A Genetic Algorithm (GA) Approach to the Portfolio Design Based on Market Movements and Asset Valuations

Abstract

Vulnerable nature of price forecasts, such as an unpredictability of future and numbers of socio-economic factors that affect market stability, often makes investment risky. Earlier studies in Finance suggested that constructing a portfolio can promise risk-spread gains. While Fund Standardization improved the traditional theories by reducing the computational complexity and by associating every interaction in the portfolio, such a method still cannot become a winning strategy because it does not measure the current value or the relative price of each asset. Inspired by the works of finding returns per risk, we attempt to design an optimal portfolio by searching products that have potential to grow further. More specifically, we first analyze risk-adjusted returns in the previous periods and use their inertia as a momentum. However, because historic movements alone do not fully elucidate future changes nor guarantee positive returns, we scored the relative values of each stock to make more informed estimations. Using the Capital Asset Pricing Model, we measured the values of each stock and determined those undervalued. In this study, we applied a Genetic Algorithm to optimize portfolios while incorporating the momentum strategy and the asset valuations. The proposed GA model was tested in two separate markets, S&P500 and KOSPI200, and projected greater profits than that from both the previous method with momentum method and the market indexes. From the experimental results, the proposed CAPM+ method was found to be very effective in financial data analysis and to lay a groundwork for a sustainable investment execution.

Keywords

• Genetic algorithm
• machine learning
• portfolio optimization
• modern portfolio theory
• investment strategy
• Sharpe ratio