Multi-social-platform budget influence maximization method based on multi-population parallel optimization

Abstract

The invention discloses a multi-social-platform budget influence maximization method based on multi-population parallel optimization. The method comprises a model building module, a seed node selection module and a seed node identification effect evaluation module. Model establishment is a task of establishing a multi-layer network model. A seed node selection process is composed of population initialization, a repair function, a fitness function, a selection operator, a crossover operator and a mutation operator, and an individual with the highest fitness value is selected as a final solution, namely a seed node set for maximizing multilayer network influence propagation. And finally, simulating a propagation experiment through a multi-layer network, and evaluating an actual effect. According to the method, a multi-population parallel optimization framework and a repair function are designed, the search range is wider, the optimal result can be found under the condition of lower time complexity, the problem that budget limitation is exceeded possibly in the training process is effectively solved, and waste of excellent individuals is avoided. The method has great application in the aspects of economic utility and social service.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and relates to a method for maximizing the budget influence of multiple social platforms based on multi-population parallel optimization, and belongs to the study of the problem of maximizing the budget influence on a multi-layer network. The main goal of this study is to design an efficient algorithm to identify the optimal seed node set to maximize the propagation scale under the premise of determining the budget, combining the topological properties of the multi-layer network and the superior performance of the genetic algorithm. Background Art

[0002] The problem of maximizing budget influence is raised due to the needs of marketing strategies, especially in the context of the rapid development of modern business and digital technology, a strong digital marketing strategy is crucial. Individuals, brands, and organizations are taking various measures to attract more attention and fans to expand their influence. For example, merchants need to formulate reasonable marketing strategies to recommend product information to the right user groups, but aimless promotion will lead to a certain waste of resources, and merchants also need to bear high promotion costs, which is obviously not appropriate. Therefore, the study of maximizing budget influence takes users as the entry point, analyzes the network structure at a lower cost, locks in high-influence users from a local perspective, and then identifies the group combination with the maximum influence from a global perspective, providing merchants with excellent marketing strategies.

[0003] The study of maximizing influence also has broad application prospects in disease prevention and control. The problem of maximizing influence was first applied in the era of word-of-mouth marketing, when information could only be spread in a single-layer network abstracted from the relationship between relatives and friends. Today, online social networks have become a key carrier for information dissemination, social interaction, and resource sharing. Traditional single-layer network models are difficult to accurately capture the complex relationships and interactions in the real world. Therefore, multi-layer networks have attracted much attention as network models that are closer to reality. For example, networks of different layers can simulate information dissemination patterns between multiple social platforms; simulate transfer patterns in transportation systems; and the interaction patterns between proteins, genes, and diseases in biological systems.

[0004] Theoretical and methodological research on maximizing budget influence in multi-layer networks is still in its infancy. Information propagation and influence diffusion in multi-layer networks have become more complex and critical. Nodes can not only influence each other within the same layer, but also propagate information across layers, which poses new challenges and opportunities for maximizing influence. Deeply understanding the information propagation mechanism in multi-layer networks, designing more effective influence maximization algorithms, and applying them to practical scenarios have become areas that urgently need attention and exploration. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-social platform budget influence maximization method based on multi-population parallel optimization. Specifically, in order to better adapt to the task of maximizing budget influence, the present invention adds a repair module to the framework to provide strong theoretical support and practical guidance for fields such as social network marketing, disease control, and public opinion guidance, and apply it to actual scenarios to promote a comprehensive understanding and application of influence propagation in multi-layer networks.

[0006] The method of the present invention is specifically as follows: Step (1) Build a multi-layer network model ; Indicates Layer network, consisting of a set of nodes , the edge set within the layer and the node cost set constitute, , is the total number of network layers.

[0007] A set of seed nodes selected from a multi-layer network For subsequent influence propagation, node collection The cost is expressed as the sum of the costs of each node in the node set.

[0008] Step (2) Meta-heuristic method to select seed nodes: (2-1) Initialization: Use multiple populations to evolve in parallel to generate Initial population set , each population Included individuals, each of which contains a set of randomly selected nodes as the initial solution. Each individual In non-repeated nodes selected from the multi-layer network and randomly select an individual As its potential initial optimal solution, .

[0009] (2-2) Using the repair module to replace nodes with low cost-effectiveness with nodes with higher cost-effectiveness; node Cost-effective ,in and Respectively represent nodes The true propagation capacity and price of Individuals in The price is greater than the set budget ,Right now , use the repair function to repair the individual, as follows: Ⅰ. First calculate the current individual The cost performance of each node in the individual node is Replace with the node with the highest cost-performance ratio in the multi-layer network ; II. Recalculate the individuals after replacement Price ; III. If , then loop Ⅰ and Ⅱ until , and obtain the repaired individual .

[0010] (2-3) From the initialized parent population Randomly select two individuals and Perform a crossover operation to and The corresponding nodes in The probability of exchange is to ensure that there are no repeated nodes in the exchanged individuals, and finally generate two new offspring and , if the new individual and If the cost exceeds the budget, the repair module in step (2-2) needs to be further adjusted to convert the individual and Input the repair function so that the output individual and satisfy and ;pass The crossover operator operation generates Offspring individuals.

[0011] (2-4) The offspring Each node in The probability of mutation is any node in the multi-layer network node set and ensures that there are no duplicate nodes in the mutated individuals; similarly, if the cost of the mutated individual exceeds the budget limit, it still needs to be further adjusted by the repair module in step (2-2); the crossover and mutation operators generate The offspring individuals are added to the original parent population to form a new candidate population. .

[0012] (2-3) Using individuals as the initial propagation node set to perform Monte Carlo simulation on a multi-layer network, the final propagation influence range That is the fitness value of the individual, expressed as ; The roulette wheel selection method is then used for selection.

[0013] right Population After the independent evolution operation of the round, the length is The optimal solution set , where each element represents the optimal solution of the corresponding population, and finally from the optimal solution set The individual with the highest fitness value is selected as the final optimal solution of the model, that is, the set of seed nodes used to maximize the propagation of multi-layer network influence. .

[0014] Step (3) takes the optimal solution obtained as the seed node set The propagation simulation is performed on the multi-layer network to calculate the propagation influence range of the seed node set. The MIC propagation model is used to simulate the propagation dynamics of the node on the multi-layer network. The state of each node is described as two possible states: inactive and active. The active node is expressed in terms of probability. Infect its inactive neighbor nodes. Each activated node has only one chance to activate the uninfected neighbor nodes. If it is not activated, no more attempts will be made. When a node is activated in a certain layer, it is assumed that the corresponding nodes in other layers also have the ability to spread from the next time step. A simulated propagation experiment is carried out on the above MIC propagation model as the initial seed node. When the propagation is terminated, the number of nodes in the network that are in an activated state is counted as the propagation influence range, and the propagation influence range of the selected seed node is quantified.

[0015] Beneficial effects of the present invention: The present invention designs a meta-heuristic budget influence maximization algorithm, which fully considers the topological properties of the multi-layer network, and through simulating the process of biological evolution, it continuously iterates and optimizes by selection, crossover and mutation to find the best node combination to achieve the purpose of the largest propagation range. The core highlight of this method is to fill the gap in the research on the problem of maximizing the budget influence on the multi-layer network, adopt a multi-population parallel optimization framework, ensure the efficiency of the algorithm, and improve the effect of the algorithm; at the same time, the design of the repair function module solves the situation of exceeding the budget during the individual evolution process. When the way of information dissemination is no longer limited to a single social tool, the importance of multi-layer network research is self-evident, and it has great applications in both economic applications and social services. This method fits the multi-level and diversified themes of the current era. From the perspective of scientific research, it provides new ideas and methods for the study of maximizing the budget influence of multi-layer networks; from a practical perspective, in the context of the widespread use of multiple social platforms, advertising companies achieve the greatest publicity effect at the lowest cost; the government publishes information on different platforms, improves the credibility of the published information, and speeds up the dissemination of official information. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The figure is a flow chart of the overall framework of the method of the present invention. DETAILED DESCRIPTION

[0017] The present invention will be further described below by means of specific examples.

[0018] A method for maximizing budget influence on multiple social platforms based on multi-population parallel optimization. Taking into account the budget constraints in real situations, the price of the node is linked to the multi-level degree of the node. In the process of selecting seed nodes, it is necessary to consider whether the set budget value is exceeded. This method imitates the biological evolution process in nature, and obtains the optimal set of seed nodes through three training processes of selection, crossover, and mutation. During the training process, a repair function is designed to solve the situation where the total price of individuals may exceed the budget limit, enriching the method of selecting seed nodes with maximized influence on multi-layer networks. At the same time, considering that the optimal solution of the algorithm may appear in individuals of different lengths, the idea of ​​multi-population parallel evolution is adopted to design the algorithm. Specifically, Figure 1 As shown:

[0019] Step (1) Build a multi-layer network model ; Indicates Layer network, consisting of a set of nodes , the edge set within the layer and the node cost set constitute, , is the total number of network layers.

[0020] Each layer in a multi-layer network has the same nodes, represented by , Indicates nodes, .

[0021] Different layers have different edge relationships. Intra-layer edge set , Indicates Strip edge, is the number of edges.

[0022] Node cost set To represent the price of the node, Representation Node The sum of the corresponding node degree values ​​in each layer, ; , indicating a node In the The degree value of the layer network.

[0023] A set of seed nodes selected from a multi-layer network For subsequent influence propagation, node collection The cost of is expressed as the sum of the costs of each node in the node set, that is, .

[0024] Step (2) Meta-heuristic method to select seed nodes: (2-1) Initialization: Use multiple populations to evolve in parallel to generate Initial population set , each population Included individuals, each of which contains a set of randomly selected nodes as the initial solution. Each individual In non-repeated nodes selected from the multi-layer network and randomly select an individual As its potential initial optimal solution, .

[0025] (2-2) Repair function: Since the individual cost generated by initialization may exceed the preset budget ,Through the repair module to adjust the over-budget individuals, the repair module replaces the ,low-cost-effective nodes with nodes with higher cost-effectiveness, ,which reduces the individual’s selection cost while ensuring the individual ,influence effect.

[0026] node Cost-effective ,in and Respectively represent nodes The real communication capacity and price of the population can be used to more accurately replace expensive but low-influence nodes from individuals. Individuals in The price is greater than the set budget ,Right now , use the repair function to repair the individual:

[0027] Ⅰ. First calculate the current individual The cost performance of each node in the individual node is Replace with the node with the highest cost-performance ratio in the multi-layer network ; II. Recalculate the individuals after replacement Price ; III. If , then loop Ⅰ and Ⅱ until , and obtain the repaired individual ; (2-3) Crossover operator: From the initialized parent population Randomly select two individuals and Perform a crossover operation to and The corresponding nodes in The probability of exchange is to ensure that there are no repeated nodes in the exchanged individuals, and finally generate two new offspring and , if the new individual and If the cost exceeds the budget, the repair function of step (2-2) needs to be further adjusted to make the individual and Input the repair function so that the output individual and satisfy and .pass The crossover operator operation generates Offspring individuals.

[0028] (2-4) Mutation operator: In order to further expand the search space of solutions, the mutation operator generates Specifically, the offspring individuals Each node in The probability of mutation is any node in the multi-layer network node set and ensures that there are no duplicate nodes in the mutated individuals. Similarly, if the cost of the mutated individual exceeds the budget limit, it still needs to be further adjusted by the repair function in step (2-2). The offspring individuals are added to the original parent population to form a new candidate population. .

[0029] (2-3) Selection operator: The selection module aims to select Select The better individuals serve as the new parent population In order to evaluate the quality of individuals, the individuals are used as the initial propagation node set to perform Monte Carlo simulation on the multi-layer network, and the final propagation influence range is It is the fitness value of the individual, expressed as Then the roulette wheel selection method is used for selection operation, and the specific process is as follows:

[0030] a. Use the fitness function to calculate each individual in the candidate population The fitness value of ; b. Calculate each individual The survival probability , and sort the survival probability of each individual in ascending order to obtain a probability list ; c. Generate a random number , , traverse the elements in the probability list and accumulate them. When the accumulated value is greater than or equal to The individual corresponding to the survival probability at is the selected individual; d. Repeat step c until the selected individuals as the new parent population ; e. The individual with the highest fitness value and the historical optimal solution of the population Compare and choose the one with the larger fitness value as the new optimal solution, that is: .

[0031] right Population After rounds of independent evolution (including crossover, mutation, and selection), the length of The optimal solution set , where each element represents the optimal solution of the corresponding population, and finally from the optimal solution set The individual with the highest fitness value is selected as the final optimal solution of the model, that is, the set of seed nodes used to maximize the propagation of multi-layer network influence. .

[0032] Step (3) Maximize the evaluation of seed node recognition effect: Use the optimal solution obtained by the above algorithm model as the seed node set The propagation simulation is performed on the multi-layer network to calculate the propagation influence range of the seed node set. The MIC propagation model is used to simulate the propagation dynamics of the node on the multi-layer network. The state of each node is described as two possible states: inactive and active. The active node is expressed in terms of probability. Infect its inactive neighbor nodes. Each activated node has only one chance to activate uninfected neighbor nodes. If it is not activated, it will not try again. When a node is activated in a certain layer, it is assumed that the corresponding nodes in other layers also have the ability to spread from the next time step. A simulated propagation experiment is carried out on the above MIC propagation model as the initial seed node. When the propagation is terminated, the number of nodes in the network that are in an activated state is counted as the propagation influence range, and the propagation influence range of the selected seed node is quantified.

Claims

1. A multi-social platform budget influence maximization method based on multi-group parallel optimization, characterized by: Step (1) Build a multi-layer network model ; Indicates Layer network, consisting of a set of nodes , the edge set within the layer and the node cost set constitute, , is the total number of network layers; A set of seed nodes selected from a multi-layer network For subsequent influence propagation, node collection The cost is expressed as the sum of the costs of each node in the node set; Step (2) Meta-heuristic method to select seed nodes: (2-1) Initialization: Use multiple populations to evolve in parallel to generate Initial population set , each population Included individuals, each of which contains a set of randomly selected nodes as the initial solution. Each individual In non-repeated nodes selected from the multi-layer network and randomly select an individual As its potential initial optimal solution, ; (2-2) Using the repair module to replace nodes with low cost-effectiveness with nodes with higher cost-effectiveness; node Cost-effective ,in and Respectively represent nodes The true propagation capacity and price of Individuals in The price is greater than the set budget ,Right now , use the repair function to repair the individual, as follows: Ⅰ. First calculate the current individual The cost performance of each node in the individual node is Replace with the node with the highest cost-performance ratio in the multi-layer network ; II. Recalculate the individuals after replacement Price ; III. If , then loop Ⅰ and Ⅱ until , and obtain the repaired individual ; (2-3) From the initialized parent population Randomly select two individuals and Perform a crossover operation to and The corresponding nodes in The probability of exchange is to ensure that there are no repeated nodes in the exchanged individuals, and finally generate two new offspring and , if the new individual and If the cost exceeds the budget, the repair module in step (2-2) needs to be further adjusted to convert the individual and Input the repair function so that the output individual and satisfy and ;pass The crossover operator operation generates offspring individuals; (2-4) The offspring Each node in The probability of mutation is any node in the multi-layer network node set and ensures that there are no duplicate nodes in the mutated individuals; similarly, if the cost of the mutated individual exceeds the budget limit, it still needs to be further adjusted by the repair module in step (2-2); the crossover and mutation operators generate The offspring individuals are added to the original parent population to form a new candidate population. ; (2-3) Using individuals as the initial propagation node set to perform Monte Carlo simulation on a multi-layer network, the final propagation influence range That is the fitness value of the individual, expressed as ; Then the roulette wheel selection method is used for selection; right Population After the independent evolution operation of the round, the length is The optimal solution set , where each element represents the optimal solution of the corresponding population, and finally from the optimal solution set The individual with the highest fitness value is selected as the final optimal solution of the model, that is, the set of seed nodes used to maximize the propagation of multi-layer network influence. ; Step (3) takes the optimal solution obtained as the seed node set Simulate propagation on multi-layer networks and calculate the propagation influence range of seed node sets; The MIC propagation model is used to simulate the propagation dynamics of nodes in a multi-layer network. The state of each node is described as two possible states: inactive and active. The active node is expressed in terms of probability. Infect its inactive neighbor nodes. Each activated node has only one chance to activate an uninfected neighbor node. If it is not activated, it will not try again. When a node is activated in a certain layer, it is assumed that the corresponding nodes in other layers also have the ability to propagate from the next time step. A simulated propagation experiment is carried out on the above MIC propagation model as the initial seed node. When the propagation is terminated, the number of nodes in the network that are in an activated state is counted as the propagation influence range, and the propagation influence range of the selected seed node is quantified.

2. The method for maximizing budget influence of multiple social platforms based on multi-group parallel optimization according to claim 1, characterized in that: Each layer in the multi-layer network in step (1) has the same nodes, represented by , Indicates nodes, ; Different layers have different edge relationships. Intra-layer edge set , Indicates Strip edge, is the number of edges; Node cost set To represent the price of the node, Representation Node The sum of the corresponding node degree values ​​in each layer, ; , indicating a node In the The degree value of the layer network.

3. The method for maximizing budget influence of multiple social platforms based on multi-population parallel optimization according to claim 1, characterized in that: In step (2-3), the roulette wheel selection method is used for selection operation, and the specific process is as follows: a. Use the fitness function to calculate each individual in the candidate population The fitness value of ; b. Calculate each individual The survival probability , and sort the survival probability of each individual in ascending order to obtain a probability list ; c. Generate a random number , , traverse the elements in the probability list and accumulate them. When the accumulated value is greater than or equal to The individual corresponding to the survival probability at is the selected individual; d. Repeat step c until the selected individuals as the new parent population ; e. The individual with the highest fitness value and the historical optimal solution of the population Compare and choose the one with the larger fitness value as the new optimal solution, that is: .

Images