Multi-objective evolutionary algorithms for financial portfolio design

Abstract

Efficient portfolio design is a real challenge in the area of computational finance. Optimisation based on Markowitz (1959) two-objective mean-variance approach is computationally expensive for real financial world. Practical portfolio design introduces further complexity as it requires the optimisation of multiple return and risk measures. Some of these measures are non-linear and non-convex. Three well known multi-objective evolutionary algorithms, i.e., Pareto envelope-based selection algorithm, micro-genetic algorithm and multi-objective particle swarm optimisation are chosen and applied for solving the bi-objective portfolio optimisation problem which simultaneously maximise the return and minimise the associated risk. Performance comparison is obtained by carrying out using practical data. The results demonstrate that MOPSO outperforms the existing two methods for the considered test cases. � 2010 Inderscience Enterprises Ltd.