The Clustering Problem Of High-dimensional Dynamic Dependent Networks

Complex network analysis has become a major area of research and has been applied in different fields,such as population ecology,social ecology,biology and the Internet.In recent years,the research of complex dynamic network has achieved remarkable results in both the scope of empirical research and the establishment of model and theoretical analysis.The general conclusion of statistical models of random networks has been proposed in the early years.Data analysis is a commonly used method in modern scientific research.It covers communication science,computer science and biological sciences.Cluster analysis plays an important role as the basic component of data analysis.Currently,a number of tools have been created for cluster analysis.In some articles about static network analysis,it is important to focus on clustering methods,mainly to summarize the data through node classification.Previous articles on network clustering mostly supposed that observations are conditionally independent.However,we construct a novel model which combines the stochastic block model,the hidden structure in Markov process and the autoregressive model to relax this assumption.We also propose relative statistical inference and VEM algorithm to calculate the parameters iteratively.In the process of using the traditional EM algorithm,we first use the variational approximation method to find a distribution that is closest to the hidden variable instead of the actual distribution to participate in the iterative process.Then,for the selection of the initial value of the algorithm,we use a more fast clustering method to replace the k-means algorithm to obtain an initial clustering of the individual.Finally,the Monte Carlo simulations are performed well.The MSE of each parameter is relatively small,indicating the consistency and robustness of the work.We also compared the previous method,and the results also show that our method performed better than the previous method.