Social network key node identification method and device fusing high-order features

Abstract

The invention discloses a social network key node identification method and device fusing high-order features, and the method comprises the steps: obtaining content transmission relation data corresponding to social media content, and constructing a dynamic social media content transmission relation network according to the content transmission relation data; acquiring high-order features of the nodes by utilizing continuous homology under different filtering functions; fusing the high-order feature with the out-degree of the low-order feature of the node to obtain an out-degree-coherence index, and taking the out-degree-coherence index as a node importance evaluation index; and according to the node importance evaluation indexes, performing descending sort on the nodes to identify key nodes. According to the method, the high-order features of the nodes in the network are captured through continuous coherence, the network structure can be analyzed more comprehensively, local and global topological features of the network can be captured without pre-defining specific scale parameters, and robustness is achieved for noise and disturbance in data.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a method and device for identifying key nodes in a social network. Background Art

[0002] With the advent of the information age, the Internet has been widely used and popularized, greatly changing the channels through which people obtain data and information. Nowadays, online social networks are ubiquitous. Various online social platforms such as Weibo, Douyin and Xiaohongshu have become important channels for people to obtain and transmit information, and have also become an important component of people's daily lives. In various public information dissemination events, a small number of users will play a greater role in promoting the dissemination of information due to their greater influence.

[0003] The existing research on key node identification mainly focuses on the analysis of network topology. By evaluating the impact of various network substructures on node importance, the node importance is quantified into numerical values ​​and sorted to identify key nodes, which is of great significance in helping to control network attacks, suppress the spread of rumors and prevent the spread of computer viruses. Most existing models evaluate node importance through the low-order structure of the network, ignoring the impact of the high-order structure of the network, resulting in worse evaluation results. Summary of the invention

[0004] In view of the above problems, the present invention is proposed to provide a method and device for identifying key nodes in a social network that overcomes the above problems or at least partially solves the above problems, and the method and device include:

[0005] Obtain the content propagation relationship data corresponding to the social media content, build a dynamic social media content propagation relationship network based on the content propagation relationship data, and use continuous coherence under different filtering functions to obtain high-order features of nodes;

[0006] According to the calculated high-order features of the node, the node's low-order feature out-degree is fused to obtain the node importance evaluation index: out-degree-coherence index;

[0007] According to the node importance evaluation index, the nodes are sorted in descending order to identify the key nodes.

[0008] Optionally, the method of obtaining high-order features of nodes by continuous coherence under different filter functions comprises: simplifying and dividing the dynamic social media content dissemination relationship network to obtain a first-order neighbor subgraph and a second-order neighbor subgraph of each node;

[0009] According to the first-order neighbor subgraph and the second-order neighbor subgraph of the node, Dowker source filtering and VR filtering are used to filter respectively to obtain the Dowker complex cluster and VR complex cluster of the node;

[0010] According to the Dowker complex cluster and VR complex cluster of the node, the node Dowker homology features and VR homology features containing high-order features are calculated on each subgraph, and they are fused to form the high-order features of the node;

[0011] Optionally, according to the dynamic social media content propagation relationship network, the step of obtaining a first-order neighbor subgraph and a second-order neighbor subgraph of each node includes:

[0012] Convert dynamic social media content propagation relationship network into an undirected network G ′ , remove the network G ′ 1-shell in the network G S ′ 2;

[0013] The network G is analyzed in nodes S ′ 2 is divided to obtain the first-order nearest neighbor subgraph and the second-order nearest neighbor subgraph of the node.

[0014] Optionally, the step of filtering using Dowker source filtering and VR filtering respectively according to the first-order nearest neighbor subgraph and the second-order nearest neighbor subgraph of the node to obtain the Dowker complex cluster and the VR complex cluster of the node includes:

[0015] The first-order neighbor subgraph of the node is filtered using the Dowker filtering function to obtain the Dowker complex cluster of the node;

[0016] The second-order neighbor subgraph of the node is filtered using the VR filtering function to obtain the VR complex cluster of the node.

[0017] Optionally, according to the Dowker complex cluster and VR complex cluster of the node, the node Dowker homology features and VR homology features containing high-order features are calculated on each subgraph, and they are fused to form the high-order features of the node, specifically including:

[0018] For the Dowker complex cluster and VR complex cluster of the node, use continuous homology to calculate the existence time of its one-dimensional homology group, and obtain the Dowker homology characteristics and VR homology characteristics of the node;

[0019] The Dowker homology features and VR homology features of the nodes are normalized using the sigmoid function, and adjustable parameters are added to control their range to obtain the high-order features of the nodes.

[0020] Optionally, the step of fusing the calculated high-order feature of the node with the out-degree of the low-order feature of the node to obtain the node importance evaluation index includes:

[0021] The node out-degree is normalized using the maximum and minimum normalization function as the low-order feature out-degree, which is multiplied by the node's high-order feature to obtain the node importance evaluation indicator: out-degree-coherence index.

[0022] Optionally, the step of sorting the nodes in descending order according to the node importance evaluation index to identify the key nodes therein includes:

[0023] The out-degree-coherence index of each node in the network is calculated, and the nodes are sorted in descending order according to the calculation results, and the nodes with the highest ranking are taken as key nodes.

[0024] According to another aspect of the present invention, a device for identifying key nodes in a social network integrating high-order features is provided, comprising:

[0025] The network preprocessing module is used to obtain the content propagation relationship data corresponding to the social media content, build a dynamic social media content propagation relationship network based on the content propagation relationship data, simplify and divide the network, and obtain the first-order neighbor subgraph and second-order neighbor subgraph of each node.

[0026] The filtering module is used to filter the first-order neighbor subgraph and the second-order neighbor subgraph of the node using Dowker source filtering and VR filtering respectively to obtain the Dowker complex cluster and VR complex cluster of the node.

[0027] The homology calculation module is used to calculate the Dowker homology features and VR homology features of the nodes containing high-order features on each subgraph according to the Dowker complex cluster and VR complex cluster of the nodes, and the two are fused to form the high-order features of the nodes.

[0028] The feature fusion module is used to fuse the high-order features of the node calculated by the homology calculation module with the low-order feature out-degree of the node to obtain the out-degree-homology index as a node importance evaluation indicator.

[0029] The key node identification module is used to sort the nodes in descending order according to the node importance evaluation index out-degree-coherence index to identify the key nodes.

[0030] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0031] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned method for identifying key nodes in a social network.

[0032] According to another aspect of the present invention, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned method for identifying key nodes in a social network.

[0033] Beneficial effects:

[0034] (1) According to the method and device for identifying key nodes in a social network that integrates high-order features provided by an embodiment of the present invention, content dissemination relationship data is constructed into a dynamic social media content dissemination relationship network, and high-order features of nodes are obtained by continuous coherence under different filtering functions, and combined with the out-degree of the low-order features of the nodes to form the out-degree-coherence index, an evaluation index of node importance. Key nodes in social networks can be identified, which facilitates the control of the dissemination of public information, helps platforms notice emergencies on the network, and promotes the development of social media platforms. Unlike existing key node identification methods, the present invention uses continuous coherence to capture high-order features of nodes in the network, can perform a more comprehensive analysis of the network structure, can capture local and global topological features of the network without pre-defining specific scale parameters, and is robust to noise and disturbances in the data.

[0035] (2) The interactions between multiple nodes in a social network form a variety of high-order structures, such as triangular structures, quadrilateral structures, pentagonal structures, and other ring structures. Dowker source filtering and VR filtering can use the time attributes of the edges in the social network as filtering parameters, thereby effectively analyzing the changes of high-order structures in the node network over time and obtaining high-order characteristics of the network.

[0036] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0038] Figure 1 A flow chart of a method for identifying key nodes in a social network provided by an embodiment of the present invention is shown;

[0039] Figure 2 A schematic diagram of a VR filter function provided by an embodiment of the present invention is shown;

[0040] Figure 3A schematic diagram of a Dowker source filtering function provided by an embodiment of the present invention is shown;

[0041] Figure 4 A schematic diagram showing the structure of a device for identifying key nodes in a social network provided by an embodiment of the present invention is shown;

[0042] Figure 5 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0043] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0044] The present invention provides a method for identifying key nodes in social networks. The method constructs content dissemination relationship data into a dynamic social media content dissemination relationship network, uses continuous coherence to capture high-order features of nodes in the network, and combines them with the node's low-order feature out-degree to form a node importance evaluation index out-degree-coherence index, thereby identifying key nodes in social networks and facilitating the control of the dissemination of public information. The present invention mines the high-order features of nodes in the network, can perform a more comprehensive analysis of the network structure, can capture local and global topological features of the network without pre-defining specific scale parameters, and is robust to noise and disturbances in the data.

[0045] Figure 1 A flowchart of an embodiment of a method for identifying key nodes in a social network according to the present invention is shown. Figure 1 As shown, the method comprises the following steps:

[0046] Step S101: Obtain content propagation relationship data corresponding to social media content, construct a dynamic social media content propagation relationship network based on the content propagation relationship data, simplify and divide the network, and obtain a first-order neighbor subgraph and a second-order neighbor subgraph for each node.

[0047] In this embodiment, a key node identification model is used to identify key nodes based on the content propagation relationship data between users. Specifically, the content propagation relationship between users refers to the interaction between users in social media, including forwarding, liking and commenting. The content propagation relationship of users can be stored using a graph data structure. In the graph data, each node represents a single user, and the edges represent the relationship characteristics between users. In particular, in this embodiment, the relationship characteristics all have time attributes. According to the time attribute of the edge, a filter function can be used to filter the sub-network later.

[0048] According to the content dissemination relationship between users, a directed and weighted network G = (V, E, W) can be constructed, where V is the node set of the network, E is the edge set of the network, and the weight function W is: Defined on the edge. Since the subsequent selected filtering function can discover more network structures for the undirected network, the network is converted into an undirected network G ′ =(V,E ′ ,W ′ ). In an undirected network, the k-shell core number of a node is a method to measure the position and importance of a node in the network. It reveals the core degree of a node in the network structure by layering the network. The higher the k-shell core number of a node, the greater the role and influence of the node in the network.

[0049] The number of k-shell cores of nodes that form a high-order topological structure must be greater than or equal to 2. In order to remove nodes that do not have a high-order topological structure nearby and reduce the time complexity of subsequent subgraph partitioning and continuous homology, the network G ′ Remove the 1-shell, that is, the nodes with only one neighbor, and get the network G S ′ 2=(V S2 ,E S ′ 2,W S ′ 2).

[0050] The time and space complexity of continuous synchronization on a large-scale network are very high, and this embodiment extracts high-order topological features from nodes, and the corresponding high-order topological structure exists near the node. In order to reduce the complexity, this embodiment only performs continuous synchronization on the node neighbor subgraph. In the network G S ′ 2. Take the node v i As the center, divide the first-order neighbor subgraph Where V i 1 The first-order neighbor subgraph of contains v i and with v i Directly connected nodes, edges Contains the edges between these nodes in the network, with weight W i 1 is the weight corresponding to the edge in the nearest neighbor subgraph. Where V i 2 Contains all nodes directly connected to node V i 1The connected nodes, the first-order neighbor nodes of these first-order neighbor nodes, the edges Contains nodes V in the network i 2 The weight W i 2 is the weight corresponding to the edge in the neighbor subgraph. n is V S2 The number of nodes in .

[0051] Step S102: filtering the first-order neighbor subgraph and the second-order neighbor subgraph of the node using Dowker source filtering and VR filtering respectively to obtain the Dowker complex cluster and VR complex cluster of the node.

[0052] Different forms of node interaction will form different high-order topological structures. In order to discover these high-order structures in a targeted manner, different filtering functions need to be used to filter the node neighbor subgraphs. Filter using the Dowker filtering function to obtain the Dowker complex cluster in δ is the filtering parameter, δ1<δ2<…<δ m Similarly, for the second-order nearest neighbor subgraph Use VR filter function to filter and get VR complex cluster Dowker complex cluster Used to discover triangular structures, and VR complex clusters Used to discover tetragonal and pentagonal structures.

[0053] The filter function divides the topological data into a series of subsets according to the restrictions of the filter parameters (when the topological data is a network, the filter parameters are usually the weights of the edges), and then represents each topological data subset as a corresponding simplicial complex through the simplicial complex composition rule. Since the filter function is formed by adding filter parameters on the basis of the simplicial complex composition rule, it is usually named after the simplicial complex composition rule. This embodiment mainly uses two filter functions: VR filtering (Vietoris-Rips Filtration) and Dowker filtering (Dowker Sink and Source Filtration), and their use examples are as follows: Figure 2 and Figure 3 shown.

[0054] Step S103: According to the Dowker complex cluster and VR complex cluster of the node, the node Dowker homology features and VR homology features containing high-order features are calculated on each subgraph, and they are merged to form the high-order features of the node.

[0055] For simplicial complex clusters For each simplicial complex X in , compute its k-dimensional homology group H k (X), and obtain the result of continuous synchronization:

[0056]

[0057] Among them, δ represents the scale parameter, X δ represents a simplicial complex with a scale parameter of δ, where the scale parameter δ is between 0 and δ ′ changes between them, forming a complex cluster. That is, the appearance and disappearance time of each topological structure in the complex cluster is expressed in the form of a continuous barcode:

[0058]

[0059] Where b is the appearance time of the topological structure, d is the disappearance time of the topological structure, J k,i for The number of k-dimensional homology groups in the network. Some topological structures will not disappear after they appear. When calculating, their disappearance time is usually set to the maximum value of the filtering parameter δ m .

[0060] In this embodiment, the coherence features of each high-order topological structure near the node are taken as the high-order topological features of the node, and the existence time of the topological structure is summed up by node. The result is called the coherence value (Persistent Value, PV), that is,

[0061]

[0062] Let k = 1, we can i The one-dimensional hole nearby, i.e. the homology feature pv1(v i ) to perform the calculation. Use Calculated Dowker homology characteristics Corresponding to the high-order topological features of each triangular structure near the node, use Calculated VR coherence features It corresponds to the high-order topological features of the quadrilateral and pentagonal structures near the node. The larger the coherence value, the greater the frequency of node interaction, the longer the interaction time, and the more obvious its high-order topological features.

[0063] The homology features calculated by the sigmoid function for the two filter functions are obtained and Normalize

[0064]

[0065] Among them, α and β are adjustable parameters and can be set to 0 in this embodiment.

[0066] Step S104: The calculated high-order features of the node are fused with the out-degree of the low-order features of the node to obtain the out-degree-coherence index, which is an indicator for evaluating the importance of the node.

[0067] Considering that the possible value ranges of node out-degree and homology value are usually very different, the maximum and minimum normalization functions are used to normalize the node out-degree to adjust the relative size of the node high-order features and low-order features to a suitable range.

[0068]

[0069] Among them, od is the out-degree of the node, and V is the set of nodes in the network.

[0070] Step 105: Sort the nodes in descending order according to the node importance evaluation indicator out-degree-coherence index to identify the key nodes therein.

[0071] Combining the coherence value that represents the high-order topological characteristics of the node with the out-degree that represents the low-order topological characteristics of the node, we can construct an evaluation index for node propagation capability: the Outdegree-Persistent Index (OPI). The calculation method for the outdegree-persistent index of node v is:

[0072] OPI(v)=OD(v)·PV(v)

[0073] Calculate the out-degree-coherence index of all nodes in the network in turn, and sort the nodes in descending order accordingly. Take the top k nodes in the sorting, i.e., the Top@k nodes, to get the key nodes in the network. The number of nodes k can be selected by yourself.

[0074] Figure 2 The VR filter function used in the embodiment of the social network key node identification device of the present invention is shown. The VR filter function is defined as follows: G = (V, E, W) is an undirected weighted network, where V = {v1, v2, ..., v n} is the node set of the network, is the edge set of the network, and the weight function W: Defined on the edge. For any δ∈R, let the subnetwork Edge Set E δ ∈E contains only edges with weights less than or equal to δ, then for any δ∈R, the VR complex can be defined as a subnetwork G δ The faction complex Cl(G δ ), define VR filtering as {Cl(G δ )→Cl(G δ′ )} 0≤δ≤δ′ .

[0075] The clique complex Cl(G) is a complex in which each simplex corresponds to a complete subgraph in G. In this filtering process, the node set is first added to the VR complex, and then the edges are added to the VR complex according to the change of the parameter δ from small to large according to the edge weight. Figure 2 As shown, let's first look at the 0-dimensional homology features. When δ = 0, the network is represented as a simple complex composed of 5 independent nodes, that is, there are 5 connected branches in the complex, so there are 5 homology features generated in the 0-dimensional persistent barcode at the beginning. When δ = 1, two edges are added to the complex, which makes the 3 nodes combined into 1 connected branch. Therefore, in the 0-dimensional persistent barcode, two homology features disappear when δ = 1. When δ = 2, two more edges are added to the complex, which makes the entire complex only have 1 connected branch. Therefore, when δ = 1, two 0-dimensional homology features disappear again. After that, the 0-dimensional homology features do not change, so there is a homology feature that exists permanently, corresponding to the interval with arrows in the 0-dimensional persistent barcode.

[0076] For the 1D homology feature, when δ=3, a 1D hole appears in the simplicial complex, so a 1D homology feature is generated, corresponding to the interval with an arrow in the 1D persistence barcode. When δ=4, although the nodes that make up the loop hole have changed (from {A,B,D,E,C} to {B,D,E,C}) because nodes A,B,C form a 2D simplex, the loop hole still exists as a topological feature, so its persistence is not affected and it exists permanently after generation. If there is an edge CD with a weight of 5 in the network in the graph, so that the next step of filtering is to form nodes B,C,D and nodes C,D,E into 2D simplexes respectively, then the original 1D loop hole will disappear, and its existence time can be expressed as [3,5].

[0077] When δ≤4, nodes A, B, and C form a complete subgraph and are therefore combined into a 2-dimensional simplex. Since there are no edges between nodes B, E and nodes C, D, nodes B, D, E, and C do not form a 3-dimensional simplex, but only a 1-dimensional simplicial complex, forming a 1-dimensional hole.

[0078] Figure 3 The Dowker filter function used in the embodiment of the social network key node identification device of the present invention is shown. The definition of the Dowker filter function is: let G = (V, E, W) be a directed weighted network, if there is a target node v ′ YV, so that any v i (1≤i≤n) satisfies W(v i ,v ′ )≤δ, then there is a Dowkerδ-complex And thus define Dowker Sink Filtration as Correspondingly, if there is a source node v ′ ∈V, so that any v i (1≤i≤n) satisfies W(x ′ ,x i )≤δ, then there is a Dowkerδ-source complex Dowker Source Filtration is

[0079] like Figure 3 As shown in the figure, when δ = 1, the Dowker source complex is constructed only by connecting edges S1A and S2D, so only nodes A and D form the complex. When δ = 2, nodes A and B have the same source node S1, so they form a 1-dimensional simplex. The same is true for nodes D and E, and together they form a Dowker source complex. When δ = 3, nodes A, B, and C have the same source node, so they form a 2-dimensional simplex. When δ = 4, as the 1-dimensional simplex CE is added to the source complex, the 1-dimensional simplicial complex {B, D, E, C} forms a 1-dimensional ring hole.

[0080] Figure 4 FIG. 2 is a schematic diagram showing a structure of an embodiment of a social media content propagation prediction device according to the present invention. Figure 4 As shown, the device includes: a network preprocessing module 401, a filtering module 402, a coherence calculation module 403, a feature fusion module 404 and a key node identification module 405.

[0081] The network preprocessing module 401 is used to obtain content propagation relationship data corresponding to social media content, construct a dynamic social media content propagation relationship network based on the content propagation relationship data, simplify and divide the network, and obtain a first-order neighbor subgraph and a second-order neighbor subgraph for each node.

[0082] In an optional manner, the network preprocessing module 401 is further used to: analyze user social data and construct a user dynamic social media content dissemination relationship network;

[0083] In an optional manner, the network preprocessing module 401 is further used to: convert the network into an undirected network, and remove the 1-shell in the network;

[0084] In an optional manner, the network preprocessing module 401 is further used to: divide the network into nodes as units to obtain a first-order neighbor subgraph and a second-order neighbor subgraph of the nodes.

[0085] The filtering module 402 is used to filter the first-order neighbor subgraph and the second-order neighbor subgraph of the node using Dowker source filtering and VR filtering respectively to obtain the Dowker complex cluster and VR complex cluster of the node.

[0086] In an optional manner, the filtering module 402 is further used to: filter the first-order neighbor subgraph of the node using a Dowker filtering function to obtain a Dowker complex cluster of the node;

[0087] In an optional manner, the filtering module 402 is further used to: filter the second-order neighbor subgraph of the node using a VR filtering function to obtain a VR complex cluster of the node.

[0088] The homology calculation module 403 is used to calculate the node Dowker homology features and VR homology features containing high-order features on each subgraph according to the Dowker complex cluster and VR complex cluster of the node, and merge them to form the high-order features of the node.

[0089] In an optional manner, the homology calculation module 403 is further used to: for the Dowker complex cluster and VR complex cluster of the node, use continuous homology to calculate the existence time of its one-dimensional homology group to obtain the Dowker homology feature and VR homology feature of the node;

[0090] In an optional manner, the homology calculation module 403 is further used to: normalize the Dowker homology features and VR homology features of the node using a sigmoid function, and add adjustable parameters to control their ranges to obtain high-order features of the node.

[0091] The feature fusion module 404 is used to fuse the calculated high-order node features with the node low-order feature out-degree to obtain the node importance evaluation indicator out-degree-coherence index.

[0092] In an optional manner, the feature fusion module 404 is further used to: normalize the node out-degree using a maximum and minimum normalization function, and multiply it with the node high-order feature to obtain the node importance evaluation indicator out-degree-coherence index.

[0093] The key node identification module 405 is used to sort the nodes in descending order according to the node importance evaluation index out-degree-coherence index to identify the key nodes therein.

[0094] In an optional manner, the key node identification module 405 is further used to: calculate the out-degree-coherence index of each node in the network, sort the nodes in descending order according to the calculation results, and take the top-ranked nodes as key nodes.

[0095] An embodiment of the present invention provides a non-volatile computer storage medium, which stores at least one executable instruction. The computer executable instruction can execute a method for identifying key nodes in a social network in any of the above method embodiments.

[0096] The executable instructions may be specifically used to cause the processor to perform the following operations:

[0097] Obtain content propagation relationship data corresponding to social media content, build a dynamic social media content propagation relationship network based on the content propagation relationship data, simplify and divide the network, and obtain the first-order neighbor subgraph and second-order neighbor subgraph of each node;

[0098] According to the first-order neighbor subgraph and the second-order neighbor subgraph of the node, Dowker source filtering and VR filtering are used to filter respectively to obtain the Dowker complex cluster and VR complex cluster of the node;

[0099] According to the Dowker complex cluster and VR complex cluster of the node, the node Dowker homology features and VR homology features containing high-order features are calculated on each subgraph, and they are fused to form the high-order features of the node;

[0100] According to the calculated high-order features of the node, the node's low-order feature out-degree is fused to obtain the node importance evaluation index out-degree-coherence index;

[0101] According to the node importance evaluation index out-degree-coherence index, the nodes are sorted in descending order to identify the key nodes.

[0102] Figure 5 The schematic diagram of the structure of the computing device embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0103] like Figure 5 As shown, the computing device may include:

[0104] Processor, Communications Interface, Memory, and Communications Bus.

[0105] Wherein: the processor, the communication interface, and the memory communicate with each other through a communication bus. The communication interface is used to communicate with other devices such as a client or other server network elements. The processor is used to execute a program, specifically, it can execute the relevant steps in the above-mentioned embodiment of the method for identifying key nodes in a social network.

[0106] Specifically, the program may include program codes including computer operation instructions.

[0107] The processor may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the server may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0108] The memory is used to store programs. The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0109] The program can be specifically used to cause the processor to perform the following operations:

[0110] Obtain content propagation relationship data corresponding to social media content, build a dynamic social media content propagation relationship network based on the content propagation relationship data, simplify and divide the network, and obtain the first-order neighbor subgraph and second-order neighbor subgraph of each node;

[0111] According to the first-order neighbor subgraph and the second-order neighbor subgraph of the node, Dowker source filtering and VR filtering are used to filter respectively to obtain the Dowker complex cluster and VR complex cluster of the node;

[0112] According to the Dowker complex cluster and VR complex cluster of the node, the node Dowker homology features and VR homology features containing high-order features are calculated on each subgraph, and they are fused to form the high-order features of the node;

[0113] According to the calculated high-order features of the node, the node's low-order feature out-degree is fused to obtain the node importance evaluation index out-degree-coherence index;

[0114] According to the node importance evaluation index out-degree-coherence index, the nodes are sorted in descending order to identify the key nodes.

[0115] The algorithm or display provided here is not inherently related to any specific computer, virtual system or other equipment. Various general systems can also be used together with the teaching based on this. According to the above description, it is obvious to construct the structure required for this type of system. In addition, the embodiment of the present invention is not directed to any specific programming language yet. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described here, and the description of the above specific language is for disclosing the best mode of the present invention.

[0116] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0117] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the following intention: the claimed invention requires more features than those explicitly recited in each claim. Rather, as reflected in the claims, the inventive aspects lie in less than all the features of the individual embodiments previously disclosed. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

[0118] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0119] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, any one of the claimed embodiments may be used in any combination.

[0120] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all functions of some or all components according to embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., computer program and computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0121] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present invention may be implemented by means of hardware comprising several different elements and by means of appropriately programmed computers. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A method for identifying key nodes in a social network by integrating high-order features, characterized in that: include: Obtain content dissemination relationship data corresponding to social media content, and build a dynamic social media content dissemination relationship network based on the content dissemination relationship data; Use continuous coherence under different filter functions to obtain high-order features of nodes; The high-order feature is combined with the low-order feature out-degree of the node to obtain the out-degree-coherence index as a node importance evaluation index; According to the node importance evaluation index, the nodes are sorted in descending order to identify key nodes.

2. The method according to claim 1, characterized in that The high-order features of nodes obtained by continuous coherence under different filtering functions are: According to the dynamic social media content dissemination relationship network, the first-order neighbor subgraph and the second-order neighbor subgraph of each node are obtained; According to the first-order neighbor subgraph and the second-order neighbor subgraph of the node, Dowker source filtering and VR filtering are used to filter respectively to obtain the Dowker complex cluster and VR complex cluster of the node; According to the Dowker complex cluster and VR complex cluster of the node, the Dowker homology feature and VR homology feature of the node are calculated, and the two are fused to form the high-order feature of the node.

3. The method according to claim 2, characterized in that The step of obtaining a first-order neighbor subgraph and a second-order neighbor subgraph of each node according to the dynamic social media content propagation relationship network includes: Convert dynamic social media content propagation relationship network into an undirected network G ′ , remove the 1-shell in the undirected network G′, and get the network G′ S2 ; The network G′ is analyzed in nodes S2 Partition is performed to obtain the first-order nearest neighbor subgraph and the second-order nearest neighbor subgraph of the node.

4. The method according to claim 2, characterized in that: According to the first-order neighbor subgraph and the second-order neighbor subgraph of the node, the steps of filtering using Dowker source filtering and VR filtering respectively to obtain the Dowker complex cluster and VR complex cluster of the node include: The first-order neighbor subgraph of the node is filtered using the Dowker filtering function to obtain the Dowker complex cluster of the node; The second-order neighbor subgraph of the node is filtered using the VR filtering function to obtain the VR complex cluster of the node.

5. The method according to claim 2, characterized in that: The method of calculating the Dowker homology feature and VR homology feature of the node according to the Dowker complex cluster and VR complex cluster of the node, and fusing the two to form the high-order feature of the node, specifically includes: For the Dowker complex cluster and VR complex cluster of the node, use continuous homology to calculate the existence time of the one-dimensional homology group, and obtain the Dowker homology characteristics and VR homology characteristics of the node; The Dowker homology features and VR homology features of the nodes are normalized using the sigmoid function, and adjustable parameters are added to control their range to obtain the high-order features of the nodes.

6. The method according to claim 1, characterized in that The step of fusing the high-order feature with the low-order feature out-degree of the node to obtain the out-degree-coherence index includes: The node out-degree is normalized using the maximum and minimum normalization function as the low-order feature out-degree, which is multiplied by the node high-order feature to obtain the out-degree-coherence index.

7. The method according to claim 1, characterized in that According to the node importance evaluation index, the steps of sorting the nodes in descending order to identify the key nodes include: Calculate the out-degree-coherence index of each node in the network, sort the nodes in descending order according to the calculation results, and take the top-ranked nodes as key nodes.

8. A device for identifying key nodes in a social network by integrating high-order features, characterized in that: include: The network preprocessing module is used to obtain the content propagation relationship data corresponding to the social media content, build a dynamic social media content propagation relationship network based on the content propagation relationship data, and perform network partitioning to obtain the first-order neighbor subgraph and the second-order neighbor subgraph of each node; A filtering module is used to filter the first-order neighbor subgraph and the second-order neighbor subgraph of the node using Dowker source filtering and VR filtering to obtain the Dowker complex cluster and VR complex cluster of the node; The homology calculation module is used to calculate the Dowker homology features and VR homology features of the nodes containing high-order features on each subgraph according to the Dowker complex cluster and VR complex cluster of the nodes, and the two are fused to form the high-order features of the nodes; The feature fusion module is used to fuse the high-order features of the node calculated by the coherence calculation module with the out-degree of the low-order features of the node to obtain the out-degree-coherence index as a node importance evaluation indicator; The key node identification module is used to sort the nodes in descending order according to the node importance evaluation index to identify the key nodes therein.

9. A computing device, characterized in that include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to a method for identifying key nodes in a social network as described in any one of claims 1-7.

10. A computer storage medium, characterized in that: The storage medium stores at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the method for identifying key nodes in a social network as described in any one of claims 1-7.

Images