Establishing a data-mining environment for wartime event prediction with an object-oriented command and control database

Abstract

The paper documents progress to date on a research project, the goal of which is wartime event prediction. It describes the operational concept, the data mining environment, and data mining techniques that use Bayesian networks for classification. Key steps in the research plan are as follows: (1) implement machine learning; (2) test the trained networks; and (3) use the technique to support a battlefield commander by predicting enemy attacks. Data for training and testing the technique can be extracted from the object oriented database that supports the Integrated Marine Multi-Agent Command and Control System (IMMACCS). These data were derived from message traffic generated during US Marine Corps exercises. The class structure in the IMMACCS data model is especially well suited to support attack classification.