Date of Award
Master of Science
Department of Operational Sciences
Brian J. Lunday, PhD
The objective of this research is to develop procedures that estimate selected, unknown parameters over an adversary's investment portfolio across a set of new or existing technologies. To solve for the selected unknown parameters, it is assumed that the adversary is maximizing the portfolio optimization problem and investing along the efficient frontier. The first technique is when an unknown risk attitude exists but all other parameters were known (i.e. expected return, variance, covariance). An adaptive line search technique that iteratively solved the portfolio optimization problem until the adversary's risk parameter was found. The second problem that was solved was when there are unknown parameters for a new investment option but all other information is known. Alternatively, two variants of a mesh-based grid search were implemented over a three-dimensional space to visualize the feasible region yielding optimal solutions. Additionally, these techniques revealed that there are multiple optimal solutions for the same portfolio allocation in a subregion, which evidence shows may be bounded within a convex region.
DTIC Accession Number
Batista, Keith, "Inverse Optimization: Inferring Unknown Instance Parameters from Observed Decisions" (2022). Theses and Dissertations. 5336.