Finanças Quantitativas e Machine Learning: Uma Análise Preditiva do Risco Sistêmico em Mercados Emergentes

This scientific article investigates the role of machine learning algorithms, such as deep neural networks and random forests, in anticipating systemic risks in emerging financial markets. Using high-frequency time series data and robust statistical models, we compare the performance of traditional techniques (GARCH, VAR) with machine learning-based approaches for estimating Value-at-Risk (VaR), Expected Shortfall, and CoVaR. Monte Carlo simulations and sensitivity analyses are employed to validate the models. Furthermore, algorithmic bias and ethical implications of automating decisions in volatile markets are discussed. The study draws on theorists such as Nouriel Roubini, Nassim Taleb, John Hull, and Marcos López de Prado, aiming to broaden the debate between predictive efficiency, financial risk, and algorithmic responsibility.