Portfolio optimization with MOPSO-Shrinkage hybrid model

Abstract

This paper introduces a novel framework for portfolio optimization that integrates Multi-Objective Particle Swarm Optimization (MOPSO) with shrinkage covariance estimators, referred to as the MOPSO-Shrinkage hybrid model. The main contribution of this study lies in combining the adaptive search capabilities of evolutionary algorithms with robust covariance estimation techniques to enhance portfolio allocation in mature financial markets. Unlike traditional shrinkage covariance models, which struggle in highly dynamic environments, our hybrid model optimally selects stocks and improves risk-adjusted returns. Empirical analysis on US stock market data from 2013 to 2023 demonstrates that MOPSO-Shrinkage models consistently outperform traditional shrinkage models, achieving higher returns, lower volatility, and superior Sharpe ratios. Among the hybrid models, MOPSO-SSIM exhibits the best performance, with an average annual return of 18.86% and a Sharpe ratio of 1.27, while significantly reducing portfolio risk. Rigorous statistical tests confirm the robustness of the model, showing that MOPSO-Shrinkage significantly outperforms traditional methods. These findings suggest that the proposed approach is well-suited for traders seeking higher risk-adjusted returns and portfolio stability in volatile markets.