Applying Spatial - Temporal Model and Game Theory to Asymmetric Threat Prediction

Applying Spatial-Temporal Model and Game Theory to Asymmetric Threat Prediction Accurate predictions of enemy course of actions (ECOA) are important to the command and control optimization strategies in long-lasting battles. In most Command and Control (C2) applications, the existing techniques, such as spatial-temporal point models for ECOA prediction or Discrete Choice Model (DCM), assume that insurgent attack features/patterns, or at least the trends of behavior patterns, are static. However, this static assumption is no longer true for intelligent and organized insurgents in recent antiterrorism war. These insurgents sometimes deliberately violate [Rational???] probability theory predictions so they can apply surprise attacks to create more casualties and spread terror. In this paper, a new game theoretic framework is proposed for modeling dynamic changes of enemy behavior features and predicting future threats. This framework semantically combines several different approaches; namely, a feature prediction game, higher level hybrid data fusion, techniques to provide concrete spatial-temporal modeling and prediction, emotion analysis of adversary rationality and non-rationality, deception identification and modeling, hierarchical knowledge representation, and a non-zero sum stochastic adversarial Markov game. We mainly describe the modification of existing spatial-temporal point models, the fusion of dynamic game feature selection technique and dynamic cohesiveness feature selection technique, the ontology about selected/unselected features, and construction of probability predictions.