Research On Mining Methods Of Relatively Important Nodes In Complex Networks

At present,the research method of node importance is mainly to judge the node importance by doing a global ranking for all nodes.However,"relative to a specific node or a group of specific nodes,which nodes are important?" This kind of question reminds us that the relative importance of nodes also have strong practical significance,especially when the network is in large scale.An efficient method for mining relatively important nodes will greatly help us discover hidden nodes with the same characteristics in the network.This paper discussed the mining of relatively significant nodes in complex networks,and put forward three methods for mining relatively significant nodes in the network from three different perspectives.Experimental results proved the applicability and precision of the dished methods.The main research contents are as follows:(1)Based on random walk,a method of neighbor layer diffusion mining relatively important node is proposed.This method hierarchizes the network,and then spreads the semaphore of known important nodes,and takes the final semaphore of unknown important nodes as the relative importance score of the nodes.By comparing with other relatively significant node mining methods on the real network data set,the experimental results show that the Neighbor Layout Diffuse algorithm has a better effect on the relatively important node mining.(2)Based on network topology,a method of the greedy strategy of edge importance mining relatively important node is proposed.This method considers the importance of the edge to represent the degree of association between two connected nodes,and then uses the greedy strategy to find the node corresponding to the most important edge of the known important node in the network every time.By comparing with other relatively significant node mining methods on real network data sets,the experimental results show that the Edge Greedy Strategy algorithm has certain advantages in mining relatively important nodes.(3)Based on machine learning,a method of label propagation mining relatively important node is proposed.This method firstly uses the network representation learning method to map the network into a vector representation.Then,in order to mine relatively important nodes,the network represented by the vector is used as the input of the machine learning algorithm,and the label propagation algorithm in machine learning is used to classify the nodes.Through comparative experiments on real networks,it is proved that the network representation learning ? label propagation algorithm has certain accuracy and applicability in relatively important node mining.