Network key node mining method based on discount strategy and improved discrete crow search algorithm

Abstract

This invention relates to the field of computer complex network optimization technology, and discloses a method for mining key nodes in a network based on a discount strategy and an improved discrete crow search algorithm. The method includes: first, preprocessing the citation network to convert it into an adjacency matrix, and then reversing the network to obtain a reverse network; then using the LRDiscount algorithm to initially screen the nodes in the reverse network to obtain a candidate node set. C Next, the candidate node set is optimized according to the local optimization process of the improved discrete crow search algorithm. C Optimization is performed; finally, the optimal set is selected from the optimized node set, and the node influence is evaluated to obtain the final result. k Key seed nodes. Compared with existing technologies, this invention combines the influence discount strategy of network nodes with an improved discrete crow search algorithm. It spreads influence by updating node positions in the process of mimicking crow search, and finds key nodes by the marginal gain generated by individual crows walking in the citation network.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of complex network influence maximization, and particularly relates to a network key node mining method based on a discount strategy and an improved discrete crow search algorithm. BACKGROUND

[0002] With the penetration of various mobile social services into human life and socialization, social networks play an important role in information sharing and information dissemination. The huge scale of social networks has brought great challenges to the research of traditional influence maximization problem, and also made the research of the problem have greater practical significance. IM (Influence Maximization, IM) aims to solve K influential nodes (nodes represent social media users) in a social network, and use the "word-of-mouth" effect to spread information, so that the influence range of these nodes reaches the maximum. Therefore, how to select K nodes in the network under the condition of ensuring time complexity and propagation effect is the main problem faced by the influence maximization problem.

[0003] For the influence maximization problem, the key is how to select the key nodes in the network, which has been concerned by researchers since the 20th century. In the field of information science, with the in-depth study of information dissemination and diffusion in social media, these key nodes play an extremely important role in the dynamic evolution of network structure, propagation control and other researches. Some researchers apply meta-heuristic search algorithm to network key node mining, but the disadvantage is that the algorithm executes slowly and has high time complexity, which cannot be applied in large-scale networks. Therefore, based on the research in this field, Zhang et al. consider that the neighbor nodes will play an important role in measuring the influence of nodes, and therefore propose a heuristic algorithm PRDiscount combined with Pagerank, which explicitly discounts the influence of all individuals with social relationships with the selected seeds. Although the meta-heuristic algorithm improves the optimization efficiency of key node mining to some extent, the single solution algorithm has only one solution in the iteration process, which has the advantages of simplicity and speed in solving the key node mining problem of small-scale networks, and is easy to fall into local optimum, thereby leading to network information redundancy. Gong et al. then propose a key node identification method based on discrete particle swarm optimization algorithm, which defines the position of the particle as the node number and the speed of the particle as the symbol of judging whether the node is updated. After multiple iterations, the global optimal particle Gbest under the search condition is found, and the position of the particle is the optimal seed node set. Therefore, in order to overcome the shortcomings of the algorithm, the current research on influence maximization of key node searching is combined with swarm intelligence optimization algorithm, which simulates the cooperative behavior of biological population or the evolution process of physical phenomenon. Due to its strong heuristic search thinking and global search ability, it has been widely used in optimization problems of influence maximization in recent years.

[0004] Therefore, how to design a method with high accuracy and low cost to obtain seed nodes in large-scale networks is extremely important. SUMMARY

[0005] The purpose of the application is to solve the problems of large network size and large data volume in traditional network key node mining, which leads to low efficiency and high time complexity of direct key node mining. The application proposes a network key node mining method based on discount strategy and improved discrete crow search algorithm, which combines the influence discount strategy of network nodes with the improved discrete crow search algorithm to update the node position in the crow search process to diffuse the influence, and finds the key nodes through the marginal gain generated by the crow individuals walking in the citation network.

[0006] TECHNICAL SCHEME: The application proposes a network key node mining method based on discount strategy and improved discrete crow search algorithm, which includes the following steps:

[0007] S1, pre-processing the citation network, converting the target network to obtain an adjacency matrix of the target network, and performing reverse operation on the citation network to obtain a reverse network;

[0008] S2, according to the discount policy algorithm LRDiscount, discounting the influence of each node in the reverse network in S1 pointing to the seed node to obtain the discounted node influence, and sequentially selecting the node with the maximum value and adding it to the candidate node set C;

[0009] S3, using the local optimization process of the improved discrete crow search algorithm to optimize the candidate node set C, the candidate node set being the candidate crow group, the improved discrete crow search algorithm increasing parallel iteration processing on the basis of the traditional discrete crow search algorithm, that is, the discrete crow search only compares the memory vector corresponding to the crow individual at each iteration, thereby updating the memory vector to satisfy the update of the crow individual position vector, to obtain the optimized node set C * ;

[0010] S4, selecting the optimal set from the optimized node set C * , and performing node influence evaluation to obtain the final k key seed nodes.

[0011] Further, the specific steps of obtaining the candidate node set C in step S2 are:

[0012] S2.1, adding a background node bg to the reverse network, connecting it to all nodes in the network, thereby obtaining a new network of N+1 nodes with strong connectivity;

[0013] S2.2, assigning 1 unit of LR value to the N nodes other than the background node bg, and the LR value of the background node bg being 0;

[0014] S2.3, evenly distributing the 1 unit of LR value to the directly connected out-neighbor nodes, and constantly iterating until reaching a stable state:

[0015]

[0016] wherein, is the out-degree of node v j , w ji is an adjacency matrix element, indicating that there is an edge between node v j and node v i , then w ji =1, otherwise 0;

[0017] S2.4, when the iteration ends, the LR value LR bg of the background node bg is t c) is divided into all nodes in the network, and the final LR value of node v i is obtained;

[0018]

[0019] S2.5, the discount policy algorithm LRDiscount discounts the influence of each neighbor node of each seed node, that is, in the reverse network, the influence of each node pointing to the seed node is discounted, so as to obtain the discounted node influence:

[0020]

[0021] Wherein, S is a set of seed nodes, represents the ratio of the number of seed nodes in the neighbor nodes of node v i to the total number of neighbor nodes;

[0022] S2.6, according to the final obtained node influence, the node with the maximum value in INF is selected in turn and added to the candidate node set C.

[0023] Further, the specific steps of using the local optimization process of the improved discrete crow search algorithm to optimize the candidate node set C in the step S3 are:

[0024] S3.1, initialize each data in the discrete crow search algorithm framework IDCSA, wherein the crow population size is N, the seed node set to be solved is k, the maximum iteration number t max , the initial parameters of the perception probability AP and the local search node neighborhood range S;

[0025] S3.2, according to the candidate node set C obtained in step S2 as a candidate crow population, and initialize the position vector x i =(node1, node2,..., node n ) of the crow population, the memory vector Memory t-1 =[m1, m2,..., m n ] -1 ; and select the initial optimal solution position vector X* from the initialized population;

[0026] S3.3, based on the node coding and the discrete representation of the crow population position vector and the memory vector, construct the discrete search rule of network space:

[0027]

[0028] Wherein, R(r is) is a local search mechanism, and the symbol "∩" defines a logical intersection operation, which aims to compare whether there are duplicate nodes between two position vectors;

[0029] S3.4, based on the node pool of the candidate node set C generated in step S2, define a target function to calculate the function fitting value of N crow individual, use local influence evaluation function LIE to approximate the evaluation of the influence of seed node in crow position vector evolution:

[0030]

[0031] S3.5, the 2-order neighbor nodes of each node in the candidate node set C are locally optimized according to the improved discrete crow search algorithm, if the marginal benefit value of a 2-order neighbor node is relatively large compared with the marginal benefit of the node itself, the 2-order neighbor node is used to replace the node in the current optimal solution, and the execution is repeated until the maximum iteration number t max is reached.

[0032] Further, the specific steps of the local optimization search of the node set according to the improved crow search algorithm in step S3.5 are:

[0033] 1) Calculate the optimal position vector difference between the current crow individual i and the tracked crow individual j On this basis, the cross operation is carried out to obtain the decision vector V node Then judge whether to perform local search optimization;

[0034] 2) Store the first-order direct neighbor nodes of node x i in the node set Neighbors, and then traverse the nodes one by one to find the 2-hop neighbor node set and include it in the node set NodeSet, and after the traversal of the first-order neighbor nodes of each node is completed, remove the duplicate nodes in the node set NodeSet to ensure that there are no duplicate nodes in the 2-hop neighbor nodes;

[0035] 3) Calculate the LIE fitness value of the seed set position vector x i after the corresponding position nodes are replaced by their neighbor nodes, and select the node in the neighbor node set NodeSet that can bring the maximum benefit to the x i vector, and replace the corresponding node in the x i vector.

[0036] Further, after the local search optimization iteration number reaches t mmax in step S4, the node set C * is obtained, and the nodes in the node set C * are selected by the influence maximization algorithm to obtain a key node set k.

[0037] Beneficial effects:

[0038] The application mines key nodes based on a discount strategy and an improved discrete crow search algorithm, solves the problems of slow convergence of the algorithm and low comprehensive influence of the optimal seed set, and specifically shows as follows:

[0039] (1) The mutual influence between nodes in the citation network is "discounted" through the LRDiscount algorithm, and the network topology structure and node attribute information are fully utilized.

[0040] (2) In the local search optimization of the discrete crow search algorithm, the network node discount strategy is considered, so that the initial active state seed set is screened out, which avoids the shortcomings of the initialization seed node set on the convergence speed and the result of the algorithm.

[0041] (3) The crossover operation is added in the discrete crow search algorithm, which not only maintains the richness of the population in the search process and avoids falling into a local optimal solution, but also ensures that the updated best node has no repeated nodes in the vector xi.

[0042] The application aims at the key node mining problem, first converts the target problem into an optimization problem, then screens the initial seed set by using the proposed network node discount strategy, and then optimizes and solves the final seed set by using the improved discrete crow search algorithm. The method optimizes the research on the future influence maximization problem to find key nodes in complex networks, and can achieve better results under the same conditions. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 is the overall flowchart of the application;

[0044] Figure 2 is the LRDiscount algorithm sub-flowchart in the application; Figure 1

[0045] is the local optimization process sub-flowchart of the improved discrete crow search algorithm in the application. Figure 3 Figure 1 is the local optimization process sub-flowchart of the improved discrete crow search algorithm in the application. DETAILED DESCRIPTION

[0046] The application will be further illustrated below in combination with the drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the application and not to limit the scope of the application, and various equivalent modifications of the application made by those skilled in the art after reading the application all fall within the scope defined by the appended claims.

[0047] As shown in the application, the specific steps of the network key node mining method based on the discount strategy and the improved discrete crow search algorithm are as follows: Figure 1

[0048] ​​S1, pre-process the citation network, convert the target network, get an adjacency matrix of the target network, and do the reverse operation on the citation network to get the reverse network.

[0049] S2, according to the discount strategy algorithm LRDiscount, discount the influence of each node pointing to the seed node in the reverse network in S1, get the discounted node influence, and select the node with the maximum value in turn and add it to the candidate node set C, the specific steps are as follows:

[0050] S2.1, in the reverse network obtained in step S1, a background node bg is added to connect all nodes in the network, thereby obtaining a strongly connected new network of N+1 nodes.

[0051] S2.2, assign 1 unit of LR value to the N nodes other than the background node bg, and the LR value of the background node bg is 0:

[0052]

[0053] S2.3, the 1 unit of LR value is evenly distributed to its directly connected out-neighbor nodes, and the iteration is continued until the stable state is reached:

[0054]

[0055] wherein, is the out-degree of node v j , w ji is the adjacency matrix element, indicating that there is an edge between node v j and node v i , then w ji =1, otherwise 0.

[0056] S2.4, when the iteration is finished, the LR value LR bg (t c ) of the background node bg is evenly distributed to all nodes in the network to obtain the final LR value of node v i :

[0057]

[0058] S2.5, the discount strategy algorithm LRDiscount discounts the influence of each neighbor node of the seed node, that is, in the reverse network, the influence of each node pointing to the seed node is discounted, thereby obtaining the discounted node influence:

[0059]

[0060] wherein, S is the seed node set, indicates the node vi The ratio of the number of seed nodes in the neighbor nodes to the total number of neighbor nodes.

[0061] S2.6, according to the finally obtained node influence, the node with the maximum value is selected in INF in turn and added to the candidate node set C.

[0062] S3, using the local optimization process of the improved discrete crow search algorithm to optimize the candidate node set C, that is, the candidate crow group, the improved discrete crow search algorithm increases parallel iteration processing on the basis of the traditional discrete crow search algorithm, that is, the discrete crow search only compares with the memory vector corresponding to the crow individual at each iteration, so as to update the memory vector to meet the update of the position vector of the crow individual, and obtains the optimized node set C * , as shown in the accompanying drawings, and the specific steps are as follows: Figure 2

[0063] S3.1, initializing each data in the discrete crow search algorithm framework IDCSA, wherein the crow population size is N, the seed node set to be solved is k, the maximum iteration number is t max , the initial parameters such as the perception probability AP and the local search node neighborhood range S.

[0064] S3.2, initializing the position vector x i =(node1,node2,...,node n ) of the crow population according to the candidate node set C obtained in step S2, the memory vector Memory t-1 =[m1,m2,...,m n ] -1 ; and selecting the initial optimal solution position vector X * from the initialized population.

[0065] S3.3, based on the node coding and the discrete representation of the crow population position vector and the memory vector, constructing the discrete search rule of the network space:

[0066]

[0067] Wherein, R(r i , s) is a local search mechanism, and the symbol "∩" defines a logical intersection operation, which aims to compare whether there are repeated nodes in the two position vectors.

[0068] S3.4, based on the node pool of the candidate node set C generated in step S2, defining the objective function to calculate the function fitting value of the N crow individuals, and using the local influence evaluation function LIE to approximately evaluate the influence of the seed nodes in the evolution of the crow position vector: ​

[0069]

[0070] S3.5, the 2-order neighbor nodes of each node in the candidate node set C are searched according to the improved discrete crow search algorithm, if the marginal benefit value of a 2-order neighbor node is relatively large compared with the marginal benefit of the node itself, the 2-order neighbor node is used to replace the node in the current optimal solution, and the execution is repeated until the maximum iteration number t max is reached. The specific process is as follows:

[0071] First, the optimal position vector difference between the current crow individual i and the tracked crow individual j is calculated On this basis, the cross operation is performed to obtain the decision vector V node Then, it is judged whether to perform local search optimization.

[0072] Then, the first-order direct neighbor nodes of the node x i are stored in the node set Neighbors, and then the node is traversed in turn, the 2-hop neighbor node set is found and included in the node set NodeSet, and after the traversal of the first-order neighbor nodes of each node is completed, the node set NodeSet is de-duplicated to ensure that there is no repeated node in the 2-hop neighbor node set.

[0073] Finally, the seed set position vector x i is calculated in turn, the LIE fitness value of the node corresponding to the position after being replaced by its neighbor node, and the node in the neighbor node set NodeSet that can bring the maximum benefit to the x i vector is selected, and the node in the x i vector is replaced.

[0074] S4, the optimal set is selected from the optimized node set C * , and the node influence is evaluated, so that the final k key seed nodes are obtained.

[0075] The present application can be combined with a computer system, so as to complete the mining of seed nodes.

[0076] The network key node mining method based on discount strategy and improved discrete crow search algorithm disclosed by the present application can be used for mining key nodes in different scale complex networks.

[0077] The above embodiments are only for illustrating the technical concept and characteristics of the present application, the purpose is to enable those skilled in the art to understand the content of the present application and implement it, and cannot limit the protection scope of the present application. Any equivalent transformation or modification made according to the spirit and essence of the present application should be covered within the protection scope of the present application.

Claims

1. A method for mining key network nodes based on a discount strategy and an improved discrete crow search algorithm, characterized in that, Includes the following steps: S1. Preprocess the citation network, transform the target network to obtain an adjacency matrix of the target network, and perform the reverse operation on the citation network to obtain the reverse network. S2. In the reverse network described in S1, the influence of the nodes in the network is discounted according to the LRDiscount algorithm to obtain the discounted influence of the nodes, and the node with the largest value is selected and added to the candidate node set C. S3. The candidate node set C is optimized using the local optimization process of the improved discrete crow search algorithm. The candidate node set is the candidate crow flock. The improved discrete crow search algorithm adds parallel iterative processing to the traditional discrete crow search algorithm. That is, in each iteration, the discrete crow search only compares with the memory vector corresponding to the current crow individual, thereby updating the memory vector to meet the update of the crow individual's position vector, and obtaining the optimized node set C. * ; S3.1 Initialize the data in the Discrete Crow Search (IDCSA) algorithm framework, where the crow population size is N, the seed node set to be solved is k, and the maximum number of iterations is t. max Initial parameters for the perception probability AP and the nearest neighbor range S of the local search node; S3.

2. Based on the candidate node set C obtained in step S2, use it as the candidate crow flock to initialize the crow population's position vector x. i = (node1, node2, ..., node n ), Memory vector t-1 =[m1,m2,…,m n ] -1 ; and select the initial optimal solution position vector X from the initialized population. * ; S3.

3. Based on node encoding and the discretized representation of the crow flock's position vector and memory vector, a discretized search rule for the network space is constructed: Where R(r) i ,s) is a local search mechanism, and the symbol "∩" is defined as a logical cross operation, the purpose of which is to compare whether there are duplicate nodes in two positional vectors; S3.

4. Based on the node pool of candidate node set C generated in step S2, define an objective function to calculate the function fit value of N individual crows, and use the Local Influence Evaluation Function (LIE) to approximate the influence of the seed node on the evolution of the crow position vector: S3.

5. For each node in the candidate node set C, perform local optimization search on its second-order neighbors using the improved discrete crow search algorithm. If the marginal benefit of a certain second-order neighbor is greater than the marginal benefit of the node itself, then replace that node in the current optimal solution with the second-order neighbor. Repeat this process until the maximum number of iterations t is reached. max Reached the upper limit; S4. Finally, from the optimized node set C * We select the optimal set and evaluate the node influence to obtain the final k key seed nodes.

2. The network key node mining method based on discount strategy and improved discrete crow search algorithm according to claim 1, characterized in that, The specific steps for obtaining the candidate seed set C in step S2 are as follows: S2.1 Add a background node bg to the reverse network and connect it to all nodes in the network to obtain a new network with N+1 strongly connected nodes. S2.2 Assign 1 unit of LR value to N nodes other than the background node bg, and the LR value of the background node bg is 0; S2.

3. Distribute this 1 unit of LR value equally among its directly connected outgoing neighbor nodes, iterating until a stable state is reached: in, For node v j The degree of outgoing, w ji These are elements of the adjacency matrix, representing node v. j With node v i If there is an edge, then w ji =1, otherwise 0; S2.4 After the iteration is complete, change the LR value of the background node bg to LR. bg (t c Distribute the remainder evenly among all nodes in the network to obtain node v. i The final LR value; S2.5, the discount strategy algorithm LRDiscount discounts the influence of each seed node's neighboring nodes. That is, in the reverse network, the influence of each node pointing to the seed node is discounted to obtain the discounted node influence: Where S is the set of seed nodes, Represents node v i The ratio of the number of seed nodes among the neighboring nodes to the total number of neighboring nodes; S2.

6. Based on the final node influence, select the node with the largest value in INF and add it to the candidate node set C.

3. The network key node mining method based on discount strategy and improved discrete crow search algorithm according to claim 1, characterized in that, The specific steps for performing local optimization search on the node set according to the improved crow search algorithm in step S3.5 are as follows: 1) Calculate the optimal position vector difference between the current crow individual i and the tracked crow individual j. Based on this, a crossover operation is performed to obtain the decision vector V. node Then determine whether to perform local search optimization; 2) Move node x i The first-order direct neighbor nodes are stored in the node set Neighbors. Then, the node is traversed in turn to find its two-hop nearest neighbor node set and add it to the node set NodeSet. After the first-order neighbor nodes of each node have been traversed, the node set NodeSet is deduplicated to ensure that there are no duplicate two-hop nearest neighbor nodes. 3) Calculate the seed set position vector x in sequence. i The LIE adaptation value after the node at the corresponding position is replaced by its nearest neighbor node, and the node set NodeSet of these nearest neighbors that can be found in x i The node in the vector that brings the maximum benefit will be the corresponding x. i Replace the nodes in the vector.

4. The network key node mining method based on discount strategy and improved discrete crow search algorithm according to claim 1, characterized in that, In S4, the number of iterations for local search optimization reaches t. max The resulting node set C * Node set C * The nodes in the algorithm select a set of k key nodes using the influence maximization algorithm.

Images