Aplicação de redes neurais LSTM para previsão de séries temporais financeiras

Abstract

Making forecasts in the financial market can bring huge gains for institutions, governments, in vestors, among others. However, it is not a trivial task, due to the chaotic and unpredictable nature of the financial market. Several approaches have already been studied in order to achieve satisfactory results, such as machine learning and neural networks Long Short-Term Memory networks (LSTM) due to their effectiveness working with time series. In this work, our proposal is to use an LSTM neural network and from the closing price using renko chart, and perform the bullish or bearish forecast for the next market movement. Results show that our proposal presents satisfactory results compared to the traditional Buy and Hold technique.