Influence Maximization Based On Community And Path Entropy In Heterogeneous Information Networks

The development of social networks has brought new potential for the rapid dissemination of information,and the determination of influential nodes in the network is regarded as the key factor for this potential to be put into action,so the problem of maximizing influence is put forward.The influence maximization problem aims to find a fixed size seed set in a given network,and then make the final information diffusion reach the maximum through a specific propagation model.Because of its great application potential in the commercial field,influence maximization is favored by researchers.At present,most of the researches on influence maximization focus on the homogenous information network,but ignore the relationship between different types of nodes and the potential semantic information.Compared with the homogeneous information network with a single object and relationship type,heterogeneous information network with rich object types and relationship types can simulate the relationship between objects in the real world more comprehensively.Therefore,influence maximization research on heterogeneous information networks can be closer to the real world.However,the existing influence maximization algorithms for heterogeneous information networks usually extract heterogeneous links between objects in the network based on meta-path and model them as homo-proton graphs to evaluate node influence.This method has the following shortcomings: on the one hand,using a single element path can only capture simple relationships between nodes,ignoring the complex relationships between nodes;On the other hand,the study of influence maximization only on the same proton graph will cause the loss of the original structure information of the network.In order to make up for the above deficiencies,the main work of this thesis is as follows:(1)A meta-structure-based influence maximization algorithm(Meta-structure-based influence maximization,MSIM)for heterogeneous information networks is proposed.Firstly,the algorithm constructs a meta structure for the target nodes in the network to retain the rich heterogeneous information between different objects in the network and the local structure information of nodes.Then,path entropy(The path entropy,PE)and structure entropy(The structure entropy,SE)are proposed to evaluate the influence of a node in the network.Finally,the seed set is selected effectively based on the influence of the node.(2)An influence maximization algorithm(Influence maximization based on community and structure entropy,CMIM)based on community and structure entropy is proposed for heterogeneous information networks.The algorithm firstly divides the target nodes into overlapping community structures,and measures the global influence of nodes based on the community structure of nodes,and appropriately improves the weight ratio of community edge nodes.Then the local influence of nodes is measured based on MSIM algorithm.Finally,the seed nodes are selected by combining the local and global influences of the nodes.(3)In order to verify the effectiveness and efficiency of MSIM algorithm and CMIM algorithm,a large number of experiments are carried out on real data sets.The influence of weight combination of different meta paths on experimental results is explored,and the configuration relationship of related parameters in MSIM algorithm and CMIM algorithm is analyzed.Experimental results show that both MSIM algorithm and MSIM algorithm have good performance and efficiency.