CN111047453A - Detection method and device for decomposing large-scale social network community based on high-order tensor - Google Patents
Detection method and device for decomposing large-scale social network community based on high-order tensor Download PDFInfo
- Publication number
- CN111047453A CN111047453A CN201911228464.7A CN201911228464A CN111047453A CN 111047453 A CN111047453 A CN 111047453A CN 201911228464 A CN201911228464 A CN 201911228464A CN 111047453 A CN111047453 A CN 111047453A
- Authority
- CN
- China
- Prior art keywords
- social network
- community
- network
- module
- information
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 25
- 238000000034 method Methods 0.000 claims abstract description 37
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 24
- 238000004458 analytical method Methods 0.000 claims abstract description 23
- 230000006835 compression Effects 0.000 claims abstract description 23
- 238000011156 evaluation Methods 0.000 claims abstract description 21
- 238000007906 compression Methods 0.000 claims abstract description 17
- 239000011159 matrix material Substances 0.000 claims description 18
- 238000005070 sampling Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 6
- 238000013075 data extraction Methods 0.000 claims description 5
- 238000007418 data mining Methods 0.000 claims description 5
- 230000010365 information processing Effects 0.000 claims description 5
- 238000012546 transfer Methods 0.000 claims description 5
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 238000005065 mining Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
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)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of detection of social network communities, and discloses a detection method and a detection device for decomposing a large-scale social network community based on a high-order tensor, wherein the detection device for decomposing the large-scale social network community based on the high-order tensor comprises the following steps: the system comprises a network information acquisition module, a central control module, a community search module, a community sharing module, a network compression module, a community analysis module, an influence evaluation module and a display module. According to the method, the network compression module adopts a block item tensor decomposition method to construct a block item tensor layer so as to replace a full connection layer in an original social network, and the characteristics of symmetry and index expression capability of the block item tensor layer are utilized, so that the parameter quantity of the full connection layer can be greatly compressed, and the classification precision of the original network can be kept; meanwhile, the network topology structure can be better utilized through the community analysis module, and the analysis accuracy is high; and realizing the distinguishing of the types of the core nodes.
Description
Technical Field
The invention belongs to the technical field of detection of social network communities, and particularly relates to a detection method and device for decomposing a large-scale social network community based on high-order tensor.
Background
The social network covers all network service forms taking human social as a core, the internet is an interactive platform capable of communicating with each other, communicating with each other and participating in each other, the development of the internet exceeds the military and technical purposes of ARPANET at first, and the social network enables the internet to be expanded into a tool for human social from research departments, schools, governments and commercial application platforms. The network social contact expands the range of the social contact to the field of mobile phone platforms, and the mobile phone becomes a new social network carrier by means of universality of the mobile phone and application of a wireless network and by means of various kinds of software such as friend-making/instant messaging/mail transceivers and the like. Social network, i.e. network + social meaning. People are connected through the carrier of the network, thereby forming a community with a certain characteristic. However, existing social networks have low compression capabilities; meanwhile, the existing dynamic social network community evolution analysis does not well utilize a network topology structure, and the accuracy needs to be improved; different core nodes have different characteristics and contribute to detection of evolution events differently, but the prior art does not distinguish the types of the core nodes.
In summary, the problems of the prior art are as follows:
the existing social network has low compression capacity; meanwhile, the existing dynamic social network community evolution analysis does not well utilize a network topology structure, and the accuracy needs to be improved; different core nodes have different characteristics and contribute to detection of evolution events differently, but the prior art does not distinguish the types of the core nodes.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a detection method and a detection device for decomposing a large-scale social network community based on a high-order tensor.
The invention is realized in such a way that a detection method for decomposing a large-scale social network community based on a high-order tensor comprises the following steps:
the method comprises the following steps that firstly, communities and discovery are achieved by researching the difference between the characteristics of a real network topological structure and an assumed network model and by means of mathematical tools such as Bayesian inference and the like and maximizing likelihood probability;
determining search characteristics of social network community information, namely which social network community characteristic factors need to be acquired; according to the determined social network community characteristic factors, corresponding data are carried out on different social network objects by using a certain program, and the corresponding program is called to an application platform;
thirdly, similarity search is carried out on the collected data through a PathSim algorithm and the meta path on the relevant information collected by the network information module, and the hidden rich semanteme in the social heterogeneous information network is mined; analyzing the social network community evolution, and evaluating the influence of the social network community through an evaluation program;
step four, compressing the searched data and the evaluation data result to the social network; displaying the acquired social network community information, the search result, the shared information, the analysis result and the evaluation result through a display; sharing social network community information resources through a sharing program;
in the fourth step, the social network performs compression processing, and the specific process is as follows:
A. acquiring a social network frame through a compression program;
B. converting a weight matrix W and an input vector X in a full connection layer of a social network into a high-order tensor W and a high-order tensor X respectively;
C. carrying out block item tensor decomposition processing on the high-order tensor W in the step B;
D. replacing the fully-connected layer of the social network with a block item tensor layer according to the high-order tensor X in the step B and the high-order tensor W decomposed by the block item tensor in the step C;
E. d, training the social network replaced in the step D by adopting a back propagation algorithm;
in the step B, converting the weight matrix W in the fully-connected layer of the social network into the high-order tensor W specifically includes: uniformly dividing the row dimension and the column dimension of the weight matrix W to represent the weight matrix W in a hierarchical block matrix form, and converting the divided weight matrix W into a high-order tensor W.
Further, the step B of converting the input vector X in the fully-connected layer of the social network into the higher-order tensor X specifically includes: and expressing the dimensionality of the input vector X into a tensor form, and converting the tensor form into a higher-order tensor X.
Further, the step C of performing block entry tensor decomposition processing on the high-order tensor W in the step B specifically includes: rearranging the dimensionality of the high-order tensor W to enable the input and output corresponding to the row dimensionality and the column dimensionality to be in coupled arrangement in a paired mode, and decomposing the high-order tensor W into the sum of a plurality of tack decompositions.
Further, the step D of replacing the full connection layer of the social network with the block item tensor layer according to the high-order tensor X in the step B and the high-order tensor W decomposed by the block item tensor in the step C specifically includes: and multiplying the high-order tensor X by the high-order tensor W after the decomposition of the block item tensor, and replacing the full connection layer of the social network as a block item tensor layer.
Further, in the third step, the method for analyzing social network community evolution is as follows:
(1) aiming at a given dynamic social network through an analysis program, dividing a community structure corresponding to a time slice network for each time slice from a first time slice;
(2) calculating a superspeader set of each time slice network and a superblocker set of each community corresponding to the time slice network according to the community structure division result;
(3) determining the type of an evolution event 1 of each community aiming at the superspeader set; possible types of the evolution event 1 include production, merge, and expand events;
(4) and determining the evolution event 2 type of each community aiming at the superblocker set: possible types of evolution events 2 include vanishing, splitting, and reducing events.
Further, the dividing, for a given dynamic social network, the community structure corresponding to the time slice network for each time slice from the first time slice includes the following steps:
for a given dynamic social network, obtaining the neighbor relation between nodes in each time slice network from the first time slice;
and according to the neighbor relation, dividing the community structure corresponding to each time slice network through a QCA algorithm.
Further, the calculating of the superspreder set of each time slice network and the superblocker set of each community corresponding to the time slice network includes the following steps:
obtaining a superspeader set of each time slice network through a DegreeDiscount algorithm;
obtaining a superblocker set of each community corresponding to each time slice network through a CoreHD algorithm;
the method for determining the type of the evolution event 1 of each community aiming at the superspeader set comprises the following steps:
judging the type of an evolution event 1 according to a calculation model of the generated event, if a superspeader node of a current time slice t does not exist in a previous time slice t-1 or is not a node in a superspeader set, judging that the evolution event 1 is the generated event, wherein a community represented by the superspeader node is a newly generated community of the current time slice t;
judging the type of an evolution event 1 according to a calculation model of a merging event, if two superspears in the same community of a current time slice t belong to different communities in the previous time slice t-1, judging that the evolution event 1 is the merging event, and merging the communities represented by the two superspears at the current time slice t;
and judging the type of the evolution event 1 according to a calculation model of the expansion event, if the scale of the superspeader node in a certain community of the current time slice t is larger than that of the superspeader node in the community corresponding to the previous time slice t-1, judging that the evolution event 1 is the expansion event, and expanding the community at the current time slice t.
Further, in the second step, the specific process of the network information collection module for collecting the social network community information data is as follows:
firstly, determining the search characteristics of social network community information, namely which social network community characteristic factors need to be acquired;
according to the determined social network community characteristic factor, corresponding data are carried out on different social network objects by using a certain program;
in the process of acquiring the data information of the social network, scheduling and acquiring are carried out, and the data information is updated and acquired synchronously with the corresponding nodes;
and isomerizing the acquired social network data into isomorphism, and calling the isomorphism to an application platform by using a corresponding program.
Further, in the fourth step, the method for evaluating community influence is as follows:
according to the information data of each social network community searched by other modules of the system, removing data which is unfavorable for the community in the information data of the social network community;
carrying out data mining and extraction on the transfer capacity of the user and the comments of the user in the social network community information data to obtain the influence of the user;
calculating the scale of a social network and the amount of users according to the influence of the users; sampling information in a social network of a certain scale, and estimating the influence of one social network;
and sampling and estimating the influence of the social network at a higher level according to the influence of the social network.
Another object of the present invention is to provide a detection apparatus for decomposing a large-scale social network community based on a higher order tensor, which implements the detection method for decomposing a large-scale social network community based on a higher order tensor, the detection apparatus for decomposing a large-scale social network community based on a higher order tensor includes:
the network information acquisition module is connected with the central control module and is used for determining the search characteristics of the social network community information, namely which social network community characteristic factors need to be acquired; according to the determined social network community characteristic factor, corresponding data are carried out on different social network objects by using a certain program; in the process of acquiring the data information of the social network, scheduling and acquiring are carried out, and the data information is updated and acquired synchronously with the corresponding nodes; isomerizing the acquired social network data into isomorphism, and calling the isomorphism to an application platform by using a corresponding program;
the heterogeneous network information processing module is connected with the central control module and used for carrying out similarity search on the related information acquired by the network information module through a meta path by utilizing a PathSim algorithm and mining the implicit rich semanteme in the social heterogeneous information network;
the central control module is connected with the network information acquisition module, the community search module, the community sharing module, the network compression module, the community analysis module, the influence evaluation module and the display module and is used for controlling each module to normally work through the host;
the community searching module is connected with the central control module, and realizes communities and discovery by researching the difference between the characteristics of a real network topological structure and a hypothetical network model and by using mathematical tools such as Bayesian inference and the like and maximizing likelihood probability;
the community sharing module is connected with the central control module and used for sharing social network community information resources through a sharing program;
the network compression module is connected with the central control module and is used for compressing the social network through a compression program;
the community analysis module is connected with the central control module and is used for analyzing social network community evolution;
the influence evaluation module is connected with the central control module and is used for removing data which are unfavorable for the community in the social network community information data according to the social network community information data searched by other modules of the system; carrying out data mining and extraction on the transfer capacity of the user and the comments of the user in the social network community information data to obtain the influence of the user; calculating the scale of a social network and the amount of users according to the influence of the users; sampling information in a social network of a certain scale, and estimating the influence of one social network; according to the influence of a social network, sampling and estimating the influence of the social network at a higher level;
and the display module is connected with the central control module and used for displaying the acquired social network community information, the search result, the shared information, the analysis result and the evaluation result through the display.
The invention has the advantages and positive effects that:
according to the method, the network compression module adopts a block item tensor decomposition method to construct a block item tensor layer so as to replace a full connection layer in an original social network, and the characteristics of symmetry and index expression capability of the block item tensor layer are utilized, so that the parameter quantity of the full connection layer can be greatly compressed, and the classification precision of the original network can be kept; meanwhile, the network topology structure can be better utilized through the community analysis module, and the analysis accuracy is high; and realizing the distinguishing of the types of the core nodes. The method for acquiring the social network community information data by the network information acquisition module for acquiring the social network community information data can enlarge the acquisition range according to the hierarchy of the social network community and improve the comprehensiveness of the social network community information data. According to the method for evaluating the community influence by the influence evaluation module for evaluating the social network community influence through the evaluation program, the correctness of the evaluation of the social network community influence can be improved.
Drawings
Fig. 1 is a flowchart of a detection method for decomposing a large-scale social network community based on a high-order tensor according to an embodiment of the present invention.
Fig. 2 is a block diagram of a structure of a detection apparatus for decomposing a large-scale social network community based on a higher-order tensor according to an embodiment of the present invention.
In fig. 2: 1. a network information acquisition module; 2. a heterogeneous network information processing module; 3. a central control module; 4. a community search module; 5. a community sharing module; 6. a network compression module; 7. a community analysis module; 8. an influence evaluation module; 9. and a display module.
Detailed Description
In order to further understand the contents, features and effects of the present invention, the following embodiments are illustrated and described in detail with reference to the accompanying drawings.
The structure of the present invention will be described in detail below with reference to the accompanying drawings.
As shown in fig. 1, a detection method for decomposing a large-scale social network community based on a higher-order tensor provided by an embodiment of the present invention includes the following steps:
s101: by researching the difference between the characteristics of the real network topological structure and the assumed network model, and by using mathematical tools such as Bayesian inference and the like, the community and discovery are realized by maximizing the likelihood probability.
S102: determining search characteristics of social network community information, namely which social network community characteristic factors need to be acquired; and according to the determined social network community characteristic factors, carrying out corresponding data on different social network objects by using a certain program, and calling the corresponding data to the application platform by using the corresponding program.
S103: similarity search is carried out on the collected data through a PathSim algorithm and the related information collected by the network information module through a meta path, and the hidden rich semanteme in the social heterogeneous information network is mined; and analyzing the social network community evolution, and evaluating the influence of the social network community through an evaluation program.
S104: compressing the searched data and the evaluation data result to the social network; displaying the acquired social network community information, the search result, the shared information, the analysis result and the evaluation result through a display; and sharing the social network community information resources through a sharing program.
As shown in fig. 2, a detection apparatus for decomposing a large-scale social network community based on a higher-order tensor according to an embodiment of the present invention includes: the system comprises a network information acquisition module 1, a heterogeneous network information processing module 2, a central control module 3, a community search module 4, a community sharing module 5, a network compression module 6, a community analysis module 7, an influence evaluation module 8 and a display module 9.
And the network information acquisition module 1 is connected with the central control module 3 and is used for acquiring social network community information data.
And the heterogeneous network information processing module 2 is connected with the central control module 3 and is used for performing similarity search on the related information acquired by the network information module through a meta path by utilizing a PathSim algorithm and mining the implicit rich semantic meaning in the social heterogeneous information network.
The central control module 3 is connected with the network information acquisition module 1, the community search module 4, the community sharing module 5, the network compression module 6, the community analysis module 7, the influence evaluation module 8 and the display module 9, and is used for controlling the normal work of each module through a host.
And the community searching module 4 is connected with the central control module 3, and realizes communities and discovery by researching the difference between the real network topological structure characteristics and the assumed network model and by using mathematical tools such as Bayesian inference and the like and maximizing the likelihood probability.
And the community sharing module 5 is connected with the central control module 3 and is used for sharing the social network community information resources through a sharing program.
And the network compression module 6 is connected with the central control module 3 and is used for compressing the social network through a compression program.
And the community analysis module 7 is connected with the central control module 3 and is used for analyzing social network community evolution.
And the influence evaluation module 8 is connected with the central control module 3 and is used for evaluating the influence of the social network community through an evaluation program.
And the display module 9 is connected with the central control module 3 and is used for displaying the acquired social network community information, the search result, the shared information, the analysis result and the evaluation result through a display.
The specific process of the network information acquisition module which is connected with the central control module 3 and used for acquiring the social network community information data, provided by the invention, for acquiring the social network community information data is as follows:
firstly, determining the search characteristics of social network community information, namely which social network community characteristic factors need to be acquired;
according to the determined social network community characteristic factor, corresponding data are carried out on different social network objects by using a certain program;
in the process of acquiring the data information of the social network, scheduling and acquiring are carried out, and the data information is updated and acquired synchronously with the corresponding nodes;
and isomerizing the acquired social network data into isomorphism, and calling the isomorphism to an application platform by using a corresponding program.
The compression method of the network compression module 6 provided by the embodiment of the invention is as follows:
A. acquiring a social network frame through a compression program;
B. converting a weight matrix W and an input vector X in a full connection layer of a social network into a high-order tensor W and a high-order tensor X respectively;
C. carrying out block item tensor decomposition processing on the high-order tensor W in the step B;
D. replacing the fully-connected layer of the social network with a block item tensor layer according to the high-order tensor X in the step B and the high-order tensor W decomposed by the block item tensor in the step C;
E. and D, training the social network replaced in the step D by adopting a back propagation algorithm.
In step B provided by the embodiment of the present invention, converting the weight matrix W in the full connection layer of the social network into the high-order tensor W specifically includes: uniformly dividing the row dimension and the column dimension of the weight matrix W to represent the weight matrix W in a hierarchical block matrix form, and converting the divided weight matrix W into a high-order tensor W.
In step B provided by the embodiment of the present invention, converting the input vector X in the full connection layer of the social network into the high-order tensor X specifically includes: and expressing the dimensionality of the input vector X into a tensor form, and converting the tensor form into a higher-order tensor X.
The step C provided in the embodiment of the present invention specifically performs, on the high-order tensor W in the step B, a block entry tensor decomposition process as follows: rearranging the dimensionality of the high-order tensor W to enable the input and output corresponding to the row dimensionality and the column dimensionality to be in coupled arrangement in a paired mode, and decomposing the high-order tensor W into the sum of a plurality of tack decompositions.
The step D provided in the embodiment of the present invention, according to the high-order tensor X in the step B and the high-order tensor W after the block item tensor decomposition in the step C, replacing the full connection layer of the social network with the block item tensor layer specifically is: and multiplying the high-order tensor X by the high-order tensor W after the decomposition of the block item tensor, and replacing the full connection layer of the social network as a block item tensor layer.
The community analysis module 7 provided by the embodiment of the invention has the following analysis method:
(1) aiming at a given dynamic social network through an analysis program, dividing a community structure corresponding to a time slice network for each time slice from a first time slice;
(2) calculating a superspeader set of each time slice network and a superblocker set of each community corresponding to the time slice network according to the community structure division result;
(3) determining the type of an evolution event 1 of each community aiming at the superspeader set; possible types of the evolution event 1 include production, merge, and expand events;
(4) and determining the evolution event 2 type of each community aiming at the superblocker set: possible types of evolution events 2 include vanishing, splitting, and reducing events.
The method for dividing the community structure corresponding to the time slice network for the given dynamic social network from the first time slice comprises the following steps:
for a given dynamic social network, obtaining the neighbor relation between nodes in each time slice network from the first time slice;
and according to the neighbor relation, dividing the community structure corresponding to each time slice network through a QCA algorithm.
The method for calculating the superspeaker set of each time slice network and the supersblocker set of each community corresponding to the time slice network, provided by the embodiment of the invention, comprises the following steps:
obtaining a superspeader set of each time slice network through a DegreeDiscount algorithm;
and obtaining a superblocker set of each community corresponding to each time slice network through a CoreHD algorithm.
The method for determining the evolution event 1 type of each community aiming at the superspeader set comprises the following steps:
judging the type of an evolution event 1 according to a calculation model of the generated event, if a superspeader node of a current time slice t does not exist in a previous time slice t-1 or is not a node in a superspeader set, judging that the evolution event 1 is the generated event, wherein a community represented by the superspeader node is a newly generated community of the current time slice t;
judging the type of an evolution event 1 according to a calculation model of a merging event, if two superspears in the same community of a current time slice t belong to different communities in the previous time slice t-1, judging that the evolution event 1 is the merging event, and merging the communities represented by the two superspears at the current time slice t;
and judging the type of the evolution event 1 according to a calculation model of the expansion event, if the scale of the superspeader node in a certain community of the current time slice t is larger than that of the superspeader node in the community corresponding to the previous time slice t-1, judging that the evolution event 1 is the expansion event, and expanding the community at the current time slice t.
The invention provides a method for evaluating community influence by an influence evaluation module 8 which is connected with a central control module 3 and used for evaluating the community influence of a social network through an evaluation program, which comprises the following steps:
according to the information data of each social network community searched by other modules of the system, removing data which is unfavorable for the community in the information data of the social network community;
carrying out data mining and extraction on the transfer capacity of the user and the comments of the user in the social network community information data to obtain the influence of the user;
calculating the scale of a social network and the amount of users according to the influence of the users; sampling information in a social network of a certain scale, and estimating the influence of one social network;
and sampling and estimating the influence of the social network at a higher level according to the influence of the social network.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and all simple modifications, equivalent changes and modifications made to the above embodiment according to the technical spirit of the present invention are within the scope of the technical solution of the present invention.
Claims (10)
1. The detection method for decomposing the large-scale social network community based on the higher order tensor according to claim 1, wherein the detection method for decomposing the large-scale social network community based on the higher order tensor comprises the following steps:
the method comprises the following steps that firstly, communities and discovery are achieved by researching the difference between the characteristics of a real network topological structure and an assumed network model and by means of mathematical tools such as Bayesian inference and the like and maximizing likelihood probability;
determining search characteristics of social network community information, namely which social network community characteristic factors need to be acquired; according to the determined social network community characteristic factors, corresponding data are carried out on different social network objects by using a certain program, and the corresponding program is called to an application platform;
thirdly, similarity search is carried out on the collected data through a PathSim algorithm and the meta path on the relevant information collected by the network information module, and the hidden rich semanteme in the social heterogeneous information network is mined; analyzing the social network community evolution, and evaluating the influence of the social network community through an evaluation program;
step four, compressing the searched data and the evaluation data result to the social network; displaying the acquired social network community information, the search result, the shared information, the analysis result and the evaluation result through a display; sharing social network community information resources through a sharing program;
in the fourth step, the social network performs compression processing, and the specific process is as follows:
A. acquiring a social network frame through a compression program;
B. converting a weight matrix W and an input vector X in a full connection layer of a social network into a high-order tensor W and a high-order tensor X respectively;
C. carrying out block item tensor decomposition processing on the high-order tensor W in the step B;
D. replacing the fully-connected layer of the social network with a block item tensor layer according to the high-order tensor X in the step B and the high-order tensor W decomposed by the block item tensor in the step C;
E. d, training the social network replaced in the step D by adopting a back propagation algorithm;
in the step B, converting the weight matrix W in the fully-connected layer of the social network into the high-order tensor W specifically includes: uniformly dividing the row dimension and the column dimension of the weight matrix W to represent the weight matrix W in a hierarchical block matrix form, and converting the divided weight matrix W into a high-order tensor W.
2. The method for detecting large-scale social network community decomposition based on higher-order tensor as claimed in claim 1, wherein the step B of converting the input vector X in the fully-connected layer of the social network into the higher-order tensor X specifically includes: and expressing the dimensionality of the input vector X into a tensor form, and converting the tensor form into a higher-order tensor X.
3. The method for detecting the decomposition of the large-scale social network community based on the higher-order tensor according to claim 1, wherein the step C of performing the block entry tensor decomposition on the higher-order tensor W in the step B specifically comprises: rearranging the dimensionality of the high-order tensor W to enable the input and output corresponding to the row dimensionality and the column dimensionality to be in coupled arrangement in a paired mode, and decomposing the high-order tensor W into the sum of a plurality of tack decompositions.
4. The method for detecting the decomposition of the large-scale social network community based on the higher-order tensor according to claim 1, wherein the step D replaces the fully-connected layer of the social network with the block item tensor layer according to the higher-order tensor X in the step B and the higher-order tensor W after the block item tensor decomposition in the step C, specifically: and multiplying the high-order tensor X by the high-order tensor W after the decomposition of the block item tensor, and replacing the full connection layer of the social network as a block item tensor layer.
5. The method for detecting the decomposition of the large-scale social network community based on the high-order tensor as set forth in claim 1, wherein in the third step, the method for analyzing the evolution of the social network community comprises the following steps:
(1) aiming at a given dynamic social network through an analysis program, dividing a community structure corresponding to a time slice network for each time slice from a first time slice;
(2) calculating a superspeader set of each time slice network and a superblocker set of each community corresponding to the time slice network according to the community structure division result;
(3) determining the type of an evolution event 1 of each community aiming at the superspeader set; possible types of the evolution event 1 include production, merge, and expand events;
(4) and determining the evolution event 2 type of each community aiming at the superblocker set: possible types of evolution events 2 include vanishing, splitting, and reducing events.
6. The method for detecting community decomposition of large-scale social networks based on higher-order tensors as claimed in claim 5, wherein said dividing the community structure corresponding to the time slice network for each time slice starting from the first time slice for a given dynamic social network comprises the following steps:
for a given dynamic social network, obtaining the neighbor relation between nodes in each time slice network from the first time slice;
and according to the neighbor relation, dividing the community structure corresponding to each time slice network through a QCA algorithm.
7. The method for detecting the decomposition of the large-scale social network communities based on the higher-order tensor as claimed in claim 5, wherein the step of calculating the superspreder set of each time slice network and the superblocker set of each community corresponding to the time slice network comprises the following steps:
obtaining a superspeader set of each time slice network through a DegreeDiscount algorithm;
obtaining a superblocker set of each community corresponding to each time slice network through a CoreHD algorithm;
the method for determining the type of the evolution event 1 of each community aiming at the superspeader set comprises the following steps:
judging the type of an evolution event 1 according to a calculation model of the generated event, if a superspeader node of a current time slice t does not exist in a previous time slice t-1 or is not a node in a superspeader set, judging that the evolution event 1 is the generated event, wherein a community represented by the superspeader node is a newly generated community of the current time slice t;
judging the type of an evolution event 1 according to a calculation model of a merging event, if two superspears in the same community of a current time slice t belong to different communities in the previous time slice t-1, judging that the evolution event 1 is the merging event, and merging the communities represented by the two superspears at the current time slice t;
and judging the type of the evolution event 1 according to a calculation model of the expansion event, if the scale of the superspeader node in a certain community of the current time slice t is larger than that of the superspeader node in the community corresponding to the previous time slice t-1, judging that the evolution event 1 is the expansion event, and expanding the community at the current time slice t.
8. The method for detecting the decomposition of the large-scale social network community based on the higher-order tensor as set forth in claim 1, wherein in the second step, the specific process of collecting the social network community information data by the network information collecting module for collecting the social network community information data is as follows:
firstly, determining the search characteristics of social network community information, namely which social network community characteristic factors need to be acquired;
according to the determined social network community characteristic factor, corresponding data are carried out on different social network objects by using a certain program;
in the process of acquiring the data information of the social network, scheduling and acquiring are carried out, and the data information is updated and acquired synchronously with the corresponding nodes;
and isomerizing the acquired social network data into isomorphism, and calling the isomorphism to an application platform by using a corresponding program.
9. The method for detecting the decomposition of the large-scale social network community based on the higher-order tensor as set forth in claim 1, wherein in the fourth step, the influence of the community is evaluated as follows:
according to the information data of each social network community searched by other modules of the system, removing data which is unfavorable for the community in the information data of the social network community;
carrying out data mining and extraction on the transfer capacity of the user and the comments of the user in the social network community information data to obtain the influence of the user;
calculating the scale of a social network and the amount of users according to the influence of the users; sampling information in a social network of a certain scale, and estimating the influence of one social network;
and sampling and estimating the influence of the social network at a higher level according to the influence of the social network.
10. A detection apparatus for decomposing a large-scale social network community based on a higher order tensor, which implements the detection method for decomposing a large-scale social network community based on a higher order tensor according to claims 1-9, wherein the detection apparatus for decomposing a large-scale social network community based on a higher order tensor comprises:
the network information acquisition module is connected with the central control module and is used for determining the search characteristics of the social network community information, namely which social network community characteristic factors need to be acquired; according to the determined social network community characteristic factor, corresponding data are carried out on different social network objects by using a certain program; in the process of acquiring the data information of the social network, scheduling and acquiring are carried out, and the data information is updated and acquired synchronously with the corresponding nodes; isomerizing the acquired social network data into isomorphism, and calling the isomorphism to an application platform by using a corresponding program;
the heterogeneous network information processing module is connected with the central control module and used for carrying out similarity search on the related information acquired by the network information module through a meta path by utilizing a PathSim algorithm and mining the implicit rich semanteme in the social heterogeneous information network;
the central control module is connected with the network information acquisition module, the community search module, the community sharing module, the network compression module, the community analysis module, the influence evaluation module and the display module and is used for controlling each module to normally work through the host;
the community searching module is connected with the central control module, and realizes communities and discovery by researching the difference between the characteristics of a real network topological structure and a hypothetical network model and by using mathematical tools such as Bayesian inference and the like and maximizing likelihood probability;
the community sharing module is connected with the central control module and used for sharing social network community information resources through a sharing program;
the network compression module is connected with the central control module and is used for compressing the social network through a compression program;
the community analysis module is connected with the central control module and is used for analyzing social network community evolution;
the influence evaluation module is connected with the central control module and is used for removing data which are unfavorable for the community in the social network community information data according to the social network community information data searched by other modules of the system; carrying out data mining and extraction on the transfer capacity of the user and the comments of the user in the social network community information data to obtain the influence of the user; calculating the scale of a social network and the amount of users according to the influence of the users; sampling information in a social network of a certain scale, and estimating the influence of one social network; according to the influence of a social network, sampling and estimating the influence of the social network at a higher level;
and the display module is connected with the central control module and used for displaying the acquired social network community information, the search result, the shared information, the analysis result and the evaluation result through the display.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911228464.7A CN111047453A (en) | 2019-12-04 | 2019-12-04 | Detection method and device for decomposing large-scale social network community based on high-order tensor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911228464.7A CN111047453A (en) | 2019-12-04 | 2019-12-04 | Detection method and device for decomposing large-scale social network community based on high-order tensor |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111047453A true CN111047453A (en) | 2020-04-21 |
Family
ID=70234588
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911228464.7A Pending CN111047453A (en) | 2019-12-04 | 2019-12-04 | Detection method and device for decomposing large-scale social network community based on high-order tensor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111047453A (en) |
Citations (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070239707A1 (en) * | 2006-04-03 | 2007-10-11 | Collins John B | Method of searching text to find relevant content |
US20100185935A1 (en) * | 2009-01-21 | 2010-07-22 | Nec Laboratories America, Inc. | Systems and methods for community detection |
CN102946331A (en) * | 2012-10-10 | 2013-02-27 | 北京交通大学 | Detecting method and device for zombie users of social networks |
CN103325061A (en) * | 2012-11-02 | 2013-09-25 | 中国人民解放军国防科学技术大学 | Community discovery method and system |
CN103631970A (en) * | 2013-12-20 | 2014-03-12 | 百度在线网络技术(北京)有限公司 | Method and device for mining associated relationship between attributes and entities |
CN103810260A (en) * | 2014-01-27 | 2014-05-21 | 西安理工大学 | Complex network community discovery method based on topological characteristics |
CN104268629A (en) * | 2014-09-15 | 2015-01-07 | 西安电子科技大学 | Complex network community detecting method based on prior information and network inherent information |
CN104484825A (en) * | 2014-12-05 | 2015-04-01 | 上海师范大学 | Evaluation algorithm of community influence of social networks |
CN105095988A (en) * | 2015-07-01 | 2015-11-25 | 中国科学院计算技术研究所 | Method and system for detecting social network information explosion |
CN105677844A (en) * | 2016-01-06 | 2016-06-15 | 北京摩比万思科技有限公司 | Mobile advertisement big data directional pushing and user cross-screen recognition method |
CN105825430A (en) * | 2016-01-08 | 2016-08-03 | 南通弘数信息科技有限公司 | Heterogeneous social network-based detection method |
CN106682142A (en) * | 2016-12-21 | 2017-05-17 | 兰州交通大学 | Method for excavating user emotions and analyzing propagation features under specific event situation |
CN106789285A (en) * | 2016-12-28 | 2017-05-31 | 西安交通大学 | A kind of multiple dimensioned community discovery method of online community network |
CN106778537A (en) * | 2016-11-28 | 2017-05-31 | 中国科学院心理研究所 | A kind of collection of animal social network structure and analysis system and its method based on image procossing |
CN106780053A (en) * | 2015-11-20 | 2017-05-31 | 香港中文大学深圳研究院 | A kind of overlap community discovery method and system based on node connection preference |
CN106802956A (en) * | 2017-01-19 | 2017-06-06 | 山东大学 | A kind of film based on weighting Heterogeneous Information network recommends method |
CN107153713A (en) * | 2017-05-27 | 2017-09-12 | 合肥工业大学 | Overlapping community detection method and system based on similitude between node in social networks |
CN107527295A (en) * | 2017-08-24 | 2017-12-29 | 中南大学 | Dynamics community of Academic Teams based on tense coauthorship network finds method and its method for evaluating quality |
CN107944556A (en) * | 2017-12-12 | 2018-04-20 | 电子科技大学 | Deep neural network compression method based on block item tensor resolution |
CN108062360A (en) * | 2017-11-29 | 2018-05-22 | 广东技术师范学院 | A kind of method, system and device of large-scale complex community structure detection |
CN108304496A (en) * | 2018-01-11 | 2018-07-20 | 上海交通大学 | Node similarity relation detection method based on composite unit path in Heterogeneous Information net |
CN108334580A (en) * | 2018-01-25 | 2018-07-27 | 重庆邮电大学 | A kind of community discovery method of combination link and attribute information |
CN108899898A (en) * | 2018-07-12 | 2018-11-27 | 四川大学 | A kind of reactive voltage partition method |
CN109460299A (en) * | 2018-11-08 | 2019-03-12 | 中山大学 | A kind of distributed parallel multi-source social network data acquisition system and method |
CN109493246A (en) * | 2018-11-22 | 2019-03-19 | 中国矿业大学 | A kind of dynamic myspace evolution analysis method and its system |
CN109829089A (en) * | 2018-12-12 | 2019-05-31 | 中国科学院计算技术研究所 | Social network user method for detecting abnormality and system based on association map |
CN110334264A (en) * | 2019-06-27 | 2019-10-15 | 北京邮电大学 | A kind of community detection method and device for isomery dynamic information network |
CN110443265A (en) * | 2018-05-04 | 2019-11-12 | 北京京东尚科信息技术有限公司 | A kind of behavioral value method and apparatus based on corporations |
-
2019
- 2019-12-04 CN CN201911228464.7A patent/CN111047453A/en active Pending
Patent Citations (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070239707A1 (en) * | 2006-04-03 | 2007-10-11 | Collins John B | Method of searching text to find relevant content |
US20100185935A1 (en) * | 2009-01-21 | 2010-07-22 | Nec Laboratories America, Inc. | Systems and methods for community detection |
CN102946331A (en) * | 2012-10-10 | 2013-02-27 | 北京交通大学 | Detecting method and device for zombie users of social networks |
CN103325061A (en) * | 2012-11-02 | 2013-09-25 | 中国人民解放军国防科学技术大学 | Community discovery method and system |
CN103631970A (en) * | 2013-12-20 | 2014-03-12 | 百度在线网络技术(北京)有限公司 | Method and device for mining associated relationship between attributes and entities |
CN103810260A (en) * | 2014-01-27 | 2014-05-21 | 西安理工大学 | Complex network community discovery method based on topological characteristics |
CN104268629A (en) * | 2014-09-15 | 2015-01-07 | 西安电子科技大学 | Complex network community detecting method based on prior information and network inherent information |
CN104484825A (en) * | 2014-12-05 | 2015-04-01 | 上海师范大学 | Evaluation algorithm of community influence of social networks |
CN105095988A (en) * | 2015-07-01 | 2015-11-25 | 中国科学院计算技术研究所 | Method and system for detecting social network information explosion |
CN106780053A (en) * | 2015-11-20 | 2017-05-31 | 香港中文大学深圳研究院 | A kind of overlap community discovery method and system based on node connection preference |
CN105677844A (en) * | 2016-01-06 | 2016-06-15 | 北京摩比万思科技有限公司 | Mobile advertisement big data directional pushing and user cross-screen recognition method |
CN105825430A (en) * | 2016-01-08 | 2016-08-03 | 南通弘数信息科技有限公司 | Heterogeneous social network-based detection method |
CN106778537A (en) * | 2016-11-28 | 2017-05-31 | 中国科学院心理研究所 | A kind of collection of animal social network structure and analysis system and its method based on image procossing |
CN106682142A (en) * | 2016-12-21 | 2017-05-17 | 兰州交通大学 | Method for excavating user emotions and analyzing propagation features under specific event situation |
CN106789285A (en) * | 2016-12-28 | 2017-05-31 | 西安交通大学 | A kind of multiple dimensioned community discovery method of online community network |
CN106802956A (en) * | 2017-01-19 | 2017-06-06 | 山东大学 | A kind of film based on weighting Heterogeneous Information network recommends method |
CN107153713A (en) * | 2017-05-27 | 2017-09-12 | 合肥工业大学 | Overlapping community detection method and system based on similitude between node in social networks |
CN107527295A (en) * | 2017-08-24 | 2017-12-29 | 中南大学 | Dynamics community of Academic Teams based on tense coauthorship network finds method and its method for evaluating quality |
CN108062360A (en) * | 2017-11-29 | 2018-05-22 | 广东技术师范学院 | A kind of method, system and device of large-scale complex community structure detection |
CN107944556A (en) * | 2017-12-12 | 2018-04-20 | 电子科技大学 | Deep neural network compression method based on block item tensor resolution |
CN108304496A (en) * | 2018-01-11 | 2018-07-20 | 上海交通大学 | Node similarity relation detection method based on composite unit path in Heterogeneous Information net |
CN108334580A (en) * | 2018-01-25 | 2018-07-27 | 重庆邮电大学 | A kind of community discovery method of combination link and attribute information |
CN110443265A (en) * | 2018-05-04 | 2019-11-12 | 北京京东尚科信息技术有限公司 | A kind of behavioral value method and apparatus based on corporations |
CN108899898A (en) * | 2018-07-12 | 2018-11-27 | 四川大学 | A kind of reactive voltage partition method |
CN109460299A (en) * | 2018-11-08 | 2019-03-12 | 中山大学 | A kind of distributed parallel multi-source social network data acquisition system and method |
CN109493246A (en) * | 2018-11-22 | 2019-03-19 | 中国矿业大学 | A kind of dynamic myspace evolution analysis method and its system |
CN109829089A (en) * | 2018-12-12 | 2019-05-31 | 中国科学院计算技术研究所 | Social network user method for detecting abnormality and system based on association map |
CN110334264A (en) * | 2019-06-27 | 2019-10-15 | 北京邮电大学 | A kind of community detection method and device for isomery dynamic information network |
Non-Patent Citations (2)
Title |
---|
李立: "基于元路径选择和融合的异构信息网络社区挖掘算法研究" * |
陈硕: ""基于张量分解的微博社区发现研究" * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107229708B (en) | Personalized travel service big data application system and method | |
Ranshous et al. | Anomaly detection in dynamic networks: a survey | |
CN103678670B (en) | Micro-blog hot word and hot topic mining system and method | |
JP5092165B2 (en) | Data construction method and system | |
CN104820708B (en) | A kind of big data clustering method and device based on cloud computing platform | |
CN111611488A (en) | Information recommendation method and device based on artificial intelligence and electronic equipment | |
CN110555172A (en) | user relationship mining method and device, electronic equipment and storage medium | |
CN110647995A (en) | Rule training method, device, equipment and storage medium | |
CN116455861B (en) | Big data-based computer network security monitoring system and method | |
CN113011471A (en) | Social group dividing method, social group dividing system and related devices | |
CN115795329B (en) | Power utilization abnormal behavior analysis method and device based on big data grid | |
CN116881430B (en) | Industrial chain identification method and device, electronic equipment and readable storage medium | |
CN111831706A (en) | Mining method and device for association rules among applications and storage medium | |
Souravlas et al. | Probabilistic community detection in social networks | |
CN112765313B (en) | False information detection method based on original text and comment information analysis algorithm | |
CN108345620B (en) | Brand information processing method, brand information processing device, storage medium and electronic equipment | |
CN105512270B (en) | Method and device for determining related objects | |
CN115329078B (en) | Text data processing method, device, equipment and storage medium | |
CN116450938A (en) | Work order recommendation realization method and system based on map | |
Song et al. | Inferring explicit and implicit social ties simultaneously in mobile social networks | |
CN111047453A (en) | Detection method and device for decomposing large-scale social network community based on high-order tensor | |
CN111107493B (en) | Method and system for predicting position of mobile user | |
CN113987126A (en) | Retrieval method and device based on knowledge graph | |
CN114003803A (en) | Method and system for discovering media account in specific region on social platform | |
CN112446542A (en) | Social network link prediction method based on attention neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200421 |