Finding approximately similar patterns in social networks

Social network analysis has gained considerable momentum due to its importance to investigative and intelligence analysts. Social networks can provide a wealth of information about behavioral patterns of individuals and groups, and can be successfully deployed to identify individuals or anomalous groups engaged in unlawful activities. Most of the tools at analysts' disposal today employ state of the art visual and statistical techniques using which they explore the data to identify potential regions of interest within a vast network and gradually zero-in on targets. However, reusing this knowledge to find approximately similar patterns in the same or another network requires going through the same process all over again. In this paper, we present an efficient searching mechanism for automated detection of approximately similar patterns which not only exhibit similar structure but also have similar attributes. We show that the proposed methods can help analysis of large social networks much more efficiently than pure visual techniques.