Detecting Patterns of Life and Anomalous Results (POLAR) in Big Graphs

Sailors and Marines are responsible for conducting missions such as maritime security operations, embassy protection, non-combatant evacuation, and disaster relief. To plan missions effectively, Warfighters need to understand normal patterns of life (POL) and anomalies in those patterns. Analyzing this data within the context of a graph representation supports understanding of how relationships among entities and concepts contribute to normal and anomalous patterns. Merging heterogeneous relational data from multiple sources results in large graphs that require new, scalable analysis methods. To address this challenge, we propose to design and demonstrate an approach for detecting Patterns of Life and Anomalous Results (POLAR) in big graphs. The focus of our effort is to design scalable patterns of life calculations that consider rich details about entities and their relationships. First, we will employ an innovative variation of data preprocessing to reduce graph dimensionality while simultaneously making it easier for analysis techniques to identify patterns of life. Second, to detect patterns of life in the graph, we will represent POLs as labeled sub-graphs of the aggregated graph. We will also augment POL extraction with anomaly detection so operational users can determine when POLs in an area are diverging from expected behavior.