Forecasting Dynamic Group Behavior in Social Media

This proposal is to develop a new system for Recognition of Communities and Heterogeneous Analysis of Group Interactions and Dynamics (RC-HAGID) for Social Media Forecasting. Social media enables individuals and groups to quickly coordinate, recruit and direct people to take action in support of a particular objective. This includes terrorist groups and criminal organizations. Accordingly, for full national security it is essential that our intelligence analysts be able to identify and understand the formation and dissolution of groups, interactions between and within groups, and changes in"state"of groups (i.e. from recruitment to carrying out some collaborative action). However, existing techniques for supporting this type of work are limited. In our new and innovative approach, we combine a number of sophisticated techniques that have been developed to address each of the above limitations, and combine them into one integrated system. The resulting tool will be able to monitor social media, identify distinct groups, apply a set of complimentary analyses in order to characterize group"state"and inter- and intra-group interactions based on individual level characteristics in recognition of their heterogeneous nature, forecast changes in group state, and identify the key factors involved with those changes in state.