Evolution computation for investment portfolio optimization

Abstract

This study investigates the application of genetic algorithms (GA) in portfolio optimization, with a focus on NASDAQ 100 and S&P 500 datasets. The aim was to surpass traditional methods, such as Modern Portfolio Theory (MPT), in adapting to the complex and dynamic financial markets. Our approach involved optimizing key GA parameters and exploring the adaptability of GA to different market scenarios using windowing techniques. A significant finding was the variance in GA performance across the NASDAQ 100 and S&P 500 datasets, with the NASDAQ 100 showing more volatility due to its tech-heavy composition. This influenced the effectiveness of symmetric windows like Quarter-to- Quarter (QTQ) and Month-to-Month (MTM) in capturing rapid market trends. In contrast, the broader S&P 500 index reflected steadier trends, where QTQ notably outperformed MTM, suggesting that window selection should be customized to dataset characteristics. The results demonstrate GA’ potential as an effective alternative to traditional portfolio management methods, highlighting the need for considering market dynamics and dataset specificities in their application. This study opens avenues for future research into real-time data analysis and adaptive strategies in financial portfolio optimization.