Geographic profiling (GP) techniques for crime analysis have proven useful for identifying the locations where serial killers dwell. In this paper we examine the application of geographic profiling techniques to an organized group of individuals, such as drug dealers and insurgents; in particular, tackling the problem of predicting which facilities in an urban area might support clandestine activities such as drug processing or bomb making. GP techniques assume a single perpetrator whose only observable actions are punctuated killings. In contrast, clandestine organizations involve several distributed individuals who communicate, coordinate, make plans, and execute. Most of their actions, potentially observable such as phone calls, are seemingly innocuous. Through the use of a simulated intelligence stream, we combine GP techniques with plan recognition technology. We advocate a recognition approach which exploits a wide range of knowledge about the group, including the methods of operation, preferences, constraints, and relationships with other like-minded groups. In turn, GP techniques can be augmented with more sophisticated distance metrics using derived geo-spatial attributes, such as cost-of-travel and perceived route risk. We then discuss approaches to fuse all information into a predictive model for each group. This model estimates the risk of future activity based on current observations of group presence. This estimated risk is used to generate actionable products such as security force search paths and prioritization of intelligence collection requests. Finally, we evaluate accuracy of the approach in the presence of noise and incomplete data.