Research On The Key Technologies Of The Construction Of Social Networks Based On Exponential Random Graph

As a specific kind of complex network, social networks play an important role in the research of social interaction among people, and are also key components of the artificial society. As one kind of agent-based modeling and simulation methods, artificial society must describe, model, construct and analysis different kinds of social networks. In order to construct social networks more clearly, briefly and realistically, a framework of constructing social networks in artificial society based on ERGM(exponential random graph models) is verified. ERGM can include configurations both for heterogeneous mixing on exogenous attributes, as well as endogenous structural dependencies. At first, some network configurations should be chosen to model certain social network according to the aim of modelers. Secondly, the parameters should be estimated using Monte Carlo Markov chain maximum likelihood estimation(MCMCMLE) method. Thirdly, the required social network is generated based on special configurations and parameters. At last, the generated network should be verified by some “goodness of fit” methods.The main research work and contributions of the paper are listed as followed:(1) The theoretic of ERGM was introduced to social network construction in artificial society. Traditional modeling methods or social networks, which are based on network mechanism models, such as regular networks, random networks, small-world networks and scale-free networks, couldn’t satisfy the desire of modeling and simulation in artificial society. With its perfective features, ERGM can play a significant role in this situation. The purpose of ERGM, in a nutshell, is to describe parsimoniously the local selection forces that shape the global structure of a network. The possible ties among nodes of a network are regarded as random variables, and assumptions about dependencies among these random tie variables determine the general form of the ERGM for the network. When capturing observed network structure, ERGM can include terms both for heterogeneous mixing on exogenous attributes as well as endogenous structural dependencies. The information gleaned from use of ERGM may then be used to understand a particular phenomenon or to simulate new random realizations of networks that retain the essential properties of the original.(2) An ERGM based social network construct framework was introduced to artificial society. In real networks, there are similar node-level attributes and same structural dependencies in social networks of same type. Therefore ERGM could be well used in artificial society. We introduce a general process of using ERGMs to construct social networks in artificial society. The process contains five steps: observed network acquirement, model selection, parameter evaluation, model construction and model evaluation.(3) An integration framework between ERGM based social network module and artificial society platform was introduced. Only in this way could ERGM based social network models be made full use of. There are two different integration form was produced: integrated by shared document and “peer to peer” likely integration between agent in artificial society and node in ERGM based social network module.(4) Two examples were implemented by using statnet package based on R platform. At first, friendship network in artificial classroom was constructed based ERGM. Secondly, social network between Twitter platform Chinese users who have more than 1000 followers was constructed based ERGM. In conclusion, to construct social network based on ERGM in artificial society is feasible and effective.