Automated population of dynamic Bayes nets for pre-conflict analysis and forecasting

Conflict-ridden nation-states and the security threats they pose $regional instability, WMD proliferation, drug and terrorism among others $continue to grow in importance to the United States and other global powers. Pre-conflict analysis has traditionally been a labor-intensive effort for skilled social scientists laboring through data for social indicators that precipitate hostility and conflict. Sentient beings have not been able to keep pace with these demands. As a result, the time that sentient beings require to analyze events and build report precludes developing rapid warnings of near events. Our goal is to decrease the time from the acquisition of data to utilization by planners/analysts to within days by augmenting human cognitive capacity (via automated information extraction), and analytical capabilities using social indicators, a generalized framework for conflict analysis and forecasting models based on hierarchical, hybrid and dynamic graphical models. This paper demonstrates that automated transform-based categorizers and linguistic pattern extraction tools, combined with hidden Markov and Bayesian network modeling techniques can automate the most arduous aspects of conflict analysis and forecasting