Análise do desempenho de regra de negociação via redes neurais artificiais em operações day trade.

Abstract

Predict whether a buy and sell operation will be profitable, is a very hot topic for the daytraders. So, many of them have been developing rules and logical procedures known as trading rules. The principle of trading rules has the premise that the patterns in market prices to recur in the future. Thanks to the rapid development of science, predict whether the operation will be profitable is advancing. Several studies have been developing by using different ways now. As an example, we can including machine learning, deep learning and so on. This paper proposes to measure the accuracy of the trading rule known as moving average crossing. For this task, we proposed a Deep Neural Network (DNN) model based on the binary classification. The financial asset chosen for research was Mini U.S. Dollar, from the second semester of 2019. For the DNN model, we set 22 features,i.e., close price, opening price, high price, low price. Deriving from the expert advisor, the features will serve as the input of the DNN architecture and then run and test the program. The result has shown that it is reliable to use the DNN method to predict whether the operation will be profitable or not.