Analysis On Node Influence And Heterogeneity Of Social Network

With the rapid development of social economy,science and technology,explosively expanding data becomes a common opportunity and challenge that many industries are confronted with,which makes big data analysis become a hot issue nowadays.Network data analysis is an important part of big data analysis.The analysis on node influence and heterogeneity of social network is the main concern of this thesis.The vital nodes usually play important roles in accelerating or controlling the spread-ing.Node influence ranking is to sort nodes based on influential score or centrality of nodes generated by reasonable and accurate measure method.The difference in influence capability of nodes in the network demonstrates the heterogeneity of the network.The heterogeneity analysis of the network is to measure the degree of the difference.As a consequence,these two problems studied in this thesis are closely related.First,a novel method based on improved h-index is presented for evaluating node influence in this thesis.The proposed method provides an additional distinguishing value on the basis of original h-index to differentiate the influence capability of nodes.Although the proposed method has the same computational complexity with the original h-index,it performs better in terms of the accuracy and distinguishing ability.For instance,in Netscience network,the cumulative number of infected nodes by top-10 nodes in the ranking list of the improved h-index approaches 260,while the cumulative number of infected nodes by top-10 nodes in the ranking list of the original h-index is less than 30.Second,a new neighbors and node'5 location based method for evaluating node in-fluence is proposed in this thesis,which takes into account a node's influence on its direct and farther neighbors(h-index and semi-local centrality)and the node's position in the network(the improved k-core).Therefore,combining both of local and global information of node together makes the proposed method a reasonable strategy and appropriate to the practice in evaluating node influence.The improved k-core can effectively distinguish the nodes with the same k-core value and iteration times compared with original k-core.We verify the accuracy of two presented methods above in evaluating node influ-ence with the help of the Susceptible-Infected-Recovered(SIR)model and Kendall's tau correlation coefficient.Additionally,the complementary cumulative distribution function(CCDF)is employed to evaluate the monotonicity of the node influence ranking.The extensive simulation results in several real social networks demonstrate the effectiveness of the proposed methods.At last,real networks usually exhibit heterogeneous nature,since different nodes play far different roles in structure and function.Therefore,not only the identification of vital nodes is of great significant,but also the measure of the heterogeneity of social network is very necessary.Heterogeneity measures of social network based on Laplacian centrality(HLC)and degree ratio(HDR)and local neighborhood(HLN)are presented in this thesis.Furthermore,heterogeneity measures of social network based on the commu-nity's size,edge abundant degree and density are explored,respectively.These proposed heterogeneity measures of network are analyzed on the basis of real social networks.