Method, apparatus, device and readable storage medium for reconstructing propagation network structure

Abstract

The present application relates to a method, apparatus, device, and readable storage medium for reconstructing a propagation network structure, including starting propagation from an infected node set to obtain infection status data of each node; constructing a target parent node set of each node based on the infection status data; updating an edge set based on the target parent node set and the node set to generate a target edge set; constructing a target network topology structure diagram based on the node set and the target edge set; if the number of times the target network topology structure diagram is constructed does not reach a preset number of constructions, re-screening multiple nodes based on the similarity between the network structures corresponding to the constructed target network topology structure diagram and the self-information of each node to update to the infected node set to generate a new infected node set for new propagation; otherwise, reconstructing the propagation network structure based on the final generated target network topology structure diagram. The present application can effectively solve the problem of limited coverage of reconstruction results and improve reconstruction efficiency and accuracy.

Description

Technical Field

[0001] The present application relates to the field of information dissemination technology, and in particular to a method, device, equipment and readable storage medium for reconstructing a dissemination network structure. Background Art

[0002] Information diffusion is ubiquitous in social networks. It is the exchange and transmission of information between individuals within society. Information diffusion pathways can be abstracted as networks, and these networks are present in all aspects of our lives, such as information dissemination on social networks and the mutual influence of individuals in interpersonal interactions. Analyzing how information diffuses within networks can help us understand the potential influence relationships between individuals, thus guiding related research. By analyzing how information diffuses and reconstructing the structure of diffusion networks, researchers can gain an intuitive understanding of the patterns of information transmission, enabling predictions about the future and potentially facilitating or preventing potential information diffusion processes. Reconstructing the diffusion network structure requires obtaining and analyzing historical data on diffusion outcomes to reconstruct the network's topology (i.e., the mutual influence relationships between the individual nodes that comprise the network).

[0003] The observation data used in existing reconstruction methods are mainly obtained through passive observation, without screening out data that is more helpful to the inference process. This may lead to certain limitations in the coverage of the obtained data, and then the inferred propagation network structure also has problems such as limited coverage, low accuracy and efficiency. Summary of the Invention

[0004] The present application provides a method, apparatus, device and readable storage medium for reconstructing a propagation network structure to solve the problems in related technologies of limited coverage of the propagation network structure, poor reconstruction accuracy and low efficiency caused by reconstructing the propagation network structure through observation data obtained through passive observation.

[0005] In a first aspect, a method for reconstructing a propagation network structure is provided, comprising the following steps:

[0006] Starting from an infected node set and propagating for a preset number of times, obtaining infection status data of each node, wherein the infected node set includes a plurality of nodes selected from the node set;

[0007] Construct the target parent node set of each node based on the infection status data;

[0008] Update the edge set based on the target parent node set and the node set to generate the target edge set;

[0009] Constructing a target network topology structure graph based on the node set and the target edge set, and determining whether the number of times the target network topology structure graph is constructed reaches a preset number of construction times;

[0010] If not, then based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, M nodes are selected from the node set as target initial infected nodes and updated to the infected node set to generate a new infected node set. Based on the new infected node set, the step of starting to propagate from the infected node set and propagating a preset number of times is performed, where M is a positive integer;

[0011] If so, the propagation network structure is reconstructed according to the target network topology diagram finally generated.

[0012] In some embodiments, before the step of starting to propagate from the infected node set and propagating a preset number of times, the method further includes:

[0013] Filter multiple nodes from the node set based on the initial infected node ratio and the total number of nodes;

[0014] The multiple nodes are added as initial infected nodes to the infected node set.

[0015] In some embodiments, the method of selecting M nodes from a node set as target initial infected nodes and updating the infected node set to generate a new infected node set based on the similarity between the network structures corresponding to the constructed target network topology structure graph and the self-information of each node to generate a new infected node set includes:

[0016] After each α propagation, the precision and recall between the network structures corresponding to the constructed target network topology graph are calculated based on the first calculation formula. The similarity between the network structures is represented by Precision and Recall, where α is a positive integer.

[0017] After each β propagation, the self-information of each node is calculated based on the second calculation formula, where β is a positive integer;

[0018] Based on the accuracy, recall and self-information, M nodes are selected from the node set as the target initial infection nodes;

[0019] Replace the nodes in the infected node set with the target initial infected node to generate a new infected node set;

[0020] The first calculation formula is:

[0021]

[0022]

[0023] Where N TP N represents the number of edges that exist in both the first network G' obtained by current inference and the second network G" obtained by the previous inference. FP N represents the number of edges that exist in G' but not in G", FN represents the number of edges that do not exist in G' but exist in G";

[0024] The second calculation formula is:

[0025] SI(v i )=-log2p(v i )

[0026]

[0027] Where, SI(v i ) represents node v i Self-information, p(v i ) represents node v i The probability of being selected as the initial infected node, N i It represents the time from the first propagation to the calculation of SI(v i ) up to node v i The number of times a node is selected as the initial infected node.

[0028] In some embodiments, selecting M nodes from a node set as target initial infection nodes based on accuracy, recall, and self-information includes:

[0029] The initial infected node ratio is updated according to the relationship between precision and recall and the preset adjustment value to obtain the target initial infected node ratio;

[0030] The value of M is determined based on the target initial infection ratio and the total number of nodes;

[0031] Sort the nodes in the node set in descending order of self-information to obtain a sorted queue;

[0032] The nodes ranked in the top M positions in the sorting queue are selected as target initial infection nodes.

[0033] In some embodiments, constructing a target parent node set for each node based on infection status data includes:

[0034] Calculate the activation mutual information between nodes based on the infection status data;

[0035] The target parent node set of each node is constructed based on the activation mutual information.

[0036] In some embodiments, calculating activation mutual information between nodes based on infection status data includes:

[0037] Calculate the occurrence probability P(X) of each node based on the infection status data i ) and the joint probability P(X i ,X j );

[0038] The activation mutual information between nodes is calculated according to the occurrence probability and joint probability by the following formula:

[0039]

[0040] in, MI(X i ,X j ) is the node v i and node v j The component of mutual information between infection states, P(X i ) is the node v i In state X i The probability of X i Represents node v i The infection status, P(X i ,X j ) represents node v i In state X i , while node v j In state X j The joint probability of AMI(X i ,X j ) represents the activation mutual information, MI(X i =0,X j =0) indicates that node v i and node v j The correlation component when both are in the uninfected state belongs to the positive infection association, MI(X i =1,X j =1) indicates node v i and node v j The correlation component when all are in the infected state belongs to positive infection association, |MI(X i =0,X j =1)| is node v i In the uninfected state, node v j The correlation component when in the infected state belongs to negative infection association, |MI(X i =1,X j =0)| is node v i In the infected state, node v j The correlation component when in the uninfected state is a negative infection association.

[0041] In some embodiments, updating the edge set based on the target parent node set and the node set to generate the target edge set includes:

[0042] Point each node in the target parent node set to the corresponding node in the node set, generating the corresponding target directed edge;

[0043] Add the target directed edge to the edge set to generate the target edge set.

[0044] In a second aspect, a propagation network structure reconstruction device is provided, comprising:

[0045] a propagation module, configured to propagate from an infected node set and propagate a preset number of times to obtain infection status data of each node, wherein the infected node set includes a plurality of nodes selected from the node set;

[0046] A construction module, which is used to construct a target parent node set for each node based on the infection status data;

[0047] An update module, which is used to update the edge set based on the target parent node set and the node set to generate a target edge set;

[0048] A graph construction module is used to construct a target network topology graph based on a node set and a target edge set, and to determine whether the number of times the target network topology graph has been constructed has reached a preset number of constructions;

[0049] a screening module, which is used to, if not, screen N nodes from the node set as target initial infected nodes based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, and update them to the infected node set to generate a new infected node set, so that the propagation module executes the step of starting propagation from the infected node set and propagating a preset number of times based on the new infected node set, where N is a positive integer;

[0050] The reconstruction module is used to reconstruct the propagation network structure according to the target network topology diagram finally generated if yes.

[0051] In a third aspect, a propagation network structure reconstruction device is provided, comprising: a memory and a processor, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the aforementioned propagation network structure reconstruction method.

[0052] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer storage medium stores a computer program, and when the computer program is executed by a processor, the aforementioned propagation network structure reconstruction method is implemented.

[0053] The present application provides a method, apparatus, device and readable storage medium for reconstructing a propagation network structure, comprising starting propagation from an infected node set and propagating a preset number of times to obtain infection status data of each node, wherein the infected node set includes multiple nodes screened from a node set; constructing a target parent node set of each node based on the infection status data; updating an edge set based on the target parent node set and the node set to generate a target edge set; constructing a target network topology structure graph based on the node set and the target edge set, and determining whether the number of times the target network topology structure graph is constructed reaches a preset number of constructions; if not, selecting M nodes from the node set as target initial infection nodes based on the similarity between the network structures corresponding to the constructed target network topology structure graph and the self-information of each node and updating them to the infected node set to generate a new infected node set, and executing the steps of starting propagation from the infected node set and propagating a preset number of times based on the new infected node set, wherein M is a positive integer; if so, reconstructing the propagation network structure according to the target network topology structure graph finally generated. Through this application, the idea of active learning is combined to optimize the initial infection nodes of the propagation process, avoid the random selection of initial nodes during network inference, and obtain data that helps to reconstruct the topological structure of the propagation network, so that each observation data obtained is as valuable as possible, bringing more information gain, thereby effectively solving the limitations of the coverage of the reconstruction results and improving the reconstruction efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0055] Figure 1 A flowchart of a method for reconstructing a propagation network structure provided in an embodiment of the present application;

[0056] Figure 2 A schematic diagram of the structure of a propagation network structure reconstruction device provided in an embodiment of the present application;

[0057] Figure 3 A structural diagram of a propagation network structure reconstruction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0059] The embodiments of the present application provide a method, apparatus, device and readable storage medium for reconstructing a propagation network structure, which can solve the problems in related technologies such as the limitation of the coverage of the propagation network structure, poor reconstruction accuracy and low efficiency caused by reconstructing the propagation network structure through observation data obtained through passive observation.

[0060] An embodiment of the present application provides a method for reconstructing the structure of a propagation network. In this embodiment, five artificial networks are generated using the LFR algorithm (as shown in Table 1, including networks Net1, Net2, Net3, Net4 and Net5, and the network average degree refers to the average number of edges connected to each point); the propagation data generation process of each network is as follows: assuming that the degree of influence between the nodes in the network is the same, the initial infected node ratio is set to 0.15, that is, at the beginning of the first propagation process, 15% of the nodes are in the "infected" state, and the subsequent infected node ratio is determined by the algorithm, and an independent cascade model (IC model) is used to simulate the process of infection propagation, and after each propagation simulation, the infection status of each node is recorded as a propagation trajectory data, and it is added to the observation data set.

[0061] Table 1 Experimental network

[0062] LFR Network Number Number of nodes Average Net1 100 4 Net2 150 4 Net3 200 4 Net4 250 4 Net5 300 4

[0063] See also Figure 1 As shown, a method for reconstructing a propagation network structure includes the following steps:

[0064] Step S10: Starting from the infected node set and propagating a preset number of times to obtain infection status data of each node, the infected node set includes multiple nodes screened from the node set;

[0065] For example, in order to more actively screen out the required information, thereby more accurately reconstructing the propagation network structure and reducing the limitations of the reconstruction results, this embodiment combines the ideas and methods of active learning. By optimizing the initial infected nodes of each propagation process, the random selection of initial nodes during network inference is avoided, and data that helps to reconstruct the topological structure of the propagation network is obtained.

[0066] It should be noted that this embodiment needs to define the potential influence relationship diagram (i.e., network topology diagram) in the propagation network as G(V,E), where V = {v1, v2, ..., v n}, which is a set of n nodes that make up the network; E = {e1, e2, …, e m}, which represents a set of m directed edges in the network, and the directed edge e(v i ,v j ) represents node v i For node v j Has an influential relationship, that is, when node v i When infected, there is a certain probability that node v j Infected, also known as node v i is node v j In addition, using the set C = {C 1 ,C 2 ,…,C r}Record the results after the observed r propagation processes, where, represents the infection status set of each node after the lth propagation process ends, The value is 0 or 1, indicating two states. If it is 0, it means that the node v i After the lth propagation process is completed, it is in an uninfected state. If it is 1, it means that the node v i After the lth transmission process is completed, it is in the infected state.

[0067] It should be understood that before each construction of the network topology structure diagram, a preset number of propagation processes will be carried out, for example, a composition will be performed every 10 propagations, that is, a network topology structure diagram will be constructed every 10 propagations. However, the specific number of propagations can be determined according to actual needs and is not limited here.

[0068] Before propagation, this embodiment needs to initialize an empty set of infected nodes A and an empty set of directed edges E, and perform screening of the initial infected nodes, that is, select multiple nodes from the node set V as initial infected nodes and add them to the infected node set A; then propagate from the infected node set and propagate a preset number of times to obtain the infection status data of each node (that is, whether each node is infected after each propagation), and then obtain the final infection status set C l , and add it to the historical observation data set C, where C l Represents the result of the lth transmission outbreak.

[0069] Furthermore, before the step of starting to propagate from the infected node set and propagating a preset number of times, the method further includes:

[0070] Filter multiple nodes from the node set based on the initial infected node ratio and the total number of nodes;

[0071] The multiple nodes are added as initial infected nodes to the infected node set.

[0072] For example, in this embodiment, before the first composition is performed, it is necessary to randomly select multiple nodes from the node set V as initial infection nodes according to the initial infection node ratio rate and the total number of nodes n in the propagation network and add them to the infection node set A, that is, randomly select rate×n nodes from the node set V to add to the infection node set A; where rate represents the ratio of the initial number of infected nodes to the total number of nodes at the beginning of each infection process; and n represents the total number of nodes in the propagation network.

[0073] It should be understood that if this is not the first time the mapping is being done, then infected node set A already contains multiple initial infected nodes. In this case, it is necessary to reselect multiple initial infected nodes from the node set based on the similarity between the network structures obtained after a preset number of propagations and the self-information of each node to replace the existing initial infected nodes in infected node set A, thereby forming a new infected node set A. Propagation then begins from this new infected node set A, thus starting a new mapping. It should be noted that the initial setting of the rate value can be determined based on actual needs. For example, the initial value of the rate can be set to 0.15, but subsequent updates and adjustments to the rate value will need to be determined based on the similarity between the network structures obtained after a preset number of propagations.

[0074] Step S20: constructing a target parent node set for each node based on the infection status data;

[0075] Exemplarily, in this embodiment, it should be understood that the parent node corresponding to each node can be obtained through the status data of whether each node is infected, and then the target parent node set of each node can be constructed.

[0076] Furthermore, the target parent node set of each node is constructed based on the infection status data, including:

[0077] Calculate the activation mutual information between nodes based on the infection status data;

[0078] The target parent node set of each node is constructed based on the activation mutual information.

[0079] For example, in this embodiment, to create a node v i The target parent node set is used as an example to illustrate: First, an empty set F is initialized. i , F i Represents node v iThe parent node set of each node can then be calculated based on the status data of whether each node is infected to obtain the activation mutual information AMI (X i ,X j ), and normalize all values of AMI<1; perform K-means algorithm (where K=2) on all activation mutual information values, fix the cluster center of one group to 1, and record the maximum activation mutual information value in the cluster with a mean close to 1 as λ; for node v i , if node v j (i≠j) and its corresponding activation mutual information AMI(X i ,X j ) is greater than λ, then the node v j Join node v i The parent node set F i Otherwise, it will not be added, and then the generated node v is constructed. i The target parent node set.

[0080] Furthermore, the calculation of activation mutual information between nodes based on infection status data includes:

[0081] Calculate the occurrence probability P(X) of each node based on the infection status data i ) and the joint probability P(X i ,X j );

[0082] The activation mutual information between nodes is calculated according to the occurrence probability and joint probability by the following formula:

[0083]

[0084] in, |MI(X i ,X j ) is the node v i and node v j The component of mutual information between infection states, P(X i ) is the node v i In state X i The probability of X i Represents node v i The infection status, P(X j ) is the node v j In state X j The probability of X j Represents node v j The infection status, P(X i ,X j ) represents node v i In state X i , while node vj In state X j The joint probability of AMI(X i ),X j ) represents the activation mutual information, MI(X i =0,X j =0) indicates that node v i and node v j The correlation component when both are in the uninfected state belongs to the positive infection association, MI(X i =1,X j =1) indicates node v i and node v j The correlation component when all are in the infected state belongs to positive infection association, |MI(X i =0,X j =1)| is node v i In the uninfected state, node v j The correlation component in the infected state belongs to the negative infection association, |MI(X i =1,X j =0)| is node v i In the infected state, node v j The correlation component when in the uninfected state is a negative infection association.

[0085] For example, in this embodiment, each node in the propagation network is considered in turn, and the activation mutual information between each node and its corresponding nodes is calculated. i and node v j The activation mutual information between the two is used as an example to illustrate: According to the infection status data, the node v is calculated by the following formula i The probability of occurrence P(X i ) and its relationship with each remaining node v j The joint probability P(X i ,X j );

[0086]

[0087]

[0088] Where, X i and X j Represents node v i and node v j The infection status of the node (a value of 0 indicates that the node is not infected, and a value of 1 indicates that the node is infected), N i Represents the node v observed in all infection results i In X iThe number of states, N i,j Represents the node v observed in all infection results i In X i state and node v j In X j The number of states, N represents the total number of observation data records. j ) calculation method and principle are similar to P(X i ) is the same as ).

[0089] According to P(X i )、P(X j ) and P(X i ,X j ) and the node v can be calculated by the following formula i and node v j The mutual information of activations between

[0090]

[0091] in,

[0092] Step S30: updating the edge set based on the target parent node set and the node set to generate a target edge set;

[0093] Exemplarily, in this embodiment, it should be understood that multiple directed edges can be generated based on the parent nodes in the target parent node set and each node in the node set, and the target edge set can be generated by updating the edge set through the multiple directed edges.

[0094] Furthermore, the edge set is updated based on the target parent node set and the node set to generate a target edge set, including:

[0095] Point each node in the target parent node set to the corresponding node in the node set, generating the corresponding target directed edge;

[0096] Add the target directed edge to the edge set to generate the target edge set.

[0097] For example, in this embodiment, it can be understood that the target edge set is generated by pointing each parent node in the target parent node set to the corresponding node in the node set, forming corresponding directed edges and adding them to the edge set E. For example, the node v i Each parent node v in the corresponding target parent node set j Points to the node v in the node collection i , you can generate the node v i Corresponding multiple directed edges, similarly generate directed edges corresponding to other nodes.

[0098] Step S40: constructing a target network topology graph based on the node set and the target edge set, and determining whether the number of times the target network topology graph has been constructed has reached a preset number of times;

[0099] For example, in this embodiment, it should be understood that by forming the node set V and the target edge set E into an influence relationship graph G, the target network topology structure graph obtained by the current iteration can be constructed; at this time, it is necessary to further determine whether the upper bound of the iteration is reached, that is, whether the number of times the target network topology structure graph is constructed reaches the preset number of constructions, and based on the judgment result, determine whether it is necessary to re-screen new initial infected nodes for starting new propagation.

[0100] Step S50: If not, then based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, N nodes are selected from the node set as target initial infected nodes and updated to the infected node set to generate a new infected node set. Based on the new infected node set, the step of starting propagation from the infected node set and propagating a preset number of times is performed, where N is a positive integer;

[0101] For example, in this embodiment, if the current number of iterations has not reached the upper bound, mapping must continue. This requires selecting new initial infected nodes to initiate new propagation and construct a new target network topology. Specifically, after each preset number of propagations, self-information for each node is generated, and mapping is performed again to obtain the corresponding network topology. Based on the similarity between the network structure constructed in the current network topology and the network structure constructed in the previous network topology, as well as the self-information of each node, initial infected nodes for the next propagation can be again selected from the node set.

[0102] Furthermore, the method selects M nodes from the node set as target initial infected nodes and updates them to the infected node set based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, so as to generate a new infected node set, including:

[0103] After each α propagation, the precision and recall between the network structures corresponding to the constructed target network topology graph are calculated based on the first calculation formula. The similarity between the network structures is represented by Precision and Recall, where α is a positive integer.

[0104] After each β propagation, the self-information of each node is calculated based on the second calculation formula, where β is a positive integer;

[0105] Based on the accuracy, recall and self-information, M nodes are selected from the node set as the target initial infection nodes;

[0106] Replace the nodes in the infected node set with the target initial infected node to generate a new infected node set;

[0107] The first calculation formula Evaluation(G',G") is:

[0108]

[0109]

[0110] Where N TP N represents the number of edges that exist in both the first network G' obtained by current inference and the second network G" obtained by the previous inference. FP N represents the number of edges that exist in G' but not in G", FN represents the number of edges that do not exist in G' but exist in G";

[0111] The second calculation formula is:

[0112] SI(v i )=-log2p(v i )

[0113]

[0114] Where, SI(v i ) represents node v i Self-information, p(v i ) represents node v i The probability of being selected as the initial infected node, N i It represents the time from the first propagation to the calculation of SI(v i ) up to node v i The number of times a node is selected as the initial infected node.

[0115] For example, in this embodiment, after each α propagation process, the Evaluation(G i ,G i+α) evaluates the similarity between the network structures predicted at different stages, and adjusts the proportion of initial infected nodes in the next infection process according to the evaluation results. Specifically, the Precision and Recall between the network structures are calculated by the first calculation formula above (i.e., Evaluation(Gi',Gi”)), and then the similarity between the network structures is evaluated based on the size relationship between Precision and Recall; after each β propagation process, the self-information SI of each node is calculated based on the relevant information of each node selected as the initial infected node and the second calculation formula above; through the similarity between the network structures and the self-information of each node, multiple nodes can be screened from the node set as the target initial infected nodes to update to the infected node set, thereby generating a new infected node set. It should be noted that the specific values of α and β can be determined according to actual needs, such as α∈[3,10], β∈[5,15].

[0116] Furthermore, M nodes are selected from the node set as target initial infection nodes based on accuracy, recall rate, and self-information, including:

[0117] The initial infected node ratio is updated according to the relationship between precision and recall and the preset adjustment value to obtain the target initial infected node ratio;

[0118] The value of M is determined based on the target initial infection ratio and the total number of nodes;

[0119] Sort the nodes in the node set in descending order of self-information to obtain a sorted queue;

[0120] The nodes ranked in the top M positions in the sorting queue are selected as target initial infection nodes.

[0121] Exemplarily, in this embodiment, the initial infection node ratio of the next infection process is adjusted according to the evaluation result of the similarity degree, that is, when Precision>Recall, the target initial infection node ratio rate' is set to the initial infection node ratio rate+preset adjustment value Δ, otherwise, rate' is set to rate-Δ; wherein, the value range of rate can be [0.05,0.25]. If it exceeds the boundary value, the corresponding boundary value is taken, and Δ can be taken as 0.005. It should be noted that the values of rate and Δ can be determined according to actual needs and are not limited here; since the total number of nodes in the propagation network is fixed, after determining the target initial infection node rate', it can be determined how many nodes need to be re-screened from the node set as the target initial infection nodes based on the total number of nodes n in the propagation network and the target initial infection node rate', that is, M=rate'×n, where M is the number of target initial infection nodes.

[0122] The nodes in the node set are sorted in descending order of self-information. The M nodes with the largest self-information are selected from the node set as the target initial infected nodes to replace the nodes in the infected node set A, thereby generating a new infected node set A'. Propagation then begins from this new infected node set A', leading to a new round of mapping. By optimizing the initial infected nodes during the propagation process, random selection of initial nodes during network inference is avoided, thereby obtaining data that helps reconstruct the topology of the propagation network.

[0123] Step S60: If yes, reconstruct the propagation network structure according to the target network topology diagram finally generated.

[0124] For example, in this embodiment, if the current number of iterations has reached the upper bound, the propagation network structure is directly reconstructed based on the target network topology diagram generated in the last round, thereby generating a reconstructed propagation network. This shows that this embodiment, when reconstructing the propagation network topology, can incorporate active learning to select better initial infected nodes for the network structure inference algorithm. This can effectively address issues such as the excessive number of iterations required by inference algorithms based on passive observation and the limited coverage of the reconstruction results, thereby improving the efficiency and accuracy of propagation network structure reconstruction.

[0125] In summary, this application introduces the concept of active learning into the inference of propagation network structure, utilizing historical information for active learning and optimizing the initial infected nodes in the propagation process to avoid random selection of initial nodes during network inference, thus bringing more information gain. Therefore, through this application, the number of iterations in the inference process can be effectively reduced, the coverage limitations of the reconstruction results can be reduced, and the inference efficiency and accuracy of the reconstructed network can be improved.

[0126] See also Figure 2 As shown, an embodiment of the present application provides a propagation network structure reconstruction device, including:

[0127] a propagation module, configured to propagate from an infected node set and propagate a preset number of times to obtain infection status data of each node, wherein the infected node set includes a plurality of nodes selected from the node set;

[0128] A construction module, which is used to construct a target parent node set for each node based on the infection status data;

[0129] An update module, which is used to update the edge set based on the target parent node set and the node set to generate a target edge set;

[0130] A graph construction module is used to construct a target network topology graph based on a node set and a target edge set, and to determine whether the number of times the target network topology graph has been constructed has reached a preset number of constructions;

[0131] a screening module, which is used to, if not, screen N nodes from the node set as target initial infected nodes based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, and update them to the infected node set to generate a new infected node set, so that the propagation module executes the step of starting propagation from the infected node set and propagating a preset number of times based on the new infected node set, where N is a positive integer;

[0132] The reconstruction module is used to reconstruct the propagation network structure according to the target network topology diagram finally generated if yes.

[0133] Furthermore, the propagation module is further configured to:

[0134] Filter multiple nodes from the node set based on the initial infected node ratio and the total number of nodes;

[0135] The multiple nodes are added as initial infected nodes to the infected node set.

[0136] Furthermore, the screening module is specifically used to:

[0137] After each α propagation, the precision and recall between the network structures corresponding to the constructed target network topology graph are calculated based on the first calculation formula. The similarity between the network structures is represented by Precision and Recall, where α is a positive integer.

[0138] After each β propagation, the self-information of each node is calculated based on the second calculation formula, where β is a positive integer;

[0139] Based on the accuracy, recall and self-information, M nodes are selected from the node set as the target initial infection nodes;

[0140] Replace the nodes in the infected node set with the target initial infected node to generate a new infected node set;

[0141] The first calculation formula is:

[0142]

[0143]

[0144] Where N TP N represents the number of edges that exist in both the first network G' obtained by current inference and the second network G" obtained by the previous inference. FP N represents the number of edges that exist in G' but not in G", FN represents the number of edges that do not exist in G' but exist in G";

[0145] The second calculation formula is:

[0146] SI(v i )=-log2p(v i )

[0147]

[0148] Where, SI(v i ) represents node v i Self-information, p(v i ) represents node v i The probability of being selected as the initial infected node, N i It represents the time from the first propagation to the calculation of SI(v i ) up to node v i The number of times a node is selected as the initial infected node.

[0149] Furthermore, the screening module is further configured to:

[0150] The initial infected node ratio is updated according to the relationship between precision and recall and the preset adjustment value to obtain the target initial infected node ratio;

[0151] The value of M is determined based on the target initial infection ratio and the total number of nodes;

[0152] Sort the nodes in the node set in descending order of self-information to obtain a sorted queue;

[0153] The nodes ranked in the top M positions in the sorting queue are selected as target initial infection nodes.

[0154] Furthermore, the building blocks are specifically used to:

[0155] Calculate the activation mutual information between nodes based on the infection status data;

[0156] The target parent node set of each node is constructed based on the activation mutual information.

[0157] Furthermore, the building block is further configured to:

[0158] Calculate the occurrence probability P(X) of each node based on the infection status data i ) and the joint probability P(X i ,X j );

[0159] The activation mutual information between nodes is calculated according to the occurrence probability and joint probability by the following formula:

[0160]

[0161] in, MI(X i ,X j ) is the node v i and node v j The component of mutual information between infection states, P(X i ) is the node v i In state X i The probability of X i Represents node v i The infection status, P(X i ,X j ) represents node v i In state X i , while node v j In state X j The joint probability of AMI(X i ,X j ) represents the activation mutual information, MI(X i =0,X j =0) indicates that node v i and node v j The correlation component when both are in the uninfected state belongs to the positive infection association, MI(X i =1,X j =1) indicates node v i and node v j The correlation component when all are in the infected state belongs to positive infection association, |MI(X i =0,Xj =1)| is node v i In the uninfected state, node v j The correlation component in the infected state belongs to the negative infection association, |(MI(X i =1,X j =0)| is node v i In the infected state, node v j The correlation component when in the uninfected state is a negative infection association.

[0162] Furthermore, the update module is specifically used to:

[0163] Point each node in the target parent node set to the corresponding node in the node set, generating the corresponding target directed edge;

[0164] Add the target directed edge to the edge set to generate the target edge set.

[0165] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and units can refer to the corresponding processes in the aforementioned embodiment of the propagation network structure reconstruction method, and will not be repeated here.

[0166] The apparatus provided in the above embodiment can be implemented in the form of a computer program. The computer program can be used in Figure 3 The propagation network structure shown runs on the reconstructed device.

[0167] An embodiment of the present application also provides a propagation network structure reconstruction device, comprising: a memory, a processor, and a network interface connected via a system bus, wherein at least one instruction is stored in the memory, and at least one instruction is loaded and executed by the processor to implement all or part of the steps of the aforementioned propagation network structure reconstruction method.

[0168] Among them, the network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0169] The processor may be a CPU, other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor, or any conventional processor. The processor is the control center of a computer device, connecting various parts of the entire computer device using various interfaces and lines.

[0170] The memory can be used to store computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application required for a function (such as a video playback function, an image playback function, etc.), etc.; the data storage area can store data created based on the use of the mobile phone (such as video data, image data, etc.). In addition, the memory can include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0171] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, all or part of the steps of the aforementioned propagation network structure reconstruction method are implemented.

[0172] The embodiments of the present application implement all or part of the aforementioned processes, and may also be completed by instructing related hardware through a computer program. The computer program may be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of each of the above methods may be implemented. Among them, the computer program includes computer program code, and the computer program code may be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0173] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, servers, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage and optical storage, etc.) containing computer-usable program code.

[0174] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0175] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0176] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the broadest scope consistent with the principles and novel features of the present application.

Claims

1. A method for reconstructing a propagation network structure, wherein the propagation network is applied to information propagation on a social network, characterized in that: The following steps are involved: Starting from an infected node set and propagating for a preset number of times, obtaining infection status data of each node, wherein the infected node set includes a plurality of nodes selected from the node set; Construct the target parent node set of each node based on the infection status data; Update the edge set based on the target parent node set and the node set to generate the target edge set; Constructing a target network topology structure graph based on the node set and the target edge set, and determining whether the number of times the target network topology structure graph is constructed reaches a preset number of constructions; If not, then based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, M nodes are selected from the node set as target initial infected nodes and updated to the infected node set to generate a new infected node set. Based on the new infected node set, the step of starting to propagate from the infected node set and propagating a preset number of times is performed, where M is a positive integer; If so, the propagation network structure is reconstructed according to the target network topology diagram finally generated; Before the step of starting to propagate from the infected node set and propagating a preset number of times, the method further includes: Filtering multiple nodes from the node set based on the initial infected node ratio and the total number of nodes; Adding the multiple nodes as initial infected nodes to the infected node set; The method of selecting M nodes from the node set as target initial infected nodes and updating them to the infected node set to generate a new infected node set based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, includes: Every time you pass After the first propagation, the accuracy of the network structure corresponding to the constructed target network topology diagram is calculated based on the first calculation formula. and recall ,pass and Characterizes the similarity between network structures, where is a positive integer; Every time you pass After the first propagation, the self-information of each node is calculated based on the second calculation formula, where is a positive integer; Based on the accuracy, recall and self-information, M nodes are selected from the node set as the target initial infection nodes; Replace the nodes in the infected node set with the target initial infected node to generate a new infected node set; The first calculation formula is: Where, Represents the first network currently inferred And the second network obtained by the last inference The number of edges that exist in Indicates exists in but The number of edges that do not exist in , Indicates does not exist in The number of edges in ; The second calculation formula is: Where, Representation node Self-information, Representation node The probability of being selected as the initial infected node, Indicates the time from the first propagation to the calculation So far, the node The number of times a node is selected as the initial infected node.

2. The method for reconstructing a propagation network structure according to claim 1, wherein: The method of selecting M nodes from the node set as target initial infection nodes based on the accuracy, recall rate and self-information includes: The initial infected node ratio is updated according to the relationship between precision and recall and the preset adjustment value to obtain the target initial infected node ratio; The value of M is determined based on the target initial infection ratio and the total number of nodes; Sort the nodes in the node set in descending order of self-information to obtain a sorted queue; The nodes ranked in the top M positions in the sorting queue are selected as target initial infection nodes.

3. The method for reconstructing a propagation network structure according to claim 1, wherein: The target parent node set of each node is constructed according to the infection status data, including: Calculate the activation mutual information between nodes based on the infection status data; The target parent node set of each node is constructed based on the activation mutual information.

4. The method for reconstructing a propagation network structure according to claim 3, wherein: The step of calculating activation mutual information between nodes based on infection status data includes: Calculate the occurrence probability of each node based on infection status data and the joint probability between each node ; The activation mutual information between nodes is calculated according to the occurrence probability and joint probability by the following formula: in, , For nodes and nodes The component of mutual information between infection states, For nodes In state The probability of Representation node infection status, Representation node In state , while nodes In state The joint probability of represents the activation mutual information, Representation node and nodes The correlation component when both are in the uninfected state belongs to the positive infection association. Representation node and nodes The correlation component when both are in the infected state belongs to the positive infection association. For nodes In the uninfected state, the node The correlation component when in the infected state belongs to negative infection association, For nodes In the infected state, the node The correlation component when in the uninfected state is a negative infection association.

5. The method for reconstructing a propagation network structure according to claim 1, wherein: The updating of the edge set based on the target parent node set and the node set to generate the target edge set includes: Point each node in the target parent node set to the corresponding node in the node set, generating the corresponding target directed edge; Add the target directed edge to the edge set to generate the target edge set.

6. A device for reconstructing a communication network structure, wherein the communication network is applied to information dissemination on a social network, characterized in that: include: a propagation module, configured to propagate from an infected node set and propagate a preset number of times to obtain infection status data of each node, wherein the infected node set includes a plurality of nodes selected from the node set; A construction module, which is used to construct a target parent node set for each node based on the infection status data; An update module, which is used to update the edge set based on the target parent node set and the node set to generate a target edge set; A graph construction module is used to construct a target network topology graph based on a node set and a target edge set, and to determine whether the number of times the target network topology graph has been constructed has reached a preset number of constructions; a screening module, which is used to, if not, screen N nodes from the node set as target initial infected nodes based on the similarity between the network structures corresponding to the constructed target network topology diagram and the self-information of each node, and update them to the infected node set to generate a new infected node set, so that the propagation module executes the step of starting propagation from the infected node set and propagating a preset number of times based on the new infected node set, where N is a positive integer; A reconstruction module is used to reconstruct the propagation network structure according to the target network topology diagram finally generated if the propagation network structure is correct. The propagation module is further configured to: select multiple nodes from the node set according to the initial infected node ratio and the total number of nodes; and add the multiple nodes as initial infected nodes to the infected node set; The screening module is specifically used for: Every time you pass After the first propagation, the accuracy of the network structure corresponding to the constructed target network topology diagram is calculated based on the first calculation formula. and recall ,pass and Characterizes the similarity between network structures, where is a positive integer; Every time you pass After the first propagation, the self-information of each node is calculated based on the second calculation formula, where is a positive integer; Based on the accuracy, recall and self-information, M nodes are selected from the node set as the target initial infection nodes; Replace the nodes in the infected node set with the target initial infected node to generate a new infected node set; The first calculation formula is: Where, Represents the first network currently inferred And the second network obtained by the last inference The number of edges that exist in Indicates exists in but The number of edges that do not exist in , Indicates does not exist in The number of edges in ; The second calculation formula is: Where, Representation node Self-information, Representation node The probability of being selected as the initial infected node, Indicates the time from the first propagation to the calculation So far, the node The number of times a node is selected as the initial infected node.

7. A propagation network structure reconstruction device, characterized in that: include: A memory and a processor, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the propagation network structure reconstruction method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for reconstructing the propagation network structure according to any one of claims 1 to 5 is implemented.

Images