Multi-Attribute Portfolio Selection with Genetic Optimization Algorithms

Abstract

The traditional portfolio theory first proposed by Markowitz only provides a solution to capital allocation to a pre-determined set of assets, regardless of asset quality. To remedy this gap, a multi-attribute asset quality analysis, before asset allocation, is proposed. Thus a two-stage multi-attribute portfolio selection framework that considers asset quality, as well as asset allocation, is formulated. For solving the proposed portfolio selection problem, this study applies genetic algorithms for multi-attribute portfolio selection and analysis. In the first stage, i.e. asset quality evaluation, a genetic algorithm is used to identify good quality assets in terms of asset ranking. In the asset allocation stage, allocation of capital to individual high-quality assets is optimized using another genetic algorithm based on Markowitz's mean-variance theory. Through the two-stage asset evaluation and allocation process, an optimal portfolio can be determined in the context of considering both multiple asset return attributes and risk exposures. Experimental results reveal that the proposed multi-attribute portfolio selection framework provides a very feasible and useful tool to assist investors in planning their investment strategy and constructing their portfolios.