The Research On Maximizing The Influence Of Heterogeneous Networks Based On Weighted PageRank And Structural Similarity

Social influence refers to the phenomenon that an individual’s opinion or behavior is affected by the behavior of others,which is mainly reflected through the interaction between people.The research of social influence has a long history in the fields of economic sociology and marketing,which has laid a certain foundation for the study of Influence Maximization(IM)in the dissemination of social networks.In the social network,everyone does not exist independently.They are connected with each other and have more or less influence on each other.The depth and scope of their influence depend on many factors.IM problem is an important part of social impact analysis.Most social networks contain many different types of nodes and connecting edges,which essentially belong to heterogeneous information networks.However,previous studies about IM mostly stay in homogeneous information networks,which consider a single type of nodes,connecting edges and connecting paths,This is different from the real social network.For heterogeneous information networks,the key to maximizing their influence lies in how to use the heterogeneous information in the network to identify the most influential nodes.The main work and innovations of this paper are as follows:(1)An influence maximization algorithm based on weighted PageRank(Comprehensive Weighted PageRank,(CWPR))is proposed.The algorithm decomposes the heterogeneous information network into several networks with only one connection type,and then assigns the weight of the corresponding edge according to the number of connection relationships between nodes.The decomposition of the network simplifies the complex network structure,and the distribution of weight distinguishes the importance of the connection relationship between nodes,which is helpful to accurately measure the influence between different nodes.The measurement of influence considers the direct and indirect influence of various types of nodes,so as to better describe the complexity and heterogeneity of node influence in the network,comprehensively retain the heterogeneous information in the information network,and make the seed nodes found have high influence.(2)An influence maximization algorithm based on meta-Path Structure Similarity(MPSS)is proposed.The shortest meta path containing different types of nodes is selected according to the topology and node type of nodes.In heterogeneous information networks,the selection of path covers different node types and connection edge types,comprehensively retains the information in heterogeneous information networks,and effectively captures the heterogeneity of nodes.At the same time,considering the global structure of nodes,the high influence seed set selected is more convincing.(3)The effectiveness of CWPR algorithm and MPSS algorithm is verified on two public real data sets.CWPR evaluates the performance of the algorithm from three aspects: propagation range,parameter influence and edge weight,and MPSS evaluates the performance from propagation path,range and influence parameters.Experimental results verify that the proposed methods are better than the baseline method.