COMTIF: computational method to treat, identify and forecast natural dynamic systems

Abstract

This work proposes a computational method to treat, identify and forecast natural dynamic systems applied in temporal data series. This method uses the exponential smoothing statistic technique on data treatment, the Nonlinear Auto Regressive Moving Average with eXogenous inputs - NARMAX statistic technique integrated to a feedforward neural network to identify, simulate and forecast temporal data series without manual treatment or choice of analitical models. The method was validated as a prototype applied in a real case study in the Amazonian Basin scenario.