Threat Indications for Guiding an Effective Response (TIGER)

The ability to collect and store such surveillance data over large areas and long time periods provides a rich database to mine for patterns of IED and other threat activity, but current data mining is human-intensive and limited to non-real-time forensic analysis of threat events. We propose to extend current capabilities to predictive and real-time analysis by developing an automated tool which estimates threat patterns based on historical data, and detects and extrapolates emerging instances of these patterns from real time data. Our predictive clustering algorithm computes a list of predicted threat events based on real time data, prioritized by likelihood and characterized by statistical uncertainty of location and time of each event. Complementing the predictive clustering algorithm, a forensic clustering algorithm processes historical sensor data and threat events to estimate parameters in the predictive clustering model. The Phase I effort developed an efficient clustering algorithm and demonstrated its ability to detect abnormal maneuvers using both simulated and real moving target data. Phase II will extend the algorithms to use more realistic models of threat patterns and normal background, and evaluate performance using both simulated data to determine performance operating characteristics and real GMTI data to evaluate military utility. BENEFIT: The research will produce automated algorithms to predict threat events from multi-source data. In particular, it will output a statistical estimate of the event likelihood as a function of event location and time. This estimate can be used in military planning to avoid or counter potential threats. In addition, the predictive clustering method proposed could be applied to many commercial problems of predictive analysis such as fraud detection, risk analysis, and analytical customer relationship management.