Decision support from learning multiple boundaries on military operational plans from simulation data

In this paper we provide decision support regarding robustness during execution of a military operational plan. We learn boundaries from simulated data from alternative plan instances of an expeditionary operation beyond which drastic changes can occur. These are boundaries that an operation must not move beyond without risk of failure. We receive simulated and evaluated plan instances from a simulation-based decision support system. These patterns are clustered by an unsupervised neural Potts spin clustering method into clusters where the instances in each cluster have similar characteristics and outcomes. This gives all plans a classification. We use a belief function based model screening method where all actions of the plan are evaluated as to their differentiating capacity between the two sets of plan instances. All plan instances are projected from their full representation to a subset of actions with high differentiating capacity. We apply supervised learning by support vector machine using the previous classification to learn support vectors for all pair of clusters given the reduced plans from the model screening. From these support vectors we derive a lower dimension hyperplane. One hyperplane between each pair of clusters will make up a full set of boundaries for this operational plan. We provide decision support during execution of an operational plan by continuously calculating the distance from the plan to the closest hyperplane step-by-step as action-by-action is being executed. This way a commander may observe if the operation is approaching a boundary during execution of the plan, beyond which it should not move without risk of failure.