Assessing and Comparing by Specific Metrics the Performance of 15 Multiobjective Optimization Metaheuristics when Solving the Portfolio Optimization Problem

Abstract

The financial market has undergone a major revolution in recent decades with the advance and spread of multiobjective optimization metaheuristics, which can be more successfully applied to various aspects of financial decisions. Portfolio optimization is one of them. The purpose of this paper is to adapt, implement in Matlab, assess and compare the performance of 15 metaheuristics belonging to four different classes (NSGA, MOPSO, MOEA/D and SPEA classes) when applying to the Markowitz's Portfolio Optimization Problem. For comparing the performance of these algorithms, we use several specific metrics quantifying the convergence and/or the diversity of the approximate Pareto front against the true Pareto front. The optimal portfolios in the sense of Pareto are selected from a universe of 20 assets listed on the Bucharest Stock Exchange.

An Evaluation Study on Investment Efficiency: A Predictive Machine Learning Approach

“…Apipie and Georgescu [9] applied multiobjective optimization metaheuristics to solve a portfolio optimization problem. By analyzing the characteristics of listed real estate companies and the ideological basis of the support vector machine theory, they designed the evaluation index selection principle, formed the evaluation index system composed of ten indexes, and established the support vector machine regression model.…”