Unsupervised Abnormal Nodes Detection in Semantic Graphs

Now a days, homeland security is the most concerned issue for every country. But identification of abnormal entities in large dataset is an important problem in case of homeland security. Though there exist methods of data mining and social network analysis which focus on finding pattern or central nodes from networks or numerical datasets. But none of them make use of semantic information associated with the network. In this paper we describe an unsupervised framework to identify abnormal instances.