A Systematic Literature Review on Decomposition Approaches to Estimate Time Series Components

Abstract

The study and modeling of systems have called the attention of several researchers, who are interested in estimating rules to describe data behavior. However, before proceeding with this estimation, it is necessary to understand the intrinsic features embedded in data. When such features are not correctly analyzed, the model accuracy tends to decrease. A well-known way to perform this analysis is by the study of time series behavior according to their stochastic and deterministic components. Nevertheless, the time series decomposition into these components is not a simple task. In order to address this issue, we conducted a rigorous and well-structured search for scientific papers in different repositories. By analyzing the recovered papers, we drew relevant conclusions such as: which methods are commonly used to decompose time series; the frequency of published papers per year; and the gaps of each method. Moreover, we have also classified the most suitable studies to estimate the determinism and stochastic- ity present in time series. After conducting this study, we concluded the development of methods to decompose time series into stochastic and deterministic components is still an open problem.