Multitask stage key node identification method based on complex network

Abstract

The invention discloses a multi-task stage key node identification method based on a complex network. The method comprises the following steps: firstly, inputting communication information flow data of nodes in different task stages of a municipal anti-uncommanded system, constructing a network structure topological graph at the same time, then performing contribution degree calculation according to behavior characteristics of the nodes, and calculating an initial weight of each node according to an entropy weight method; dynamic weight distribution is carried out according to the business volume proportion of each node in different task stages; performing activeness calculation according to structural characteristics of the nodes, firstly calculating related indexes of the nodes, and then performing comprehensive judgment through an improved gravitation model; and finally, multiplying the contribution degree and the activeness to obtain importance values of the nodes, and sorting the importance values from large to small, wherein the nodes sorted in the front are key nodes in the complex network. According to the method, the problems that a traditional key node recognition algorithm is high in static performance, key nodes in different task stages cannot be recognized and the like are solved, and in addition, compared with a traditional gravitation model, the improved gravitation model has a more accurate recognition effect.

Description

Technical Field

[0001] The invention belongs to the technical field of complex networks, and in particular to a method for identifying key nodes in multi-task stages based on complex networks. Background Art

[0002] In recent years, the rapid development of drone technology has greatly lowered the threshold for its use, allowing more and more ordinary consumers to easily own and operate drones. Whether it is tourism photography, aerial video, or agricultural plant protection and environmental monitoring, drones have shown great application potential. However, with the popularization of drone use, illegal flights (i.e. "black flights") have become increasingly serious, and phenomena such as smuggling prohibited items and sensitive candid photography have emerged in an endless stream, which not only poses a potential threat to national and social security, but also arouses widespread public concern about drone management. In response to this challenge, the municipal anti-unmanned command system came into being. The municipal anti-unmanned command system refers to the use of radar detection, radio interference suppression and other technologies to conduct reconnaissance and identification of drones, threat judgment and guidance and disposal. Under distributed confrontation, the municipal anti-unmanned command system presents four stages: perception, processing, decision-making and execution. When a key node in any stage fails, it will greatly affect the stage advancement and thus affect the stable operation of the entire system. Therefore, it is crucial to clarify the key nodes at different stages, which can help commanders prioritize the allocation of limited resources to protect and restore the key parts of the system and improve the response efficiency and effectiveness of the system. However, traditional key node identification methods have problems such as strong staticity, low recognition accuracy and inability to perform multi-task stage identification. Therefore, designing a fast and efficient multi-task stage key node identification method is a research hotspot in the field of complex networks.

[0003] At present, there are several effective ways to identify key nodes in complex networks, including classical algorithms, intelligent optimization algorithms, methods based on deep reinforcement learning, methods based on game theory, and methods based on physical models.

[0004] Commonly used physical models are mainly gravity model and spring model. The gravity model is based on the law of universal gravitation in physics. Inspired by the law of universal gravitation, Ma et al. used the k-core value as the mass and the shortest path between nodes as the distance to propose a gravity model that can sort important nodes.

[0005] The traditional gravity model is simple in principle and easy to implement, but it only considers the structural characteristics of the node and ignores the behavioral characteristics of the node, so it is impossible to identify the key nodes in the multi-task stage. At the same time, it only considers the single indicator of the node and ignores the role of the gravitational constant in the gravity model, which affects the accuracy of key node identification. Summary of the invention

[0006] In view of the deficiencies of the above technologies, the present invention provides a method for identifying key nodes in multi-task stages based on complex networks, which can solve the problems such as the low recognition accuracy of traditional gravity models and the inability to give key nodes in multi-task stages.

[0007] The technical solution of the present invention to solve the technical problems is as follows: A method for identifying key nodes in multi-task stages based on complex networks, comprising the following steps:

[0008] S1: Input the communication information flow data between nodes in different task stages of the municipal anti-illegal command system, and simultaneously construct a network structure topology graph G(V, E);

[0009] S2: Assign an initial weight y to the nodes in different task stages according to the entropy weight method;

[0010] S3: Determine the traffic volume ψ and traffic volume ratio Φ of the nodes in different task stages;

[0011] S4: Determine the contribution degree W of the nodes according to the initial weight y and the traffic volume ratio Φ;

[0012] S5: Calculate the degree value, k-core value, eigenvector centrality value, clustering coefficient, shortest path between nodes, and network average shortest path of the nodes according to the network topology structure;

[0013] S6: Calculate the attenuation factor, mass, distance, gravitational coefficient, and cut-off radius of the improved gravity model according to the topological indexes of the nodes;

[0014] S7: Calculate the activity G of the nodes through the improved gravity model;

[0015] S8: Calculate the importance value I of the nodes according to the contribution degree W and the activity G, and sort them from large to small according to the normalized result of I. The nodes with the top-ranked results are the key nodes in the complex network.

[0016] Compared with the prior art, the present invention has the following remarkable advantages: According to the requirements for identifying key nodes in multi-task stages of complex networks in the information age, the present invention respectively proposes the concepts of contribution degree and activity from the behavioral characteristics and structural characteristics of the nodes, solves the problem that traditional key node identification methods cannot identify key nodes in multi-task stages, and at the same time, for the traditional gravity model, introduces a variety of topological indexes to correct the mass, gravitational coefficient, and cut-off radius, greatly improving the identification accuracy of key nodes. Brief Description of the Drawings

[0017] Figure 1 It is the overall process framework diagram of the method for identifying key nodes in multi-task stages based on complex networks of the present invention.

[0018] Figure 2 It is the network structure topology graph of the present invention. Detailed implementation manners

[0019] To better understand the present invention, the implementation manners of the present invention will be explained in detail below with reference to the accompanying drawings and embodiments.

[0020] A method for identifying key nodes in multi-task phases based on complex networks according to the present invention, the overall process framework of which is as Figure 1 shown, and includes the following steps:

[0021] S1: Input the communication information flow data between nodes in different task phases of the municipal anti-illegal command system, and simultaneously construct a network structure topology graph G(V, E);

[0022] S2: Assign an initial weight y to the nodes in different task phases according to the entropy weight method;

[0023] S3: Determine the traffic volume ψ and traffic volume ratio Φ of the nodes in different task phases;

[0024] S4: Determine the contribution degree W of the nodes according to the initial weight y and the traffic volume ratio Φ;

[0025] S5: Calculate the degree value, k-core value, eigenvector centrality value, clustering coefficient, shortest path between nodes, and network average shortest path of the nodes according to the network topology structure;

[0026] S6: Calculate the weakening factor, mass, distance, gravitational coefficient, and cut-off radius of the improved gravitational model according to the topological indexes of the nodes;

[0027] S7: Calculate the activity G of the nodes through the improved gravitational model;

[0028] S8: Calculate the importance value I of the nodes according to the contribution degree W and the activity G, and sort the nodes from large to small according to the normalized result of I. The nodes in the top 20% of the sorting results are the key nodes in the complex network;

[0029] Further, in the step S1, the information flow data refers to the information interaction relationship between nodes in different task phases, specifically referring to the information flow from the source node to the sink node, and the size is represented by the message density per minute. The network topology structure graph G(V, E) refers to a network structure composed of m nodes and n edges, where V represents the node set and E represents the edge set.

[0030] Further, in the step S2, the information entropy of the sample data in different task phases is calculated by the entropy weight method, and the formula is:

[0031]

[0032] Among them, H(X) represents the information entropy, m represents the number of samples in the normalized dataset, pi Indicates the proportion of each characteristic index.

[0033] Calculate the weight of the characteristic index through information entropy:

[0034]

[0035] Among them, w i represents the weight of the i-th index, H(X i ) represents the information entropy of the i-th index, and n represents the total number of indexes.

[0036] The initial weight of the node can be expressed as:

[0037]

[0038] Among them, y(i) represents the initial weight of the node, w i represents the weight of the i-th index of the node, b i represents the value of the i-th index, and n represents the total number of indexes. Further, the traffic volume ψ of the node in step S3 refers to the total information flow passing through the node in different task stages, and the traffic volume ratio Φ refers to the traffic volume of the node divided by the total traffic volume of all nodes in this stage. The formula is:

[0039]

[0040] Among them, ψ(i) represents the traffic volume of node i, m represents the total number of nodes, and φ(i) represents the traffic volume ratio of node i.

[0041] Further, in step S4, the contribution degree W is the product of the initial weight y in step S2 and the traffic volume ratio Φ in S3. The formula is:

[0042] W(i) = y(i) × φ(i)

[0043] Among them, y(i) represents the initial weight of node i, φ(i) represents the traffic volume ratio of node i, and W(i) represents the contribution degree of node i.

[0044] Further, in step S5, the degree value of the node represents the number of neighbors of the node. The formula is:

[0045]

[0046] Among them, k d (i) is the degree value of node i, a ij is the edge connecting node i and j. If there is an edge, then a ij = 1. If not, then a ij = 0.

[0047] The k-core value represents the core position of a node in the network. The process is as follows: first, calculate the initial degree of all nodes in the network. Then, start iteratively peeling from the node with the smallest degree value. Remove all nodes in the current network with degrees equal to the minimum value and assign them to the shell (k-shell) corresponding to the current k value. At the same time, update the degree values of the remaining nodes. Repeat this process until the degrees of the remaining nodes are all greater than the current k value, then increment the k value by 1 and continue peeling the next layer until all network nodes are assigned to the corresponding shells.

[0048] The eigenvector centrality value reflects both the number of a node's neighbors and the influence of the neighbors. Its formula is:

[0049]

[0050] where k e (i) is the eigenvector centrality of node i, c is the reciprocal of the largest eigenvalue of the network adjacency matrix, and a ij is the edge connecting node i and j.

[0051] The clustering coefficient reflects the degree of clustering of a node's neighbors. Its formula is:

[0052]

[0053] where cl i is the clustering coefficient of node i, e i is the number of edges existing among the neighbor nodes of node i, and k d (i) is the degree value of node i.

[0054] The shortest path between nodes is the minimum value of all path sets between nodes. The network average shortest path is the average value of all shortest paths between nodes. All topological metrics are obtained by the calculation module.

[0055] Furthermore, in step S6, the weakening factor is the weight of the k-core index, which is used to weaken the influence of the k-core index. Its formula is:

[0056]

[0057] where k dmid 、k emid and k smi respectively represent the median of the degree value, k-core value, and eigenvector centrality value, and k dmax 、k emax and k sma respectively represent the maximum of the degree value, k-core value, and eigenvector centrality value. Taking is to prevent the role of the k-core index from being overly weakened.

[0058] The quality is jointly determined by the degree value, k-core value, eigenvector centrality value, and weakening factor of the node, and the formula is as follows:

[0059]

[0060] Among them, k d (i), k s (i), k e (i), M(i) are the degree value, k-core value, eigenvector centrality value, and quality of node i respectively.

[0061] The gravitational coefficient is determined by the clustering coefficient of the node, and the formula is as follows:

[0062]

[0063] Among them, cl i represents the clustering coefficient of node i, and C(i, j) represents the gravitational coefficient between nodes i and j.

[0064] The truncation radius R is half of the average shortest path of the network and is used to determine the influence range of the node.

[0065] Furthermore, in the step S7, the gravitational value between nodes is the product of the masses of the two nodes and the gravitational coefficient between the nodes divided by the square of the shortest path between the nodes. The activity G of a node is the sum of the gravitational values from this node to all other nodes, and the formula is as follows:

[0066]

[0067] Among them, G(i) represents the activity of node i, C(i, j) represents the gravitational coefficient between nodes i and j, M(i) represents the mass of node i, and d(i, j) represents the shortest path between nodes i and j.

[0068] Furthermore, in the step S8, the importance value I(i) of node i is the product of the contribution degree W(i) of node i in step S4 and the activity G(i) of node i in step S7, and the formula is as follows:

[0069] I(i) = W(i) × G(i)

[0070] The importance values of the nodes are normalized to the maximum value and then sorted from large to small. The top 20% of the sorted results are the key nodes.

[0071] For the convenience of understanding the present invention, the identification of key nodes in the multi-task stage of the present invention will be described below with specific embodiments:

[0072] Example 1: Demonstration of contribution degree calculation

[0073] Construct the information flow data of two phases shown in Table 1, with a total of 4 pieces of information flow data in each phase.

[0074] Table 1 Information flow data of different phases

[0075]

[0076] Calculate the initial weights of each phase node according to the entropy weight method. Phase 1: {Node 1: 0.9, Node 2: 0.9, Node 3: 0.06, Node 4: 0.06, Node 5: 0.04}; Phase 2: {Node 1: 0.05, Node 2: 0.05, Node 3: 0.9, Node 4: 0.9, Node 5: 0.05}.

[0077] The node traffic volume shown in Table 2 can be obtained through the total traffic volume flowing through the nodes in each phase.

[0078] Table 2 Traffic volume of nodes in different phases

[0079]

[0080] Taking the contribution degree of Node 1 in Phase 1 as an example, its value is:

[0081]

[0082] Example 2: Instance verification based on the municipal anti-disorder command system

[0083] In this example, a typical municipal anti-disorder command system including 45 nodes (covering different types such as perception, processing, decision-making, and execution) and 56 edges is constructed as Figure 2 shown. A method for identifying key nodes in multi-task phases based on complex networks, the steps of the method include:

[0084] S1: Input the communication information flow data of the nodes in different task phases of the municipal anti-disorder command system and Figure 2 the network structure topology diagram shown, and part of the information flow data is shown in Table 3;

[0085] Table 3 Part of the information flow data of different task phases

[0086]

[0087]

[0088] S2: Calculate the initial weight y of the node in different task phases according to the entropy weight method;

[0089] S3: Determine the traffic volume ψ and traffic volume ratio Φ of the node according to the information flow data between the nodes in different task phases in Table 3 in step S1;

[0090] S4: Determine the contribution degree W of the node according to the initial weight y in step S2 and the traffic volume ratio Φ in S3;

[0091] S5: Calculate the degree value, k-core value, eigenvector centrality value, clustering coefficient, shortest path between nodes, and average shortest path of the network according to the Figure 2 shown network topology structure;

[0092] S6: Calculate the weakening factor, mass, distance, gravitational coefficient, and cut-off radius of the improved gravitational model according to the topological indexes of the nodes in step S5;

[0093] S7: Calculate the activity G of the node by using the improved gravitational model according to the parameters in step S6;

[0094] S8: Calculate the importance value I of the node according to the contribution degree W and activity G calculated in step S4, sort from large to small according to the normalization result of I, and the top 20% of the sorted results are key nodes. Table 4 shows the importance of nodes in the signal processing and command decision-making stages;

[0095] Table 4 Importance of Nodes in Different Task Stages

[0096]

[0097]

[0098] Although the specific implementation manners of the invention are described above in combination with the accompanying drawings and embodiments, it is not a limitation to the protection scope of the invention. Based on the technical solutions of the invention, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the invention.

Claims

1. A method for identifying key nodes in multi-task stages based on complex networks, characterized in that: The steps include: S1: Input the communication information flow data between nodes in different task stages of the city-level anti-unmanned command system, and construct the network structure topology graph G(V, E); S2: assign initial weights y to nodes at different task stages according to the entropy weight method; S3: Determine the traffic volume ψ and traffic volume ratio Φ of the nodes in different task stages; S4: Determine the contribution W of the node according to the initial weight y and the traffic volume ratio Φ; S5: Calculate the node degree value, k-core value, eigenvector centrality value, clustering coefficient, shortest path between nodes and average shortest path of the network according to the network topology structure; S6: Calculate the weakening factor, mass, distance, gravity coefficient and cutoff radius of the improved gravity model according to the topological indicators of the nodes; S7: Calculate the activity G of the node through the improved gravity model; S8: Calculate the importance value I of the node according to the contribution W and the activity G, and sort the nodes from large to small according to the normalized result of I. The nodes with the highest ranking are the key nodes in the complex network.

2. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1 is characterized by: In step S1, the communication information flow data refers to the information interaction relationship between nodes in different task stages, specifically refers to the information flow from the source node to the sink node, and the size is expressed by the message density per minute; The network topology graph G(V, E) refers to a network structure consisting of m nodes and n edges, where V represents the node set and E represents the edge set.

3. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1 is characterized by: In step S2, the information entropy of the sample data at different task stages is calculated by the entropy weight method, and the formula is: Among them, H(X) represents information entropy, m represents the number of samples in the normalized data set, and p i Indicates the proportion of each characteristic indicator; The weight of the feature index is calculated by information entropy: Among them, w i represents the weight of the i-th indicator, H(X i ) represents the information entropy of the i-th indicator, and n represents the total number of indicators; The initial weight of the node can be expressed as: Among them, y(i) represents the initial weight of the node, w i represents the weight of the node i-th index, b i represents the value of the i-th indicator, and n represents the total number of indicators.

4. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1 is characterized by: In step S3, the traffic volume ψ of a node refers to the total information flow through the node in different task stages, and the traffic volume ratio Φ refers to the traffic volume of a node divided by the total traffic volume of all nodes in the stage, and the formula is: Among them, ψ(i) represents the business volume of node i, m represents the total number of nodes, and φ(i) represents the business volume ratio of node i.

5. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1 is characterized by: In step S4, the contribution W is the product of the initial weight y in step S2 and the business volume ratio Φ in step S3, and its formula is: W(i)=y(i)×φ(i) Among them, y(i) represents the initial weight of node i, φ(i) represents the traffic proportion of node i, and W(i) represents the contribution of node i.

6. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1 is characterized by: In step S5, the degree of a node represents the number of neighbors of the node, and its formula is: Among them, k d (i) is the degree of node i, a ij is the edge between nodes i and j. If there is an edge, then a ij =1, if it does not exist, then a ij =0; The k-core value represents the core position of the node in the network. The process is to first calculate the initial degree of all nodes in the network, and then iteratively peel off from the node with the smallest degree value, remove all nodes with the minimum degree in the current network and assign them to the shell k-shell corresponding to the current k value, and update the degree values ​​of the remaining nodes at the same time; repeat this process until the degrees of the remaining nodes are all greater than the current k value, then add 1 to the k value and continue to peel off the next layer until all network nodes are divided into the corresponding shell layers; The eigenvector centrality value reflects both the number of node neighbors and the influence of the neighbors, and its formula is: Among them, k e (i) is the eigenvector centrality of node i, c is the inverse of the maximum eigenvalue of the network adjacency matrix, a ij is the edge connecting nodes i and j; The clustering coefficient reflects the degree of aggregation of node neighbors, and its formula is: Among them, cl i is the clustering coefficient of node i, e i is the number of edges between node i and its neighbor nodes, k d (i) is the degree value of node i; The shortest path between nodes is the minimum value of all path sets between nodes, the average shortest path of the network is the average value of all shortest paths between nodes, and all topological indicators are obtained by the calculation module.

7. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1 is characterized by: In step S6, the weakening factor is the weight of the k-core index, and its formula is: Among them, k dmid , k emid and k smid Represents the median of degree value, k-core value and eigenvector centrality value, k dmax , k emax and k smax Respectively represent the maximum values ​​of degree value, k-core value and eigenvector centrality value; The quality is determined by the node's degree, k-core value, eigenvector centrality value, and weakening factor, and the formula is: Among them, k d (i), k s (i), k e (i), M(i) are the degree value, k-core value, eigenvector centrality value and quality of node i respectively; The gravitational coefficient is determined by the clustering coefficient of the node, and its formula is: Among them, cl i represents the clustering coefficient of node i, C(i,j) represents the attraction coefficient between nodes i and j; The cutoff radius R is half of the average shortest path in the network.

8. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1 is characterized by: In step S7, the gravitational force between nodes is the product of the mass of the two nodes and the gravitational force coefficient between the nodes divided by the square of the shortest path between the nodes. The activity G of a node is the sum of the gravitational force from the node to all other nodes, and the formula is: Among them, G(i) represents the activity of node i, G(i,j) represents the gravitational coefficient between nodes i and j, M(i) represents the mass of node i, and d(i,j) represents the shortest path between nodes i and j.

9. The method for identifying key nodes in multiple task phases based on complex networks according to claim 1, characterized in that: In step S8, the importance value I(i) of node i is the product of the contribution W(i) of node i in step S4 and the activity G(i) of node i in step S7, and its formula is: I(i)=W(i)×G(i) The importance values ​​of the nodes are normalized to the maximum value and then sorted from large to small. The top 20% of the nodes in the sorting result are key nodes.

Images