The recent change in emphasis from reactive to proactive law enforcement, as evidenced by such concepts as community-oriented policing, has resulted in the need for tools to support these efforts. While geographic information systems (GISs) have been very successful at tracking criminal activity, proactive law enforcement requires systems that anticipate the emergence of criminal activity. One such system under development at Carnegie Mellon University and the Pittsburgh Bureau of Police is an early warning system that incorporates a GIS and artificial neural networks to predict the emergence or flare ups of drug hot-spot areas. The system obtains its input from cell-aggregated GIS-based data, processes the data with a previously trained artificial neural network and outputs the results to a choropleth map indicating those areas for which the network has predicted a relatively high number of 911 calls for service for drugs. The chapter describes the system¦s development, explains some of the underlying theory behind the neural network, and compares the network's performance to some of the more traditional geographic forecasting methods. Figures, tables, references