Research On Influence Seeding Strategies From The Perspective Of Social Distance

With the rapid development of social media,research on influence seeding strategies in social networks has attracted extensive attention from experts and scholars,and has also been applied in different fields such as business marketing,content recommendation and public opinion control.As the scale of social networks and the data of information transmission increase with each passing day,the structural characteristics and relational attributes of network individuals also become increasingly complex.These characteristics make the study of influence seeding diffusion in social networks more challenging.However,most of the current researches on seeding diffusion remain in the context of a single structure of social networks,and seldom consider the significant role of individual relationship characteristics and network structure complexity in the study of influence seeding strategies.Therefore,from the perspective of individual’s social distance in different dimensions in social network,this paper studies influence seeding strategy in social network by integrating model construction,agent-based simulation experiment design,data analysis and other methods and technologies.The key research is as follows:(1)research on influence seeding strategy in single-layer social network considering neighborhood relations;(2)research on influence seeding strategy considering multiple relationships in multi-layer social network;(3)research on influence seeding strategy considering interlayer relationship in multi-layer social network.Specifically,the innovative research content of this paper mainly includes the following three aspects:(1)Firstly,from the perspective of social distance between individuals in a single-layer social network,combined with the role of individual neighborhood structure,complexity and uncertainty of relationship characteristics in analyzing network node centrality,the third chapter of this paper proposes a measure of social influence centrality based on information entropy,which considers influence factors of social distance such as intimacy and trust degree of the relationship within the range of strong connection.In addition,the effectiveness of the proposed centrality index is compared with the traditional centrality index through the influence diffusion evaluation methods such as single-stage seeding strategy and sequential seeding strategy,and the performance of the proposed centrality index is verified.Finally,the influence of model parameters on seeding strategy is analyzed.(2)Secondly,considering the factors of social distance change caused by multiple social relationships between individuals in multi-layer social networks,the fourth chapter of this paper studies the construction of centrality of individual influence in weighted multi-layer networks and the effectiveness evaluation based on different seeding strategies.By extending the classical gravity model,a centrality metric is proposed to measure the influence of individuals in multi-layer networks,in which the number of neighbors is the individual’s "quality" and the social distance between two individuals is the individuals’ "distance".Social distance considers two characteristics of social relations: the superposition of social relations between individuals in different network layers will reduce their social distance,and the strength of social relations between individuals will decrease with the increase of distance.The effectiveness of the proposed centrality measure is verified by comparing the correlation between influence sequences of different centralities and the real influence sequence,and analyzing the influence coverage and diffusion rate of centralities under different seeding strategies.Finally,the relationship between distance radius parameter and the correlation of influence ranking is discussed.(3)Finally,in view of the fact that there exist conversion costs of influence between different layers in actual multi-layer social networks,in the fifth chapter of this paper,a novel two-stage seeding strategy is proposed to measure layer centrality and node centrality in weighted multiplex networks from the perspective of individuals’ social distance between different layers,and the effectiveness of different centrality is evaluated by diffusion dynamics process.First,the target layer and influential seed nodes located in it are determined based on the layer centrality and seeding centrality respectively,and then the validity of the propsed centralities is verified on different multi-layer network data sets by using a redefined independent cascade model.Finally,the sensitivity of neighborhood parameters to target seeding layer selection and influence diffusion is also discussed.This thesis conducts in-depth research on the seeding strategy of influence in social networks,which not only expands the research scope for the diffusion of social influence,but also provides reference for relevant practical application fields.