Research On The Identification Method Of Important Nodes In Complex Networks With Temporal Characteristic

Identification of important nodes in complex networks is one of the core tasks in complex network research.Accurately identifying the important nodes in the network is helpful for researchers to further understand the topology of the network and explore the dynamics of the network.At present,the research on identification of important nodes in traditional static complex networks have achieved a series of rich achievements.However,with the deepening of the research,researchers found that the topology of the actual network in real life often changes with time,showing the temporal characteristics.The traditional static network model can not describe the temporal characteristics of the complex network.In order to overcome the defects of traditional network models,researchers proposed the concept of the temporal network to describe temporal characteristics of complex network systems.The study of temporal networks is also considered as the second revolution in the field of complex networks.Temporal networks,high-order networks,adaptive networks and dynamic processes on networks are considered as four hot topics in the future complex network researchThis thesis focuses on the modeling method of temporal networks and the identification of important nodes of the temporal networks.The main research contents are as follows:(1)Research on the modeling method of temporal networks.It is found that the mainstream super-adjacency matrix models usually use only a single parameter or only local topological information when describing the coupling relationship between network snapshots.Based on this,a super-adjacency matrix model based on restart random walk is proposed.The ability of restart random walk to obtain the global topology is used to describe the coupling relationship between different time layers of the temporal network more accurately,which lays a foundation for the identification of important nodes in the temporal network.(2)It is difficult to measure the importance of nodes from a global perspective in the temporal eigenvector centrality.In order to solve this problem,this thesis on the basis of improved super-adjacency matrix model,the introduction of widely used in multi-source information fusion of D-S evidence theory.The global importance of each node was obtained by multi-source information fusion,and the eigenvector centrality of temporal network based on D-S evidence theory(ERSAM)was proposed.In order to verify the effectiveness of ERSAM algorithm,a comparative experiment is carried out on the public temporal network data set.The accuracy of ERSAM algorithm and other four important node identification methods is compared by correlation coefficient.The experimental results show that,ERSAM algorithm can significantly improve the accuracy of important node identification in temporal networks.(3)Due to the high computational complexity of current identification methods for important nodes in temporal networks,they cannot be applied to large-scale networks.In this thesis,the tensor representation of multi-layer graph model is extended to the temporal network,and an improved PageRank centrality based on partial random walk is proposed according to the temporal characteristics of temporal network.The stability coefficient between two nodes is defined according to the duration of edge appearance between two nodes in the network.The coefficient is introduced into the traditional random walk as a deviation parameter to control the transition probability of the random walk,and two transition probability tensors and a set of tensor equations are constructed.Then,the fixed point iterative method is used to solve the tensor equation,and the two stable probability vectors are the importance vector of nodes and the importance vector of layers.Finally,a comparative experiment is conducted on a large scale data set of public temporal networks.Experimental results show that compared with other important node identification methods in temporal networks,the improved PageRank centrality based on topological band random walk can identify important nodes in the network more accurately.