Estudo de caso da metodologia Dynamic Time Scan Forecasting para previsão da série histórica de preços da ação preferencial da Petrobrás

Abstract

Motivated by the capacity of the Dynamic Time Scan Forecasting (DTSF) method in identifying patterns through statistical similarity functions in order to build forecasts in time series, this work proposes to study its efficiency in generating results in the stock market, using the series of prices of Petrobras preferred stock (PETR4) as a case study. For the evaluations, periods of stability of price variations, moments of strong fluctuations between price increases and falls, and moments of atypical events with a strong impact on market price volatility and, in particular, on Petrobras shares, were selected. The methodology was also compared with methodologies already quite common in the analysis of time series, these being the ARIMA, ETS and TBATS modeling.