Multiobjective robustness for portfolio optimization in volatile environments

Multiobjective methods are ideal for evolving a set of portfolio optimisation solutions that span a range from highreturn/high-risk to low-return/low-risk, and an investor can choose her preferred point on the risk-return frontier. However, there are no guarantees that a low-risk solution will remain low-risk — if the environment changes, the relative positions of previously identified solutions may alter. A lowrisk solution may become high-risk and vice versa. The robustness of a Multiobjective Genetic Programming (MOGP) algorithm such as SPEA2 is vitally important in the context of the real-world problem of portfolio optimisation. We explore robustness in this context, providing new definitions and a statistical measure to quantify the robustness of solutions. A new robustness measure is incorporated into a MOGP fitness function to bias evolution towards more robust solutions. This new system (“R-SPEA2”) is compared against the original SPEA2 and we present our results. Cat...

• Type: Conference
• Year: 2008
• Where: GECCO
• Authors: Ghada Hassan, Christopher D. Clack
• Keywords: GECCO 2008, Optimization, Portfolio Optimisation, Portfolio Optimisation Solutions, Robustness