Dimension Reduction EA for Portfolio Replication Problem: Search Space Change by Fixing Important Variables

Abstract

The portfolio optimization problem is viewed as one of high-dimensional problems. It is difficult to find an optimal solution in the high-dimensional problem because evolutionary algorithms have to determine many design variables for a solution. In this paper, we propose a dimension reduction approach in an evolutionary algorithm for this problem. Our approach changes the search space by the dimension reduction which fixes important design variables. In the numerical experiments, we show that our approach is very useful for the high-dimensional portfolio replication problem.