K-edge connected component mining system and mining method applied to social networks

By adopting a k-edge connected component mining system with local edge connection detection and segmentation tree pruning strategy in social networks, the problem of incorrect calculation and low calculation efficiency in the mining process of k-edge connected component in the existing technology is solved, and high Accuracy and high efficiency mining effect.

CN111899117BActive Publication Date: 2024-11-01ZHEJIANG LAB
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010747667.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-07-29
Publication Date
2024-11-01
Estimated Expiration
2040-07-29

AI Technical Summary

Technical Problem

The prior art has problems of incorrect calculation and low computing efficiency in the mining process of k-edge-side connectivity components in social networks, especially in the lack of effective pruning conditions in the process of node merging and graph segmentation, resulting in large amounts of calculations and many iterations.

Method used

Adopt a k-edge connection component mining system based on local edge connection degree detection, including traversal sorting module, graph segmentation module, local edge connection degree detection module, node merging module and sub-graph pruning module, the maximum through semantic perception Adjacent search sorting strategy and hierarchical graph segmentation calculations, extract the k-core structure for detection, and node merging and sub-graph pruning are performed based on the detection results to reduce unnecessary calculations.

Benefits of technology

The accuracy and efficiency of k-edge connected component mining is improved, the number of iterations is reduced, the calculation speed and accuracy is improved, and the calculation amount is reduced through local edge connection detection and segmentation tree pruning strategies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111899117B_ABST
    Figure CN111899117B_ABST
Patent Text Reader

Abstract

The present invention provides a k-edge connected component mining system based on local edge connectivity detection for social networks, which includes a hierarchical graph segmentation calculation module, a traversal sorting module, a graph segmentation module, a local edge connectivity detection module, a node merging module and a subgraph pruning module. The hierarchical graph segmentation calculation module maintains a supernode index. The traversal sorting module initializes a node sequence and traverses the nodes of the target graph. The graph segmentation module segments the target graph or subgraph according to the sorting state of the node sequence. The local edge connectivity detection module detects the edge connectivity value of the k-core structure of the current local subgraph. The node merging module confirms whether to merge nodes based on the detection result of the edge connectivity value. The subgraph pruning module performs subgraph deletion and pruning operations based on the calculation result of the node merging module. At the same time, the present invention also provides a mining method using the above-mentioned mining system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of social network, and in particular to a k-edge connected component mining system and mining method based on local edge connectivity detection applied to social network. Background Art

[0002] Graphs in computer science are an abstract data structure with a wide range of applications. Many complex relationships in the real world can be represented by graphs, such as the molecular structure of proteins, social networks, circuit wiring, etc. Graph-related applications cover many fields. Emerging social network technology analysis and semantic Web technology analysis also use graph data structures to solve complex relationships between multiple entities.

[0003] With the development trend of big data, the application of social network technology is becoming more and more extensive. Social networks often require precise analysis of the relationship between multiple entities, such as Weibo, WeChat Moments, ResearchGate academic network and other fields. Users are used as entities and the relationship between entities is used as the degree of association. From the perspective of abstract data structure in computer science, it is defined that each node in social network technology corresponds to a user, the edge between the nodes corresponds to the connection relationship between users, and the two ends of the edge correspond to two nodes respectively. By analyzing the complex relationship between the nodes and the edges, the structural characteristics of a social network can be described.

[0004] The "community structure characteristics" in complex social networks refer to the fact that "nodes in the same community are closely connected to each other, while nodes in different communities are sparsely connected to each other" in complex networks. With the continuous expansion and development of application fields, community structure has different new connotations: communities in social networks represent people with certain similar characteristics. The so-called community refers to a small group of users that are locally clustered. Dense subgraphs are used to represent locally clustered users (communities). Among them, the important indicator for measuring the local clustering of users is the edge connectivity of the subgraph. The k-edge connected component is defined as a local clustered user group with a correlation of not less than k, that is, a community. In this community, if any k-1 relationships between user pairs are disconnected, users in the community can still find another user through at least one relationship path. The purpose of k-edge connected component mining is to detect and reveal the inherent community structure in complex networks. The research on this issue has very important theoretical and practical significance and has been widely used.

[0005] According to the existing graph theory research, the following definitions and theorems are given:

[0006] Given an undirected graph G=(V,E), where V is a node set, E is an edge set, and the edge set E is a set of unordered tuples consisting of elements in the node set V.

[0007] Node merging: Given two nodes v,u∈V, then e v,u ∈E means that there is an edge connecting v and u in the graph G. If the node set Perform a merge operation on the nodes in S, that is, construct a super node v S , for the edge connecting the nodes inside the node set S and the nodes outside the node set S Use supernode v S Replace the node v, and replace the nodes {v|v∈S} and edges {e v,u |v∈S,u∈S} are all deleted.

[0008] Minimum cut: Assume that the node set SΥT = V, Then C = (S, T) represents the set of edges connecting nodes in S and nodes in T, which is a cut. Among all the cuts in graph G, the one with the smallest weight is the minimum cut.

[0009] Minimum st cut: Given two nodes s, t∈V, if a cut Cs, t = (S, T), where s∈S, t∈T, then Cs, t is a st cut. Among all st cuts, the one with the smallest weight is called the minimum st cut.

[0010] Max st flow: Given two nodes s, t∈V, each edge has a specified capacity, the path from s to t with the maximum capacity is the maximum st flow.

[0011] Max Flow Min Cut Theorem: Given two nodes s, t∈V, the minimum st cut is equal to the maximum st flow.

[0012] k-core: A subgraph where the degree of any node is not less than k.

[0013] The k-core lemma: a k-edge connector Figure 1 It must be a k-core, but the opposite is not necessarily true.

[0014] k-core component: If the degree of each node of a subgraph is not less than k, it means that the subgraph is a k-core. If a k-core is not contained by other k-cores, it is a k-core connected component.

[0015] k-edge connected component: A subgraph is k-edge connected if its minimum cut is k. A k-edge connected subgraph is a k-edge connected component if it is not contained by any other k-edge connected subgraph.

[0016] In the prior art, there is a node merging and graph splitting framework, which uses a Maximum Adjacency Search (MAS) sorting strategy to traverse and sort the node set of the target graph. The working principle of the MAS sorting strategy is: given an undirected graph G = (V, E), the undirected graph G maintains a set node sequence, initially a node is randomly selected from the node set V to join the set node sequence L, and then each time the node with the largest edge connection with the current node sequence L is selected from the remaining node set VL and added to the tail of L until all the remaining nodes are added to the node sequence L.

[0017] According to the existing MAS theoretical research, the following theorems and inferences can be obtained:

[0018] MAS Theorem: The global minimum st-cut of the last two nodes s and t in the node sequence generated by the MAS sorting strategy is the cut C(V\{t},{t}) between t and other nodes.

[0019] MAS inference: When generating the node sequence L using the MAS sorting strategy, if the degree of connectivity between the newly added node t and the node sequence L is x, then the global minimum st-cut between t and the last node s in the current L is x.

[0020] In the above implementation, the prior art uses the MAS sorting strategy to traverse and sort the nodes of the target graph, and mines out st node pairs with a minimum cut less than k (hereinafter referred to as "st node pairs"), and merges the st node pairs as nodes belonging to the same k-edge connected component. However, in some cases, the st node pairs do not belong to the same k-edge connected component, so directly merging the st node pairs will lead to erroneous calculation results.

[0021] Please refer to Figure 1 and Figure 2 ,in Figure 1 It is a schematic diagram of a target graph structure in the prior art. Figure 2 Yes Figure 1 Schematic diagram of the principle of traversal and sorting of the target graph structure shown. In the target graph 10, k=3 is set, then there is only one k-edge connected component {b, g, h, i} in the target graph. The node sequence L={a, b, c, d, e, f, g, h, i} generated by the MAS sorting strategy, although according to MAS inference, e and f are a node pair with a global st cut equal to 3, that is, e and f are an st node pair, but e and f do not belong to any k-edge connected component. In the prior art, the DecB-LMSD method is used to construct the node sequence L for the first time. After judging that e and f are an st node pair according to MAS inference, e and f are merged, resulting in an erroneous result.

[0022] In addition, in any traversal of the maximum adjacency search, the prior art will definitely produce at least one of the two situations of subgraph splitting or node merging. For example, a target graph containing multiple nodes is provided. When the MAS sorting strategy traverses to the last node t of the target graph, if the degree of connection between node t and the previous node sequence L is less than k, node t and L will be split into two subgraphs. Otherwise, if the degree of connection between node t and the current node sequence L is greater than or equal to k, node t will be merged with the last node s in the current L to generate a new node sequence, which includes node t. Therefore, the k-edge connected component mining method in the prior art needs to split the target graph into the smallest granularity, that is, each leaf node of the segmentation tree is either a k-edge connected component or a single node. When the subgraph does not meet the k-edge connected component or single node conditions, the segmentation operation will be continued in a loop, which results in a large amount of calculation for constructing the segmentation tree. For subgraphs composed of nodes that do not belong to any k-edge connected component, the prior art lacks effective pruning conditions to delete them in advance.

[0023] exist Figure 1 In the target graph 10 shown, k=3 is set, and the node set {b, g, h, i} is a k-edge connected component. However, in the process of constructing the DecB-LMSD method, only one new node is merged in each round of traversal, and the k-edge connected component of the four nodes {b, g, h, i} is completely merged after three rounds of traversal. It can be seen that in the above-mentioned method of mining k-edge connected components in the prior art, each traversal can only merge a part of the k-edge connected component, and a k-edge connected component can only be completely merged after multiple rounds of traversal, and the calculation efficiency is relatively low.

[0024] In view of the above defects, the present invention provides an approximate k-edge connected component mining algorithm for social networks, which has both efficiency and high accuracy, and the algorithm can converge within a shorter number of iterations. Summary of the invention

[0025] The present invention aims to provide a k-edge connected component mining system based on local edge connectivity detection which is applied to social networks quickly and with high accuracy.

[0026] At the same time, a k-edge connected component mining method using the above-mentioned k-edge connected component mining system based on local edge connectivity detection is also provided.

[0027] A k-edge connected component mining system based on local edge connectivity detection for social networks includes a traversal sorting module, a graph segmentation module, a local edge connectivity detection module, a node merging module and a graph subgraph pruning module. The traversal sorting module initializes a node sequence, receives a target graph including multiple nodes, and uses a MAS sorting strategy to traverse multiple nodes of the target graph to construct a new node sequence. The graph segmentation module confirms whether to segment the target graph to obtain a subgraph after segmentation based on the relationship between the sorting state of the node sequence constructed by the traversal sorting module and the set edge connectivity value. The local edge connectivity detection module extracts a k-core structure from the local target graph for detection, and determines the edge connectivity value of the k-core structure based on the detection result. The node merging module confirms whether to merge nodes based on the relationship between the detection result of the local edge connectivity detection module and the set edge connectivity value.

[0028] Furthermore, the mining system also includes a hierarchical graph partitioning calculation module for maintaining a supernode index.

[0029] Furthermore, the traversal sorting model uses a semantically-aware MAS sorting strategy to traverse the nodes of the target graph and construct a new node sequence.

[0030] Furthermore, it also includes a subgraph pruning module, which confirms whether to perform a subgraph deletion and pruning operation based on the calculation result of the node merging module.

[0031] A k-edge connected component mining method based on local edge connectivity detection applied to a social network comprises the following steps: providing a traversal sorting module, the traversal sorting module initializing a node sequence; receiving a target graph comprising a plurality of nodes, the traversal sorting module traversing the nodes of the target graph using a semantically-aware MAS sorting strategy and constructing a new node sequence; providing a graph segmentation module, the graph segmentation module performing a graph segmentation operation according to the sorting state of the node sequence constructed by the traversal sorting module; providing a local k-edge connectivity detection module, performing a local edge connectivity detection on the target graph; providing a node merging module, the node merging module performing a node merging operation according to the detection result of the local k-edge connectivity detection module to obtain a subgraph set.

[0032] Furthermore, before traversing by the traversal sorting module, the following steps are also included: providing a hierarchical graph partitioning calculation module, wherein the hierarchical graph partitioning calculation module maintains a super node index.

[0033] Furthermore, after performing the merging operation on the nodes, the method further includes a pruning and screening step for the subgraph in the target graph.

[0034] Furthermore, the step of constructing a node sequence using the semantically-aware maximum adjacency search sorting strategy also includes the following steps: initializing an empty node sequence; selecting a node with the largest connection with the node set of the initialized node sequence from the unsorted nodes and placing it at the end of the node sequence; when there are multiple nodes in the unsorted nodes that have the same maximum edge connectivity as the generated node sequence, the principle of "the most recently updated candidate node is given priority" is used to select the next node.

[0035] Furthermore, when constructing a node sequence, for each node that will be added to the node sequence after being selected, the edge connectivity value of the node and the current node sequence before it is added is calculated, and the relationship between the connectivity value and the set value k is determined. When the connectivity value is greater than or equal to the set value k, a local edge connectivity detection operation is performed to detect the edge connectivity between the node and the node before it.

[0036] Furthermore, when the edge connectivity values ​​of a plurality of consecutive nodes and their previous node sequences are greater than or equal to a set value k, a local edge connectivity detection operation is performed only when the last node in the consecutive nodes is added to the node sequence.

[0037] Furthermore, when performing the local edge connectivity detection operation, the following steps are included:

[0038] Define a local subgraph: define the node that is newly added to the node sequence as a t-node, and use the connected component formed by the t-node and its previous nodes as the current local subgraph;

[0039] Extracting a k-core structure to obtain an ST node pair: extracting a k-core structure containing a t node from the current local subgraph. If the k-core structure is empty, the local edge connectivity detection operation is terminated. If the k-core structure is not empty, the node farthest from the t node is selected in the k-core structure as an S node to obtain an ST node pair.

[0040] Calculate the minimum ST cut: calculate the maximum ST flow to obtain the minimum ST cut of the k-core structure;

[0041] Merge nodes: Determine the relationship between the minimum ST cut value and the set value. If the minimum ST cut value is greater than or equal to the set value, merge all the nodes in the k-core structure in the target graph, and record the merged supernodes and the corresponding original nodes in the supernode index, and then end the local edge connectivity detection operation; if the minimum ST cut value is less than the set value, divide the k-core structure into two subgraphs by deleting the edges included in the minimum ST cut, and use the subgraph containing the t node in the two subgraphs as the current local subgraph, and continue to extract the k-core structure and calculate the minimum ST cut of the k-core structure.

[0042] Furthermore, when selecting the farthest S node, the hierarchical path search strategy and the semantically-aware maximum adjacent search and sorting strategy are used to obtain two S nodes. For the two S nodes, their maximum ST flows with the T node are calculated respectively, and the smaller maximum ST flow value is used as the minimum ST cut of the K core.

[0043] Furthermore, when constructing a node sequence, each time a node is added to the end of the node sequence, the edge connectivity between the set of sorted nodes and the set of unsorted nodes is calculated, and the relationship between the connectivity and a set value k is determined. When the connectivity is less than the set value, the target graph is split by deleting the edges connecting the two sets.

[0044] Furthermore, the pruning and screening method first determines the number of nodes or super nodes in the subgraph. If the number is 1, the subgraph should be pruned, and then determines whether there are any node mergers in the subgraph during this round of traversal. If there are no node mergers in this round, the subgraph should be pruned.

[0045] Furthermore, if the output residual graph is not empty, the graph is input as the target graph into the hierarchical graph segmentation calculation module for a new round of calculation; when the target graph is empty, the calculation result of the k-edge connected component is recorded in the supernode index, and the k-edge connected component mining algorithm based on local edge connectivity detection stops.

[0046] Compared with the related art, the k-edge connected component mining method based on local edge connectivity detection provided by the present invention has the following beneficial effects:

[0047] First, in the k-edge connected component mining system of the present invention, the local edge connectivity detection module is added to extract the local k-core structure and detect the edge connectivity therein before merging the st node pairs with global k connectivity, thereby improving the accuracy of the merging results.

[0048] Secondly, when taking the local edge connectivity detection calculation, the method of calculating the minimum st cut is used to replace the calculation of the k-core minimum cut, thereby reducing the time of local edge connectivity detection. S nodes are selected by hierarchical path search and maximum neighbor search, and the minimum st cut is calculated respectively. The above two feasible s-point search strategies increase the probability that the calculated minimum st cut is the k-core minimum cut, thereby improving the accuracy of local edge connectivity detection. When calculating the minimum st cut, the maximum st flow algorithm is used to replace the calculation of the minimum st cut according to the maximum flow minimum cut theorem to improve the algorithm speed.

[0049] Furthermore, when the mining method is used to mine a k-edge connected subgraph, according to the k-core lemma, when the last of several consecutive k-connected nodes is added to L, a k-core containing the last node is extracted from L to detect the local edge connectivity of the k-core structure. When constructing a new node sequence, if several consecutive nodes with a connectivity greater than k are added to L, a local k-edge connectivity detection calculation is performed when the last node is added. If the edge connectivity of the k-core structure is greater than or equal to k, a maximum merging strategy is used to merge all nodes at once, thereby reducing the number of local k-edge connectivity detection and merging, and improving computational efficiency.

[0050] More importantly, based on the results of connectivity detection, an effective partition tree pruning strategy is proposed. In the process of hierarchical decomposition, the nodes and edges that are judged not to belong to any k-edge connected components are deleted in advance, so that the algorithm reduces a lot of unnecessary calculations and converges within a shorter number of iterations. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of a target graph structure in the prior art;

[0052] Figure 2 The existing method is Figure 1 Schematic diagram of traversing the target graph shown;

[0053] Figure 3 It is a structural schematic diagram of a k-edge connected component mining system based on local edge connectivity detection of the present invention;

[0054] Figure 4 Is adopted Figure 3 The flowchart of the k-edge connected component mining system based on local edge connectivity detection is shown;

[0055] Figure 5 Yes Yes Figure 4 A schematic flow chart of a method for performing k-edge connected component mining by a k-edge connected component mining system;

[0056] Figure 6 A queue data structure with a head index is used to manage candidate nodes and their connection values ​​with L; and

[0057] Figure 7 It is a flowchart of a local edge connectivity detection method using a local edge connectivity detection module. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] The k-edge connected component mining system and mining method based on local edge connectivity detection are intended to detect and mine the association between local clustered user groups with different densities in a social network. Taking Weibo and WeChat Moments as examples, the social network is abstracted as a graph in a computer, where a "graph" is an abstract data structure; users in the social network are abstracted as "nodes"; the association or interaction relationship between different users is abstracted as an "edge" between "nodes"; user groups with a certain association are defined as a "subgraph", and the abstract data structure is a set of graphs, which includes multiple subgraphs and or single nodes; the association parameter value is set to "k", and the k value represents the edge connectivity of the subgraph with a local clustering degree. Specifically, the k-edge connected component represents a local user group with an association not less than the k value. In the subgraph, when any k-1 edge relationships between the user pairs are disconnected, users in the local group of users can still find another user through at least one relationship path. The larger the k value, the higher the user association in the local group of users.

[0060] In the present invention, the undirected target graph is taken as G = (V, E), where V is a node set, E is an edge set, and the node set V includes a plurality of nodes {v1, v2, ..., v n}, where n is the number of nodes. In the target graph G, there may be multiple connected components with edge connectivity not less than k. The k-edge connected component mining system based on local edge connectivity detection uses the k-edge connected component mining method to mine connected components with edge connectivity not less than k, quickly discover local clustered user groups, and then assist in commercial behaviors such as advertising, interest recommendation, and influential community mining.

[0061] See also Figure 3 , is a structural block diagram of a k-edge connected component mining system based on local edge connectivity detection of the present invention. In the k-edge connected component mining system and mining method based on local edge connectivity detection of the present invention, a semantically-aware Maximum Adjacency Search (MAS) sorting strategy is used to detect and mine k-edge connected components of the target graph G. The k-edge connected component mining system 20 is a k-edge connected component mining system based on local edge connectivity detection applied to social networks, which performs local edge connectivity detection on the target graph G to mine k-edge connected components.

[0062] The k-edge connected component mining system 20 includes a hierarchical graph segmentation calculation module 21 , a traversal sorting module 23 , a graph segmentation module 24 , a local edge connectivity detection module 25 , a node merging module 27 and a subgraph pruning module 29 .

[0063] The hierarchical graph segmentation calculation module 21 maintains a supernode index. When the k-edge connected component mining system 20 is working, the hierarchical graph segmentation calculation module 21 performs hierarchical segmentation calculations on the target graph G, the subgraph after the target graph G is segmented, and the subgraph after the subgraph is segmented. Specifically, when the target graph G is first subjected to k-edge connected component mining, the hierarchical graph segmentation calculation module 21 performs hierarchical calculations on the target graph G; when the target graph G is segmented into several subgraphs, the hierarchical graph segmentation calculation module 21 performs hierarchical graph segmentation calculations on the several subgraphs after segmentation, and the target graph in this process may be a subgraph. In the hierarchical graph segmentation calculation process, if there is a subgraph that meets specific conditions, a pruning operation is performed on the subgraph until the maintained target graph is empty.

[0064] A node sequence L is initialized in the traversal sorting module 23. Initially, a target graph G including multiple nodes is provided. The traversal sorting module 23 receives nodes from the target graph G, and traverses the nodes of the target graph G using the MAS sorting strategy to form a new node sequence. The so-called semantic-aware maximum adjacency search sorting strategy refers to giving an undirected graph G = (V, E) and initializing a node sequence L, where V represents a node set and E represents the edges between nodes. The nodes that are not added to the node set V are defined as remaining nodes, and the nodes in the remaining node set VL that are connected to the node sequence L are called candidate nodes. The remaining nodes are merged into the node sequence L to form a new node sequence. Initially, a node is randomly selected from the node set V to be added to the node sequence L, and then each time the node with the largest edge connection degree with the current node sequence L is selected from the candidate nodes to be added to the tail of the node sequence L until all nodes are added to the node sequence L. As the sorting nodes in the node sequence L increase, the candidate nodes change continuously. If there are multiple remaining nodes among the candidate nodes whose connectivity has the same maximum value as the connectivity of the current sequence L, the remaining nodes of the next order are selected according to the principle of "the most recently updated candidate nodes are selected first".

[0065] The graph segmentation module 24 determines whether to segment the target graph or subgraph according to the node sequence state constructed by the traversal sorting module 23. Specifically, if the edge connectivity between the sorted node set and the unsorted node set is less than a set value, the target graph is segmented by deleting the edge between the two node sets to obtain two segmented subgraphs.

[0066] As for the graph segmentation module 24, the graph segmentation module 24 can perform segmentation operations on the target graph or the sub-graphs decomposed from the target graph.

[0067] The local edge connectivity detection module 25 is used to extract a k-core structure from the current local subgraph and detect and determine the edge connectivity value of the k-core structure. Specifically, if the k-core structure is empty, the local edge connectivity detection operation is terminated; if the k-core structure is not empty, the node farthest from the t-node is selected in the k-core structure as the s-node to obtain the minimum st-cut of the k-core structure.

[0068] The node merging module 27 determines whether to merge nodes according to the relationship between the detection result of the local edge connectivity detection module 25 and the set edge connectivity value k. Specifically, if the minimum st cut value is greater than or equal to the set value k, all nodes in the k core are merged in the target graph G.

[0069] The subgraph pruning module 29 determines whether to prune the subgraph according to the merging result of the node merging module 27. Specifically, if the subgraph does not perform node merging in this round, the subgraph is pruned; in addition, if the number of (super) nodes in the subgraph is 1, the subgraph should be pruned. Otherwise, no pruning operation is performed.

[0070] Please refer to Figure 4 and Figure 5 ,in Figure 4 Is adopted Figure 3 The flowchart of the k-edge connected component mining system for k-edge connected component mining is shown in FIG. Figure 5 yes Figure 4 The schematic flow chart of the method for mining k-edge connected components by the k-edge connected component mining system shown in FIG. When the method for detecting and mining k-edge connected components on the target graph G is performed by using the k-edge connected component mining system 20, the method comprises the following steps:

[0071] Step S01, providing the hierarchical graph partitioning calculation module 21, wherein the hierarchical graph partitioning calculation module 21 maintains a supernode index;

[0072] Initially, when the k-edge-connected component mining system is used to mine k-edge-connected components on the target graph G, a target graph G including a plurality of nodes is initially provided.

[0073] Step S02, providing a traversal sorting module 23, wherein the traversal sorting module 23 initializes a node sequence L;

[0074] Step S03, receiving a target graph G including a plurality of nodes, the traversal sorting module 23 traverses the nodes of the target graph G using a MAS sorting strategy, and constructs a new node sequence;

[0075] In this step, when constructing the node sequence, each time after visiting a node, further calculate the edge connectivity between the node and the set consisting of all nodes before it and the set consisting of all nodes after it, and judge the relationship between the connectivity and the set value k, and execute the next step based on the judgment result, that is, the graph splitting operation.

[0076] Step S04, when the traversal sorting module 23 is used to traverse and sort the nodes of the target graph G to construct a new node sequence, a graph segmentation module 24 is provided, and the graph segmentation module 24 performs segmentation of the target graph G according to the constructed node sequence state;

[0077] In this step, when specifically executed, the so-called constructed node sequence state means that when the connectivity value is less than the set value, the target graph or subgraph is split by deleting the edges connected between the two sets. When the connectivity value is greater than the set value, the operation of splitting the target graph or subgraph is not performed.

[0078] It should be noted that when step S04 is the first time to perform a segmentation operation, the graph segmentation module 24 performs a segmentation operation on the target graph G; when step S04 is a segmentation operation on a sub-graph after segmentation of the target graph G, the segmentation operation on the sub-graph is performed.

[0079] Step S05, when the traversal sorting module 23 is used to traverse and sort the nodes of the target graph G to construct a new node sequence, a local k-edge connectivity detection module 25 is provided, and the local k-edge connectivity detection module 25 performs local edge connectivity detection on the target graph G;

[0080] In this step, by adding the local edge connectivity detection module 25, the local k-core structure is extracted and the edge connectivity therein is detected before merging the st node pairs of the global k connectivity.

[0081] Step S06, providing a node merging module 27, wherein the node merging module 27 performs a node merging operation according to the detection result of the local k-edge connectivity detection module 25;

[0082] Specifically, the edge connectivity value is set to k. If the minimum st cut value is greater than or equal to the set value k, then in step S05, all nodes of the k-core structure are merged in the target graph G, and the merged supernodes and the corresponding original nodes are recorded in the supernode index in step S01, and then the local edge connectivity detection operation is terminated. If the minimum st cut value is less than the set value, the k-core is divided into two subgraphs by deleting the edges included in the minimum st cut, and the subgraph containing the t node in the two subgraphs is used as the current local subgraph to continue to perform the local k-edge connectivity detection operation;

[0083] Step S07: providing a subgraph pruning module 29, wherein the subgraph pruning module 29 performs pruning and screening on each subgraph in the target graph G.

[0084] In steps S02-S03, if Figure 6 As shown, the present invention uses a queue data structure with a head index to manage candidate nodes and their connection values ​​with L. The length of the queue is 4*n, and any node v in V i In the data structure shown, it occupies 4 consecutive positions: 4i-4, 4i-3, 4i-2, 4i-1. These 4 positions, connectivity, priority, previous, and next, respectively record v i Connectivity with L, v i The priority of v i The starting position of the previous node with the same connectivity, and the i The starting position of the next node with the same connectivity. The larger the value of the priority, the higher the priority of the node. i If it is not a candidate node, then v i The corresponding 4 positions are all set to zero. The data structure shown maintains a linked list head index. Each linked list head in the index contains a number x, which points to the starting position of the node with the highest priority and the connection degree x. If there are multiple nodes with the connection degree x, a bidirectional linked list is formed in the order of priority: the next bit of the higher priority node records the starting position of the lower priority node, and the previous bit of the lower priority node records the starting position of the higher priority node. The previous bit of the highest priority node and the next bit of the lowest priority node are both set to zero.

[0085] The data structure shown manages nodes through special operations to achieve semantically aware sorting, involving four operations: node selection, node deletion, node update, and node clearing. Specifically, when the MAS sorting algorithm selects the next node from the candidate nodes, the program calls the node selection operation to obtain the node, then calls the node deletion operation to delete the node from the data structure, and then calls the node update operation to batch update the nodes connected to the node and not added to the node sequence L in the data structure shown. If the algorithm divides the target graph according to the sorted node sequence L and the unsorted node set VL, the node clearing operation needs to be called.

[0086] (1) Node selection: Read the linked list head with the largest value from the head index of the data structure shown, and obtain the node position pointed to by the linked list head. The node corresponding to the position is the next node in sequence.

[0087] (2) Node deletion: Set all four positions corresponding to the node to be deleted to zero, and connect the two nodes before and after the node on the doubly linked list. If the next bit of the node does not point to other nodes, delete the linked list head corresponding to the node from the index.

[0088] (3) Node update: Sort the nodes that need to be updated according to connectivity and priority: the nodes with smaller connectivity are ranked higher; if the connectivity is the same, the nodes with smaller priority are ranked higher. Update these nodes in the sorted order: call the node deletion operation to delete the node from the original doubly linked list; set the value of the node's connectivity bit to the updated connectivity y; insert the node to the first position of the doubly linked list with the value of the linked list head y, and set the node's priority bit to the priority value of the next node in the doubly linked list plus 1. If the linked list head with the value y does not exist, create a new linked list head with the value y and point it to the first position of the node's address, and set the node's priority bit to 1.

[0089] (4) Node clearing: Traverse the linked list heads in the head index one by one. For each linked list head, set all four positions corresponding to each node in the doubly linked list it points to to zero, and then delete the linked list head.

[0090] In the present invention, the data structure shown is used to manage candidate nodes, which can implement the sorting principle of "preferential selection of the candidate nodes most recently updated", and each selection, deletion, and update operation can be completed within a constant time complexity, and the algorithm efficiency is relatively high.

[0091] When constructing the node sequence described in step S03, each newly added node in the sequence is processed in turn: for each newly added node, the edge connectivity between the node and the previous node sequence is calculated, and the relationship between the connectivity and the set value k is determined. When the connectivity is greater than or equal to the set value, a local edge connectivity detection operation can be performed to detect the edge connectivity between the node and the previous node.

[0092] Furthermore, when the edge connectivity between a plurality of consecutive nodes and their previous node sequences is greater than or equal to a set value, the local edge connectivity detection operation is performed only after the last node of the consecutive nodes is added to the node sequence.

[0093] like Figure 7 As shown, under the above conditions, the local edge connectivity detection method using the local edge connectivity detection module 25 includes the following steps:

[0094] Step S41, defining a local subgraph;

[0095] The node that is most recently added to the node sequence is defined as a t-node, and the connected subgraph formed by the t-node and its previous nodes is used as the current local subgraph.

[0096] Step S42, extracting the k-core structure to obtain st node pairs;

[0097] A maximum k-core containing a t-node is extracted from the current local subgraph. If the k-core is empty, the local edge connectivity detection operation is terminated. If the k-core is not empty, the node farthest from the t-node is selected in the k-core as the s-node to obtain an st-node pair.

[0098] Specifically, when selecting the farthest S node, the hierarchical path search strategy and the semantic-aware maximum adjacent search and sorting strategy are used to obtain two S nodes. For the two S nodes, their maximum ST flows with the T node are calculated respectively, and the smaller maximum ST flow value is used as the minimum ST cut of the K core.

[0099] Hierarchical path search method: divide the nodes in the k-core into several levels: H0, H1, ..., H m , where the subscript of the level represents the distance of the node from the t node. The smallest level only contains the t node, and the next level contains the neighboring nodes of the t node. The adjacent nodes that are not included in the upper level are placed in the next level, and the node s is selected from the level with the largest subscript.

[0100] MAS search method: When selecting the s node, a maximum neighbor search algorithm based on semantic perception is used to traverse, and a node sequence is constructed with the t node as the initial node and the k core, and the last node of the node sequence is selected as the s node.

[0101] Step S43, calculating the minimum st cut;

[0102] The maximum ST flow method is used to calculate the maximum ST flow to obtain the minimum ST cut of the k cores.

[0103] Step S44, merging nodes.

[0104] Determine the relationship between the minimum st cut value and the set value k. If the minimum st cut value is greater than or equal to the set value k, merge all the nodes in the k-core in the target graph, and record the merged supernode and the corresponding original node in the supernode index in step S01, and then end the local edge connectivity detection operation. If the minimum st cut value is less than the set value, divide the k-core into two subgraphs by deleting the edges included in the minimum st cut, and use the subgraph containing the t node in the two subgraphs as the current local subgraph, and continue to execute steps S42 to S44.

[0105] In step S07, the pruning and screening method for each subgraph in the target graph G is as follows: first, the number of (super) nodes in the subgraph is determined. If the number is 1, the subgraph should be pruned. Then, it is determined whether there are any node merges in the subgraph during this round of traversal. If there are no node merges in this round, the subgraph should be pruned.

[0106] After deleting the subgraph to be pruned, if the output remaining graph is empty, the subgraph set described in step S01 is output as the result; if the output remaining graph is not empty, the graph is used as the target graph, and steps S02 to S07 are continued.

[0107] In the method flow of performing k-edge connected component mining in the above-mentioned k-edge connected component mining system, when initially performing mining on the global target graph G, the method starts from step S01, and traverses and sorts the target graph G, and performs merging, splitting and pruning operations.

[0108] It should be noted that in the method flow of performing k-edge connected component mining in the above-mentioned k-edge connected component mining system, when steps S02 to S07 are executed in a loop, there is no need to traverse and sort the target graph G, merge nodes, split the target graph, and perform pruning operations. Instead, traverse and sort, merge nodes, split the target graph, and perform pruning operations directly on the subgraphs after the target graph G is split, that is, directly execute steps S02 to S07.

[0109] Furthermore, the subgraph splitting operation in step S04 is not a necessary step. The subgraph splitting operation is performed only when the edge connectivity between the sorted node set and the unsorted node set is less than a set value.

[0110] Similarly, the node merging operation in step S06 is not a necessary step. Only when the minimum st cut value is greater than or equal to the set value k, all nodes in the k cores are merged in the target graph G.

[0111] It can be seen that step S04 and step S06 do not have to be calculated every time iteratively, but are selected and executed according to the conditions met. Step S04 and step S06 can be executed simultaneously or successively. Of course, as an improvement of the implementation method, step S06 and step S04 can be executed interchangeably.

[0112] Compared with the prior art, in the k-edge connected component mining system 20 of the present invention, a local edge connectivity detection module 25 is added to extract the connected subgraph composed of the node before merging and the node before it as the current local subgraph, and then determine the minimum ST cut of the k-core, so that a relatively accurate detection result can be quickly calculated. On the other hand, when taking the local edge connectivity detection calculation, the S node is selected by hierarchical path search and maximum adjacent search, so that the probability that the calculated minimum ST cut is the minimum cut of the k-core is increased, thereby improving the accuracy of the local edge connectivity detection.

[0113] Furthermore, when the mining method is used to mine a k-edge connected subgraph, when constructing a new node sequence, if several consecutive nodes with a connectivity greater than k are added to L, a local edge connectivity detection calculation is performed when the last node is added, thereby reducing the number of local edge connectivity detections. According to the k-core lemma, when the last of several consecutive k-connected nodes is added to L, a k-core containing the last node is extracted from L, and the local edge connectivity of the k-core is detected. If the edge connectivity of the k-core is greater than or equal to k, a maximum merging strategy is adopted to merge all nodes to improve computational efficiency.

[0114] In addition, based on the results of connectivity detection, an effective partition tree pruning strategy is proposed to delete in advance the nodes and edges that are judged not to belong to any k-edge connected components, so that the algorithm can reduce a large amount of unnecessary calculations and converge within a shorter number of iterations.

[0115] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A k-edge connected component mining system based on local edge connectivity detection applied to social networks, characterized in that: include: A traversal sorting module, in which a node sequence is initialized, a target graph including a plurality of nodes is received, and a MAS sorting strategy is adopted to traverse the plurality of nodes of the target graph to construct a new node sequence; A graph segmentation module, which determines whether to segment the target graph to obtain a segmented subgraph according to the relationship between the sorting state of the node sequence constructed by the traversal sorting module and the set edge connectivity value; A local edge connectivity detection module extracts a k-core structure from a local target graph for detection, and determines an edge connectivity value of the k-core structure based on the detection result; and A node merging module determines whether to merge nodes according to the relationship between the detection result of the local edge connectivity detection module and the set edge connectivity value. When the local edge connectivity detection module performs the local edge connectivity detection operation, the following steps are included: Define a local subgraph: define the node that is newly added to the node sequence as a t-node, and use the connected component formed by the t-node and its previous nodes as the current local subgraph; Extracting a k-core structure to obtain an ST node pair: extracting a k-core structure containing a t node from the current local subgraph. If the k-core structure is empty, the local edge connectivity detection operation is terminated. If the k-core structure is not empty, the node farthest from the t node is selected in the k-core structure as an S node to obtain an ST node pair. Calculate the minimum ST cut: Calculate the maximum ST flow to obtain the minimum ST cut of the k-core structure.

2. The k-edge connected component mining system based on local edge connectivity detection applied to social networks according to claim 1, characterized in that: The mining system also includes a hierarchical graph partitioning computing module for maintaining a supernode index.

3. The k-edge connected component mining system based on local edge connectivity detection applied to social networks according to claim 1, characterized in that: The traversal sorting module uses a semantically-aware MAS sorting strategy to traverse the nodes of the target graph and construct a new node sequence.

4. The k-edge connected component mining system based on local edge connectivity detection according to claim 1, characterized in that: It also includes a subgraph pruning module, which confirms whether to perform a subgraph deletion and pruning operation based on the calculation result of the node merging module.

5. A k-edge connected component mining method based on local edge connectivity detection applied to social networks, characterized in that: The steps include: Providing a traversal sorting module, wherein the traversal sorting module initializes a node sequence; A target graph including a plurality of nodes is received, and the traversal sorting module traverses the nodes of the target graph using a MAS sorting strategy based on semantic perception, and constructs a new node sequence; Providing a graph partitioning module, wherein the graph partitioning module performs a graph partitioning operation according to the sorting state of the node sequence constructed by the traversal sorting module; Providing a local k-edge connectivity detection module to perform local edge connectivity detection on the target graph; and Providing a node merging module, wherein the node merging module performs a node merging operation according to the detection result of the local k-edge connectivity detection module to obtain a subgraph set, When performing a local edge connectivity check operation, the following steps are included: Define a local subgraph: define the node that is newly added to the node sequence as a t-node, and use the connected component formed by the t-node and its previous nodes as the current local subgraph; Extracting a k-core structure to obtain an ST node pair: extracting a k-core structure containing a t node from the current local subgraph. If the k-core structure is empty, the local edge connectivity detection operation is terminated. If the k-core structure is not empty, the node farthest from the t node is selected in the k-core structure as an S node to obtain an ST node pair. Calculate the minimum ST cut: Calculate the maximum ST flow to obtain the minimum ST cut of the k-core structure.

6. The k-edge connected component mining method based on local edge connectivity detection applied to social networks according to claim 5, characterized in that: Before traversing by the traversal sorting module, the following steps are also included: providing a hierarchical graph partitioning calculation module, wherein the hierarchical graph partitioning calculation module maintains a super node index.

7. The k-edge connected component mining method based on local edge connectivity detection applied to social networks according to claim 6, characterized in that: After the merging operation is performed on the nodes, the method further includes a pruning and screening step for the subgraph in the target graph.

8. The k-edge connected component mining method based on local edge connectivity detection applied to social networks according to claim 6, characterized in that: The step of constructing a node sequence using a semantically-aware maximum adjacency search sorting strategy also includes the following steps: initializing an empty node sequence; selecting a node with the largest connection with the node set of the initialized node sequence from the unsorted nodes and placing it at the end of the node sequence; when there are multiple nodes in the unsorted nodes that have the same maximum edge connectivity as the generated node sequence, the principle of "the most recently updated candidate node is given priority" is used to select the next node.

9. The k-edge connected component mining method based on local edge connectivity detection applied to social networks according to claim 8, characterized in that: When constructing a node sequence, for each node that will be added to the node sequence after being selected, the edge connectivity value of the node and the current node sequence before it is added is calculated, and the relationship between the connectivity value and the set value k is determined. When the connectivity value is greater than or equal to the set value k, a local edge connectivity detection operation is performed to detect the edge connectivity between the node and the node before it.

10. The k-edge connected component mining method based on local edge connectivity detection applied to social networks according to claim 9, characterized in that: When the edge connectivity value of a plurality of consecutive nodes and their previous node sequence is greater than or equal to a set value k, a local edge connectivity detection operation is performed only when the last node of the plurality of consecutive nodes is added to the node sequence.

11. The k-edge connected component mining method based on local edge connectivity detection applied to social networks according to claim 9 or 10, characterized in that: When performing a local edge connectivity check operation, the following steps are included: Merge nodes: Determine the relationship between the minimum ST cut value and the set value. If the minimum ST cut value is greater than or equal to the set value, merge all the nodes in the k-core structure in the target graph, and record the merged supernodes and the corresponding original nodes in the supernode index, and then end the local edge connectivity detection operation; if the minimum ST cut value is less than the set value, divide the k-core structure into two subgraphs by deleting the edges included in the minimum ST cut, and use the subgraph containing the t node in the two subgraphs as the current local subgraph, and continue to extract the k-core structure and calculate the minimum ST cut of the k-core structure.

12. The local edge connectivity detection operation according to claim 11, characterized in that: When selecting the farthest S node, the hierarchical path search strategy and the semantically-aware maximum adjacent search and sorting strategy are respectively used to obtain two S nodes. For the two S nodes, their maximum ST flows with the T node are respectively calculated, and the smaller maximum ST flow value is used as the minimum ST cut of the K core.

13. The k-edge connected component mining method based on local edge connectivity detection according to claim 6, wherein the segmentation is characterized in that: When constructing a node sequence, each time a node is added to the end of the node sequence, the edge connectivity between the set of sorted nodes and the set of unsorted nodes is calculated, and the relationship between the connectivity and a set value k is determined. When the connectivity is less than the set value, the target graph is split by deleting the edges connecting the set of sorted nodes and the set of unsorted nodes.

14. The k-edge connected component mining method based on local edge connectivity detection according to claim 7, characterized in that: The pruning and screening method first determines the number of nodes or super nodes in the subgraph. If the number is 1, the subgraph should be pruned. Then, it is determined whether there is a node merger in the subgraph during this round of traversal access. If there is no node merger in this round, the subgraph should be pruned.

15. The k-edge connected component mining method based on local edge connectivity detection according to claim 6, characterized in that: If the output residual graph is not empty, the graph is input as the target graph into the hierarchical graph segmentation calculation module for a new round of calculation; when the target graph is empty, the calculation result of the k-edge connected component is recorded in the supernode index, and the k-edge connected component mining algorithm based on local edge connectivity detection stops.