Predictive Anomaly Resolution via Cases (PARSEC)

Predictive anomaly resolution via cases (PARSEC) is part of the Defense Advanced Research Project Agency’s PANDA (Predictive Analysis for Naval Deployment Activities) program. PANDA provides a transformational capability for maritime domain awareness by supporting automated monitoring of ships anywhere in the world.

Most ships follow predictable routes, but sometimes deviate in space (such as a new port) and time (such as loitering). PANDA learns kinematic models for each ship and detects deviations from normal patterns. While most deviations are legitimate and do not warrant analyst attention, a small subset can be attributed to piracy, smuggling, or other activities of interest to analysts. PARSEC is responsible for determining which deviations merit analysts’ attention.

PARSEC uses probabilistic case-based reasoning (PCBR) to explain and classify each new deviation as threatening or benign. PCBR uses previous and similar cases and current context information, such as weather, port closures, ownership changes, and commodity prices, to classify cases and provide possible explanations for deviations (for example, avoiding a hurricane or changing destination due to a port closure).

By automatically filtering benign deviations, PARSEC enables analysts to monitor 10,000 to 100,000 vessels, an increase of several orders of magnitude over current capabilities.