Forecasting time series combining Holt-Winters and bootstrap approaches

Abstract

Exponential smoothing methods are the most used in time series modeling and forecasting, due to their versatility and the vast model option they integrate. Also, within the computing statistical area, Bootstrap methodology is widely applied in statistical inference concerning time series. Therefore, this study's main objective is to analyse Holt-Winters exponential smoothing method's performance associated to Bootstrap methodology, as an alternative procedure for modeling and forecasting in time series. The Bootstrap methodology combined with Holt-Winters methodology is applied to a study case on an environmental time series concerning a surface water quality variable, Dissolved Oxygen (DO). The proposed procedure allows to obtaining better point forecasts and interval forecasts with less amplitude than those obtained by means of the usual methods.