Self-constructing clusters in dynamic heterogeneous multi typed network

Abstract

As dynamic networks such as social and information networks are more ubiquitous, clustering the data on the networks can provide the structure of data in various different models. As well clustering the data in different time windows dynamically, provide the evolution behavior which helps in analyzing the features of the network. For example in the information network such as DBLP which contains multiple types of objects such as author, paper, conference and terms, clustering gives us overall view of evolutionary structure such as continue, merge, split, appearance and disappearance of the multiple objects in heterogeneous networks. In this paper we use Probabilistic generative model along with conditional probability, to generate efficient clusters. The number of clusters is not predefined as well it is not fixed and a prior parameter is added to define the number of clusters dynamically.