Handling the complexities of the multi-constrained portfolio optimization problem with the support of a novel MOEA

Abstract

The incorporation of additional constraints to the basic mean–variance (MV) model adds realism to the model, but simultaneously makes the problem difficult to be solved with exact approaches. In this paper we address the challenges that have arisen by the multi-constrained portfolio optimization problem with the assistance of a novel specially engineered multi-objective evolutionary algorithm (MOEA). The proposed algorithm incorporates a new efficient representation scheme and specially designed mutation and recombination operators alongside with efficient algorithmic approaches for the correct incorporation of complex real-world constraints into the MV model. We test the algorithm’s performance in comparison with two well-known MOEAs by using a wide range of test problems up to 1317 stocks. For all examined cases the proposed algorithm outperforms the other two MOEAs in terms of performance and processing speed.

Keywords:

• Multiobjective optimization
• evolutionary algorithms
• representation scheme
• portfolio optimization
• cardinality
• class constraint
• round-lots
• pre-assignment constraint

Notes

Please note this paper has been re-typeset by Taylor & Francis from the manuscript originally provided to the previous publisher.