Análise de tendência em gráfico de velas com redes neurais para a ação PETR4.SA (petróleo brasileiro - Petrobras)

Abstract

Artificial Neural Networks (ANNs) are part of the field of machine learning, which corresponds to the ability to teach a machine certain patterns or activities from data. Inspired by the functioning of the human nervous system, ANNs have a high capacity to process large volumes of information and identify complex patterns, being widely used in time series analysis, graphs and, especially, in the financial market. In this sense, the present work focuses on the analysis of trends in the financial market, with a focus on the PETR4.SA (Petrobras) stock. Thus, as a main objective, this research proposes to predict the upward or downward movement of this asset in certain time windows, facilitating the visualization of possible opportunities and reducing risks. The methodology consists of using historical data ranging from January 2001 to December 2024, making use of LSTM (Long Short-Term Memory) neural networks due to their effectiveness in dealing with time series problems. Layer adjustments were made to the network, and performance evaluation was conducted using metrics such as accuracy, precision, and F1-Score, aiming for more reliable results. The final considerations will indicate to what extent the constructed model can assist investors and researchers in understanding the movements of the analyzed asset.