CN104008163A - Trust based social network maximum influence node calculation method - Google Patents

Abstract

The invention particularly discloses a trust based social network maximum influence node calculation method and relates to the technical field of social networking. The trust based social network maximum influence node calculation method comprises calculating the trust degree between nodes due to analysis of transaction conditions between the nodes, calculating the influence value between the nodes according to the distance between the nodes, synthesizing the trust degree and the influence value of the nodes, obtaining the node comprehensive influence, synthesizing the node comprehensive influence and the number of nodes which can be activated to find out the potential node having the largest influence and adding the selected node into node concentration; selecting the node having the largest influence; enabling the selected node to be used for developing a social network. The trust based social network maximum influence node calculation method has practical application prospects in aspects such as the internet network safety due to being relates to aspects such as direct trust, indirect trust, direct influence, indirect influence, comprehensive influence and social network influence between the users.

Description

Maximize the computing method that affect node based on the community network of trusting

Technical field

The present invention relates to social networks technical field, in particular, is to maximize based on the community network of trusting the computing method that affect node.

Background technology

In recent years, along with the universal and develop rapidly of Internet, a large amount of community networks are shown one's talent, as the Friendster based on love affair, and the LinkedIn based on work relationship etc.Community network is by the complex network forming because of the interactive relation forming between Social Individual and individuality.The community network of this complexity, the propagation to information and diffusion play vital effect, and community network is a double-edged sword simultaneously.Utilize on the one hand the interchange that community network is convenient for people to and diffuse information, such as, community network has huge effect at the real idea fermentation of reflection broad masses.Meanwhile, under the supervision of a few thing that makes government by community network in the people, play a part positive for building a sound social environment.On the other hand, utilize community network to make the harmful thing of the country and people with chance also to some lawless persons.Therefore, in the time utilizing social networks, how to accomplish to maximize favourable factors and minimize unfavourable ones, effectively supervision, is also the subject matter of community network development now.Huge in view of the number of members of some large-scale community networks, and relation is each other more complicated, finally can be on people's behavior generation impact.Their appearance maximizes and has brought huge challenge to the society of legacy network.There is the research of many related fields.Traditional greedy algorithm of naturally climbing the mountain proposing as Kemple and Kleinberg, each step is all chosen the current influential node of tool, but too high for large-scale social networks complexity, and does not consider the degree of belief problem between node, and impracticable.The people such as farming family's hall have proposed a kind of new-type hybrid having the greatest impact algorithm based on linear threshold, this algorithm is divided into two stages: heuristics phase is chosen the node of potential impact maximum, the greed stage is chosen the node having the greatest impact, but do not consider the different different problems of node activation threshold, equally the degree of belief between node is not considered yet.The community network based on threshold value that the people such as Chen Hao propose maximizes impact, the problem that different node activation thresholds is different is improved, but do not consider equally the importance of the degree of belief between node, ignored the impact that its maximization on community network produces.

As previously mentioned, the social maximization problems research of current social network, is all generally based on linear threshold model and independent cascade model.For the research of having the greatest impact of community network, mainly concentrate under independent cascade model and utilize time module feature to reduce in the complexity of greedy algorithm at present.But all just consider the impact between node, the degree of belief between node had not been considered.Due to the limitation of considering, cause the imperfection of research method, the community network that makes finally to choose maximize affect the obtained coverage of set of node can be relatively little.

Summary of the invention

Object of the present invention, for the disappearance existing in prior art and deficiency, proposes to maximize based on the community network of trusting the computing method that affect node.

The present invention is the research of the relevant issues of carrying out under social networks, because social networks is an abstract and complicated object, analyzes for convenient, has introduced the method for graph theory, by the social networks imagery of abstract.And introduce and trust this element, consider to trust the material impact to activating node, by the degree of belief between node and influence value combination, draw combined influence value, then be worth having the greatest impact of community network node according to combined influence.

The present invention, proposes a kind of heuritic approach (as shown in Figure 1) based on trusting between node based on linear threshold model, its basic procedure:

First, calculate the degree of belief between node by the transaction situation between analysis node, calculate internodal influence value according to the distance of distance between node, degree of belief to node and influence value carry out comprehensively, obtain the combined influence of egress, the combined influence of node and the current nodes that can activate of this node are carried out comprehensively, finding out the node of potential impact maximum, the node of choosing is joined in set of node.

Secondly, choose the node having the greatest impact.

Finally, by the set of node of choosing for developing social networks.

The community network that the present invention is based on trust maximizes the computing method that affect node, and concrete steps are:

A. defining social networks and social networks maximizes and affects the discovery of node, chooses

B. heuristics phase: (1), if un-activation node v is the out-degree node that has activated node u, v is direct trust to the trust of u, calculates directly and trusts; If v is reached at the node of u, v is indirect trust to the trust of u, calculates indirectly and trusts.(2) if un-activation node v is the out-degree node that has activated node u, u is direct impact on the impact of v, calculates directly impact; If v is reached at the node of u, u is remote effect on the impact of v, calculates remote effect.(3) influence value calculating in conjunction with above-mentioned two situations and trust value, the combined influence of computing node u.(4) the node combined influence value drawing according to above step and can activate interstitial content, the community network that calculates node affects.(5) node that the community network of heuristics phase being chosen has the greatest impact joins in having the greatest impact of community network node set.

C. the greed stage: from remaining node, choose the current node that can activate at most node that has, join in having the greatest impact of community network node set.

D. the generation of having the greatest impact of community network node set.

Wherein, the detailed process of described steps A is:

Described social networks turns to two orderly tuple <V of a figure G with form, E>,

|V|＝n,V＝{v ₁,v ₂,......v _n},|E|＝m,E＝{e ₁,e ₂,e ₃......e _m}

Each user in social networks is used as to a node, is expressed as:

G＝(V,E)

(1) V is a non-NULL finite set, and the set being made up of the user node in social networks, is called vertex set;

(2) E represents the finite multiple subset of the ordered set V & V of the relation between user in social networks, and the set that element repeats, is called limit collection, referred to as limit.

Described node is the major part that forms social networks, regards the each user in social networks as a node, is divided into according to the state of node:

(1) activate node, in social networks, if a certain user node has been bought a certain commodity or accepted a certain idea, be called the node of activation about these commodity or idea.

(2) un-activation node, in social networks, if a certain user node is not bought a certain commodity or do not accepted a certain idea, is called the un-activation node about these commodity or idea.

Mutual relationship between described node is divided into:

(1) out-degree node, in G=(V, E), the node of limit using u as initial point terminal pointed, is called the out-degree node of node u.

(2) in-degree node, in G=(V, E), the node of the corresponding initial point in limit using u as terminal, is called the in-degree node of node u.

(3) can reach node, in G=(V, E), u ₁can arrive u by specific path ₂, and u ₂not u ₁out-degree node, claim u ₂u ₁reached at node.

(4) unreachable node, in G=(V, E), u ₁do not exist path to arrive u ₂, claim u ₂u ₁unreachable node.

Described social networks maximizes affect node discovery, finds in social networks the influential member of tool.

Described k social networks maximizes affects choosing of node, makes the number finally buying or diffuse information maximum, and introduces heuristic factor c, chooses in heuristics phase individual having the greatest impact of community network node, the greed stage is chosen individual having the greatest impact of community network node.

Wherein, the detailed process of described step B is:

(1) if un-activation node v is the out-degree node that has activated node u, the direct trust that v produces u, its computing method are:

${\mathrm{DT}}_{\mathrm{uv}}=\frac{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_s^k\&times;W_s^k+\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_c^k\&times;W_c^k}{N}+\mathrm{RTN}\left(u\right)---\left(1\right)$

Wherein, $\mathrm{RTN}\left(u\right)=\frac{\underset{w\&Element;N\left(u\right)}{\mathrm{\&Sigma;}}\mathrm{RTN}(u\&RightArrow;w)}{\mathrm{out}\left(u\right)}$

DT _uvfor the trust of v to u.Wherein, O _srepresent the recommendation success of u to v, be illustrated in and recommend for the k time successfully, while recommending successfully, O _sbe 1, recommend failed O _sbe 0.

W _sweights while representing to recommend successfully, be illustrated in the weights while recommending for the k time successfully;

O _crepresent the weights of evaluating, and O _c∈ [0,1], represent the weights of the evaluation providing while recommendation for the k time, $\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{W}_s^i+\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{W}_c^i=1.$

RTN (u) is the prestige of node u.N (u) is the set of the out-degree node of u, RTN (u → w) be u go out the trust value of limit neighbours w to it, out (u) represents the out-degree of u, i.e. the trust value RTN (u) of u depends on its trust of all limit neighbours of going out to it.

For each node, a given initial trust value.

If v is reached at the node of u, v produces indirectly and trusts u, and its computing method are:

${\mathrm{IT}}_{\mathrm{uv}}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{1}{2^{h_i}}\&times;\mathrm{RTN}\left(u\right),h_i\&Element;H\left[n\right]---\left(2\right)$

IT _uvthe trust of v to u.U is n bar to the path of v, and traversal u is to each paths of v, h _ibe u to the level at v place on the i path of v, be equivalent to the height of v on binary tree, array H[n] in deposit the level of each u to v place on the path of v.

(2) if un-activation node v is the out-degree node that has activated node u, u has a direct impact v, and its computing method are:

${\mathrm{DI}}_{\mathrm{uv}}=\frac{\mathrm{\&lambda;}\&times;W(u,v)\&times;\mathrm{out}\left(u\right)}{\underset{q\&Element;O\left(v\right)}{\mathrm{\&Sigma;}}\frac{1}{\mathrm{out}\left(q\right)}}---\left(3\right)$

Wherein, DI _uvfor the direct impact that u produces v, λ is decay factor, and W (u, v) is the weight on limit (u, v), and out (u) is the out-degree of node u, and set O (v) is the set that goes out limit neighbours of node v.

If v is reached at the node of u, u produces remote effect to v, and its computing method are:

${\mathrm{IE}}_{\mathrm{uv}}=\underset{q_i=\mathrm{PA}\left[i\right]\left[n\right],q_{i_{\mathrm{active}}=0}}{\overset{n}{\mathrm{\&Sigma;}}}\left(\underset{p_i=\mathrm{PA}\left[i\right]\left[j\right]}{\overset{n}{\mathrm{\&Pi;}}}\frac{\mathrm{\&lambda;}\&times;w}{\mathrm{out}\left(p_i\right)}\right)---\left(4\right)$

IE _uvfor the remote effect of u to v generation.Wherein, PA[n] [n] be u to the set of the node of process in all paths of v, for each road is through distributing a two-dimensional array.PA[i] storage is j node in i paths in [j]; PA[i] [n] be every paths i, total n paths, every u is to the path effects sum of v; represent that i paths did not travel through; W is the weights that connect the limit of two adjacent nodes in a paths; λ represents the reduction of weights effect; Out (p _i) expression node p _iout-degree.

(3) suppose to activate node u, un-activation node is v, and u can arrive v by specific path, introduces multi-stress θ, θ ∈ [0,1].

If 1. un-activation node v is the out-degree node that has activated node u, the computing method of the combined influence of u to v are:

CE _uv＝θDI _uv+(1-θ)DT _uv (5)

If 2. v is reached at the node of u, the computing method of the combined influence of u to v are:

CE _uv＝θIE _uv+(1-θ)IT _uv (6)

(4) heuristics phase is chosen social network influence to maximize the standard of node is combined influence and the current nodes that activates of node of node.By both combinations, can find more accurately community network to maximize affects node, and joins in the set of having the greatest impact of community network node.The node that heuristics phase is selected, although not necessarily have maximum coverage, is but containing huge potential impact.

Social network influence node is calculated as follows:

${\mathrm{MISN}}_u=\frac{\underset{r\&Element;R(u\&RightArrow;v),r_{\mathrm{active}}=0}{\mathrm{\&Sigma;}}{\mathrm{CE}}_{\mathrm{uv}}+N}{\mathrm{out}\left(u\right)}---\left(7\right)$

Wherein, MISN _uit is the community network impact of u; R (u → v) represent that u can arrive v through specific path; R represents a certain paths of u to v; r _active=0 represents not live through this paths; N is the current nodes being activated by u; Out (u) is the out-degree of node u.

(5) according to the community network impact of the node having calculated, therefrom choose the node that community network has the greatest impact, join in having the greatest impact of community network node set.

Wherein, the detailed process of described step C is:

The greed stage is exactly from remaining node, chooses the current node that can activate at most node that has, and joins in having the greatest impact of community network node set.

Wherein, the detailed process of described step D is:

The generation of having the greatest impact of community network node set.

As mentioned above, compared with prior art, what the present invention proposed maximizes based on the community network of trusting the computing method that affect node, and can choose more accurately community network maximization affects node, thereby realizes as much as possible having the greatest impact of community network.The present invention, from direct trust, trusts indirectly, directly impact, and remote effect, combined influence, the aspects such as community network impact, realize choosing of having the greatest impact of community network node.

Brief description of the drawings

Fig. 1 maximizes based on the community network of trusting the FB(flow block) that affects node computing method in the present invention;

Fig. 2 is that the community network that the present invention is based on trust maximizes the network node figure of an example that affects node computing method.

Embodiment

Below in conjunction with drawings and Examples, the present invention is further described

First the present invention maximizes and affects the discovery of node and define social networks and social networks:

Define 1 social networks.A social networks can form turn to a figure G, and figure G is two orderly tuple <V, E>, wherein,

(1) V is a non-NULL finite set, and the set being made up of the user node in social networks, is called vertex set.

(2) E represents the finite multiple subset of the ordered set V & V of the relation between user in social networks, and the set that element repeats, is called limit collection, referred to as limit.

Define 2 social network influence and maximize node discovery.Find the influential member of tool in social networks.Select k social networks to maximize to affect node to make the number finally buying or diffuse information maximum.

In the present invention, the each user in social networks is used as to a node, can be expressed as: G=(V, E), V represents the user node set in social networks, E represents the limit between user node in social networks.| V|=n, V={v ₁, v ₂... v _n, choose from figure G in heuristics phase individual node, adds figure

The complex network being made up of the relation between user and user due to social networks, so the present invention has carried out related definition to concepts such as nodes.

Define 3 nodes.Node is the major part that forms social networks, regards the each user in social networks as a node in this article, and each node has the attribute of himself.Meanwhile, according to the state of node, node can be divided into again: activated node and un-activation node.

(1) activated node.In social networks, if a certain user node has been bought a certain commodity or accepted a certain idea, be called the node of activation about these commodity or idea.

(2) un-activation node.In social networks, if a certain user node is not bought a certain commodity or do not accepted a certain idea, be called the un-activation node about these commodity or idea.

Definition 4 is according to the position relationship between node, and the mutual relationship between node is divided into: out-degree node, in-degree node, can reach node, unreachable node.

(1) out-degree node.In G=(V, E), the node of limit using u as initial point terminal pointed, is called the out-degree node of node u.

(2) in-degree node.In G=(V, E), the node of the corresponding initial point in limit using u as terminal, is called the in-degree node of node u.

(3) can reach node.In G=(V, E), u ₁can arrive u by specific path ₂, and u ₂not u ₁out-degree node, claim u ₂u ₁reached at node.

(4) unreachable node.In G=(V, E), u ₁do not exist path to arrive u ₂, claim u ₂u ₁unreachable node.

Provide community network of the present invention below, and maximize the computing method that affect node:

Maximize the computing method that affect node based on the community network of trusting, the concrete steps of these computing method are:

A. defining social networks and social networks maximizes and affects the discovery of node, chooses

B. heuristics phase: (1), if un-activation node v is the out-degree node that has activated node u, v is direct trust to the trust of u, calculates directly and trusts; If v is reached at the node of u, v is indirect trust to the trust of u, calculates indirectly and trusts.(2) if un-activation node v is the out-degree node that has activated node u, u is direct impact on the impact of v, calculates directly impact; If v is reached at the node of u, u is remote effect on the impact of v, calculates remote effect.(3) influence value calculating in conjunction with above-mentioned two situations and trust value, the combined influence value of computing node u.(4) the node combined influence value drawing according to above step and can activate interstitial content, the community network that calculates node affects.(5) node that the community network of heuristics phase being chosen has the greatest impact joins in having the greatest impact of community network node set.

C. the greed stage: from remaining node, choose the current node that can activate at most node that has, join in having the greatest impact of community network node set.

D. the generation of having the greatest impact of community network node set.

The research for having the greatest impact of community network node in the past, has all just considered the impact between node, it is considered herein that, the degree of belief between node has vital effect to the node in activation social networks as much as possible.Therefore,, with respect to research in the past, the present invention is more comprehensive.

(1) calculating of directly trusting and indirectly trusting

User node in social networks has been divided into and has activated node and un-activation node, division according to being whether user node is bought certain commodity or whether accepted certain idea.And whether buy certain commodity or whether accept certain idea, depend on whether user is subject to others impact and the trusting degree of user to others' recommendation.

Definition 5 is trusted.Un-activation node in social networks according to self in the past with activate the situation that contacts (bought other commodity in the past or accepted the situation of other idea under the recommendation that activates node) of node or activated the prestige of node, produce trust to a certain degree to activating node.According to the source difference of trusting, will trust again point in order directly to trust and indirect trust.

A. directly trust.Supposing to activate node is u, and un-activation node is v, in the time of out-degree node that v is u, illustrates and between u and v, has contact, and according to contact situation in the past and the prestige of node u itself, v will produce directly trust to u, abbreviation DT.

B. indirectly trust.Supposing to activate node is u, and un-activation node is v, and in the time of reached at node that v is u, the prestige of node u is through the transmission of n layer, and v will produce indirectly and trust u, is called for short IT.

If un-activation node v is the out-degree node that has activated node u, v directly trusts as follows to u generation:

${\mathrm{DT}}_{\mathrm{uv}}=\frac{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_s^k\&times;W_s^k+\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_c^k\&times;W_c^k}{N}+\mathrm{RTN}\left(u\right),---\left(1\right)$

Wherein, $\mathrm{RTN}\left(u\right)=\frac{\underset{w\&Element;N\left(u\right)}{\mathrm{\&Sigma;}}\mathrm{RTN}(u\&RightArrow;w)}{\mathrm{out}\left(u\right)}$

DT _uvfor the trust of v to u.Wherein, O _srepresent the recommendation success of u and v, be illustrated in and recommend for the k time successfully, while recommending successfully, O _sbe 1, recommend failed O _sbe 0.W _sweights while representing to recommend successfully, be illustrated in the weights while recommending for the k time successfully; O _crepresent the weights of evaluating, and O _c∈ [0,1], represent the weights of the evaluation providing while recommendation for the k time, $\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{W}_s^i+\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{W}_c^i=1.$

RTN (u) is the prestige of node u.N (u) is the set of the out-degree node of u, RTN (u → w) be u go out the trust value of limit neighbours w to it, out (u) represents the out-degree of u.Be the trust value RTN (u) of u, depend on its trust of all limit neighbours of going out to it.For each node, a given initial trust value.

If v is reached at the node of u, v produces indirectly and trusts u, and its computing method are:

${\mathrm{IT}}_{\mathrm{uv}}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{1}{2^{h_i}}\&times;\mathrm{RTN}\left(u\right),h_i\&Element;H\left[n\right]---\left(2\right)$

IT _uvthe trust of v to u.U is n bar to the path of v, and traversal u is to each paths of v, h _ibe u to the level at v place on the i path of v, be equivalent to the height of v on binary tree, array H[n] in deposit the level of each u to v place on the path of v.

(2) the directly calculating of impact and remote effect

Not only there is trusting relationship in the user node in social networks,, also influence each other meanwhile.There is certain impact to unactivated node in the node having activated.

Definition 6 impacts.In social networks, un-activation node is subject to activating the impact of node, can produce and activate similar the reacting of node certain commodity or certain idea.According to the source difference of impact, impact is divided into direct impact and remote effect.

A. directly impact.Supposing to activate node is u, and un-activation node is v, and in the time of out-degree node that v is u, u will exert an influence to v, and the namely direct impact of u on v is called for short DI.

B. remote effect.Supposing to activate node is u, and un-activation node is v, and in the time of reached at node that v is u, u, by transmission successively, will exert an influence to v, and the namely remote effect of u to v are called for short IE.

If un-activation node v is the out-degree node that has activated node u, u will have a direct impact as follows to v:

${\mathrm{DI}}_{\mathrm{uv}}=\frac{\mathrm{\&lambda;}\&times;W(u,v)\&times;\mathrm{out}\left(u\right)}{\underset{q\&Element;O\left(v\right)}{\mathrm{\&Sigma;}}\frac{1}{\mathrm{out}\left(q\right)}}---\left(3\right)$

DI _uvfor the direct impact of u on v generation.Wherein λ is decay factor, and W (u, v) is the weight on limit (u, v), and out (u) is the out-degree of node u, and set O (v) is the set that goes out limit neighbours of node v.

If v is reached at the node of u, will to produce remote effect to v as follows for u:

${\mathrm{IE}}_{\mathrm{uv}}=\underset{q_i=\mathrm{PA}\left[i\right]\left[n\right],q_{i_{\mathrm{active}}=0}}{\overset{n}{\mathrm{\&Sigma;}}}\left(\underset{p_i=\mathrm{PA}\left[i\right]\left[j\right]}{\overset{n}{\mathrm{\&Pi;}}}\frac{\mathrm{\&lambda;}\&times;w}{\mathrm{out}\left(p_i\right)}\right)---\left(4\right)$

IE _uvfor the remote effect of u to v generation.Wherein PA[n] be u in [n] to the set of the node of process in the path of v, for each road is through distributing a two-dimensional array; PA[i] storage is j node in i paths in [j]; PA[i] [n] be every paths i, total n paths, every u is to the path effects sum of v; represent that i paths did not travel through; W is the weights that connect the limit of two adjacent nodes in a paths; λ represents the reduction of weights effect; Out (p _i) expression node p _iout-degree.

(3) calculating of combined influence

In order to obtain the comprehensive unified impact of egress u on v, define combined influence.Because node v exists direct trust and indirectly trusts u, and u exists direct impact and remote effect to v, for being maximized, the community network of postorder affects node (The Max Influence of Social Networking Node, MISN) the work of choosing is carried out fast and effectively, by these four effective combinations of factor, can obtain the combined influence of node u to v, be called for short CE.

Suppose to activate node u, un-activation node is v, and u can arrive v by certain path, introduces multi-stress θ, θ ∈ [0,1].

If a. un-activation node v is the out-degree node that has activated node u, combined influence is:

CE _uv＝θDI _uv+(1-θ)DT _uv (5)

If b. v is reached at the node of u, combined influence is:

CE _uv＝θIE _uv+(1-θ)IT _uv (6)

(4) calculating of the community network of node impact

Heuristics phase is chosen social networks, and to maximize the standard that affects node be combined influence and the current nodes that activates of node of node.By both combinations, can find more accurately having the greatest impact of community network node, and join in having the greatest impact of community network node set.The node that heuristics phase is selected, although not necessarily have maximum coverage, is but containing huge potential impact.

Social network influence node is calculated as follows:

${\mathrm{MISN}}_u=\frac{\underset{r\&Element;R(u\&RightArrow;v),r_{\mathrm{active}}=0}{\mathrm{\&Sigma;}}{\mathrm{CE}}_{\mathrm{uv}}+N}{\mathrm{out}\left(u\right)}---\left(7\right)$

SN _uit is the combined influence of u.Wherein R (u → v) represent that u can arrive v through certain path; R represents the wherein paths of u to v, r _active=0 represents not live through this paths; N is the number of the current node being activated by u; Out (u) is the out-degree of node u.

Embodiment:

Provide the community network impact calculating (as shown in Figure 2) that following embodiment illustrates node.V={u ₁, u ₂,, u ₃, u ₄,, u ₅, u ₆, u ₇, u ₈, E={ (u ₁, u ₂), (u ₁, u ₃), (u ₁, u ₄), (u ₂, u ₅), (u ₂, u ₆), (u ₃, u ₇), (u ₄, u ₅), (u ₄, u ₇), (u ₅, u ₆), (u ₄, u ₇), (u ₆, u ₈), (u ₇, u ₈), wherein (u ₁, u ₂) represent u ₁, u ₂between limit.

This embodiment has 8 nodes, suppose therefrom to choose 4 nodes as having the greatest impact of community network node, heuristic factor C=0.5, heuristics phase should be chosen two nodes and adds having the greatest impact of community network node set, and the greed stage should be chosen two nodes and add having the greatest impact of community network node set.

Suppose u ₁for activating node, u ₂, u ₃, u ₄, u ₅, u ₆, u ₇, u ₈for un-activation node.

1. u ₁→ u ₂, calculate direct trust and directly impact according to formula (1) (3), then calculate combined influence according to formula (5).

2. u ₁→ u ₃, calculate direct trust and directly impact according to formula (1) (3), then calculate combined influence according to formula (5).

3. u ₁→ u ₄, calculate direct trust and directly impact according to formula (1) (3), then calculate combined influence according to formula (5).

4. u ₁→ u ₅, have two paths: u ₁→ u ₂→ u ₅, u ₁→ u ₄→ u ₅.Calculate respectively u according to formula (2) (4) ₁→ u ₂→ u ₅and u ₁→ u ₄→ u ₅indirect trust and remote effect, then calculate combined influence separately according to formula (6).

5. u ₁→ u ₆, have two paths: u ₁→ u ₂→ u ₆, u ₁→ u ₂→ u ₅→ u ₆.Calculate respectively u according to formula (2) (4) ₁→ u ₂→ u ₆and u ₁→ u ₂→ u ₅→ u ₆indirect trust and remote effect, then calculate combined influence separately according to formula (6).

6. u ₁→ u ₇, have two paths: u ₁→ u ₄→ u ₇, u ₁→ u ₃→ u ₇.Calculate respectively u according to formula (2) (4) ₁→ u ₄→ u ₇and u ₁→ u ₃→ u ₇indirect trust and remote effect, then calculate combined influence separately according to formula (6).

7. u ₁→ u ₈, have five paths u ₁→ u ₂→ u ₅→ u ₆→ u ₈, u ₁→ u ₂→ u ₆→ u ₈, u ₁→ u ₄→ u ₅→ u ₆→ u ₈, u ₁→ u ₄→ u ₇→ u ₈, u ₁→ u ₃→ u ₇→ u ₈.Calculate respectively u according to formula (2) (4) ₂→ u ₅→ u ₆→ u ₈, u ₁→ u ₂→ u ₆→ u ₈, u ₁→ u ₄→ u ₅→ u ₆→ u ₈, u ₁→ u ₄→ u ₇→ u ₈, u ₁→ u ₃→ u ₇→ u ₈indirect trust and remote effect, then calculate combined influence separately according to formula (6).

8. by all combined influence substitution formula (7) that calculate above, calculate node u ₁community network impact.

Suppose respectively successively u ₂, u ₃, u ₄, u ₅, u ₆, u ₇, u ₈for un-activation node, calculate the direct trust of each node and indirectly trust, directly impact and remote effect, combined influence, finally calculate the community network impact of each node, after calculating finishes, therefrom choose the node that two community networks have the greatest impact, add the set of having the greatest impact of community network.

Calculate the nodes that remaining unchecked node can activate at most, choose two nodes that can activate maximum nodes, add the set of having the greatest impact of community network.

Having the greatest impact of community network node set generates.

In sum, the present invention is to be the research of carrying out relevant issues under a relatively abstract and complicated object condition at social networks itself.For convenience of analyzing, introduce the method for graph theory, by the social networks imagery of abstract.And introduce and trust this element, consider the material impact of trusting activating node, the degree of belief between node and influence value are carried out to combination, draw combined influence, then draw having the greatest impact of community network node according to combined influence.Compared with prior art, the computing method of having the greatest impact of the community network node based on trusting that the present invention proposes, can choose more accurately having the greatest impact of community network node, thereby realize as much as possible having the greatest impact of community network.The present invention, from direct trust, trusts indirectly, directly impact, and remote effect, combined influence, the aspects such as community network impact, realize choosing of having the greatest impact of community network node.As can be seen here, in current Renren Network, in the social networks such as microblogging, for better diffusing information and finding the most influential user etc. to there is the prospect of practical application, make influential user propagate positive information, make great efforts to build a positive network environment.Meanwhile, due to the trust and the impact that the present invention relates between user, so safely etc. also there is the prospect of practical application in aspect at Internet.

Claims (9)

1. maximize based on the community network of trusting computing method that affect node, it is characterized in that, based on Graph-theoretical Approach, its computing method basic step is:

The first step, definition social networks and social networks maximize affects the discovery of node, choose;

Second step, heuristics phase;

The 3rd step, the greed stage;

The 4th step, the generation of having the greatest impact of community network node set.

2. the community network based on trusting according to claim 1 maximizes the computing method that affect node, it is characterized in that, described first step definition, comprises step:

Described social networks turns to two orderly tuple <V of a figure G with form, E>,

|V|＝n,V＝{v ₁,v ₂,......v _n},|E|＝m,E＝{e ₁,e ₂,e ₃......e _m}

Each user in social networks is used as to a node, is expressed as:

G＝(V,E)

(1) V is a non-NULL finite set, and the set being made up of the user node in social networks, is called vertex set;

(2) E represents the finite multiple subset of the ordered set V & V of the relation between user in social networks, and the set that element repeats, is called limit collection, referred to as limit;

Described node is the major part that forms social networks, regards the each user in social networks as a node, is divided into according to the state of node:

(1) activate node, in social networks, if a certain user node has been bought a certain commodity or accepted a certain idea, be called the node of activation about these commodity or idea;

(2) un-activation node, in social networks, if a certain user node is not bought a certain commodity or do not accepted a certain idea, is called the un-activation node about these commodity or idea;

Mutual relationship between described node is divided into:

(1) out-degree node, in G=(V, E), the node of limit using u as initial point terminal pointed, is called the out-degree node of node u;

(2) in-degree node, in G=(V, E), the node of the corresponding initial point in limit using u as terminal, is called the in-degree node of node u;

(3) can reach node, in G=(V, E), u ₁can arrive u by specific path ₂, and u ₂not u ₁out-degree node, claim u ₂u ₁reached at node;

(4) unreachable node, in G=(V, E), u ₁do not exist path to arrive u ₂, claim u ₂u ₁unreachable node;

Described social networks maximizes affect node discovery, finds in social networks the influential member of tool;

Described k social networks maximizes affects choosing of node, makes the number finally buying or diffuse information maximum, and introduces heuristic factor c, chooses in heuristics phase individual having the greatest impact of community network node, the greed stage is chosen individual having the greatest impact of community network node.

3. the community network based on trusting according to claim 1 maximizes the computing method that affect node, it is characterized in that, described second step heuristics phase, comprises step:

(1) directly trust and indirectly trust;

(2) directly impact and remote effect;

(3) node combined influence;

(4) community network affects node calculating;

(5) choose having the greatest impact of community network node.

4. the community network based on trusting according to claim 1 maximizes the computing method that affect node, it is characterized in that, described the 3rd step greed stage, comprise step: from remaining node, choose the current node that can activate at most node that has, join in having the greatest impact of community network node set.

5. the community network based on trusting according to claim 3 maximizes the computing method that affect node, it is characterized in that, described (one) directly trusts and indirectly trusts, and also comprises step:

If un-activation node v is the out-degree node that has activated node u, v is direct trust to the trust of u, and direct trust is calculated:

Its computing method are:

${\mathrm{DT}}_{\mathrm{uv}}=\frac{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_s^k\&times;W_s^k+\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_c^k\&times;W_c^k}{N}+\mathrm{RTN}\left(u\right),---\left(1\right)$

Wherein, $\mathrm{RTN}\left(u\right)=\frac{\underset{w\&Element;N\left(u\right)}{\mathrm{\&Sigma;}}\mathrm{RTN}(u\&RightArrow;w)}{\mathrm{out}\left(u\right)}$

DT _uvfor the trust of v to u;

Wherein, O _srepresent the recommendation success of u to v, be illustrated in and recommend for the k time successfully, while recommending successfully, O _sbe 1, recommend failed O _sbe 0;

W _sweights while representing to recommend successfully, be illustrated in the weights while recommending for the k time successfully;

O _crepresent the weights of evaluating, and O _c∈ [0,1], represent the weights of the evaluation providing while recommendation for the k time, $\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{W}_s^i+\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{W}_c^i=1;$

RTN (u) is the prestige of node u; N (u) is the set of the out-degree node of u, RTN (u → w) be u go out the trust value of limit neighbours w to it, out (u) represents the out-degree of u;

Be the trust value RTN (u) of u, depend on its trust of all limit neighbours of going out to it;

For each node, a given initial trust value;

If v is reached at the node of u, v is indirect trust to the trust of u, and indirect trust is calculated, and its computing method are:

${\mathrm{IT}}_{\mathrm{uv}}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{1}{2^{h_i}}\&times;\mathrm{RTN}\left(u\right),h_i\&Element;H\left[n\right]---\left(2\right)$

IT _uvthe trust of v to u; U is n bar to the path of v, and traversal u is to each paths h of v _ibe u to the level at v place on the i path of v, be equivalent to the height of v on binary tree, array H[n] in deposit the level of each u to v place on the path of v.

6. the community network based on trusting according to claim 3 maximizes the computing method that affect node, it is characterized in that, described (two) directly affect and remote effect, also comprise step:

If un-activation node v is the out-degree node that has activated node u, u has a direct impact v, and direct impact is calculated, and its computing method are:

${\mathrm{DI}}_{\mathrm{uv}}=\frac{\mathrm{\&lambda;}\&times;W(u,v)\&times;\mathrm{out}\left(u\right)}{\underset{q\&Element;O\left(v\right)}{\mathrm{\&Sigma;}}\frac{1}{\mathrm{out}\left(q\right)}}---\left(3\right)$

Wherein, DI _uvfor the direct impact that u produces v, λ is decay factor, and W (u, v) is the weight on limit (u, v), and out (u) is the out-degree of node u, and set O (v) is the set that goes out limit neighbours of node v;

If v is reached at the node of u, u produces remote effect to v, and remote effect are calculated, and its computing method are:

${\mathrm{IE}}_{\mathrm{uv}}=\underset{q_i=\mathrm{PA}\left[i\right]\left[n\right],q_{i_{\mathrm{active}}=0}}{\overset{n}{\mathrm{\&Sigma;}}}\left(\underset{p_i=\mathrm{PA}\left[i\right]\left[j\right]}{\overset{n}{\mathrm{\&Pi;}}}\frac{\mathrm{\&lambda;}\&times;w}{\mathrm{out}\left(p_i\right)}\right)---\left(4\right)$

IE _uvfor the remote effect of u to v generation;

Wherein, PA[n] [n] be u to the set of the node of process in all paths of v, for each road is through distributing a two-dimensional array;

PA[i] storage is j node in i paths in [j];

PA[i] [n] be every paths i, total n paths, every u is to the path effects sum of v;

represent that i paths did not travel through;

W is the weights that connect the limit of two adjacent nodes in a paths;

λ represents the reduction of weights effect; Out (p _i) expression node p _iout-degree.

7. the community network based on trusting according to claim 3 maximizes the computing method that affect node, it is characterized in that, described (three) node combined influence, also comprise step: to above-mentioned directly and the influence value that calculates of indirect two kinds of situations and trust value to carry out result comprehensive, calculate the combined influence value of node u;

Described combined influence, because node v exists direct trust and indirectly trusts u, and u exists direct impact and remote effect to v, for the work of choosing that makes the community network maximization of postorder affect node is carried out fast and effectively, by these four effective combinations of factor, can obtain the combined influence of node u to v, be called for short CE;

Suppose to activate node u, un-activation node is v, and u can arrive v by certain path, introduces multi-stress θ, θ ∈ [0,1];

(1) if un-activation node v is the out-degree node that has activated node u, combined influence is:

CE _uv＝θDI _uv+(1-θ)DT _uv (5)

(2) if v is reached at the node of u, combined influence is:

CE _uv＝θIE _uv+(1-θ)IT _uv (6)。

8. the community network based on trusting according to claim 3 maximizes the computing method that affect node, it is characterized in that, described (four) community network affects the calculating of node, also comprises step:

The node combined influence value drawing according to above step and the activated interstitial content of node, calculate the community network impact of node, is calculated as follows:

${\mathrm{MISN}}_u=\frac{\underset{r\&Element;R(u\&RightArrow;v),r_{\mathrm{active}}=0}{\mathrm{\&Sigma;}}{\mathrm{CE}}_{\mathrm{uv}}+N}{\mathrm{out}\left(u\right)}---\left(7\right)$

MISN _uit is the community network impact of u;

Wherein R (u → v) represent that u can arrive v through certain path;

R represents the wherein paths of u to v, r _active=0 represents not live through this paths;

N is the number of the current node being activated by u;

Out (u) is the out-degree of node u.

9. the community network based on trusting according to claim 3 maximizes the computing method that affect node, it is characterized in that, described (five) choose having the greatest impact of community network node, also comprise step: the node that community network is had the greatest impact joins in the set of having the greatest impact of community network node.