7.6 Extrapolating Military Data: A Case Study

Using a rule-generating software package from WizWhy, a branch of the U.S. Navy developed intelligent applications for planning, discovery, and knowledge acquisition and recently received a grant from the Office of Naval Research. The grant gave this branch not only the ability to study new data surveillance and prediction software tools available on the market, but also the authority to determine their price/performance capabilities. Specifically, the Navy wanted to investigate a way to fuse volumes of data in a meaningful way in order to permit data cleansing, analysis, modeling, and prediction. However, the grant provided only a modest budget. This meant they needed to source a COTS product—the Navy's term for a commercial off-the-shelf product.

The Navy recognized that data mining is useful for condensing and extrapolating military information. Data mining helped them to make sense out of arrays of complex distributed heterogeneous sensors. As with many data mining analyses, their project began by collecting data and creating a database. Then, a data mining application was applied to reveal hidden patterns and relationships within the database. A set of rules was revealed explaining the data and issuing predictions for new cases. Based on the discovered rules, unexpected phenomena in the data were discovered.

WizWhy was used to mine those features and to explore, by feedback, the most important features. This in turn provided a better understanding of their working domain and, thereby, permitted the evolution of more complex systems that would otherwise have been impossible. Rule-based tools such as this one can be used to crack the knowledge acquisition bottleneck, which has impeded the growth and dissemination of knowledge-based technologies.