Estudo sobre o uso de análise técnica e XGBoost em operações de day-trade

Abstract

The objective of this work is to evaluate the effects of Technical Analysis indicators to predict the price of the mini futures Dollar and the mini Ibovespa futures contracts, using XGBoost as a classification model in day-trade operations. In this context, predictor variables derived from Arithmetic Moving Averages, Exponential Moving Averages, Moving Average Convergence-Divergence, and Stochastic were used. The results of these variables at time 𝑡 were used to indicate whether the movement of asset prices at time 𝑡+1 was High (appreciation) or Low (devaluation). The classification models are built from XGBoost and evaluated through Accuracy and the simulated financial result. The best model for the mini futures Dollar presented an Accuracy of 50.1% and was able to generate a profit of R$ 11,940, while the model used as its benchmark generated a loss of R$ 15.135. The accuracy of the best model for the mini Ibovespa was 51.8% and its financial result was a profit of R$ 10,627, calculated in the same period in which the model used as its benchmark generated a loss of R$ 8.769.