Comparison of Robust and Data Driven Optimization

Abstract

In this industrial report, we use a new method to solve a portfolio optimization problem in investment banking devision application. In corporate advising service, investment bank devisions need to find an optimal debt structure, in various currencies and payment schemes. Essentially, this is a optimization problem under uncertainty. The objective of the problem is to minimise the expected borrowing cost subject to some risk constraints. In the robust optimization formulation, we use the Conditional Value at Risk (CVaR) as risk measure. In the robust utility optimization formulation, we incorporate the risk preference into the utility functions. We show that the data driven optimization formulation does not work well when the parameter inputs are highly volatile. In that case, robust utility function generate more stable result.