CN110321493A - A kind of abnormality detection of social networks and optimization method, system and computer equipment - Google Patents
A kind of abnormality detection of social networks and optimization method, system and computer equipment Download PDFInfo
- Publication number
- CN110321493A CN110321493A CN201910549606.3A CN201910549606A CN110321493A CN 110321493 A CN110321493 A CN 110321493A CN 201910549606 A CN201910549606 A CN 201910549606A CN 110321493 A CN110321493 A CN 110321493A
- Authority
- CN
- China
- Prior art keywords
- social networks
- matrix
- link
- likelihood
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention belongs to the secret protection technical fields of social relationships sensitive in social networks; specially a kind of abnormality detection of social networks and optimization method, system and equipment; the method includes converting diagram data for social networks; and reinforcing denoising is carried out to diagram data using restructing algorithm, obtain reconstruct social networks；Definition rule measurement is horizontal come the systematicness for assessing reconstruct social networks, calculates the likelihood of network link, and judges the critical link that network link is also missing from for false link；By the limited critical link of selection and the reconfigurability based on their regulating networks, social networks is optimized, the social networks regulated and controled on demand；One aspect of the present invention can identify dummy account, on the other hand can be optimized again to reconstructed network, promote its validity and accuracy.
Description
Technical field
The invention belongs to the technical field of safety protection of sensitive relations in complicated social networks and the excavations of critical link
Analysis and optimising and adjustment technology, specially a kind of abnormality detection of social networks regulatory mechanism and optimization method, system and calculating
Machine equipment.
Background technique
In online social networks, potential commercial interest results in the generation and diffusion of dummy account.In social networks
There is a large amount of dummy accounts, by taking microblogging as an example, have a large amount of robot account to forward various advertisements or serve as waterborne troops, also
Behaviors, these accounts such as pornographic video, publication rumour, swindle, which are peddled, with the presence of some accounts belongs to dummy account.False account
The presence at family causes significant threat to safety.Also, noise (the mistake of different level is constantly present in real social networks
Or distracter), the systematicness and mode of social networks are just not obvious enough, and it is unfavorable to generate to the effect of abnormality detection and regulation
Influence, it is therefore necessary to accurately reconstructed to social networks, and necessary control methods are taken based on critical link
Reconstructed network is optimized to reach.
But traditional reconstructing method can not measure the inherent re-configurability of network, reduce evaluation to network and
Optimization, lays particular emphasis on modeling rule component as accurately as possible, clearly captures the inherent pass between minor structure and network reconfiguration
System.Therefore, effect of the microcosmos network element in macro network analysis is indefinite, and it is poor to eventually lead to interpretability.And structure
Application of the pattern learning technology in network reconfiguration is also little affected by attention, and it is lower that this will lead to reconstruction accuracy, to influence me
To the abnormality detection of dummy account.If can accurately quantify network structure regularity, pass through disturbing for social networks link
Dynamic regulation, changes tactic pattern relevant to target object, can also provide a new think of for optimizing for social networks
Road.
Summary of the invention
In order to identify, there are a large amount of dummy accounts on social networks, and the social networks of optimal reconfiguration, the present invention propose
A kind of abnormality detection based on social networks regulatory mechanism and optimization method, system and computer equipment, such as Fig. 1, the method
Comprising the following specific steps
S1, according to social networks construct diagram data, obtain the adjacency matrix of diagram data, using restructing algorithm to diagram data into
Row strengthens denoising, obtains reconstruct social networks；
S2, definition rule level are horizontal come the systematicness for assessing reconstruct social networks, calculate the likelihood of network link；
S3, in the social networks of reconstruct, be ranked up by likelihood of the ascending order to link, and obtain ranked list；
S4, ranked list is successively traversed since the first item of ranked list, if calculating social networks after removal currentitem
Systematicness it is horizontal, if current systematicness level increases, remove the corresponding link of this；
The social networks of S5, output after step S4 optimization.
Further, step S1 includes:
It S11, is that side constructs a non-directed graph as figure number using relationship of the user in social networks between node, user
According to, and diagram data is indicated with adjacency matrix；
S12, building low-rank from indicating network model, using augmentation Lagrange multiplier method obtain optimal representing matrix and
Error matrix；
S13, the link of reconstruct social networks is obtained by optimal representing matrix and basic matrix, and there are Likelihood matrix SM；
S14, matrix SM is divided into positive component SM with entry symbol+With negative component SM-；
If S15, positive component SM+In entry exist in adjacency matrix X, then exclude the entry, link in remaining entry
Likelihood be higher than threshold value entry be lack link；
If S16, negative component SM-In entry do not have in adjacency matrix X, i.e., respective items are 0, then the entry, for falseness
Link.
Further, network low-rank includes: from expression modeling
Wherein, E is error term, and X is the adjacency matrix of social networks, and Z is the representing matrix of social networks, rank () table
Show that rank of matrix, λ are balance parameters, | | | |2,1Representing matrix norm.
Further, the adjacency matrix is expressed as X ∈ Rn×m, i.e. adjacency matrix is the matrix that a m × n is tieed up, adjacent square
Each column of battle array are considered as a partial structurtes, and the partial structurtes of the i-th column indicate are as follows:
Wherein, X:,iIndicate the partial structurtes that adjacency matrix i-th arranges；D:,kIndicate partial structurtes X:,kBasic matrix, Zk,iIt is
Basic matrix D:,kWeight.
Further, there is a possibility that square by the link that the representing matrix of optimization and basic matrix obtain reconstruct social networks
Battle array SM include:
SM=XZ*+(XZ*)T；
Wherein, X is basic matrix, Z*For by the representing matrix of optimization, subscript T indicates transposed matrix.
Further, systematicness measurement representation are as follows:
Wherein, σrFor systematicness measurement, n is the dimension of representing matrix, and r is the order of representing matrix, and a is in representing matrix
The quantity of non-zero entry.
Further, the likelihood for calculating network link includes: that the likelihood of network link passes through the section of its head and end
The reconstruct likelihood of point is estimated, indicates are as follows:
Ui,j=RC (i) × RC (j)；
Wherein, Ui,jIndicate the likelihood of the link between node i and node j；RC (k) indicates the reconstruct of network node k seemingly
Right property, Zk,iIndicate basic matrix D:,kWeight, | | indicate absolute value, the dimension of n representing matrix.
The present invention can be by using the difference between reconstructed network and " true " social networks, i.e., remaining link, as
Test set is assessed, and one aspect of the present invention can excavate necessary data and be used to identify dummy account, on the other hand energy again
Reconstructed network is optimized, its validity and accuracy are promoted, time performance and complexity are also all optimized.
Detailed description of the invention
Fig. 1 is a kind of flow chart of abnormality detection and optimization method based on social networks regulatory mechanism of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The present invention proposes the abnormality detection and optimization method, system and computer equipment of a kind of social networks, the method
Comprising the following specific steps
S1, according to social networks construct diagram data, obtain the adjacency matrix of diagram data, using restructing algorithm to diagram data into
Row strengthens denoising, obtains reconstruct social networks；
S2, definition rule level are horizontal come the systematicness for assessing reconstruct social networks, calculate the likelihood of network link；
S3, in the social networks of reconstruct, be ranked up by likelihood of the ascending order to link, and obtain ranked list；
S4, ranked list is successively traversed since the first item of ranked list, if calculating social networks after removal currentitem
Systematicness it is horizontal, if current systematicness level increases, remove the corresponding link of this；
The social networks of S5, output after step S4 optimization.
Under data publication environment, the present invention collects the social network relationships number of needs from large amount of complex network first
According to then being pre-processed to data, identify sensitive relations, be abstracted as diagram data, obtain its adjacency matrix, and with this net
Network modeling quantifies its regularity, calculates reconstruct likelihood, then adjusts its reconfigurability, realizes the identification and again of dummy account
Network forming network is optimized, finally, by obtained optimization social networks in the form of network data again in issue back complex network
It goes.
In the present embodiment, individual or account are just abstracted as node, and account and account, interpersonal relationship are with regard to table
It is shown as link.Its corresponding adjacency matrix can easily be obtained by this diagram data very much.In order to reinforce social networks systematicness
And tactic pattern, convenient for abnormality detection and optimization, in the reconstruct of social networks, the present invention allows X ∈ Rn×mIndicate social networks
Adjacency matrix.Each column of matrix X are considered as a partial structurtes X:, i, therefore X includes m partial structurtes, that is, [X:,1,
X:,2,…,X:,m].Give a complete basic matrix D=[D:,1,D:,2,…,D:,m]∈Rn×m, each partial structurtes can be by
One group of linear combination of base is expressed as follows:
Wherein Zk,iCorresponding base D:,kWeight.Therefore, the adjacency matrix X of network can be reconstructed in a manner of X=DZ, wherein
Z∈Rn×mIt is representing matrix.In order to recognize the organisation of network, the optimal candidate of basic matrix D is adjacency matrix X.Based on above-mentioned
It discusses, social networks can indicate to model certainly with low-rank are as follows:
Wherein E is error term, and λ >=0 is the tradeoff parameter for balancing different item, and Z is representing matrix.
One common way is the nuclear norm with representing matrix | | Z | |*, i.e., the sum of singular value of rank (z) come generation
For the order (i.e. LRNR restructing algorithm) of Z.It will be appreciated, however, that the target of the network model proposed is to find institute by joint
Having the lowest class of data indicates to identify immanent structure mode；In order to solve social networks reconstruction, the present invention is focused on
Bottom-layer network is inferred in a manner of accurate, therefore, the present invention replaces nuclear norm with Frobenius norm (abbreviation F- norm),
LFNR restructing algorithm is just obtained, this can substantially reduce the computation complexity of network modelling, guarantee network reconfiguration high-precision
Under the premise of, it can efficiently reconstructed network；In order to solve this problem, present invention introduces auxiliary variable J to make objective function can
Separation can then be handled by solving following augmentation Lagrange's multiplier (ALM) problem:
Wherein, | | | |*Nuclear norm is sought in expression,Indicate F2 norm, | | | |2,1Representing matrix norm；Y1And Y2It is
Lagrange's multiplier, and μ > 0 is punishment parameter, the transposed matrix of subscript T representing matrix.Then J, Z and E are minimized respectively.It examines
Consider solution efficiency, the present invention has selected inaccurate ALM method to solve.The adjacency matrix of social networks and tradeoff parameter is defeated
Entering to this algorithm can be obtained by representing matrix and Error Matrix.Then optimal representing matrix Z solution obtained*With group moment
X is with SM=XZ for battle array*+(XZ*)TMode combine and can reconstruct the social networks link of " true " there are Likelihood matrixes
SM, probability existing for the element representation therein link, 0 indicates centainly to be not present, and 1 expression certainly exists, the table between 0 and 1
Show its existing probability, and negative then expression falseness link.In a sense, the effect of the matrix is equal to adjacency matrix,
Signify " true " social networks, that is, the social networks after reconstructing.After obtaining score matrix, according to entry symbol, it is by SM points
Positive component SM+With negative component SM-, by by positive component SM+In entry entry will be relatively deleted compared with adjacency matrix X, it is remaining
It is the link lacked that the likelihood of link, which is higher than the entry of threshold value, in entry, by by negative component SM-In entry and adjacent square
Battle array X is compared, to SM-The likelihood of entry is ranked up, and likelihood is false chain lower than the entry of threshold value in remaining entry
Road.
In order to which the systematicness for quantifying to reconstruct social networks is horizontal, the likelihood of its link is calculated, and identify dummy account, this
Invention is directly used in the effect analyzed the common point of partial structurtes and characterize network link from expression network model for what is proposed.
Wherein networking rule is intended to measure the degree that complex network can be modeled and reconstruct, and can actually pass through partial structurtes
Common point captures.Certain partial structurtes of network may be identical, their ratio can use representing matrix Z*Order come
It characterizes, identical partial structurtes are more in network, and representing matrix is fewer from the order in expression model.Even if not identical
Partial structurtes, some partial structurtes also may be constructed such that the combination of other structures.In this case, the canonical of partial structurtes
Property can be characterized with the quantity of the non-zero entry of representing matrix.Being then based on representing matrix, we define σrTo measure network
Systematicness:
Wherein n, r and a are the quantity of the dimension of representing matrix, sum of ranks non-zero entry respectively.(n-r)/n indicates phase in network
With the ratio of partial structurtes, a/ (nr) characterizes the regularity of partial structurtes.And for link likelihood, the of representing matrix
K row represents the contribution of the reconstruct of the partial structurtes of node k to other structures.The quantity of non-zero entry is more, the section for reconstruct
The frequency of the partial structurtes of point k is higher.Then, for network node k, it reconstructs likelihood are as follows:
The larger value of node reconstruct likelihood means that frequency of the relevant link of node for network reconfiguration is higher.So
The likelihood of network link can be estimated by the reconstruct likelihood of its endpoint node:
Ui,j=RC (i) × RC (j).
Wherein Ui,jIt is exactly node i, the likelihood of the link between j.Its biggish value is more likely regular link.Institute
There is unobservable link to be all ranked up according to their score, and the high link of score has and high there is a possibility that.
Similarly, all links observed all are sorted, and the lower link of score is more likely false link, and false link pair
What is answered is exactly dummy account, to realize the abnormality detection of social networks dummy account.
The present invention also proposes a kind of abnormality detection and optimization system based on social networks regulatory mechanism, including the system
Social networks including being sequentially connected obtains module, social networks conversion module, adjacency matrix generation unit, basic matrix and generates list
Member, social networks reconstructed module, false link judgment module and reconstructed network adjustment module；It is characterized by:
The social networks obtains the social networks and user information that module user obtains user in social networks；
The social networks conversion module user converts diagram data for the social networks of user in social networks and indicates；
The adjacency matrix generation unit is used to diagram data being converted to adjacency matrix；
The basic matrix generation unit user generates basic matrix according to adjacency matrix；Generating process is to will abut against matrix decomposition
For the product of basic matrix and basic matrix weight；
The social networks reconstructed module user carries out reinforcing denoising to diagram data, the social networks reconstructed；
The falseness link judgment module user calculates the likelihood of network link, and is sentenced by the likelihood of network link
Whether the network link that breaks is false link；
The reconstructed network adjustment module user deletes the false link in network link.
Further, the false link judgment module includes that low sequence is bright from expression network model subelement, augmentation glug
Day multiplier unit, entry symbol decision subelement, false link decision device；It is utilized according to low sequence from expression network model subelement
Augmentation Lagrange multiplier unit solves representing matrix and error matrix；It is social that reconstruct is obtained according to representing matrix and basic matrix
There are Likelihood matrix SM for the link of network, have used matrix multiplier and matrix adder in this step, and link is deposited
SM=XZ is expressed as in Likelihood matrix SM*+(XZ*)T；It with entry symbol is according to by chain in entry symbol decision subelement
Road is positive component SM there are Likelihood matrix SM points+With negative component SM-And false link decision device is inputted, it is adjudicated in false link
In device, by negative component SM-In entry scanned in adjacency matrix, if the entry is not present in adjacency matrix, should
The corresponding link of entry is falseness.
Further, the reconstructed network adjustment module includes systematicness measuring calculator, link likelihood calculator, row
Sequence subelement and subelement is deleted, link likelihood calculator calculates the likelihood of link, and sorting unit is by the likelihood liter of link
Sequence arranges to obtain ranked list；Using the element in systematicness measuring calculator successively calculations list, if after removing currentElement
The systematicness of reconstructed network, which is measured, to be risen, then using the corresponding link of subelement deletion currentElement is deleted, so that reconstructed network
Systematicness level rise.
A kind of abnormality detection based on social networks regulatory mechanism and optimizing computer equipment, including memory, processor
And the computer program that can be run on a memory and on a processor is stored, the computer program that the processor executes
Realize aforementioned any method.
The present invention can measure systematicness level using the systematicness measurement proposed.It is evident that social networks
Systematicness it is higher, predictability is higher, and reconfigurability is higher, and the precision of reconstruct is also higher.So can be by mentioning
The systematicness of high social networks changes tactic pattern relevant to target object, to optimize to reconstructed network.Although
The embodiment of the present invention has shown and described, for the ordinary skill in the art, it is possible to understand that do not departing from this hair
A variety of change, modification, replacement and modification, model of the invention can be carried out to these embodiments in the case where bright principle and spirit
It encloses and is defined by the appended claims and the equivalents thereof.
Claims (10)
1. the abnormality detection and optimization method of a kind of social networks, which comprises the following steps:
S1, diagram data is constructed according to social networks, obtains the adjacency matrix of diagram data, diagram data carried out using restructing algorithm strong
Change denoising, obtains reconstruct social networks；
S2, definition rule level are horizontal come the systematicness for assessing reconstruct social networks, calculate the likelihood of network link；
S3, in the social networks of reconstruct, be ranked up by likelihood of the ascending order to link, and obtain ranked list；
S4, ranked list is successively traversed since the first item of ranked list, if calculating the rule of social networks after removal currentitem
Then property is horizontal, if current systematicness level increases, removes the corresponding link of this；
The social networks of S5, output after step S4 optimization.
2. the abnormality detection and optimization method of a kind of social networks according to claim 1, which is characterized in that step S1 tool
Body includes:
It S11, is that side constructs a non-directed graph as diagram data using relationship of the user in social networks between node, user,
And diagram data is indicated with adjacency matrix；
S12, building low-rank obtain optimal representing matrix and error using augmentation Lagrange multiplier method from network model is indicated
Matrix；
S13, the link of reconstruct social networks is obtained by optimal representing matrix and basic matrix, and there are Likelihood matrix SM；
S14, matrix SM is divided into positive component SM with entry symbol+With negative component SM-；
If S15, positive component SM+In entry exist in adjacency matrix X, then exclude the entry, the likelihood of link in remaining entry
Property be higher than threshold value entry be lack link；
If S16, negative component SM-In entry do not have in adjacency matrix X, i.e., respective items are 0, then the corresponding link of the entry
For false link.
3. the abnormality detection and optimization method of a kind of social networks according to claim 2, which is characterized in that the network
Low-rank indicates to model certainly are as follows:
Wherein, E is error term；X is that the adjacency matrix adjacency matrix of social networks is expressed as X ∈ Rn×m, i.e. adjacency matrix is one
The matrix of m × n dimension, each column of adjacency matrix are considered as a partial structurtes, and the partial structurtes of the i-th column indicate are as follows:X:,iIndicate the partial structurtes that adjacency matrix i-th arranges；D:,kIt indicates
Partial structurtes X:,kBasic matrix, Zk,iIt is basic matrix D:,kWeight；Z is the representing matrix of social networks, and rank () indicates square
Rank of matrix, λ are balance parameters, | | | |2,1Representing matrix norm.
4. the abnormality detection and optimization method of a kind of social networks according to claim 2, which is characterized in that pass through optimization
Representing matrix and basic matrix obtain the link of reconstruct social networks there are Likelihood matrix SM to include:
SM=XZ*+(XZ*)T；
Wherein, X is basic matrix, Z*To pass through the representing matrix of optimization, the transposed matrix of subscript T representing matrix.
5. the abnormality detection and optimization method of a kind of social networks according to claim 1, which is characterized in that the rule
Property level is expressed as:
Wherein, σrIndicate that systematicness is horizontal, n is the dimension of representing matrix, and r is the order of representing matrix, and a is non-in representing matrix
The quantity of zero entry.
6. the abnormality detection and optimization method of a kind of social networks according to claim 1, which is characterized in that calculate network
The likelihood of link includes: that the likelihood of network link is estimated by the reconstruct likelihood of the node of its head and end, is indicated
Are as follows:
Ui,j=RC (i) × RC (j)；
Wherein, Ui,jIndicate the likelihood of the link between node i and node j；The reconstruct likelihood of RC (k) expression network node k
Property, Zk,iIndicate basic matrix D:,kWeight, | | indicate absolute value, the dimension of n representing matrix.
7. the abnormality detection and optimization system of a kind of social networks, the system comprises the social networks being sequentially connected to obtain mould
Block, social networks conversion module, adjacency matrix generation unit, basic matrix generation unit, social networks reconstructed module, false link
Judgment module and reconstructed network adjustment module；It is characterized by:
The social networks obtains the social networks and user information that module user obtains user in social networks；
The social networks conversion module user converts diagram data for the social networks of user in social networks and indicates；
The adjacency matrix generation unit is used to diagram data being converted to adjacency matrix；
The basic matrix generation unit user generates basic matrix according to adjacency matrix；
The social networks reconstructed module user carries out reinforcing denoising to diagram data, the social networks reconstructed；
The falseness link judgment module is used to judge whether the network link reconstructed in social networks to be false link；
The reconstructed network adjustment module is used to calculate the likelihood of link and systematicness measurement in reconstruct social networks, and is based on
The false link in network link is deleted in likelihood and the systematicness measurement of link.
8. the abnormality detection and optimization system of a kind of social networks according to claim 8, which is characterized in that the falseness
Link judgment module includes that low sequence is sub from expression network model subelement, augmentation Lagrange multiplier unit, entry symbol decision
Unit, false link decision device；It is solved from expression network model subelement using augmentation Lagrange multiplier unit according to low sequence
Representing matrix and error matrix out；It obtains there is a possibility that according to the link that representing matrix and basic matrix obtain reconstruct social networks
Matrix SM with entry symbol is according to by link, there are Likelihood matrix SM to divide for positive component in entry symbol decision subelement
SM+With negative component SM-And false link decision device is inputted, in false link decision device, by negative component SM-In entry in neighbour
It connects in matrix and scans for, if the entry is not present in adjacency matrix, the corresponding link of the entry is false link.
9. the abnormality detection and optimization system of a kind of social networks according to claim 8, which is characterized in that the reconstruct
Network adjustment module includes systematicness measuring calculator, link likelihood calculator, sorting subunit and deletes subelement, link
Likelihood calculator calculates the likelihood of link, and sorting unit arranges the likelihood ascending order of link to obtain ranked list；Utilize rule
The then property measuring calculator successively element in calculations list, rises if the systematicness for removing reconstructed network after currentElement is measured,
Then the corresponding link of currentElement is deleted using deletion subelement.
10. a kind of abnormality detection of social networks and optimizing computer equipment, which is characterized in that including memory, processor with
And the computer program that can be run on a memory and on a processor is stored, the computer program that the processor executes is real
Any method described in existing claim 1~6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910549606.3A CN110321493B (en) | 2019-06-24 | 2019-06-24 | Abnormity detection and optimization method and system of social network and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910549606.3A CN110321493B (en) | 2019-06-24 | 2019-06-24 | Abnormity detection and optimization method and system of social network and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110321493A true CN110321493A (en) | 2019-10-11 |
| CN110321493B CN110321493B (en) | 2021-09-28 |
Family
ID=68120150
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910549606.3A Active CN110321493B (en) | 2019-06-24 | 2019-06-24 | Abnormity detection and optimization method and system of social network and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110321493B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110768826A (en) * | 2019-10-18 | 2020-02-07 | 重庆邮电大学 | Power information network key information transmission link identification method and system |
| CN112073298A (en) * | 2020-08-26 | 2020-12-11 | 重庆理工大学 | Social network link abnormity prediction system integrating stacked generalization and cost sensitive learning |
| CN112256941A (en) * | 2020-10-20 | 2021-01-22 | 北京工商大学 | Social network reconstruction method based on local information |
| CN112445957A (en) * | 2020-11-05 | 2021-03-05 | 西安电子科技大学 | Social network abnormal user detection method, system, medium, equipment and terminal |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295907A (en) * | 2016-08-23 | 2017-01-04 | 浙江工业大学 | A kind of method that prediction network the unknown based on second order local community information connects limit |
| US20170019495A1 (en) * | 2014-03-10 | 2017-01-19 | Nokia Solutions And Networks Oy | Distribution of popular content between user nodes of a social network community via direct proximity-based communication |
| CN107977745A (en) * | 2017-11-30 | 2018-05-01 | 武汉大学 | One kind is used for community network node Abnormal Evolution behavioral value and appraisal procedure |
| CN109800573A (en) * | 2019-01-17 | 2019-05-24 | 西安电子科技大学 | Based on the anonymous social networks guard method with link disturbance of degree |
-
2019
- 2019-06-24 CN CN201910549606.3A patent/CN110321493B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170019495A1 (en) * | 2014-03-10 | 2017-01-19 | Nokia Solutions And Networks Oy | Distribution of popular content between user nodes of a social network community via direct proximity-based communication |
| CN106295907A (en) * | 2016-08-23 | 2017-01-04 | 浙江工业大学 | A kind of method that prediction network the unknown based on second order local community information connects limit |
| CN107977745A (en) * | 2017-11-30 | 2018-05-01 | 武汉大学 | One kind is used for community network node Abnormal Evolution behavioral value and appraisal procedure |
| CN109800573A (en) * | 2019-01-17 | 2019-05-24 | 西安电子科技大学 | Based on the anonymous social networks guard method with link disturbance of degree |
Non-Patent Citations (1)
| Title |
|---|
| 祖宝开: "分块低秩图的遥感影像半监督分类应用", 《计算机科学与探索》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110768826A (en) * | 2019-10-18 | 2020-02-07 | 重庆邮电大学 | Power information network key information transmission link identification method and system |
| CN110768826B (en) * | 2019-10-18 | 2022-06-03 | 重庆邮电大学 | Power information network key information transmission link identification method and system |
| CN112073298A (en) * | 2020-08-26 | 2020-12-11 | 重庆理工大学 | Social network link abnormity prediction system integrating stacked generalization and cost sensitive learning |
| CN112073298B (en) * | 2020-08-26 | 2021-08-17 | 重庆理工大学 | Social network link abnormity prediction system integrating stacked generalization and cost sensitive learning |
| CN112256941A (en) * | 2020-10-20 | 2021-01-22 | 北京工商大学 | Social network reconstruction method based on local information |
| CN112256941B (en) * | 2020-10-20 | 2023-10-13 | 北京工商大学 | Social network reconstruction method based on local information |
| CN112445957A (en) * | 2020-11-05 | 2021-03-05 | 西安电子科技大学 | Social network abnormal user detection method, system, medium, equipment and terminal |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110321493B (en) | 2021-09-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107480341B (en) | A kind of dam safety comprehensive method based on deep learning | |
| CN110321493A (en) | A kind of abnormality detection of social networks and optimization method, system and computer equipment | |
| CN110427654B (en) | Landslide prediction model construction method and system based on sensitive state | |
| Helbig et al. | Archive management for dynamic multi-objective optimisation problems using vector evaluated particle swarm optimisation | |
| CN110232434A (en) | A kind of neural network framework appraisal procedure based on attributed graph optimization | |
| Bai et al. | Entropic dynamic time warping kernels for co-evolving financial time series analysis | |
| Chen et al. | Exploiting structural and temporal evolution in dynamic link prediction | |
| Lahdelma et al. | Stochastic multicriteria acceptability analysis (SMAA) | |
| Rufo et al. | Log-linear pool to combine prior distributions: A suggestion for a calibration-based approach | |
| Ma et al. | Decomposition-based multiobjective evolutionary algorithm for community detection in dynamic social networks | |
| CN114297036A (en) | Data processing method and device, electronic equipment and readable storage medium | |
| CN109686402A (en) | Based on key protein matter recognition methods in dynamic weighting interactive network | |
| CN107045659A (en) | Predict the method and device of photovoltaic generation power | |
| Wright et al. | Neural network architecture selection analysis with application to cryptography location | |
| Rodrigues et al. | A study of generalization and fitness landscapes for neuroevolution | |
| Wang et al. | Multivariate ordinal categorical process control based on log-linear modeling | |
| Ramasamy et al. | Application of artificial neural network for software reliability growth modeling with testing effort | |
| CN114648235A (en) | Industrial data quality quantitative analysis report generation method based on data image | |
| CN117078114B (en) | Water quality evaluation method and system for water-bearing lakes under influence of diversion engineering | |
| CN116842459B (en) | Electric energy metering fault diagnosis method and diagnosis terminal based on small sample learning | |
| CN109977131A (en) | A kind of house type matching system | |
| CN108830407B (en) | Sensor distribution optimization method in structure health monitoring under multi-working condition | |
| Li et al. | Kennard-Stone balance algorithm for time-series big data stream mining | |
| CN114578087B (en) | Wind speed uncertainty measurement method based on non-dominant sorting and stochastic simulation algorithm | |
| Pattnaik et al. | A survey on machine learning techniques used for software quality prediction |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |