This project is my undergraduate thesis, written with Dyego Soares de Araujo at the University of Brasilia in electrical engineering under the original Portuguese title “Aprendizado por Reforço Aplicado ao Mercado Financeiro”.
The work explores whether a reinforcement learning system can use classic financial indicators to outperform traditional stock market trading strategies. The implementation used SARSA for the buy-side decision process while keeping sell strategies comparable to classical baselines.
According to the thesis abstract, the intelligent strategy achieved more stable returns and in some tests reached results up to six times higher than the classical strategies evaluated.
It is also one of the earliest pieces of work connecting the themes that kept showing up later in my career: AI, finance, and real-world decision systems.