Robust Sensitivity Analysis of Courses of Action Using an Additive Value Model

摘要

The Department of Defense (DoD) requires the ability to quantifiably measure progress in arenas that are complex and difficult to measure, such as the stability of a region. Therefore, the DoD works diligently to predict the effect of operations and sponsors research to improve prediction and analysis. They desire a repeatable, systematic methodology to aid in the selection of courses of action (COA) that efficiently meet stated objectives and quantitatively measure the degree of accomplishment of these objectives. The author proposes a value-focused thinking (VFT) decision analysis (DA) approach to this problem. This methodology not only aids in selection of possible COAs, but provides a framework to compare the effectiveness of implemented actions via key indicators. Due to the complex nature of COA selection and assessment, weights within the DA model are often fluid. Sensitivity analysis provides the justification of COA selection in such an environment. This thesis focuses on conducting further analysis of the ranked alternatives through a robust sensitivity analysis technique. Sensitivity analysis begins with the examination of the top ranked alternative by varying one weight at a time, one- way sensitivity. The author then proposes a more robust examination of multiple weight sensitivity using five unique measures and optimization via linear and non-linear programming. The measures reveal the alternatives sensitive to small simultaneous variations of multiple weights within the model, n-way sensitivity. Small measure values indicate sensitive alternatives, and indicate to a field commander where to more closely examine the consequences of a selected COA.