CN105915376A - Log information network structuring method and log information network structuring system based on P2P program requesting system - Google Patents

Abstract

The invention discloses a log information network structuring method based on a P2P program requesting system. The method comprises the following steps of a first step, acquiring log information of a user; a second step, according to the log information of a node, constructing a community network structure in which the users are used as nodes, relationships among the users are used as sides and weight is represented through node bandwidth; a third step, performing community structure dividing on the community network structure according to a LabelRank algorithm; and a fourth step, combining community structure dividing results in the third steps; wherein when data transmission is required, data transmission is performed by means of effective nodes in a corresponding community which is obtained after combining. Furthermore the invention discloses a log information network structuring system based on the P2P program requesting system. The log information network structuring method and the log information network structuring system can effectively handle node testing in crossed communities, thereby forming a more reasonable, more effective and more stable P2P network.

Description

Log information network structure method based on P2P VOD system and system thereof

Technical field

The present invention relates to P2P network and detection algorithm field, community, a kind of log information network structure method and system based on P2P VOD system.

Background technology

Along with the high speed development of the Internet, multimedia also obtains fulminant growth in the Internet, people gradually from the traditional media such as TV, radio be relayed to the Internet watch audio frequency and video.This fulminant growth does not only result in the bandwidth traffic cost of resource provider and increases severely, but also requires network equipment provider continuous upgrading network, improves network transfer speeds.And P2P network is proposing the most just to cause the hugest vibrations of the Internet as a kind of new network transmission mode, it is not only as a kind of software architecture form, is also the embodiment of a kind of social pattern.The client-server mode that P2P network breaks traditions, formed a kind of with Client-client the ability of server (client have), client is converted into the form of role server, thus improves bandwidth resources utilization rate and alleviate the pressure of resource service provider.But in actual applications, P2P network usually shows not fully up to expectations in stability, effectiveness, one is unstable, often results in user and interrupts during request resource.Two is effectiveness, and user asks the waiting time of resource to be slower than server most of the time, and this is that the hardware such as bandwidth and processor disposal ability of client causes.Shortcoming like this causes P2P development scale limited.

Community's detection division methods is a study hotspot in recent years, academia starts one upsurge to the research of complex network, complex network refers to, by numerous individualities connecting each other, interacting, the network structure produced according to certain contact, wherein contain associated abundant information.Complex network network does not has specific field, and it relates to the numerous areas such as mathematics, physics, sociology, computer, and obtains the concern of the researcher of various fields.In order to study and obtain the effective information of complex network structures, need to use algorithm that it is carried out structure division, such as based on modularity division methods, division methods based on hierarchical structure etc., these algorithms are devoted to the rational community structure solving and finding in complex network, thus let us preferably recognizes the characteristic of complex network.And P2P network is as one of complex network, community's detection division methods is utilized it to be calculated, divides, it appeared that effective key message of network, thus improve stability and the effectiveness of network.But while community's detection division methods obtains concern and the research of substantial amounts of researcher, also obtain sizable achievement, but for various informative network structure, still there are some the most unsolved basic problems.Although such as some find that algorithm it is can obtain satisfied result, but generally require to sacrifice calculation cost as structure；The algorithm that some performances are the most superior, sacrificing accuracy the most at most is cost.These problems are required for the most perfect.

In prior art, for the problem solving the overlapping community in complicated P2P network, propose copra algorithm to tackle, detailed in China's documents and materials " the community mining algorithm research summary propagated based on label " describe overlapping copra algorithm and innovatory algorithm research thereof, particularly propose in the improved method to copra algorithm setting threshold value to control the number of tags of each node so that new algorithm need not initially.

In addition, disclose a kind of overlapping community detection method propagated based on multi-tag in Chinese patent application CN201510076028.8, comprise the following steps: step A, construct social network diagram: read network data, structure is with user as node, and customer relationship is the social network diagram on limit；Step B, analyzes the coarse core of network: according to social network diagram, and the degree of each node, analyzes coarse core set RoughCore of social networks；Step C, init Tag set: calculate the structure weights of each limit two node, integrating step B gained RoughCore result in social networks, initialize the tag set of each node, and judge each joint core state CoreStatus；Step D, performs label and propagates: according to link density in whole social networks, calculates the new tag set of each node, filters less degree of membership label according to joint core state CoreStatus simultaneously, obtains preliminary overlapping community result；Step E, decomposes discontinuous community: in preliminary overlapping community result, discontinuous community is decomposed into Duo Gezi community, obtains final social networks overlap community structure.

But reasonability, effectiveness and stability that community structure of the prior art divides are to be improved.

Summary of the invention

It is an object of the invention to provide a kind of log information network structure method and system based on P2P VOD system, the method and system and the detection of the node effectively processing cross-community can be formed more reasonable, effective and stable P2P network.

The concrete technical scheme of the present invention is: a kind of log information network structure method based on P2P VOD system, comprises the following steps:

Step 1: collect the log information of user；

Step 2: according to the log information of node, structure is with user as node, and the relation between user is limit, is represented the community network structure of weight by node bandwidth；

Step 3: community network structure is carried out community structure division according to LabelRank algorithm, obtains multiple community；Wherein, LabelRank algorithm comprises an operation operator: transmission operates；

Described transmission operation is particularly as follows: preserve label distribution at the vectorial P of each one 1 × n of node definition, and n is the quantity of node；Define an adjacency matrix A and store network structure, then P_iC () represents the probability that node i belongs to label c, the collection of label share C and represents, and the number of label is initialized as number n of node, calculates every time and is required for updating the vectorial P of each node, and formula is as follows:

$P_i^{,}\left(c\right)=\underset{j\∈Nb\left(i\right)}{\Σ}{P}_j\left(c\right)/k_i,\∀c\∈C\mathrm{...}\left(1\right)$

Wherein, Nb (i) is the set of the neighbor node of node i, k_i=| Nb (i) | is the number of nodes of neighbor node set, P '_iC () represents and updates the probability that posterior nodal point i belongs to label c, being then followed by calculating can be expressed as by matrix A and vector P:

A×P.............(2)

Firstly the need of initialization vector P in transmission operation operator, initialized method particularly includes: the probability P of the list of labels of each node_ijIt is initialized as the inverse of the weight sum of neighbor node；As follows:

P_ij=w/k_i............(3)

The node i that represents w arrives the weight of node j；

The vectorial P of each node is obtained after above-mentioned steps；

In the renewal process every time calculated of transmission operation operator, threshold value λ is set, when the probability of the affiliated label of node is more than λ, then retains, less than then giving up；And when label probability all of in node is both less than λ, then randomly select a reservation；

Step 4: the node in the community in step 3 is merged and processes to remove the invalid node in community.

In above-mentioned log information network structure method based on P2P VOD system, described expansion factor is particularly as follows: use expansion factor Γ in vector P_inAffecting label transmission, wherein in is a real number value, by using expansion factor to carry out Decoupling network structure, calculates Γ every time_inP, P_icIn will be increased^thPower, shown in equation below:

${\mathrm{\Γ}}_{in}{P}_i\left(c\right)=P_i{\left(c\right)}^{in}/\underset{j\∈C}{\Σ}{P}_i{\left(j\right)}^{in}$

After calculating, it can make the numerical value of the label probability of high probability in node increase, and the numerical value of the label probability of low probability reduces.

In above-mentioned log information network structure method based on P2P VOD system, described cutting coefficient is particularly as follows: introduce operator Φ in vector P_rDefinition threshold value r ∈ [0,1], if the label probability in vector P is less than r during calculating, deletes the label in vector.

In above-mentioned log information network structure method based on P2P VOD system, described condition updates particularly as follows: one condition of definition updates operator Θ_q, its implication is when in algorithmic procedure, only node just updates this node time unequal with the label of its neighbor node vector, is not the most updated operation；The most in each iteration, when node meets following equation, just it is updated operation；

$\underset{j\∈Nb\left(i\right)}{\Σ}isSubset(C_i^{*},C_j^{*})\≤{\mathrm{qk}}_i$

It is the node i tag set of maximum probability, k in conventional calculation procedure_iRepresenting the degree of node i, q is the span [0,1] of a real number value q.

In above-mentioned log information network structure method based on P2P VOD system, the Peer number in same group, User IP, group interior other IDs, user bandwidth when described log information is ID, video URL, broadcast mode, Connection Service device.

In above-mentioned log information network structure method based on P2P VOD system, described step 4 is specially

The result that community structure in S41: obtaining step 3 divides；

S42: the effective node in the result divide community structure merges, and removes invalid node；

S43: the data after storage merges in the form of a file.

The present invention also provides for a kind of system for realizing above-mentioned log information network structure method based on P2P VOD system simultaneously, including following structure:

Log collection module, for collecting the log information of user；

Community network structure constructing module, for the log information according to user, structure is with user as node, and the relation between user is limit, is represented the community network structure of weight by node bandwidth；

Community structure divides module, for community network structure being carried out community structure division according to LabelRank algorithm, obtains multiple community；

Merge module: the community structure division result obtained for community structure divides module merges and processes to remove the invalid node in community.

In above-mentioned log information network structure system based on P2P VOD system, described merging module includes following submodule:

Result obtains submodule, divides the result of the community structure division that module obtains for obtaining community structure；

Merging submodule, the effective node in the result dividing community structure merges, and removes invalid node；

Sub module stored, the data after storage merges in the form of a file.

Compared with prior art, the beneficial effects of the present invention is:

The present invention carries out community structure division by using LabelRank algorithm to community network structure, in operating operation operator-transmission, the probability to the label of each node arranges threshold value, each node can be only present in limited community, the most each node retains limited label, the node belonging to multiple label is exported, reaches to process the purpose of cross-community.

In the present invention, carry out community's detection by research community detection method, analysis P2P daily record the P2P network that it is formed and divide, find out the key message in P2P network, form more reasonable, effective and stable P2P network.

Accompanying drawing explanation

Fig. 1 is the flow chart of the embodiment of the present invention 1；

Fig. 2 is the flow chart of the embodiment of the present invention 1；

Fig. 3 and 4 is the schematic diagram of the transmission operation of the embodiment of the present invention 1 and 2；

Fig. 5-7 is the embodiment of the present invention 1 and 2 network structure without merging treatment；

Fig. 8 is the embodiment of the present invention 1 network structure through merging treatment；

Fig. 9 is the block diagram of the embodiment of the present invention 1.

Detailed description of the invention

Below in conjunction with detailed description of the invention, technical scheme is described in further detail, but does not constitute any limitation of the invention.

Embodiment 1

As illustrated in fig. 1 and 2, a kind of log information network structure method based on P2P VOD system, comprise the following steps:

Step 1: collect the log information of user；

Step 2: according to the log information of node, structure is with user as node, and the relation between user is limit, is represented the community network structure of weight by node bandwidth；

Annexation between node is calculated by following algorithm flow:

1) ID is mark, can count on Peer number and the ID of Peer of broadcasting of a certain moment same video URL, thus can form the network structure relation not possessing weight between node.

2) set weight coefficient with the bandwidth relationship of user, thus form the network structure possessing weight.

Above-mentioned is the calculation process of coarseness, on the basis of above-mentioned calculation process, needs in view of situations below in the case of Practical Calculation:

1) statistics neighbours' Peer node rule.

When a certain node and other neighbor node network consistings, we need to arrange a threshold value, when the number of nodes of the network formed exceedes threshold value, we just can be in this network statistics to network structure, so can filter out some mininets, because generally mininet is less to the probability forming P2P network.

2) contact between Peer node is calculated.

When a certain node and other nodes are formed and contact, generally requiring and weigh the tightness degree of relation between them by a quantitative value, this tightness degree then represents by weight, and this technology is to weigh the network weight of formation from the network bandwidth relation of node.

Acquisition node bandwidth method:

The method calculating node bandwidth, because P2P VOD system is sing on web form, therefore we are to utilize JavaScript to carry out testing the speed and record, thus reach to record the bandwidth information of user, and we mainly utilize JavaScript to test the bandwidth of client.By obtaining the bandwidth information of each node, we can be with the weight information between the Form generation node of passing ratio, such as node A and node B are related with video C, the bandwidth of node A, B is respectively 100KB, 200KB, then node A and node B and video C contact respectively 1/3 and 2/3, be normalized.By bandwidth we can with the certain weight of algorithm, but effectively weight also needs to consider the network speed speed that node is concrete, if network speed speed is less than certain parameter value set, then weight is reduced into 0.

Step 3: community network structure is carried out community structure division according to LabelRank algorithm, obtains multiple community；Wherein, LabelRank algorithm comprises four operation operators: transmission operation, expansion factor, cutting coefficient, condition update；

1, transmission operation

Preserving label distribution at the vectorial P of each one 1 × n of node definition, n is the quantity of node；Define an adjacency matrix A and store network structure, then P_iC () represents the probability that node i belongs to label c, the collection of label share C and represents, and the number of label is initialized as number n of node, calculates every time and is required for updating the vectorial P of each node, and formula is as follows:

$P_i^{,}\left(c\right)=\underset{j\∈Nb\left(i\right)}{\Σ}{P}_j\left(c\right)/k_i,\∀c\∈C\mathrm{...}\left(1\right)$

Wherein, Nb (i) is the set of the neighbor node of node i, k_i=| Nb (i) | is the number of nodes of neighbor node set, P '_iC () represents and updates the probability that posterior nodal point i belongs to label c, the label of any node is the most all unique, and being then followed by calculating can be expressed as by matrix A and vector P:

A×P.............(2)

Firstly the need of initialization vector P in transmission operation operator, initialized method particularly includes: the probability P of the list of labels of each node_ijIt is initialized as the inverse of the weight sum of neighbor node；As follows:

P_ij=w/k_i............(3)

The node i that represents w arrives the weight of node j；

The vectorial P of each node is obtained after above-mentioned steps；

In the renewal process every time calculated of transmission operation operator, threshold value λ is set, when the probability of the affiliated label of node is more than λ, then retains, less than then giving up；And when label probability all of in node is both less than λ, then randomly select a reservation；

Transmission operation is as it is shown on figure 3, initialize each node, and as a example by node a, what (b, 1/4) represented is, and node a belongs to the probability of label b is 1/4, because node a has 4 neighbor nodes, the weight of acquiescence each edge is 1, therefore obtains 1/4.Concurrently set threshold value λ=1/2, then retain when probability is more than or equal to 1/2；Less than then giving up；When node all both less than 1/2 time, then randomly select one, and probability be set to 1.Available result as shown in Figure 4.

2, expansion factor

In the algorithm, we use expansion factor Γ in vector P_inAffecting label transmission, wherein in is a real number value.We use expansion factor to carry out Decoupling network structure.Calculate Γ every time_inP, P_icIn will be increased^thPower, shown in equation below:

${\mathrm{\Γ}}_{in}{P}_i\left(c\right)=P_i{\left(c\right)}^{in}/\underset{j\∈C}{\Σ}{P}_i{\left(j\right)}^{in}$

After calculating, it can make the numerical value of the label probability of high probability in node increase, and the label probability numerical value of low probability reduces few.Such as, two labels are initialized as 0.6 and 0.4.When in is 2, the label numerical value after calculating is respectively 0.6923 and 0.4.

3, cutting coefficient

In order to avoid the problem that EMS memory occupation is excessive, we introduce operator Φ in vector P_rDefinition threshold value r ∈ [0,1], if the label probability in vector P is less than r during calculating, deletes the label in vector.Operator Φ_rCan well be used in combination with expansion factor, reduce the number of label in vector.If it is demonstrated experimentally that r=0.1, the mean number of the label of the most each node can be less than 3.

In the present invention, having threshold value λ equally in transmission operation operator, it is different from this cutting coefficient operator effect；Specifically, threshold value λ is to make transmission operation operator be optimized, and makes each node have one or arrange the label limiting quantity number, it is achieved the multi-tag output of overlapping community.

But the effect of this cutting coefficient operator is primarily to remove the label that in LabelRank algorithm, label probability is the least, reduces EMS memory occupation, improves and calculates speed.

4, condition updates

By three above operations factor network structure calculated the performance that still can not well ensure algorithm, this is because above three condition is extremely difficult to convergence to the process that community is detected, and divides community's poor quality out.Therefore one condition of definition updates operator Θ_q, with it, algorithm being improved, its implication is when in algorithmic procedure, only node just updates this node time unequal with the label of its neighbor node vector, is not the most updated operation.This can make algorithm just be updated when only meeting certain condition and continue community's detection operation, in each iteration, when node meets following equation, is just updated operation.

$\underset{j\∈Nb\left(i\right)}{\Σ}isSubset(C_i^{*},C_j^{*})\≤{\mathrm{qk}}_i$

It is the node i tag set of maximum probability, k in conventional calculation procedure_iRepresenting the degree of node i, q is a real number value (span [0,1]).IfisSubset(s₁,s₂) return value is 1 to be otherwise 0, updates the label of node i when meeting above-mentioned formula.Formula is considered as into the similarity of two nodes of measurement.

Algorithm flow such as table 1 below:

Table 1

Step 4: the community structure division result in step 3 is merged and processes to remove the invalid node in community；When needs carry out data transmission, the effective node in the corresponding community obtained after using merging treatment carries out data transmission.

LabelRank algorithm is for the input in different moment, network structure is divided, thus produce the most corresponding result, it is contemplated that the P2P VOD system of reality, mainly to capture for a certain resource, whether this node can be as resource service provider, it is possible to the resource of stable effective transmission.Therefore we are required for the division result in each moment and merge process, the effective node belonging to same resource is mainly merged by merging treatment, and remove some invalid node, so-called invalid node refers to cannot function as the node of resource provider, and whether can be node bandwidth as the primary concern factor of resource provider, the namely weight in the present embodiment, merging main step is:

The result that community structure in S41: obtaining step 3 divides；

S42: the effective node in the result divide community structure merges, and removes invalid node；

S43: the data after storage merges in the form of a file.

Merge result to remove and be combined result in the form of a file and store, can be conducive to VOD system that result is read out, and quickly position the situation of present node, thus judge whether this node adds P2P network as effective node.

As shown in Fig. 5,6,7, that numeral 1-4 represents is user, and what numeral more than 10000 represented is video id, Fig. 5,6,7 is 3 results exported according to moment 1-3 by LabelRank, can be apparent from sees, what the result in figure can not be apparent from finds out network structure result.After merging the result in above-mentioned 3 moment, result is as shown in Figure 8.

By above-mentioned method, the network transmission performance of P2P can be strengthened, improve the effectiveness of P2P resource transmission.By community's detection partitioning technology being applied to the division of log information file, system can be from conventional ruuning situation, automatically quickly distinguishing the network performance of active user, thus user node is converted to effective P2P node, the reliability of P2P network is ensured.

It is of special importance that use for reference and mixed the thought of COPRA in the LabelRank algorithm of the present invention, in the transmission of the present invention operates, threshold value λ is set, when the probability of the affiliated label of node is more than λ, then retains, less than then giving up；There is an exceptional case, i.e. when label probability all of in node is both less than λ, then randomly select a reservation, so can make the present invention that LabelRank algorithm can be overcome cannot to process the defect of node of overlapping community, make a node be present in limited community, in transmission operation, limited label of each node can be retained, the node belonging to multiple label is exported, reaches to process the purpose of cross-community.

Embodiment 2

The present embodiment provides the system of a kind of log information network structure method based on P2P VOD system for realizing described in embodiment 1, including following structure:

Log collection module 1, for collecting the log information of user；

Community network structure constructing module 2, for the log information according to user, structure is with user as node, and the relation between user is limit, is represented the community network structure of weight by node bandwidth；

Community structure divides module 3, for community network structure being carried out community structure division according to LabelRank algorithm, obtains multiple community；

LabelRank algorithm depends on 4 operation operators:

1, transmission operation

Described transmission operation is particularly as follows: preserve label distribution at the vectorial P of each one 1 × n of node definition, and n is the quantity of node；Define an adjacency matrix A and store network structure, then P_iC () represents the probability that node i belongs to label c, the collection of label share C and represents, and the number of label is initialized as number n of node, calculates every time and is required for updating the vectorial P of each node, and formula is as follows:

$P_i^{,}\left(c\right)=\underset{j\∈Nb\left(i\right)}{\Σ}{P}_j\left(c\right)/k_i,\∀c\∈C\mathrm{...}\left(1\right)$

Wherein, Nb (i) is the set of the neighbor node of node i, k_i=| Nb (i) | is the number of nodes of neighbor node set, P '_iC () represents and updates the probability that posterior nodal point i belongs to label c, being then followed by calculating can be expressed as by matrix A and vector P:

A×P.............(2)

Firstly the need of initialization vector P in transmission operation operator, initialized method particularly includes: the probability P of the list of labels of each node_ijIt is initialized as the inverse of the weight sum of neighbor node；As follows:

P_ij=w/k_i............(3)

The node i that represents w arrives the weight of node j；

The vectorial P of each node is obtained after above-mentioned steps；

In the renewal process every time calculated of transmission operation operator, threshold value λ is set, when the probability of the affiliated label of node is more than λ, then retains, less than then giving up；And when label probability all of in node is both less than λ, then randomly select a reservation；

Transmission operation is as it is shown on figure 3, initialize each node, and as a example by node a, what (b, 1/4) represented is, and node a belongs to the probability of label b is 1/4, because node a has 4 neighbor nodes, the weight of acquiescence each edge is 1, therefore obtains 1/4.Concurrently set threshold value λ=1/2, then retain when probability is more than or equal to 1/2；Less than then giving up；When node all both less than 1/2 time, then randomly select one, and probability be set to 1.Available result as shown in Figure 4.

2, expansion factor

In the algorithm, we use expansion factor Γ in vector P_inAffecting label transmission, wherein in is a real number value.We use expansion factor to carry out Decoupling network structure.Calculate Γ every time_inP, P_icIn will be increased^thPower, shown in equation below:

${\mathrm{\Γ}}_{in}{P}_i\left(c\right)=P_i{\left(c\right)}^{in}/\underset{j\∈C}{\Σ}{P}_i{\left(j\right)}^{in}$

After calculating, it can make the numerical value of the label probability of high probability in node increase, and the label probability numerical value of low probability reduces.Such as, two labels are initialized as 0.6 and 0.4.When in is 2, the label numerical value after calculating is respectively 0.6923 and 0.4.

3, cutting coefficient

In order to avoid the problem that EMS memory occupation is excessive, we introduce operator Φ in vector P_rDefinition threshold value r ∈ [0,1], if the label probability in vector P is less than r during calculating, deletes the label in vector.Operator Φ_rCan well be used in combination with expansion factor, reduce the number of label in vector.If it is demonstrated experimentally that r=0.1, the mean number of the label of the most each node can be less than 3.

4, condition updates

By three above operations factor network structure calculated the performance that still can not well ensure algorithm, this is because above three condition is extremely difficult to convergence to the process that community is detected, and divides community's poor quality out.Therefore one condition of definition updates operator Θ_q, with it, algorithm being improved, its implication is when in algorithmic procedure, only node just updates this node time unequal with the label of its neighbor node vector, is not the most updated operation.This can make algorithm just be updated when only meeting certain condition and continue community's detection operation, in each iteration, when node meets following equation, is just updated operation.

$\underset{j\∈Nb\left(i\right)}{\Σ}isSubset(C_i^{*},C_j^{*})\≤{\mathrm{qk}}_i$

It is the node i tag set of maximum probability, k in conventional calculation procedure_iRepresenting the degree of node i, q is a real number value (span [0,1]).IfisSubset(s₁,s₂) return value is 1 to be otherwise 0, updates the label of node i when meeting above-mentioned formula.Formula is considered as into the similarity of two nodes of measurement.

Algorithm flow such as table 2 below:

Table 2

Merge module 4: merge process to remove the invalid node in community for community structure being divided the community structure division result that obtains of module, and carried out data transmission by the effective node in the corresponding community that obtains after merging treatment.

Described merging module 4 includes following submodule:

Result obtains submodule 41, divides the result of the community structure division that module obtains for obtaining community structure；

Merging submodule 42, the effective node in the result dividing community structure merges, and removes invalid node；

Sub module stored 43, the data after storage merges in the form of a file.

Merge result to remove and be combined result in the form of a file and store, can be conducive to VOD system that result is read out, and quickly position the situation of present node, thus judge whether this node adds P2P network as effective node.

As shown in Fig. 5,6,7, that numeral 1-4 represents is user, and what numeral more than 10000 represented is video id, Fig. 5,6,7 is 3 results exported according to moment 1-3 by LabelRank, can be apparent from sees, what the result in figure can not be apparent from finds out network structure result.After merging the result in above-mentioned 3 moment, result is as shown in Figure 8.

By above-mentioned system, the network transmission performance of P2P can be strengthened, improve the effectiveness of P2P resource transmission.By community's detection partitioning technology being applied to the division of log information file, system can be from conventional ruuning situation, automatically quickly distinguishing the network performance of active user, thus user node is converted to effective P2P node, the reliability of P2P network is ensured.

Above-described only presently preferred embodiments of the present invention, all any amendment, equivalent and improvement etc. made in the range of the spirit and principles in the present invention, should be included within the scope of the present invention.

Claims (8)

1. a log information network structure method based on P2P VOD system, it is characterised in that comprise the following steps:

Step 1: collect the log information of user；

Step 2: according to the log information of node, structure is with user as node, and the relation between user is limit, by node band The wide community network structure representing weight；

Step 3: community network structure is carried out community structure division according to LabelRank algorithm, obtains multiple community；Wherein, LabelRank algorithm comprises an operation operator: transmission operates；

Described transmission operation is particularly as follows: preserve label distribution at the vectorial P of each one 1 × n of node definition, and n is node Quantity；Define an adjacency matrix A and store network structure, then P_iC () represents the probability that node i belongs to label c, the collection of label Share C to represent, and the number of label is initialized as number n of node, calculate every time and be required for updating the vector of each node P, formula is as follows:

$P_i^{,}\left(c\right)=\underset{j\∈Nb\left(i\right)}{\mathrm{\Σ}}{P}_j\left(c\right)/k_i,\∀c\∈C\mathrm{...}\left(1\right)$

Wherein, Nb (i) is the set of the neighbor node of node i, k_i=| Nb (i) | is the number of nodes of neighbor node set, P '_i(c) generation Table updates the probability that posterior nodal point i belongs to label c, and being then followed by calculating can be expressed as by matrix A and vector P:

A×P.............(2)

Firstly the need of initialization vector P in transmission operation operator, initialized method particularly includes: the list of labels of each node Probability P_ijIt is initialized as the inverse of the weight sum of neighbor node；As follows:

P_ij=w/k_i............(3)

The node i that represents w arrives the weight of node j；

The vectorial P of each node is obtained after above-mentioned steps；

In the renewal process every time calculated of transmission operation operator, threshold value λ is set, when the probability of the affiliated label of node During more than λ, then retain, less than then giving up；And when label probability all of in node is both less than λ, then randomly select one Retain；

Step 4: the node in the community in step 3 is merged and processes to remove the invalid node in community.

Log information network structure method based on P2P VOD system the most according to claim 1, it is characterised in that Described LabelRank algorithm also includes an operation operator: expansion factor；

Described expansion factor is particularly as follows: use expansion factor Γ in vector P_inAffecting label transmission, wherein in is a reality Numerical value, by using expansion factor to carry out Decoupling network structure, calculates Γ every time_inP, P_icIn will be increased^thPower is as follows Shown in formula:

${\mathrm{\Γ}}_{in}{P}_i\left(c\right)=P_i{\left(c\right)}^{in}/\underset{j\∈C}{\mathrm{\Σ}}{P}_i{\left(j\right)}^{in}$

After calculating, it can make the numerical value of the label probability of high probability in node increase, and the numerical value of the label probability of low probability reduces.

Log information network structure method based on P2P VOD system the most according to claim 1, it is characterised in that Described LabelRank algorithm also includes an operation operator: cutting coefficient；

Described cutting coefficient is particularly as follows: introduce operator Φ in vector P_rDefinition threshold value r ∈ [0,1], was calculating If the label probability in vector P is less than r in journey, the label in vector is deleted.

Log information network structure method based on P2P VOD system the most according to claim 1, it is characterised in that Described LabelRank algorithm also includes an operation operator: condition updates；

Described condition updates particularly as follows: one condition of definition updates operator Θ_q, its implication is when only having node in algorithmic procedure Just update this node time unequal with the label of its neighbor node vector, be not the most updated operation；The most in each iteration, When node meets following equation, just it is updated operation；

$\underset{j\∈Nb\left(i\right)}{\mathrm{\Σ}}isSubset(C_i^{*},C_j^{*})\≤{\mathrm{qk}}_i$

It is the node i tag set of maximum probability, k in conventional calculation procedure_iRepresenting the degree of node i, q is a real number value, The span [0,1] of q.

Log information network structure method based on P2P VOD system the most according to claim 1, it is characterised in that Peer number in same group when described log information is ID, video URL, video playback mode, Connection Service device, Other IDs, user bandwidth in User IP, group.

Log information network structure method based on P2P VOD system the most according to claim 1, it is characterised in that Described step 4 is specially

The result that community structure in S41: obtaining step 3 divides；

S42: the effective node in the result divide community structure merges, and removes invalid node；

S43: the data after storage merges in the form of a file.

7. for realizing a system for the log information network structure method based on P2P VOD system described in claim 1, It is characterized in that, including following structure:

Log collection module, for collecting the log information of user；

Community network structure constructing module, for the log information according to user, the structure pass with user as node, between user System is limit, is represented the community network structure of weight by node bandwidth；

Community structure divides module, for community network structure being carried out community structure division according to LabelRank algorithm, obtains Multiple communities；

Merge module: the community structure division result obtained for community structure divides module merges process to remove community In invalid node.

Log information network structure system based on P2P VOD system the most according to claim 7, it is characterised in that Described merging module includes following submodule:

Result obtains submodule, divides the result of the community structure division that module obtains for obtaining community structure；

Merging submodule, the effective node in the result dividing community structure merges, and removes invalid node；

Sub module stored, the data after storage merges in the form of a file.