Automatic Time Series Forecasting with Ata Method in R: ATAforecasting Package

Abstract

Ata method is a new univariate time series forecasting method that provides innovative solutions to issues faced during the initialization and optimization stages of existing methods. The Ata method's forecasting performance is superior to existing methods in terms of easy implementation and accurate forecasting. It can be applied to non-seasonal or deseasonalized time series, where the deseasonalization can be performed via any preferred decomposition method. The R package ATAforecasting was developed as a comprehensive toolkit for automatic time series forecasting. It focuses on modeling all types of time series components with any preferred Ata methods and handling seasonality patterns by utilizing some popular decomposition techniques. The ATAforecasting package allows researchers to model seasonality with STL, STLplus, TBATS, stR, and TRAMO/SEATS, and power family transformation and analyze the any time series with a simple Ata method and additive, multiplicative, damped trend the Ata methods and level fixed Ata trended methods. It offers functions for researchers and data analysts to model any type of time series data sets without requiring specialization. However, an expert user may use the functions that can model all possible time series behaviors. The package also incorporates types of model specifications and their graphs, uses different accuracy measures that surely increase the Ata method's performance.

Comparison of the two hybrid models, Wavelet-ARIMA and Wavelet-ES, to predict the prices of the US dollar index

“…ES models have many formulas according to the nomenclature of Hyndman et al 2008. These models are named with three capital letters, the first letter describing the error, the second letter describing the general trend, and the third letter describing seasonality, [3], [5]. In this research, given that there is no general trend or seasonal trend in the time series resulting from the wavelet transformation, so a simple exponential smoothing will be employed with two models: the first model (A,N,N) which is the errors model, and the second model (M,N,N) which is the double error model.…”