指数随机图(p*)模型不同描述的对比研究

Comparative analysis of different descriptions in exponential random graph models(p*)

以pajek软件中的3组密度不同、大小不同的数据为例,运用Markov随机图模型和根据Snijders的新描述改进的2个指数随机图模型,分别对上述3组数据进行对比分析。结果表明:Markov随机图模型存在退化问题,并且网络越大、密度越高,这种退化问题越严重;改进后的模型拟合结果优于Markov随机图模型,因此Markov随机图模型可以被改进后的指数随机图模型替代。

Three sets of data in the pajek program were taken as an example,with different data densities and sizes. The Markov random graph model and two exponential random graph models based on Snijders＇s new specification were used to make comparative analysis on the three sets of data.The results show that the decay problem occurs in the Markov random graph model and that the degradation becomes more serious with increasing size and density of the network.The results also indicate that the exponential random models have higher goodness of fit compared with the Markov random model,therefore,the Markov random model can be replaced by the improved exponential random model.