Study Of Influential Nodes Mining Algorithms In The Complex Networks

The inherent heterogeneity of the real networks leads to the different influences of nodes in network structure and functions,while finding out and protecting the key nodes in the network is essential to maintain the normal operation of the network.From the perspective of transmission dynamics,identifying influential nodes and controlling them can prevent the rapid spread of infectious diseases in the network and avoid rumors and other inappropriate comments from spreading to the whole network;from the stability of the network structure,protecting vital nodes can prevent urban traffic from being paralyzed and redistribute limited power resources to ensure normal power consumption in the city.Based on graph theory,it is of great practical significance to study the influential nodes mining algorithm in complex networks and accurately identifying the key nodes is useful to restrain the spread of infectious diseases,control the spread of rumors,improve congested traffic,and rationally allocating power resources.From the perspective of network invulnerability and information dissemination,this paper combines the characteristics of the network topology and the attributes of the individuals in the network.Then,based on the mutual attraction between nodes and the multi-attribute decision-making scheme,the importance of nodes in complex networks is studied.The node's importance ranking algorithm is also used to find key users in social networks and the main work is as follows:1.An algorithm is proposed for evaluating the importance of nodes in complex network based on mutual attraction between nodes-global similarity centrality.The diversity of user's needs determines that people will only pay attention to messages which are of interest,and information that is irrelevant to them will not be forwarded or transmitted.Therefore,users who are far from the information source are less likely to obtain message.Then,by analyzing the relative position of pair of nodes in the network designs a node importance evaluation algorithm that considers the indispensability of the node's dissemination of news.The algorithm obtains the relative position relationship between nodes,and calculates the similarity of the constructed node distribution vector and distance vector to characterize the different roles of nodes in the process of transmitting information.Then from the perspective of the global attributes of the network,the weight of nodes is calculated according to the characteristics of each node and its similarity to other nodes,and the important nodes in the network are found based on this.The proposed method has been applied to a variety of networks of different scales and compared with existing algorithms in performance.The experimental results show that the algorithm has high accuracy and low time complexity,and is suitable for large-scale networks.2.A node influence ranking algorithm based on node similarity and multi-attribute decision making scheme is designed-TOPSIS based on relative entropy.Multi-attribute decision making scheme is a method suitable for evaluating multiple indicators,by calculating the difference between the proposed scheme and the best-worst scheme to evaluate the pros and cons of the scheme.However,in the specific calculation process,the scale of the scheme is ignored so that the final result is often only applicable to a specific network and the actual situation is quite different.Based on this,Jensen-Shannon entropy is used to construct the similarity matrix of the network,which can reflect the similarity degree of each node in the process of information transmission.The similarity matrix is applied to replace the traditional adjacency matrix,and the concepts of Pearson correlation coefficient and relative entropy are introduced to solve the defects of the multi-attribute decision making scheme.Above this,the algorithm also comprehensively considers the amount of information provided by each indicator and the degree of overlap of information,so as to determine the objective weight of each indicator,which can more specifically reflect the objective information of the node.Experiments in multiple real networks and artificial networks show that the accuracy of the proposed algorithm is better than existing algorithms.3.Mining important nodes in complex networks is applied to social systems and a node ranking method based on H-index and cumulative distribution function is proposed to find the most key users in the system.When traditional H-index index evaluates the spreading capability of nodes in the network,only the degree value of neighboring nodes is considered and the mutual attraction between nodes is ignored.Moreover,nodes with different influence are often assigned the same weight,which leads to the low feasibility of the algorithm.Inspired by this idea,we put forward an Extended Mixing H-index Centrality( ),which improves the accuracy of algorithm by analyzing the correlation between nodes and neighboring nodes,and according to the similar degree between nodes using the cumulative distribution function to assign a larger weight to the important neighboring nodes,so that this type of nodes have a larger proportion and the common nodes only have a smaller ratio.Simulation experiments in several classic real social networks show that the proposed method can rank the spreading capability of nodes more accurately than existing algorithms.