Previsão de preços no mercado de ações brasileiro por meio de técnidas de inteligência artificial

Abstract

In January 2021, the brazilian financial market moved R$35 billions in a single day.That’s might be happened because a recente raise of participants in the Market with an intradayoperator profile (buy and sell assets in the same day). However, even being attractive, there aresome issues to obtain consistente profits. At the same time, had a raise in the studies tha uses technical analysis and fundamentalism with machine learning to forecast price trends, and the neural network the technic more use. This studies analysis the quality to predict 15 minutes foward price trend, using Petrobrás stocks beetween 2014 and 2020. With More than 100 technical indicators, implemented a variable extractor base on eXtreme Gradient Boosting (XGboost) to reduce noise in the sample. A neural network like Long Short Term Memory(LSTM) was applied to predict the trend price, as a result met 56% accuracy. Emphasisthat the best results ocurred to predict downward trend (57%) and upward trend (51%). In particular, it was found that there were improvements in relation to results previously presentedin the literature.