| With the development of online social networks,more and more enterprises are using viral marketing strategies to promote products and disseminate information.The issue of maximizing social influence is common in activities such as product diffusion and information dissemination.Research on this issue focuses on how to select initial users(seeds)to maximize the scope of influence dissemination.This research is of great significance for in-depth understanding of the laws of network and information dissemination,as well as solving practical problems such as product promotion.Current research is mainly based on linear threshold models and independent cascade models,combining different management objectives to design extended models to solve problems.However,there is still a lack of research on introducing individual user sensitivity into communication models,which cannot accurately describe the impact of individual user sensitivity on adoption rates and reflect the individual heterogeneity of users.Users with high individual sensitivity are more likely to accept information and product promotion,while users with low sensitivity are difficult to rely on social communication to receive products and information.Therefore,comprehensive consideration of multiple factors such as individual user sensitivity and the effectiveness of information dissemination to improve the effectiveness of decision-making has become an urgent issue for enterprises to address.In order to address the above challenges,this dissertation proposes an extension problem,namely,the influence maximization problem based on individual sensitivity.By introducing individual sensitivity into the propagation model,an improved propagation model and influence maximization optimization scheme are constructed,and a sensitivity update greedy algorithm for solving large-scale instances is designed to help implement the application of this problem in large-scale networks.In addition,this article explores the role of network parameters in the diffusion of influence based on individual sensitivity,with the aim of assisting enterprises in analyzing product diffusion strategies in different network environments,and providing targeted recommendations for government public opinion control.The specific research content is as follows:(1)Given the initial user,an influence propagation model based on individual sensitivity is constructed.This model introduces individual sensitivity into traditional linear threshold models to depict a more realistic information dissemination process.Firstly,it analyzes the characteristics of individual sensitivity in the dissemination of social influence.Secondly,an influence propagation model based on individual sensitivity is constructed,and model validation verifies that the model conforms to the laws of information propagation in reality.Finally,simulation experiments were conducted with traditional linear threshold models.The experimental results show that the model constructed in this dissertation has a smaller propagation range and shorter diffusion period than the traditional linear threshold model,indicating that many users are not activated under the influence of individual sensitivity,thereby limiting the dissemination effect of information,consistent with the conclusion of user heterogeneity in information dissemination.The experimental results further demonstrate that the model is reasonable and reliable in describing the impact of individual user sensitivity on information dissemination,laying a foundation for solving the problem of maximizing impact.(2)Construct an integer programming model and solution algorithm for the influence maximization problem.Based on the improved propagation model,this dissertation constructs an integer programming model to solve the influence maximization problem,and designs a sensitivity update greedy algorithm(SUG)to solve the model,aiming at maximizing the diffusion range of influence.The results of algorithm testing on real network datasets such as Facebook show that the SUG algorithm designed in this dissertation significantly improves the effect of influence diffusion compared to other algorithms.Specifically,SUG algorithm improves the diffusion range by 50%compared to DD algorithm,saves more than 95% of the calculation time compared to GA algorithm,and achieves a diffusion quality close to that of GA algorithm.With the increase of the diffusion stage,SUG algorithm has always been in the leading position in terms of solution quality.At the initial stage,SUG algorithm has significant advantages in solution quality,and when the threshold range is high,SUG algorithm’s diffusion scale tends to stabilize;In terms of cost calculation,when the influence diffusion scale is the same,the SUG algorithm selects a smaller number of initial users,which is approximately similar to the GA algorithm.Therefore,the SUG algorithm is suitable for the promotion of new products to reduce the initial investment cost.SUG algorithm has shown good performance and strong adaptability in different social networks,providing a valuable reference for larger and more complex social network applications.(3)Explore the effect of network parameters on the diffusion of influence based on individual sensitivity.Based on the constructed influence maximization model and algorithm,this dissertation compares the effects of influence diffusion under four different networks,such as small world networks and scale-free networks,and selects relevant network parameters for sensitivity analysis.The experimental results show that small world networks and scale-free networks have the best diffusion effects,and the smaller the reconnection probability of small world networks,the better the diffusion effect of influence based on individual sensitivity;The higher the threshold range of individual sensitivity,the smaller the range of information dissemination;The larger the network scale,the worse the diffusion effect;The larger the average aggregation coefficient,the larger the range of information dissemination;Due to the dual effects of both increasing the ability to disseminate information and decreasing the allocation weight,the relationship between the average degree and the diffusion effect is not significant.In summary,the models and algorithms constructed in this article can effectively solve the problem of maximizing social influence,while fully considering the role of network parameters in the diffusion effect of influence under individual sensitivity.This study enriches the research content in the field of social influence maximization,and provides new ideas and analytical tools for enterprises to use online simulation computing tools to predict the diffusion effect of information and products,provide effective suggestions for government public opinion control. |
