Strategic Threat Probabilities with Hierarchical Spatiotemporal Predictions

After many years involved in unconventional warfare and an increase in fiscal constraints, quantitative support for decision making is more important today than in the past. Decisions focused on risk and strategic objectives are critical for changing how we approach the global terrorism threat. This article examines a framework for threat likelihood prediction models based on the underlying patterns in the multivariate, multilevel, spatial, and temporal correlations from previous incidents. The models in this research focus on logistic and discrete choice formulations and use a time-correlated cross-validation evaluation method to measure predictive performance. The information gained from these models serves to inform where national instruments of power could aid in deterring the global terrorism threat.

Methodology

To predict incident probability in a given region within a specified time window, we must first outline a framework, including how it handles data.: [3] We can then approach probability formulations with two models: logistic and discrete choice.