CN109146700A - A kind of influence power feature extracting method for social networks leader - Google Patents
A kind of influence power feature extracting method for social networks leader Download PDFInfo
- Publication number
- CN109146700A CN109146700A CN201810923109.0A CN201810923109A CN109146700A CN 109146700 A CN109146700 A CN 109146700A CN 201810923109 A CN201810923109 A CN 201810923109A CN 109146700 A CN109146700 A CN 109146700A
- Authority
- CN
- China
- Prior art keywords
- factor
- coefficient
- indicate
- user
- behavior
- 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
- 238000000034 method Methods 0.000 title claims abstract description 29
- 210000003813 thumb Anatomy 0.000 claims abstract description 33
- 230000003993 interaction Effects 0.000 claims abstract description 22
- 239000000284 extract Substances 0.000 claims abstract description 10
- 238000000605 extraction Methods 0.000 claims abstract description 10
- 230000009471 action Effects 0.000 claims description 13
- 230000001186 cumulative effect Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000005484 gravity Effects 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract description 3
- 238000000691 measurement method Methods 0.000 abstract description 3
- 238000013139 quantization Methods 0.000 abstract description 3
- 230000006399 behavior Effects 0.000 description 69
- 230000000694 effects Effects 0.000 description 7
- 230000008901 benefit Effects 0.000 description 3
- 239000010410 layer Substances 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000003012 network analysis Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000004750 isotope dilution mass spectroscopy Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Telephonic Communication Services (AREA)
Abstract
The invention discloses a kind of influence power feature extracting methods for social networks leader, the present invention passes through the quantization to behavior factor, extract the impact factor coefficient of seven dimensional characteristics, the factor is acted including reply, thumb up the movement factor, concern acts the factor, forward behavior factor, quote behavior factor, dispatch and refers to the movement factor at behavior factor, it substantially avoids impact factor in the prior art and extracts incomplete problem, by getting close to coefficient, attention coefficient, propagation coefficient calculates super side attribute with interaction liveness coefficient, it avoids instruction and calculates the incomplete problem of Consideration brought by propagation information and user's itself affect, provide the measurement and evaluation method of more comprehensively complete user force, and opinion leader is differentiated with this, accuracy with higher network leader's influence power feature extraction.
Description
Technical field
The invention belongs to social network data digging technology fields, and in particular to a kind of influence for social networks leader
Power feature extracting method.
Background technique
Currently, having become current ten shunting similar to the microblogging social networks of twitter with the development of network technology
A kind of capable social networks interactive system, user is all over the world, and with its high speed development, netizen's height participates in content and produces
Raw topic discussion, any social event are likely to fast propagation amplification, and network becomes the main carriers of reflection social public opinion
One of, it plays an important role in public sentiment propagation.If be not good to network public-opinion guidance, negative public sentiment will be to the public of society
Safety forms biggish threat.Opinion leader is the center leader of opinion crowd in human communication network, it can effectively be represented pair
Public sentiment topic guiding has leading effect, and leader of opinion plays an important role in communication effect, it means that leader of opinion
The information of propagation is more more valuable than other stages on source.Therefore, when carrying out public sentiment monitoring, in preparatory discovery crowd
Leader, and it is targetedly monitored, the accurate of public sentiment monitoring will be improved, improve its working efficiency.
The method that domestic and foreign scholars study social networks leader influence power feature at present has following several:
(1) it according to topic publisher, reply person and the reply relationship between them in social platform, establishes and corresponds to them
Online recommendation network, using improved influence power diffusion model IDMS calculate based on content of text excavation, posting person's characteristic,
The influence power ranking of the opinion leader of money order receipt to be signed and returned to the sender person characteristic and mutually reply relationship, is finally filtered out using scientific and effective method
Opinion leader in social platform, to establish the model of opinion leader in identification social platform.This method is to be with netizen
Node, the reply relationship constituted between netizen are that side constructs community network, do not consider how external information generates shadow to netizen
The problems such as sound, how the psychological driving force that netizen issues speech acts on, therefore it is inadequate for being solely focused on reply behavior.
(2) the network public opinion supernetwork model that social activity, environment, psychology and viewpoint four-layer network describe is established, it will be existing right
The single layer social network analysis that network public opinion carries out is extended to super-network research, on the basis of the model of super-network, proposes one
The new super side sort algorithm of kind, is ranked up by the super side that the algorithm participates in being formed to each public opinion main body in social network,
And then excavate network public opinion leader.Design grid only considers between the public opinion main body participated in discussion in network public opinion in this method
Reply relationship is discussed, and discusses not comprehensive enough, lacks the discussion for the movement factor such as referring to thumbing up.Environment net only considered
Information issues the influence to public opinion in information transmitting, lacks the discussion to behavior factors such as forwarding, references;Super side attribute meter
When calculation, only consider that information propagation effect degree, the psychology conversion degree of association and viewpoint similarity degree come into calculating.Only from information
Transmitting is set out with the influence factor of user itself, is accounted for the relationship between user, such as: user's is closed
Note degree interacts and gets close to coefficient etc. between liveness and user.
Summary of the invention
For above-mentioned deficiency in the prior art, the influence power feature extraction provided by the invention for social networks leader
Method solve existing social networks public opinion Effetiveness factor extract not comprehensively, the super side attribute calculating of social networks it is unspecific
Problem.
In order to achieve the above object of the invention, a kind of the technical solution adopted by the present invention are as follows: shadow for social networks leader
Ring power feature extracting method, comprising the following steps:
S1, in social networks, according to the behavior act of user, determine the sequence of operation of user;
S2, according to the sequence of operation, determine the user time penalty factor for issuing action behavior;
All historical operations of S3, counting user, according to determining time penalty factor, all behaviour for the historical operation that adds up
Make as a result, determining cumulative effect coefficient;
S4, according to user time penalty factor and cumulative effect coefficient, quantify simultaneously to determine the action behavior in social networks
Impact factor;
S5, established respectively according to action behavior impact factor get close to coefficient net, attention coefficient net, propagation coefficient net and
Interact liveness net, and calculate separately it is corresponding get close to coefficient, attention coefficient, propagation coefficient and interaction liveness coefficient;
S6, basis get close to coefficient, are concerned coefficient, propagation coefficient and interaction liveness coefficient, establish influence power super-network
Model;
S7, according to influence power supernetwork model, extract the influence power feature of social networks leader.
Further, in the step S1, the information in the sequence of operation includes the user for issuing behavior act, the use
Family operates specific behavior act, the user executes the object user being associated in operating process and user carries out the tool of the operation
The body time;
In the step S4, the action behavior impact factor include reply movement the factor, forward behavior factor, thumb up it is dynamic
Make the factor, reference behavior factor, refer to the movement factor, the concern movement factor and dispatch behavior factor.
Further, in the step S2, the user time penalty factor is calculated in the following manner:
Wherein, fp(T) time penalty factor of the user's operation in T time difference is indicated;
V indicates the original influence coefficient value for not considering time punishment;
G indicates the gravity factor.
Further, in the step S3, the cumulative effect coefficient is obtained by following calculation:
Wherein, SijIt indicates in t-t0Time difference in, the cumulative effect force coefficient under the operation from user i to j;
VtijIndicate that i is to the number of the same operation of j in certain day;
VtIndicate that user i agrees to the number operated in certain day.
Further, in the step S5,
It is described to get close in coefficient net:
The factor is acted by concern, refer to the movement factor and thumbs up movement factor building and gets close to coefficient net;
It is described to get close to coefficient are as follows:
In formula, JijPay close attention to the movement factor;
mijIndicate that i refers to the movement factor to j;
mjiIndicate that j refers to the movement factor to i;
lijIndicate that i thumbs up the movement factor to j;
ljiIndicate that j thumbs up the movement factor to i;
In the attention coefficient net:
The factor is acted by the concern, thumbs up the movement factor, forwarding behavior factor and dispatch behavior factor building quilt
Attention rate net;
The attention coefficient are as follows:
In formula, Ji,jPay close attention to the movement factor;
piIndicate dispatch behavior factor;
rtjiIndicate j to the forwarding behavior factor of i;
lijIndicate that i thumbs up the movement factor to j;
ljiIndicate that j thumbs up the movement factor to i;
In the propagation coefficient net:
The factor, forwarding behavior factor, dispatch behavior factor and reference behavior factor building is acted by the reply to propagate
Coefficient net;
The propagation coefficient are as follows:
In formula, rtjiIndicate that j is the forwarding behavior factor of i;
rpjiIndicate j to the forwarding behavior factor of i;
qjiIndicate that j acts the factor to the reply of i;
piIndicate dispatch behavior factor;
In the interaction liveness net:
By referring to the movement factor, thumbing up the movement factor, the reply movement factor, reference behavior factor and dispatch behavior factor
Building interaction liveness net;
The interaction liveness coefficient are as follows:
In formula, li,jIndicate that i thumbs up the movement factor to j;
mijIndicate that i refers to the movement factor to j;
rpijIndicate that i acts the factor to the reply of j;
qijIndicate i to the reference behavior factor of j;
pjIndicate dispatch behavior factor.
Further, in the step S6, the influence power supernetwork model are as follows:
Wherein, HyperRank (ui) indicate user node uiSuper side influence power Rank value;
FociIndicate the attention total value of user i;
N is number of nodes total in influence power super-network;
M(ui) it is all user node uiThe set for other nodes being linked to;
L(ui) indicate user node uiHinged node number;
FociIndicate the attention coefficient of user i;
ClsijIt presses the flesh coefficient;
Spri,jIndicate propagation coefficient;
ActijIndicate i to the liveness coefficient of j;
σ (x) is activation primitive;
ujIndicate user node uj。
Further, in the step S7, the method for the influence power feature of extraction social networks leader specifically:
Threshold value threshold is set, when super side influence power Rank value of user's node in influence power super-network is more than threshold
When value threshold, using user's node as the key event in influence power super-network, and saved as the crowd of network leader
Point set completes the extraction of the influence power feature of social networks leader.
The invention has the benefit that the present invention extracts the shadow of seven dimensional characteristics by the quantization to behavior factor
Ring factor coefficient, including reply acts the factor, thumbs up the movement factor, concern acts the factor, forward behavior factor, reference behavior because
Son, hair push away literary behavior factor, refer to the movement factor, substantially avoid impact factor in the prior art and extract incomplete problem,
Super side attribute is calculated by getting close to coefficient, attention coefficient, propagation coefficient and interaction liveness coefficient, avoids indicating gage
It calculates and propagates the incomplete problem of Consideration brought by information and user's itself affect, provide more comprehensively complete user's shadow
The measurement and evaluation method of power are rung, and opinion leader is differentiated with this, there is higher network leader influence power feature extraction
Accuracy.
Detailed description of the invention
Fig. 1 is the influence power feature extracting method implementation flow chart of social networks leader in embodiment provided by the invention.
Fig. 2 is that influence power super-network evaluates ranking process flow diagram in embodiment provided by the invention.
Fig. 3 is that 4 layers of influence power in embodiment provided by the invention under proximities, attention, propagation degree, liveness are super
Network model figure.
Fig. 4 is that embodiment partial user influence power super-network provided by the invention gets close to property coefficient side visual presentation
Figure.
Specific embodiment
A specific embodiment of the invention is described below, in order to facilitate understanding by those skilled in the art this hair
It is bright, it should be apparent that the present invention is not limited to the ranges of specific embodiment, for those skilled in the art,
As long as various change is in the spirit and scope of the present invention that the attached claims limit and determine, these variations are aobvious and easy
See, all are using the innovation and creation of present inventive concept in the column of protection.
As shown in Figure 1, a kind of influence power feature extracting method for social networks leader, which is characterized in that including with
Lower step:
S1, in social networks, according to the behavior act of user, determine the sequence of operation of user;
In social networks, according to natural person's behavior act, the sequence of operation of a set of user behavior is defined, each user's
Action behavior can be made of the sequence of one group of same format, the sequence of operation are as follows:
Uid, method, target, datetime
Uid is the user for issuing behavior act;
Method is the specific behavior act of the user's operation;
Target is that the user executes the object user being associated in operating process;
Datetime is the specific time that user carries out the operation;
S2, according to the sequence of operation, determine the user time penalty factor for issuing action behavior;
Consider time passage, consider nearest connection and operation as far as possible between two users, defines time penalty factor and adopt
With the time penalty factor algorithm in Hacker News, therefore in above-mentioned steps S2, the user time penalty factor are as follows:
Wherein, fp(T) time penalty factor of the user's operation in T time difference is indicated;
V indicates the original influence coefficient value for not considering time punishment;
G indicates the gravity factor, indicates constantly to drop in impression as time go on, according in Hacker News, G value
It is defaulted as 1.2;
All historical operations of S3, counting user, according to determining time penalty factor, all behaviour for the historical operation that adds up
Make as a result, determining cumulative effect coefficient;
Every kind of operation is all the set of one group of multiple time series, thus since considering history all operations of the user it is tired
Product influences to accumulate all operating results as a result, so on the basis of time penalty factor, define the accumulation of aforesaid operations sequence
Influencing coefficient is
Wherein, SijIt indicates in t-t0Time difference in, the cumulative effect force coefficient under the operation from user i to j;
VtijIndicate that i is to the number of the same operation of j in certain day;
VtIndicate that user i agrees to the number operated in certain day.
S4, according to user time penalty factor and cumulative effect coefficient, quantify simultaneously to determine the action behavior in social networks
Impact factor;
Pass through total sequence of operation, it is easy to all perpetual object set of certain user are obtained, is defined as:
Fi={ Pit0, Pit1, Pit2... Pit, t ∈ (t0, t0+T)
Wherein, FiIndicate all perpetual objects of user i;
PitPerpetual object caused by the operation P for being i in time t, in perpetual object on face, it may be considered that by two
User as the coefficient in concern dimension, therefore uses Jaccard Index in the gas similitude that respectively perpetual object collection closes
Outstanding person blocks German number to measure the similitude between two set, the similitude are as follows:
Thus 7 dimensional characteristics are obtained and are quantified as impact factor coefficient respectively, above-mentioned action behavior impact factor includes back
Factor rp is made in double actionij, forwarding behavior factor rtij, thumb up movement factor lij, reference behavior factor qij, refer to movement factor mij、
Concern acts factor JijWith dispatch behavior factor pi;
S5, established respectively according to action behavior impact factor get close to coefficient net, attention coefficient net, propagation coefficient net and
Interact liveness net, and calculate separately it is corresponding get close to coefficient, attention coefficient, propagation coefficient and interaction liveness coefficient;
It establishes above-mentioned when getting close to coefficient net:
Coefficient is got close to indicate the correlation degree of the build-in attributes such as interpersonal relationships between user, intimate behavior, original 7
It pays close attention to, thumb up in a dimension and referring to three behaviors, being the familarity generated after more being got close between user, therefore closing
Infuse, thumb up and refer to the proximities that these three factors can preferably reflect between user.So concern acts the factor, refers to
It the movement factor and thumbs up the movement factor and constructs jointly and get close to coefficient net;
Coefficient net is got close to according to foundation, coefficient is got close in definition are as follows:
In formula, JijPay close attention to the movement factor;
mijIndicate that i refers to the movement factor to j;
mjiIndicate that j refers to the movement factor to i;
lijIndicate that i thumbs up the movement factor to j;
ljiIndicate that j thumbs up the movement factor to i;
Wherein, if the behavioral difference that thumbs up mutually of both sides is larger, it is not especially to lie in that showing in fact, which has a side,
Other side, coefficient should be got close in the case of this comprising both sides by, which calculating, certain punishment decaying;Conversely, when both sides thumb up mutually degree
It is very high, and almost behavior never has special height difference, then shows that both sides are very close, and certain increasing should be given on getting close to coefficient
Benefit;It is all identical to get close to weight mutual between any two node in coefficient net, therefore, gets close to coefficient net and is considered as non-directed graph
Net.
When establishing above-mentioned attention rate coefficient net:
In the degree for indicating to be concerned about other side between user in social networks, what attention considered is whether both sides value
Or be concerned about and pay attention to another party, whether this usually shows close personal circle, if is concerned about other side's thing of concern, if
The message for paying attention to other side's hair is same as above the decaying and gain performance that thumb up behavior.So concern acts the factor, thumbs up movement
The factor, forwarding behavior factor and dispatch behavior factor have constructed attention coefficient net jointly.
According to above-mentioned attention coefficient net, attention coefficient is defined as:
In formula, JijPay close attention to the movement factor;
piIndicate dispatch behavior factor;
rtjiIndicate j to the forwarding behavior factor of i;
lijIndicate that i thumbs up the movement factor to j;
ljiIndicate that j thumbs up the movement factor to i;
When establishing above-mentioned propagation coefficient net:
In social network, propagation coefficient is defined to indicate that each user propagates other side's message each other in social subnet
Diffusion a, it is contemplated that whether user and his audience in social networks generate front to his information and promote and ring
Answer, audience reply, forwarding, reference original text the case where it is more positive, show that the transmission capacity of the user in this regard is stronger.
So the reply movement factor, forwarding behavior factor, dispatch behavior factor and reference behavior factor have constructed propagation system jointly
Number net.
According to the propagation coefficient net of above-mentioned building, propagation coefficient is defined as:
In formula, rtjiIndicate that j is the forwarding behavior factor of i;
rpjiIndicate that j acts the factor to the reply of i;
qjiIndicate j to the reference behavior factor of i;
piIndicate dispatch behavior factor;
When constructing above-mentioned interaction liveness net:
The interaction degree of various aspects message, foundation judge the mathematical modulo whether user participates in discussion between any two users
Type, still, when the participation between user and surrounding the same group is higher, messaging interaction is more active, more can be shown that this person is strong
It is strong to belong in personal circle representated by this group, both sides can be substantially judged by the behaviors such as replying, thumbing up, refer to, quote
Whether keep in touch, whether current interaction is actively and active.So dispatch behavior factor, refer to movement the factor, thumb up movement because
Son replys behavior factor and quotes behavior factor and constructed interaction liveness net jointly.
According to the interaction liveness net of above-mentioned building, liveness coefficient is interacted is defined as:
In formula, li,jIndicate that i thumbs up the movement factor to j;
mijIndicate that i refers to the movement factor to j;
rpijIndicate that i acts the factor to the reply of j;
qijIndicate i to the reference behavior factor of j;
pjIndicate dispatch behavior factor.
S6, basis get close to coefficient, are concerned coefficient, propagation coefficient and interaction liveness coefficient, establish influence power super-network
Model;
As shown in Fig. 2, the information propagation degree in super-network, the Research foundation of psychological transfer correlation degree, viewpoint similarity,
And to social activity, content, topic, viewpoint on the Research foundation of super-network analysis application, to get close to coefficient net, attention
Net, propagation coefficient net and interaction liveness net are characterized subnet, and super side sort algorithm is to iteration in improvement influence power super-network
Formula finally obtains supernetwork model by verifying and amendment are as follows:
Wherein, HyperRank (ui) indicate user node uiSuper side influence power Rank value;
FociIndicate the attention total value of user i, i.e. the sum of the attention coefficient value of i and all association users;
N is number of nodes total in influence power super-network;
M(ui) it is all user node uiThe set for other nodes being linked to;
L(ui) indicate user node uiHinged node number;
FociIndicate the attention coefficient of user i;
ClsijIt presses the flesh coefficient;
Spri,jIndicate propagation coefficient;
ActijIndicate i to the liveness coefficient of j;
σ (x) is activation primitive;As the clip functions to attention rate, by export-restriction to [0,1) in, usually using σ
(x)=tanh (x).
ujIndicate user node uj。
As shown in figure 3, illustrating 4 layers of influence power supernetwork model under proximities, attention, propagation degree, liveness;
The initial value of a minimum value and super-network node has been given in above-mentioned supernetwork model to the node of no any linkGuarantee subsequent calculating to be to be not in that zero calculates.Propagation coefficient Spr acts herein as resistance coefficient, and meaning is,
At any time, message reaches after certain user node and will continue to the probability spread backward, in the case where 1-Spr, Xia Yilian
The user being connected to will abandon spreading news.Degree of getting close to Cls between two users is higher, should get during next iteration
More weights.And user uiTo user ujInteraction liveness may be at the relationship of opposition, but interact active at direction trend
When, it should include certain attenuation relation, conversely, certain gain effect suitably can be provided to influence power.
According to Twitter user's propagation characteristic, when user can touch the information of user, super phase while super with other
Associated probability is bigger, in close nature sub-network it is super while the intimate factor surpass closer to another while interacting activity, and anticipate
Justice is closer, and the super of user indicates bigger to the similitude of path profile when indicating and another is super, the super side of user indicate and
Similitude between other super side expressions is bigger, and meaning similar intensity.The representative on the super side of user is to pushing away literary path and other
Super side representative is bigger to path distribution similarity, and the weight of the super side and other super frontier junctures connection acquisitions is bigger.
S7, according to influence power supernetwork model, extract the influence power feature of social networks leader.
In above-mentioned steps S7, the method for the influence power feature of extraction social networks leader specifically:
Threshold value threshold is set, when super side influence power Rank value of user's node in influence power super-network is more than threshold
When value threshold, using user's node as the key event in influence power super-network, and saved as the crowd of network leader
Point set completes the extraction of the influence power feature of social networks leader.As shown in figure 4, illustrating the super net of certain customers' influence power
Network gets close to property coefficient.
The invention has the benefit that the present invention extracts the shadow of seven dimensional characteristics by the quantization to behavior factor
Ring factor coefficient, including reply acts the factor, thumbs up the movement factor, concern acts the factor, forward behavior factor, reference behavior because
Son, behavior factor of sending the documents, refers to the movement factor, substantially avoids impact factor in the prior art and extracts incomplete problem, leads to
It crosses and gets close to coefficient, attention coefficient, propagation coefficient and interaction liveness coefficient to calculate super side attribute, avoid instruction and calculate
The incomplete problem of Consideration brought by information and user's itself affect is propagated, more comprehensively complete customer impact is provided
The measurement and evaluation method of power, and opinion leader is differentiated with this, with higher network leader influence power feature extraction
Accuracy.
Claims (7)
1. a kind of influence power feature extracting method for social networks leader, which comprises the following steps:
S1, in social networks, according to the behavior act of user, determine the sequence of operation of user;
S2, according to the sequence of operation, determine the user time penalty factor for issuing action behavior;
All historical operations of S3, counting user, according to determining time penalty factor, all operation knots for the historical operation that adds up
Fruit determines cumulative effect coefficient;
S4, according to user time penalty factor and cumulative effect coefficient, quantify simultaneously to determine that the action behavior in social networks influences
The factor;
S5, it is established respectively according to action behavior impact factor and gets close to coefficient net, attention coefficient net, propagation coefficient net and interaction
Liveness net, and calculate separately it is corresponding get close to coefficient, attention coefficient, propagation coefficient and interaction liveness coefficient;
S6, basis get close to coefficient, are concerned coefficient, propagation coefficient and interaction liveness coefficient, establish influence power supernetwork model;
S7, according to influence power supernetwork model, extract the influence power feature of social networks leader.
2. the influence power feature extracting method according to claim 1 for social networks leader, which is characterized in that described
In step S1, the information in the sequence of operation include issue the user of behavior act, the specific behavior act of the user's operation,
The user executes the object user being associated in operating process and user carries out the specific time of the operation;
In the step S4, the action behavior impact factor include reply movement the factor, forward behavior factor, thumb up movement because
Son, reference behavior factor refer to the movement factor, the concern movement factor and dispatch behavior factor.
3. the influence power feature extracting method according to claim 2 for social networks leader, which is characterized in that described
In step S2, the user time penalty factor is calculated in the following manner:
Wherein, fp(T) time penalty factor of the user's operation in T time difference is indicated;
V indicates the original influence coefficient value for not considering time punishment;
G indicates the gravity factor.
4. the influence power feature extracting method according to claim 3 for social networks leader, which is characterized in that described
In step S3, the cumulative effect coefficient is calculated in the following manner:
Wherein, SijIt indicates in t-t0Time difference in, the cumulative effect force coefficient under the operation from user i to j;
VtijIt indicates in certain day to the number of the same operation of j;
VtIndicate that user i agrees to the number operated in certain day.
5. the influence power feature extracting method according to claim 2 for social networks leader, which is characterized in that described
In step S5,
It is described to get close in coefficient net:
The factor is acted by concern, refer to the movement factor and thumbs up movement factor building and gets close to coefficient net;
It is described to get close to coefficient are as follows:
In formula, Ji,jPay close attention to the movement factor;
mijIndicate that i refers to the movement factor to j;
mjiIndicate that j refers to the movement factor to i;
lijIndicate that i thumbs up the movement factor to j;
ljiIndicate that j thumbs up the movement factor to i;
In the attention coefficient net:
By the concern act the factor, thumb up movement the factor, forwarding behavior factor and dispatch behavior factor building be concerned
Spend net;
The attention coefficient are as follows:
In formula, JijPay close attention to the movement factor;
piIndicate dispatch behavior factor;
rtjiIndicate j to the forwarding behavior factor of i;
lijIndicate that i thumbs up the movement factor to j;
ljiIndicate that j thumbs up the movement factor to i;
In the propagation coefficient net:
The factor, forwarding behavior factor, dispatch behavior factor and reference behavior factor, which are acted, by the reply constructs propagation coefficient
Net;
The propagation coefficient are as follows:
In formula, rtjiIndicate that j is the forwarding behavior factor of i;
rpjiIndicate j to the forwarding behavior factor of i;
qjiIndicate that j acts the factor to the reply of i;
piIndicate dispatch behavior factor;
In the interaction liveness net:
By referring to the movement factor, thumbing up the movement factor, the reply movement factor, reference behavior factor and dispatch behavior factor building
Interact liveness net;
The interaction liveness coefficient are as follows:
In formula, li,jIndicate that i thumbs up the movement factor to j;
mijIndicate that i refers to the movement factor to j;
rpijIndicate that i acts the factor to the reply of j;
qijIndicate i to the reference behavior factor of j;
pjIndicate dispatch behavior factor.
6. the influence power feature extracting method according to claim 5 for social networks leader, which is characterized in that described
In step S6, the influence power supernetwork model are as follows:
Wherein, HyperRank (ui) indicate user node uiSuper side influence power Rank value;
FociIndicate the attention total value of user i;
N is number of nodes total in influence power super-network;
M(ui) it is all user node uiThe set for other nodes being linked to;
L(ui) indicate user node uiHinged node number;
FociIndicate the attention coefficient of user i;
ClsijIt presses the flesh coefficient;
Spri,jIndicate propagation coefficient;
ActijIndicate i to the liveness coefficient of j;
σ (x) is activation primitive;
ujIndicate user node uj。
7. the influence power feature extracting method according to claim 6 for social networks leader, which is characterized in that described
In step S7, the method for the influence power feature of extraction social networks leader specifically:
Threshold value threshold is set, when super side influence power Rank value of user's node in influence power super-network is more than threshold value
When threshold, using user's node as the key event in influence power super-network, and crowd's node as network leader
Set completes the extraction of the influence power feature of social networks leader.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810923109.0A CN109146700B (en) | 2018-08-14 | 2018-08-14 | Method for extracting influence characteristics of social network leader |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810923109.0A CN109146700B (en) | 2018-08-14 | 2018-08-14 | Method for extracting influence characteristics of social network leader |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109146700A true CN109146700A (en) | 2019-01-04 |
CN109146700B CN109146700B (en) | 2021-11-02 |
Family
ID=64792964
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810923109.0A Active CN109146700B (en) | 2018-08-14 | 2018-08-14 | Method for extracting influence characteristics of social network leader |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109146700B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112508726A (en) * | 2020-12-25 | 2021-03-16 | 东北电力大学 | False public opinion identification system based on information spreading characteristics and processing method thereof |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102662956A (en) * | 2012-03-05 | 2012-09-12 | 西北工业大学 | Method for identifying opinion leaders in social network based on topic link behaviors of users |
CN103150333A (en) * | 2013-01-26 | 2013-06-12 | 安徽博约信息科技有限责任公司 | Opinion leader identification method in microblog media |
US20140108184A1 (en) * | 2012-10-16 | 2014-04-17 | Rimedio, Inc. | Transaction-driven social network |
CN103886105A (en) * | 2014-04-11 | 2014-06-25 | 北京工业大学 | User influence analysis method based on social network user behaviors |
CN104331817A (en) * | 2014-10-29 | 2015-02-04 | 深圳先进技术研究院 | User feature extraction method and system of e-commerce recommendation model |
CN106127590A (en) * | 2016-06-21 | 2016-11-16 | 重庆邮电大学 | A kind of information Situation Awareness based on node power of influence and propagation management and control model |
CN106570763A (en) * | 2016-11-09 | 2017-04-19 | 福建中金在线信息科技有限公司 | User influence evaluation method and system |
CN106682208A (en) * | 2016-12-30 | 2017-05-17 | 桂林电子科技大学 | Prediction method of micro-blog forwarding behavior based on fusion feature selection and random forest |
CN107633260A (en) * | 2017-08-23 | 2018-01-26 | 上海师范大学 | A kind of social network opinion leader method for digging based on cluster |
CN107909496A (en) * | 2017-07-28 | 2018-04-13 | 北京百分点信息科技有限公司 | User influence in social network analysis method, device and electronic equipment |
-
2018
- 2018-08-14 CN CN201810923109.0A patent/CN109146700B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102662956A (en) * | 2012-03-05 | 2012-09-12 | 西北工业大学 | Method for identifying opinion leaders in social network based on topic link behaviors of users |
US20140108184A1 (en) * | 2012-10-16 | 2014-04-17 | Rimedio, Inc. | Transaction-driven social network |
CN103150333A (en) * | 2013-01-26 | 2013-06-12 | 安徽博约信息科技有限责任公司 | Opinion leader identification method in microblog media |
CN103886105A (en) * | 2014-04-11 | 2014-06-25 | 北京工业大学 | User influence analysis method based on social network user behaviors |
CN104331817A (en) * | 2014-10-29 | 2015-02-04 | 深圳先进技术研究院 | User feature extraction method and system of e-commerce recommendation model |
CN106127590A (en) * | 2016-06-21 | 2016-11-16 | 重庆邮电大学 | A kind of information Situation Awareness based on node power of influence and propagation management and control model |
CN106570763A (en) * | 2016-11-09 | 2017-04-19 | 福建中金在线信息科技有限公司 | User influence evaluation method and system |
CN106682208A (en) * | 2016-12-30 | 2017-05-17 | 桂林电子科技大学 | Prediction method of micro-blog forwarding behavior based on fusion feature selection and random forest |
CN107909496A (en) * | 2017-07-28 | 2018-04-13 | 北京百分点信息科技有限公司 | User influence in social network analysis method, device and electronic equipment |
CN107633260A (en) * | 2017-08-23 | 2018-01-26 | 上海师范大学 | A kind of social network opinion leader method for digging based on cluster |
Non-Patent Citations (3)
Title |
---|
YI-CHENG CHEN等: "Mining Opinion Leaders in Big Social Network", 《2017 IEEE 31ST INTERNATIONAL CONFERENCE ON ADVANCED INFORMATION NETWORKING AND APPLICATIONS (AINA)》 * |
张琛等: "社交网络用户影响力的模糊综合评价", 《计算机系统应用》 * |
陈志雄等: "基于情感倾向性分析的微博意见领袖识别模型", 《计算机科学》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112508726A (en) * | 2020-12-25 | 2021-03-16 | 东北电力大学 | False public opinion identification system based on information spreading characteristics and processing method thereof |
CN112508726B (en) * | 2020-12-25 | 2022-04-19 | 东北电力大学 | False public opinion identification system based on information spreading characteristics and processing method thereof |
Also Published As
Publication number | Publication date |
---|---|
CN109146700B (en) | 2021-11-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Pinna et al. | A petri nets model for blockchain analysis | |
Adali et al. | Measuring behavioral trust in social networks | |
Zhang et al. | Dynamics of information diffusion and its applications on complex networks | |
CN104239385B (en) | For inferring the method and system of the relation between theme | |
CN103279887B (en) | A kind of microblogging based on information theory propagates visual analysis method | |
CN103179198B (en) | Based on the topic influence individual method for digging of many relational networks | |
CN104134159B (en) | A kind of method that spread scope is maximized based on stochastic model information of forecasting | |
Ienco et al. | The meme ranking problem: Maximizing microblogging virality | |
CN103077172B (en) | A kind of method and apparatus for excavating cheating user | |
CN106682770A (en) | Friend circle-based dynamic microblog forwarding behavior prediction system and method | |
CN110825948B (en) | Rumor propagation control method based on rumor-splitting message and representation learning | |
CN103064917A (en) | Specific-tendency high-influence user group discovering method orienting microblog | |
Chen et al. | Influencerank: An efficient social influence measurement for millions of users in microblog | |
CN105956184A (en) | Method for identifying collaborative and organized junk information release team in micro-blog social network | |
Zhu et al. | Information dissemination model for social media with constant updates | |
CN103457799B (en) | Microblog zombie user detection method based on graph of a relation | |
CN110807556A (en) | Method and device for predicting propagation trend of microblog rumors or/and dagger rumors | |
CN109213953A (en) | A kind of modeling method of social networks multi information propagation model | |
Liu et al. | Improving fraud detection via hierarchical attention-based graph neural network | |
CN109146700A (en) | A kind of influence power feature extracting method for social networks leader | |
CN110851684B (en) | Social topic influence recognition method and device based on ternary association graph | |
Guo et al. | Influence maximization algorithm based on group trust and local topology structure | |
Lu et al. | Pricing game of celebrities in sponsored viral marketing in online social networks with a greedy advertising platform | |
CN102750288B (en) | A kind of internet content recommend method and device | |
De Choudhury et al. | Dynamic prediction of communication flow using social context |
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 |