Research On The Method Of Random Walks In Social Networks

Social network refers to the relatively stable relationship between the individuals of the society due to the interaction.With the development of the information technology,social network has gradually been becoming a new cross-cutting research hotspot.Random walks is an ideal mathematical state of Brownian motion.Because it has no rules and can not predict the future steps and development,it is widely used to standardize modeling methods in the social network.Google's page ranking algorithm,called PageRank,is a well-known application in computer field.In the social network analysis,random walk method not only reveals the structure of the network,but also simulates the flow and the trendency of the information in the real network,which can help us to deeply and comprehensively analyze the robustness and the stability of the network.However,for the large-scale undirected network,the iterative computation time of the transfer matrix convergence process in the random walks is very large.For the large-scale directed network,the common random walks can not be carried out in this network.This paper studies the traditional random walks and the biased random wlks,for undirected social network and directed social network,and respectively proposes an algorithm to study the characteristics of the structural and the information diffusion of the social network,called hierarchical community detection algorithm based on partial matrix approximation convergence(PMAC)on random walks and community-based information diffusion(CID)using biased random walks.In hierarchical community detection algorithm based on partial matrix approximation convergence(PMAC)on random walks,since the network has natural power-law distribution and community structure,we use the core nodes to replace all nodes in the network.Then,we use the error function of the global convergence of the transfer matrix in network to determine the steps random walks.Finally,the initial communities are formed centering the core nodes.Using the evaluation function of the community to assess and iterative merge communities can get the last reasonable community division,which greatly improves the efficiency of the algorithm.The experiment uses three real network datasets to analyze,in the large-scale network,compared with the other two classic algorithms,the best step of the algorithm gets the best results.In community-based information diffusion(CID)using biased random walks,which significantly reduces the computational complexity.Because the traditional random walks is not applicable to the directed network,it uses the biased function of the the random walk to control the direction of random walks,that is called information diffusion direction.Finally,it links the total increasing function of the influence to independent cascade model to carry out the information diffusion,it not only makes a wider range but also reduces the running time.The experiment uses two datasets to analyze,in a large-scale network,compared with the other two classic algorithms,biased random walks to get a better result.