The Research On Influence Measurement And Influence Maximization In Heterogeneous Information Networks

Influence refers to the case when individuals change their behaviors under the influence of others.Influence analysis in social networks can not only understand people's behaviors of people from the angle of sociology,but also deepen their understanding of online social networks,and provide assistance for applications such as information acquisition,product promotion,and public opinion regulation.At present,many social influence analysis methods have been proposed,mainly including two aspects: influence strength measurement and influence maximization.But most of these studies are based on the fact that the underlying network is homogeneous,ignoring the types of nodes and links in online social networks,and few studies have systematically studied how to mine the influence strength and influence dissemination among nodes in heterogeneous information networks.Compared with the analysis of influence in homogeneous networks,heterogeneous links between different types of objects in heterogeneous information network contain different semantics.In the process of influence diffusion,different links have different functions.The process of information diffusion is more complicated than in a homogeneous network.Based on the characteristics of multiple types of nodes and links in heterogeneous information networks that are related to each other and affect each other,this thesis fully considers the different roles of different types of objects and links in heterogeneous information networks.This dissertation investigates two key research issues of influence spreading in heterogeneous information network: influence measurement and influence maximization problem.The main contribution of this thesis includes the following aspects:(1)A meta path-based information entropy for modeling social influence in heterogeneous information networks(MPIE)was proposed,it not only flexibly integrates heterogeneous information,but also obtains potential link information to measure the influence of nodes.(2)Although MPIE obtains rich information in HIN by setting meta path,and comprehensively evaluates the influence between the same type of nodes in HIN,MPIE needs to select meta path by the user-guide way,and cannot directly measures the influence between different types of nodes in HIN.Thus,influence maximization based on network embedding(IMNE)in HIN was proposed,which not only achieving influence maximization,but also directly measure the influence between different types of nodes.(3)This thesis evaluates the performance of MPIE and IMNE on three real heterogeneous information network datasets.The experimental evaluation of the MPIE mainly includes the diffusion range of influence,the ranking of influence,and parameter analysis.The experimental results show that the measurement of the impact of MPIE is more accurate and comprehensive.The infection rate,infection time,and the parameter analysis have shown that IMNE can reach the maximum infection rate in a short period of time compared with baselines.