Search

CN-122022786-A - Method, system, equipment and medium for maximizing opinion of social network

CN122022786ACN 122022786 ACN122022786 ACN 122022786ACN-122022786-A

Abstract

The invention provides a method, a system, equipment and a medium for maximizing opinion of a social network, which are used for constructing an optimization problem with the double targets of maximizing the average final opinion value of the network and minimizing the total cost of seed node selection based on modeling results by modeling the social network, so that the method is more in line with the core requirement of realizing optimal propagation under limited resources in practical application; by constructing a multi-objective evolution optimization framework, on one hand, diversity of population foundations can be guaranteed, excessive focusing of certain types of nodes in an initial stage is avoided, on the other hand, the initial population has global coverage and contains community propagation dominant genes, a high-quality foundation is laid for subsequent evolution optimization, the problem of insufficient coverage or uneven propagation force of a traditional sampling community is solved, when the population is subjected to iterative optimization, the seed node selection cost is greatly reduced on the premise that the propagation effect is not seriously reduced, and the defect of evolution operators in local cost fine adjustment is overcome.

Inventors

  • LI SHUO
  • HAN CHUNLEI
  • CHEN YAN
  • GAO CHENJIE
  • YI KAI

Assignees

  • 中国电子科技集团公司第二十研究所

Dates

Publication Date
20260512
Application Date
20251229

Claims (10)

  1. 1. A method for maximizing opinion in a social network, comprising the steps of: Step S1, modeling a social network, and constructing an optimization problem with the total cost selected by a maximized network average final opinion value and a minimized seed node as a double target based on a modeling result; S2, constructing a multi-objective evolutionary optimization framework, wherein the multi-objective evolutionary optimization framework comprises a population initialization method based on a mixed strategy, an evolutionary operator aiming at the optimization problem, and an iterative optimization mechanism combined with cost optimization local search; and S3, carrying out iterative optimization on the population by using a non-dominant ranking genetic algorithm NSGA-II framework and combining the multi-objective evolutionary optimization framework, outputting a group of pareto optimal solutions, and taking the pareto optimal solutions as an optimal seed node scheme containing a plurality of opinion-cost balances.
  2. 2. The method for maximizing opinion of a social network according to claim 1, wherein when modeling the social network in step S1, the social network is abstracted into an undirected graph : ; In the formula, The node set is used for representing users in the social network, each node corresponds to an individual with independent mutual capability and opinion expression capability, and the total number of the nodes is ; The method comprises the steps that an edge set is used for representing interaction relation among nodes, if at least one effective interaction exists between the nodes within a set time window, edges exist, and potential paths of opinion propagation are determined by the existence of the edges; Setting an adjacency matrix For edge collection Is a mathematical mapping of adjacency matrix Is of the matrix dimension of (a) If there is an edge Otherwise 0; Defining binary decision variable vectors : ; In the formula, Representing nodes Is selected as the seed node of the seed, Indicating that it is not selected, co-selected Obtaining seed node sets by using seed nodes The method comprises the following steps: ; based on Friedkin-Johnsen opinion propagation model, if Intrinsic opinion Fix to 1, otherwise Randomly initializing; By solving a linear system Obtaining balanced opinion vectors In which, in the process, A diagonal matrix of node degrees; Is a unit matrix, and the intrinsic opinion vector s is the intrinsic opinion of all nodes The column vectors are arranged in node order.
  3. 3. The method for maximizing opinion in a social network according to claim 2, wherein when the optimization problem is obtained in step S1, the method comprises the following steps: Based on the binary decision variable vector And balance opinion vectors Obtaining the average final opinion value : ; In the formula, Is the i-th component of the balance opinion vector z, and corresponds to the balance opinion value of the i-th node in the network; PageRank-based seed node selection total cost : ; In the formula, Representing seed nodes PageRank value of (2) for reflecting seed nodes Global influence in a social network; based on the mean final opinion value And the seed node selects the total cost Constructing an optimization problem with the double targets of maximizing the average final opinion value of the network and minimizing the total cost of seed node selection: ; In the formula, Representing constraints.
  4. 4. The method for maximizing opinion in a social network according to claim 1, wherein the method for initializing the population based on the mixed strategy in step S2 is as follows: dividing the initial population into two parts, wherein each part accounts for one half of the population scale; The first part adopts a random sampling strategy, and completely randomly selects nodes as seed nodes; The second part adopts a roulette sampling strategy based on a community propagator ranking index CSR, constructs probability distribution according to the CSR value of the node, and preferentially selects the node with high CSR value as a seed node, wherein the community propagator ranking index CSR is expressed as: ; In the formula, Is a node Is used for the degree of (a), Representing nodes A collection of connected communities that are connected to each other, Representing communities Opposite node The contribution of the diversity of the community is that, Representing communities Is used for the ratio of the scale of (a), Representing communities Is used for the internal density of the steel sheet, The method is used for comprehensively reflecting the trans-community propagation potential of the nodes in the network structure.
  5. 5. The method of maximizing opinion in a social network according to claim 4, wherein said constructing an evolutionary operator for the optimization problem in step S2, and an iterative optimization mechanism incorporating cost optimization local search, comprises: generating offspring individuals by adopting uniform crossing operators; A mutation operator turned over according to the position is adopted, and mutation probability is preset; constructing and executing a two-hop neighbor cost optimization local search operator, wherein the executing process comprises the following steps: step A1, screening a two-hop neighbor node set of each seed node as a candidate node aiming at each seed node in a current individual, wherein the two-hop neighbor node comprises a direct neighbor of the seed node and a neighbor of the direct neighbor, and removing the seed node; A2, for each candidate node, simulating an opinion propagation process after the candidate node is replaced by the current seed node, and calculating a replaced average final opinion value Sum total cost ; Step A3, setting opinion effect protection threshold Sum cost optimization threshold If the following conditions are satisfied: ≤ And (2) and ; The candidate node is included in the active candidate set, wherein Mean final opinion value before replacement; Selecting total cost for the seed node before replacement, otherwise, not including a valid candidate set; a4, selecting the total cost from the effective candidate set corresponding to each seed node And reducing the candidate node with the largest amplitude as the optimal replacement node, and executing replacement iterative optimization.
  6. 6. The method for maximizing opinion in a social network as recited in claim 5, wherein the probability of execution of the two-hop neighbor cost optimized local search operator is preset to 0.6, the opinion effect protection threshold Preset to 20%, the cost optimizes the threshold value Preset to 40%.
  7. 7. The method for maximizing opinion in a social network according to claim 1, wherein in step S3, when outputting a set of pareto optimal solutions, the method comprises the following steps: the method comprises the steps of (1) layering a population according to a double target of non-dominant sorting of opinion propagation effects and total cost, wherein individuals at the same level are mutually non-dominant, and dividing the individuals into a plurality of non-dominant levels; Calculating the crowding degree of individuals in the same non-dominant level, and preferentially reserving the individuals with sparse distribution; Adopting a binary tournament selection strategy based on comparison of non-dominant levels and crowdedness, randomly selecting two individuals from a current parent population each time, comparing the non-dominant levels of the two individuals, and selecting an individual with a better non-dominant level if the non-dominant levels of the two individuals are different; The parent population is subjected to local search by adopting a uniform crossing operator, a mutation operator turned according to the position and two-hop neighbor cost optimization, a child population is generated, and the child population is combined into a temporary population with the scale of 2 n; And screening out a new generation of father population through non-dominant sorting and crowding degree calculation, and performing iterative execution until the preset iterative times are reached to obtain a group of pareto optimal solutions, and taking the pareto optimal solutions as an optimal seed node scheme containing a plurality of opinion-cost balances.
  8. 8. A system for maximizing social networking opinion, characterized in that it comprises, based on the method of maximizing social networking opinion of any of claims 1-7: the initialization unit is configured to model the social network, and construct an optimization problem with the maximum network average final opinion value and the minimum total cost selected by the seed nodes as double targets based on the modeling result; The optimization unit is configured to construct a multi-objective evolutionary optimization framework, wherein the multi-objective evolutionary optimization framework comprises a population initialization method based on a mixed strategy, an evolutionary operator aiming at the optimization problem, and an iterative optimization mechanism combined with cost optimization local search; And the output unit is configured to use a non-dominant ranking genetic algorithm NSGA-II framework, perform iterative optimization on the population in combination with the multi-objective evolutionary optimization framework, output a group of pareto optimal solutions and serve as an optimal seed node scheme comprising a plurality of opinion-cost balances.
  9. 9. An electronic device, characterized in that, the electronic device includes: at least one processor, and A memory communicatively coupled to the at least one processor, wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of maximizing social networking opinion of any of claims 1-7.
  10. 10. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method of maximizing social networking opinion of any of claims 1-7.

Description

Method, system, equipment and medium for maximizing opinion of social network Technical Field The invention belongs to the technical field of social network information processing, and particularly relates to a method, a system, equipment and a medium for maximizing social network opinion. Background The current main opinion maximization method is mainly divided into two types, namely a static screening method based on greedy strategies, such as a classical K-center and degree-center algorithm, and a random searching method based on a simple evolutionary algorithm, wherein the core defect of single-objective optimization limitation exists, opinion maximization is only used as a unique objective, and other factors in practical application, such as seed node selection cost, resource input limitation and the like are not combined to perform collaborative optimization, and meanwhile community information utilization depth is insufficient, so that the requirements of social network practical scenes are difficult to meet. Social network nature has the community characteristic, and node propagation influence is showing by community boundary restriction, and cross-community propagation is because of factors such as user interest difference, interaction frequency are low the degree of difficulty is far higher than the intra-community propagation, and current method that contains community consideration is mostly only with the community as node grouping label, does not go deep into the collaborative value of high propagation power node and cross-community bridge node in the mining community, but neglects the preference of community bridge node through evenly distributing seed node to each community, leads to the suggestion to be difficult to break through the community barrier, through the high propagation power node in the overfocusing single community, causes seed node to concentrate on a few communities and forms the propagation blind area. Therefore, the application expects a method capable of ensuring that the seed node achieves the maximum opinion and simultaneously achieves the cooperative balance of the propagation effect and the resource investment while simultaneously achieving the minimum cost. Disclosure of Invention In order to overcome the defects in the prior art in the utilization depth of community information, the requirement of a social network actual scene is difficult to meet, and the problem that only a single optimization target is considered and the cost required for achieving opinion maximization is ignored, the invention provides a method, a system, equipment and a medium for realizing the opinion maximization of the social network. In order to achieve the above purpose, the present invention provides the following technical solutions: In a first aspect, embodiments of the present disclosure provide a method for maximizing social networking opinion, comprising the steps of: Step S1, modeling a social network, and constructing an optimization problem with the total cost selected by a maximized network average final opinion value and a minimized seed node as a double target based on a modeling result; S2, constructing a multi-objective evolutionary optimization framework, wherein the multi-objective evolutionary optimization framework comprises a population initialization method based on a mixed strategy, an evolutionary operator aiming at the optimization problem, and an iterative optimization mechanism combined with cost optimization local search; and S3, carrying out iterative optimization on the population by using a non-dominant ranking genetic algorithm NSGA-II framework and combining the multi-objective evolutionary optimization framework, outputting a group of pareto optimal solutions, and taking the pareto optimal solutions as an optimal seed node scheme containing a plurality of opinion-cost balances. Further, when modeling the social network in the step S1, the social network is abstracted into an undirected graph: ; In the formula,The node set is used for representing users in the social network, each node corresponds to an individual with independent mutual capability and opinion expression capability, and the total number of the nodes is;The method comprises the steps that an edge set is used for representing interaction relation among nodes, if at least one effective interaction exists between the nodes within a set time window, edges exist, and potential paths of opinion propagation are determined by the existence of the edges; Setting an adjacency matrix For edge collectionIs a mathematical mapping of adjacency matrixIs of the matrix dimension of (a)If there is an edgeOtherwise0; Defining binary decision variable vectors: ; In the formula,Representing nodesIs selected as the seed node of the seed,Indicating that it is not selected, co-selectedObtaining seed node sets by using seed nodesThe method comprises the following steps: ; based on Friedkin-Johnsen opinion propagation mo