Digital Evolution of ETF Portfolios with Multiobjective Genetic Algorithms and Filtering by Financial Metrics

The construction of efficient personal portfolios required balancing conflicting objectives, such as maximizing returns and minimizing risk, while respecting operational constraints, investor profiles, and investment horizons. This study proposed an integrated decision-support model that combined asset screening based on financial metrics with multiobjective optimization using the NSGA-II Genetic Algorithm to build Exchange Traded Fund (ETF) portfolios in the U.S. market. The research addressed the growing complexity of investment selection in a market with more than 5,000 ETFs traded in the United States. Initially, ETFs were screened according to performance indicators, including risk, liquidity, diversification, robustness, and reliability. Subsequently, NSGA-II was applied to generate efficient portfolios by balancing expected-return maximization and risk reduction. Rather than recommending a single portfolio composition, the method identified a frontier of diversified solutions and demonstrated the risk-return trade-off through the Sharpe ratio. Its applicability was demonstrated by the generation of portfolios aligned with conservative, moderate, and aggressive investor profiles. For short-term investors, the solutions prioritized higher liquidity and lower exposure to volatility; for medium-term horizons, they sought a balance between risk and return; and for long-term horizons, they allowed greater exposure to assets with appreciation potential and diversification. It was concluded that integrating financial screening and multiobjective optimization offered a structured and replicable approach to support ETF investment decisions, making asset selection more consistent amid the wide variety available in the U.S. market.