Análise da eficácia das estratégias baseadas em padrões de candlesticks com o uso de modelos de machine learning: um estudo sobre a lucratividade e efetividade no mercado financeiro

Abstract

The Brazilian investment market has shown significant growth, driven by the increasing interest of investors. In this context, technical analysis is widely used by those seeking short-term opportunities. Strategies based on candlestick patterns and classical indicators, such as support, resistance, and trends, are commonly applied, although their effectiveness and profitability are often questioned, particularly in assets such as futures contracts and stocks. This study aimed to evaluate the effectiveness of these strategies, both individually and in combination with traditional technical indicators, integrating machine learning models. A comprehensive analysis of trade management in the context of day trading was conducted, focusing on identifying the most favorable timeframes for achieving profitable results. Among the findings, the strategy using 500-point stop gain and stop loss parameters combined with the Random Forest model demonstrated strong performance and consistent profitability in day trading. The results contribute to a critical evaluation of traditional strategies and highlight the potential for optimization through computational approaches.