Research On Community Detection Methods In Online Social Networks

Since the fourth wave of technological revolution,various social platforms have gradually emerged and even developed in a short period of time.They have greatly changed people's traditional social ways,and large-scale and complex online social networks have also come into being.As one of the branches of online social network research,community discovery can further reveal the hidden structure of the network,and also benefit other areas of research.This thesis mainly focuses on the state of online social networks to carry out relevant work,respectively from the static and dynamic aspects of online social networks modeling analysis and theoretical research.A series of scientific experiments were designed in this thesis to test the scientificity and reliability of the proposed model.Based on this,the main research work of this thesis is as follows:(1)In view of the fact that most community discovery methods in online social networks do not consider node self-migration and node deviation,they are unable to extract graph features effectively,leading to unsatisfactory community segmentation results.Therefore,this thesis redesigned the static community discovery algorithm.Firstly,the Laplacian matrix model is established by integrating the basic principle of matrix decomposition.Secondly,considering the high cost of global online social network information acquisition and calculation,a community discovery model combined with local distance is designed.Finally,the node rank optimization function is used to select the best community division.At the same time,11 different social networks are verified and analyzed.Experimental results show that in real networks,the overall performance of the proposed algorithm is improved by nearly 7% compared with the seven most advanced optimization methods.This algorithm is reasonable and effective,and can be extended to multi-scale community discovery.(2)At present,most of the community discovery algorithms are still static,and some of the existing dynamic community discovery algorithms are faced with the problem that the dynamic network model construction can not efficiently represent the whole dynamic network change process,so this thesis designs a phylogenetic planted partition model.Firstly,time dimension is introduced into the typical migration partition model,all states are treated as variables,and the observation equation is constructed.Then,this thesis uses graph optimization strategy to take the observation equation of the whole dynamic social network as the constraint between variables,and constructs the error function,and minimizes the quadratic form of the error function.Finally,this thesis uses Levenberg-Marquardt idea to calculate the gradient of error function,and updates it continuously according to its direction.At the same time,simulation experiments are carried out in the experimental environment of artificial network and real network.The experimental results show that the accuracy of the proposed model is 5% and 3% higher than that of the four advanced methods,respectively,which proves that the proposed model is robust,reasonable and effective.To sum up,this thesis makes an in-depth analysis of the network state characteristics of online social networks and constructs a reasonable and effective mathematical model,which has been verified in multiple artificial networks and real networks.Therefore,the conclusions obtained in this thesis also have a wide range of application value.