The Application Of Gaussian Mixture Model In The Detection Of Community Structure Of Networks

Based upon the principle that things of a kind come together, cluster analysis is the process of grouping a set of physical or abstract objects into classes of similar objects that share common features. As one of the key techniques of data mining, cluster analysis is not only of great theoretical significance, but has very extensive prospect of applications. It helps to reveal the relations among data. Normal distribution Gaussian mixture model, as an important vector clustering, plays an important role in networks, biology, medicine, economy, etc.According to the relationship between network clustering and vector clustering, we can detect the community structure of complex networks by using vector clustering. In this paper, we use singular value decomposition to transform the network to vector and then apply Gaussian mixture model to detect the community structure. Experiments show that it has very high accuracy. We also build up a framework that may incorporate other clustering methods.For the autoregressive data, we bring up an autoregressive mixing model and solve it with Expectation Maximization Algorithm. Through derivation, we come up with the analytic forms of estimate of the parameters of the model and the detailed numerical solutions. Compared with other algorithms such as K-means, Mclust, etc, this algorithm can reach better results and lower error rate in the clustering of the autoregressive data.