The Research On Influence Maximization Of Heterogeneous Network Based On Adjacency Entropy And Higher-Order Structure

Information network is ubiquitous in real life,such as social network,economic cooperation network,transportation network and so on.Networks are changing and reshaping People’s daily lives.As an important research field in network analysis,influence maximization problem aims to find the most influential seed node set in information network as the initial information transmission source,so as to maximize the information diffusion range of these seed nodes combined together.The research on this issue has important theoretical significance and practical value for controlling public opinion,developing marketing strategy and preventing disease outbreak.At present,most studies on influence maximization are oriented to homogeneous information networks containing only one type of nodes and a single relationship between nodes.However,heterogeneous information networks with various node types and link relations retain more information and realize a more complete and natural abstraction of the real world.Therefore,the introduction of heterogeneous information network is conducive to the study of influence maximization.The existing influence maximization algorithms of heterogeneous networks also have shortcomings such as low time efficiency,inaccurate influence measurement and ignoring high-order structural relations between nodes.In order to deal with the above shortcomings,the main work of this thesis is as follows:(1)The concepts of link entropy and path interaction entropy were proposed to measure the influence of nodes.The link entropy calculation measures the direct influence of a node.The path interaction entropy captures the indirect influence brought by the semantic and structural information of heterogeneous networks.(2)A meta path based adjacency information entropy model(MPAIE)and a meta graph based adjacency information entropy model(MGAIE)were proposed to maximize the influence of heterogeneous networks.(3)Considering the characteristics of local small groups in heterogeneous networks and the attenuation of node influence with the increase of propagation distance,a higherorder structure-based heterogeneous information network influence maximization model(HSIM)was further proposed to maximize the influence of heterogeneous networks.(4)The performance of MPAIE,MGAIE and HSIM models were verified from three aspects of influence range,time efficiency and parameter analysis on two real heterogeneous information network data sets belonging to different fields.The experimental results show that MPAIE and MGAIE models can measure node influence more accurately than other baseline methods,and the computational efficiency of the model is also significantly improved compared with the existing heterogeneous network influence maximization methods.Compared with MPAIE and MGAIE models,HSIM model further improves the accuracy of influence measurement and further reduces the time complexity.