CN103617279A - Method for achieving microblog information spreading influence assessment model on basis of Pagerank method - Google Patents

Abstract

The invention discloses a method for assessing microblog information spreading influence on the basis of a Pagerank method. According to the method for assessing the microblog information spreading influence, data analysis is performed on large-scale increment microblog information, and a comprehensive assessment model for direct influence and indirect influence of microblog information spreading is provided by studying the spreading characteristics of the microblog information. The method comprises the steps of first performing crawling of real microblog network information data needed in an experiment from a network; obtaining individual activeness by standardized measurement according to the number of original microblogs averagely issued by an individual at the position of some node within some time period, the number of forwarded microblogs and the number of microblogs participating in comments; shrinking quantity difference of user fan quantities, and working out the concern degree of nodes according to the browsing number, the comment number, the forwarding number and the praising number of the microblogs; finally enabling the microblog spreading capability to serve as a transmission factor affecting an influence value, and constructing the microblog spreading influence model on the basis of the Pagerank method.

Description

The implementation method of the micro-blog information propagation effect force estimation model based on Pagerank method

Technical field

The present invention relates to data mining technology field, particularly a kind of implementation method of the micro-blog information propagation effect force estimation model based on Pagerank method.

Background technology

Microblogging is one of current internet internet exchange platform the most prevailing, and it is with convenience, strong interactivity, and the features such as instantaneity have caused huge impact to conventional information communications media.Microblogging has become novel public topic as a kind of channel based on individual freedom expression speech and social interactions and has propagated platform.It is revolutionary innovation in a kind of Information Communication pattern, has greatly changed the personal expression mode of popular script, becomes gradually one of the most important self-expression of people, obtaining information and social mode.

Analyze the feature in microblogging transmission distribution, and the accurate rule of finding wherein, excavate valuable microblogging and user thereof, for understand user's dissemination microeffect and social effect, understand the inherent law that much-talked-about topic forms and instruct spin to have important value.Micro-blog information propagation effect masterpiece is an emerging research topic, and also there is attention rate widely in academia at home and abroad.

Not only in ， law circle of academia, also investigate this and take the impact on the leading directions of news of network environment that microblogging is representative.China Supreme People's Court, Supreme Procuratorate clearly stipulate in handling the explanation > > that utilizes information network to implement the some problems of criminal case governing law such as calumny at < <, utilize information network to calumniate other people, same calumny information reality is clicked, number of visits reaches more than 5000 times, or is forwarded number of times and reaches 500 times and above will be regarded as forming the crime of defamation.

How better to weigh the influence power of the micro-blog information of propagating in network, find out a N with strongest influence power node, make the node that is affected in final social networks maximum, Information Communication scope is maximum, rather than the size of its pageview of single tolerance and transfer amount.And the present invention can solve problem above well.

Summary of the invention

The object of the invention has been to design a kind of propagation effect power recognition methods based on micro-blog information, the method is in the extensive enterprising line number of increment micro-blog information according to one's analysis, by the propagation characteristic of research micro-blog information, the direct influence of micro-blog information propagation and the Integrated Evaluation Model of indirect influence are proposed.

The technical solution adopted for the present invention to solve the technical problems is: a kind of implementation method of the micro-blog information propagation effect force estimation model based on Pagerank method, the method is weighed the weight of propagating node by mark micro-blog information influence power and the importance of node, the weights of node are larger, and influence power is just larger.When calculating the weight of each node, the authority (whether being authenticated as large V by official) of node will be considered, bean vermicelli quantity and quality thereof that node has, and the many factors such as the interactive relation of node and linking relationship.The method comprises the following steps:

Step 1, write microblogging data reptile program, from network, crawl out real microblogging network information data, therefrom extract the network topology structure information that micro-blog information is propagated.

The definition of step 2, individual liveness

The definition of individual liveness L comprises three aspects, and it comprises, the original microblogging quantity O of average issue every day of (as one month) this individuality in section sometime, and the microblogging quantity F of forwarding, participates in the microblogging quantity C of comment.Its quantizating index is as follows:

L=O * W _o+ F * W _f+ C * W _cformula (1)

Wherein, W _ofor the shared liveness weight of original microblogging model, W _f, W _cbe respectively forwarding microblogging, evaluate the shared liveness weight of microblogging.By these three kinds of Measure Indexes are standardized, bring again formula into and calculate, can prevent preferably that the problem that weight difference is excessive from appearring in attribute large and less initial value.The concrete mode that standardization and parameter are determined is shown in embodiment.

Step 3, in order to get rid of in microblog users the interference to customer impact degree of a large amount of " corpse beans vermicelli " that exist, and dwindle the order of magnitude difference of user's bean vermicelli quantity, bean vermicelli number user less but that influence power is stronger is not left in the basket, the present invention defines w (p, q) and represents the weight of user q to user p.Through type (2) calculates the value of w (p, q):

$w(p,q)=\sqrt{L_{\mathrm{pq}}\&times;f\left(q\right)}$ Formula (2)

Wherein, if user q is authenticated, make L _pq>=1.L _pqspecifically by formula (3), obtained:

formula (3)

F (q) represents the bean vermicelli number of user q.

Step 4, the concerned degree of node

If the attention rate that the microblogging of this Nodes is subject to is higher, influence power is stronger, considers following three indexs for this reason: this microblogging browse several S, comment number P, forwards number M and point and praises several Z.

Therefore, the concerned degree of node is as the formula (4):

C=S * W _s+ P * W _p+ M * W _m+ Z * W _zformula (4)

Wherein, W _s, W _p, W _m, W _zrepresent to browse respectively number, comment number, forwards number, and point is praised several weight ratios.

Step 5, microblogging propagation effect power

By individual liveness, the bean vermicelli after two indexs of the concerned degree of node and normalization is counted three's integrated definition and is gone out microblogging propagation effect power index.Expression formula is:

$R=L\&times;C\&times;\underset{q\&Element;E}{\mathrm{\&Sigma;}}w(p,q)$ Formula (5)

The bean vermicelli set that wherein E is individual p.

Step 6, the microblogging propagation effect power model based on Pagerank

Because the PR value of the page in Pagerank method is average being delivered on each page that chain picks out, thereby ignored the difference of the page self importance.Therefore,, when the influence power of the present invention when Twitter message is transmitted analyzed, in order more comprehensively to evaluate Different Individual influence power, the present invention adds individual behavioural characteristic in method and thought to as Consideration.Expression formula is:

$\mathrm{VPR}\left(p\right)=d-(1-d)\underset{q\&Element;E}{\mathrm{\&Sigma;}}R(p,q)\mathrm{VPR}\left(q\right)$ Formula (6)

Wherein d is the ratio of damping between 0 to 1, for guaranteeing last numerical convergence, the bean vermicelli set that E is individual p, R (p, q), for individual q distributes to the ratio of the VPR value of p, the size that is taken all good friend's propagation effect power sums of family q by the microblogging propagation effect power of individual p determines.Suppose that q has N good friend, q distributes to the VPR value ratio of p and is so:

$R(p,q)=\frac{R_p}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{S}_i}$ Formula (7)

Through after iteration repeatedly, the VPR value in network can be tending towards convergence, can obtain the VPR value of all nodes in microblogging communication network, is worth larger propagation effect power larger.

Beneficial effect:

1, from authenticity, the angle of concerned property and activity is set out, and respectively to individual liveness, the concerned degree of node is analyzed respectively, and the method for dwindling the order of magnitude difference of user's bean vermicelli quantity reduces rubbish user's interference, comprehensively draws microblogging propagation effect power.

2, on the basis of Pagerank method, modify, consider the influence power degree of node itself, be delivered on the node that chain goes out the PR value of each node is pro rata, considered fully the characteristic of node self, more accurately objectively draw influence power value.

Accompanying drawing explanation

Fig. 1 is method flow diagram of the present invention.

Embodiment

Below in conjunction with Figure of description, the invention is described in further detail.

As shown in Figure 1, the invention provides a kind of implementation method of the micro-blog information propagation effect force estimation model based on Pagerank method, the method comprises the steps:

Step 1, write microblogging data reptile program, from network, crawl out real microblogging network information data, therefrom extract the network topology structure information that micro-blog information is propagated.It mainly comprises the authority (whether being large V, whether by the personal authentication of Sina) of Nodes microblogging bloger individuality; Bean vermicelli quantity and quality thereof that node has; The interactive relation of node and linking relationship, for example average every day issuing microblog quantity, and in original microblogging, forward the ratio of microblogging, and Nodes microblogging browse number, comment number, forwards number and point and praises number.

Step 2, individual liveness

The definition of individuality liveness A comprises three aspects, and it comprises, the original microblogging quantity V that in section t, (as one month) this Nodes individuality is on average issued every day sometime _o, the microblogging quantity V of forwarding _f, participate in the microblogging quantity V commenting on _c.Its quantizating index is as follows:

A=V _o* W _o+ V _f* W _f+ V _c* W _cformula (1)

Wherein, W _ofor the shared liveness weight of original microblogging model, W _f, W _cbe respectively forwarding microblogging, evaluate the shared liveness weight of microblogging.By these three kinds of Measure Indexes are standardized, bring again formula into and calculate, can prevent preferably that the problem that weight difference is excessive from appearring in attribute large and less initial value.Concrete method for normalizing is:

${V_O}^{\&prime;}=\frac{V_O-{\mathrm{Min}}_O}{{\mathrm{Max}}_O-{\mathrm{Min}}_O}({{\mathrm{Max}}_O}^{\&prime;}-{{\mathrm{Min}}_O}^{\&prime;})+{{\mathrm{Min}}_O}^{\&prime;}$ Formula (2)

${V_F}^{\&prime;}=\frac{V_F-{\mathrm{Min}}_F}{{\mathrm{Max}}_F-{\mathrm{Min}}_F}({{\mathrm{Max}}_F}^{\&prime;}-{{\mathrm{Min}}_F}^{\&prime;})+{{\mathrm{Min}}_F}^{\&prime;}$ Formula (3)

${V_C}^{\&prime;}=\frac{V_C-{\mathrm{Min}}_C}{{\mathrm{Max}}_C-{\mathrm{Min}}_C}({{\mathrm{Max}}_C}^{\&prime;}-{{\mathrm{Min}}_C}^{\&prime;})+{{\mathrm{Min}}_C}^{\&prime;}$ Formula (4)

Max wherein _o, Min _obe illustrated respectively in the time period of collection maximal value and the minimum value of the original microblogging quantity of average issue; Max _f, Min _frepresent respectively average maximal value and the minimum value that forwards microblogging quantity in time interval; Max _c, Min _crepresent respectively maximal value and the minimum value of comment.In the process of calculating, for the result of above-mentioned formula is normalized in interval [0,1], make Max _o'=1, Min _o'=0, Max _f'=1, Min _f'=0, Max _c'=1, Min _c'=0.Therefore after formula (1) standard, obtain the expression-form of formula (5):

A=V _o' * W _o+ V _f' * W _f+ V _c' * W _cformula (5)

Step 3, in order to get rid of in microblog users the interference to customer impact degree of a large amount of " corpse beans vermicelli " that exist, and dwindle the order of magnitude difference of user's bean vermicelli quantity, bean vermicelli number user less but that influence power is stronger is not left in the basket, the present invention defines w (p, q) and represents the weight of user q to user p.Through type (6) calculates the value of w (p, q):

$w(p,q)=\sqrt{L_{\mathrm{pq}}\&times;f\left(q\right)}$ Formula (6)

Wherein, if user q is authenticated, make L _pq>=1.L _pqspecifically by formula (7), obtained:

formula (7)

F (q) represents the bean vermicelli number of user q.

Step 4, the concerned degree of node

If the attention rate that the microblogging of this Nodes is subject to is higher, influence power is stronger, considers following three indexs for this reason: this microblogging browse several V _s, comment number V _p, forward number V _mand point is praised several V _z.

Therefore, the concerned degree of node is as the formula (8):

C=V _s* W _s+ V _p* W _p+ V _m* W _m+ V _z* W _zformula (8)

Wherein, W _s, W _p, W _m, W _zrepresent to browse respectively number, comment number, forwards number, and point is praised several weight ratios.

Step 5, microblogging propagation effect power evaluation model

By individual liveness, the bean vermicelli after two indexs of the concerned degree of node and normalization is counted three's integrated definition and is gone out microblogging propagation effect power index.Expression formula is:

$R=A\&times;C\&times;\underset{q\&Element;E}{\mathrm{\&Sigma;}}w(p,q)$ Formula (9)

The bean vermicelli set that wherein E is individual p.

Step 6, the microblogging propagation effect power model based on Pagerank

Pagerank method is for weighing the classical way of each node significance level of network in webpage, its core concept is according to the backward chaining quantity of each node, the value of each node to be had influence on each node equably, and the PR value of each node is average comprehensive of its all of its neighbor node contribution.In procedure, Pagerank not only considers the poll that web page joint obtains, also can include the importance that participates in the adjacent node of ballot in reference, the ticket that launch those the important websites often value of a ticket of the website more general than influence power is high, and this is just in time similar to the influence power factor of propagating node in microblogging communication process.

But, because the PR value of the page in Pagerank method is average being delivered on each page that chain picks out, thereby ignored the difference of the page self importance.Therefore,, when the influence power of the present invention when Twitter message is transmitted analyzed, in order more comprehensively to evaluate Different Individual influence power, the present invention adds individual behavioural characteristic in method and thought to as Consideration.

The basic thought of Pagerank method is the transmission factor using microblogging transmission capacity defined above as assignment affects power value, the user that microblogging transmission capacity is high can obtain higher influence power value, the influence power value that the user that corresponding microblogging transmission capacity is low obtains is lower, with regard to having solved influence power value, be the problem of evenly transmitting like this, overcome annexation between simple dependence user and carried out the deficiency of rank, made model can reflect more objectively actual conditions.Expression formula is:

$\mathrm{VPR}\left(p\right)=d-(1-d)\underset{q\&Element;E}{\mathrm{\&Sigma;}}R(p,q)\mathrm{VPR}\left(q\right)$ Formula (10)

Wherein d is the ratio of damping between 0 to 1, for guaranteeing last numerical convergence, the bean vermicelli set that E is individual p, R (p, q), for individual q distributes to the ratio of the VPR value of p, the size that is taken all good friend's propagation effect power sums of family q by the microblogging propagation effect power of individual p determines.Suppose that q has N good friend, q distributes to the VPR value ratio of p and is so:

$R(p,q)=\frac{R_p}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{S}_i}$ Formula (11)

Through after iteration repeatedly, the VPR value in network can be tending towards convergence, can obtain the VPR value of all nodes in microblogging communication network, is worth larger propagation effect power larger.

Claims (6)

1. the implementation method of the micro-blog information propagation effect force estimation model based on Pagerank method, is characterized in that, described method comprises the steps:

Step 1: write web crawlers;

1. write microblogging data reptile program, from network, crawl out real microblogging network information data;

2. therefrom extract the network topology structure information that micro-blog information is propagated;

Step 2: the definition of individual liveness;

The definition of individual liveness L comprises three aspects, and it comprises;

1. the original microblogging quantity O of (as one month) this individuality average every day of issue in section sometime;

2. the microblogging quantity F forwarding;

3. participate in the microblogging quantity C of comment;

Step 3: in order to get rid of the interference of a large amount of " corpse beans vermicelli " that exist to customer impact degree in microblog users, and dwindle the order of magnitude difference of user's bean vermicelli quantity, bean vermicelli number user less but that influence power is stronger is not left in the basket, and the present invention has redefined user q the weight of user p has been represented;

Step 4: the concerned degree of node;

If the attention rate that the microblogging of certain Nodes is subject to is higher, influence power is stronger, comprises following three indexs: this microblogging browse several S, comment number P, forwards number M and point and praises several Z;

Therefore, the concerned degree of node is shown below:

C＝S×W _S+P×W _P+M×W _M+Z×W _Z

Wherein, W _s, W _p, W _m, W _zrepresent to browse respectively number, comment number, forwards number, and point is praised several weight ratios;

Step 5: microblogging propagation effect power;

By individual liveness, the bean vermicelli after two indexs of the concerned degree of node and normalization is counted three's integrated definition and is gone out microblogging propagation effect power index, and expression formula is:

$R=L\&times;C\&times;\underset{q\&Element;E}{\mathrm{\&Sigma;}}w(p,q)$

The bean vermicelli set that wherein E is individual p;

Step 6: the microblogging propagation effect power model based on Pagerank;

In Pagerank method, the PR value of the page is average being delivered on each page that chain picks out, and when the influence power when Twitter message is transmitted is analyzed, is that the individual behavioural characteristic of microblogging platform is added in the method as Consideration.

2. the implementation method of a kind of micro-blog information propagation effect force estimation model based on Pagerank method according to claim 1, is characterized in that: in the step 2 of described method, described individual liveness is added to quantizating index, its quantizating index is as follows:

A＝V _O×W _O+V _F×W _F+V _C×W _C

Wherein, W _ofor the shared liveness weight of original microblogging model, W _f, W _cbe respectively forwarding microblogging, evaluate the shared liveness weight of microblogging, by these three kinds of Measure Indexes are standardized, bring again formula into and calculate.

3. the implementation method of a kind of micro-blog information propagation effect force estimation model based on Pagerank method according to claim 1, is characterized in that: in the step 2 of described method, concrete method for normalizing is:

${V_O}^{\&prime;}=\frac{V_O-{\mathrm{Min}}_O}{{\mathrm{Max}}_O-{\mathrm{Min}}_O}({{\mathrm{Max}}_O}^{\&prime;}-{{\mathrm{Min}}_O}^{\&prime;})+{{\mathrm{Min}}_O}^{\&prime;}$

${V_F}^{\&prime;}=\frac{V_F-{\mathrm{Min}}_F}{{\mathrm{Max}}_F-{\mathrm{Min}}_F}({{\mathrm{Max}}_F}^{\&prime;}-{{\mathrm{Min}}_F}^{\&prime;})+{{\mathrm{Min}}_F}^{\&prime;}$

${V_C}^{\&prime;}=\frac{V_C-{\mathrm{Min}}_C}{{\mathrm{Max}}_C-{\mathrm{Min}}_C}({{\mathrm{Max}}_C}^{\&prime;}-{{\mathrm{Min}}_C}^{\&prime;})+{{\mathrm{Min}}_C}^{\&prime;}$

Therefore the expression-form obtaining after formula (1) standard is:

A＝V _O'×W _O+V _F'×W _F+V _C'×W _C

。

4. the implementation method of a kind of micro-blog information propagation effect force estimation model based on Pagerank method according to claim 1, it is characterized in that: in the step 3 of described method, the present invention defines w (p, q) represent the weight of user q to user p, by following formula, calculate the value of w (p, q):

$w(p,q)=\sqrt{L_{\mathrm{pq}}\&times;f\left(q\right)}.$

5. the implementation method of a kind of micro-blog information propagation effect force estimation model based on Pagerank method according to claim 1, it is characterized in that: in the step 6 of described method, the basic thought of Pagerank method is the transmission factor using microblogging transmission capacity defined above as assignment affects power value, the user that microblogging transmission capacity is high can obtain higher influence power value, the influence power value that the user that corresponding microblogging transmission capacity is low obtains is lower, and its expression formula is:

$\mathrm{VPR}\left(p\right)=d-(1-d)\underset{q\&Element;E}{\mathrm{\&Sigma;}}R(p,q)\mathrm{VPR}\left(q\right).$

6. the implementation method of a kind of micro-blog information propagation effect force estimation model based on Pagerank method according to claim 1, it is characterized in that: in the step 6 of described method, the size that is taken all good friend's propagation effect power sums of family q by the microblogging propagation effect power of individual p determines, suppose that q has N good friend, q distributes to the VPR value ratio of p and is so:

$R(p,q)=\frac{R_p}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{S}_i}$

Through after iteration repeatedly, the VPR value in network can be tending towards convergence, can obtain the VPR value of all nodes in microblogging communication network, is worth larger propagation effect power larger.

Images