " are meant that two or more.
Exemplary embodiment according to the present invention, it is provided that a kind of detection method based on isomery community network, mainly comprises the steps that
Isomery community network is mapped to multi-dimensional matrix;
Determine transition probability and the transition probability of relation of described multi-dimensional matrix interior joint;
Utilize Random Walk Algorithm, it is thus achieved that the equiblibrium mass distribution of node and the equiblibrium mass distribution of relation;
Equiblibrium mass distribution according to node and the equiblibrium mass distribution of relation, it is thus achieved that single tie society network of Weight;And
Utilize single tie society Web Community detection algorithm, single tie society network based on described Weight and obtain community's testing result of isomery community network.
Below with reference to accompanying drawing, above-mentioned steps is illustrated one by one.
Isomery community network is mapped to multi-dimensional matrix:
The community network of m relation is typically defined as form (V, the E of figure group(d)), d=1,2 ..., m, wherein v represents the node set containing n element.E(d)It it is the adjacency matrix fastening Undirected networks d pass.E(d)An actually binary matrix, if having a limit, then between node i and jOtherwise
One isomery community network can be represented by a cubical tensor form of n × n × m.If node i is connected on node j by the d article relation, then (i, j d) are non-zero to the entry in tensor.Definition R is an entity, S=(si,j,d) it is a substantial connection tensor, si,j,d∈ R represents node i and node j relation weight under the d relation.And the value in multi-dimensional matrix represents node i and the weight of node j synthesis under the d relation, wherein 1≤i, j≤n, 1≤d≤m, m and n is the positive integer more than or equal to 2, the weight of described synthesis is the sum of products of node i and node j tensor under d relation and relation weight, specifically can be calculated by below equation:
Determine transition probability and the transition probability of relation of described multi-dimensional matrix interior joint:
Usually, the tensor S=[s in three directions is definedi,j,d] (1≤i, j≤n, 1≤d≤m) representation relation and the synthesis of entity.For the community network of synthesis, S=Α is a set with m relation, and S contains m matrix, is equivalent to the optimization object function that all kinds are mutual.For feature synthesizes, S represents the feature synthesis with point, and the detection method generally by overlapping community carries out each Relation extraction.If we define furtherWithBeing respectively many relational networks interior joint and the weight of relation, the weight of synthesis can be defined as:
The mission critical of iteration is exactly the weight vectors of calculated relationshipFor many relational networks, it is interactional between node and relation, it would be desirable to obtaining the sequence of combining of a node and relation, definition R is a real world, two vectorsWith
We assume that random walk be applied in many relational networks, two transition probability tensor O=[o so can be constructedi,j,d] and R=[ri,j,d], represent the transition probability of node and the transition probability of relation respectively.By standardizing substantial connection tensor S, the transition probability of O and R is as follows:
Fig. 1 is from the transition probability R of the transition probability O and relation spatially illustrating node based on A, specifically, oi,j,dBe fasten the d pass, the i-th row carries out normalization process, r to substantial connection tensor S horizontal directioni,j,dBe i-th with longitudinally carry out normalization process to substantial connection tensor S is vertical on j node.Definition XtAnd YtBeing the stochastic variable accessing any one node and any one relation in t respectively, therefore, we can obtain:
oi,j,d=p (Xt=i | Xt-1=j, Yt=d) (6)
oi,j,d=p (Yt=d | Xt=i, Xt-1=j) (7)
It is apparent that the order (X of variable immediatelyt,Yt: t=0,1...) it is a Markov Chain, the algorithm combining sequence can calculate the transition probability O of node and the transition probability R of relation.
Utilize Random Walk Algorithm, it is thus achieved that the equiblibrium mass distribution of node and the equiblibrium mass distribution of relation:
WithBeing the equilibrium of node and relation or stable probability distribution respectively, if the random walk model of PageRank is applied in heterogeneous network by we, when t is infinitely-great time, p and q can reach equilibrium.Therefore, it can obtain:
By upper we can analyze, calculate Prob [Xt-1=j, Yt=d] and Prob [Xt=i, Xt-1=j] it is the equiblibrium mass distribution determining nodeEquiblibrium mass distribution with relationCommitted step.
Equiblibrium mass distribution according to node and the equiblibrium mass distribution of relation, it is thus achieved that single tie society network of Weight:
In heterogeneous network, different nodes and different relations all show its importance, are described in detail below the most how associating sort algorithm obtains the probability distribution of node and relation, then show its unique probability distribution existed.
The present invention proposes use conditional probability to carry out the modeling of joint probability distribution.By two joint probability Prob [Xt-1=j, Yt=d] and Prob [Xt=i, Xt-1=j] deform, can be represented by the formula form of conditional probability, can obtain:
Prob[Xt-1=j, Yt=d]=Prob [Xt-1=j] Prob [Yt=d | Xt-1=j] (11)
Prob[Xt=i, Xt-1=j]=Prob [Xt-1=j] Prob [Xt=i | Xt-1=j] (12).
Fig. 2 illustrates the calculating in hyperspace of the both the above conditional probability:
Formula (13) illustrates the transition probability from node j to node i, but and Prob unlike PageRankn(j | i) is not always maintained at constant, and it is affected by relation weight.In heterogeneous network, the weight of node is not merely affected by its neighboring node, and the relationship strength impact of different weight is also awarded.A given node j, selects node i conditional probability to represent as follows from node j:
Additionally, invention demonstrates a method given node j to select the conditional probability of d relation:
Therefore, using conditional probability, formula (11) and formula (12) can be with the forms being written as:
The iterative computation of associating sort algorithm is completed by formula (15) and formula (16), but, and random walk model is similarly, associating sort algorithm can also be represented by simple and clear vector matrix form.In view of this, we establish the matrix of two auxiliaryWithJ=1 ..., the dimension of n.V and U vector is all n × m, and it is respectively by the vector of m × 1 dimensionVector with n × 1 dimensionComposition.WithIt is defined as follows:
If V and U vector is carried out row normalization operation by us, just there is an equation below:
Prob[Xt-1=j | Yt=d]=vj,d,Prob[Xt=i | Xt-1=j]=uj,i(18)
Additionally, associating sort algorithm have also contemplated that the impact that the prior probability of node and relation produces, in conjunction with above formula, we use following iterative formula to calculate the ranking value of node and relation simultaneously:
HereWithBeing the prior distribution of node and relation respectively, α and β is used to balance network structure and the Dynamic gene of priori.In ideal conditions, prior distribution is the importance being calculated node and relation by this domain expert.Now, it is assumed that random walk rests on node i (that is, X at presentt=i), we can pass through formula (19) can calculate the probability selecting node j, can be calculated the probability of choice relation d by formula (20).Pass through ptAnd qtIterative computation pt+1And qt+1, finally we obtain the equiblibrium mass distribution of node and relation, and the false code of associating sort algorithm is as shown in Figure 3.
Utilize single tie society Web Community detection algorithm, single tie society network based on described Weight and obtain community's testing result of isomery community network:
Carry out community detection to merging with single relational matrix of weight with tradition list relation community detection method, such as, use Kmeans, GMM and GMM-NK algorithm to carry out community's detection, obtain the community divided.
In sum, according to the exemplary embodiment of the present invention, first implementation process builds heterogeneous network;Secondly the similarity matrix fastened in each pass according to the property calculation of heterogeneous network interior joint;Reuse associating sort algorithm to be iterated calculating, obtain merging the single relation similarity matrix with weight;Finally carry out community's detection with the single relation community detection method of tradition to merging the single relational matrix with weight.
A kind of based on isomery community network the detection method that exemplary embodiment according to the present invention provides, it may include below step:
Step 1: first build heterogeneous network in implementation process;Secondly the similarity matrix fastened in each pass according to the property calculation of heterogeneous network interior joint.
Definition Pi,dAnd pj,dIt is respectively node i, the value of j in relation d.By distance spatially for the degree matrix in close relations in constructive formula (21):
Step 2: input the degree matrix in close relations that multiple pass is fastened, uses associating sort algorithm to be iterated calculating, obtains merging the single relation similarity matrix with weight, can obtain the weighted value of heterogeneous network interior joint and relation simultaneously.Such as explanation without exception, hereinafter α=β=0.5.
Step 3: carry out community's detection with the single relation community detection method of tradition to merging the single relational matrix with weight, such as, use Kmeans, GMM and GMM-NK algorithm to carry out community's detection, obtain the community divided.
In order to verify the effectiveness of the detection method based on isomery community network according to the inventive method exemplary embodiment, choosing Iris data set to test as the data set of synthesis network, it is significant that final experimental data indicates the associating effect that detects for isomery community network of sort algorithm that the present invention proposes.The attribute of Iris data set is as shown in table 1:
Table 1.Iris data set
Because the community relations of a priori (namely GroundTruth) is known, then we use conventional normalized mutual information (NMI) as evaluation criterion.NMI is defined as follows:
In order to verify the effectiveness in isomery community network of associating sort algorithm, we calculate at P respectively with formula (21)i,dAnd pj,dIn the similarity that pass is fastened, thus constituting a matrix Tendor_d of degree closely in relation d, Tendor_d matrix is a symmetrical matrix, and on matrix, each element represents P respectivelyi,dAnd pj,dThe value of the degree closely in relation d.We represent P with lightnessi,dAnd pj,dTightness degree, the brightest compactness represented between node is the highest, therefore the figure constructed is exactly probably potential community in bright square region, node in these bright square regions defines the community combined closely in relation d, and the most each different bright square region reflects the relation d significance level to respective community.
Fig. 4 respectively show the similarity matrix of 4 relations on Iris data set.
As shown in Figure 4, for Iris data set, relation 3 and relation 4 contain the most obvious bright square region, therefore relation 3 and relation 4 detect for the community of Iris data set and have more meaning, accordingly, relation 3 and relation 4 have bigger weight compared to relation 1 and relation 2 at Iris data set.All in all, if Iris data set to be done community's detection, the relation 3 of Iris data set and relation 4 by us, importing Iris below, application associating sort algorithm calculates the equiblibrium mass distribution about relation respectively.
The relation equivalence obtained by associating sort algorithm iterative computation is distributed as it is shown in figure 5, substantially matched with the matrix of degree closely of Fig. 4 reflection by analyzing the relation equivalence distribution understood by combining the Iris that sort algorithm iterative computation obtains.
Fig. 6 illustrates the contrast situation of the NMI value calculated on the heterogeneous network of the two data set, obviously, no matter we use any clustering algorithm, and the performance shown on the network through the synthesis of associating sort algorithm is considerably beyond the performance on single relational network.Experiment shows, heterogeneous network converged can be become high-quality single relational network by associating sort algorithm, and the community structure single relational network more original than any one in single relational network of this synthesis is the most a lot.And, understanding through chart data analysis, the performance that GMM-NK algorithm shows is above GMM's and Kmeans.
Fig. 7 illustrates associating sort algorithm convergence on generated data collection.With reference to Fig. 7, we are it will be clear that the change of equiblibrium mass distribution of node and relation, | | pt-pt-1||2+||qt-qt-1||2Value in limited iterations rapid underground fall, be finally gradually reduced through iteration continuously, when ε=10-5Being one and restrain sufficient standard value, on Iris and Breast generated data collection, the iterations of continuous sort algorithm is 11 to 17 to take turns (less than 20) respectively.
The detection method based on isomery community network that the present invention proposes all influences each other in view of the distribution of node and relation and couples, so using conditional probability to carry out the modeling of joint probability distribution, thus obtain the equiblibrium mass distribution of heterogeneous network relation, the single relational network with weight can be obtained, it is also possible to obtain the weight of heterogeneous network interior joint and relation by iterative computation.
The above, the only detailed description of the invention of the present invention, but; protection scope of the present invention is not limited to this; any those familiar with the art in the technical scope that the invention discloses, the change that can readily occur in or replacement, all should contain within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with described scope of the claims.
Claims (6)
1. a detection method based on isomery community network, including:
Isomery community network is mapped to multi-dimensional matrix;
Determine transition probability and the transition probability of relation of described multi-dimensional matrix interior joint;
Utilize Random Walk Algorithm, it is thus achieved that the equiblibrium mass distribution of node and the equiblibrium mass distribution of relation;
Equiblibrium mass distribution according to node and the equiblibrium mass distribution of relation, it is thus achieved that single tie society network of Weight;And
Utilize single tie society Web Community detection algorithm, single tie society network based on described Weight and obtain community's testing result of isomery community network.
Detection method the most according to claim 1, wherein said multi-dimensional matrix is the matrix of n × n × m, and the value in multi-dimensional matrix represents node i and the weight of node j synthesis under the d relation, wherein 1≤i, j≤n, 1≤d≤m, m and n is the positive integer more than or equal to 2.
Detection method the most according to claim 2, wherein determines that the described transition probability of multi-dimensional matrix interior joint and the transition probability of relation include:
Tensor S=[the s that definition is three-dimensionali,j,d], it represents relation and the synthesis of node;And
The transition probability determining node is O=[oi,j,d] and the transition probability of relation be R=[ri,j,d], wherein
Detection method the most according to claim 3, wherein utilizes Random Walk Algorithm, it is thus achieved that according to the equiblibrium mass distribution of node and the equiblibrium mass distribution of relation
Determine equiblibrium mass distribution and the equiblibrium mass distribution of described relation of described node, whereinWithBeing the prior distribution of node and relation respectively, α and β is Dynamic gene.
Detection method the most according to claim 2, the weight of wherein said synthesis is the sum of products of node i and node j tensor under d relation and relation weight.
6., according to the arbitrary described detection method of claim 1-5, wherein said single tie society Web Community detection algorithm includes at least one in Kmeans algorithm, GMM algorithm and GMM-NK algorithm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610011812.5A CN105825430A (en) | 2016-01-08 | 2016-01-08 | Heterogeneous social network-based detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610011812.5A CN105825430A (en) | 2016-01-08 | 2016-01-08 | Heterogeneous social network-based detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105825430A true CN105825430A (en) | 2016-08-03 |
Family
ID=56987350
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610011812.5A Pending CN105825430A (en) | 2016-01-08 | 2016-01-08 | Heterogeneous social network-based detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105825430A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109002683A (en) * | 2018-06-12 | 2018-12-14 | 淮海工学院 | Community network finds algorithm checking system and method |
CN109492132A (en) * | 2018-10-26 | 2019-03-19 | 广州市香港科大霍英东研究院 | Method, system, terminal and the storage medium of Heterogeneous Information internet startup disk |
CN110519368A (en) * | 2019-08-27 | 2019-11-29 | 腾讯科技(深圳)有限公司 | A kind of method and relevant device obtaining communication network interior joint technorati authority |
CN111047453A (en) * | 2019-12-04 | 2020-04-21 | 兰州交通大学 | Detection method and device for decomposing large-scale social network community based on high-order tensor |
CN111104722A (en) * | 2018-10-10 | 2020-05-05 | 华北电力大学(保定) | Electric power communication network modeling method considering overlapping communities |
CN111931485A (en) * | 2020-08-12 | 2020-11-13 | 北京建筑大学 | Multi-mode heterogeneous associated entity identification method based on cross-network representation learning |
CN112256801A (en) * | 2020-10-10 | 2021-01-22 | 深圳力维智联技术有限公司 | Method, system and storage medium for extracting key entities in entity relationship graph |
-
2016
- 2016-01-08 CN CN201610011812.5A patent/CN105825430A/en active Pending
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109002683A (en) * | 2018-06-12 | 2018-12-14 | 淮海工学院 | Community network finds algorithm checking system and method |
CN109002683B (en) * | 2018-06-12 | 2021-07-20 | 淮海工学院 | Social network discovery algorithm verification system and method |
CN111104722A (en) * | 2018-10-10 | 2020-05-05 | 华北电力大学(保定) | Electric power communication network modeling method considering overlapping communities |
CN109492132A (en) * | 2018-10-26 | 2019-03-19 | 广州市香港科大霍英东研究院 | Method, system, terminal and the storage medium of Heterogeneous Information internet startup disk |
CN110519368A (en) * | 2019-08-27 | 2019-11-29 | 腾讯科技(深圳)有限公司 | A kind of method and relevant device obtaining communication network interior joint technorati authority |
CN110519368B (en) * | 2019-08-27 | 2021-09-07 | 腾讯科技(深圳)有限公司 | Method for obtaining authority degree of nodes in propagation network and related equipment |
CN111047453A (en) * | 2019-12-04 | 2020-04-21 | 兰州交通大学 | Detection method and device for decomposing large-scale social network community based on high-order tensor |
CN111931485A (en) * | 2020-08-12 | 2020-11-13 | 北京建筑大学 | Multi-mode heterogeneous associated entity identification method based on cross-network representation learning |
CN112256801A (en) * | 2020-10-10 | 2021-01-22 | 深圳力维智联技术有限公司 | Method, system and storage medium for extracting key entities in entity relationship graph |
CN112256801B (en) * | 2020-10-10 | 2024-04-09 | 深圳力维智联技术有限公司 | Method, system and storage medium for extracting key entity in entity relation diagram |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105825430A (en) | Heterogeneous social network-based detection method | |
Ren et al. | Pythagorean fuzzy TODIM approach to multi-criteria decision making | |
Samoilenko et al. | Using Data Envelopment Analysis (DEA) for monitoring efficiency-based performance of productivity-driven organizations: Design and implementation of a decision support system | |
Wang et al. | Effective lossless condensed representation and discovery of spatial co-location patterns | |
Kizielewicz et al. | Comparison of Fuzzy TOPSIS, Fuzzy VIKOR, Fuzzy WASPAS and Fuzzy MMOORA methods in the housing selection problem | |
CN111932386B (en) | User account determining method and device, information pushing method and device, and electronic equipment | |
CN104731962A (en) | Method and system for friend recommendation based on similar associations in social network | |
CN108304380A (en) | A method of scholar's name disambiguation of fusion academic | |
Wang et al. | A back‐propagation neural network‐based approach for multi‐represented feature matching in update propagation | |
CN110473592A (en) | The multi-angle of view mankind for having supervision based on figure convolutional network cooperate with lethal gene prediction technique | |
CN107633093A (en) | A kind of structure and its querying method of DECISION KNOWLEDGE collection of illustrative plates of powering | |
CN111931505A (en) | Cross-language entity alignment method based on subgraph embedding | |
CN105574541A (en) | Compactness sorting based network community discovery method | |
Shi et al. | Network embedding via community based variational autoencoder | |
CN110990718A (en) | Social network model building module of company image improving system | |
CN110704694A (en) | Organization hierarchy dividing method based on network representation learning and application thereof | |
Lu et al. | Multiple-kernel combination fuzzy clustering for community detection | |
Santiago et al. | A methodology for the characterization of flow conductivity through the identification of communities in samples of fractured rocks | |
Zhu et al. | Loan default prediction based on convolutional neural network and LightGBM | |
Guo et al. | Distribution network topology identification based on gradient boosting decision tree and attribute weighted naive Bayes | |
CN110992194A (en) | User reference index algorithm based on attribute-containing multi-process sampling graph representation learning model | |
Sundaram et al. | A Delaunay diagram‐based min–max CP‐tree algorithm for spatial data analysis | |
CN107424026A (en) | Businessman's reputation evaluation method and device | |
CN115204171A (en) | Document-level event extraction method and system based on hypergraph neural network | |
CN113095361A (en) | Comparable learning object generation method and system based on graph matching network |
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: 20160803 |
|
RJ01 | Rejection of invention patent application after publication |